summaryrefslogtreecommitdiffstats
path: root/third_party/rust/rusqlite
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-19 00:47:55 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-19 00:47:55 +0000
commit26a029d407be480d791972afb5975cf62c9360a6 (patch)
treef435a8308119effd964b339f76abb83a57c29483 /third_party/rust/rusqlite
parentInitial commit. (diff)
downloadfirefox-26a029d407be480d791972afb5975cf62c9360a6.tar.xz
firefox-26a029d407be480d791972afb5975cf62c9360a6.zip
Adding upstream version 124.0.1.upstream/124.0.1
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'third_party/rust/rusqlite')
-rw-r--r--third_party/rust/rusqlite/.cargo-checksum.json1
-rw-r--r--third_party/rust/rusqlite/Cargo.lock1024
-rw-r--r--third_party/rust/rusqlite/Cargo.toml259
-rw-r--r--third_party/rust/rusqlite/LICENSE19
-rw-r--r--third_party/rust/rusqlite/README.md248
-rw-r--r--third_party/rust/rusqlite/benches/cache.rs18
-rw-r--r--third_party/rust/rusqlite/benches/exec.rs17
-rw-r--r--third_party/rust/rusqlite/examples/load_extension.rs22
-rw-r--r--third_party/rust/rusqlite/examples/loadable_extension.rs50
-rw-r--r--third_party/rust/rusqlite/src/backup.rs428
-rw-r--r--third_party/rust/rusqlite/src/blob/mod.rs550
-rw-r--r--third_party/rust/rusqlite/src/blob/pos_io.rs272
-rw-r--r--third_party/rust/rusqlite/src/busy.rs170
-rw-r--r--third_party/rust/rusqlite/src/cache.rs350
-rw-r--r--third_party/rust/rusqlite/src/collation.rs215
-rw-r--r--third_party/rust/rusqlite/src/column.rs248
-rw-r--r--third_party/rust/rusqlite/src/config.rs163
-rw-r--r--third_party/rust/rusqlite/src/context.rs75
-rw-r--r--third_party/rust/rusqlite/src/error.rs476
-rw-r--r--third_party/rust/rusqlite/src/functions.rs1071
-rw-r--r--third_party/rust/rusqlite/src/hooks.rs806
-rw-r--r--third_party/rust/rusqlite/src/inner_connection.rs462
-rw-r--r--third_party/rust/rusqlite/src/lib.rs2200
-rw-r--r--third_party/rust/rusqlite/src/limits.rs163
-rw-r--r--third_party/rust/rusqlite/src/load_extension_guard.rs46
-rw-r--r--third_party/rust/rusqlite/src/params.rs455
-rw-r--r--third_party/rust/rusqlite/src/pragma.rs454
-rw-r--r--third_party/rust/rusqlite/src/raw_statement.rs240
-rw-r--r--third_party/rust/rusqlite/src/row.rs588
-rw-r--r--third_party/rust/rusqlite/src/serialize.rs167
-rw-r--r--third_party/rust/rusqlite/src/session.rs933
-rw-r--r--third_party/rust/rusqlite/src/statement.rs1360
-rw-r--r--third_party/rust/rusqlite/src/trace.rs184
-rw-r--r--third_party/rust/rusqlite/src/transaction.rs778
-rw-r--r--third_party/rust/rusqlite/src/types/chrono.rs319
-rw-r--r--third_party/rust/rusqlite/src/types/from_sql.rs366
-rw-r--r--third_party/rust/rusqlite/src/types/mod.rs447
-rw-r--r--third_party/rust/rusqlite/src/types/serde_json.rs135
-rw-r--r--third_party/rust/rusqlite/src/types/time.rs426
-rw-r--r--third_party/rust/rusqlite/src/types/to_sql.rs541
-rw-r--r--third_party/rust/rusqlite/src/types/url.rs82
-rw-r--r--third_party/rust/rusqlite/src/types/value.rs142
-rw-r--r--third_party/rust/rusqlite/src/types/value_ref.rs258
-rw-r--r--third_party/rust/rusqlite/src/unlock_notify.rs117
-rw-r--r--third_party/rust/rusqlite/src/util/mod.rs9
-rw-r--r--third_party/rust/rusqlite/src/util/param_cache.rs60
-rw-r--r--third_party/rust/rusqlite/src/util/small_cstr.rs170
-rw-r--r--third_party/rust/rusqlite/src/util/sqlite_string.rs239
-rw-r--r--third_party/rust/rusqlite/src/version.rs27
-rw-r--r--third_party/rust/rusqlite/src/vtab/array.rs223
-rw-r--r--third_party/rust/rusqlite/src/vtab/csvtab.rs387
-rw-r--r--third_party/rust/rusqlite/src/vtab/mod.rs1333
-rw-r--r--third_party/rust/rusqlite/src/vtab/series.rs341
-rw-r--r--third_party/rust/rusqlite/src/vtab/vtablog.rs297
-rw-r--r--third_party/rust/rusqlite/test.csv6
-rw-r--r--third_party/rust/rusqlite/tests/config_log.rs34
-rw-r--r--third_party/rust/rusqlite/tests/deny_single_threaded_sqlite_config.rs20
-rw-r--r--third_party/rust/rusqlite/tests/vtab.rs100
58 files changed, 20591 insertions, 0 deletions
diff --git a/third_party/rust/rusqlite/.cargo-checksum.json b/third_party/rust/rusqlite/.cargo-checksum.json
new file mode 100644
index 0000000000..73492ab78a
--- /dev/null
+++ b/third_party/rust/rusqlite/.cargo-checksum.json
@@ -0,0 +1 @@
+{"files":{"Cargo.lock":"7cabcce9576f0c112313e3859f9b98c57b7716ac3a54fddd5fec021f2ddf1ef3","Cargo.toml":"e004de86867ad91d8a07a44645aec126ab4dc345b4956878d161ff0e7e2d06c1","LICENSE":"f59ba65550f2a5adff98ec6478d783402c8e0a3eb515025c0b3438f3d30dc39e","README.md":"3ee43dc667cbc03092311fa1ad8c9baa8fee82a6afc11289cfa0d4b1d8cc173e","benches/cache.rs":"30459d8d7f65dd684909a53590f9be32e2e545a0775cd93ee0ce85dd08410955","benches/exec.rs":"a9f5e209dbfc73d1ee574608e0105a6c317b83f73d873ea8afd8b60e07cbc8d7","examples/load_extension.rs":"122195ec44c040ed9f7fb254035cc341a5d0d27120ba7540b8b6fa9ece6023d8","examples/loadable_extension.rs":"82dbcae9aa5ff73f7b36d1c8a296c858f8482e49aa9a32215a1e3cd738250a6d","src/backup.rs":"d4b91a5d103e7f6d4607bd186108f3460c698b6740366a8e7224778ba3c3a883","src/blob/mod.rs":"57ce4e741ec6b611cff5a56a6220628fd99154ae72ad41e6d7522d2b4e660caa","src/blob/pos_io.rs":"688513c4ca0d64d79555cb6fd8e75c0b41dceae7b10ae95e37c106ee93411dac","src/busy.rs":"39bf1447e5624b1b50bdd28601caa8a0f1559eaaac657b53eb03e186da6b72be","src/cache.rs":"07c7631eb1aef16e9bbc6d4512ef5831e37d42d18cac7862c0c2a7ed2af37e62","src/collation.rs":"5918eaa233acb6f205131a9ee0fc933d7b26a0c7f5965f30e8e2f9a0f34f9230","src/column.rs":"4ba6b52593e7343432c0fa95ded635515dd04d04e2e14c414ba896d435fc971d","src/config.rs":"d3c796cfc3760751d3c16629c989c446d19efee39a53f11081927458458b0f46","src/context.rs":"8dd0235e61c2c70617c14c8136dc8ae4ea3e61e6aca1177ce8c04edfbcbeebc3","src/error.rs":"8798596ed7c18b147a0ccf03f495c39b61b547083c7a584c0761371ab6c7ec0b","src/functions.rs":"e45647b4c3574f18aaf998982828b67933608239e39b499634486d7d95fa04ba","src/hooks.rs":"f21757bed19237c8ff99b255410ecb5e945383fc3250543afdd6a83c9309f8f0","src/inner_connection.rs":"b8512b7462bb007bf39ace4952066ddb043e400f76788661c35741ae3c55d086","src/lib.rs":"dfc6e23a3e6a9da3ec1a59db5b9f4e2aae1c9bdc452cc64b018a05e18f7835c5","src/limits.rs":"febcf841b690d82dd74c1704bc507c34d91f331714ef18cbba1b0236dc4f9897","src/load_extension_guard.rs":"9a75592bb26ccf22d4165b14779ca64a2f07d3358efbea60f7dea4b8c398e00a","src/params.rs":"c54d679576f795cacb8775d4eaefbe2b951d869c986e8245f37a98d9af0582ab","src/pragma.rs":"dc0c04f442612832c772f80a4e167744cd04f26e4d659b4b4e82038ff6e3039d","src/raw_statement.rs":"3bd688e59ad8285ea9a696a515b50f2a96468a1272d55e902a699616fbd746e3","src/row.rs":"0246bbe36634b2ca6b7aa05824bef33063b21debbb21cb4b293a6285a54581c6","src/serialize.rs":"d3f90888abbed15e48a85667d059065ec62338131bfef9480c3872637991217b","src/session.rs":"82a5f5409c83117ac0df4db51aaf0c2d4f524d378d37f5181c1978c36d9f2532","src/statement.rs":"6d260cdd3c443d09b890086cf8f94b0a4656fc3151265261180b2a104ab29a58","src/trace.rs":"1d8759dfd780fed70167255e12da27fd1f8cc942f8cd0ebcfcd69b90d7d4c31b","src/transaction.rs":"826f70decf2db3df40d38f68f483f56372834fde94755337e4cec4ab8e628349","src/types/chrono.rs":"d924bbc2b6177d51ed02bb0f444a079b991040b6485dfaea11859c936ac679e6","src/types/from_sql.rs":"ffe30f5d48ce27b725ecf18671238ae31037db4a6405703c25173edf6210a9f8","src/types/mod.rs":"91e7bdc021398b200c58cc8527c1c7bb598c59347b7d599cf7eb3a2d75e75ee6","src/types/serde_json.rs":"a50ce89fc59ae038ab4ff04f0dbdda467fb925b584419a727e04c9a293c6fe9b","src/types/time.rs":"93b5a613d25607db88d81ed3cbb1f38057a9b63e9ce59a864c661e0bf6188790","src/types/to_sql.rs":"a78faa2bee1293d7c66b10679eb959622905351a37e08d7a78476bfe5f02b993","src/types/url.rs":"0633f2cad5910df7c5ce2d709c345b422618cb83bda0c91e67326e82b0bdeb6a","src/types/value.rs":"d39f51cfe31f57e6030927ed70e87150b466f4826e79281562645a574a4bc314","src/types/value_ref.rs":"f5d26052164f440046cddddc752706120495945917cc0b8233bf7e702ae94091","src/unlock_notify.rs":"c5a7b3413d863b4c87b0ee42fa817c04a277022cd24facfcc0dd2c641c2ca3af","src/util/mod.rs":"3915f19ddc31558e5a1d0fb8e2cb414ea2d2323131fe0d8c0fd83797f775b031","src/util/param_cache.rs":"efb480749cd74c09aeca28be074050e3a9aed7c8ed1371ca562342cba9c408dd","src/util/small_cstr.rs":"634f43a03d41ecb6515ff39303e01242b9dc774c17de9177a0a07a5812c0f777","src/util/sqlite_string.rs":"7f281776180449f7e6cd0169cf68658ce9324c275411105caaef21267de901a3","src/version.rs":"392b327ecd9acc1d289f92abb671104b065e232368456b0bbd5bfaca96d90950","src/vtab/array.rs":"a51ce401ed4029a22a93ac03f235a9349087581338edd46a8582e5bf869e4cc8","src/vtab/csvtab.rs":"086a87f7b4be32dbadb318ded1870d1b4a4a1c23bcc73ca77b71d26863c4501f","src/vtab/mod.rs":"cf1cdd7dec42b64e27825089dbcfaf5ae5137eec496ff84239f53c758692b97b","src/vtab/series.rs":"7b93cf3ecca98f1e72945c0c8132042b7cbc935f9ddbdcedea3b295513ecc3da","src/vtab/vtablog.rs":"0835dd517e02875ffde0080a64b765e4cea0e0c9c9645c3c58ddb86e664151da","test.csv":"3f5649d7b9f80468999b80d4d0e747c6f4f1bba80ade792689eeb4359dc1834a","tests/config_log.rs":"e786bb1a0ca1b560788b24a2091e7d294753a7bee8a39f0e6037b435809267df","tests/deny_single_threaded_sqlite_config.rs":"5ce873b83ae38e26f07cd6b93094e144484da22356d7d7c683eb2f608ea1b4f3","tests/vtab.rs":"a66ee6d79ea156585a10d770fff31f122d968c2f2f20790aa83a73b5cc8e6405"},"package":"a78046161564f5e7cd9008aff3b2990b3850dc8e0349119b98e8f251e099f24d"} \ No newline at end of file
diff --git a/third_party/rust/rusqlite/Cargo.lock b/third_party/rust/rusqlite/Cargo.lock
new file mode 100644
index 0000000000..2b04c3f0a0
--- /dev/null
+++ b/third_party/rust/rusqlite/Cargo.lock
@@ -0,0 +1,1024 @@
+# This file is automatically @generated by Cargo.
+# It is not intended for manual editing.
+version = 3
+
+[[package]]
+name = "ahash"
+version = "0.8.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "91429305e9f0a25f6205c5b8e0d2db09e0708a7a6df0f42212bb56c32c8ac97a"
+dependencies = [
+ "cfg-if",
+ "once_cell",
+ "version_check",
+ "zerocopy",
+]
+
+[[package]]
+name = "aho-corasick"
+version = "1.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b2969dcb958b36655471fc61f7e416fa76033bdd4bfed0678d8fee1e2d07a1f0"
+dependencies = [
+ "memchr",
+]
+
+[[package]]
+name = "allocator-api2"
+version = "0.2.16"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0942ffc6dcaadf03badf6e6a2d0228460359d5e34b57ccdc720b7382dfbd5ec5"
+
+[[package]]
+name = "android-tzdata"
+version = "0.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e999941b234f3131b00bc13c22d06e8c5ff726d1b6318ac7eb276997bbb4fef0"
+
+[[package]]
+name = "android_system_properties"
+version = "0.1.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "819e7219dbd41043ac279b19830f2efc897156490d7fd6ea916720117ee66311"
+dependencies = [
+ "libc",
+]
+
+[[package]]
+name = "autocfg"
+version = "1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d468802bab17cbc0cc575e9b053f41e72aa36bfa6b7f55e3529ffa43161b97fa"
+
+[[package]]
+name = "bencher"
+version = "0.1.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7dfdb4953a096c551ce9ace855a604d702e6e62d77fac690575ae347571717f5"
+
+[[package]]
+name = "bindgen"
+version = "0.69.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9ffcebc3849946a7170a05992aac39da343a90676ab392c51a4280981d6379c2"
+dependencies = [
+ "bitflags 2.4.1",
+ "cexpr",
+ "clang-sys",
+ "lazy_static",
+ "lazycell",
+ "peeking_take_while",
+ "proc-macro2",
+ "quote",
+ "regex",
+ "rustc-hash",
+ "shlex",
+ "syn",
+]
+
+[[package]]
+name = "bitflags"
+version = "1.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bef38d45163c2f1dde094a7dfd33ccf595c92905c8f8f4fdc18d06fb1037718a"
+
+[[package]]
+name = "bitflags"
+version = "2.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "327762f6e5a765692301e5bb513e0d9fef63be86bbc14528052b1cd3e6f03e07"
+
+[[package]]
+name = "bumpalo"
+version = "3.14.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7f30e7476521f6f8af1a1c4c0b8cc94f0bee37d91763d0ca2665f299b6cd8aec"
+
+[[package]]
+name = "cc"
+version = "1.0.83"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f1174fb0b6ec23863f8b971027804a42614e347eafb0a95bf0b12cdae21fc4d0"
+dependencies = [
+ "libc",
+]
+
+[[package]]
+name = "cexpr"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6fac387a98bb7c37292057cffc56d62ecb629900026402633ae9160df93a8766"
+dependencies = [
+ "nom",
+]
+
+[[package]]
+name = "cfg-if"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "baf1de4339761588bc0619e3cbc0120ee582ebb74b53b4efbf79117bd2da40fd"
+
+[[package]]
+name = "chrono"
+version = "0.4.31"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7f2c685bad3eb3d45a01354cedb7d5faa66194d1d58ba6e267a8de788f79db38"
+dependencies = [
+ "android-tzdata",
+ "iana-time-zone",
+ "num-traits",
+ "windows-targets",
+]
+
+[[package]]
+name = "clang-sys"
+version = "1.6.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c688fc74432808e3eb684cae8830a86be1d66a2bd58e1f248ed0960a590baf6f"
+dependencies = [
+ "glob",
+ "libc",
+ "libloading",
+]
+
+[[package]]
+name = "core-foundation-sys"
+version = "0.8.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e496a50fda8aacccc86d7529e2c1e0892dbd0f898a6b5645b5561b89c3210efa"
+
+[[package]]
+name = "csv"
+version = "1.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ac574ff4d437a7b5ad237ef331c17ccca63c46479e5b5453eb8e10bb99a759fe"
+dependencies = [
+ "csv-core",
+ "itoa",
+ "ryu",
+ "serde",
+]
+
+[[package]]
+name = "csv-core"
+version = "0.1.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5efa2b3d7902f4b634a20cae3c9c4e6209dc4779feb6863329607560143efa70"
+dependencies = [
+ "memchr",
+]
+
+[[package]]
+name = "deranged"
+version = "0.3.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0f32d04922c60427da6f9fef14d042d9edddef64cb9d4ce0d64d0685fbeb1fd3"
+dependencies = [
+ "powerfmt",
+]
+
+[[package]]
+name = "doc-comment"
+version = "0.3.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fea41bba32d969b513997752735605054bc0dfa92b4c56bf1189f2e174be7a10"
+
+[[package]]
+name = "equivalent"
+version = "1.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5443807d6dff69373d433ab9ef5378ad8df50ca6298caf15de6e52e24aaf54d5"
+
+[[package]]
+name = "errno"
+version = "0.3.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7c18ee0ed65a5f1f81cac6b1d213b69c35fa47d4252ad41f1486dbd8226fe36e"
+dependencies = [
+ "libc",
+ "windows-sys",
+]
+
+[[package]]
+name = "fallible-iterator"
+version = "0.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2acce4a10f12dc2fb14a218589d4f1f62ef011b2d0cc4b3cb1bba8e94da14649"
+
+[[package]]
+name = "fallible-streaming-iterator"
+version = "0.1.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7360491ce676a36bf9bb3c56c1aa791658183a54d2744120f27285738d90465a"
+
+[[package]]
+name = "fastrand"
+version = "2.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "25cbce373ec4653f1a01a31e8a5e5ec0c622dc27ff9c4e6606eefef5cbbed4a5"
+
+[[package]]
+name = "form_urlencoded"
+version = "1.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a62bc1cf6f830c2ec14a513a9fb124d0a213a629668a4186f329db21fe045652"
+dependencies = [
+ "percent-encoding",
+]
+
+[[package]]
+name = "getrandom"
+version = "0.2.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "fe9006bed769170c11f845cf00c7c1e9092aeb3f268e007c3e760ac68008070f"
+dependencies = [
+ "cfg-if",
+ "libc",
+ "wasi",
+]
+
+[[package]]
+name = "glob"
+version = "0.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d2fabcfbdc87f4758337ca535fb41a6d701b65693ce38287d856d1674551ec9b"
+
+[[package]]
+name = "hashbrown"
+version = "0.14.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f93e7192158dbcda357bdec5fb5788eebf8bbac027f3f33e719d29135ae84156"
+dependencies = [
+ "ahash",
+ "allocator-api2",
+]
+
+[[package]]
+name = "hashlink"
+version = "0.8.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e8094feaf31ff591f651a2664fb9cfd92bba7a60ce3197265e9482ebe753c8f7"
+dependencies = [
+ "hashbrown",
+]
+
+[[package]]
+name = "iana-time-zone"
+version = "0.1.58"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8326b86b6cff230b97d0d312a6c40a60726df3332e721f72a1b035f451663b20"
+dependencies = [
+ "android_system_properties",
+ "core-foundation-sys",
+ "iana-time-zone-haiku",
+ "js-sys",
+ "wasm-bindgen",
+ "windows-core",
+]
+
+[[package]]
+name = "iana-time-zone-haiku"
+version = "0.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f31827a206f56af32e590ba56d5d2d085f558508192593743f16b2306495269f"
+dependencies = [
+ "cc",
+]
+
+[[package]]
+name = "idna"
+version = "0.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7d20d6b07bfbc108882d88ed8e37d39636dcc260e15e30c45e6ba089610b917c"
+dependencies = [
+ "unicode-bidi",
+ "unicode-normalization",
+]
+
+[[package]]
+name = "indexmap"
+version = "2.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d530e1a18b1cb4c484e6e34556a0d948706958449fca0cab753d649f2bce3d1f"
+dependencies = [
+ "equivalent",
+ "hashbrown",
+]
+
+[[package]]
+name = "itoa"
+version = "1.0.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "af150ab688ff2122fcef229be89cb50dd66af9e01a4ff320cc137eecc9bacc38"
+
+[[package]]
+name = "js-sys"
+version = "0.3.65"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "54c0c35952f67de54bb584e9fd912b3023117cbafc0a77d8f3dee1fb5f572fe8"
+dependencies = [
+ "wasm-bindgen",
+]
+
+[[package]]
+name = "lazy_static"
+version = "1.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e2abad23fbc42b3700f2f279844dc832adb2b2eb069b2df918f455c4e18cc646"
+
+[[package]]
+name = "lazycell"
+version = "1.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "830d08ce1d1d941e6b30645f1a0eb5643013d835ce3779a5fc208261dbe10f55"
+
+[[package]]
+name = "libc"
+version = "0.2.150"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "89d92a4743f9a61002fae18374ed11e7973f530cb3a3255fb354818118b2203c"
+
+[[package]]
+name = "libloading"
+version = "0.7.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b67380fd3b2fbe7527a606e18729d21c6f3951633d0500574c4dc22d2d638b9f"
+dependencies = [
+ "cfg-if",
+ "winapi",
+]
+
+[[package]]
+name = "libsqlite3-sys"
+version = "0.27.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cf4e226dcd58b4be396f7bd3c20da8fdee2911400705297ba7d2d7cc2c30f716"
+dependencies = [
+ "bindgen",
+ "cc",
+ "openssl-sys",
+ "pkg-config",
+ "prettyplease",
+ "quote",
+ "syn",
+ "vcpkg",
+]
+
+[[package]]
+name = "linux-raw-sys"
+version = "0.4.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "969488b55f8ac402214f3f5fd243ebb7206cf82de60d3172994707a4bcc2b829"
+
+[[package]]
+name = "log"
+version = "0.4.20"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b5e6163cb8c49088c2c36f57875e58ccd8c87c7427f7fbd50ea6710b2f3f2e8f"
+
+[[package]]
+name = "memchr"
+version = "2.6.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f665ee40bc4a3c5590afb1e9677db74a508659dfd71e126420da8274909a0167"
+
+[[package]]
+name = "minimal-lexical"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "68354c5c6bd36d73ff3feceb05efa59b6acb7626617f4962be322a825e61f79a"
+
+[[package]]
+name = "nom"
+version = "7.1.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d273983c5a657a70a3e8f2a01329822f3b8c8172b73826411a55751e404a0a4a"
+dependencies = [
+ "memchr",
+ "minimal-lexical",
+]
+
+[[package]]
+name = "num-traits"
+version = "0.2.17"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "39e3200413f237f41ab11ad6d161bc7239c84dcb631773ccd7de3dfe4b5c267c"
+dependencies = [
+ "autocfg",
+]
+
+[[package]]
+name = "once_cell"
+version = "1.18.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dd8b5dd2ae5ed71462c540258bedcb51965123ad7e7ccf4b9a8cafaa4a63576d"
+
+[[package]]
+name = "openssl-src"
+version = "300.1.6+3.1.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "439fac53e092cd7442a3660c85dde4643ab3b5bd39040912388dcdabf6b88085"
+dependencies = [
+ "cc",
+]
+
+[[package]]
+name = "openssl-sys"
+version = "0.9.95"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "40a4130519a360279579c2053038317e40eff64d13fd3f004f9e1b72b8a6aaf9"
+dependencies = [
+ "cc",
+ "libc",
+ "openssl-src",
+ "pkg-config",
+ "vcpkg",
+]
+
+[[package]]
+name = "peeking_take_while"
+version = "0.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "19b17cddbe7ec3f8bc800887bab5e717348c95ea2ca0b1bf0837fb964dc67099"
+
+[[package]]
+name = "percent-encoding"
+version = "2.3.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9b2a4787296e9989611394c33f193f676704af1686e70b8f8033ab5ba9a35a94"
+
+[[package]]
+name = "phf"
+version = "0.11.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ade2d8b8f33c7333b51bcf0428d37e217e9f32192ae4772156f65063b8ce03dc"
+dependencies = [
+ "phf_shared",
+]
+
+[[package]]
+name = "phf_codegen"
+version = "0.11.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e8d39688d359e6b34654d328e262234662d16cc0f60ec8dcbe5e718709342a5a"
+dependencies = [
+ "phf_generator",
+ "phf_shared",
+]
+
+[[package]]
+name = "phf_generator"
+version = "0.11.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "48e4cc64c2ad9ebe670cb8fd69dd50ae301650392e81c05f9bfcb2d5bdbc24b0"
+dependencies = [
+ "phf_shared",
+ "rand",
+]
+
+[[package]]
+name = "phf_shared"
+version = "0.11.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "90fcb95eef784c2ac79119d1dd819e162b5da872ce6f3c3abe1e8ca1c082f72b"
+dependencies = [
+ "siphasher",
+ "uncased",
+]
+
+[[package]]
+name = "pkg-config"
+version = "0.3.27"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "26072860ba924cbfa98ea39c8c19b4dd6a4a25423dbdf219c1eca91aa0cf6964"
+
+[[package]]
+name = "powerfmt"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "439ee305def115ba05938db6eb1644ff94165c5ab5e9420d1c1bcedbba909391"
+
+[[package]]
+name = "prettyplease"
+version = "0.2.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ae005bd773ab59b4725093fd7df83fd7892f7d8eafb48dbd7de6e024e4215f9d"
+dependencies = [
+ "proc-macro2",
+ "syn",
+]
+
+[[package]]
+name = "proc-macro2"
+version = "1.0.69"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "134c189feb4956b20f6f547d2cf727d4c0fe06722b20a0eec87ed445a97f92da"
+dependencies = [
+ "unicode-ident",
+]
+
+[[package]]
+name = "quote"
+version = "1.0.33"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5267fca4496028628a95160fc423a33e8b2e6af8a5302579e322e4b520293cae"
+dependencies = [
+ "proc-macro2",
+]
+
+[[package]]
+name = "rand"
+version = "0.8.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "34af8d1a0e25924bc5b7c43c079c942339d8f0a8b57c39049bef581b46327404"
+dependencies = [
+ "rand_core",
+]
+
+[[package]]
+name = "rand_core"
+version = "0.6.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ec0be4795e2f6a28069bec0b5ff3e2ac9bafc99e6a9a7dc3547996c5c816922c"
+
+[[package]]
+name = "redox_syscall"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4722d768eff46b75989dd134e5c353f0d6296e5aaa3132e776cbdb56be7731aa"
+dependencies = [
+ "bitflags 1.3.2",
+]
+
+[[package]]
+name = "regex"
+version = "1.10.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "380b951a9c5e80ddfd6136919eef32310721aa4aacd4889a8d39124b026ab343"
+dependencies = [
+ "aho-corasick",
+ "memchr",
+ "regex-automata",
+ "regex-syntax",
+]
+
+[[package]]
+name = "regex-automata"
+version = "0.4.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5f804c7828047e88b2d32e2d7fe5a105da8ee3264f01902f796c8e067dc2483f"
+dependencies = [
+ "aho-corasick",
+ "memchr",
+ "regex-syntax",
+]
+
+[[package]]
+name = "regex-syntax"
+version = "0.8.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c08c74e62047bb2de4ff487b251e4a92e24f48745648451635cec7d591162d9f"
+
+[[package]]
+name = "rusqlite"
+version = "0.30.0"
+dependencies = [
+ "bencher",
+ "bitflags 2.4.1",
+ "chrono",
+ "csv",
+ "doc-comment",
+ "fallible-iterator",
+ "fallible-streaming-iterator",
+ "hashlink",
+ "lazy_static",
+ "libsqlite3-sys",
+ "regex",
+ "rusqlite-macros",
+ "serde_json",
+ "smallvec",
+ "tempfile",
+ "time",
+ "unicase",
+ "url",
+ "uuid",
+]
+
+[[package]]
+name = "rusqlite-macros"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3d92a8272e44aab115714abdf828c98edca44c8fc69861b44465cc77cfb56f2d"
+dependencies = [
+ "fallible-iterator",
+ "sqlite3-parser",
+]
+
+[[package]]
+name = "rustc-hash"
+version = "1.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "08d43f7aa6b08d49f382cde6a7982047c3426db949b1424bc4b7ec9ae12c6ce2"
+
+[[package]]
+name = "rustix"
+version = "0.38.21"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2b426b0506e5d50a7d8dafcf2e81471400deb602392c7dd110815afb4eaf02a3"
+dependencies = [
+ "bitflags 2.4.1",
+ "errno",
+ "libc",
+ "linux-raw-sys",
+ "windows-sys",
+]
+
+[[package]]
+name = "ryu"
+version = "1.0.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1ad4cc8da4ef723ed60bced201181d83791ad433213d8c24efffda1eec85d741"
+
+[[package]]
+name = "serde"
+version = "1.0.192"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bca2a08484b285dcb282d0f67b26cadc0df8b19f8c12502c13d966bf9482f001"
+dependencies = [
+ "serde_derive",
+]
+
+[[package]]
+name = "serde_derive"
+version = "1.0.192"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d6c7207fbec9faa48073f3e3074cbe553af6ea512d7c21ba46e434e70ea9fbc1"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
+[[package]]
+name = "serde_json"
+version = "1.0.108"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3d1c7e3eac408d115102c4c24ad393e0821bb3a5df4d506a80f85f7a742a526b"
+dependencies = [
+ "itoa",
+ "ryu",
+ "serde",
+]
+
+[[package]]
+name = "shlex"
+version = "1.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a7cee0529a6d40f580e7a5e6c495c8fbfe21b7b52795ed4bb5e62cdf92bc6380"
+
+[[package]]
+name = "siphasher"
+version = "0.3.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "38b58827f4464d87d377d175e90bf58eb00fd8716ff0a62f80356b5e61555d0d"
+
+[[package]]
+name = "smallvec"
+version = "1.11.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4dccd0940a2dcdf68d092b8cbab7dc0ad8fa938bf95787e1b916b0e3d0e8e970"
+
+[[package]]
+name = "sqlite3-parser"
+version = "0.12.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9add252f9b70a7d493b03127524ed06cdf7480b3dc8c1b2ccfda89384d90a8b7"
+dependencies = [
+ "bitflags 2.4.1",
+ "cc",
+ "fallible-iterator",
+ "indexmap",
+ "log",
+ "memchr",
+ "phf",
+ "phf_codegen",
+ "phf_shared",
+ "smallvec",
+ "uncased",
+]
+
+[[package]]
+name = "syn"
+version = "2.0.39"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "23e78b90f2fcf45d3e842032ce32e3f2d1545ba6636271dcbf24fa306d87be7a"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "unicode-ident",
+]
+
+[[package]]
+name = "tempfile"
+version = "3.8.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7ef1adac450ad7f4b3c28589471ade84f25f731a7a0fe30d71dfa9f60fd808e5"
+dependencies = [
+ "cfg-if",
+ "fastrand",
+ "redox_syscall",
+ "rustix",
+ "windows-sys",
+]
+
+[[package]]
+name = "time"
+version = "0.3.30"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c4a34ab300f2dee6e562c10a046fc05e358b29f9bf92277f30c3c8d82275f6f5"
+dependencies = [
+ "deranged",
+ "itoa",
+ "powerfmt",
+ "serde",
+ "time-core",
+ "time-macros",
+]
+
+[[package]]
+name = "time-core"
+version = "0.1.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ef927ca75afb808a4d64dd374f00a2adf8d0fcff8e7b184af886c3c87ec4a3f3"
+
+[[package]]
+name = "time-macros"
+version = "0.2.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4ad70d68dba9e1f8aceda7aa6711965dfec1cac869f311a51bd08b3a2ccbce20"
+dependencies = [
+ "time-core",
+]
+
+[[package]]
+name = "tinyvec"
+version = "1.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "87cc5ceb3875bb20c2890005a4e226a4651264a5c75edb2421b52861a0a0cb50"
+dependencies = [
+ "tinyvec_macros",
+]
+
+[[package]]
+name = "tinyvec_macros"
+version = "0.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1f3ccbac311fea05f86f61904b462b55fb3df8837a366dfc601a0161d0532f20"
+
+[[package]]
+name = "uncased"
+version = "0.9.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9b9bc53168a4be7402ab86c3aad243a84dd7381d09be0eddc81280c1da95ca68"
+dependencies = [
+ "version_check",
+]
+
+[[package]]
+name = "unicase"
+version = "2.7.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f7d2d4dafb69621809a81864c9c1b864479e1235c0dd4e199924b9742439ed89"
+dependencies = [
+ "version_check",
+]
+
+[[package]]
+name = "unicode-bidi"
+version = "0.3.13"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "92888ba5573ff080736b3648696b70cafad7d250551175acbaa4e0385b3e1460"
+
+[[package]]
+name = "unicode-ident"
+version = "1.0.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "3354b9ac3fae1ff6755cb6db53683adb661634f67557942dea4facebec0fee4b"
+
+[[package]]
+name = "unicode-normalization"
+version = "0.1.22"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5c5713f0fc4b5db668a2ac63cdb7bb4469d8c9fed047b1d0292cc7b0ce2ba921"
+dependencies = [
+ "tinyvec",
+]
+
+[[package]]
+name = "url"
+version = "2.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "143b538f18257fac9cad154828a57c6bf5157e1aa604d4816b5995bf6de87ae5"
+dependencies = [
+ "form_urlencoded",
+ "idna",
+ "percent-encoding",
+]
+
+[[package]]
+name = "uuid"
+version = "1.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "88ad59a7560b41a70d191093a945f0b87bc1deeda46fb237479708a1d6b6cdfc"
+dependencies = [
+ "getrandom",
+]
+
+[[package]]
+name = "vcpkg"
+version = "0.2.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "accd4ea62f7bb7a82fe23066fb0957d48ef677f6eeb8215f372f52e48bb32426"
+
+[[package]]
+name = "version_check"
+version = "0.9.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "49874b5167b65d7193b8aba1567f5c7d93d001cafc34600cee003eda787e483f"
+
+[[package]]
+name = "wasi"
+version = "0.11.0+wasi-snapshot-preview1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9c8d87e72b64a3b4db28d11ce29237c246188f4f51057d65a7eab63b7987e423"
+
+[[package]]
+name = "wasm-bindgen"
+version = "0.2.88"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7daec296f25a1bae309c0cd5c29c4b260e510e6d813c286b19eaadf409d40fce"
+dependencies = [
+ "cfg-if",
+ "wasm-bindgen-macro",
+]
+
+[[package]]
+name = "wasm-bindgen-backend"
+version = "0.2.88"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e397f4664c0e4e428e8313a469aaa58310d302159845980fd23b0f22a847f217"
+dependencies = [
+ "bumpalo",
+ "log",
+ "once_cell",
+ "proc-macro2",
+ "quote",
+ "syn",
+ "wasm-bindgen-shared",
+]
+
+[[package]]
+name = "wasm-bindgen-macro"
+version = "0.2.88"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5961017b3b08ad5f3fe39f1e79877f8ee7c23c5e5fd5eb80de95abc41f1f16b2"
+dependencies = [
+ "quote",
+ "wasm-bindgen-macro-support",
+]
+
+[[package]]
+name = "wasm-bindgen-macro-support"
+version = "0.2.88"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c5353b8dab669f5e10f5bd76df26a9360c748f054f862ff5f3f8aae0c7fb3907"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+ "wasm-bindgen-backend",
+ "wasm-bindgen-shared",
+]
+
+[[package]]
+name = "wasm-bindgen-shared"
+version = "0.2.88"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0d046c5d029ba91a1ed14da14dca44b68bf2f124cfbaf741c54151fdb3e0750b"
+
+[[package]]
+name = "winapi"
+version = "0.3.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5c839a674fcd7a98952e593242ea400abe93992746761e38641405d28b00f419"
+dependencies = [
+ "winapi-i686-pc-windows-gnu",
+ "winapi-x86_64-pc-windows-gnu",
+]
+
+[[package]]
+name = "winapi-i686-pc-windows-gnu"
+version = "0.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ac3b87c63620426dd9b991e5ce0329eff545bccbbb34f3be09ff6fb6ab51b7b6"
+
+[[package]]
+name = "winapi-x86_64-pc-windows-gnu"
+version = "0.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "712e227841d057c1ee1cd2fb22fa7e5a5461ae8e48fa2ca79ec42cfc1931183f"
+
+[[package]]
+name = "windows-core"
+version = "0.51.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f1f8cf84f35d2db49a46868f947758c7a1138116f7fac3bc844f43ade1292e64"
+dependencies = [
+ "windows-targets",
+]
+
+[[package]]
+name = "windows-sys"
+version = "0.48.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "677d2418bec65e3338edb076e806bc1ec15693c5d0104683f2efe857f61056a9"
+dependencies = [
+ "windows-targets",
+]
+
+[[package]]
+name = "windows-targets"
+version = "0.48.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9a2fa6e2155d7247be68c096456083145c183cbbbc2764150dda45a87197940c"
+dependencies = [
+ "windows_aarch64_gnullvm",
+ "windows_aarch64_msvc",
+ "windows_i686_gnu",
+ "windows_i686_msvc",
+ "windows_x86_64_gnu",
+ "windows_x86_64_gnullvm",
+ "windows_x86_64_msvc",
+]
+
+[[package]]
+name = "windows_aarch64_gnullvm"
+version = "0.48.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2b38e32f0abccf9987a4e3079dfb67dcd799fb61361e53e2882c3cbaf0d905d8"
+
+[[package]]
+name = "windows_aarch64_msvc"
+version = "0.48.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dc35310971f3b2dbbf3f0690a219f40e2d9afcf64f9ab7cc1be722937c26b4bc"
+
+[[package]]
+name = "windows_i686_gnu"
+version = "0.48.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a75915e7def60c94dcef72200b9a8e58e5091744960da64ec734a6c6e9b3743e"
+
+[[package]]
+name = "windows_i686_msvc"
+version = "0.48.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8f55c233f70c4b27f66c523580f78f1004e8b5a8b659e05a4eb49d4166cca406"
+
+[[package]]
+name = "windows_x86_64_gnu"
+version = "0.48.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "53d40abd2583d23e4718fddf1ebec84dbff8381c07cae67ff7768bbf19c6718e"
+
+[[package]]
+name = "windows_x86_64_gnullvm"
+version = "0.48.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0b7b52767868a23d5bab768e390dc5f5c55825b6d30b86c844ff2dc7414044cc"
+
+[[package]]
+name = "windows_x86_64_msvc"
+version = "0.48.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ed94fce61571a4006852b7389a063ab983c02eb1bb37b47f8272ce92d06d9538"
+
+[[package]]
+name = "zerocopy"
+version = "0.7.25"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8cd369a67c0edfef15010f980c3cbe45d7f651deac2cd67ce097cd801de16557"
+dependencies = [
+ "zerocopy-derive",
+]
+
+[[package]]
+name = "zerocopy-derive"
+version = "0.7.25"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c2f140bda219a26ccc0cdb03dba58af72590c53b22642577d88a927bc5c87d6b"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn",
+]
diff --git a/third_party/rust/rusqlite/Cargo.toml b/third_party/rust/rusqlite/Cargo.toml
new file mode 100644
index 0000000000..e16fc07c52
--- /dev/null
+++ b/third_party/rust/rusqlite/Cargo.toml
@@ -0,0 +1,259 @@
+# THIS FILE IS AUTOMATICALLY GENERATED BY CARGO
+#
+# When uploading crates to the registry Cargo will automatically
+# "normalize" Cargo.toml files for maximal compatibility
+# with all versions of Cargo and also rewrite `path` dependencies
+# to registry (e.g., crates.io) dependencies.
+#
+# If you are reading this file be aware that the original Cargo.toml
+# will likely look very different (and much more reasonable).
+# See Cargo.toml.orig for the original contents.
+
+[package]
+edition = "2021"
+name = "rusqlite"
+version = "0.30.0"
+authors = ["The rusqlite developers"]
+exclude = [
+ "/.github/*",
+ "/.gitattributes",
+ "/appveyor.yml",
+ "/Changelog.md",
+ "/clippy.toml",
+ "/codecov.yml",
+]
+description = "Ergonomic wrapper for SQLite"
+documentation = "https://docs.rs/rusqlite/"
+readme = "README.md"
+keywords = [
+ "sqlite",
+ "database",
+ "ffi",
+]
+categories = ["database"]
+license = "MIT"
+repository = "https://github.com/rusqlite/rusqlite"
+
+[package.metadata.docs.rs]
+all-features = false
+default-target = "x86_64-unknown-linux-gnu"
+features = [
+ "modern-full",
+ "rusqlite-macros",
+]
+no-default-features = true
+rustdoc-args = [
+ "--cfg",
+ "docsrs",
+]
+
+[package.metadata.playground]
+all-features = false
+features = ["bundled-full"]
+
+[lib]
+name = "rusqlite"
+
+[[example]]
+name = "loadable_extension"
+crate-type = ["cdylib"]
+required-features = [
+ "loadable_extension",
+ "modern_sqlite",
+ "functions",
+ "vtab",
+ "trace",
+]
+
+[[example]]
+name = "load_extension"
+required-features = [
+ "load_extension",
+ "bundled",
+ "functions",
+ "vtab",
+ "trace",
+]
+
+[[test]]
+name = "config_log"
+harness = false
+
+[[test]]
+name = "deny_single_threaded_sqlite_config"
+
+[[test]]
+name = "vtab"
+
+[[bench]]
+name = "cache"
+harness = false
+
+[[bench]]
+name = "exec"
+harness = false
+
+[dependencies.bitflags]
+version = "2.0"
+
+[dependencies.chrono]
+version = "0.4"
+features = ["clock"]
+optional = true
+default-features = false
+
+[dependencies.csv]
+version = "1.1"
+optional = true
+
+[dependencies.fallible-iterator]
+version = "0.3"
+
+[dependencies.fallible-streaming-iterator]
+version = "0.1"
+
+[dependencies.hashlink]
+version = "0.8"
+
+[dependencies.libsqlite3-sys]
+version = "0.27.0"
+
+[dependencies.rusqlite-macros]
+version = "0.1.0"
+optional = true
+
+[dependencies.serde_json]
+version = "1.0"
+optional = true
+
+[dependencies.smallvec]
+version = "1.6.1"
+
+[dependencies.time]
+version = "0.3.0"
+features = [
+ "formatting",
+ "macros",
+ "parsing",
+]
+optional = true
+
+[dependencies.url]
+version = "2.1"
+optional = true
+
+[dependencies.uuid]
+version = "1.0"
+optional = true
+
+[dev-dependencies.bencher]
+version = "0.1"
+
+[dev-dependencies.doc-comment]
+version = "0.3"
+
+[dev-dependencies.lazy_static]
+version = "1.4"
+
+[dev-dependencies.regex]
+version = "1.5.5"
+
+[dev-dependencies.tempfile]
+version = "3.1.0"
+
+[dev-dependencies.unicase]
+version = "2.6.0"
+
+[dev-dependencies.uuid]
+version = "1.0"
+features = ["v4"]
+
+[features]
+array = ["vtab"]
+backup = []
+blob = []
+buildtime_bindgen = ["libsqlite3-sys/buildtime_bindgen"]
+bundled = [
+ "libsqlite3-sys/bundled",
+ "modern_sqlite",
+]
+bundled-full = [
+ "modern-full",
+ "bundled",
+]
+bundled-sqlcipher = [
+ "libsqlite3-sys/bundled-sqlcipher",
+ "bundled",
+]
+bundled-sqlcipher-vendored-openssl = [
+ "libsqlite3-sys/bundled-sqlcipher-vendored-openssl",
+ "bundled-sqlcipher",
+]
+bundled-windows = ["libsqlite3-sys/bundled-windows"]
+collation = []
+column_decltype = []
+csvtab = [
+ "csv",
+ "vtab",
+]
+extra_check = []
+functions = []
+hooks = []
+i128_blob = []
+in_gecko = [
+ "modern_sqlite",
+ "libsqlite3-sys/in_gecko",
+]
+limits = []
+load_extension = []
+loadable_extension = ["libsqlite3-sys/loadable_extension"]
+modern-full = [
+ "array",
+ "backup",
+ "blob",
+ "modern_sqlite",
+ "chrono",
+ "collation",
+ "column_decltype",
+ "csvtab",
+ "extra_check",
+ "functions",
+ "hooks",
+ "i128_blob",
+ "limits",
+ "load_extension",
+ "serde_json",
+ "series",
+ "time",
+ "trace",
+ "unlock_notify",
+ "url",
+ "uuid",
+ "vtab",
+ "window",
+]
+modern_sqlite = ["libsqlite3-sys/bundled_bindings"]
+release_memory = []
+serialize = ["modern_sqlite"]
+series = ["vtab"]
+session = [
+ "libsqlite3-sys/session",
+ "hooks",
+]
+sqlcipher = ["libsqlite3-sys/sqlcipher"]
+trace = []
+unlock_notify = ["libsqlite3-sys/unlock_notify"]
+vtab = []
+wasm32-wasi-vfs = ["libsqlite3-sys/wasm32-wasi-vfs"]
+window = ["functions"]
+winsqlite3 = ["libsqlite3-sys/winsqlite3"]
+with-asan = ["libsqlite3-sys/with-asan"]
+
+[badges.appveyor]
+repository = "rusqlite/rusqlite"
+
+[badges.codecov]
+repository = "rusqlite/rusqlite"
+
+[badges.maintenance]
+status = "actively-developed"
diff --git a/third_party/rust/rusqlite/LICENSE b/third_party/rust/rusqlite/LICENSE
new file mode 100644
index 0000000000..9e5b9f721c
--- /dev/null
+++ b/third_party/rust/rusqlite/LICENSE
@@ -0,0 +1,19 @@
+Copyright (c) 2014-2021 The rusqlite developers
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.
diff --git a/third_party/rust/rusqlite/README.md b/third_party/rust/rusqlite/README.md
new file mode 100644
index 0000000000..13db5ff67d
--- /dev/null
+++ b/third_party/rust/rusqlite/README.md
@@ -0,0 +1,248 @@
+# Rusqlite
+
+[![Latest Version](https://img.shields.io/crates/v/rusqlite.svg)](https://crates.io/crates/rusqlite)
+[![Documentation](https://docs.rs/rusqlite/badge.svg)](https://docs.rs/rusqlite)
+[![Build Status (GitHub)](https://github.com/rusqlite/rusqlite/workflows/CI/badge.svg)](https://github.com/rusqlite/rusqlite/actions)
+[![Build Status (AppVeyor)](https://ci.appveyor.com/api/projects/status/github/rusqlite/rusqlite?branch=master&svg=true)](https://ci.appveyor.com/project/rusqlite/rusqlite)
+[![Code Coverage](https://codecov.io/gh/rusqlite/rusqlite/branch/master/graph/badge.svg)](https://codecov.io/gh/rusqlite/rusqlite)
+[![Dependency Status](https://deps.rs/repo/github/rusqlite/rusqlite/status.svg)](https://deps.rs/repo/github/rusqlite/rusqlite)
+[![Discord Chat](https://img.shields.io/discord/927966344266256434.svg?logo=discord)](https://discord.gg/nFYfGPB8g4)
+
+Rusqlite is an ergonomic wrapper for using SQLite from Rust.
+
+Historically, the API was based on the one from [`rust-postgres`](https://github.com/sfackler/rust-postgres). However, the two have diverged in many ways, and no compatibility between the two is intended.
+
+## Usage
+
+In your Cargo.toml:
+
+```toml
+[dependencies]
+# `bundled` causes us to automatically compile and link in an up to date
+# version of SQLite for you. This avoids many common build issues, and
+# avoids depending on the version of SQLite on the users system (or your
+# system), which may be old or missing. It's the right choice for most
+# programs that control their own SQLite databases.
+#
+# That said, it's not ideal for all scenarios and in particular, generic
+# libraries built around `rusqlite` should probably not enable it, which
+# is why it is not a default feature -- it could become hard to disable.
+rusqlite = { version = "0.30.0", features = ["bundled"] }
+```
+
+Simple example usage:
+
+```rust
+use rusqlite::{Connection, Result};
+
+#[derive(Debug)]
+struct Person {
+ id: i32,
+ name: String,
+ data: Option<Vec<u8>>,
+}
+
+fn main() -> Result<()> {
+ let conn = Connection::open_in_memory()?;
+
+ conn.execute(
+ "CREATE TABLE person (
+ id INTEGER PRIMARY KEY,
+ name TEXT NOT NULL,
+ data BLOB
+ )",
+ (), // empty list of parameters.
+ )?;
+ let me = Person {
+ id: 0,
+ name: "Steven".to_string(),
+ data: None,
+ };
+ conn.execute(
+ "INSERT INTO person (name, data) VALUES (?1, ?2)",
+ (&me.name, &me.data),
+ )?;
+
+ let mut stmt = conn.prepare("SELECT id, name, data FROM person")?;
+ let person_iter = stmt.query_map([], |row| {
+ Ok(Person {
+ id: row.get(0)?,
+ name: row.get(1)?,
+ data: row.get(2)?,
+ })
+ })?;
+
+ for person in person_iter {
+ println!("Found person {:?}", person.unwrap());
+ }
+ Ok(())
+}
+```
+
+### Supported SQLite Versions
+
+The base `rusqlite` package supports SQLite version 3.14.0 or newer. If you need
+support for older versions, please file an issue. Some cargo features require a
+newer SQLite version; see details below.
+
+### Optional Features
+
+Rusqlite provides several features that are behind [Cargo
+features](https://doc.rust-lang.org/cargo/reference/manifest.html#the-features-section). They are:
+
+* [`load_extension`](https://docs.rs/rusqlite/~0/rusqlite/struct.LoadExtensionGuard.html)
+ allows loading dynamic library-based SQLite extensions.
+* [`backup`](https://docs.rs/rusqlite/~0/rusqlite/backup/index.html)
+ allows use of SQLite's online backup API. Note: This feature requires SQLite 3.6.11 or later.
+* [`functions`](https://docs.rs/rusqlite/~0/rusqlite/functions/index.html)
+ allows you to load Rust closures into SQLite connections for use in queries.
+ Note: This feature requires SQLite 3.7.3 or later.
+* `window` for [window function](https://www.sqlite.org/windowfunctions.html) support (`fun(...) OVER ...`). (Implies `functions`.)
+* [`trace`](https://docs.rs/rusqlite/~0/rusqlite/trace/index.html)
+ allows hooks into SQLite's tracing and profiling APIs. Note: This feature
+ requires SQLite 3.6.23 or later.
+* [`blob`](https://docs.rs/rusqlite/~0/rusqlite/blob/index.html)
+ gives `std::io::{Read, Write, Seek}` access to SQL BLOBs. Note: This feature
+ requires SQLite 3.7.4 or later.
+* [`limits`](https://docs.rs/rusqlite/~0/rusqlite/struct.Connection.html#method.limit)
+ allows you to set and retrieve SQLite's per connection limits.
+* `chrono` implements [`FromSql`](https://docs.rs/rusqlite/~0/rusqlite/types/trait.FromSql.html)
+ and [`ToSql`](https://docs.rs/rusqlite/~0/rusqlite/types/trait.ToSql.html) for various
+ types from the [`chrono` crate](https://crates.io/crates/chrono).
+* `serde_json` implements [`FromSql`](https://docs.rs/rusqlite/~0/rusqlite/types/trait.FromSql.html)
+ and [`ToSql`](https://docs.rs/rusqlite/~0/rusqlite/types/trait.ToSql.html) for the
+ `Value` type from the [`serde_json` crate](https://crates.io/crates/serde_json).
+* `time` implements [`FromSql`](https://docs.rs/rusqlite/~0/rusqlite/types/trait.FromSql.html)
+ and [`ToSql`](https://docs.rs/rusqlite/~0/rusqlite/types/trait.ToSql.html) for various
+ types from the [`time` crate](https://crates.io/crates/time).
+* `url` implements [`FromSql`](https://docs.rs/rusqlite/~0/rusqlite/types/trait.FromSql.html)
+ and [`ToSql`](https://docs.rs/rusqlite/~0/rusqlite/types/trait.ToSql.html) for the
+ `Url` type from the [`url` crate](https://crates.io/crates/url).
+* `bundled` uses a bundled version of SQLite. This is a good option for cases where linking to SQLite is complicated, such as Windows.
+* `sqlcipher` looks for the SQLCipher library to link against instead of SQLite. This feature overrides `bundled`.
+* `bundled-sqlcipher` uses a bundled version of SQLCipher. This searches for and links against a system-installed crypto library to provide the crypto implementation.
+* `bundled-sqlcipher-vendored-openssl` allows using bundled-sqlcipher with a vendored version of OpenSSL (via the `openssl-sys` crate) as the crypto provider.
+ - As the name implies this depends on the `bundled-sqlcipher` feature, and automatically turns it on.
+ - If turned on, this uses the [`openssl-sys`](https://crates.io/crates/openssl-sys) crate, with the `vendored` feature enabled in order to build and bundle the OpenSSL crypto library.
+* `hooks` for [Commit, Rollback](http://sqlite.org/c3ref/commit_hook.html) and [Data Change](http://sqlite.org/c3ref/update_hook.html) notification callbacks.
+* `unlock_notify` for [Unlock](https://sqlite.org/unlock_notify.html) notification.
+* `vtab` for [virtual table](https://sqlite.org/vtab.html) support (allows you to write virtual table implementations in Rust). Currently, only read-only virtual tables are supported.
+* `series` exposes [`generate_series(...)`](https://www.sqlite.org/series.html) Table-Valued Function. (Implies `vtab`.)
+* [`csvtab`](https://sqlite.org/csv.html), CSV virtual table written in Rust. (Implies `vtab`.)
+* [`array`](https://sqlite.org/carray.html), The `rarray()` Table-Valued Function. (Implies `vtab`.)
+* `i128_blob` allows storing values of type `i128` type in SQLite databases. Internally, the data is stored as a 16 byte big-endian blob, with the most significant bit flipped, which allows ordering and comparison between different blobs storing i128s to work as expected.
+* `uuid` allows storing and retrieving `Uuid` values from the [`uuid`](https://docs.rs/uuid/) crate using blobs.
+* [`session`](https://sqlite.org/sessionintro.html), Session module extension. Requires `buildtime_bindgen` feature. (Implies `hooks`.)
+* `extra_check` fail when a query passed to execute is readonly or has a column count > 0.
+* `column_decltype` provides `columns()` method for Statements and Rows; omit if linking to a version of SQLite/SQLCipher compiled with `-DSQLITE_OMIT_DECLTYPE`.
+* `collation` exposes [`sqlite3_create_collation_v2`](https://sqlite.org/c3ref/create_collation.html).
+* `winsqlite3` allows linking against the SQLite present in newer versions of Windows
+
+## Notes on building rusqlite and libsqlite3-sys
+
+`libsqlite3-sys` is a separate crate from `rusqlite` that provides the Rust
+declarations for SQLite's C API. By default, `libsqlite3-sys` attempts to find a SQLite library that already exists on your system using pkg-config, or a
+[Vcpkg](https://github.com/Microsoft/vcpkg) installation for MSVC ABI builds.
+
+You can adjust this behavior in a number of ways:
+
+* If you use the `bundled`, `bundled-sqlcipher`, or `bundled-sqlcipher-vendored-openssl` features, `libsqlite3-sys` will use the
+ [cc](https://crates.io/crates/cc) crate to compile SQLite or SQLCipher from source and
+ link against that. This source is embedded in the `libsqlite3-sys` crate and
+ is currently SQLite 3.44.0 (as of `rusqlite` 0.30.0 / `libsqlite3-sys`
+ 0.27.0). This is probably the simplest solution to any build problems. You can enable this by adding the following in your `Cargo.toml` file:
+ ```toml
+ [dependencies.rusqlite]
+ version = "0.30.0"
+ features = ["bundled"]
+ ```
+* When using any of the `bundled` features, the build script will honor `SQLITE_MAX_VARIABLE_NUMBER` and `SQLITE_MAX_EXPR_DEPTH` variables. It will also honor a `LIBSQLITE3_FLAGS` variable, which can have a format like `"-USQLITE_ALPHA -DSQLITE_BETA SQLITE_GAMMA ..."`. That would disable the `SQLITE_ALPHA` flag, and set the `SQLITE_BETA` and `SQLITE_GAMMA` flags. (The initial `-D` can be omitted, as on the last one.)
+* When using `bundled-sqlcipher` (and not also using `bundled-sqlcipher-vendored-openssl`), `libsqlite3-sys` will need to
+ link against crypto libraries on the system. If the build script can find a `libcrypto` from OpenSSL or LibreSSL (it will consult `OPENSSL_LIB_DIR`/`OPENSSL_INCLUDE_DIR` and `OPENSSL_DIR` environment variables), it will use that. If building on and for Macs, and none of those variables are set, it will use the system's SecurityFramework instead.
+
+* When linking against a SQLite (or SQLCipher) library already on the system (so *not* using any of the `bundled` features), you can set the `SQLITE3_LIB_DIR` (or `SQLCIPHER_LIB_DIR`) environment variable to point to a directory containing the library. You can also set the `SQLITE3_INCLUDE_DIR` (or `SQLCIPHER_INCLUDE_DIR`) variable to point to the directory containing `sqlite3.h`.
+* Installing the sqlite3 development packages will usually be all that is required, but
+ the build helpers for [pkg-config](https://github.com/alexcrichton/pkg-config-rs)
+ and [vcpkg](https://github.com/mcgoo/vcpkg-rs) have some additional configuration
+ options. The default when using vcpkg is to dynamically link,
+ which must be enabled by setting `VCPKGRS_DYNAMIC=1` environment variable before build.
+ `vcpkg install sqlite3:x64-windows` will install the required library.
+* When linking against a SQLite (or SQLCipher) library already on the system, you can set the `SQLITE3_STATIC` (or `SQLCIPHER_STATIC`) environment variable to 1 to request that the library be statically instead of dynamically linked.
+
+
+### Binding generation
+
+We use [bindgen](https://crates.io/crates/bindgen) to generate the Rust
+declarations from SQLite's C header file. `bindgen`
+[recommends](https://github.com/servo/rust-bindgen#library-usage-with-buildrs)
+running this as part of the build process of libraries that used this. We tried
+this briefly (`rusqlite` 0.10.0, specifically), but it had some annoyances:
+
+* The build time for `libsqlite3-sys` (and therefore `rusqlite`) increased
+ dramatically.
+* Running `bindgen` requires a relatively-recent version of Clang, which many
+ systems do not have installed by default.
+* Running `bindgen` also requires the SQLite header file to be present.
+
+As of `rusqlite` 0.10.1, we avoid running `bindgen` at build-time by shipping
+pregenerated bindings for several versions of SQLite. When compiling
+`rusqlite`, we use your selected Cargo features to pick the bindings for the
+minimum SQLite version that supports your chosen features. If you are using
+`libsqlite3-sys` directly, you can use the same features to choose which
+pregenerated bindings are chosen:
+
+* `min_sqlite_version_3_14_0` - SQLite 3.14.0 bindings (this is the default)
+
+If you use any of the `bundled` features, you will get pregenerated bindings for the
+bundled version of SQLite/SQLCipher. If you need other specific pregenerated binding
+versions, please file an issue. If you want to run `bindgen` at buildtime to
+produce your own bindings, use the `buildtime_bindgen` Cargo feature.
+
+If you enable the `modern_sqlite` feature, we'll use the bindings we would have
+included with the bundled build. You generally should have `buildtime_bindgen`
+enabled if you turn this on, as otherwise you'll need to keep the version of
+SQLite you link with in sync with what rusqlite would have bundled, (usually the
+most recent release of SQLite). Failing to do this will cause a runtime error.
+
+## Contributing
+
+Rusqlite has many features, and many of them impact the build configuration in
+incompatible ways. This is unfortunate, and makes testing changes hard.
+
+To help here: you generally should ensure that you run tests/lint for
+`--features bundled`, and `--features "bundled-full session buildtime_bindgen"`.
+
+If running bindgen is problematic for you, `--features bundled-full` enables
+bundled and all features which don't require binding generation, and can be used
+instead.
+
+### Checklist
+
+- Run `cargo fmt` to ensure your Rust code is correctly formatted.
+- Ensure `cargo clippy --workspace --features bundled` passes without warnings.
+- Ensure `cargo clippy --workspace --features "bundled-full session buildtime_bindgen"` passes without warnings.
+- Ensure `cargo test --workspace --features bundled` reports no failures.
+- Ensure `cargo test --workspace --features "bundled-full session buildtime_bindgen"` reports no failures.
+
+## Author
+
+Rusqlite is the product of hard work by a number of people. A list is available
+here: https://github.com/rusqlite/rusqlite/graphs/contributors
+
+## Community
+
+Feel free to join the [Rusqlite Discord Server](https://discord.gg/nFYfGPB8g4) to discuss or get help with `rusqlite` or `libsqlite3-sys`.
+
+## License
+
+Rusqlite and libsqlite3-sys are available under the MIT license. See the LICENSE file for more info.
+
+### Licenses of Bundled Software
+
+Depending on the set of enabled cargo `features`, rusqlite and libsqlite3-sys will also bundle other libraries, which have their own licensing terms:
+
+- If `--features=bundled-sqlcipher` is enabled, the vendored source of [SQLcipher](https://github.com/sqlcipher/sqlcipher) will be compiled and statically linked in. SQLcipher is distributed under a BSD-style license, as described [here](libsqlite3-sys/sqlcipher/LICENSE).
+
+- If `--features=bundled` is enabled, the vendored source of SQLite will be compiled and linked in. SQLite is in the public domain, as described [here](https://www.sqlite.org/copyright.html).
+
+Both of these are quite permissive, have no bearing on the license of the code in `rusqlite` or `libsqlite3-sys` themselves, and can be entirely ignored if you do not use the feature in question.
diff --git a/third_party/rust/rusqlite/benches/cache.rs b/third_party/rust/rusqlite/benches/cache.rs
new file mode 100644
index 0000000000..dd3683ec72
--- /dev/null
+++ b/third_party/rust/rusqlite/benches/cache.rs
@@ -0,0 +1,18 @@
+use bencher::{benchmark_group, benchmark_main, Bencher};
+use rusqlite::Connection;
+
+fn bench_no_cache(b: &mut Bencher) {
+ let db = Connection::open_in_memory().unwrap();
+ db.set_prepared_statement_cache_capacity(0);
+ let sql = "SELECT 1, 'test', 3.14 UNION SELECT 2, 'exp', 2.71";
+ b.iter(|| db.prepare(sql).unwrap());
+}
+
+fn bench_cache(b: &mut Bencher) {
+ let db = Connection::open_in_memory().unwrap();
+ let sql = "SELECT 1, 'test', 3.14 UNION SELECT 2, 'exp', 2.71";
+ b.iter(|| db.prepare_cached(sql).unwrap());
+}
+
+benchmark_group!(cache_benches, bench_no_cache, bench_cache);
+benchmark_main!(cache_benches);
diff --git a/third_party/rust/rusqlite/benches/exec.rs b/third_party/rust/rusqlite/benches/exec.rs
new file mode 100644
index 0000000000..b95cb354ee
--- /dev/null
+++ b/third_party/rust/rusqlite/benches/exec.rs
@@ -0,0 +1,17 @@
+use bencher::{benchmark_group, benchmark_main, Bencher};
+use rusqlite::Connection;
+
+fn bench_execute(b: &mut Bencher) {
+ let db = Connection::open_in_memory().unwrap();
+ let sql = "PRAGMA user_version=1";
+ b.iter(|| db.execute(sql, []).unwrap());
+}
+
+fn bench_execute_batch(b: &mut Bencher) {
+ let db = Connection::open_in_memory().unwrap();
+ let sql = "PRAGMA user_version=1";
+ b.iter(|| db.execute_batch(sql).unwrap());
+}
+
+benchmark_group!(exec_benches, bench_execute, bench_execute_batch);
+benchmark_main!(exec_benches);
diff --git a/third_party/rust/rusqlite/examples/load_extension.rs b/third_party/rust/rusqlite/examples/load_extension.rs
new file mode 100644
index 0000000000..9e52bb202f
--- /dev/null
+++ b/third_party/rust/rusqlite/examples/load_extension.rs
@@ -0,0 +1,22 @@
+//! Ensure loadable_extension.rs works.
+
+use rusqlite::{Connection, Result};
+
+fn main() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ unsafe {
+ db.load_extension_enable()?;
+ #[cfg(not(windows))]
+ db.load_extension("target/debug/examples/libloadable_extension", None)?;
+ #[cfg(windows)]
+ db.load_extension("target/debug/examples/loadable_extension", None)?;
+ db.load_extension_disable()?;
+ }
+
+ let str = db.query_row("SELECT rusqlite_test_function()", [], |row| {
+ row.get::<_, String>(0)
+ })?;
+ assert_eq!(&str, "Rusqlite extension loaded correctly!");
+ Ok(())
+}
diff --git a/third_party/rust/rusqlite/examples/loadable_extension.rs b/third_party/rust/rusqlite/examples/loadable_extension.rs
new file mode 100644
index 0000000000..e913240573
--- /dev/null
+++ b/third_party/rust/rusqlite/examples/loadable_extension.rs
@@ -0,0 +1,50 @@
+//! Adaptation of https://sqlite.org/loadext.html#programming_loadable_extensions
+use std::os::raw::{c_char, c_int};
+
+use rusqlite::ffi;
+use rusqlite::functions::FunctionFlags;
+use rusqlite::types::{ToSqlOutput, Value};
+use rusqlite::{to_sqlite_error, Connection, Result};
+
+/// # build
+/// ```sh
+/// cargo build --example loadable_extension --features "loadable_extension modern_sqlite functions vtab trace"
+/// ```
+/// # test
+/// ```sh
+/// sqlite> .log on
+/// sqlite> .load target/debug/examples/libloadable_extension.so
+/// (28) Rusqlite extension initialized
+/// sqlite> SELECT rusqlite_test_function();
+/// Rusqlite extension loaded correctly!
+/// ```
+#[allow(clippy::not_unsafe_ptr_arg_deref)]
+#[no_mangle]
+pub extern "C" fn sqlite3_extension_init(
+ db: *mut ffi::sqlite3,
+ pz_err_msg: *mut *mut c_char,
+ p_api: *mut ffi::sqlite3_api_routines,
+) -> c_int {
+ if p_api.is_null() {
+ return ffi::SQLITE_ERROR;
+ } else if let Err(err) = extension_init(db, p_api) {
+ return unsafe { to_sqlite_error(&err, pz_err_msg) };
+ }
+ ffi::SQLITE_OK
+}
+
+fn extension_init(db: *mut ffi::sqlite3, p_api: *mut ffi::sqlite3_api_routines) -> Result<()> {
+ let db = unsafe { Connection::extension_init2(db, p_api)? };
+ db.create_scalar_function(
+ "rusqlite_test_function",
+ 0,
+ FunctionFlags::SQLITE_DETERMINISTIC,
+ |_ctx| {
+ Ok(ToSqlOutput::Owned(Value::Text(
+ "Rusqlite extension loaded correctly!".to_string(),
+ )))
+ },
+ )?;
+ rusqlite::trace::log(ffi::SQLITE_WARNING, "Rusqlite extension initialized");
+ Ok(())
+}
diff --git a/third_party/rust/rusqlite/src/backup.rs b/third_party/rust/rusqlite/src/backup.rs
new file mode 100644
index 0000000000..f28ae9a8ac
--- /dev/null
+++ b/third_party/rust/rusqlite/src/backup.rs
@@ -0,0 +1,428 @@
+//! Online SQLite backup API.
+//!
+//! To create a [`Backup`], you must have two distinct [`Connection`]s - one
+//! for the source (which can be used while the backup is running) and one for
+//! the destination (which cannot). A [`Backup`] handle exposes three methods:
+//! [`step`](Backup::step) will attempt to back up a specified number of pages,
+//! [`progress`](Backup::progress) gets the current progress of the backup as of
+//! the last call to [`step`](Backup::step), and
+//! [`run_to_completion`](Backup::run_to_completion) will attempt to back up the
+//! entire source database, allowing you to specify how many pages are backed up
+//! at a time and how long the thread should sleep between chunks of pages.
+//!
+//! The following example is equivalent to "Example 2: Online Backup of a
+//! Running Database" from [SQLite's Online Backup API
+//! documentation](https://www.sqlite.org/backup.html).
+//!
+//! ```rust,no_run
+//! # use rusqlite::{backup, Connection, Result};
+//! # use std::path::Path;
+//! # use std::time;
+//!
+//! fn backup_db<P: AsRef<Path>>(
+//! src: &Connection,
+//! dst: P,
+//! progress: fn(backup::Progress),
+//! ) -> Result<()> {
+//! let mut dst = Connection::open(dst)?;
+//! let backup = backup::Backup::new(src, &mut dst)?;
+//! backup.run_to_completion(5, time::Duration::from_millis(250), Some(progress))
+//! }
+//! ```
+
+use std::marker::PhantomData;
+use std::path::Path;
+use std::ptr;
+
+use std::os::raw::c_int;
+use std::thread;
+use std::time::Duration;
+
+use crate::ffi;
+
+use crate::error::error_from_handle;
+use crate::{Connection, DatabaseName, Result};
+
+impl Connection {
+ /// Back up the `name` database to the given
+ /// destination path.
+ ///
+ /// If `progress` is not `None`, it will be called periodically
+ /// until the backup completes.
+ ///
+ /// For more fine-grained control over the backup process (e.g.,
+ /// to sleep periodically during the backup or to back up to an
+ /// already-open database connection), see the `backup` module.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the destination path cannot be opened
+ /// or if the backup fails.
+ pub fn backup<P: AsRef<Path>>(
+ &self,
+ name: DatabaseName<'_>,
+ dst_path: P,
+ progress: Option<fn(Progress)>,
+ ) -> Result<()> {
+ use self::StepResult::{Busy, Done, Locked, More};
+ let mut dst = Connection::open(dst_path)?;
+ let backup = Backup::new_with_names(self, name, &mut dst, DatabaseName::Main)?;
+
+ let mut r = More;
+ while r == More {
+ r = backup.step(100)?;
+ if let Some(f) = progress {
+ f(backup.progress());
+ }
+ }
+
+ match r {
+ Done => Ok(()),
+ Busy => Err(unsafe { error_from_handle(ptr::null_mut(), ffi::SQLITE_BUSY) }),
+ Locked => Err(unsafe { error_from_handle(ptr::null_mut(), ffi::SQLITE_LOCKED) }),
+ More => unreachable!(),
+ }
+ }
+
+ /// Restore the given source path into the
+ /// `name` database. If `progress` is not `None`, it will be
+ /// called periodically until the restore completes.
+ ///
+ /// For more fine-grained control over the restore process (e.g.,
+ /// to sleep periodically during the restore or to restore from an
+ /// already-open database connection), see the `backup` module.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the destination path cannot be opened
+ /// or if the restore fails.
+ pub fn restore<P: AsRef<Path>, F: Fn(Progress)>(
+ &mut self,
+ name: DatabaseName<'_>,
+ src_path: P,
+ progress: Option<F>,
+ ) -> Result<()> {
+ use self::StepResult::{Busy, Done, Locked, More};
+ let src = Connection::open(src_path)?;
+ let restore = Backup::new_with_names(&src, DatabaseName::Main, self, name)?;
+
+ let mut r = More;
+ let mut busy_count = 0_i32;
+ 'restore_loop: while r == More || r == Busy {
+ r = restore.step(100)?;
+ if let Some(ref f) = progress {
+ f(restore.progress());
+ }
+ if r == Busy {
+ busy_count += 1;
+ if busy_count >= 3 {
+ break 'restore_loop;
+ }
+ thread::sleep(Duration::from_millis(100));
+ }
+ }
+
+ match r {
+ Done => Ok(()),
+ Busy => Err(unsafe { error_from_handle(ptr::null_mut(), ffi::SQLITE_BUSY) }),
+ Locked => Err(unsafe { error_from_handle(ptr::null_mut(), ffi::SQLITE_LOCKED) }),
+ More => unreachable!(),
+ }
+ }
+}
+
+/// Possible successful results of calling
+/// [`Backup::step`].
+#[derive(Copy, Clone, Debug, PartialEq, Eq)]
+#[non_exhaustive]
+pub enum StepResult {
+ /// The backup is complete.
+ Done,
+
+ /// The step was successful but there are still more pages that need to be
+ /// backed up.
+ More,
+
+ /// The step failed because appropriate locks could not be acquired. This is
+ /// not a fatal error - the step can be retried.
+ Busy,
+
+ /// The step failed because the source connection was writing to the
+ /// database. This is not a fatal error - the step can be retried.
+ Locked,
+}
+
+/// Struct specifying the progress of a backup. The
+/// percentage completion can be calculated as `(pagecount - remaining) /
+/// pagecount`. The progress of a backup is as of the last call to
+/// [`step`](Backup::step) - if the source database is modified after a call to
+/// [`step`](Backup::step), the progress value will become outdated and
+/// potentially incorrect.
+#[derive(Copy, Clone, Debug)]
+pub struct Progress {
+ /// Number of pages in the source database that still need to be backed up.
+ pub remaining: c_int,
+ /// Total number of pages in the source database.
+ pub pagecount: c_int,
+}
+
+/// A handle to an online backup.
+pub struct Backup<'a, 'b> {
+ phantom_from: PhantomData<&'a Connection>,
+ to: &'b Connection,
+ b: *mut ffi::sqlite3_backup,
+}
+
+impl Backup<'_, '_> {
+ /// Attempt to create a new handle that will allow backups from `from` to
+ /// `to`. Note that `to` is a `&mut` - this is because SQLite forbids any
+ /// API calls on the destination of a backup while the backup is taking
+ /// place.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying `sqlite3_backup_init` call returns
+ /// `NULL`.
+ #[inline]
+ pub fn new<'a, 'b>(from: &'a Connection, to: &'b mut Connection) -> Result<Backup<'a, 'b>> {
+ Backup::new_with_names(from, DatabaseName::Main, to, DatabaseName::Main)
+ }
+
+ /// Attempt to create a new handle that will allow backups from the
+ /// `from_name` database of `from` to the `to_name` database of `to`. Note
+ /// that `to` is a `&mut` - this is because SQLite forbids any API calls on
+ /// the destination of a backup while the backup is taking place.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying `sqlite3_backup_init` call returns
+ /// `NULL`.
+ pub fn new_with_names<'a, 'b>(
+ from: &'a Connection,
+ from_name: DatabaseName<'_>,
+ to: &'b mut Connection,
+ to_name: DatabaseName<'_>,
+ ) -> Result<Backup<'a, 'b>> {
+ let to_name = to_name.as_cstring()?;
+ let from_name = from_name.as_cstring()?;
+
+ let to_db = to.db.borrow_mut().db;
+
+ let b = unsafe {
+ let b = ffi::sqlite3_backup_init(
+ to_db,
+ to_name.as_ptr(),
+ from.db.borrow_mut().db,
+ from_name.as_ptr(),
+ );
+ if b.is_null() {
+ return Err(error_from_handle(to_db, ffi::sqlite3_errcode(to_db)));
+ }
+ b
+ };
+
+ Ok(Backup {
+ phantom_from: PhantomData,
+ to,
+ b,
+ })
+ }
+
+ /// Gets the progress of the backup as of the last call to
+ /// [`step`](Backup::step).
+ #[inline]
+ #[must_use]
+ pub fn progress(&self) -> Progress {
+ unsafe {
+ Progress {
+ remaining: ffi::sqlite3_backup_remaining(self.b),
+ pagecount: ffi::sqlite3_backup_pagecount(self.b),
+ }
+ }
+ }
+
+ /// Attempts to back up the given number of pages. If `num_pages` is
+ /// negative, will attempt to back up all remaining pages. This will hold a
+ /// lock on the source database for the duration, so it is probably not
+ /// what you want for databases that are currently active (see
+ /// [`run_to_completion`](Backup::run_to_completion) for a better
+ /// alternative).
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying `sqlite3_backup_step` call returns
+ /// an error code other than `DONE`, `OK`, `BUSY`, or `LOCKED`. `BUSY` and
+ /// `LOCKED` are transient errors and are therefore returned as possible
+ /// `Ok` values.
+ #[inline]
+ pub fn step(&self, num_pages: c_int) -> Result<StepResult> {
+ use self::StepResult::{Busy, Done, Locked, More};
+
+ let rc = unsafe { ffi::sqlite3_backup_step(self.b, num_pages) };
+ match rc {
+ ffi::SQLITE_DONE => Ok(Done),
+ ffi::SQLITE_OK => Ok(More),
+ ffi::SQLITE_BUSY => Ok(Busy),
+ ffi::SQLITE_LOCKED => Ok(Locked),
+ _ => self.to.decode_result(rc).map(|_| More),
+ }
+ }
+
+ /// Attempts to run the entire backup. Will call
+ /// [`step(pages_per_step)`](Backup::step) as many times as necessary,
+ /// sleeping for `pause_between_pages` between each call to give the
+ /// source database time to process any pending queries. This is a
+ /// direct implementation of "Example 2: Online Backup of a Running
+ /// Database" from [SQLite's Online Backup API documentation](https://www.sqlite.org/backup.html).
+ ///
+ /// If `progress` is not `None`, it will be called after each step with the
+ /// current progress of the backup. Note that is possible the progress may
+ /// not change if the step returns `Busy` or `Locked` even though the
+ /// backup is still running.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if any of the calls to [`step`](Backup::step) return
+ /// `Err`.
+ pub fn run_to_completion(
+ &self,
+ pages_per_step: c_int,
+ pause_between_pages: Duration,
+ progress: Option<fn(Progress)>,
+ ) -> Result<()> {
+ use self::StepResult::{Busy, Done, Locked, More};
+
+ assert!(pages_per_step > 0, "pages_per_step must be positive");
+
+ loop {
+ let r = self.step(pages_per_step)?;
+ if let Some(progress) = progress {
+ progress(self.progress());
+ }
+ match r {
+ More | Busy | Locked => thread::sleep(pause_between_pages),
+ Done => return Ok(()),
+ }
+ }
+ }
+}
+
+impl Drop for Backup<'_, '_> {
+ #[inline]
+ fn drop(&mut self) {
+ unsafe { ffi::sqlite3_backup_finish(self.b) };
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use super::Backup;
+ use crate::{Connection, DatabaseName, Result};
+ use std::time::Duration;
+
+ #[test]
+ fn test_backup() -> Result<()> {
+ let src = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER);
+ INSERT INTO foo VALUES(42);
+ END;";
+ src.execute_batch(sql)?;
+
+ let mut dst = Connection::open_in_memory()?;
+
+ {
+ let backup = Backup::new(&src, &mut dst)?;
+ backup.step(-1)?;
+ }
+
+ let the_answer: i64 = dst.one_column("SELECT x FROM foo")?;
+ assert_eq!(42, the_answer);
+
+ src.execute_batch("INSERT INTO foo VALUES(43)")?;
+
+ {
+ let backup = Backup::new(&src, &mut dst)?;
+ backup.run_to_completion(5, Duration::from_millis(250), None)?;
+ }
+
+ let the_answer: i64 = dst.one_column("SELECT SUM(x) FROM foo")?;
+ assert_eq!(42 + 43, the_answer);
+ Ok(())
+ }
+
+ #[test]
+ fn test_backup_temp() -> Result<()> {
+ let src = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TEMPORARY TABLE foo(x INTEGER);
+ INSERT INTO foo VALUES(42);
+ END;";
+ src.execute_batch(sql)?;
+
+ let mut dst = Connection::open_in_memory()?;
+
+ {
+ let backup =
+ Backup::new_with_names(&src, DatabaseName::Temp, &mut dst, DatabaseName::Main)?;
+ backup.step(-1)?;
+ }
+
+ let the_answer: i64 = dst.one_column("SELECT x FROM foo")?;
+ assert_eq!(42, the_answer);
+
+ src.execute_batch("INSERT INTO foo VALUES(43)")?;
+
+ {
+ let backup =
+ Backup::new_with_names(&src, DatabaseName::Temp, &mut dst, DatabaseName::Main)?;
+ backup.run_to_completion(5, Duration::from_millis(250), None)?;
+ }
+
+ let the_answer: i64 = dst.one_column("SELECT SUM(x) FROM foo")?;
+ assert_eq!(42 + 43, the_answer);
+ Ok(())
+ }
+
+ #[test]
+ fn test_backup_attached() -> Result<()> {
+ let src = Connection::open_in_memory()?;
+ let sql = "ATTACH DATABASE ':memory:' AS my_attached;
+ BEGIN;
+ CREATE TABLE my_attached.foo(x INTEGER);
+ INSERT INTO my_attached.foo VALUES(42);
+ END;";
+ src.execute_batch(sql)?;
+
+ let mut dst = Connection::open_in_memory()?;
+
+ {
+ let backup = Backup::new_with_names(
+ &src,
+ DatabaseName::Attached("my_attached"),
+ &mut dst,
+ DatabaseName::Main,
+ )?;
+ backup.step(-1)?;
+ }
+
+ let the_answer: i64 = dst.one_column("SELECT x FROM foo")?;
+ assert_eq!(42, the_answer);
+
+ src.execute_batch("INSERT INTO foo VALUES(43)")?;
+
+ {
+ let backup = Backup::new_with_names(
+ &src,
+ DatabaseName::Attached("my_attached"),
+ &mut dst,
+ DatabaseName::Main,
+ )?;
+ backup.run_to_completion(5, Duration::from_millis(250), None)?;
+ }
+
+ let the_answer: i64 = dst.one_column("SELECT SUM(x) FROM foo")?;
+ assert_eq!(42 + 43, the_answer);
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/blob/mod.rs b/third_party/rust/rusqlite/src/blob/mod.rs
new file mode 100644
index 0000000000..c9e797bc0f
--- /dev/null
+++ b/third_party/rust/rusqlite/src/blob/mod.rs
@@ -0,0 +1,550 @@
+//! Incremental BLOB I/O.
+//!
+//! Note that SQLite does not provide API-level access to change the size of a
+//! BLOB; that must be performed through SQL statements.
+//!
+//! There are two choices for how to perform IO on a [`Blob`].
+//!
+//! 1. The implementations it provides of the `std::io::Read`, `std::io::Write`,
+//! and `std::io::Seek` traits.
+//!
+//! 2. A positional IO API, e.g. [`Blob::read_at`], [`Blob::write_at`] and
+//! similar.
+//!
+//! Documenting these in order:
+//!
+//! ## 1. `std::io` trait implementations.
+//!
+//! `Blob` conforms to `std::io::Read`, `std::io::Write`, and `std::io::Seek`,
+//! so it plays nicely with other types that build on these (such as
+//! `std::io::BufReader` and `std::io::BufWriter`). However, you must be careful
+//! with the size of the blob. For example, when using a `BufWriter`, the
+//! `BufWriter` will accept more data than the `Blob` will allow, so make sure
+//! to call `flush` and check for errors. (See the unit tests in this module for
+//! an example.)
+//!
+//! ## 2. Positional IO
+//!
+//! `Blob`s also offer a `pread` / `pwrite`-style positional IO api in the form
+//! of [`Blob::read_at`], [`Blob::write_at`], [`Blob::raw_read_at`],
+//! [`Blob::read_at_exact`], and [`Blob::raw_read_at_exact`].
+//!
+//! These APIs all take the position to read from or write to from as a
+//! parameter, instead of using an internal `pos` value.
+//!
+//! ### Positional IO Read Variants
+//!
+//! For the `read` functions, there are several functions provided:
+//!
+//! - [`Blob::read_at`]
+//! - [`Blob::raw_read_at`]
+//! - [`Blob::read_at_exact`]
+//! - [`Blob::raw_read_at_exact`]
+//!
+//! These can be divided along two axes: raw/not raw, and exact/inexact:
+//!
+//! 1. Raw/not raw refers to the type of the destination buffer. The raw
+//! functions take a `&mut [MaybeUninit<u8>]` as the destination buffer,
+//! where the "normal" functions take a `&mut [u8]`.
+//!
+//! Using `MaybeUninit` here can be more efficient in some cases, but is
+//! often inconvenient, so both are provided.
+//!
+//! 2. Exact/inexact refers to to whether or not the entire buffer must be
+//! filled in order for the call to be considered a success.
+//!
+//! The "exact" functions require the provided buffer be entirely filled, or
+//! they return an error, whereas the "inexact" functions read as much out of
+//! the blob as is available, and return how much they were able to read.
+//!
+//! The inexact functions are preferable if you do not know the size of the
+//! blob already, and the exact functions are preferable if you do.
+//!
+//! ### Comparison to using the `std::io` traits:
+//!
+//! In general, the positional methods offer the following Pro/Cons compared to
+//! using the implementation `std::io::{Read, Write, Seek}` we provide for
+//! `Blob`:
+//!
+//! 1. (Pro) There is no need to first seek to a position in order to perform IO
+//! on it as the position is a parameter.
+//!
+//! 2. (Pro) `Blob`'s positional read functions don't mutate the blob in any
+//! way, and take `&self`. No `&mut` access required.
+//!
+//! 3. (Pro) Positional IO functions return `Err(rusqlite::Error)` on failure,
+//! rather than `Err(std::io::Error)`. Returning `rusqlite::Error` is more
+//! accurate and convenient.
+//!
+//! Note that for the `std::io` API, no data is lost however, and it can be
+//! recovered with `io_err.downcast::<rusqlite::Error>()` (this can be easy
+//! to forget, though).
+//!
+//! 4. (Pro, for now). A `raw` version of the read API exists which can allow
+//! reading into a `&mut [MaybeUninit<u8>]` buffer, which avoids a potential
+//! costly initialization step. (However, `std::io` traits will certainly
+//! gain this someday, which is why this is only a "Pro, for now").
+//!
+//! 5. (Con) The set of functions is more bare-bones than what is offered in
+//! `std::io`, which has a number of adapters, handy algorithms, further
+//! traits.
+//!
+//! 6. (Con) No meaningful interoperability with other crates, so if you need
+//! that you must use `std::io`.
+//!
+//! To generalize: the `std::io` traits are useful because they conform to a
+//! standard interface that a lot of code knows how to handle, however that
+//! interface is not a perfect fit for [`Blob`], so another small set of
+//! functions is provided as well.
+//!
+//! # Example (`std::io`)
+//!
+//! ```rust
+//! # use rusqlite::blob::ZeroBlob;
+//! # use rusqlite::{Connection, DatabaseName};
+//! # use std::error::Error;
+//! # use std::io::{Read, Seek, SeekFrom, Write};
+//! # fn main() -> Result<(), Box<dyn Error>> {
+//! let db = Connection::open_in_memory()?;
+//! db.execute_batch("CREATE TABLE test_table (content BLOB);")?;
+//!
+//! // Insert a BLOB into the `content` column of `test_table`. Note that the Blob
+//! // I/O API provides no way of inserting or resizing BLOBs in the DB -- this
+//! // must be done via SQL.
+//! db.execute("INSERT INTO test_table (content) VALUES (ZEROBLOB(10))", [])?;
+//!
+//! // Get the row id off the BLOB we just inserted.
+//! let rowid = db.last_insert_rowid();
+//! // Open the BLOB we just inserted for IO.
+//! let mut blob = db.blob_open(DatabaseName::Main, "test_table", "content", rowid, false)?;
+//!
+//! // Write some data into the blob. Make sure to test that the number of bytes
+//! // written matches what you expect; if you try to write too much, the data
+//! // will be truncated to the size of the BLOB.
+//! let bytes_written = blob.write(b"01234567")?;
+//! assert_eq!(bytes_written, 8);
+//!
+//! // Move back to the start and read into a local buffer.
+//! // Same guidance - make sure you check the number of bytes read!
+//! blob.seek(SeekFrom::Start(0))?;
+//! let mut buf = [0u8; 20];
+//! let bytes_read = blob.read(&mut buf[..])?;
+//! assert_eq!(bytes_read, 10); // note we read 10 bytes because the blob has size 10
+//!
+//! // Insert another BLOB, this time using a parameter passed in from
+//! // rust (potentially with a dynamic size).
+//! db.execute(
+//! "INSERT INTO test_table (content) VALUES (?1)",
+//! [ZeroBlob(64)],
+//! )?;
+//!
+//! // given a new row ID, we can reopen the blob on that row
+//! let rowid = db.last_insert_rowid();
+//! blob.reopen(rowid)?;
+//! // Just check that the size is right.
+//! assert_eq!(blob.len(), 64);
+//! # Ok(())
+//! # }
+//! ```
+//!
+//! # Example (Positional)
+//!
+//! ```rust
+//! # use rusqlite::blob::ZeroBlob;
+//! # use rusqlite::{Connection, DatabaseName};
+//! # use std::error::Error;
+//! # fn main() -> Result<(), Box<dyn Error>> {
+//! let db = Connection::open_in_memory()?;
+//! db.execute_batch("CREATE TABLE test_table (content BLOB);")?;
+//! // Insert a blob into the `content` column of `test_table`. Note that the Blob
+//! // I/O API provides no way of inserting or resizing blobs in the DB -- this
+//! // must be done via SQL.
+//! db.execute("INSERT INTO test_table (content) VALUES (ZEROBLOB(10))", [])?;
+//! // Get the row id off the blob we just inserted.
+//! let rowid = db.last_insert_rowid();
+//! // Open the blob we just inserted for IO.
+//! let mut blob = db.blob_open(DatabaseName::Main, "test_table", "content", rowid, false)?;
+//! // Write some data into the blob.
+//! blob.write_at(b"ABCDEF", 2)?;
+//!
+//! // Read the whole blob into a local buffer.
+//! let mut buf = [0u8; 10];
+//! blob.read_at_exact(&mut buf, 0)?;
+//! assert_eq!(&buf, b"\0\0ABCDEF\0\0");
+//!
+//! // Insert another blob, this time using a parameter passed in from
+//! // rust (potentially with a dynamic size).
+//! db.execute(
+//! "INSERT INTO test_table (content) VALUES (?1)",
+//! [ZeroBlob(64)],
+//! )?;
+//!
+//! // given a new row ID, we can reopen the blob on that row
+//! let rowid = db.last_insert_rowid();
+//! blob.reopen(rowid)?;
+//! assert_eq!(blob.len(), 64);
+//! # Ok(())
+//! # }
+//! ```
+use std::cmp::min;
+use std::io;
+use std::ptr;
+
+use super::ffi;
+use super::types::{ToSql, ToSqlOutput};
+use crate::{Connection, DatabaseName, Result};
+
+mod pos_io;
+
+/// Handle to an open BLOB. See
+/// [`rusqlite::blob`](crate::blob) documentation for in-depth discussion.
+pub struct Blob<'conn> {
+ conn: &'conn Connection,
+ blob: *mut ffi::sqlite3_blob,
+ // used by std::io implementations,
+ pos: i32,
+}
+
+impl Connection {
+ /// Open a handle to the BLOB located in `row_id`,
+ /// `column`, `table` in database `db`.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if `db`/`table`/`column` cannot be converted to a
+ /// C-compatible string or if the underlying SQLite BLOB open call
+ /// fails.
+ #[inline]
+ pub fn blob_open<'a>(
+ &'a self,
+ db: DatabaseName<'_>,
+ table: &str,
+ column: &str,
+ row_id: i64,
+ read_only: bool,
+ ) -> Result<Blob<'a>> {
+ let c = self.db.borrow_mut();
+ let mut blob = ptr::null_mut();
+ let db = db.as_cstring()?;
+ let table = super::str_to_cstring(table)?;
+ let column = super::str_to_cstring(column)?;
+ let rc = unsafe {
+ ffi::sqlite3_blob_open(
+ c.db(),
+ db.as_ptr(),
+ table.as_ptr(),
+ column.as_ptr(),
+ row_id,
+ !read_only as std::os::raw::c_int,
+ &mut blob,
+ )
+ };
+ c.decode_result(rc).map(|_| Blob {
+ conn: self,
+ blob,
+ pos: 0,
+ })
+ }
+}
+
+impl Blob<'_> {
+ /// Move a BLOB handle to a new row.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite BLOB reopen call fails.
+ #[inline]
+ pub fn reopen(&mut self, row: i64) -> Result<()> {
+ let rc = unsafe { ffi::sqlite3_blob_reopen(self.blob, row) };
+ if rc != ffi::SQLITE_OK {
+ return self.conn.decode_result(rc);
+ }
+ self.pos = 0;
+ Ok(())
+ }
+
+ /// Return the size in bytes of the BLOB.
+ #[inline]
+ #[must_use]
+ pub fn size(&self) -> i32 {
+ unsafe { ffi::sqlite3_blob_bytes(self.blob) }
+ }
+
+ /// Return the current size in bytes of the BLOB.
+ #[inline]
+ #[must_use]
+ pub fn len(&self) -> usize {
+ self.size().try_into().unwrap()
+ }
+
+ /// Return true if the BLOB is empty.
+ #[inline]
+ #[must_use]
+ pub fn is_empty(&self) -> bool {
+ self.size() == 0
+ }
+
+ /// Close a BLOB handle.
+ ///
+ /// Calling `close` explicitly is not required (the BLOB will be closed
+ /// when the `Blob` is dropped), but it is available so you can get any
+ /// errors that occur.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite close call fails.
+ #[inline]
+ pub fn close(mut self) -> Result<()> {
+ self.close_()
+ }
+
+ #[inline]
+ fn close_(&mut self) -> Result<()> {
+ let rc = unsafe { ffi::sqlite3_blob_close(self.blob) };
+ self.blob = ptr::null_mut();
+ self.conn.decode_result(rc)
+ }
+}
+
+impl io::Read for Blob<'_> {
+ /// Read data from a BLOB incrementally. Will return Ok(0) if the end of
+ /// the blob has been reached.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite read call fails.
+ #[inline]
+ fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
+ let max_allowed_len = (self.size() - self.pos) as usize;
+ let n = min(buf.len(), max_allowed_len) as i32;
+ if n <= 0 {
+ return Ok(0);
+ }
+ let rc = unsafe { ffi::sqlite3_blob_read(self.blob, buf.as_mut_ptr().cast(), n, self.pos) };
+ self.conn
+ .decode_result(rc)
+ .map(|_| {
+ self.pos += n;
+ n as usize
+ })
+ .map_err(|err| io::Error::new(io::ErrorKind::Other, err))
+ }
+}
+
+impl io::Write for Blob<'_> {
+ /// Write data into a BLOB incrementally. Will return `Ok(0)` if the end of
+ /// the blob has been reached; consider using `Write::write_all(buf)`
+ /// if you want to get an error if the entirety of the buffer cannot be
+ /// written.
+ ///
+ /// This function may only modify the contents of the BLOB; it is not
+ /// possible to increase the size of a BLOB using this API.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite write call fails.
+ #[inline]
+ fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
+ let max_allowed_len = (self.size() - self.pos) as usize;
+ let n = min(buf.len(), max_allowed_len) as i32;
+ if n <= 0 {
+ return Ok(0);
+ }
+ let rc = unsafe { ffi::sqlite3_blob_write(self.blob, buf.as_ptr() as *mut _, n, self.pos) };
+ self.conn
+ .decode_result(rc)
+ .map(|_| {
+ self.pos += n;
+ n as usize
+ })
+ .map_err(|err| io::Error::new(io::ErrorKind::Other, err))
+ }
+
+ #[inline]
+ fn flush(&mut self) -> io::Result<()> {
+ Ok(())
+ }
+}
+
+impl io::Seek for Blob<'_> {
+ /// Seek to an offset, in bytes, in BLOB.
+ #[inline]
+ fn seek(&mut self, pos: io::SeekFrom) -> io::Result<u64> {
+ let pos = match pos {
+ io::SeekFrom::Start(offset) => offset as i64,
+ io::SeekFrom::Current(offset) => i64::from(self.pos) + offset,
+ io::SeekFrom::End(offset) => i64::from(self.size()) + offset,
+ };
+
+ if pos < 0 {
+ Err(io::Error::new(
+ io::ErrorKind::InvalidInput,
+ "invalid seek to negative position",
+ ))
+ } else if pos > i64::from(self.size()) {
+ Err(io::Error::new(
+ io::ErrorKind::InvalidInput,
+ "invalid seek to position past end of blob",
+ ))
+ } else {
+ self.pos = pos as i32;
+ Ok(pos as u64)
+ }
+ }
+}
+
+#[allow(unused_must_use)]
+impl Drop for Blob<'_> {
+ #[inline]
+ fn drop(&mut self) {
+ self.close_();
+ }
+}
+
+/// BLOB of length N that is filled with zeroes.
+///
+/// Zeroblobs are intended to serve as placeholders for BLOBs whose content is
+/// later written using incremental BLOB I/O routines.
+///
+/// A negative value for the zeroblob results in a zero-length BLOB.
+#[derive(Copy, Clone)]
+pub struct ZeroBlob(pub i32);
+
+impl ToSql for ZeroBlob {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ let ZeroBlob(length) = *self;
+ Ok(ToSqlOutput::ZeroBlob(length))
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use crate::{Connection, DatabaseName, Result};
+ use std::io::{BufRead, BufReader, BufWriter, Read, Seek, SeekFrom, Write};
+
+ fn db_with_test_blob() -> Result<(Connection, i64)> {
+ let db = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE test (content BLOB);
+ INSERT INTO test VALUES (ZEROBLOB(10));
+ END;";
+ db.execute_batch(sql)?;
+ let rowid = db.last_insert_rowid();
+ Ok((db, rowid))
+ }
+
+ #[test]
+ fn test_blob() -> Result<()> {
+ let (db, rowid) = db_with_test_blob()?;
+
+ let mut blob = db.blob_open(DatabaseName::Main, "test", "content", rowid, false)?;
+ assert_eq!(4, blob.write(b"Clob").unwrap());
+ assert_eq!(6, blob.write(b"567890xxxxxx").unwrap()); // cannot write past 10
+ assert_eq!(0, blob.write(b"5678").unwrap()); // still cannot write past 10
+
+ blob.reopen(rowid)?;
+ blob.close()?;
+
+ blob = db.blob_open(DatabaseName::Main, "test", "content", rowid, true)?;
+ let mut bytes = [0u8; 5];
+ assert_eq!(5, blob.read(&mut bytes[..]).unwrap());
+ assert_eq!(&bytes, b"Clob5");
+ assert_eq!(5, blob.read(&mut bytes[..]).unwrap());
+ assert_eq!(&bytes, b"67890");
+ assert_eq!(0, blob.read(&mut bytes[..]).unwrap());
+
+ blob.seek(SeekFrom::Start(2)).unwrap();
+ assert_eq!(5, blob.read(&mut bytes[..]).unwrap());
+ assert_eq!(&bytes, b"ob567");
+
+ // only first 4 bytes of `bytes` should be read into
+ blob.seek(SeekFrom::Current(-1)).unwrap();
+ assert_eq!(4, blob.read(&mut bytes[..]).unwrap());
+ assert_eq!(&bytes, b"78907");
+
+ blob.seek(SeekFrom::End(-6)).unwrap();
+ assert_eq!(5, blob.read(&mut bytes[..]).unwrap());
+ assert_eq!(&bytes, b"56789");
+
+ blob.reopen(rowid)?;
+ assert_eq!(5, blob.read(&mut bytes[..]).unwrap());
+ assert_eq!(&bytes, b"Clob5");
+
+ // should not be able to seek negative or past end
+ blob.seek(SeekFrom::Current(-20)).unwrap_err();
+ blob.seek(SeekFrom::End(0)).unwrap();
+ blob.seek(SeekFrom::Current(1)).unwrap_err();
+
+ // write_all should detect when we return Ok(0) because there is no space left,
+ // and return a write error
+ blob.reopen(rowid)?;
+ blob.write_all(b"0123456789x").unwrap_err();
+ Ok(())
+ }
+
+ #[test]
+ fn test_blob_in_bufreader() -> Result<()> {
+ let (db, rowid) = db_with_test_blob()?;
+
+ let mut blob = db.blob_open(DatabaseName::Main, "test", "content", rowid, false)?;
+ assert_eq!(8, blob.write(b"one\ntwo\n").unwrap());
+
+ blob.reopen(rowid)?;
+ let mut reader = BufReader::new(blob);
+
+ let mut line = String::new();
+ assert_eq!(4, reader.read_line(&mut line).unwrap());
+ assert_eq!("one\n", line);
+
+ line.truncate(0);
+ assert_eq!(4, reader.read_line(&mut line).unwrap());
+ assert_eq!("two\n", line);
+
+ line.truncate(0);
+ assert_eq!(2, reader.read_line(&mut line).unwrap());
+ assert_eq!("\0\0", line);
+ Ok(())
+ }
+
+ #[test]
+ fn test_blob_in_bufwriter() -> Result<()> {
+ let (db, rowid) = db_with_test_blob()?;
+
+ {
+ let blob = db.blob_open(DatabaseName::Main, "test", "content", rowid, false)?;
+ let mut writer = BufWriter::new(blob);
+
+ // trying to write too much and then flush should fail
+ assert_eq!(8, writer.write(b"01234567").unwrap());
+ assert_eq!(8, writer.write(b"01234567").unwrap());
+ writer.flush().unwrap_err();
+ }
+
+ {
+ // ... but it should've written the first 10 bytes
+ let mut blob = db.blob_open(DatabaseName::Main, "test", "content", rowid, false)?;
+ let mut bytes = [0u8; 10];
+ assert_eq!(10, blob.read(&mut bytes[..]).unwrap());
+ assert_eq!(b"0123456701", &bytes);
+ }
+
+ {
+ let blob = db.blob_open(DatabaseName::Main, "test", "content", rowid, false)?;
+ let mut writer = BufWriter::new(blob);
+
+ // trying to write_all too much should fail
+ writer.write_all(b"aaaaaaaaaabbbbb").unwrap();
+ writer.flush().unwrap_err();
+ }
+
+ {
+ // ... but it should've written the first 10 bytes
+ let mut blob = db.blob_open(DatabaseName::Main, "test", "content", rowid, false)?;
+ let mut bytes = [0u8; 10];
+ assert_eq!(10, blob.read(&mut bytes[..]).unwrap());
+ assert_eq!(b"aaaaaaaaaa", &bytes);
+ Ok(())
+ }
+ }
+}
diff --git a/third_party/rust/rusqlite/src/blob/pos_io.rs b/third_party/rust/rusqlite/src/blob/pos_io.rs
new file mode 100644
index 0000000000..d970ab735b
--- /dev/null
+++ b/third_party/rust/rusqlite/src/blob/pos_io.rs
@@ -0,0 +1,272 @@
+use super::Blob;
+
+use std::convert::TryFrom;
+use std::mem::MaybeUninit;
+use std::slice::from_raw_parts_mut;
+
+use crate::ffi;
+use crate::{Error, Result};
+
+impl<'conn> Blob<'conn> {
+ /// Write `buf` to `self` starting at `write_start`, returning an error if
+ /// `write_start + buf.len()` is past the end of the blob.
+ ///
+ /// If an error is returned, no data is written.
+ ///
+ /// Note: the blob cannot be resized using this function -- that must be
+ /// done using SQL (for example, an `UPDATE` statement).
+ ///
+ /// Note: This is part of the positional I/O API, and thus takes an absolute
+ /// position write to, instead of using the internal position that can be
+ /// manipulated by the `std::io` traits.
+ ///
+ /// Unlike the similarly named [`FileExt::write_at`][fext_write_at] function
+ /// (from `std::os::unix`), it's always an error to perform a "short write".
+ ///
+ /// [fext_write_at]: https://doc.rust-lang.org/std/os/unix/fs/trait.FileExt.html#tymethod.write_at
+ #[inline]
+ pub fn write_at(&mut self, buf: &[u8], write_start: usize) -> Result<()> {
+ let len = self.len();
+
+ if buf.len().saturating_add(write_start) > len {
+ return Err(Error::BlobSizeError);
+ }
+ // We know `len` fits in an `i32`, so either:
+ //
+ // 1. `buf.len() + write_start` overflows, in which case we'd hit the
+ // return above (courtesy of `saturating_add`).
+ //
+ // 2. `buf.len() + write_start` doesn't overflow but is larger than len,
+ // in which case ditto.
+ //
+ // 3. `buf.len() + write_start` doesn't overflow but is less than len.
+ // This means that both `buf.len()` and `write_start` can also be
+ // losslessly converted to i32, since `len` came from an i32.
+ // Sanity check the above.
+ debug_assert!(i32::try_from(write_start).is_ok() && i32::try_from(buf.len()).is_ok());
+ self.conn.decode_result(unsafe {
+ ffi::sqlite3_blob_write(
+ self.blob,
+ buf.as_ptr().cast(),
+ buf.len() as i32,
+ write_start as i32,
+ )
+ })
+ }
+
+ /// An alias for `write_at` provided for compatibility with the conceptually
+ /// equivalent [`std::os::unix::FileExt::write_all_at`][write_all_at]
+ /// function from libstd:
+ ///
+ /// [write_all_at]: https://doc.rust-lang.org/std/os/unix/fs/trait.FileExt.html#method.write_all_at
+ #[inline]
+ pub fn write_all_at(&mut self, buf: &[u8], write_start: usize) -> Result<()> {
+ self.write_at(buf, write_start)
+ }
+
+ /// Read as much as possible from `offset` to `offset + buf.len()` out of
+ /// `self`, writing into `buf`. On success, returns the number of bytes
+ /// written.
+ ///
+ /// If there's insufficient data in `self`, then the returned value will be
+ /// less than `buf.len()`.
+ ///
+ /// See also [`Blob::raw_read_at`], which can take an uninitialized buffer,
+ /// or [`Blob::read_at_exact`] which returns an error if the entire `buf` is
+ /// not read.
+ ///
+ /// Note: This is part of the positional I/O API, and thus takes an absolute
+ /// position to read from, instead of using the internal position that can
+ /// be manipulated by the `std::io` traits. Consequently, it does not change
+ /// that value either.
+ #[inline]
+ pub fn read_at(&self, buf: &mut [u8], read_start: usize) -> Result<usize> {
+ // Safety: this is safe because `raw_read_at` never stores uninitialized
+ // data into `as_uninit`.
+ let as_uninit: &mut [MaybeUninit<u8>] =
+ unsafe { from_raw_parts_mut(buf.as_mut_ptr().cast(), buf.len()) };
+ self.raw_read_at(as_uninit, read_start).map(|s| s.len())
+ }
+
+ /// Read as much as possible from `offset` to `offset + buf.len()` out of
+ /// `self`, writing into `buf`. On success, returns the portion of `buf`
+ /// which was initialized by this call.
+ ///
+ /// If there's insufficient data in `self`, then the returned value will be
+ /// shorter than `buf`.
+ ///
+ /// See also [`Blob::read_at`], which takes a `&mut [u8]` buffer instead of
+ /// a slice of `MaybeUninit<u8>`.
+ ///
+ /// Note: This is part of the positional I/O API, and thus takes an absolute
+ /// position to read from, instead of using the internal position that can
+ /// be manipulated by the `std::io` traits. Consequently, it does not change
+ /// that value either.
+ #[inline]
+ pub fn raw_read_at<'a>(
+ &self,
+ buf: &'a mut [MaybeUninit<u8>],
+ read_start: usize,
+ ) -> Result<&'a mut [u8]> {
+ let len = self.len();
+
+ let read_len = match len.checked_sub(read_start) {
+ None | Some(0) => 0,
+ Some(v) => v.min(buf.len()),
+ };
+
+ if read_len == 0 {
+ // We could return `Ok(&mut [])`, but it seems confusing that the
+ // pointers don't match, so fabricate a empty slice of u8 with the
+ // same base pointer as `buf`.
+ let empty = unsafe { from_raw_parts_mut(buf.as_mut_ptr().cast::<u8>(), 0) };
+ return Ok(empty);
+ }
+
+ // At this point we believe `read_start as i32` is lossless because:
+ //
+ // 1. `len as i32` is known to be lossless, since it comes from a SQLite
+ // api returning an i32.
+ //
+ // 2. If we got here, `len.checked_sub(read_start)` was Some (or else
+ // we'd have hit the `if read_len == 0` early return), so `len` must
+ // be larger than `read_start`, and so it must fit in i32 as well.
+ debug_assert!(i32::try_from(read_start).is_ok());
+
+ // We also believe that `read_start + read_len <= len` because:
+ //
+ // 1. This is equivalent to `read_len <= len - read_start` via algebra.
+ // 2. We know that `read_len` is `min(len - read_start, buf.len())`
+ // 3. Expanding, this is `min(len - read_start, buf.len()) <= len - read_start`,
+ // or `min(A, B) <= A` which is clearly true.
+ //
+ // Note that this stuff is in debug_assert so no need to use checked_add
+ // and such -- we'll always panic on overflow in debug builds.
+ debug_assert!(read_start + read_len <= len);
+
+ // These follow naturally.
+ debug_assert!(buf.len() >= read_len);
+ debug_assert!(i32::try_from(buf.len()).is_ok());
+ debug_assert!(i32::try_from(read_len).is_ok());
+
+ unsafe {
+ self.conn.decode_result(ffi::sqlite3_blob_read(
+ self.blob,
+ buf.as_mut_ptr().cast(),
+ read_len as i32,
+ read_start as i32,
+ ))?;
+
+ Ok(from_raw_parts_mut(buf.as_mut_ptr().cast::<u8>(), read_len))
+ }
+ }
+
+ /// Equivalent to [`Blob::read_at`], but returns a `BlobSizeError` if `buf`
+ /// is not fully initialized.
+ #[inline]
+ pub fn read_at_exact(&self, buf: &mut [u8], read_start: usize) -> Result<()> {
+ let n = self.read_at(buf, read_start)?;
+ if n != buf.len() {
+ Err(Error::BlobSizeError)
+ } else {
+ Ok(())
+ }
+ }
+
+ /// Equivalent to [`Blob::raw_read_at`], but returns a `BlobSizeError` if
+ /// `buf` is not fully initialized.
+ #[inline]
+ pub fn raw_read_at_exact<'a>(
+ &self,
+ buf: &'a mut [MaybeUninit<u8>],
+ read_start: usize,
+ ) -> Result<&'a mut [u8]> {
+ let buflen = buf.len();
+ let initted = self.raw_read_at(buf, read_start)?;
+ if initted.len() != buflen {
+ Err(Error::BlobSizeError)
+ } else {
+ Ok(initted)
+ }
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use crate::{Connection, DatabaseName, Result};
+ // to ensure we don't modify seek pos
+ use std::io::Seek as _;
+
+ #[test]
+ fn test_pos_io() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE test_table(content BLOB);")?;
+ db.execute("INSERT INTO test_table(content) VALUES (ZEROBLOB(10))", [])?;
+
+ let rowid = db.last_insert_rowid();
+ let mut blob = db.blob_open(DatabaseName::Main, "test_table", "content", rowid, false)?;
+ // modify the seek pos to ensure we aren't using it or modifying it.
+ blob.seek(std::io::SeekFrom::Start(1)).unwrap();
+
+ let one2ten: [u8; 10] = [1u8, 2, 3, 4, 5, 6, 7, 8, 9, 10];
+ blob.write_at(&one2ten, 0).unwrap();
+
+ let mut s = [0u8; 10];
+ blob.read_at_exact(&mut s, 0).unwrap();
+ assert_eq!(&s, &one2ten, "write should go through");
+ blob.read_at_exact(&mut s, 1).unwrap_err();
+
+ blob.read_at_exact(&mut s, 0).unwrap();
+ assert_eq!(&s, &one2ten, "should be unchanged");
+
+ let mut fives = [0u8; 5];
+ blob.read_at_exact(&mut fives, 0).unwrap();
+ assert_eq!(&fives, &[1u8, 2, 3, 4, 5]);
+
+ blob.read_at_exact(&mut fives, 5).unwrap();
+ assert_eq!(&fives, &[6u8, 7, 8, 9, 10]);
+ blob.read_at_exact(&mut fives, 7).unwrap_err();
+ blob.read_at_exact(&mut fives, 12).unwrap_err();
+ blob.read_at_exact(&mut fives, 10).unwrap_err();
+ blob.read_at_exact(&mut fives, i32::MAX as usize)
+ .unwrap_err();
+ blob.read_at_exact(&mut fives, i32::MAX as usize + 1)
+ .unwrap_err();
+
+ // zero length writes are fine if in bounds
+ blob.read_at_exact(&mut [], 10).unwrap();
+ blob.read_at_exact(&mut [], 0).unwrap();
+ blob.read_at_exact(&mut [], 5).unwrap();
+
+ blob.write_all_at(&[16, 17, 18, 19, 20], 5).unwrap();
+ blob.read_at_exact(&mut s, 0).unwrap();
+ assert_eq!(&s, &[1u8, 2, 3, 4, 5, 16, 17, 18, 19, 20]);
+
+ blob.write_at(&[100, 99, 98, 97, 96], 6).unwrap_err();
+ blob.write_at(&[100, 99, 98, 97, 96], i32::MAX as usize)
+ .unwrap_err();
+ blob.write_at(&[100, 99, 98, 97, 96], i32::MAX as usize + 1)
+ .unwrap_err();
+
+ blob.read_at_exact(&mut s, 0).unwrap();
+ assert_eq!(&s, &[1u8, 2, 3, 4, 5, 16, 17, 18, 19, 20]);
+
+ let mut s2: [std::mem::MaybeUninit<u8>; 10] = [std::mem::MaybeUninit::uninit(); 10];
+ {
+ let read = blob.raw_read_at_exact(&mut s2, 0).unwrap();
+ assert_eq!(read, &s);
+ assert!(std::ptr::eq(read.as_ptr(), s2.as_ptr().cast()));
+ }
+
+ let mut empty = [];
+ assert!(std::ptr::eq(
+ blob.raw_read_at_exact(&mut empty, 0).unwrap().as_ptr(),
+ empty.as_ptr().cast(),
+ ));
+ blob.raw_read_at_exact(&mut s2, 5).unwrap_err();
+
+ let end_pos = blob.stream_position().unwrap();
+ assert_eq!(end_pos, 1);
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/busy.rs b/third_party/rust/rusqlite/src/busy.rs
new file mode 100644
index 0000000000..18fa7e2907
--- /dev/null
+++ b/third_party/rust/rusqlite/src/busy.rs
@@ -0,0 +1,170 @@
+//! Busy handler (when the database is locked)
+use std::convert::TryInto;
+use std::mem;
+use std::os::raw::{c_int, c_void};
+use std::panic::catch_unwind;
+use std::ptr;
+use std::time::Duration;
+
+use crate::ffi;
+use crate::{Connection, InnerConnection, Result};
+
+impl Connection {
+ /// Set a busy handler that sleeps for a specified amount of time when a
+ /// table is locked. The handler will sleep multiple times until at
+ /// least "ms" milliseconds of sleeping have accumulated.
+ ///
+ /// Calling this routine with an argument equal to zero turns off all busy
+ /// handlers.
+ ///
+ /// There can only be a single busy handler for a particular database
+ /// connection at any given moment. If another busy handler was defined
+ /// (using [`busy_handler`](Connection::busy_handler)) prior to calling this
+ /// routine, that other busy handler is cleared.
+ ///
+ /// Newly created connections currently have a default busy timeout of
+ /// 5000ms, but this may be subject to change.
+ pub fn busy_timeout(&self, timeout: Duration) -> Result<()> {
+ let ms: i32 = timeout
+ .as_secs()
+ .checked_mul(1000)
+ .and_then(|t| t.checked_add(timeout.subsec_millis().into()))
+ .and_then(|t| t.try_into().ok())
+ .expect("too big");
+ self.db.borrow_mut().busy_timeout(ms)
+ }
+
+ /// Register a callback to handle `SQLITE_BUSY` errors.
+ ///
+ /// If the busy callback is `None`, then `SQLITE_BUSY` is returned
+ /// immediately upon encountering the lock. The argument to the busy
+ /// handler callback is the number of times that the
+ /// busy handler has been invoked previously for the
+ /// same locking event. If the busy callback returns `false`, then no
+ /// additional attempts are made to access the
+ /// database and `SQLITE_BUSY` is returned to the
+ /// application. If the callback returns `true`, then another attempt
+ /// is made to access the database and the cycle repeats.
+ ///
+ /// There can only be a single busy handler defined for each database
+ /// connection. Setting a new busy handler clears any previously set
+ /// handler. Note that calling [`busy_timeout()`](Connection::busy_timeout)
+ /// or evaluating `PRAGMA busy_timeout=N` will change the busy handler
+ /// and thus clear any previously set busy handler.
+ ///
+ /// Newly created connections default to a
+ /// [`busy_timeout()`](Connection::busy_timeout) handler with a timeout
+ /// of 5000ms, although this is subject to change.
+ pub fn busy_handler(&self, callback: Option<fn(i32) -> bool>) -> Result<()> {
+ unsafe extern "C" fn busy_handler_callback(p_arg: *mut c_void, count: c_int) -> c_int {
+ let handler_fn: fn(i32) -> bool = mem::transmute(p_arg);
+ c_int::from(catch_unwind(|| handler_fn(count)).unwrap_or_default())
+ }
+ let c = self.db.borrow_mut();
+ let r = match callback {
+ Some(f) => unsafe {
+ ffi::sqlite3_busy_handler(c.db(), Some(busy_handler_callback), f as *mut c_void)
+ },
+ None => unsafe { ffi::sqlite3_busy_handler(c.db(), None, ptr::null_mut()) },
+ };
+ c.decode_result(r)
+ }
+}
+
+impl InnerConnection {
+ #[inline]
+ fn busy_timeout(&mut self, timeout: c_int) -> Result<()> {
+ let r = unsafe { ffi::sqlite3_busy_timeout(self.db, timeout) };
+ self.decode_result(r)
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use std::sync::atomic::{AtomicBool, Ordering};
+ use std::sync::mpsc::sync_channel;
+ use std::thread;
+ use std::time::Duration;
+
+ use crate::{Connection, ErrorCode, Result, TransactionBehavior};
+
+ #[test]
+ fn test_default_busy() -> Result<()> {
+ let temp_dir = tempfile::tempdir().unwrap();
+ let path = temp_dir.path().join("test.db3");
+
+ let mut db1 = Connection::open(&path)?;
+ let tx1 = db1.transaction_with_behavior(TransactionBehavior::Exclusive)?;
+ let db2 = Connection::open(&path)?;
+ let r: Result<()> = db2.query_row("PRAGMA schema_version", [], |_| unreachable!());
+ assert_eq!(
+ r.unwrap_err().sqlite_error_code(),
+ Some(ErrorCode::DatabaseBusy)
+ );
+ tx1.rollback()
+ }
+
+ #[test]
+ #[ignore] // FIXME: unstable
+ fn test_busy_timeout() {
+ let temp_dir = tempfile::tempdir().unwrap();
+ let path = temp_dir.path().join("test.db3");
+
+ let db2 = Connection::open(&path).unwrap();
+ db2.busy_timeout(Duration::from_secs(1)).unwrap();
+
+ let (rx, tx) = sync_channel(0);
+ let child = thread::spawn(move || {
+ let mut db1 = Connection::open(&path).unwrap();
+ let tx1 = db1
+ .transaction_with_behavior(TransactionBehavior::Exclusive)
+ .unwrap();
+ rx.send(1).unwrap();
+ thread::sleep(Duration::from_millis(100));
+ tx1.rollback().unwrap();
+ });
+
+ assert_eq!(tx.recv().unwrap(), 1);
+ let _ = db2
+ .query_row("PRAGMA schema_version", [], |row| row.get::<_, i32>(0))
+ .expect("unexpected error");
+
+ child.join().unwrap();
+ }
+
+ #[test]
+ #[ignore] // FIXME: unstable
+ fn test_busy_handler() {
+ static CALLED: AtomicBool = AtomicBool::new(false);
+ fn busy_handler(_: i32) -> bool {
+ CALLED.store(true, Ordering::Relaxed);
+ thread::sleep(Duration::from_millis(100));
+ true
+ }
+
+ let temp_dir = tempfile::tempdir().unwrap();
+ let path = temp_dir.path().join("test.db3");
+
+ let db2 = Connection::open(&path).unwrap();
+ db2.busy_handler(Some(busy_handler)).unwrap();
+
+ let (rx, tx) = sync_channel(0);
+ let child = thread::spawn(move || {
+ let mut db1 = Connection::open(&path).unwrap();
+ let tx1 = db1
+ .transaction_with_behavior(TransactionBehavior::Exclusive)
+ .unwrap();
+ rx.send(1).unwrap();
+ thread::sleep(Duration::from_millis(100));
+ tx1.rollback().unwrap();
+ });
+
+ assert_eq!(tx.recv().unwrap(), 1);
+ let _ = db2
+ .query_row("PRAGMA schema_version", [], |row| row.get::<_, i32>(0))
+ .expect("unexpected error");
+ assert!(CALLED.load(Ordering::Relaxed));
+
+ child.join().unwrap();
+ }
+}
diff --git a/third_party/rust/rusqlite/src/cache.rs b/third_party/rust/rusqlite/src/cache.rs
new file mode 100644
index 0000000000..05ddb875f9
--- /dev/null
+++ b/third_party/rust/rusqlite/src/cache.rs
@@ -0,0 +1,350 @@
+//! Prepared statements cache for faster execution.
+
+use crate::raw_statement::RawStatement;
+use crate::{Connection, PrepFlags, Result, Statement};
+use hashlink::LruCache;
+use std::cell::RefCell;
+use std::ops::{Deref, DerefMut};
+use std::sync::Arc;
+
+impl Connection {
+ /// Prepare a SQL statement for execution, returning a previously prepared
+ /// (but not currently in-use) statement if one is available. The
+ /// returned statement will be cached for reuse by future calls to
+ /// [`prepare_cached`](Connection::prepare_cached) once it is dropped.
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn insert_new_people(conn: &Connection) -> Result<()> {
+ /// {
+ /// let mut stmt = conn.prepare_cached("INSERT INTO People (name) VALUES (?1)")?;
+ /// stmt.execute(["Joe Smith"])?;
+ /// }
+ /// {
+ /// // This will return the same underlying SQLite statement handle without
+ /// // having to prepare it again.
+ /// let mut stmt = conn.prepare_cached("INSERT INTO People (name) VALUES (?1)")?;
+ /// stmt.execute(["Bob Jones"])?;
+ /// }
+ /// Ok(())
+ /// }
+ /// ```
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if `sql` cannot be converted to a C-compatible string
+ /// or if the underlying SQLite call fails.
+ #[inline]
+ pub fn prepare_cached(&self, sql: &str) -> Result<CachedStatement<'_>> {
+ self.cache.get(self, sql)
+ }
+
+ /// Set the maximum number of cached prepared statements this connection
+ /// will hold. By default, a connection will hold a relatively small
+ /// number of cached statements. If you need more, or know that you
+ /// will not use cached statements, you
+ /// can set the capacity manually using this method.
+ #[inline]
+ pub fn set_prepared_statement_cache_capacity(&self, capacity: usize) {
+ self.cache.set_capacity(capacity);
+ }
+
+ /// Remove/finalize all prepared statements currently in the cache.
+ #[inline]
+ pub fn flush_prepared_statement_cache(&self) {
+ self.cache.flush();
+ }
+}
+
+/// Prepared statements LRU cache.
+// #[derive(Debug)] // FIXME: https://github.com/kyren/hashlink/pull/4
+pub struct StatementCache(RefCell<LruCache<Arc<str>, RawStatement>>);
+
+#[allow(clippy::non_send_fields_in_send_ty)]
+unsafe impl Send for StatementCache {}
+
+/// Cacheable statement.
+///
+/// Statement will return automatically to the cache by default.
+/// If you want the statement to be discarded, call
+/// [`discard()`](CachedStatement::discard) on it.
+pub struct CachedStatement<'conn> {
+ stmt: Option<Statement<'conn>>,
+ cache: &'conn StatementCache,
+}
+
+impl<'conn> Deref for CachedStatement<'conn> {
+ type Target = Statement<'conn>;
+
+ #[inline]
+ fn deref(&self) -> &Statement<'conn> {
+ self.stmt.as_ref().unwrap()
+ }
+}
+
+impl<'conn> DerefMut for CachedStatement<'conn> {
+ #[inline]
+ fn deref_mut(&mut self) -> &mut Statement<'conn> {
+ self.stmt.as_mut().unwrap()
+ }
+}
+
+impl Drop for CachedStatement<'_> {
+ #[allow(unused_must_use)]
+ #[inline]
+ fn drop(&mut self) {
+ if let Some(stmt) = self.stmt.take() {
+ self.cache.cache_stmt(unsafe { stmt.into_raw() });
+ }
+ }
+}
+
+impl CachedStatement<'_> {
+ #[inline]
+ fn new<'conn>(stmt: Statement<'conn>, cache: &'conn StatementCache) -> CachedStatement<'conn> {
+ CachedStatement {
+ stmt: Some(stmt),
+ cache,
+ }
+ }
+
+ /// Discard the statement, preventing it from being returned to its
+ /// [`Connection`]'s collection of cached statements.
+ #[inline]
+ pub fn discard(mut self) {
+ self.stmt = None;
+ }
+}
+
+impl StatementCache {
+ /// Create a statement cache.
+ #[inline]
+ pub fn with_capacity(capacity: usize) -> StatementCache {
+ StatementCache(RefCell::new(LruCache::new(capacity)))
+ }
+
+ #[inline]
+ fn set_capacity(&self, capacity: usize) {
+ self.0.borrow_mut().set_capacity(capacity);
+ }
+
+ // Search the cache for a prepared-statement object that implements `sql`.
+ // If no such prepared-statement can be found, allocate and prepare a new one.
+ //
+ // # Failure
+ //
+ // Will return `Err` if no cached statement can be found and the underlying
+ // SQLite prepare call fails.
+ fn get<'conn>(
+ &'conn self,
+ conn: &'conn Connection,
+ sql: &str,
+ ) -> Result<CachedStatement<'conn>> {
+ let trimmed = sql.trim();
+ let mut cache = self.0.borrow_mut();
+ let stmt = match cache.remove(trimmed) {
+ Some(raw_stmt) => Ok(Statement::new(conn, raw_stmt)),
+ None => conn.prepare_with_flags(trimmed, PrepFlags::SQLITE_PREPARE_PERSISTENT),
+ };
+ stmt.map(|mut stmt| {
+ stmt.stmt.set_statement_cache_key(trimmed);
+ CachedStatement::new(stmt, self)
+ })
+ }
+
+ // Return a statement to the cache.
+ fn cache_stmt(&self, stmt: RawStatement) {
+ if stmt.is_null() {
+ return;
+ }
+ let mut cache = self.0.borrow_mut();
+ stmt.clear_bindings();
+ if let Some(sql) = stmt.statement_cache_key() {
+ cache.insert(sql, stmt);
+ } else {
+ debug_assert!(
+ false,
+ "bug in statement cache code, statement returned to cache that without key"
+ );
+ }
+ }
+
+ #[inline]
+ fn flush(&self) {
+ let mut cache = self.0.borrow_mut();
+ cache.clear();
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use super::StatementCache;
+ use crate::{Connection, Result};
+ use fallible_iterator::FallibleIterator;
+
+ impl StatementCache {
+ fn clear(&self) {
+ self.0.borrow_mut().clear();
+ }
+
+ fn len(&self) -> usize {
+ self.0.borrow().len()
+ }
+
+ fn capacity(&self) -> usize {
+ self.0.borrow().capacity()
+ }
+ }
+
+ #[test]
+ fn test_cache() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let cache = &db.cache;
+ let initial_capacity = cache.capacity();
+ assert_eq!(0, cache.len());
+ assert!(initial_capacity > 0);
+
+ let sql = "PRAGMA schema_version";
+ {
+ let mut stmt = db.prepare_cached(sql)?;
+ assert_eq!(0, cache.len());
+ assert_eq!(0, stmt.query_row([], |r| r.get::<_, i64>(0))?);
+ }
+ assert_eq!(1, cache.len());
+
+ {
+ let mut stmt = db.prepare_cached(sql)?;
+ assert_eq!(0, cache.len());
+ assert_eq!(0, stmt.query_row([], |r| r.get::<_, i64>(0))?);
+ }
+ assert_eq!(1, cache.len());
+
+ cache.clear();
+ assert_eq!(0, cache.len());
+ assert_eq!(initial_capacity, cache.capacity());
+ Ok(())
+ }
+
+ #[test]
+ fn test_set_capacity() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let cache = &db.cache;
+
+ let sql = "PRAGMA schema_version";
+ {
+ let mut stmt = db.prepare_cached(sql)?;
+ assert_eq!(0, cache.len());
+ assert_eq!(0, stmt.query_row([], |r| r.get::<_, i64>(0))?);
+ }
+ assert_eq!(1, cache.len());
+
+ db.set_prepared_statement_cache_capacity(0);
+ assert_eq!(0, cache.len());
+
+ {
+ let mut stmt = db.prepare_cached(sql)?;
+ assert_eq!(0, cache.len());
+ assert_eq!(0, stmt.query_row([], |r| r.get::<_, i64>(0))?);
+ }
+ assert_eq!(0, cache.len());
+
+ db.set_prepared_statement_cache_capacity(8);
+ {
+ let mut stmt = db.prepare_cached(sql)?;
+ assert_eq!(0, cache.len());
+ assert_eq!(0, stmt.query_row([], |r| r.get::<_, i64>(0))?);
+ }
+ assert_eq!(1, cache.len());
+ Ok(())
+ }
+
+ #[test]
+ fn test_discard() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let cache = &db.cache;
+
+ let sql = "PRAGMA schema_version";
+ {
+ let mut stmt = db.prepare_cached(sql)?;
+ assert_eq!(0, cache.len());
+ assert_eq!(0, stmt.query_row([], |r| r.get::<_, i64>(0))?);
+ stmt.discard();
+ }
+ assert_eq!(0, cache.len());
+ Ok(())
+ }
+
+ #[test]
+ fn test_ddl() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch(
+ r#"
+ CREATE TABLE foo (x INT);
+ INSERT INTO foo VALUES (1);
+ "#,
+ )?;
+
+ let sql = "SELECT * FROM foo";
+
+ {
+ let mut stmt = db.prepare_cached(sql)?;
+ assert_eq!(Ok(Some(1i32)), stmt.query([])?.map(|r| r.get(0)).next());
+ }
+
+ db.execute_batch(
+ r#"
+ ALTER TABLE foo ADD COLUMN y INT;
+ UPDATE foo SET y = 2;
+ "#,
+ )?;
+
+ {
+ let mut stmt = db.prepare_cached(sql)?;
+ assert_eq!(
+ Ok(Some((1i32, 2i32))),
+ stmt.query([])?.map(|r| Ok((r.get(0)?, r.get(1)?))).next()
+ );
+ }
+ Ok(())
+ }
+
+ #[test]
+ fn test_connection_close() -> Result<()> {
+ let conn = Connection::open_in_memory()?;
+ conn.prepare_cached("SELECT * FROM sqlite_master;")?;
+
+ conn.close().expect("connection not closed");
+ Ok(())
+ }
+
+ #[test]
+ fn test_cache_key() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let cache = &db.cache;
+ assert_eq!(0, cache.len());
+
+ //let sql = " PRAGMA schema_version; -- comment";
+ let sql = "PRAGMA schema_version; ";
+ {
+ let mut stmt = db.prepare_cached(sql)?;
+ assert_eq!(0, cache.len());
+ assert_eq!(0, stmt.query_row([], |r| r.get::<_, i64>(0))?);
+ }
+ assert_eq!(1, cache.len());
+
+ {
+ let mut stmt = db.prepare_cached(sql)?;
+ assert_eq!(0, cache.len());
+ assert_eq!(0, stmt.query_row([], |r| r.get::<_, i64>(0))?);
+ }
+ assert_eq!(1, cache.len());
+ Ok(())
+ }
+
+ #[test]
+ fn test_empty_stmt() -> Result<()> {
+ let conn = Connection::open_in_memory()?;
+ conn.prepare_cached("")?;
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/collation.rs b/third_party/rust/rusqlite/src/collation.rs
new file mode 100644
index 0000000000..c1fe3f7837
--- /dev/null
+++ b/third_party/rust/rusqlite/src/collation.rs
@@ -0,0 +1,215 @@
+//! Add, remove, or modify a collation
+use std::cmp::Ordering;
+use std::os::raw::{c_char, c_int, c_void};
+use std::panic::{catch_unwind, UnwindSafe};
+use std::ptr;
+use std::slice;
+
+use crate::ffi;
+use crate::{str_to_cstring, Connection, InnerConnection, Result};
+
+// FIXME copy/paste from function.rs
+unsafe extern "C" fn free_boxed_value<T>(p: *mut c_void) {
+ drop(Box::from_raw(p.cast::<T>()));
+}
+
+impl Connection {
+ /// Add or modify a collation.
+ #[inline]
+ pub fn create_collation<C>(&self, collation_name: &str, x_compare: C) -> Result<()>
+ where
+ C: Fn(&str, &str) -> Ordering + Send + UnwindSafe + 'static,
+ {
+ self.db
+ .borrow_mut()
+ .create_collation(collation_name, x_compare)
+ }
+
+ /// Collation needed callback
+ #[inline]
+ pub fn collation_needed(
+ &self,
+ x_coll_needed: fn(&Connection, &str) -> Result<()>,
+ ) -> Result<()> {
+ self.db.borrow_mut().collation_needed(x_coll_needed)
+ }
+
+ /// Remove collation.
+ #[inline]
+ pub fn remove_collation(&self, collation_name: &str) -> Result<()> {
+ self.db.borrow_mut().remove_collation(collation_name)
+ }
+}
+
+impl InnerConnection {
+ fn create_collation<C>(&mut self, collation_name: &str, x_compare: C) -> Result<()>
+ where
+ C: Fn(&str, &str) -> Ordering + Send + UnwindSafe + 'static,
+ {
+ unsafe extern "C" fn call_boxed_closure<C>(
+ arg1: *mut c_void,
+ arg2: c_int,
+ arg3: *const c_void,
+ arg4: c_int,
+ arg5: *const c_void,
+ ) -> c_int
+ where
+ C: Fn(&str, &str) -> Ordering,
+ {
+ let r = catch_unwind(|| {
+ let boxed_f: *mut C = arg1.cast::<C>();
+ assert!(!boxed_f.is_null(), "Internal error - null function pointer");
+ let s1 = {
+ let c_slice = slice::from_raw_parts(arg3.cast::<u8>(), arg2 as usize);
+ String::from_utf8_lossy(c_slice)
+ };
+ let s2 = {
+ let c_slice = slice::from_raw_parts(arg5.cast::<u8>(), arg4 as usize);
+ String::from_utf8_lossy(c_slice)
+ };
+ (*boxed_f)(s1.as_ref(), s2.as_ref())
+ });
+ let t = match r {
+ Err(_) => {
+ return -1; // FIXME How ?
+ }
+ Ok(r) => r,
+ };
+
+ match t {
+ Ordering::Less => -1,
+ Ordering::Equal => 0,
+ Ordering::Greater => 1,
+ }
+ }
+
+ let boxed_f: *mut C = Box::into_raw(Box::new(x_compare));
+ let c_name = str_to_cstring(collation_name)?;
+ let flags = ffi::SQLITE_UTF8;
+ let r = unsafe {
+ ffi::sqlite3_create_collation_v2(
+ self.db(),
+ c_name.as_ptr(),
+ flags,
+ boxed_f.cast::<c_void>(),
+ Some(call_boxed_closure::<C>),
+ Some(free_boxed_value::<C>),
+ )
+ };
+ let res = self.decode_result(r);
+ // The xDestroy callback is not called if the sqlite3_create_collation_v2()
+ // function fails.
+ if res.is_err() {
+ drop(unsafe { Box::from_raw(boxed_f) });
+ }
+ res
+ }
+
+ fn collation_needed(
+ &mut self,
+ x_coll_needed: fn(&Connection, &str) -> Result<()>,
+ ) -> Result<()> {
+ use std::mem;
+ #[allow(clippy::needless_return)]
+ unsafe extern "C" fn collation_needed_callback(
+ arg1: *mut c_void,
+ arg2: *mut ffi::sqlite3,
+ e_text_rep: c_int,
+ arg3: *const c_char,
+ ) {
+ use std::ffi::CStr;
+ use std::str;
+
+ if e_text_rep != ffi::SQLITE_UTF8 {
+ // TODO: validate
+ return;
+ }
+
+ let callback: fn(&Connection, &str) -> Result<()> = mem::transmute(arg1);
+ let res = catch_unwind(|| {
+ let conn = Connection::from_handle(arg2).unwrap();
+ let collation_name = {
+ let c_slice = CStr::from_ptr(arg3).to_bytes();
+ str::from_utf8(c_slice).expect("illegal collation sequence name")
+ };
+ callback(&conn, collation_name)
+ });
+ if res.is_err() {
+ return; // FIXME How ?
+ }
+ }
+
+ let r = unsafe {
+ ffi::sqlite3_collation_needed(
+ self.db(),
+ x_coll_needed as *mut c_void,
+ Some(collation_needed_callback),
+ )
+ };
+ self.decode_result(r)
+ }
+
+ #[inline]
+ fn remove_collation(&mut self, collation_name: &str) -> Result<()> {
+ let c_name = str_to_cstring(collation_name)?;
+ let r = unsafe {
+ ffi::sqlite3_create_collation_v2(
+ self.db(),
+ c_name.as_ptr(),
+ ffi::SQLITE_UTF8,
+ ptr::null_mut(),
+ None,
+ None,
+ )
+ };
+ self.decode_result(r)
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use crate::{Connection, Result};
+ use fallible_streaming_iterator::FallibleStreamingIterator;
+ use std::cmp::Ordering;
+ use unicase::UniCase;
+
+ fn unicase_compare(s1: &str, s2: &str) -> Ordering {
+ UniCase::new(s1).cmp(&UniCase::new(s2))
+ }
+
+ #[test]
+ fn test_unicase() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ db.create_collation("unicase", unicase_compare)?;
+
+ collate(db)
+ }
+
+ fn collate(db: Connection) -> Result<()> {
+ db.execute_batch(
+ "CREATE TABLE foo (bar);
+ INSERT INTO foo (bar) VALUES ('Maße');
+ INSERT INTO foo (bar) VALUES ('MASSE');",
+ )?;
+ let mut stmt = db.prepare("SELECT DISTINCT bar COLLATE unicase FROM foo ORDER BY 1")?;
+ let rows = stmt.query([])?;
+ assert_eq!(rows.count()?, 1);
+ Ok(())
+ }
+
+ fn collation_needed(db: &Connection, collation_name: &str) -> Result<()> {
+ if "unicase" == collation_name {
+ db.create_collation(collation_name, unicase_compare)
+ } else {
+ Ok(())
+ }
+ }
+
+ #[test]
+ fn test_collation_needed() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.collation_needed(collation_needed)?;
+ collate(db)
+ }
+}
diff --git a/third_party/rust/rusqlite/src/column.rs b/third_party/rust/rusqlite/src/column.rs
new file mode 100644
index 0000000000..b18eb8a912
--- /dev/null
+++ b/third_party/rust/rusqlite/src/column.rs
@@ -0,0 +1,248 @@
+use std::str;
+
+use crate::{Error, Result, Statement};
+
+/// Information about a column of a SQLite query.
+#[cfg(feature = "column_decltype")]
+#[cfg_attr(docsrs, doc(cfg(feature = "column_decltype")))]
+#[derive(Debug)]
+pub struct Column<'stmt> {
+ name: &'stmt str,
+ decl_type: Option<&'stmt str>,
+}
+
+#[cfg(feature = "column_decltype")]
+#[cfg_attr(docsrs, doc(cfg(feature = "column_decltype")))]
+impl Column<'_> {
+ /// Returns the name of the column.
+ #[inline]
+ #[must_use]
+ pub fn name(&self) -> &str {
+ self.name
+ }
+
+ /// Returns the type of the column (`None` for expression).
+ #[inline]
+ #[must_use]
+ pub fn decl_type(&self) -> Option<&str> {
+ self.decl_type
+ }
+}
+
+impl Statement<'_> {
+ /// Get all the column names in the result set of the prepared statement.
+ ///
+ /// If associated DB schema can be altered concurrently, you should make
+ /// sure that current statement has already been stepped once before
+ /// calling this method.
+ pub fn column_names(&self) -> Vec<&str> {
+ let n = self.column_count();
+ let mut cols = Vec::with_capacity(n);
+ for i in 0..n {
+ let s = self.column_name_unwrap(i);
+ cols.push(s);
+ }
+ cols
+ }
+
+ /// Return the number of columns in the result set returned by the prepared
+ /// statement.
+ ///
+ /// If associated DB schema can be altered concurrently, you should make
+ /// sure that current statement has already been stepped once before
+ /// calling this method.
+ #[inline]
+ pub fn column_count(&self) -> usize {
+ self.stmt.column_count()
+ }
+
+ /// Check that column name reference lifetime is limited:
+ /// https://www.sqlite.org/c3ref/column_name.html
+ /// > The returned string pointer is valid...
+ ///
+ /// `column_name` reference can become invalid if `stmt` is reprepared
+ /// (because of schema change) when `query_row` is called. So we assert
+ /// that a compilation error happens if this reference is kept alive:
+ /// ```compile_fail
+ /// use rusqlite::{Connection, Result};
+ /// fn main() -> Result<()> {
+ /// let db = Connection::open_in_memory()?;
+ /// let mut stmt = db.prepare("SELECT 1 as x")?;
+ /// let column_name = stmt.column_name(0)?;
+ /// let x = stmt.query_row([], |r| r.get::<_, i64>(0))?; // E0502
+ /// assert_eq!(1, x);
+ /// assert_eq!("x", column_name);
+ /// Ok(())
+ /// }
+ /// ```
+ #[inline]
+ pub(super) fn column_name_unwrap(&self, col: usize) -> &str {
+ // Just panic if the bounds are wrong for now, we never call this
+ // without checking first.
+ self.column_name(col).expect("Column out of bounds")
+ }
+
+ /// Returns the name assigned to a particular column in the result set
+ /// returned by the prepared statement.
+ ///
+ /// If associated DB schema can be altered concurrently, you should make
+ /// sure that current statement has already been stepped once before
+ /// calling this method.
+ ///
+ /// ## Failure
+ ///
+ /// Returns an `Error::InvalidColumnIndex` if `idx` is outside the valid
+ /// column range for this row.
+ ///
+ /// # Panics
+ ///
+ /// Panics when column name is not valid UTF-8.
+ #[inline]
+ pub fn column_name(&self, col: usize) -> Result<&str> {
+ self.stmt
+ .column_name(col)
+ // clippy::or_fun_call (nightly) vs clippy::unnecessary-lazy-evaluations (stable)
+ .ok_or(Error::InvalidColumnIndex(col))
+ .map(|slice| {
+ str::from_utf8(slice.to_bytes()).expect("Invalid UTF-8 sequence in column name")
+ })
+ }
+
+ /// Returns the column index in the result set for a given column name.
+ ///
+ /// If there is no AS clause then the name of the column is unspecified and
+ /// may change from one release of SQLite to the next.
+ ///
+ /// If associated DB schema can be altered concurrently, you should make
+ /// sure that current statement has already been stepped once before
+ /// calling this method.
+ ///
+ /// # Failure
+ ///
+ /// Will return an `Error::InvalidColumnName` when there is no column with
+ /// the specified `name`.
+ #[inline]
+ pub fn column_index(&self, name: &str) -> Result<usize> {
+ let bytes = name.as_bytes();
+ let n = self.column_count();
+ for i in 0..n {
+ // Note: `column_name` is only fallible if `i` is out of bounds,
+ // which we've already checked.
+ if bytes.eq_ignore_ascii_case(self.stmt.column_name(i).unwrap().to_bytes()) {
+ return Ok(i);
+ }
+ }
+ Err(Error::InvalidColumnName(String::from(name)))
+ }
+
+ /// Returns a slice describing the columns of the result of the query.
+ ///
+ /// If associated DB schema can be altered concurrently, you should make
+ /// sure that current statement has already been stepped once before
+ /// calling this method.
+ #[cfg(feature = "column_decltype")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "column_decltype")))]
+ pub fn columns(&self) -> Vec<Column> {
+ let n = self.column_count();
+ let mut cols = Vec::with_capacity(n);
+ for i in 0..n {
+ let name = self.column_name_unwrap(i);
+ let slice = self.stmt.column_decltype(i);
+ let decl_type = slice.map(|s| {
+ str::from_utf8(s.to_bytes()).expect("Invalid UTF-8 sequence in column declaration")
+ });
+ cols.push(Column { name, decl_type });
+ }
+ cols
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use crate::{Connection, Result};
+
+ #[test]
+ #[cfg(feature = "column_decltype")]
+ fn test_columns() -> Result<()> {
+ use super::Column;
+
+ let db = Connection::open_in_memory()?;
+ let query = db.prepare("SELECT * FROM sqlite_master")?;
+ let columns = query.columns();
+ let column_names: Vec<&str> = columns.iter().map(Column::name).collect();
+ assert_eq!(
+ column_names.as_slice(),
+ &["type", "name", "tbl_name", "rootpage", "sql"]
+ );
+ let column_types: Vec<Option<String>> = columns
+ .iter()
+ .map(|col| col.decl_type().map(str::to_lowercase))
+ .collect();
+ assert_eq!(
+ &column_types[..3],
+ &[
+ Some("text".to_owned()),
+ Some("text".to_owned()),
+ Some("text".to_owned()),
+ ]
+ );
+ Ok(())
+ }
+
+ #[test]
+ fn test_column_name_in_error() -> Result<()> {
+ use crate::{types::Type, Error};
+ let db = Connection::open_in_memory()?;
+ db.execute_batch(
+ "BEGIN;
+ CREATE TABLE foo(x INTEGER, y TEXT);
+ INSERT INTO foo VALUES(4, NULL);
+ END;",
+ )?;
+ let mut stmt = db.prepare("SELECT x as renamed, y FROM foo")?;
+ let mut rows = stmt.query([])?;
+ let row = rows.next()?.unwrap();
+ match row.get::<_, String>(0).unwrap_err() {
+ Error::InvalidColumnType(idx, name, ty) => {
+ assert_eq!(idx, 0);
+ assert_eq!(name, "renamed");
+ assert_eq!(ty, Type::Integer);
+ }
+ e => {
+ panic!("Unexpected error type: {e:?}");
+ }
+ }
+ match row.get::<_, String>("y").unwrap_err() {
+ Error::InvalidColumnType(idx, name, ty) => {
+ assert_eq!(idx, 1);
+ assert_eq!(name, "y");
+ assert_eq!(ty, Type::Null);
+ }
+ e => {
+ panic!("Unexpected error type: {e:?}");
+ }
+ }
+ Ok(())
+ }
+
+ /// `column_name` reference should stay valid until `stmt` is reprepared (or
+ /// reset) even if DB schema is altered (SQLite documentation is
+ /// ambiguous here because it says reference "is valid until (...) the next
+ /// call to sqlite3_column_name() or sqlite3_column_name16() on the same
+ /// column.". We assume that reference is valid if only
+ /// `sqlite3_column_name()` is used):
+ #[test]
+ #[cfg(feature = "modern_sqlite")]
+ fn test_column_name_reference() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE y (x);")?;
+ let stmt = db.prepare("SELECT x FROM y;")?;
+ let column_name = stmt.column_name(0)?;
+ assert_eq!("x", column_name);
+ db.execute_batch("ALTER TABLE y RENAME COLUMN x TO z;")?;
+ // column name is not refreshed until statement is re-prepared
+ let same_column_name = stmt.column_name(0)?;
+ assert_eq!(same_column_name, column_name);
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/config.rs b/third_party/rust/rusqlite/src/config.rs
new file mode 100644
index 0000000000..194ee59639
--- /dev/null
+++ b/third_party/rust/rusqlite/src/config.rs
@@ -0,0 +1,163 @@
+//! Configure database connections
+
+use std::os::raw::c_int;
+
+use crate::error::check;
+use crate::ffi;
+use crate::{Connection, Result};
+
+/// Database Connection Configuration Options
+/// See [Database Connection Configuration Options](https://sqlite.org/c3ref/c_dbconfig_enable_fkey.html) for details.
+#[repr(i32)]
+#[allow(non_snake_case, non_camel_case_types)]
+#[non_exhaustive]
+#[allow(clippy::upper_case_acronyms)]
+pub enum DbConfig {
+ //SQLITE_DBCONFIG_MAINDBNAME = 1000, /* const char* */
+ //SQLITE_DBCONFIG_LOOKASIDE = 1001, /* void* int int */
+ /// Enable or disable the enforcement of foreign key constraints.
+ SQLITE_DBCONFIG_ENABLE_FKEY = ffi::SQLITE_DBCONFIG_ENABLE_FKEY,
+ /// Enable or disable triggers.
+ SQLITE_DBCONFIG_ENABLE_TRIGGER = ffi::SQLITE_DBCONFIG_ENABLE_TRIGGER,
+ /// Enable or disable the fts3_tokenizer() function which is part of the
+ /// FTS3 full-text search engine extension.
+ SQLITE_DBCONFIG_ENABLE_FTS3_TOKENIZER = ffi::SQLITE_DBCONFIG_ENABLE_FTS3_TOKENIZER, // 3.12.0
+ //SQLITE_DBCONFIG_ENABLE_LOAD_EXTENSION = 1005,
+ /// In WAL mode, enable or disable the checkpoint operation before closing
+ /// the connection.
+ SQLITE_DBCONFIG_NO_CKPT_ON_CLOSE = 1006, // 3.16.2
+ /// Activates or deactivates the query planner stability guarantee (QPSG).
+ SQLITE_DBCONFIG_ENABLE_QPSG = 1007, // 3.20.0
+ /// Includes or excludes output for any operations performed by trigger
+ /// programs from the output of EXPLAIN QUERY PLAN commands.
+ SQLITE_DBCONFIG_TRIGGER_EQP = 1008, // 3.22.0
+ /// Activates or deactivates the "reset" flag for a database connection.
+ /// Run VACUUM with this flag set to reset the database.
+ SQLITE_DBCONFIG_RESET_DATABASE = 1009, // 3.24.0
+ /// Activates or deactivates the "defensive" flag for a database connection.
+ SQLITE_DBCONFIG_DEFENSIVE = 1010, // 3.26.0
+ /// Activates or deactivates the "writable_schema" flag.
+ #[cfg(feature = "modern_sqlite")]
+ SQLITE_DBCONFIG_WRITABLE_SCHEMA = 1011, // 3.28.0
+ /// Activates or deactivates the legacy behavior of the ALTER TABLE RENAME
+ /// command.
+ #[cfg(feature = "modern_sqlite")]
+ SQLITE_DBCONFIG_LEGACY_ALTER_TABLE = 1012, // 3.29
+ /// Activates or deactivates the legacy double-quoted string literal
+ /// misfeature for DML statements only.
+ #[cfg(feature = "modern_sqlite")]
+ SQLITE_DBCONFIG_DQS_DML = 1013, // 3.29.0
+ /// Activates or deactivates the legacy double-quoted string literal
+ /// misfeature for DDL statements.
+ #[cfg(feature = "modern_sqlite")]
+ SQLITE_DBCONFIG_DQS_DDL = 1014, // 3.29.0
+ /// Enable or disable views.
+ #[cfg(feature = "modern_sqlite")]
+ SQLITE_DBCONFIG_ENABLE_VIEW = 1015, // 3.30.0
+ /// Activates or deactivates the legacy file format flag.
+ #[cfg(feature = "modern_sqlite")]
+ SQLITE_DBCONFIG_LEGACY_FILE_FORMAT = 1016, // 3.31.0
+ /// Tells SQLite to assume that database schemas (the contents of the
+ /// sqlite_master tables) are untainted by malicious content.
+ #[cfg(feature = "modern_sqlite")]
+ SQLITE_DBCONFIG_TRUSTED_SCHEMA = 1017, // 3.31.0
+ /// Sets or clears a flag that enables collection of the
+ /// sqlite3_stmt_scanstatus_v2() statistics
+ #[cfg(feature = "modern_sqlite")]
+ SQLITE_DBCONFIG_STMT_SCANSTATUS = 1018, // 3.42.0
+ /// Changes the default order in which tables and indexes are scanned
+ #[cfg(feature = "modern_sqlite")]
+ SQLITE_DBCONFIG_REVERSE_SCANORDER = 1019, // 3.42.0
+}
+
+impl Connection {
+ /// Returns the current value of a `config`.
+ ///
+ /// - `SQLITE_DBCONFIG_ENABLE_FKEY`: return `false` or `true` to indicate
+ /// whether FK enforcement is off or on
+ /// - `SQLITE_DBCONFIG_ENABLE_TRIGGER`: return `false` or `true` to indicate
+ /// whether triggers are disabled or enabled
+ /// - `SQLITE_DBCONFIG_ENABLE_FTS3_TOKENIZER`: return `false` or `true` to
+ /// indicate whether `fts3_tokenizer` are disabled or enabled
+ /// - `SQLITE_DBCONFIG_NO_CKPT_ON_CLOSE`: return `false` to indicate
+ /// checkpoints-on-close are not disabled or `true` if they are
+ /// - `SQLITE_DBCONFIG_ENABLE_QPSG`: return `false` or `true` to indicate
+ /// whether the QPSG is disabled or enabled
+ /// - `SQLITE_DBCONFIG_TRIGGER_EQP`: return `false` to indicate
+ /// output-for-trigger are not disabled or `true` if it is
+ #[inline]
+ pub fn db_config(&self, config: DbConfig) -> Result<bool> {
+ let c = self.db.borrow();
+ unsafe {
+ let mut val = 0;
+ check(ffi::sqlite3_db_config(
+ c.db(),
+ config as c_int,
+ -1,
+ &mut val,
+ ))?;
+ Ok(val != 0)
+ }
+ }
+
+ /// Make configuration changes to a database connection
+ ///
+ /// - `SQLITE_DBCONFIG_ENABLE_FKEY`: `false` to disable FK enforcement,
+ /// `true` to enable FK enforcement
+ /// - `SQLITE_DBCONFIG_ENABLE_TRIGGER`: `false` to disable triggers, `true`
+ /// to enable triggers
+ /// - `SQLITE_DBCONFIG_ENABLE_FTS3_TOKENIZER`: `false` to disable
+ /// `fts3_tokenizer()`, `true` to enable `fts3_tokenizer()`
+ /// - `SQLITE_DBCONFIG_NO_CKPT_ON_CLOSE`: `false` (the default) to enable
+ /// checkpoints-on-close, `true` to disable them
+ /// - `SQLITE_DBCONFIG_ENABLE_QPSG`: `false` to disable the QPSG, `true` to
+ /// enable QPSG
+ /// - `SQLITE_DBCONFIG_TRIGGER_EQP`: `false` to disable output for trigger
+ /// programs, `true` to enable it
+ #[inline]
+ pub fn set_db_config(&self, config: DbConfig, new_val: bool) -> Result<bool> {
+ let c = self.db.borrow_mut();
+ unsafe {
+ let mut val = 0;
+ check(ffi::sqlite3_db_config(
+ c.db(),
+ config as c_int,
+ new_val as c_int,
+ &mut val,
+ ))?;
+ Ok(val != 0)
+ }
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use super::DbConfig;
+ use crate::{Connection, Result};
+
+ #[test]
+ fn test_db_config() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ let opposite = !db.db_config(DbConfig::SQLITE_DBCONFIG_ENABLE_FKEY)?;
+ assert_eq!(
+ db.set_db_config(DbConfig::SQLITE_DBCONFIG_ENABLE_FKEY, opposite),
+ Ok(opposite)
+ );
+ assert_eq!(
+ db.db_config(DbConfig::SQLITE_DBCONFIG_ENABLE_FKEY),
+ Ok(opposite)
+ );
+
+ let opposite = !db.db_config(DbConfig::SQLITE_DBCONFIG_ENABLE_TRIGGER)?;
+ assert_eq!(
+ db.set_db_config(DbConfig::SQLITE_DBCONFIG_ENABLE_TRIGGER, opposite),
+ Ok(opposite)
+ );
+ assert_eq!(
+ db.db_config(DbConfig::SQLITE_DBCONFIG_ENABLE_TRIGGER),
+ Ok(opposite)
+ );
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/context.rs b/third_party/rust/rusqlite/src/context.rs
new file mode 100644
index 0000000000..bcaefc9395
--- /dev/null
+++ b/third_party/rust/rusqlite/src/context.rs
@@ -0,0 +1,75 @@
+//! Code related to `sqlite3_context` common to `functions` and `vtab` modules.
+
+use std::os::raw::{c_int, c_void};
+#[cfg(feature = "array")]
+use std::rc::Rc;
+
+use crate::ffi;
+use crate::ffi::sqlite3_context;
+
+use crate::str_for_sqlite;
+use crate::types::{ToSqlOutput, ValueRef};
+#[cfg(feature = "array")]
+use crate::vtab::array::{free_array, ARRAY_TYPE};
+
+// This function is inline despite it's size because what's in the ToSqlOutput
+// is often known to the compiler, and thus const prop/DCE can substantially
+// simplify the function.
+#[inline]
+pub(super) unsafe fn set_result(ctx: *mut sqlite3_context, result: &ToSqlOutput<'_>) {
+ let value = match *result {
+ ToSqlOutput::Borrowed(v) => v,
+ ToSqlOutput::Owned(ref v) => ValueRef::from(v),
+
+ #[cfg(feature = "blob")]
+ ToSqlOutput::ZeroBlob(len) => {
+ // TODO sqlite3_result_zeroblob64 // 3.8.11
+ return ffi::sqlite3_result_zeroblob(ctx, len);
+ }
+ #[cfg(feature = "array")]
+ ToSqlOutput::Array(ref a) => {
+ return ffi::sqlite3_result_pointer(
+ ctx,
+ Rc::into_raw(a.clone()) as *mut c_void,
+ ARRAY_TYPE,
+ Some(free_array),
+ );
+ }
+ };
+
+ match value {
+ ValueRef::Null => ffi::sqlite3_result_null(ctx),
+ ValueRef::Integer(i) => ffi::sqlite3_result_int64(ctx, i),
+ ValueRef::Real(r) => ffi::sqlite3_result_double(ctx, r),
+ ValueRef::Text(s) => {
+ let length = s.len();
+ if length > c_int::MAX as usize {
+ ffi::sqlite3_result_error_toobig(ctx);
+ } else {
+ let (c_str, len, destructor) = match str_for_sqlite(s) {
+ Ok(c_str) => c_str,
+ // TODO sqlite3_result_error
+ Err(_) => return ffi::sqlite3_result_error_code(ctx, ffi::SQLITE_MISUSE),
+ };
+ // TODO sqlite3_result_text64 // 3.8.7
+ ffi::sqlite3_result_text(ctx, c_str, len, destructor);
+ }
+ }
+ ValueRef::Blob(b) => {
+ let length = b.len();
+ if length > c_int::MAX as usize {
+ ffi::sqlite3_result_error_toobig(ctx);
+ } else if length == 0 {
+ ffi::sqlite3_result_zeroblob(ctx, 0);
+ } else {
+ // TODO sqlite3_result_blob64 // 3.8.7
+ ffi::sqlite3_result_blob(
+ ctx,
+ b.as_ptr().cast::<c_void>(),
+ length as c_int,
+ ffi::SQLITE_TRANSIENT(),
+ );
+ }
+ }
+ }
+}
diff --git a/third_party/rust/rusqlite/src/error.rs b/third_party/rust/rusqlite/src/error.rs
new file mode 100644
index 0000000000..69aaf2ef21
--- /dev/null
+++ b/third_party/rust/rusqlite/src/error.rs
@@ -0,0 +1,476 @@
+use crate::types::FromSqlError;
+use crate::types::Type;
+use crate::{errmsg_to_string, ffi, Result};
+use std::error;
+use std::fmt;
+use std::os::raw::c_int;
+use std::path::PathBuf;
+use std::str;
+
+/// Enum listing possible errors from rusqlite.
+#[derive(Debug)]
+#[allow(clippy::enum_variant_names)]
+#[non_exhaustive]
+pub enum Error {
+ /// An error from an underlying SQLite call.
+ SqliteFailure(ffi::Error, Option<String>),
+
+ /// Error reported when attempting to open a connection when SQLite was
+ /// configured to allow single-threaded use only.
+ SqliteSingleThreadedMode,
+
+ /// Error when the value of a particular column is requested, but it cannot
+ /// be converted to the requested Rust type.
+ FromSqlConversionFailure(usize, Type, Box<dyn error::Error + Send + Sync + 'static>),
+
+ /// Error when SQLite gives us an integral value outside the range of the
+ /// requested type (e.g., trying to get the value 1000 into a `u8`).
+ /// The associated `usize` is the column index,
+ /// and the associated `i64` is the value returned by SQLite.
+ IntegralValueOutOfRange(usize, i64),
+
+ /// Error converting a string to UTF-8.
+ Utf8Error(str::Utf8Error),
+
+ /// Error converting a string to a C-compatible string because it contained
+ /// an embedded nul.
+ NulError(std::ffi::NulError),
+
+ /// Error when using SQL named parameters and passing a parameter name not
+ /// present in the SQL.
+ InvalidParameterName(String),
+
+ /// Error converting a file path to a string.
+ InvalidPath(PathBuf),
+
+ /// Error returned when an [`execute`](crate::Connection::execute) call
+ /// returns rows.
+ ExecuteReturnedResults,
+
+ /// Error when a query that was expected to return at least one row (e.g.,
+ /// for [`query_row`](crate::Connection::query_row)) did not return any.
+ QueryReturnedNoRows,
+
+ /// Error when the value of a particular column is requested, but the index
+ /// is out of range for the statement.
+ InvalidColumnIndex(usize),
+
+ /// Error when the value of a named column is requested, but no column
+ /// matches the name for the statement.
+ InvalidColumnName(String),
+
+ /// Error when the value of a particular column is requested, but the type
+ /// of the result in that column cannot be converted to the requested
+ /// Rust type.
+ InvalidColumnType(usize, String, Type),
+
+ /// Error when a query that was expected to insert one row did not insert
+ /// any or insert many.
+ StatementChangedRows(usize),
+
+ /// Error returned by
+ /// [`functions::Context::get`](crate::functions::Context::get) when the
+ /// function argument cannot be converted to the requested type.
+ #[cfg(feature = "functions")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "functions")))]
+ InvalidFunctionParameterType(usize, Type),
+ /// Error returned by [`vtab::Values::get`](crate::vtab::Values::get) when
+ /// the filter argument cannot be converted to the requested type.
+ #[cfg(feature = "vtab")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "vtab")))]
+ InvalidFilterParameterType(usize, Type),
+
+ /// An error case available for implementors of custom user functions (e.g.,
+ /// [`create_scalar_function`](crate::Connection::create_scalar_function)).
+ #[cfg(feature = "functions")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "functions")))]
+ #[allow(dead_code)]
+ UserFunctionError(Box<dyn error::Error + Send + Sync + 'static>),
+
+ /// Error available for the implementors of the
+ /// [`ToSql`](crate::types::ToSql) trait.
+ ToSqlConversionFailure(Box<dyn error::Error + Send + Sync + 'static>),
+
+ /// Error when the SQL is not a `SELECT`, is not read-only.
+ InvalidQuery,
+
+ /// An error case available for implementors of custom modules (e.g.,
+ /// [`create_module`](crate::Connection::create_module)).
+ #[cfg(feature = "vtab")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "vtab")))]
+ #[allow(dead_code)]
+ ModuleError(String),
+
+ /// An unwinding panic occurs in an UDF (user-defined function).
+ #[cfg(feature = "functions")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "functions")))]
+ UnwindingPanic,
+
+ /// An error returned when
+ /// [`Context::get_aux`](crate::functions::Context::get_aux) attempts to
+ /// retrieve data of a different type than what had been stored using
+ /// [`Context::set_aux`](crate::functions::Context::set_aux).
+ #[cfg(feature = "functions")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "functions")))]
+ GetAuxWrongType,
+
+ /// Error when the SQL contains multiple statements.
+ MultipleStatement,
+ /// Error when the number of bound parameters does not match the number of
+ /// parameters in the query. The first `usize` is how many parameters were
+ /// given, the 2nd is how many were expected.
+ InvalidParameterCount(usize, usize),
+
+ /// Returned from various functions in the Blob IO positional API. For
+ /// example,
+ /// [`Blob::raw_read_at_exact`](crate::blob::Blob::raw_read_at_exact) will
+ /// return it if the blob has insufficient data.
+ #[cfg(feature = "blob")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "blob")))]
+ BlobSizeError,
+ /// Error referencing a specific token in the input SQL
+ #[cfg(feature = "modern_sqlite")] // 3.38.0
+ #[cfg_attr(docsrs, doc(cfg(feature = "modern_sqlite")))]
+ SqlInputError {
+ /// error code
+ error: ffi::Error,
+ /// error message
+ msg: String,
+ /// SQL input
+ sql: String,
+ /// byte offset of the start of invalid token
+ offset: c_int,
+ },
+ /// Loadable extension initialization error
+ #[cfg(feature = "loadable_extension")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "loadable_extension")))]
+ InitError(ffi::InitError),
+}
+
+impl PartialEq for Error {
+ fn eq(&self, other: &Error) -> bool {
+ match (self, other) {
+ (Error::SqliteFailure(e1, s1), Error::SqliteFailure(e2, s2)) => e1 == e2 && s1 == s2,
+ (Error::SqliteSingleThreadedMode, Error::SqliteSingleThreadedMode) => true,
+ (Error::IntegralValueOutOfRange(i1, n1), Error::IntegralValueOutOfRange(i2, n2)) => {
+ i1 == i2 && n1 == n2
+ }
+ (Error::Utf8Error(e1), Error::Utf8Error(e2)) => e1 == e2,
+ (Error::NulError(e1), Error::NulError(e2)) => e1 == e2,
+ (Error::InvalidParameterName(n1), Error::InvalidParameterName(n2)) => n1 == n2,
+ (Error::InvalidPath(p1), Error::InvalidPath(p2)) => p1 == p2,
+ (Error::ExecuteReturnedResults, Error::ExecuteReturnedResults) => true,
+ (Error::QueryReturnedNoRows, Error::QueryReturnedNoRows) => true,
+ (Error::InvalidColumnIndex(i1), Error::InvalidColumnIndex(i2)) => i1 == i2,
+ (Error::InvalidColumnName(n1), Error::InvalidColumnName(n2)) => n1 == n2,
+ (Error::InvalidColumnType(i1, n1, t1), Error::InvalidColumnType(i2, n2, t2)) => {
+ i1 == i2 && t1 == t2 && n1 == n2
+ }
+ (Error::StatementChangedRows(n1), Error::StatementChangedRows(n2)) => n1 == n2,
+ #[cfg(feature = "functions")]
+ (
+ Error::InvalidFunctionParameterType(i1, t1),
+ Error::InvalidFunctionParameterType(i2, t2),
+ ) => i1 == i2 && t1 == t2,
+ #[cfg(feature = "vtab")]
+ (
+ Error::InvalidFilterParameterType(i1, t1),
+ Error::InvalidFilterParameterType(i2, t2),
+ ) => i1 == i2 && t1 == t2,
+ (Error::InvalidQuery, Error::InvalidQuery) => true,
+ #[cfg(feature = "vtab")]
+ (Error::ModuleError(s1), Error::ModuleError(s2)) => s1 == s2,
+ #[cfg(feature = "functions")]
+ (Error::UnwindingPanic, Error::UnwindingPanic) => true,
+ #[cfg(feature = "functions")]
+ (Error::GetAuxWrongType, Error::GetAuxWrongType) => true,
+ (Error::InvalidParameterCount(i1, n1), Error::InvalidParameterCount(i2, n2)) => {
+ i1 == i2 && n1 == n2
+ }
+ #[cfg(feature = "blob")]
+ (Error::BlobSizeError, Error::BlobSizeError) => true,
+ #[cfg(feature = "modern_sqlite")]
+ (
+ Error::SqlInputError {
+ error: e1,
+ msg: m1,
+ sql: s1,
+ offset: o1,
+ },
+ Error::SqlInputError {
+ error: e2,
+ msg: m2,
+ sql: s2,
+ offset: o2,
+ },
+ ) => e1 == e2 && m1 == m2 && s1 == s2 && o1 == o2,
+ #[cfg(feature = "loadable_extension")]
+ (Error::InitError(e1), Error::InitError(e2)) => e1 == e2,
+ (..) => false,
+ }
+ }
+}
+
+impl From<str::Utf8Error> for Error {
+ #[cold]
+ fn from(err: str::Utf8Error) -> Error {
+ Error::Utf8Error(err)
+ }
+}
+
+impl From<std::ffi::NulError> for Error {
+ #[cold]
+ fn from(err: std::ffi::NulError) -> Error {
+ Error::NulError(err)
+ }
+}
+
+const UNKNOWN_COLUMN: usize = usize::MAX;
+
+/// The conversion isn't precise, but it's convenient to have it
+/// to allow use of `get_raw(…).as_…()?` in callbacks that take `Error`.
+impl From<FromSqlError> for Error {
+ #[cold]
+ fn from(err: FromSqlError) -> Error {
+ // The error type requires index and type fields, but they aren't known in this
+ // context.
+ match err {
+ FromSqlError::OutOfRange(val) => Error::IntegralValueOutOfRange(UNKNOWN_COLUMN, val),
+ FromSqlError::InvalidBlobSize { .. } => {
+ Error::FromSqlConversionFailure(UNKNOWN_COLUMN, Type::Blob, Box::new(err))
+ }
+ FromSqlError::Other(source) => {
+ Error::FromSqlConversionFailure(UNKNOWN_COLUMN, Type::Null, source)
+ }
+ _ => Error::FromSqlConversionFailure(UNKNOWN_COLUMN, Type::Null, Box::new(err)),
+ }
+ }
+}
+
+#[cfg(feature = "loadable_extension")]
+impl From<ffi::InitError> for Error {
+ #[cold]
+ fn from(err: ffi::InitError) -> Error {
+ Error::InitError(err)
+ }
+}
+
+impl fmt::Display for Error {
+ fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+ match *self {
+ Error::SqliteFailure(ref err, None) => err.fmt(f),
+ Error::SqliteFailure(_, Some(ref s)) => write!(f, "{s}"),
+ Error::SqliteSingleThreadedMode => write!(
+ f,
+ "SQLite was compiled or configured for single-threaded use only"
+ ),
+ Error::FromSqlConversionFailure(i, ref t, ref err) => {
+ if i != UNKNOWN_COLUMN {
+ write!(f, "Conversion error from type {t} at index: {i}, {err}")
+ } else {
+ err.fmt(f)
+ }
+ }
+ Error::IntegralValueOutOfRange(col, val) => {
+ if col != UNKNOWN_COLUMN {
+ write!(f, "Integer {val} out of range at index {col}")
+ } else {
+ write!(f, "Integer {val} out of range")
+ }
+ }
+ Error::Utf8Error(ref err) => err.fmt(f),
+ Error::NulError(ref err) => err.fmt(f),
+ Error::InvalidParameterName(ref name) => write!(f, "Invalid parameter name: {name}"),
+ Error::InvalidPath(ref p) => write!(f, "Invalid path: {}", p.to_string_lossy()),
+ Error::ExecuteReturnedResults => {
+ write!(f, "Execute returned results - did you mean to call query?")
+ }
+ Error::QueryReturnedNoRows => write!(f, "Query returned no rows"),
+ Error::InvalidColumnIndex(i) => write!(f, "Invalid column index: {i}"),
+ Error::InvalidColumnName(ref name) => write!(f, "Invalid column name: {name}"),
+ Error::InvalidColumnType(i, ref name, ref t) => {
+ write!(f, "Invalid column type {t} at index: {i}, name: {name}")
+ }
+ Error::InvalidParameterCount(i1, n1) => write!(
+ f,
+ "Wrong number of parameters passed to query. Got {i1}, needed {n1}"
+ ),
+ Error::StatementChangedRows(i) => write!(f, "Query changed {i} rows"),
+
+ #[cfg(feature = "functions")]
+ Error::InvalidFunctionParameterType(i, ref t) => {
+ write!(f, "Invalid function parameter type {t} at index {i}")
+ }
+ #[cfg(feature = "vtab")]
+ Error::InvalidFilterParameterType(i, ref t) => {
+ write!(f, "Invalid filter parameter type {t} at index {i}")
+ }
+ #[cfg(feature = "functions")]
+ Error::UserFunctionError(ref err) => err.fmt(f),
+ Error::ToSqlConversionFailure(ref err) => err.fmt(f),
+ Error::InvalidQuery => write!(f, "Query is not read-only"),
+ #[cfg(feature = "vtab")]
+ Error::ModuleError(ref desc) => write!(f, "{desc}"),
+ #[cfg(feature = "functions")]
+ Error::UnwindingPanic => write!(f, "unwinding panic"),
+ #[cfg(feature = "functions")]
+ Error::GetAuxWrongType => write!(f, "get_aux called with wrong type"),
+ Error::MultipleStatement => write!(f, "Multiple statements provided"),
+ #[cfg(feature = "blob")]
+ Error::BlobSizeError => "Blob size is insufficient".fmt(f),
+ #[cfg(feature = "modern_sqlite")]
+ Error::SqlInputError {
+ ref msg,
+ offset,
+ ref sql,
+ ..
+ } => write!(f, "{msg} in {sql} at offset {offset}"),
+ #[cfg(feature = "loadable_extension")]
+ Error::InitError(ref err) => err.fmt(f),
+ }
+ }
+}
+
+impl error::Error for Error {
+ fn source(&self) -> Option<&(dyn error::Error + 'static)> {
+ match *self {
+ Error::SqliteFailure(ref err, _) => Some(err),
+ Error::Utf8Error(ref err) => Some(err),
+ Error::NulError(ref err) => Some(err),
+
+ Error::IntegralValueOutOfRange(..)
+ | Error::SqliteSingleThreadedMode
+ | Error::InvalidParameterName(_)
+ | Error::ExecuteReturnedResults
+ | Error::QueryReturnedNoRows
+ | Error::InvalidColumnIndex(_)
+ | Error::InvalidColumnName(_)
+ | Error::InvalidColumnType(..)
+ | Error::InvalidPath(_)
+ | Error::InvalidParameterCount(..)
+ | Error::StatementChangedRows(_)
+ | Error::InvalidQuery
+ | Error::MultipleStatement => None,
+
+ #[cfg(feature = "functions")]
+ Error::InvalidFunctionParameterType(..) => None,
+ #[cfg(feature = "vtab")]
+ Error::InvalidFilterParameterType(..) => None,
+
+ #[cfg(feature = "functions")]
+ Error::UserFunctionError(ref err) => Some(&**err),
+
+ Error::FromSqlConversionFailure(_, _, ref err)
+ | Error::ToSqlConversionFailure(ref err) => Some(&**err),
+
+ #[cfg(feature = "vtab")]
+ Error::ModuleError(_) => None,
+
+ #[cfg(feature = "functions")]
+ Error::UnwindingPanic => None,
+
+ #[cfg(feature = "functions")]
+ Error::GetAuxWrongType => None,
+
+ #[cfg(feature = "blob")]
+ Error::BlobSizeError => None,
+ #[cfg(feature = "modern_sqlite")]
+ Error::SqlInputError { ref error, .. } => Some(error),
+ #[cfg(feature = "loadable_extension")]
+ Error::InitError(ref err) => Some(err),
+ }
+ }
+}
+
+impl Error {
+ /// Returns the underlying SQLite error if this is [`Error::SqliteFailure`].
+ #[inline]
+ #[must_use]
+ pub fn sqlite_error(&self) -> Option<&ffi::Error> {
+ match self {
+ Self::SqliteFailure(error, _) => Some(error),
+ _ => None,
+ }
+ }
+
+ /// Returns the underlying SQLite error code if this is
+ /// [`Error::SqliteFailure`].
+ #[inline]
+ #[must_use]
+ pub fn sqlite_error_code(&self) -> Option<ffi::ErrorCode> {
+ self.sqlite_error().map(|error| error.code)
+ }
+}
+
+// These are public but not re-exported by lib.rs, so only visible within crate.
+
+#[cold]
+pub fn error_from_sqlite_code(code: c_int, message: Option<String>) -> Error {
+ Error::SqliteFailure(ffi::Error::new(code), message)
+}
+
+#[cold]
+pub unsafe fn error_from_handle(db: *mut ffi::sqlite3, code: c_int) -> Error {
+ let message = if db.is_null() {
+ None
+ } else {
+ Some(errmsg_to_string(ffi::sqlite3_errmsg(db)))
+ };
+ error_from_sqlite_code(code, message)
+}
+
+#[cold]
+#[cfg(not(feature = "modern_sqlite"))] // SQLite >= 3.38.0
+pub unsafe fn error_with_offset(db: *mut ffi::sqlite3, code: c_int, _sql: &str) -> Error {
+ error_from_handle(db, code)
+}
+
+#[cold]
+#[cfg(feature = "modern_sqlite")] // SQLite >= 3.38.0
+pub unsafe fn error_with_offset(db: *mut ffi::sqlite3, code: c_int, sql: &str) -> Error {
+ if db.is_null() {
+ error_from_sqlite_code(code, None)
+ } else {
+ let error = ffi::Error::new(code);
+ let msg = errmsg_to_string(ffi::sqlite3_errmsg(db));
+ if ffi::ErrorCode::Unknown == error.code {
+ let offset = ffi::sqlite3_error_offset(db);
+ if offset >= 0 {
+ return Error::SqlInputError {
+ error,
+ msg,
+ sql: sql.to_owned(),
+ offset,
+ };
+ }
+ }
+ Error::SqliteFailure(error, Some(msg))
+ }
+}
+
+pub fn check(code: c_int) -> Result<()> {
+ if code != crate::ffi::SQLITE_OK {
+ Err(error_from_sqlite_code(code, None))
+ } else {
+ Ok(())
+ }
+}
+
+/// Transform Rust error to SQLite error (message and code).
+/// # Safety
+/// This function is unsafe because it uses raw pointer
+pub unsafe fn to_sqlite_error(e: &Error, err_msg: *mut *mut std::os::raw::c_char) -> c_int {
+ use crate::util::alloc;
+ match e {
+ Error::SqliteFailure(err, s) => {
+ if let Some(s) = s {
+ *err_msg = alloc(s);
+ }
+ err.extended_code
+ }
+ err => {
+ *err_msg = alloc(&err.to_string());
+ ffi::SQLITE_ERROR
+ }
+ }
+}
diff --git a/third_party/rust/rusqlite/src/functions.rs b/third_party/rust/rusqlite/src/functions.rs
new file mode 100644
index 0000000000..522f1167d0
--- /dev/null
+++ b/third_party/rust/rusqlite/src/functions.rs
@@ -0,0 +1,1071 @@
+//! Create or redefine SQL functions.
+//!
+//! # Example
+//!
+//! Adding a `regexp` function to a connection in which compiled regular
+//! expressions are cached in a `HashMap`. For an alternative implementation
+//! that uses SQLite's [Function Auxiliary Data](https://www.sqlite.org/c3ref/get_auxdata.html) interface
+//! to avoid recompiling regular expressions, see the unit tests for this
+//! module.
+//!
+//! ```rust
+//! use regex::Regex;
+//! use rusqlite::functions::FunctionFlags;
+//! use rusqlite::{Connection, Error, Result};
+//! use std::sync::Arc;
+//! type BoxError = Box<dyn std::error::Error + Send + Sync + 'static>;
+//!
+//! fn add_regexp_function(db: &Connection) -> Result<()> {
+//! db.create_scalar_function(
+//! "regexp",
+//! 2,
+//! FunctionFlags::SQLITE_UTF8 | FunctionFlags::SQLITE_DETERMINISTIC,
+//! move |ctx| {
+//! assert_eq!(ctx.len(), 2, "called with unexpected number of arguments");
+//! let regexp: Arc<Regex> = ctx.get_or_create_aux(0, |vr| -> Result<_, BoxError> {
+//! Ok(Regex::new(vr.as_str()?)?)
+//! })?;
+//! let is_match = {
+//! let text = ctx
+//! .get_raw(1)
+//! .as_str()
+//! .map_err(|e| Error::UserFunctionError(e.into()))?;
+//!
+//! regexp.is_match(text)
+//! };
+//!
+//! Ok(is_match)
+//! },
+//! )
+//! }
+//!
+//! fn main() -> Result<()> {
+//! let db = Connection::open_in_memory()?;
+//! add_regexp_function(&db)?;
+//!
+//! let is_match: bool =
+//! db.query_row("SELECT regexp('[aeiou]*', 'aaaaeeeiii')", [], |row| {
+//! row.get(0)
+//! })?;
+//!
+//! assert!(is_match);
+//! Ok(())
+//! }
+//! ```
+use std::any::Any;
+use std::marker::PhantomData;
+use std::ops::Deref;
+use std::os::raw::{c_int, c_void};
+use std::panic::{catch_unwind, RefUnwindSafe, UnwindSafe};
+use std::ptr;
+use std::slice;
+use std::sync::Arc;
+
+use crate::ffi;
+use crate::ffi::sqlite3_context;
+use crate::ffi::sqlite3_value;
+
+use crate::context::set_result;
+use crate::types::{FromSql, FromSqlError, ToSql, ValueRef};
+
+use crate::{str_to_cstring, Connection, Error, InnerConnection, Result};
+
+unsafe fn report_error(ctx: *mut sqlite3_context, err: &Error) {
+ if let Error::SqliteFailure(ref err, ref s) = *err {
+ ffi::sqlite3_result_error_code(ctx, err.extended_code);
+ if let Some(Ok(cstr)) = s.as_ref().map(|s| str_to_cstring(s)) {
+ ffi::sqlite3_result_error(ctx, cstr.as_ptr(), -1);
+ }
+ } else {
+ ffi::sqlite3_result_error_code(ctx, ffi::SQLITE_CONSTRAINT_FUNCTION);
+ if let Ok(cstr) = str_to_cstring(&err.to_string()) {
+ ffi::sqlite3_result_error(ctx, cstr.as_ptr(), -1);
+ }
+ }
+}
+
+unsafe extern "C" fn free_boxed_value<T>(p: *mut c_void) {
+ drop(Box::from_raw(p.cast::<T>()));
+}
+
+/// Context is a wrapper for the SQLite function
+/// evaluation context.
+pub struct Context<'a> {
+ ctx: *mut sqlite3_context,
+ args: &'a [*mut sqlite3_value],
+}
+
+impl Context<'_> {
+ /// Returns the number of arguments to the function.
+ #[inline]
+ #[must_use]
+ pub fn len(&self) -> usize {
+ self.args.len()
+ }
+
+ /// Returns `true` when there is no argument.
+ #[inline]
+ #[must_use]
+ pub fn is_empty(&self) -> bool {
+ self.args.is_empty()
+ }
+
+ /// Returns the `idx`th argument as a `T`.
+ ///
+ /// # Failure
+ ///
+ /// Will panic if `idx` is greater than or equal to
+ /// [`self.len()`](Context::len).
+ ///
+ /// Will return Err if the underlying SQLite type cannot be converted to a
+ /// `T`.
+ pub fn get<T: FromSql>(&self, idx: usize) -> Result<T> {
+ let arg = self.args[idx];
+ let value = unsafe { ValueRef::from_value(arg) };
+ FromSql::column_result(value).map_err(|err| match err {
+ FromSqlError::InvalidType => {
+ Error::InvalidFunctionParameterType(idx, value.data_type())
+ }
+ FromSqlError::OutOfRange(i) => Error::IntegralValueOutOfRange(idx, i),
+ FromSqlError::Other(err) => {
+ Error::FromSqlConversionFailure(idx, value.data_type(), err)
+ }
+ FromSqlError::InvalidBlobSize { .. } => {
+ Error::FromSqlConversionFailure(idx, value.data_type(), Box::new(err))
+ }
+ })
+ }
+
+ /// Returns the `idx`th argument as a `ValueRef`.
+ ///
+ /// # Failure
+ ///
+ /// Will panic if `idx` is greater than or equal to
+ /// [`self.len()`](Context::len).
+ #[inline]
+ #[must_use]
+ pub fn get_raw(&self, idx: usize) -> ValueRef<'_> {
+ let arg = self.args[idx];
+ unsafe { ValueRef::from_value(arg) }
+ }
+
+ /// Returns the subtype of `idx`th argument.
+ ///
+ /// # Failure
+ ///
+ /// Will panic if `idx` is greater than or equal to
+ /// [`self.len()`](Context::len).
+ pub fn get_subtype(&self, idx: usize) -> std::os::raw::c_uint {
+ let arg = self.args[idx];
+ unsafe { ffi::sqlite3_value_subtype(arg) }
+ }
+
+ /// Fetch or insert the auxiliary data associated with a particular
+ /// parameter. This is intended to be an easier-to-use way of fetching it
+ /// compared to calling [`get_aux`](Context::get_aux) and
+ /// [`set_aux`](Context::set_aux) separately.
+ ///
+ /// See `https://www.sqlite.org/c3ref/get_auxdata.html` for a discussion of
+ /// this feature, or the unit tests of this module for an example.
+ pub fn get_or_create_aux<T, E, F>(&self, arg: c_int, func: F) -> Result<Arc<T>>
+ where
+ T: Send + Sync + 'static,
+ E: Into<Box<dyn std::error::Error + Send + Sync + 'static>>,
+ F: FnOnce(ValueRef<'_>) -> Result<T, E>,
+ {
+ if let Some(v) = self.get_aux(arg)? {
+ Ok(v)
+ } else {
+ let vr = self.get_raw(arg as usize);
+ self.set_aux(
+ arg,
+ func(vr).map_err(|e| Error::UserFunctionError(e.into()))?,
+ )
+ }
+ }
+
+ /// Sets the auxiliary data associated with a particular parameter. See
+ /// `https://www.sqlite.org/c3ref/get_auxdata.html` for a discussion of
+ /// this feature, or the unit tests of this module for an example.
+ pub fn set_aux<T: Send + Sync + 'static>(&self, arg: c_int, value: T) -> Result<Arc<T>> {
+ let orig: Arc<T> = Arc::new(value);
+ let inner: AuxInner = orig.clone();
+ let outer = Box::new(inner);
+ let raw: *mut AuxInner = Box::into_raw(outer);
+ unsafe {
+ ffi::sqlite3_set_auxdata(
+ self.ctx,
+ arg,
+ raw.cast(),
+ Some(free_boxed_value::<AuxInner>),
+ );
+ };
+ Ok(orig)
+ }
+
+ /// Gets the auxiliary data that was associated with a given parameter via
+ /// [`set_aux`](Context::set_aux). Returns `Ok(None)` if no data has been
+ /// associated, and Ok(Some(v)) if it has. Returns an error if the
+ /// requested type does not match.
+ pub fn get_aux<T: Send + Sync + 'static>(&self, arg: c_int) -> Result<Option<Arc<T>>> {
+ let p = unsafe { ffi::sqlite3_get_auxdata(self.ctx, arg) as *const AuxInner };
+ if p.is_null() {
+ Ok(None)
+ } else {
+ let v: AuxInner = AuxInner::clone(unsafe { &*p });
+ v.downcast::<T>()
+ .map(Some)
+ .map_err(|_| Error::GetAuxWrongType)
+ }
+ }
+
+ /// Get the db connection handle via [sqlite3_context_db_handle](https://www.sqlite.org/c3ref/context_db_handle.html)
+ ///
+ /// # Safety
+ ///
+ /// This function is marked unsafe because there is a potential for other
+ /// references to the connection to be sent across threads, [see this comment](https://github.com/rusqlite/rusqlite/issues/643#issuecomment-640181213).
+ pub unsafe fn get_connection(&self) -> Result<ConnectionRef<'_>> {
+ let handle = ffi::sqlite3_context_db_handle(self.ctx);
+ Ok(ConnectionRef {
+ conn: Connection::from_handle(handle)?,
+ phantom: PhantomData,
+ })
+ }
+
+ /// Set the Subtype of an SQL function
+ pub fn set_result_subtype(&self, sub_type: std::os::raw::c_uint) {
+ unsafe { ffi::sqlite3_result_subtype(self.ctx, sub_type) };
+ }
+}
+
+/// A reference to a connection handle with a lifetime bound to something.
+pub struct ConnectionRef<'ctx> {
+ // comes from Connection::from_handle(sqlite3_context_db_handle(...))
+ // and is non-owning
+ conn: Connection,
+ phantom: PhantomData<&'ctx Context<'ctx>>,
+}
+
+impl Deref for ConnectionRef<'_> {
+ type Target = Connection;
+
+ #[inline]
+ fn deref(&self) -> &Connection {
+ &self.conn
+ }
+}
+
+type AuxInner = Arc<dyn Any + Send + Sync + 'static>;
+
+/// Aggregate is the callback interface for user-defined
+/// aggregate function.
+///
+/// `A` is the type of the aggregation context and `T` is the type of the final
+/// result. Implementations should be stateless.
+pub trait Aggregate<A, T>
+where
+ A: RefUnwindSafe + UnwindSafe,
+ T: ToSql,
+{
+ /// Initializes the aggregation context. Will be called prior to the first
+ /// call to [`step()`](Aggregate::step) to set up the context for an
+ /// invocation of the function. (Note: `init()` will not be called if
+ /// there are no rows.)
+ fn init(&self, ctx: &mut Context<'_>) -> Result<A>;
+
+ /// "step" function called once for each row in an aggregate group. May be
+ /// called 0 times if there are no rows.
+ fn step(&self, ctx: &mut Context<'_>, acc: &mut A) -> Result<()>;
+
+ /// Computes and returns the final result. Will be called exactly once for
+ /// each invocation of the function. If [`step()`](Aggregate::step) was
+ /// called at least once, will be given `Some(A)` (the same `A` as was
+ /// created by [`init`](Aggregate::init) and given to
+ /// [`step`](Aggregate::step)); if [`step()`](Aggregate::step) was not
+ /// called (because the function is running against 0 rows), will be
+ /// given `None`.
+ ///
+ /// The passed context will have no arguments.
+ fn finalize(&self, ctx: &mut Context<'_>, acc: Option<A>) -> Result<T>;
+}
+
+/// `WindowAggregate` is the callback interface for
+/// user-defined aggregate window function.
+#[cfg(feature = "window")]
+#[cfg_attr(docsrs, doc(cfg(feature = "window")))]
+pub trait WindowAggregate<A, T>: Aggregate<A, T>
+where
+ A: RefUnwindSafe + UnwindSafe,
+ T: ToSql,
+{
+ /// Returns the current value of the aggregate. Unlike xFinal, the
+ /// implementation should not delete any context.
+ fn value(&self, acc: Option<&mut A>) -> Result<T>;
+
+ /// Removes a row from the current window.
+ fn inverse(&self, ctx: &mut Context<'_>, acc: &mut A) -> Result<()>;
+}
+
+bitflags::bitflags! {
+ /// Function Flags.
+ /// See [sqlite3_create_function](https://sqlite.org/c3ref/create_function.html)
+ /// and [Function Flags](https://sqlite.org/c3ref/c_deterministic.html) for details.
+ #[repr(C)]
+ pub struct FunctionFlags: ::std::os::raw::c_int {
+ /// Specifies UTF-8 as the text encoding this SQL function prefers for its parameters.
+ const SQLITE_UTF8 = ffi::SQLITE_UTF8;
+ /// Specifies UTF-16 using little-endian byte order as the text encoding this SQL function prefers for its parameters.
+ const SQLITE_UTF16LE = ffi::SQLITE_UTF16LE;
+ /// Specifies UTF-16 using big-endian byte order as the text encoding this SQL function prefers for its parameters.
+ const SQLITE_UTF16BE = ffi::SQLITE_UTF16BE;
+ /// Specifies UTF-16 using native byte order as the text encoding this SQL function prefers for its parameters.
+ const SQLITE_UTF16 = ffi::SQLITE_UTF16;
+ /// Means that the function always gives the same output when the input parameters are the same.
+ const SQLITE_DETERMINISTIC = ffi::SQLITE_DETERMINISTIC; // 3.8.3
+ /// Means that the function may only be invoked from top-level SQL.
+ const SQLITE_DIRECTONLY = 0x0000_0008_0000; // 3.30.0
+ /// Indicates to SQLite that a function may call `sqlite3_value_subtype()` to inspect the sub-types of its arguments.
+ const SQLITE_SUBTYPE = 0x0000_0010_0000; // 3.30.0
+ /// Means that the function is unlikely to cause problems even if misused.
+ const SQLITE_INNOCUOUS = 0x0000_0020_0000; // 3.31.0
+ }
+}
+
+impl Default for FunctionFlags {
+ #[inline]
+ fn default() -> FunctionFlags {
+ FunctionFlags::SQLITE_UTF8
+ }
+}
+
+impl Connection {
+ /// Attach a user-defined scalar function to
+ /// this database connection.
+ ///
+ /// `fn_name` is the name the function will be accessible from SQL.
+ /// `n_arg` is the number of arguments to the function. Use `-1` for a
+ /// variable number. If the function always returns the same value
+ /// given the same input, `deterministic` should be `true`.
+ ///
+ /// The function will remain available until the connection is closed or
+ /// until it is explicitly removed via
+ /// [`remove_function`](Connection::remove_function).
+ ///
+ /// # Example
+ ///
+ /// ```rust
+ /// # use rusqlite::{Connection, Result};
+ /// # use rusqlite::functions::FunctionFlags;
+ /// fn scalar_function_example(db: Connection) -> Result<()> {
+ /// db.create_scalar_function(
+ /// "halve",
+ /// 1,
+ /// FunctionFlags::SQLITE_UTF8 | FunctionFlags::SQLITE_DETERMINISTIC,
+ /// |ctx| {
+ /// let value = ctx.get::<f64>(0)?;
+ /// Ok(value / 2f64)
+ /// },
+ /// )?;
+ ///
+ /// let six_halved: f64 = db.query_row("SELECT halve(6)", [], |r| r.get(0))?;
+ /// assert_eq!(six_halved, 3f64);
+ /// Ok(())
+ /// }
+ /// ```
+ ///
+ /// # Failure
+ ///
+ /// Will return Err if the function could not be attached to the connection.
+ #[inline]
+ pub fn create_scalar_function<F, T>(
+ &self,
+ fn_name: &str,
+ n_arg: c_int,
+ flags: FunctionFlags,
+ x_func: F,
+ ) -> Result<()>
+ where
+ F: FnMut(&Context<'_>) -> Result<T> + Send + UnwindSafe + 'static,
+ T: ToSql,
+ {
+ self.db
+ .borrow_mut()
+ .create_scalar_function(fn_name, n_arg, flags, x_func)
+ }
+
+ /// Attach a user-defined aggregate function to this
+ /// database connection.
+ ///
+ /// # Failure
+ ///
+ /// Will return Err if the function could not be attached to the connection.
+ #[inline]
+ pub fn create_aggregate_function<A, D, T>(
+ &self,
+ fn_name: &str,
+ n_arg: c_int,
+ flags: FunctionFlags,
+ aggr: D,
+ ) -> Result<()>
+ where
+ A: RefUnwindSafe + UnwindSafe,
+ D: Aggregate<A, T> + 'static,
+ T: ToSql,
+ {
+ self.db
+ .borrow_mut()
+ .create_aggregate_function(fn_name, n_arg, flags, aggr)
+ }
+
+ /// Attach a user-defined aggregate window function to
+ /// this database connection.
+ ///
+ /// See `https://sqlite.org/windowfunctions.html#udfwinfunc` for more
+ /// information.
+ #[cfg(feature = "window")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "window")))]
+ #[inline]
+ pub fn create_window_function<A, W, T>(
+ &self,
+ fn_name: &str,
+ n_arg: c_int,
+ flags: FunctionFlags,
+ aggr: W,
+ ) -> Result<()>
+ where
+ A: RefUnwindSafe + UnwindSafe,
+ W: WindowAggregate<A, T> + 'static,
+ T: ToSql,
+ {
+ self.db
+ .borrow_mut()
+ .create_window_function(fn_name, n_arg, flags, aggr)
+ }
+
+ /// Removes a user-defined function from this
+ /// database connection.
+ ///
+ /// `fn_name` and `n_arg` should match the name and number of arguments
+ /// given to [`create_scalar_function`](Connection::create_scalar_function)
+ /// or [`create_aggregate_function`](Connection::create_aggregate_function).
+ ///
+ /// # Failure
+ ///
+ /// Will return Err if the function could not be removed.
+ #[inline]
+ pub fn remove_function(&self, fn_name: &str, n_arg: c_int) -> Result<()> {
+ self.db.borrow_mut().remove_function(fn_name, n_arg)
+ }
+}
+
+impl InnerConnection {
+ fn create_scalar_function<F, T>(
+ &mut self,
+ fn_name: &str,
+ n_arg: c_int,
+ flags: FunctionFlags,
+ x_func: F,
+ ) -> Result<()>
+ where
+ F: FnMut(&Context<'_>) -> Result<T> + Send + UnwindSafe + 'static,
+ T: ToSql,
+ {
+ unsafe extern "C" fn call_boxed_closure<F, T>(
+ ctx: *mut sqlite3_context,
+ argc: c_int,
+ argv: *mut *mut sqlite3_value,
+ ) where
+ F: FnMut(&Context<'_>) -> Result<T>,
+ T: ToSql,
+ {
+ let r = catch_unwind(|| {
+ let boxed_f: *mut F = ffi::sqlite3_user_data(ctx).cast::<F>();
+ assert!(!boxed_f.is_null(), "Internal error - null function pointer");
+ let ctx = Context {
+ ctx,
+ args: slice::from_raw_parts(argv, argc as usize),
+ };
+ (*boxed_f)(&ctx)
+ });
+ let t = match r {
+ Err(_) => {
+ report_error(ctx, &Error::UnwindingPanic);
+ return;
+ }
+ Ok(r) => r,
+ };
+ let t = t.as_ref().map(|t| ToSql::to_sql(t));
+
+ match t {
+ Ok(Ok(ref value)) => set_result(ctx, value),
+ Ok(Err(err)) => report_error(ctx, &err),
+ Err(err) => report_error(ctx, err),
+ }
+ }
+
+ let boxed_f: *mut F = Box::into_raw(Box::new(x_func));
+ let c_name = str_to_cstring(fn_name)?;
+ let r = unsafe {
+ ffi::sqlite3_create_function_v2(
+ self.db(),
+ c_name.as_ptr(),
+ n_arg,
+ flags.bits(),
+ boxed_f.cast::<c_void>(),
+ Some(call_boxed_closure::<F, T>),
+ None,
+ None,
+ Some(free_boxed_value::<F>),
+ )
+ };
+ self.decode_result(r)
+ }
+
+ fn create_aggregate_function<A, D, T>(
+ &mut self,
+ fn_name: &str,
+ n_arg: c_int,
+ flags: FunctionFlags,
+ aggr: D,
+ ) -> Result<()>
+ where
+ A: RefUnwindSafe + UnwindSafe,
+ D: Aggregate<A, T> + 'static,
+ T: ToSql,
+ {
+ let boxed_aggr: *mut D = Box::into_raw(Box::new(aggr));
+ let c_name = str_to_cstring(fn_name)?;
+ let r = unsafe {
+ ffi::sqlite3_create_function_v2(
+ self.db(),
+ c_name.as_ptr(),
+ n_arg,
+ flags.bits(),
+ boxed_aggr.cast::<c_void>(),
+ None,
+ Some(call_boxed_step::<A, D, T>),
+ Some(call_boxed_final::<A, D, T>),
+ Some(free_boxed_value::<D>),
+ )
+ };
+ self.decode_result(r)
+ }
+
+ #[cfg(feature = "window")]
+ fn create_window_function<A, W, T>(
+ &mut self,
+ fn_name: &str,
+ n_arg: c_int,
+ flags: FunctionFlags,
+ aggr: W,
+ ) -> Result<()>
+ where
+ A: RefUnwindSafe + UnwindSafe,
+ W: WindowAggregate<A, T> + 'static,
+ T: ToSql,
+ {
+ let boxed_aggr: *mut W = Box::into_raw(Box::new(aggr));
+ let c_name = str_to_cstring(fn_name)?;
+ let r = unsafe {
+ ffi::sqlite3_create_window_function(
+ self.db(),
+ c_name.as_ptr(),
+ n_arg,
+ flags.bits(),
+ boxed_aggr.cast::<c_void>(),
+ Some(call_boxed_step::<A, W, T>),
+ Some(call_boxed_final::<A, W, T>),
+ Some(call_boxed_value::<A, W, T>),
+ Some(call_boxed_inverse::<A, W, T>),
+ Some(free_boxed_value::<W>),
+ )
+ };
+ self.decode_result(r)
+ }
+
+ fn remove_function(&mut self, fn_name: &str, n_arg: c_int) -> Result<()> {
+ let c_name = str_to_cstring(fn_name)?;
+ let r = unsafe {
+ ffi::sqlite3_create_function_v2(
+ self.db(),
+ c_name.as_ptr(),
+ n_arg,
+ ffi::SQLITE_UTF8,
+ ptr::null_mut(),
+ None,
+ None,
+ None,
+ None,
+ )
+ };
+ self.decode_result(r)
+ }
+}
+
+unsafe fn aggregate_context<A>(ctx: *mut sqlite3_context, bytes: usize) -> Option<*mut *mut A> {
+ let pac = ffi::sqlite3_aggregate_context(ctx, bytes as c_int) as *mut *mut A;
+ if pac.is_null() {
+ return None;
+ }
+ Some(pac)
+}
+
+unsafe extern "C" fn call_boxed_step<A, D, T>(
+ ctx: *mut sqlite3_context,
+ argc: c_int,
+ argv: *mut *mut sqlite3_value,
+) where
+ A: RefUnwindSafe + UnwindSafe,
+ D: Aggregate<A, T>,
+ T: ToSql,
+{
+ let pac = if let Some(pac) = aggregate_context(ctx, std::mem::size_of::<*mut A>()) {
+ pac
+ } else {
+ ffi::sqlite3_result_error_nomem(ctx);
+ return;
+ };
+
+ let r = catch_unwind(|| {
+ let boxed_aggr: *mut D = ffi::sqlite3_user_data(ctx).cast::<D>();
+ assert!(
+ !boxed_aggr.is_null(),
+ "Internal error - null aggregate pointer"
+ );
+ let mut ctx = Context {
+ ctx,
+ args: slice::from_raw_parts(argv, argc as usize),
+ };
+
+ #[allow(clippy::unnecessary_cast)]
+ if (*pac as *mut A).is_null() {
+ *pac = Box::into_raw(Box::new((*boxed_aggr).init(&mut ctx)?));
+ }
+
+ (*boxed_aggr).step(&mut ctx, &mut **pac)
+ });
+ let r = match r {
+ Err(_) => {
+ report_error(ctx, &Error::UnwindingPanic);
+ return;
+ }
+ Ok(r) => r,
+ };
+ match r {
+ Ok(_) => {}
+ Err(err) => report_error(ctx, &err),
+ };
+}
+
+#[cfg(feature = "window")]
+unsafe extern "C" fn call_boxed_inverse<A, W, T>(
+ ctx: *mut sqlite3_context,
+ argc: c_int,
+ argv: *mut *mut sqlite3_value,
+) where
+ A: RefUnwindSafe + UnwindSafe,
+ W: WindowAggregate<A, T>,
+ T: ToSql,
+{
+ let pac = if let Some(pac) = aggregate_context(ctx, std::mem::size_of::<*mut A>()) {
+ pac
+ } else {
+ ffi::sqlite3_result_error_nomem(ctx);
+ return;
+ };
+
+ let r = catch_unwind(|| {
+ let boxed_aggr: *mut W = ffi::sqlite3_user_data(ctx).cast::<W>();
+ assert!(
+ !boxed_aggr.is_null(),
+ "Internal error - null aggregate pointer"
+ );
+ let mut ctx = Context {
+ ctx,
+ args: slice::from_raw_parts(argv, argc as usize),
+ };
+ (*boxed_aggr).inverse(&mut ctx, &mut **pac)
+ });
+ let r = match r {
+ Err(_) => {
+ report_error(ctx, &Error::UnwindingPanic);
+ return;
+ }
+ Ok(r) => r,
+ };
+ match r {
+ Ok(_) => {}
+ Err(err) => report_error(ctx, &err),
+ };
+}
+
+unsafe extern "C" fn call_boxed_final<A, D, T>(ctx: *mut sqlite3_context)
+where
+ A: RefUnwindSafe + UnwindSafe,
+ D: Aggregate<A, T>,
+ T: ToSql,
+{
+ // Within the xFinal callback, it is customary to set N=0 in calls to
+ // sqlite3_aggregate_context(C,N) so that no pointless memory allocations occur.
+ let a: Option<A> = match aggregate_context(ctx, 0) {
+ Some(pac) =>
+ {
+ #[allow(clippy::unnecessary_cast)]
+ if (*pac as *mut A).is_null() {
+ None
+ } else {
+ let a = Box::from_raw(*pac);
+ Some(*a)
+ }
+ }
+ None => None,
+ };
+
+ let r = catch_unwind(|| {
+ let boxed_aggr: *mut D = ffi::sqlite3_user_data(ctx).cast::<D>();
+ assert!(
+ !boxed_aggr.is_null(),
+ "Internal error - null aggregate pointer"
+ );
+ let mut ctx = Context { ctx, args: &mut [] };
+ (*boxed_aggr).finalize(&mut ctx, a)
+ });
+ let t = match r {
+ Err(_) => {
+ report_error(ctx, &Error::UnwindingPanic);
+ return;
+ }
+ Ok(r) => r,
+ };
+ let t = t.as_ref().map(|t| ToSql::to_sql(t));
+ match t {
+ Ok(Ok(ref value)) => set_result(ctx, value),
+ Ok(Err(err)) => report_error(ctx, &err),
+ Err(err) => report_error(ctx, err),
+ }
+}
+
+#[cfg(feature = "window")]
+unsafe extern "C" fn call_boxed_value<A, W, T>(ctx: *mut sqlite3_context)
+where
+ A: RefUnwindSafe + UnwindSafe,
+ W: WindowAggregate<A, T>,
+ T: ToSql,
+{
+ // Within the xValue callback, it is customary to set N=0 in calls to
+ // sqlite3_aggregate_context(C,N) so that no pointless memory allocations occur.
+ let pac = aggregate_context(ctx, 0).filter(|&pac| {
+ #[allow(clippy::unnecessary_cast)]
+ !(*pac as *mut A).is_null()
+ });
+
+ let r = catch_unwind(|| {
+ let boxed_aggr: *mut W = ffi::sqlite3_user_data(ctx).cast::<W>();
+ assert!(
+ !boxed_aggr.is_null(),
+ "Internal error - null aggregate pointer"
+ );
+ (*boxed_aggr).value(pac.map(|pac| &mut **pac))
+ });
+ let t = match r {
+ Err(_) => {
+ report_error(ctx, &Error::UnwindingPanic);
+ return;
+ }
+ Ok(r) => r,
+ };
+ let t = t.as_ref().map(|t| ToSql::to_sql(t));
+ match t {
+ Ok(Ok(ref value)) => set_result(ctx, value),
+ Ok(Err(err)) => report_error(ctx, &err),
+ Err(err) => report_error(ctx, err),
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use regex::Regex;
+ use std::os::raw::c_double;
+
+ #[cfg(feature = "window")]
+ use crate::functions::WindowAggregate;
+ use crate::functions::{Aggregate, Context, FunctionFlags};
+ use crate::{Connection, Error, Result};
+
+ fn half(ctx: &Context<'_>) -> Result<c_double> {
+ assert_eq!(ctx.len(), 1, "called with unexpected number of arguments");
+ let value = ctx.get::<c_double>(0)?;
+ Ok(value / 2f64)
+ }
+
+ #[test]
+ fn test_function_half() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.create_scalar_function(
+ "half",
+ 1,
+ FunctionFlags::SQLITE_UTF8 | FunctionFlags::SQLITE_DETERMINISTIC,
+ half,
+ )?;
+ let result: f64 = db.one_column("SELECT half(6)")?;
+
+ assert!((3f64 - result).abs() < f64::EPSILON);
+ Ok(())
+ }
+
+ #[test]
+ fn test_remove_function() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.create_scalar_function(
+ "half",
+ 1,
+ FunctionFlags::SQLITE_UTF8 | FunctionFlags::SQLITE_DETERMINISTIC,
+ half,
+ )?;
+ let result: f64 = db.one_column("SELECT half(6)")?;
+ assert!((3f64 - result).abs() < f64::EPSILON);
+
+ db.remove_function("half", 1)?;
+ let result: Result<f64> = db.one_column("SELECT half(6)");
+ result.unwrap_err();
+ Ok(())
+ }
+
+ // This implementation of a regexp scalar function uses SQLite's auxiliary data
+ // (https://www.sqlite.org/c3ref/get_auxdata.html) to avoid recompiling the regular
+ // expression multiple times within one query.
+ fn regexp_with_auxiliary(ctx: &Context<'_>) -> Result<bool> {
+ assert_eq!(ctx.len(), 2, "called with unexpected number of arguments");
+ type BoxError = Box<dyn std::error::Error + Send + Sync + 'static>;
+ let regexp: std::sync::Arc<Regex> = ctx
+ .get_or_create_aux(0, |vr| -> Result<_, BoxError> {
+ Ok(Regex::new(vr.as_str()?)?)
+ })?;
+
+ let is_match = {
+ let text = ctx
+ .get_raw(1)
+ .as_str()
+ .map_err(|e| Error::UserFunctionError(e.into()))?;
+
+ regexp.is_match(text)
+ };
+
+ Ok(is_match)
+ }
+
+ #[test]
+ fn test_function_regexp_with_auxiliary() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch(
+ "BEGIN;
+ CREATE TABLE foo (x string);
+ INSERT INTO foo VALUES ('lisa');
+ INSERT INTO foo VALUES ('lXsi');
+ INSERT INTO foo VALUES ('lisX');
+ END;",
+ )?;
+ db.create_scalar_function(
+ "regexp",
+ 2,
+ FunctionFlags::SQLITE_UTF8 | FunctionFlags::SQLITE_DETERMINISTIC,
+ regexp_with_auxiliary,
+ )?;
+
+ let result: bool = db.one_column("SELECT regexp('l.s[aeiouy]', 'lisa')")?;
+
+ assert!(result);
+
+ let result: i64 =
+ db.one_column("SELECT COUNT(*) FROM foo WHERE regexp('l.s[aeiouy]', x) == 1")?;
+
+ assert_eq!(2, result);
+ Ok(())
+ }
+
+ #[test]
+ fn test_varargs_function() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.create_scalar_function(
+ "my_concat",
+ -1,
+ FunctionFlags::SQLITE_UTF8 | FunctionFlags::SQLITE_DETERMINISTIC,
+ |ctx| {
+ let mut ret = String::new();
+
+ for idx in 0..ctx.len() {
+ let s = ctx.get::<String>(idx)?;
+ ret.push_str(&s);
+ }
+
+ Ok(ret)
+ },
+ )?;
+
+ for &(expected, query) in &[
+ ("", "SELECT my_concat()"),
+ ("onetwo", "SELECT my_concat('one', 'two')"),
+ ("abc", "SELECT my_concat('a', 'b', 'c')"),
+ ] {
+ let result: String = db.one_column(query)?;
+ assert_eq!(expected, result);
+ }
+ Ok(())
+ }
+
+ #[test]
+ fn test_get_aux_type_checking() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.create_scalar_function("example", 2, FunctionFlags::default(), |ctx| {
+ if !ctx.get::<bool>(1)? {
+ ctx.set_aux::<i64>(0, 100)?;
+ } else {
+ assert_eq!(ctx.get_aux::<String>(0), Err(Error::GetAuxWrongType));
+ assert_eq!(*ctx.get_aux::<i64>(0)?.unwrap(), 100);
+ }
+ Ok(true)
+ })?;
+
+ let res: bool =
+ db.one_column("SELECT example(0, i) FROM (SELECT 0 as i UNION SELECT 1)")?;
+ // Doesn't actually matter, we'll assert in the function if there's a problem.
+ assert!(res);
+ Ok(())
+ }
+
+ struct Sum;
+ struct Count;
+
+ impl Aggregate<i64, Option<i64>> for Sum {
+ fn init(&self, _: &mut Context<'_>) -> Result<i64> {
+ Ok(0)
+ }
+
+ fn step(&self, ctx: &mut Context<'_>, sum: &mut i64) -> Result<()> {
+ *sum += ctx.get::<i64>(0)?;
+ Ok(())
+ }
+
+ fn finalize(&self, _: &mut Context<'_>, sum: Option<i64>) -> Result<Option<i64>> {
+ Ok(sum)
+ }
+ }
+
+ impl Aggregate<i64, i64> for Count {
+ fn init(&self, _: &mut Context<'_>) -> Result<i64> {
+ Ok(0)
+ }
+
+ fn step(&self, _ctx: &mut Context<'_>, sum: &mut i64) -> Result<()> {
+ *sum += 1;
+ Ok(())
+ }
+
+ fn finalize(&self, _: &mut Context<'_>, sum: Option<i64>) -> Result<i64> {
+ Ok(sum.unwrap_or(0))
+ }
+ }
+
+ #[test]
+ fn test_sum() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.create_aggregate_function(
+ "my_sum",
+ 1,
+ FunctionFlags::SQLITE_UTF8 | FunctionFlags::SQLITE_DETERMINISTIC,
+ Sum,
+ )?;
+
+ // sum should return NULL when given no columns (contrast with count below)
+ let no_result = "SELECT my_sum(i) FROM (SELECT 2 AS i WHERE 1 <> 1)";
+ let result: Option<i64> = db.one_column(no_result)?;
+ assert!(result.is_none());
+
+ let single_sum = "SELECT my_sum(i) FROM (SELECT 2 AS i UNION ALL SELECT 2)";
+ let result: i64 = db.one_column(single_sum)?;
+ assert_eq!(4, result);
+
+ let dual_sum = "SELECT my_sum(i), my_sum(j) FROM (SELECT 2 AS i, 1 AS j UNION ALL SELECT \
+ 2, 1)";
+ let result: (i64, i64) = db.query_row(dual_sum, [], |r| Ok((r.get(0)?, r.get(1)?)))?;
+ assert_eq!((4, 2), result);
+ Ok(())
+ }
+
+ #[test]
+ fn test_count() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.create_aggregate_function(
+ "my_count",
+ -1,
+ FunctionFlags::SQLITE_UTF8 | FunctionFlags::SQLITE_DETERMINISTIC,
+ Count,
+ )?;
+
+ // count should return 0 when given no columns (contrast with sum above)
+ let no_result = "SELECT my_count(i) FROM (SELECT 2 AS i WHERE 1 <> 1)";
+ let result: i64 = db.one_column(no_result)?;
+ assert_eq!(result, 0);
+
+ let single_sum = "SELECT my_count(i) FROM (SELECT 2 AS i UNION ALL SELECT 2)";
+ let result: i64 = db.one_column(single_sum)?;
+ assert_eq!(2, result);
+ Ok(())
+ }
+
+ #[cfg(feature = "window")]
+ impl WindowAggregate<i64, Option<i64>> for Sum {
+ fn inverse(&self, ctx: &mut Context<'_>, sum: &mut i64) -> Result<()> {
+ *sum -= ctx.get::<i64>(0)?;
+ Ok(())
+ }
+
+ fn value(&self, sum: Option<&mut i64>) -> Result<Option<i64>> {
+ Ok(sum.copied())
+ }
+ }
+
+ #[test]
+ #[cfg(feature = "window")]
+ fn test_window() -> Result<()> {
+ use fallible_iterator::FallibleIterator;
+
+ let db = Connection::open_in_memory()?;
+ db.create_window_function(
+ "sumint",
+ 1,
+ FunctionFlags::SQLITE_UTF8 | FunctionFlags::SQLITE_DETERMINISTIC,
+ Sum,
+ )?;
+ db.execute_batch(
+ "CREATE TABLE t3(x, y);
+ INSERT INTO t3 VALUES('a', 4),
+ ('b', 5),
+ ('c', 3),
+ ('d', 8),
+ ('e', 1);",
+ )?;
+
+ let mut stmt = db.prepare(
+ "SELECT x, sumint(y) OVER (
+ ORDER BY x ROWS BETWEEN 1 PRECEDING AND 1 FOLLOWING
+ ) AS sum_y
+ FROM t3 ORDER BY x;",
+ )?;
+
+ let results: Vec<(String, i64)> = stmt
+ .query([])?
+ .map(|row| Ok((row.get("x")?, row.get("sum_y")?)))
+ .collect()?;
+ let expected = vec![
+ ("a".to_owned(), 9),
+ ("b".to_owned(), 12),
+ ("c".to_owned(), 16),
+ ("d".to_owned(), 12),
+ ("e".to_owned(), 9),
+ ];
+ assert_eq!(expected, results);
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/hooks.rs b/third_party/rust/rusqlite/src/hooks.rs
new file mode 100644
index 0000000000..52a53a3886
--- /dev/null
+++ b/third_party/rust/rusqlite/src/hooks.rs
@@ -0,0 +1,806 @@
+//! Commit, Data Change and Rollback Notification Callbacks
+#![allow(non_camel_case_types)]
+
+use std::os::raw::{c_char, c_int, c_void};
+use std::panic::{catch_unwind, RefUnwindSafe};
+use std::ptr;
+
+use crate::ffi;
+
+use crate::{Connection, InnerConnection};
+
+/// Action Codes
+#[derive(Clone, Copy, Debug, Eq, PartialEq)]
+#[repr(i32)]
+#[non_exhaustive]
+#[allow(clippy::upper_case_acronyms)]
+pub enum Action {
+ /// Unsupported / unexpected action
+ UNKNOWN = -1,
+ /// DELETE command
+ SQLITE_DELETE = ffi::SQLITE_DELETE,
+ /// INSERT command
+ SQLITE_INSERT = ffi::SQLITE_INSERT,
+ /// UPDATE command
+ SQLITE_UPDATE = ffi::SQLITE_UPDATE,
+}
+
+impl From<i32> for Action {
+ #[inline]
+ fn from(code: i32) -> Action {
+ match code {
+ ffi::SQLITE_DELETE => Action::SQLITE_DELETE,
+ ffi::SQLITE_INSERT => Action::SQLITE_INSERT,
+ ffi::SQLITE_UPDATE => Action::SQLITE_UPDATE,
+ _ => Action::UNKNOWN,
+ }
+ }
+}
+
+/// The context received by an authorizer hook.
+///
+/// See <https://sqlite.org/c3ref/set_authorizer.html> for more info.
+#[derive(Clone, Copy, Debug, Eq, PartialEq)]
+pub struct AuthContext<'c> {
+ /// The action to be authorized.
+ pub action: AuthAction<'c>,
+
+ /// The database name, if applicable.
+ pub database_name: Option<&'c str>,
+
+ /// The inner-most trigger or view responsible for the access attempt.
+ /// `None` if the access attempt was made by top-level SQL code.
+ pub accessor: Option<&'c str>,
+}
+
+/// Actions and arguments found within a statement during
+/// preparation.
+///
+/// See <https://sqlite.org/c3ref/c_alter_table.html> for more info.
+#[derive(Clone, Copy, Debug, Eq, PartialEq)]
+#[non_exhaustive]
+#[allow(missing_docs)]
+pub enum AuthAction<'c> {
+ /// This variant is not normally produced by SQLite. You may encounter it
+ // if you're using a different version than what's supported by this library.
+ Unknown {
+ /// The unknown authorization action code.
+ code: i32,
+ /// The third arg to the authorizer callback.
+ arg1: Option<&'c str>,
+ /// The fourth arg to the authorizer callback.
+ arg2: Option<&'c str>,
+ },
+ CreateIndex {
+ index_name: &'c str,
+ table_name: &'c str,
+ },
+ CreateTable {
+ table_name: &'c str,
+ },
+ CreateTempIndex {
+ index_name: &'c str,
+ table_name: &'c str,
+ },
+ CreateTempTable {
+ table_name: &'c str,
+ },
+ CreateTempTrigger {
+ trigger_name: &'c str,
+ table_name: &'c str,
+ },
+ CreateTempView {
+ view_name: &'c str,
+ },
+ CreateTrigger {
+ trigger_name: &'c str,
+ table_name: &'c str,
+ },
+ CreateView {
+ view_name: &'c str,
+ },
+ Delete {
+ table_name: &'c str,
+ },
+ DropIndex {
+ index_name: &'c str,
+ table_name: &'c str,
+ },
+ DropTable {
+ table_name: &'c str,
+ },
+ DropTempIndex {
+ index_name: &'c str,
+ table_name: &'c str,
+ },
+ DropTempTable {
+ table_name: &'c str,
+ },
+ DropTempTrigger {
+ trigger_name: &'c str,
+ table_name: &'c str,
+ },
+ DropTempView {
+ view_name: &'c str,
+ },
+ DropTrigger {
+ trigger_name: &'c str,
+ table_name: &'c str,
+ },
+ DropView {
+ view_name: &'c str,
+ },
+ Insert {
+ table_name: &'c str,
+ },
+ Pragma {
+ pragma_name: &'c str,
+ /// The pragma value, if present (e.g., `PRAGMA name = value;`).
+ pragma_value: Option<&'c str>,
+ },
+ Read {
+ table_name: &'c str,
+ column_name: &'c str,
+ },
+ Select,
+ Transaction {
+ operation: TransactionOperation,
+ },
+ Update {
+ table_name: &'c str,
+ column_name: &'c str,
+ },
+ Attach {
+ filename: &'c str,
+ },
+ Detach {
+ database_name: &'c str,
+ },
+ AlterTable {
+ database_name: &'c str,
+ table_name: &'c str,
+ },
+ Reindex {
+ index_name: &'c str,
+ },
+ Analyze {
+ table_name: &'c str,
+ },
+ CreateVtable {
+ table_name: &'c str,
+ module_name: &'c str,
+ },
+ DropVtable {
+ table_name: &'c str,
+ module_name: &'c str,
+ },
+ Function {
+ function_name: &'c str,
+ },
+ Savepoint {
+ operation: TransactionOperation,
+ savepoint_name: &'c str,
+ },
+ Recursive,
+}
+
+impl<'c> AuthAction<'c> {
+ fn from_raw(code: i32, arg1: Option<&'c str>, arg2: Option<&'c str>) -> Self {
+ match (code, arg1, arg2) {
+ (ffi::SQLITE_CREATE_INDEX, Some(index_name), Some(table_name)) => Self::CreateIndex {
+ index_name,
+ table_name,
+ },
+ (ffi::SQLITE_CREATE_TABLE, Some(table_name), _) => Self::CreateTable { table_name },
+ (ffi::SQLITE_CREATE_TEMP_INDEX, Some(index_name), Some(table_name)) => {
+ Self::CreateTempIndex {
+ index_name,
+ table_name,
+ }
+ }
+ (ffi::SQLITE_CREATE_TEMP_TABLE, Some(table_name), _) => {
+ Self::CreateTempTable { table_name }
+ }
+ (ffi::SQLITE_CREATE_TEMP_TRIGGER, Some(trigger_name), Some(table_name)) => {
+ Self::CreateTempTrigger {
+ trigger_name,
+ table_name,
+ }
+ }
+ (ffi::SQLITE_CREATE_TEMP_VIEW, Some(view_name), _) => {
+ Self::CreateTempView { view_name }
+ }
+ (ffi::SQLITE_CREATE_TRIGGER, Some(trigger_name), Some(table_name)) => {
+ Self::CreateTrigger {
+ trigger_name,
+ table_name,
+ }
+ }
+ (ffi::SQLITE_CREATE_VIEW, Some(view_name), _) => Self::CreateView { view_name },
+ (ffi::SQLITE_DELETE, Some(table_name), None) => Self::Delete { table_name },
+ (ffi::SQLITE_DROP_INDEX, Some(index_name), Some(table_name)) => Self::DropIndex {
+ index_name,
+ table_name,
+ },
+ (ffi::SQLITE_DROP_TABLE, Some(table_name), _) => Self::DropTable { table_name },
+ (ffi::SQLITE_DROP_TEMP_INDEX, Some(index_name), Some(table_name)) => {
+ Self::DropTempIndex {
+ index_name,
+ table_name,
+ }
+ }
+ (ffi::SQLITE_DROP_TEMP_TABLE, Some(table_name), _) => {
+ Self::DropTempTable { table_name }
+ }
+ (ffi::SQLITE_DROP_TEMP_TRIGGER, Some(trigger_name), Some(table_name)) => {
+ Self::DropTempTrigger {
+ trigger_name,
+ table_name,
+ }
+ }
+ (ffi::SQLITE_DROP_TEMP_VIEW, Some(view_name), _) => Self::DropTempView { view_name },
+ (ffi::SQLITE_DROP_TRIGGER, Some(trigger_name), Some(table_name)) => Self::DropTrigger {
+ trigger_name,
+ table_name,
+ },
+ (ffi::SQLITE_DROP_VIEW, Some(view_name), _) => Self::DropView { view_name },
+ (ffi::SQLITE_INSERT, Some(table_name), _) => Self::Insert { table_name },
+ (ffi::SQLITE_PRAGMA, Some(pragma_name), pragma_value) => Self::Pragma {
+ pragma_name,
+ pragma_value,
+ },
+ (ffi::SQLITE_READ, Some(table_name), Some(column_name)) => Self::Read {
+ table_name,
+ column_name,
+ },
+ (ffi::SQLITE_SELECT, ..) => Self::Select,
+ (ffi::SQLITE_TRANSACTION, Some(operation_str), _) => Self::Transaction {
+ operation: TransactionOperation::from_str(operation_str),
+ },
+ (ffi::SQLITE_UPDATE, Some(table_name), Some(column_name)) => Self::Update {
+ table_name,
+ column_name,
+ },
+ (ffi::SQLITE_ATTACH, Some(filename), _) => Self::Attach { filename },
+ (ffi::SQLITE_DETACH, Some(database_name), _) => Self::Detach { database_name },
+ (ffi::SQLITE_ALTER_TABLE, Some(database_name), Some(table_name)) => Self::AlterTable {
+ database_name,
+ table_name,
+ },
+ (ffi::SQLITE_REINDEX, Some(index_name), _) => Self::Reindex { index_name },
+ (ffi::SQLITE_ANALYZE, Some(table_name), _) => Self::Analyze { table_name },
+ (ffi::SQLITE_CREATE_VTABLE, Some(table_name), Some(module_name)) => {
+ Self::CreateVtable {
+ table_name,
+ module_name,
+ }
+ }
+ (ffi::SQLITE_DROP_VTABLE, Some(table_name), Some(module_name)) => Self::DropVtable {
+ table_name,
+ module_name,
+ },
+ (ffi::SQLITE_FUNCTION, _, Some(function_name)) => Self::Function { function_name },
+ (ffi::SQLITE_SAVEPOINT, Some(operation_str), Some(savepoint_name)) => Self::Savepoint {
+ operation: TransactionOperation::from_str(operation_str),
+ savepoint_name,
+ },
+ (ffi::SQLITE_RECURSIVE, ..) => Self::Recursive,
+ (code, arg1, arg2) => Self::Unknown { code, arg1, arg2 },
+ }
+ }
+}
+
+pub(crate) type BoxedAuthorizer =
+ Box<dyn for<'c> FnMut(AuthContext<'c>) -> Authorization + Send + 'static>;
+
+/// A transaction operation.
+#[derive(Clone, Copy, Debug, Eq, PartialEq)]
+#[non_exhaustive]
+#[allow(missing_docs)]
+pub enum TransactionOperation {
+ Unknown,
+ Begin,
+ Release,
+ Rollback,
+}
+
+impl TransactionOperation {
+ fn from_str(op_str: &str) -> Self {
+ match op_str {
+ "BEGIN" => Self::Begin,
+ "RELEASE" => Self::Release,
+ "ROLLBACK" => Self::Rollback,
+ _ => Self::Unknown,
+ }
+ }
+}
+
+/// [`authorizer`](Connection::authorizer) return code
+#[derive(Clone, Copy, Debug, Eq, PartialEq)]
+#[non_exhaustive]
+pub enum Authorization {
+ /// Authorize the action.
+ Allow,
+ /// Don't allow access, but don't trigger an error either.
+ Ignore,
+ /// Trigger an error.
+ Deny,
+}
+
+impl Authorization {
+ fn into_raw(self) -> c_int {
+ match self {
+ Self::Allow => ffi::SQLITE_OK,
+ Self::Ignore => ffi::SQLITE_IGNORE,
+ Self::Deny => ffi::SQLITE_DENY,
+ }
+ }
+}
+
+impl Connection {
+ /// Register a callback function to be invoked whenever
+ /// a transaction is committed.
+ ///
+ /// The callback returns `true` to rollback.
+ #[inline]
+ pub fn commit_hook<F>(&self, hook: Option<F>)
+ where
+ F: FnMut() -> bool + Send + 'static,
+ {
+ self.db.borrow_mut().commit_hook(hook);
+ }
+
+ /// Register a callback function to be invoked whenever
+ /// a transaction is committed.
+ #[inline]
+ pub fn rollback_hook<F>(&self, hook: Option<F>)
+ where
+ F: FnMut() + Send + 'static,
+ {
+ self.db.borrow_mut().rollback_hook(hook);
+ }
+
+ /// Register a callback function to be invoked whenever
+ /// a row is updated, inserted or deleted in a rowid table.
+ ///
+ /// The callback parameters are:
+ ///
+ /// - the type of database update (`SQLITE_INSERT`, `SQLITE_UPDATE` or
+ /// `SQLITE_DELETE`),
+ /// - the name of the database ("main", "temp", ...),
+ /// - the name of the table that is updated,
+ /// - the ROWID of the row that is updated.
+ #[inline]
+ pub fn update_hook<F>(&self, hook: Option<F>)
+ where
+ F: FnMut(Action, &str, &str, i64) + Send + 'static,
+ {
+ self.db.borrow_mut().update_hook(hook);
+ }
+
+ /// Register a query progress callback.
+ ///
+ /// The parameter `num_ops` is the approximate number of virtual machine
+ /// instructions that are evaluated between successive invocations of the
+ /// `handler`. If `num_ops` is less than one then the progress handler
+ /// is disabled.
+ ///
+ /// If the progress callback returns `true`, the operation is interrupted.
+ pub fn progress_handler<F>(&self, num_ops: c_int, handler: Option<F>)
+ where
+ F: FnMut() -> bool + Send + RefUnwindSafe + 'static,
+ {
+ self.db.borrow_mut().progress_handler(num_ops, handler);
+ }
+
+ /// Register an authorizer callback that's invoked
+ /// as a statement is being prepared.
+ #[inline]
+ pub fn authorizer<'c, F>(&self, hook: Option<F>)
+ where
+ F: for<'r> FnMut(AuthContext<'r>) -> Authorization + Send + RefUnwindSafe + 'static,
+ {
+ self.db.borrow_mut().authorizer(hook);
+ }
+}
+
+impl InnerConnection {
+ #[inline]
+ pub fn remove_hooks(&mut self) {
+ self.update_hook(None::<fn(Action, &str, &str, i64)>);
+ self.commit_hook(None::<fn() -> bool>);
+ self.rollback_hook(None::<fn()>);
+ self.progress_handler(0, None::<fn() -> bool>);
+ self.authorizer(None::<fn(AuthContext<'_>) -> Authorization>);
+ }
+
+ fn commit_hook<F>(&mut self, hook: Option<F>)
+ where
+ F: FnMut() -> bool + Send + 'static,
+ {
+ unsafe extern "C" fn call_boxed_closure<F>(p_arg: *mut c_void) -> c_int
+ where
+ F: FnMut() -> bool,
+ {
+ let r = catch_unwind(|| {
+ let boxed_hook: *mut F = p_arg.cast::<F>();
+ (*boxed_hook)()
+ });
+ c_int::from(r.unwrap_or_default())
+ }
+
+ // unlike `sqlite3_create_function_v2`, we cannot specify a `xDestroy` with
+ // `sqlite3_commit_hook`. so we keep the `xDestroy` function in
+ // `InnerConnection.free_boxed_hook`.
+ let free_commit_hook = if hook.is_some() {
+ Some(free_boxed_hook::<F> as unsafe fn(*mut c_void))
+ } else {
+ None
+ };
+
+ let previous_hook = match hook {
+ Some(hook) => {
+ let boxed_hook: *mut F = Box::into_raw(Box::new(hook));
+ unsafe {
+ ffi::sqlite3_commit_hook(
+ self.db(),
+ Some(call_boxed_closure::<F>),
+ boxed_hook.cast(),
+ )
+ }
+ }
+ _ => unsafe { ffi::sqlite3_commit_hook(self.db(), None, ptr::null_mut()) },
+ };
+ if !previous_hook.is_null() {
+ if let Some(free_boxed_hook) = self.free_commit_hook {
+ unsafe { free_boxed_hook(previous_hook) };
+ }
+ }
+ self.free_commit_hook = free_commit_hook;
+ }
+
+ fn rollback_hook<F>(&mut self, hook: Option<F>)
+ where
+ F: FnMut() + Send + 'static,
+ {
+ unsafe extern "C" fn call_boxed_closure<F>(p_arg: *mut c_void)
+ where
+ F: FnMut(),
+ {
+ drop(catch_unwind(|| {
+ let boxed_hook: *mut F = p_arg.cast::<F>();
+ (*boxed_hook)();
+ }));
+ }
+
+ let free_rollback_hook = if hook.is_some() {
+ Some(free_boxed_hook::<F> as unsafe fn(*mut c_void))
+ } else {
+ None
+ };
+
+ let previous_hook = match hook {
+ Some(hook) => {
+ let boxed_hook: *mut F = Box::into_raw(Box::new(hook));
+ unsafe {
+ ffi::sqlite3_rollback_hook(
+ self.db(),
+ Some(call_boxed_closure::<F>),
+ boxed_hook.cast(),
+ )
+ }
+ }
+ _ => unsafe { ffi::sqlite3_rollback_hook(self.db(), None, ptr::null_mut()) },
+ };
+ if !previous_hook.is_null() {
+ if let Some(free_boxed_hook) = self.free_rollback_hook {
+ unsafe { free_boxed_hook(previous_hook) };
+ }
+ }
+ self.free_rollback_hook = free_rollback_hook;
+ }
+
+ fn update_hook<F>(&mut self, hook: Option<F>)
+ where
+ F: FnMut(Action, &str, &str, i64) + Send + 'static,
+ {
+ unsafe extern "C" fn call_boxed_closure<F>(
+ p_arg: *mut c_void,
+ action_code: c_int,
+ p_db_name: *const c_char,
+ p_table_name: *const c_char,
+ row_id: i64,
+ ) where
+ F: FnMut(Action, &str, &str, i64),
+ {
+ let action = Action::from(action_code);
+ drop(catch_unwind(|| {
+ let boxed_hook: *mut F = p_arg.cast::<F>();
+ (*boxed_hook)(
+ action,
+ expect_utf8(p_db_name, "database name"),
+ expect_utf8(p_table_name, "table name"),
+ row_id,
+ );
+ }));
+ }
+
+ let free_update_hook = if hook.is_some() {
+ Some(free_boxed_hook::<F> as unsafe fn(*mut c_void))
+ } else {
+ None
+ };
+
+ let previous_hook = match hook {
+ Some(hook) => {
+ let boxed_hook: *mut F = Box::into_raw(Box::new(hook));
+ unsafe {
+ ffi::sqlite3_update_hook(
+ self.db(),
+ Some(call_boxed_closure::<F>),
+ boxed_hook.cast(),
+ )
+ }
+ }
+ _ => unsafe { ffi::sqlite3_update_hook(self.db(), None, ptr::null_mut()) },
+ };
+ if !previous_hook.is_null() {
+ if let Some(free_boxed_hook) = self.free_update_hook {
+ unsafe { free_boxed_hook(previous_hook) };
+ }
+ }
+ self.free_update_hook = free_update_hook;
+ }
+
+ fn progress_handler<F>(&mut self, num_ops: c_int, handler: Option<F>)
+ where
+ F: FnMut() -> bool + Send + RefUnwindSafe + 'static,
+ {
+ unsafe extern "C" fn call_boxed_closure<F>(p_arg: *mut c_void) -> c_int
+ where
+ F: FnMut() -> bool,
+ {
+ let r = catch_unwind(|| {
+ let boxed_handler: *mut F = p_arg.cast::<F>();
+ (*boxed_handler)()
+ });
+ c_int::from(r.unwrap_or_default())
+ }
+
+ if let Some(handler) = handler {
+ let boxed_handler = Box::new(handler);
+ unsafe {
+ ffi::sqlite3_progress_handler(
+ self.db(),
+ num_ops,
+ Some(call_boxed_closure::<F>),
+ &*boxed_handler as *const F as *mut _,
+ );
+ }
+ self.progress_handler = Some(boxed_handler);
+ } else {
+ unsafe { ffi::sqlite3_progress_handler(self.db(), num_ops, None, ptr::null_mut()) }
+ self.progress_handler = None;
+ };
+ }
+
+ fn authorizer<'c, F>(&'c mut self, authorizer: Option<F>)
+ where
+ F: for<'r> FnMut(AuthContext<'r>) -> Authorization + Send + RefUnwindSafe + 'static,
+ {
+ unsafe extern "C" fn call_boxed_closure<'c, F>(
+ p_arg: *mut c_void,
+ action_code: c_int,
+ param1: *const c_char,
+ param2: *const c_char,
+ db_name: *const c_char,
+ trigger_or_view_name: *const c_char,
+ ) -> c_int
+ where
+ F: FnMut(AuthContext<'c>) -> Authorization + Send + 'static,
+ {
+ catch_unwind(|| {
+ let action = AuthAction::from_raw(
+ action_code,
+ expect_optional_utf8(param1, "authorizer param 1"),
+ expect_optional_utf8(param2, "authorizer param 2"),
+ );
+ let auth_ctx = AuthContext {
+ action,
+ database_name: expect_optional_utf8(db_name, "database name"),
+ accessor: expect_optional_utf8(
+ trigger_or_view_name,
+ "accessor (inner-most trigger or view)",
+ ),
+ };
+ let boxed_hook: *mut F = p_arg.cast::<F>();
+ (*boxed_hook)(auth_ctx)
+ })
+ .map_or_else(|_| ffi::SQLITE_ERROR, Authorization::into_raw)
+ }
+
+ let callback_fn = authorizer
+ .as_ref()
+ .map(|_| call_boxed_closure::<'c, F> as unsafe extern "C" fn(_, _, _, _, _, _) -> _);
+ let boxed_authorizer = authorizer.map(Box::new);
+
+ match unsafe {
+ ffi::sqlite3_set_authorizer(
+ self.db(),
+ callback_fn,
+ boxed_authorizer
+ .as_ref()
+ .map_or_else(ptr::null_mut, |f| &**f as *const F as *mut _),
+ )
+ } {
+ ffi::SQLITE_OK => {
+ self.authorizer = boxed_authorizer.map(|ba| ba as _);
+ }
+ err_code => {
+ // The only error that `sqlite3_set_authorizer` returns is `SQLITE_MISUSE`
+ // when compiled with `ENABLE_API_ARMOR` and the db pointer is invalid.
+ // This library does not allow constructing a null db ptr, so if this branch
+ // is hit, something very bad has happened. Panicking instead of returning
+ // `Result` keeps this hook's API consistent with the others.
+ panic!("unexpectedly failed to set_authorizer: {}", unsafe {
+ crate::error::error_from_handle(self.db(), err_code)
+ });
+ }
+ }
+ }
+}
+
+unsafe fn free_boxed_hook<F>(p: *mut c_void) {
+ drop(Box::from_raw(p.cast::<F>()));
+}
+
+unsafe fn expect_utf8<'a>(p_str: *const c_char, description: &'static str) -> &'a str {
+ expect_optional_utf8(p_str, description)
+ .unwrap_or_else(|| panic!("received empty {description}"))
+}
+
+unsafe fn expect_optional_utf8<'a>(
+ p_str: *const c_char,
+ description: &'static str,
+) -> Option<&'a str> {
+ if p_str.is_null() {
+ return None;
+ }
+ std::str::from_utf8(std::ffi::CStr::from_ptr(p_str).to_bytes())
+ .unwrap_or_else(|_| panic!("received non-utf8 string as {description}"))
+ .into()
+}
+
+#[cfg(test)]
+mod test {
+ use super::Action;
+ use crate::{Connection, Result};
+ use std::sync::atomic::{AtomicBool, Ordering};
+
+ #[test]
+ fn test_commit_hook() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ static CALLED: AtomicBool = AtomicBool::new(false);
+ db.commit_hook(Some(|| {
+ CALLED.store(true, Ordering::Relaxed);
+ false
+ }));
+ db.execute_batch("BEGIN; CREATE TABLE foo (t TEXT); COMMIT;")?;
+ assert!(CALLED.load(Ordering::Relaxed));
+ Ok(())
+ }
+
+ #[test]
+ fn test_fn_commit_hook() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ fn hook() -> bool {
+ true
+ }
+
+ db.commit_hook(Some(hook));
+ db.execute_batch("BEGIN; CREATE TABLE foo (t TEXT); COMMIT;")
+ .unwrap_err();
+ Ok(())
+ }
+
+ #[test]
+ fn test_rollback_hook() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ static CALLED: AtomicBool = AtomicBool::new(false);
+ db.rollback_hook(Some(|| {
+ CALLED.store(true, Ordering::Relaxed);
+ }));
+ db.execute_batch("BEGIN; CREATE TABLE foo (t TEXT); ROLLBACK;")?;
+ assert!(CALLED.load(Ordering::Relaxed));
+ Ok(())
+ }
+
+ #[test]
+ fn test_update_hook() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ static CALLED: AtomicBool = AtomicBool::new(false);
+ db.update_hook(Some(|action, db: &str, tbl: &str, row_id| {
+ assert_eq!(Action::SQLITE_INSERT, action);
+ assert_eq!("main", db);
+ assert_eq!("foo", tbl);
+ assert_eq!(1, row_id);
+ CALLED.store(true, Ordering::Relaxed);
+ }));
+ db.execute_batch("CREATE TABLE foo (t TEXT)")?;
+ db.execute_batch("INSERT INTO foo VALUES ('lisa')")?;
+ assert!(CALLED.load(Ordering::Relaxed));
+ Ok(())
+ }
+
+ #[test]
+ fn test_progress_handler() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ static CALLED: AtomicBool = AtomicBool::new(false);
+ db.progress_handler(
+ 1,
+ Some(|| {
+ CALLED.store(true, Ordering::Relaxed);
+ false
+ }),
+ );
+ db.execute_batch("BEGIN; CREATE TABLE foo (t TEXT); COMMIT;")?;
+ assert!(CALLED.load(Ordering::Relaxed));
+ Ok(())
+ }
+
+ #[test]
+ fn test_progress_handler_interrupt() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ fn handler() -> bool {
+ true
+ }
+
+ db.progress_handler(1, Some(handler));
+ db.execute_batch("BEGIN; CREATE TABLE foo (t TEXT); COMMIT;")
+ .unwrap_err();
+ Ok(())
+ }
+
+ #[test]
+ fn test_authorizer() -> Result<()> {
+ use super::{AuthAction, AuthContext, Authorization};
+
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo (public TEXT, private TEXT)")
+ .unwrap();
+
+ let authorizer = move |ctx: AuthContext<'_>| match ctx.action {
+ AuthAction::Read {
+ column_name: "private",
+ ..
+ } => Authorization::Ignore,
+ AuthAction::DropTable { .. } => Authorization::Deny,
+ AuthAction::Pragma { .. } => panic!("shouldn't be called"),
+ _ => Authorization::Allow,
+ };
+
+ db.authorizer(Some(authorizer));
+ db.execute_batch(
+ "BEGIN TRANSACTION; INSERT INTO foo VALUES ('pub txt', 'priv txt'); COMMIT;",
+ )
+ .unwrap();
+ db.query_row_and_then("SELECT * FROM foo", [], |row| -> Result<()> {
+ assert_eq!(row.get::<_, String>("public")?, "pub txt");
+ assert!(row.get::<_, Option<String>>("private")?.is_none());
+ Ok(())
+ })
+ .unwrap();
+ db.execute_batch("DROP TABLE foo").unwrap_err();
+
+ db.authorizer(None::<fn(AuthContext<'_>) -> Authorization>);
+ db.execute_batch("PRAGMA user_version=1").unwrap(); // Disallowed by first authorizer, but it's now removed.
+
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/inner_connection.rs b/third_party/rust/rusqlite/src/inner_connection.rs
new file mode 100644
index 0000000000..b7c930fdea
--- /dev/null
+++ b/third_party/rust/rusqlite/src/inner_connection.rs
@@ -0,0 +1,462 @@
+use std::ffi::CStr;
+use std::os::raw::{c_char, c_int};
+#[cfg(feature = "load_extension")]
+use std::path::Path;
+use std::ptr;
+use std::str;
+use std::sync::atomic::AtomicBool;
+use std::sync::{Arc, Mutex};
+
+use super::ffi;
+use super::str_for_sqlite;
+use super::{Connection, InterruptHandle, OpenFlags, PrepFlags, Result};
+use crate::error::{error_from_handle, error_from_sqlite_code, error_with_offset, Error};
+use crate::raw_statement::RawStatement;
+use crate::statement::Statement;
+use crate::version::version_number;
+
+pub struct InnerConnection {
+ pub db: *mut ffi::sqlite3,
+ // It's unsafe to call `sqlite3_close` while another thread is performing
+ // a `sqlite3_interrupt`, and vice versa, so we take this mutex during
+ // those functions. This protects a copy of the `db` pointer (which is
+ // cleared on closing), however the main copy, `db`, is unprotected.
+ // Otherwise, a long running query would prevent calling interrupt, as
+ // interrupt would only acquire the lock after the query's completion.
+ interrupt_lock: Arc<Mutex<*mut ffi::sqlite3>>,
+ #[cfg(feature = "hooks")]
+ pub free_commit_hook: Option<unsafe fn(*mut std::os::raw::c_void)>,
+ #[cfg(feature = "hooks")]
+ pub free_rollback_hook: Option<unsafe fn(*mut std::os::raw::c_void)>,
+ #[cfg(feature = "hooks")]
+ pub free_update_hook: Option<unsafe fn(*mut std::os::raw::c_void)>,
+ #[cfg(feature = "hooks")]
+ pub progress_handler: Option<Box<dyn FnMut() -> bool + Send>>,
+ #[cfg(feature = "hooks")]
+ pub authorizer: Option<crate::hooks::BoxedAuthorizer>,
+ owned: bool,
+}
+
+unsafe impl Send for InnerConnection {}
+
+impl InnerConnection {
+ #[allow(clippy::mutex_atomic, clippy::arc_with_non_send_sync)] // See unsafe impl Send / Sync for InterruptHandle
+ #[inline]
+ pub unsafe fn new(db: *mut ffi::sqlite3, owned: bool) -> InnerConnection {
+ InnerConnection {
+ db,
+ interrupt_lock: Arc::new(Mutex::new(db)),
+ #[cfg(feature = "hooks")]
+ free_commit_hook: None,
+ #[cfg(feature = "hooks")]
+ free_rollback_hook: None,
+ #[cfg(feature = "hooks")]
+ free_update_hook: None,
+ #[cfg(feature = "hooks")]
+ progress_handler: None,
+ #[cfg(feature = "hooks")]
+ authorizer: None,
+ owned,
+ }
+ }
+
+ pub fn open_with_flags(
+ c_path: &CStr,
+ flags: OpenFlags,
+ vfs: Option<&CStr>,
+ ) -> Result<InnerConnection> {
+ ensure_safe_sqlite_threading_mode()?;
+
+ // Replicate the check for sane open flags from SQLite, because the check in
+ // SQLite itself wasn't added until version 3.7.3.
+ debug_assert_eq!(1 << OpenFlags::SQLITE_OPEN_READ_ONLY.bits(), 0x02);
+ debug_assert_eq!(1 << OpenFlags::SQLITE_OPEN_READ_WRITE.bits(), 0x04);
+ debug_assert_eq!(
+ 1 << (OpenFlags::SQLITE_OPEN_READ_WRITE | OpenFlags::SQLITE_OPEN_CREATE).bits(),
+ 0x40
+ );
+ if (1 << (flags.bits() & 0x7)) & 0x46 == 0 {
+ return Err(Error::SqliteFailure(
+ ffi::Error::new(ffi::SQLITE_MISUSE),
+ None,
+ ));
+ }
+
+ let z_vfs = match vfs {
+ Some(c_vfs) => c_vfs.as_ptr(),
+ None => ptr::null(),
+ };
+
+ unsafe {
+ let mut db: *mut ffi::sqlite3 = ptr::null_mut();
+ let r = ffi::sqlite3_open_v2(c_path.as_ptr(), &mut db, flags.bits(), z_vfs);
+ if r != ffi::SQLITE_OK {
+ let e = if db.is_null() {
+ error_from_sqlite_code(r, Some(c_path.to_string_lossy().to_string()))
+ } else {
+ let mut e = error_from_handle(db, r);
+ if let Error::SqliteFailure(
+ ffi::Error {
+ code: ffi::ErrorCode::CannotOpen,
+ ..
+ },
+ Some(msg),
+ ) = e
+ {
+ e = Error::SqliteFailure(
+ ffi::Error::new(r),
+ Some(format!("{msg}: {}", c_path.to_string_lossy())),
+ );
+ }
+ ffi::sqlite3_close(db);
+ e
+ };
+
+ return Err(e);
+ }
+
+ // attempt to turn on extended results code; don't fail if we can't.
+ ffi::sqlite3_extended_result_codes(db, 1);
+
+ let r = ffi::sqlite3_busy_timeout(db, 5000);
+ if r != ffi::SQLITE_OK {
+ let e = error_from_handle(db, r);
+ ffi::sqlite3_close(db);
+ return Err(e);
+ }
+
+ Ok(InnerConnection::new(db, true))
+ }
+ }
+
+ #[inline]
+ pub fn db(&self) -> *mut ffi::sqlite3 {
+ self.db
+ }
+
+ #[inline]
+ pub fn decode_result(&self, code: c_int) -> Result<()> {
+ unsafe { InnerConnection::decode_result_raw(self.db(), code) }
+ }
+
+ #[inline]
+ unsafe fn decode_result_raw(db: *mut ffi::sqlite3, code: c_int) -> Result<()> {
+ if code == ffi::SQLITE_OK {
+ Ok(())
+ } else {
+ Err(error_from_handle(db, code))
+ }
+ }
+
+ #[allow(clippy::mutex_atomic)]
+ pub fn close(&mut self) -> Result<()> {
+ if self.db.is_null() {
+ return Ok(());
+ }
+ self.remove_hooks();
+ let mut shared_handle = self.interrupt_lock.lock().unwrap();
+ assert!(
+ !shared_handle.is_null(),
+ "Bug: Somehow interrupt_lock was cleared before the DB was closed"
+ );
+ if !self.owned {
+ self.db = ptr::null_mut();
+ return Ok(());
+ }
+ unsafe {
+ let r = ffi::sqlite3_close(self.db);
+ // Need to use _raw because _guard has a reference out, and
+ // decode_result takes &mut self.
+ let r = InnerConnection::decode_result_raw(self.db, r);
+ if r.is_ok() {
+ *shared_handle = ptr::null_mut();
+ self.db = ptr::null_mut();
+ }
+ r
+ }
+ }
+
+ #[inline]
+ pub fn get_interrupt_handle(&self) -> InterruptHandle {
+ InterruptHandle {
+ db_lock: Arc::clone(&self.interrupt_lock),
+ }
+ }
+
+ #[inline]
+ #[cfg(feature = "load_extension")]
+ pub unsafe fn enable_load_extension(&mut self, onoff: c_int) -> Result<()> {
+ let r = ffi::sqlite3_enable_load_extension(self.db, onoff);
+ self.decode_result(r)
+ }
+
+ #[cfg(feature = "load_extension")]
+ pub unsafe fn load_extension(
+ &self,
+ dylib_path: &Path,
+ entry_point: Option<&str>,
+ ) -> Result<()> {
+ let dylib_str = super::path_to_cstring(dylib_path)?;
+ let mut errmsg: *mut c_char = ptr::null_mut();
+ let r = if let Some(entry_point) = entry_point {
+ let c_entry = crate::str_to_cstring(entry_point)?;
+ ffi::sqlite3_load_extension(self.db, dylib_str.as_ptr(), c_entry.as_ptr(), &mut errmsg)
+ } else {
+ ffi::sqlite3_load_extension(self.db, dylib_str.as_ptr(), ptr::null(), &mut errmsg)
+ };
+ if r == ffi::SQLITE_OK {
+ Ok(())
+ } else {
+ let message = super::errmsg_to_string(errmsg);
+ ffi::sqlite3_free(errmsg.cast::<std::os::raw::c_void>());
+ Err(error_from_sqlite_code(r, Some(message)))
+ }
+ }
+
+ #[inline]
+ pub fn last_insert_rowid(&self) -> i64 {
+ unsafe { ffi::sqlite3_last_insert_rowid(self.db()) }
+ }
+
+ pub fn prepare<'a>(
+ &mut self,
+ conn: &'a Connection,
+ sql: &str,
+ flags: PrepFlags,
+ ) -> Result<Statement<'a>> {
+ let mut c_stmt: *mut ffi::sqlite3_stmt = ptr::null_mut();
+ let (c_sql, len, _) = str_for_sqlite(sql.as_bytes())?;
+ let mut c_tail: *const c_char = ptr::null();
+ // TODO sqlite3_prepare_v3 (https://sqlite.org/c3ref/c_prepare_normalize.html) // 3.20.0, #728
+ #[cfg(not(feature = "unlock_notify"))]
+ let r = unsafe { self.prepare_(c_sql, len, flags, &mut c_stmt, &mut c_tail) };
+ #[cfg(feature = "unlock_notify")]
+ let r = unsafe {
+ use crate::unlock_notify;
+ let mut rc;
+ loop {
+ rc = self.prepare_(c_sql, len, flags, &mut c_stmt, &mut c_tail);
+ if !unlock_notify::is_locked(self.db, rc) {
+ break;
+ }
+ rc = unlock_notify::wait_for_unlock_notify(self.db);
+ if rc != ffi::SQLITE_OK {
+ break;
+ }
+ }
+ rc
+ };
+ // If there is an error, *ppStmt is set to NULL.
+ if r != ffi::SQLITE_OK {
+ return Err(unsafe { error_with_offset(self.db, r, sql) });
+ }
+ // If the input text contains no SQL (if the input is an empty string or a
+ // comment) then *ppStmt is set to NULL.
+ let tail = if c_tail.is_null() {
+ 0
+ } else {
+ let n = (c_tail as isize) - (c_sql as isize);
+ if n <= 0 || n >= len as isize {
+ 0
+ } else {
+ n as usize
+ }
+ };
+ Ok(Statement::new(conn, unsafe {
+ RawStatement::new(c_stmt, tail)
+ }))
+ }
+
+ #[inline]
+ #[cfg(not(feature = "modern_sqlite"))]
+ unsafe fn prepare_(
+ &self,
+ z_sql: *const c_char,
+ n_byte: c_int,
+ _: PrepFlags,
+ pp_stmt: *mut *mut ffi::sqlite3_stmt,
+ pz_tail: *mut *const c_char,
+ ) -> c_int {
+ ffi::sqlite3_prepare_v2(self.db(), z_sql, n_byte, pp_stmt, pz_tail)
+ }
+
+ #[inline]
+ #[cfg(feature = "modern_sqlite")]
+ unsafe fn prepare_(
+ &self,
+ z_sql: *const c_char,
+ n_byte: c_int,
+ flags: PrepFlags,
+ pp_stmt: *mut *mut ffi::sqlite3_stmt,
+ pz_tail: *mut *const c_char,
+ ) -> c_int {
+ ffi::sqlite3_prepare_v3(self.db(), z_sql, n_byte, flags.bits(), pp_stmt, pz_tail)
+ }
+
+ #[inline]
+ pub fn changes(&self) -> u64 {
+ #[cfg(not(feature = "modern_sqlite"))]
+ unsafe {
+ ffi::sqlite3_changes(self.db()) as u64
+ }
+ #[cfg(feature = "modern_sqlite")] // 3.37.0
+ unsafe {
+ ffi::sqlite3_changes64(self.db()) as u64
+ }
+ }
+
+ #[inline]
+ pub fn is_autocommit(&self) -> bool {
+ unsafe { ffi::sqlite3_get_autocommit(self.db()) != 0 }
+ }
+
+ pub fn is_busy(&self) -> bool {
+ let db = self.db();
+ unsafe {
+ let mut stmt = ffi::sqlite3_next_stmt(db, ptr::null_mut());
+ while !stmt.is_null() {
+ if ffi::sqlite3_stmt_busy(stmt) != 0 {
+ return true;
+ }
+ stmt = ffi::sqlite3_next_stmt(db, stmt);
+ }
+ }
+ false
+ }
+
+ pub fn cache_flush(&mut self) -> Result<()> {
+ crate::error::check(unsafe { ffi::sqlite3_db_cacheflush(self.db()) })
+ }
+
+ #[cfg(not(feature = "hooks"))]
+ #[inline]
+ fn remove_hooks(&mut self) {}
+
+ pub fn db_readonly(&self, db_name: super::DatabaseName<'_>) -> Result<bool> {
+ let name = db_name.as_cstring()?;
+ let r = unsafe { ffi::sqlite3_db_readonly(self.db, name.as_ptr()) };
+ match r {
+ 0 => Ok(false),
+ 1 => Ok(true),
+ -1 => Err(Error::SqliteFailure(
+ ffi::Error::new(ffi::SQLITE_MISUSE),
+ Some(format!("{db_name:?} is not the name of a database")),
+ )),
+ _ => Err(error_from_sqlite_code(
+ r,
+ Some("Unexpected result".to_owned()),
+ )),
+ }
+ }
+
+ #[cfg(feature = "modern_sqlite")] // 3.37.0
+ pub fn txn_state(
+ &self,
+ db_name: Option<super::DatabaseName<'_>>,
+ ) -> Result<super::transaction::TransactionState> {
+ let r = if let Some(ref name) = db_name {
+ let name = name.as_cstring()?;
+ unsafe { ffi::sqlite3_txn_state(self.db, name.as_ptr()) }
+ } else {
+ unsafe { ffi::sqlite3_txn_state(self.db, ptr::null()) }
+ };
+ match r {
+ 0 => Ok(super::transaction::TransactionState::None),
+ 1 => Ok(super::transaction::TransactionState::Read),
+ 2 => Ok(super::transaction::TransactionState::Write),
+ -1 => Err(Error::SqliteFailure(
+ ffi::Error::new(ffi::SQLITE_MISUSE),
+ Some(format!("{db_name:?} is not the name of a valid schema")),
+ )),
+ _ => Err(error_from_sqlite_code(
+ r,
+ Some("Unexpected result".to_owned()),
+ )),
+ }
+ }
+
+ #[inline]
+ #[cfg(feature = "release_memory")]
+ pub fn release_memory(&self) -> Result<()> {
+ self.decode_result(unsafe { ffi::sqlite3_db_release_memory(self.db) })
+ }
+}
+
+impl Drop for InnerConnection {
+ #[allow(unused_must_use)]
+ #[inline]
+ fn drop(&mut self) {
+ self.close();
+ }
+}
+
+#[cfg(not(any(target_arch = "wasm32", feature = "loadable_extension")))]
+static SQLITE_INIT: std::sync::Once = std::sync::Once::new();
+
+pub static BYPASS_SQLITE_INIT: AtomicBool = AtomicBool::new(false);
+
+// threading mode checks are not necessary (and do not work) on target
+// platforms that do not have threading (such as webassembly)
+#[cfg(target_arch = "wasm32")]
+fn ensure_safe_sqlite_threading_mode() -> Result<()> {
+ Ok(())
+}
+
+#[cfg(not(any(target_arch = "wasm32")))]
+fn ensure_safe_sqlite_threading_mode() -> Result<()> {
+ // Ensure SQLite was compiled in threadsafe mode.
+ if unsafe { ffi::sqlite3_threadsafe() == 0 } {
+ return Err(Error::SqliteSingleThreadedMode);
+ }
+
+ // Now we know SQLite is _capable_ of being in Multi-thread of Serialized mode,
+ // but it's possible someone configured it to be in Single-thread mode
+ // before calling into us. That would mean we're exposing an unsafe API via
+ // a safe one (in Rust terminology), which is no good. We have two options
+ // to protect against this, depending on the version of SQLite we're linked
+ // with:
+ //
+ // 1. If we're on 3.7.0 or later, we can ask SQLite for a mutex and check for
+ // the magic value 8. This isn't documented, but it's what SQLite
+ // returns for its mutex allocation function in Single-thread mode.
+ // 2. If we're prior to SQLite 3.7.0, AFAIK there's no way to check the
+ // threading mode. The check we perform for >= 3.7.0 will segfault.
+ // Instead, we insist on being able to call sqlite3_config and
+ // sqlite3_initialize ourself, ensuring we know the threading
+ // mode. This will fail if someone else has already initialized SQLite
+ // even if they initialized it safely. That's not ideal either, which is
+ // why we expose bypass_sqlite_initialization above.
+ if version_number() >= 3_007_000 {
+ const SQLITE_SINGLETHREADED_MUTEX_MAGIC: usize = 8;
+ let is_singlethreaded = unsafe {
+ let mutex_ptr = ffi::sqlite3_mutex_alloc(0);
+ let is_singlethreaded = mutex_ptr as usize == SQLITE_SINGLETHREADED_MUTEX_MAGIC;
+ ffi::sqlite3_mutex_free(mutex_ptr);
+ is_singlethreaded
+ };
+ if is_singlethreaded {
+ Err(Error::SqliteSingleThreadedMode)
+ } else {
+ Ok(())
+ }
+ } else {
+ #[cfg(not(feature = "loadable_extension"))]
+ SQLITE_INIT.call_once(|| {
+ use std::sync::atomic::Ordering;
+ if BYPASS_SQLITE_INIT.load(Ordering::Relaxed) {
+ return;
+ }
+
+ unsafe {
+ assert!(ffi::sqlite3_config(ffi::SQLITE_CONFIG_MULTITHREAD) == ffi::SQLITE_OK && ffi::sqlite3_initialize() == ffi::SQLITE_OK,
+ "Could not ensure safe initialization of SQLite.\n\
+ To fix this, either:\n\
+ * Upgrade SQLite to at least version 3.7.0\n\
+ * Ensure that SQLite has been initialized in Multi-thread or Serialized mode and call\n\
+ rusqlite::bypass_sqlite_initialization() prior to your first connection attempt."
+ );
+ }
+ });
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/lib.rs b/third_party/rust/rusqlite/src/lib.rs
new file mode 100644
index 0000000000..cb4b14b9dc
--- /dev/null
+++ b/third_party/rust/rusqlite/src/lib.rs
@@ -0,0 +1,2200 @@
+//! Rusqlite is an ergonomic wrapper for using SQLite from Rust.
+//!
+//! Historically, the API was based on the one from
+//! [`rust-postgres`](https://github.com/sfackler/rust-postgres). However, the
+//! two have diverged in many ways, and no compatibility between the two is
+//! intended.
+//!
+//! ```rust
+//! use rusqlite::{params, Connection, Result};
+//!
+//! #[derive(Debug)]
+//! struct Person {
+//! id: i32,
+//! name: String,
+//! data: Option<Vec<u8>>,
+//! }
+//!
+//! fn main() -> Result<()> {
+//! let conn = Connection::open_in_memory()?;
+//!
+//! conn.execute(
+//! "CREATE TABLE person (
+//! id INTEGER PRIMARY KEY,
+//! name TEXT NOT NULL,
+//! data BLOB
+//! )",
+//! (), // empty list of parameters.
+//! )?;
+//! let me = Person {
+//! id: 0,
+//! name: "Steven".to_string(),
+//! data: None,
+//! };
+//! conn.execute(
+//! "INSERT INTO person (name, data) VALUES (?1, ?2)",
+//! (&me.name, &me.data),
+//! )?;
+//!
+//! let mut stmt = conn.prepare("SELECT id, name, data FROM person")?;
+//! let person_iter = stmt.query_map([], |row| {
+//! Ok(Person {
+//! id: row.get(0)?,
+//! name: row.get(1)?,
+//! data: row.get(2)?,
+//! })
+//! })?;
+//!
+//! for person in person_iter {
+//! println!("Found person {:?}", person.unwrap());
+//! }
+//! Ok(())
+//! }
+//! ```
+#![warn(missing_docs)]
+#![cfg_attr(docsrs, feature(doc_cfg))]
+
+pub use libsqlite3_sys as ffi;
+
+use std::cell::RefCell;
+use std::default::Default;
+use std::ffi::{CStr, CString};
+use std::fmt;
+use std::os::raw::{c_char, c_int};
+
+use std::path::Path;
+use std::result;
+use std::str;
+use std::sync::atomic::Ordering;
+use std::sync::{Arc, Mutex};
+
+use crate::cache::StatementCache;
+use crate::inner_connection::{InnerConnection, BYPASS_SQLITE_INIT};
+use crate::raw_statement::RawStatement;
+use crate::types::ValueRef;
+
+pub use crate::cache::CachedStatement;
+#[cfg(feature = "column_decltype")]
+pub use crate::column::Column;
+pub use crate::error::{to_sqlite_error, Error};
+pub use crate::ffi::ErrorCode;
+#[cfg(feature = "load_extension")]
+pub use crate::load_extension_guard::LoadExtensionGuard;
+pub use crate::params::{params_from_iter, Params, ParamsFromIter};
+pub use crate::row::{AndThenRows, Map, MappedRows, Row, RowIndex, Rows};
+pub use crate::statement::{Statement, StatementStatus};
+#[cfg(feature = "modern_sqlite")]
+pub use crate::transaction::TransactionState;
+pub use crate::transaction::{DropBehavior, Savepoint, Transaction, TransactionBehavior};
+pub use crate::types::ToSql;
+pub use crate::version::*;
+#[cfg(feature = "rusqlite-macros")]
+#[doc(hidden)]
+pub use rusqlite_macros::__bind;
+
+mod error;
+
+#[cfg(feature = "backup")]
+#[cfg_attr(docsrs, doc(cfg(feature = "backup")))]
+pub mod backup;
+#[cfg(feature = "blob")]
+#[cfg_attr(docsrs, doc(cfg(feature = "blob")))]
+pub mod blob;
+mod busy;
+mod cache;
+#[cfg(feature = "collation")]
+#[cfg_attr(docsrs, doc(cfg(feature = "collation")))]
+mod collation;
+mod column;
+pub mod config;
+#[cfg(any(feature = "functions", feature = "vtab"))]
+mod context;
+#[cfg(feature = "functions")]
+#[cfg_attr(docsrs, doc(cfg(feature = "functions")))]
+pub mod functions;
+#[cfg(feature = "hooks")]
+#[cfg_attr(docsrs, doc(cfg(feature = "hooks")))]
+pub mod hooks;
+mod inner_connection;
+#[cfg(feature = "limits")]
+#[cfg_attr(docsrs, doc(cfg(feature = "limits")))]
+pub mod limits;
+#[cfg(feature = "load_extension")]
+mod load_extension_guard;
+mod params;
+mod pragma;
+mod raw_statement;
+mod row;
+#[cfg(feature = "serialize")]
+#[cfg_attr(docsrs, doc(cfg(feature = "serialize")))]
+pub mod serialize;
+#[cfg(feature = "session")]
+#[cfg_attr(docsrs, doc(cfg(feature = "session")))]
+pub mod session;
+mod statement;
+#[cfg(feature = "trace")]
+#[cfg_attr(docsrs, doc(cfg(feature = "trace")))]
+pub mod trace;
+mod transaction;
+pub mod types;
+#[cfg(feature = "unlock_notify")]
+mod unlock_notify;
+mod version;
+#[cfg(feature = "vtab")]
+#[cfg_attr(docsrs, doc(cfg(feature = "vtab")))]
+pub mod vtab;
+
+pub(crate) mod util;
+pub(crate) use util::SmallCString;
+
+// Number of cached prepared statements we'll hold on to.
+const STATEMENT_CACHE_DEFAULT_CAPACITY: usize = 16;
+
+/// A macro making it more convenient to longer lists of
+/// parameters as a `&[&dyn ToSql]`.
+///
+/// # Example
+///
+/// ```rust,no_run
+/// # use rusqlite::{Result, Connection, params};
+///
+/// struct Person {
+/// name: String,
+/// age_in_years: u8,
+/// data: Option<Vec<u8>>,
+/// }
+///
+/// fn add_person(conn: &Connection, person: &Person) -> Result<()> {
+/// conn.execute(
+/// "INSERT INTO person(name, age_in_years, data) VALUES (?1, ?2, ?3)",
+/// params![person.name, person.age_in_years, person.data],
+/// )?;
+/// Ok(())
+/// }
+/// ```
+#[macro_export]
+macro_rules! params {
+ () => {
+ &[] as &[&dyn $crate::ToSql]
+ };
+ ($($param:expr),+ $(,)?) => {
+ &[$(&$param as &dyn $crate::ToSql),+] as &[&dyn $crate::ToSql]
+ };
+}
+
+/// A macro making it more convenient to pass lists of named parameters
+/// as a `&[(&str, &dyn ToSql)]`.
+///
+/// # Example
+///
+/// ```rust,no_run
+/// # use rusqlite::{Result, Connection, named_params};
+///
+/// struct Person {
+/// name: String,
+/// age_in_years: u8,
+/// data: Option<Vec<u8>>,
+/// }
+///
+/// fn add_person(conn: &Connection, person: &Person) -> Result<()> {
+/// conn.execute(
+/// "INSERT INTO person (name, age_in_years, data)
+/// VALUES (:name, :age, :data)",
+/// named_params! {
+/// ":name": person.name,
+/// ":age": person.age_in_years,
+/// ":data": person.data,
+/// },
+/// )?;
+/// Ok(())
+/// }
+/// ```
+#[macro_export]
+macro_rules! named_params {
+ () => {
+ &[] as &[(&str, &dyn $crate::ToSql)]
+ };
+ // Note: It's a lot more work to support this as part of the same macro as
+ // `params!`, unfortunately.
+ ($($param_name:literal: $param_val:expr),+ $(,)?) => {
+ &[$(($param_name, &$param_val as &dyn $crate::ToSql)),+] as &[(&str, &dyn $crate::ToSql)]
+ };
+}
+
+/// Captured identifiers in SQL
+///
+/// * only SQLite `$x` / `@x` / `:x` syntax works (Rust `&x` syntax does not
+/// work).
+/// * `$x.y` expression does not work.
+///
+/// # Example
+///
+/// ```rust, no_run
+/// # use rusqlite::{prepare_and_bind, Connection, Result, Statement};
+///
+/// fn misc(db: &Connection) -> Result<Statement> {
+/// let name = "Lisa";
+/// let age = 8;
+/// let smart = true;
+/// Ok(prepare_and_bind!(db, "SELECT $name, @age, :smart;"))
+/// }
+/// ```
+#[cfg(feature = "rusqlite-macros")]
+#[cfg_attr(docsrs, doc(cfg(feature = "rusqlite-macros")))]
+#[macro_export]
+macro_rules! prepare_and_bind {
+ ($conn:expr, $sql:literal) => {{
+ let mut stmt = $conn.prepare($sql)?;
+ $crate::__bind!(stmt $sql);
+ stmt
+ }};
+}
+
+/// Captured identifiers in SQL
+///
+/// * only SQLite `$x` / `@x` / `:x` syntax works (Rust `&x` syntax does not
+/// work).
+/// * `$x.y` expression does not work.
+#[cfg(feature = "rusqlite-macros")]
+#[cfg_attr(docsrs, doc(cfg(feature = "rusqlite-macros")))]
+#[macro_export]
+macro_rules! prepare_cached_and_bind {
+ ($conn:expr, $sql:literal) => {{
+ let mut stmt = $conn.prepare_cached($sql)?;
+ $crate::__bind!(stmt $sql);
+ stmt
+ }};
+}
+
+/// A typedef of the result returned by many methods.
+pub type Result<T, E = Error> = result::Result<T, E>;
+
+/// See the [method documentation](#tymethod.optional).
+pub trait OptionalExtension<T> {
+ /// Converts a `Result<T>` into a `Result<Option<T>>`.
+ ///
+ /// By default, Rusqlite treats 0 rows being returned from a query that is
+ /// expected to return 1 row as an error. This method will
+ /// handle that error, and give you back an `Option<T>` instead.
+ fn optional(self) -> Result<Option<T>>;
+}
+
+impl<T> OptionalExtension<T> for Result<T> {
+ fn optional(self) -> Result<Option<T>> {
+ match self {
+ Ok(value) => Ok(Some(value)),
+ Err(Error::QueryReturnedNoRows) => Ok(None),
+ Err(e) => Err(e),
+ }
+ }
+}
+
+unsafe fn errmsg_to_string(errmsg: *const c_char) -> String {
+ let c_slice = CStr::from_ptr(errmsg).to_bytes();
+ String::from_utf8_lossy(c_slice).into_owned()
+}
+
+fn str_to_cstring(s: &str) -> Result<SmallCString> {
+ Ok(SmallCString::new(s)?)
+}
+
+/// Returns `Ok((string ptr, len as c_int, SQLITE_STATIC | SQLITE_TRANSIENT))`
+/// normally.
+/// Returns error if the string is too large for sqlite.
+/// The `sqlite3_destructor_type` item is always `SQLITE_TRANSIENT` unless
+/// the string was empty (in which case it's `SQLITE_STATIC`, and the ptr is
+/// static).
+fn str_for_sqlite(s: &[u8]) -> Result<(*const c_char, c_int, ffi::sqlite3_destructor_type)> {
+ let len = len_as_c_int(s.len())?;
+ let (ptr, dtor_info) = if len != 0 {
+ (s.as_ptr().cast::<c_char>(), ffi::SQLITE_TRANSIENT())
+ } else {
+ // Return a pointer guaranteed to live forever
+ ("".as_ptr().cast::<c_char>(), ffi::SQLITE_STATIC())
+ };
+ Ok((ptr, len, dtor_info))
+}
+
+// Helper to cast to c_int safely, returning the correct error type if the cast
+// failed.
+fn len_as_c_int(len: usize) -> Result<c_int> {
+ if len >= (c_int::MAX as usize) {
+ Err(Error::SqliteFailure(
+ ffi::Error::new(ffi::SQLITE_TOOBIG),
+ None,
+ ))
+ } else {
+ Ok(len as c_int)
+ }
+}
+
+#[cfg(unix)]
+fn path_to_cstring(p: &Path) -> Result<CString> {
+ use std::os::unix::ffi::OsStrExt;
+ Ok(CString::new(p.as_os_str().as_bytes())?)
+}
+
+#[cfg(not(unix))]
+fn path_to_cstring(p: &Path) -> Result<CString> {
+ let s = p.to_str().ok_or_else(|| Error::InvalidPath(p.to_owned()))?;
+ Ok(CString::new(s)?)
+}
+
+/// Name for a database within a SQLite connection.
+#[derive(Copy, Clone, Debug)]
+pub enum DatabaseName<'a> {
+ /// The main database.
+ Main,
+
+ /// The temporary database (e.g., any "CREATE TEMPORARY TABLE" tables).
+ Temp,
+
+ /// A database that has been attached via "ATTACH DATABASE ...".
+ Attached(&'a str),
+}
+
+/// Shorthand for [`DatabaseName::Main`].
+pub const MAIN_DB: DatabaseName<'static> = DatabaseName::Main;
+
+/// Shorthand for [`DatabaseName::Temp`].
+pub const TEMP_DB: DatabaseName<'static> = DatabaseName::Temp;
+
+// Currently DatabaseName is only used by the backup and blob mods, so hide
+// this (private) impl to avoid dead code warnings.
+impl DatabaseName<'_> {
+ #[inline]
+ fn as_cstring(&self) -> Result<SmallCString> {
+ use self::DatabaseName::{Attached, Main, Temp};
+ match *self {
+ Main => str_to_cstring("main"),
+ Temp => str_to_cstring("temp"),
+ Attached(s) => str_to_cstring(s),
+ }
+ }
+}
+
+/// A connection to a SQLite database.
+pub struct Connection {
+ db: RefCell<InnerConnection>,
+ cache: StatementCache,
+}
+
+unsafe impl Send for Connection {}
+
+impl Drop for Connection {
+ #[inline]
+ fn drop(&mut self) {
+ self.flush_prepared_statement_cache();
+ }
+}
+
+impl Connection {
+ /// Open a new connection to a SQLite database. If a database does not exist
+ /// at the path, one is created.
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn open_my_db() -> Result<()> {
+ /// let path = "./my_db.db3";
+ /// let db = Connection::open(path)?;
+ /// // Use the database somehow...
+ /// println!("{}", db.is_autocommit());
+ /// Ok(())
+ /// }
+ /// ```
+ ///
+ /// # Flags
+ ///
+ /// `Connection::open(path)` is equivalent to using
+ /// [`Connection::open_with_flags`] with the default [`OpenFlags`]. That is,
+ /// it's equivalent to:
+ ///
+ /// ```ignore
+ /// Connection::open_with_flags(
+ /// path,
+ /// OpenFlags::SQLITE_OPEN_READ_WRITE
+ /// | OpenFlags::SQLITE_OPEN_CREATE
+ /// | OpenFlags::SQLITE_OPEN_URI
+ /// | OpenFlags::SQLITE_OPEN_NO_MUTEX,
+ /// )
+ /// ```
+ ///
+ /// These flags have the following effects:
+ ///
+ /// - Open the database for both reading or writing.
+ /// - Create the database if one does not exist at the path.
+ /// - Allow the filename to be interpreted as a URI (see <https://www.sqlite.org/uri.html#uri_filenames_in_sqlite>
+ /// for details).
+ /// - Disables the use of a per-connection mutex.
+ ///
+ /// Rusqlite enforces thread-safety at compile time, so additional
+ /// locking is not needed and provides no benefit. (See the
+ /// documentation on [`OpenFlags::SQLITE_OPEN_FULL_MUTEX`] for some
+ /// additional discussion about this).
+ ///
+ /// Most of these are also the default settings for the C API, although
+ /// technically the default locking behavior is controlled by the flags used
+ /// when compiling SQLite -- rather than let it vary, we choose `NO_MUTEX`
+ /// because it's a fairly clearly the best choice for users of this library.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if `path` cannot be converted to a C-compatible string
+ /// or if the underlying SQLite open call fails.
+ #[inline]
+ pub fn open<P: AsRef<Path>>(path: P) -> Result<Connection> {
+ let flags = OpenFlags::default();
+ Connection::open_with_flags(path, flags)
+ }
+
+ /// Open a new connection to an in-memory SQLite database.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite open call fails.
+ #[inline]
+ pub fn open_in_memory() -> Result<Connection> {
+ let flags = OpenFlags::default();
+ Connection::open_in_memory_with_flags(flags)
+ }
+
+ /// Open a new connection to a SQLite database.
+ ///
+ /// [Database Connection](http://www.sqlite.org/c3ref/open.html) for a description of valid
+ /// flag combinations.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if `path` cannot be converted to a C-compatible
+ /// string or if the underlying SQLite open call fails.
+ #[inline]
+ pub fn open_with_flags<P: AsRef<Path>>(path: P, flags: OpenFlags) -> Result<Connection> {
+ let c_path = path_to_cstring(path.as_ref())?;
+ InnerConnection::open_with_flags(&c_path, flags, None).map(|db| Connection {
+ db: RefCell::new(db),
+ cache: StatementCache::with_capacity(STATEMENT_CACHE_DEFAULT_CAPACITY),
+ })
+ }
+
+ /// Open a new connection to a SQLite database using the specific flags and
+ /// vfs name.
+ ///
+ /// [Database Connection](http://www.sqlite.org/c3ref/open.html) for a description of valid
+ /// flag combinations.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if either `path` or `vfs` cannot be converted to a
+ /// C-compatible string or if the underlying SQLite open call fails.
+ #[inline]
+ pub fn open_with_flags_and_vfs<P: AsRef<Path>>(
+ path: P,
+ flags: OpenFlags,
+ vfs: &str,
+ ) -> Result<Connection> {
+ let c_path = path_to_cstring(path.as_ref())?;
+ let c_vfs = str_to_cstring(vfs)?;
+ InnerConnection::open_with_flags(&c_path, flags, Some(&c_vfs)).map(|db| Connection {
+ db: RefCell::new(db),
+ cache: StatementCache::with_capacity(STATEMENT_CACHE_DEFAULT_CAPACITY),
+ })
+ }
+
+ /// Open a new connection to an in-memory SQLite database.
+ ///
+ /// [Database Connection](http://www.sqlite.org/c3ref/open.html) for a description of valid
+ /// flag combinations.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite open call fails.
+ #[inline]
+ pub fn open_in_memory_with_flags(flags: OpenFlags) -> Result<Connection> {
+ Connection::open_with_flags(":memory:", flags)
+ }
+
+ /// Open a new connection to an in-memory SQLite database using the specific
+ /// flags and vfs name.
+ ///
+ /// [Database Connection](http://www.sqlite.org/c3ref/open.html) for a description of valid
+ /// flag combinations.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if `vfs` cannot be converted to a C-compatible
+ /// string or if the underlying SQLite open call fails.
+ #[inline]
+ pub fn open_in_memory_with_flags_and_vfs(flags: OpenFlags, vfs: &str) -> Result<Connection> {
+ Connection::open_with_flags_and_vfs(":memory:", flags, vfs)
+ }
+
+ /// Convenience method to run multiple SQL statements (that cannot take any
+ /// parameters).
+ ///
+ /// ## Example
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn create_tables(conn: &Connection) -> Result<()> {
+ /// conn.execute_batch(
+ /// "BEGIN;
+ /// CREATE TABLE foo(x INTEGER);
+ /// CREATE TABLE bar(y TEXT);
+ /// COMMIT;",
+ /// )
+ /// }
+ /// ```
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if `sql` cannot be converted to a C-compatible string
+ /// or if the underlying SQLite call fails.
+ pub fn execute_batch(&self, sql: &str) -> Result<()> {
+ let mut sql = sql;
+ while !sql.is_empty() {
+ let stmt = self.prepare(sql)?;
+ if !stmt.stmt.is_null() && stmt.step()? && cfg!(feature = "extra_check") {
+ // Some PRAGMA may return rows
+ return Err(Error::ExecuteReturnedResults);
+ }
+ let tail = stmt.stmt.tail();
+ if tail == 0 || tail >= sql.len() {
+ break;
+ }
+ sql = &sql[tail..];
+ }
+ Ok(())
+ }
+
+ /// Convenience method to prepare and execute a single SQL statement.
+ ///
+ /// On success, returns the number of rows that were changed or inserted or
+ /// deleted (via `sqlite3_changes`).
+ ///
+ /// ## Example
+ ///
+ /// ### With positional params
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection};
+ /// fn update_rows(conn: &Connection) {
+ /// match conn.execute("UPDATE foo SET bar = 'baz' WHERE qux = ?1", [1i32]) {
+ /// Ok(updated) => println!("{} rows were updated", updated),
+ /// Err(err) => println!("update failed: {}", err),
+ /// }
+ /// }
+ /// ```
+ ///
+ /// ### With positional params of varying types
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{params, Connection};
+ /// fn update_rows(conn: &Connection) {
+ /// match conn.execute(
+ /// "UPDATE foo SET bar = 'baz' WHERE qux = ?1 AND quux = ?2",
+ /// params![1i32, 1.5f64],
+ /// ) {
+ /// Ok(updated) => println!("{} rows were updated", updated),
+ /// Err(err) => println!("update failed: {}", err),
+ /// }
+ /// }
+ /// ```
+ ///
+ /// ### With named params
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn insert(conn: &Connection) -> Result<usize> {
+ /// conn.execute(
+ /// "INSERT INTO test (name) VALUES (:name)",
+ /// &[(":name", "one")],
+ /// )
+ /// }
+ /// ```
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if `sql` cannot be converted to a C-compatible string
+ /// or if the underlying SQLite call fails.
+ #[inline]
+ pub fn execute<P: Params>(&self, sql: &str, params: P) -> Result<usize> {
+ self.prepare(sql)
+ .and_then(|mut stmt| stmt.check_no_tail().and_then(|()| stmt.execute(params)))
+ }
+
+ /// Returns the path to the database file, if one exists and is known.
+ ///
+ /// Returns `Some("")` for a temporary or in-memory database.
+ ///
+ /// Note that in some cases [PRAGMA
+ /// database_list](https://sqlite.org/pragma.html#pragma_database_list) is
+ /// likely to be more robust.
+ #[inline]
+ pub fn path(&self) -> Option<&str> {
+ unsafe {
+ let db = self.handle();
+ let db_name = DatabaseName::Main.as_cstring().unwrap();
+ let db_filename = ffi::sqlite3_db_filename(db, db_name.as_ptr());
+ if db_filename.is_null() {
+ None
+ } else {
+ CStr::from_ptr(db_filename).to_str().ok()
+ }
+ }
+ }
+
+ /// Attempts to free as much heap memory as possible from the database
+ /// connection.
+ ///
+ /// This calls [`sqlite3_db_release_memory`](https://www.sqlite.org/c3ref/db_release_memory.html).
+ #[inline]
+ #[cfg(feature = "release_memory")]
+ pub fn release_memory(&self) -> Result<()> {
+ self.db.borrow_mut().release_memory()
+ }
+
+ /// Get the SQLite rowid of the most recent successful INSERT.
+ ///
+ /// Uses [sqlite3_last_insert_rowid](https://www.sqlite.org/c3ref/last_insert_rowid.html) under
+ /// the hood.
+ #[inline]
+ pub fn last_insert_rowid(&self) -> i64 {
+ self.db.borrow_mut().last_insert_rowid()
+ }
+
+ /// Convenience method to execute a query that is expected to return a
+ /// single row.
+ ///
+ /// ## Example
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Result, Connection};
+ /// fn preferred_locale(conn: &Connection) -> Result<String> {
+ /// conn.query_row(
+ /// "SELECT value FROM preferences WHERE name='locale'",
+ /// [],
+ /// |row| row.get(0),
+ /// )
+ /// }
+ /// ```
+ ///
+ /// If the query returns more than one row, all rows except the first are
+ /// ignored.
+ ///
+ /// Returns `Err(QueryReturnedNoRows)` if no results are returned. If the
+ /// query truly is optional, you can call `.optional()` on the result of
+ /// this to get a `Result<Option<T>>`.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if `sql` cannot be converted to a C-compatible string
+ /// or if the underlying SQLite call fails.
+ #[inline]
+ pub fn query_row<T, P, F>(&self, sql: &str, params: P, f: F) -> Result<T>
+ where
+ P: Params,
+ F: FnOnce(&Row<'_>) -> Result<T>,
+ {
+ let mut stmt = self.prepare(sql)?;
+ stmt.check_no_tail()?;
+ stmt.query_row(params, f)
+ }
+
+ // https://sqlite.org/tclsqlite.html#onecolumn
+ #[cfg(test)]
+ pub(crate) fn one_column<T: types::FromSql>(&self, sql: &str) -> Result<T> {
+ self.query_row(sql, [], |r| r.get(0))
+ }
+
+ /// Convenience method to execute a query that is expected to return a
+ /// single row, and execute a mapping via `f` on that returned row with
+ /// the possibility of failure. The `Result` type of `f` must implement
+ /// `std::convert::From<Error>`.
+ ///
+ /// ## Example
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Result, Connection};
+ /// fn preferred_locale(conn: &Connection) -> Result<String> {
+ /// conn.query_row_and_then(
+ /// "SELECT value FROM preferences WHERE name='locale'",
+ /// [],
+ /// |row| row.get(0),
+ /// )
+ /// }
+ /// ```
+ ///
+ /// If the query returns more than one row, all rows except the first are
+ /// ignored.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if `sql` cannot be converted to a C-compatible string
+ /// or if the underlying SQLite call fails.
+ #[inline]
+ pub fn query_row_and_then<T, E, P, F>(&self, sql: &str, params: P, f: F) -> Result<T, E>
+ where
+ P: Params,
+ F: FnOnce(&Row<'_>) -> Result<T, E>,
+ E: From<Error>,
+ {
+ let mut stmt = self.prepare(sql)?;
+ stmt.check_no_tail()?;
+ let mut rows = stmt.query(params)?;
+
+ rows.get_expected_row().map_err(E::from).and_then(f)
+ }
+
+ /// Prepare a SQL statement for execution.
+ ///
+ /// ## Example
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn insert_new_people(conn: &Connection) -> Result<()> {
+ /// let mut stmt = conn.prepare("INSERT INTO People (name) VALUES (?1)")?;
+ /// stmt.execute(["Joe Smith"])?;
+ /// stmt.execute(["Bob Jones"])?;
+ /// Ok(())
+ /// }
+ /// ```
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if `sql` cannot be converted to a C-compatible string
+ /// or if the underlying SQLite call fails.
+ #[inline]
+ pub fn prepare(&self, sql: &str) -> Result<Statement<'_>> {
+ self.prepare_with_flags(sql, PrepFlags::default())
+ }
+
+ /// Prepare a SQL statement for execution.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if `sql` cannot be converted to a C-compatible string
+ /// or if the underlying SQLite call fails.
+ #[inline]
+ pub fn prepare_with_flags(&self, sql: &str, flags: PrepFlags) -> Result<Statement<'_>> {
+ self.db.borrow_mut().prepare(self, sql, flags)
+ }
+
+ /// Close the SQLite connection.
+ ///
+ /// This is functionally equivalent to the `Drop` implementation for
+ /// `Connection` except that on failure, it returns an error and the
+ /// connection itself (presumably so closing can be attempted again).
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite call fails.
+ #[inline]
+ pub fn close(self) -> Result<(), (Connection, Error)> {
+ self.flush_prepared_statement_cache();
+ let r = self.db.borrow_mut().close();
+ r.map_err(move |err| (self, err))
+ }
+
+ /// Enable loading of SQLite extensions from both SQL queries and Rust.
+ ///
+ /// You must call [`Connection::load_extension_disable`] when you're
+ /// finished loading extensions (failure to call it can lead to bad things,
+ /// see "Safety"), so you should strongly consider using
+ /// [`LoadExtensionGuard`] instead of this function, automatically disables
+ /// extension loading when it goes out of scope.
+ ///
+ /// # Example
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn load_my_extension(conn: &Connection) -> Result<()> {
+ /// // Safety: We fully trust the loaded extension and execute no untrusted SQL
+ /// // while extension loading is enabled.
+ /// unsafe {
+ /// conn.load_extension_enable()?;
+ /// let r = conn.load_extension("my/trusted/extension", None);
+ /// conn.load_extension_disable()?;
+ /// r
+ /// }
+ /// }
+ /// ```
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite call fails.
+ ///
+ /// # Safety
+ ///
+ /// TLDR: Don't execute any untrusted queries between this call and
+ /// [`Connection::load_extension_disable`].
+ ///
+ /// Perhaps surprisingly, this function does not only allow the use of
+ /// [`Connection::load_extension`] from Rust, but it also allows SQL queries
+ /// to perform [the same operation][loadext]. For example, in the period
+ /// between `load_extension_enable` and `load_extension_disable`, the
+ /// following operation will load and call some function in some dynamic
+ /// library:
+ ///
+ /// ```sql
+ /// SELECT load_extension('why_is_this_possible.dll', 'dubious_func');
+ /// ```
+ ///
+ /// This means that while this is enabled a carefully crafted SQL query can
+ /// be used to escalate a SQL injection attack into code execution.
+ ///
+ /// Safely using this function requires that you trust all SQL queries run
+ /// between when it is called, and when loading is disabled (by
+ /// [`Connection::load_extension_disable`]).
+ ///
+ /// [loadext]: https://www.sqlite.org/lang_corefunc.html#load_extension
+ #[cfg(feature = "load_extension")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "load_extension")))]
+ #[inline]
+ pub unsafe fn load_extension_enable(&self) -> Result<()> {
+ self.db.borrow_mut().enable_load_extension(1)
+ }
+
+ /// Disable loading of SQLite extensions.
+ ///
+ /// See [`Connection::load_extension_enable`] for an example.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite call fails.
+ #[cfg(feature = "load_extension")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "load_extension")))]
+ #[inline]
+ pub fn load_extension_disable(&self) -> Result<()> {
+ // It's always safe to turn off extension loading.
+ unsafe { self.db.borrow_mut().enable_load_extension(0) }
+ }
+
+ /// Load the SQLite extension at `dylib_path`. `dylib_path` is passed
+ /// through to `sqlite3_load_extension`, which may attempt OS-specific
+ /// modifications if the file cannot be loaded directly (for example
+ /// converting `"some/ext"` to `"some/ext.so"`, `"some\\ext.dll"`, ...).
+ ///
+ /// If `entry_point` is `None`, SQLite will attempt to find the entry point.
+ /// If it is not `None`, the entry point will be passed through to
+ /// `sqlite3_load_extension`.
+ ///
+ /// ## Example
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result, LoadExtensionGuard};
+ /// fn load_my_extension(conn: &Connection) -> Result<()> {
+ /// // Safety: we don't execute any SQL statements while
+ /// // extension loading is enabled.
+ /// let _guard = unsafe { LoadExtensionGuard::new(conn)? };
+ /// // Safety: `my_sqlite_extension` is highly trustworthy.
+ /// unsafe { conn.load_extension("my_sqlite_extension", None) }
+ /// }
+ /// ```
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite call fails.
+ ///
+ /// # Safety
+ ///
+ /// This is equivalent to performing a `dlopen`/`LoadLibrary` on a shared
+ /// library, and calling a function inside, and thus requires that you trust
+ /// the library that you're loading.
+ ///
+ /// That is to say: to safely use this, the code in the extension must be
+ /// sound, trusted, correctly use the SQLite APIs, and not contain any
+ /// memory or thread safety errors.
+ #[cfg(feature = "load_extension")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "load_extension")))]
+ #[inline]
+ pub unsafe fn load_extension<P: AsRef<Path>>(
+ &self,
+ dylib_path: P,
+ entry_point: Option<&str>,
+ ) -> Result<()> {
+ self.db
+ .borrow_mut()
+ .load_extension(dylib_path.as_ref(), entry_point)
+ }
+
+ /// Get access to the underlying SQLite database connection handle.
+ ///
+ /// # Warning
+ ///
+ /// You should not need to use this function. If you do need to, please
+ /// [open an issue on the rusqlite repository](https://github.com/rusqlite/rusqlite/issues) and describe
+ /// your use case.
+ ///
+ /// # Safety
+ ///
+ /// This function is unsafe because it gives you raw access
+ /// to the SQLite connection, and what you do with it could impact the
+ /// safety of this `Connection`.
+ #[inline]
+ pub unsafe fn handle(&self) -> *mut ffi::sqlite3 {
+ self.db.borrow().db()
+ }
+
+ /// Create a `Connection` from a raw handle.
+ ///
+ /// The underlying SQLite database connection handle will not be closed when
+ /// the returned connection is dropped/closed.
+ ///
+ /// # Safety
+ ///
+ /// This function is unsafe because improper use may impact the Connection.
+ #[inline]
+ pub unsafe fn from_handle(db: *mut ffi::sqlite3) -> Result<Connection> {
+ let db = InnerConnection::new(db, false);
+ Ok(Connection {
+ db: RefCell::new(db),
+ cache: StatementCache::with_capacity(STATEMENT_CACHE_DEFAULT_CAPACITY),
+ })
+ }
+
+ /// Like SQLITE_EXTENSION_INIT2 macro
+ #[cfg(feature = "loadable_extension")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "loadable_extension")))]
+ pub unsafe fn extension_init2(
+ db: *mut ffi::sqlite3,
+ p_api: *mut ffi::sqlite3_api_routines,
+ ) -> Result<Connection> {
+ ffi::rusqlite_extension_init2(p_api)?;
+ Connection::from_handle(db)
+ }
+
+ /// Create a `Connection` from a raw owned handle.
+ ///
+ /// The returned connection will attempt to close the inner connection
+ /// when dropped/closed. This function should only be called on connections
+ /// owned by the caller.
+ ///
+ /// # Safety
+ ///
+ /// This function is unsafe because improper use may impact the Connection.
+ /// In particular, it should only be called on connections created
+ /// and owned by the caller, e.g. as a result of calling
+ /// `ffi::sqlite3_open`().
+ #[inline]
+ pub unsafe fn from_handle_owned(db: *mut ffi::sqlite3) -> Result<Connection> {
+ let db = InnerConnection::new(db, true);
+ Ok(Connection {
+ db: RefCell::new(db),
+ cache: StatementCache::with_capacity(STATEMENT_CACHE_DEFAULT_CAPACITY),
+ })
+ }
+
+ /// Get access to a handle that can be used to interrupt long running
+ /// queries from another thread.
+ #[inline]
+ pub fn get_interrupt_handle(&self) -> InterruptHandle {
+ self.db.borrow().get_interrupt_handle()
+ }
+
+ #[inline]
+ fn decode_result(&self, code: c_int) -> Result<()> {
+ self.db.borrow().decode_result(code)
+ }
+
+ /// Return the number of rows modified, inserted or deleted by the most
+ /// recently completed INSERT, UPDATE or DELETE statement on the database
+ /// connection.
+ ///
+ /// See <https://www.sqlite.org/c3ref/changes.html>
+ #[inline]
+ pub fn changes(&self) -> u64 {
+ self.db.borrow().changes()
+ }
+
+ /// Test for auto-commit mode.
+ /// Autocommit mode is on by default.
+ #[inline]
+ pub fn is_autocommit(&self) -> bool {
+ self.db.borrow().is_autocommit()
+ }
+
+ /// Determine if all associated prepared statements have been reset.
+ #[inline]
+ pub fn is_busy(&self) -> bool {
+ self.db.borrow().is_busy()
+ }
+
+ /// Flush caches to disk mid-transaction
+ pub fn cache_flush(&self) -> Result<()> {
+ self.db.borrow_mut().cache_flush()
+ }
+
+ /// Determine if a database is read-only
+ pub fn is_readonly(&self, db_name: DatabaseName<'_>) -> Result<bool> {
+ self.db.borrow().db_readonly(db_name)
+ }
+}
+
+impl fmt::Debug for Connection {
+ fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+ f.debug_struct("Connection")
+ .field("path", &self.path())
+ .finish()
+ }
+}
+
+/// Batch iterator
+/// ```rust
+/// use rusqlite::{Batch, Connection, Result};
+///
+/// fn main() -> Result<()> {
+/// let conn = Connection::open_in_memory()?;
+/// let sql = r"
+/// CREATE TABLE tbl1 (col);
+/// CREATE TABLE tbl2 (col);
+/// ";
+/// let mut batch = Batch::new(&conn, sql);
+/// while let Some(mut stmt) = batch.next()? {
+/// stmt.execute([])?;
+/// }
+/// Ok(())
+/// }
+/// ```
+#[derive(Debug)]
+pub struct Batch<'conn, 'sql> {
+ conn: &'conn Connection,
+ sql: &'sql str,
+ tail: usize,
+}
+
+impl<'conn, 'sql> Batch<'conn, 'sql> {
+ /// Constructor
+ pub fn new(conn: &'conn Connection, sql: &'sql str) -> Batch<'conn, 'sql> {
+ Batch { conn, sql, tail: 0 }
+ }
+
+ /// Iterates on each batch statements.
+ ///
+ /// Returns `Ok(None)` when batch is completed.
+ #[allow(clippy::should_implement_trait)] // fallible iterator
+ pub fn next(&mut self) -> Result<Option<Statement<'conn>>> {
+ while self.tail < self.sql.len() {
+ let sql = &self.sql[self.tail..];
+ let next = self.conn.prepare(sql)?;
+ let tail = next.stmt.tail();
+ if tail == 0 {
+ self.tail = self.sql.len();
+ } else {
+ self.tail += tail;
+ }
+ if next.stmt.is_null() {
+ continue;
+ }
+ return Ok(Some(next));
+ }
+ Ok(None)
+ }
+}
+
+impl<'conn> Iterator for Batch<'conn, '_> {
+ type Item = Result<Statement<'conn>>;
+
+ fn next(&mut self) -> Option<Result<Statement<'conn>>> {
+ self.next().transpose()
+ }
+}
+
+bitflags::bitflags! {
+ /// Flags for opening SQLite database connections. See
+ /// [sqlite3_open_v2](https://www.sqlite.org/c3ref/open.html) for details.
+ ///
+ /// The default open flags are `SQLITE_OPEN_READ_WRITE | SQLITE_OPEN_CREATE
+ /// | SQLITE_OPEN_URI | SQLITE_OPEN_NO_MUTEX`. See [`Connection::open`] for
+ /// some discussion about these flags.
+ #[derive(Clone, Copy, Debug, Eq, Hash, PartialEq)]
+ #[repr(C)]
+ pub struct OpenFlags: ::std::os::raw::c_int {
+ /// The database is opened in read-only mode.
+ /// If the database does not already exist, an error is returned.
+ const SQLITE_OPEN_READ_ONLY = ffi::SQLITE_OPEN_READONLY;
+ /// The database is opened for reading and writing if possible,
+ /// or reading only if the file is write protected by the operating system.
+ /// In either case the database must already exist, otherwise an error is returned.
+ const SQLITE_OPEN_READ_WRITE = ffi::SQLITE_OPEN_READWRITE;
+ /// The database is created if it does not already exist
+ const SQLITE_OPEN_CREATE = ffi::SQLITE_OPEN_CREATE;
+ /// The filename can be interpreted as a URI if this flag is set.
+ const SQLITE_OPEN_URI = ffi::SQLITE_OPEN_URI;
+ /// The database will be opened as an in-memory database.
+ const SQLITE_OPEN_MEMORY = ffi::SQLITE_OPEN_MEMORY;
+ /// The new database connection will not use a per-connection mutex (the
+ /// connection will use the "multi-thread" threading mode, in SQLite
+ /// parlance).
+ ///
+ /// This is used by default, as proper `Send`/`Sync` usage (in
+ /// particular, the fact that [`Connection`] does not implement `Sync`)
+ /// ensures thread-safety without the need to perform locking around all
+ /// calls.
+ const SQLITE_OPEN_NO_MUTEX = ffi::SQLITE_OPEN_NOMUTEX;
+ /// The new database connection will use a per-connection mutex -- the
+ /// "serialized" threading mode, in SQLite parlance.
+ ///
+ /// # Caveats
+ ///
+ /// This flag should probably never be used with `rusqlite`, as we
+ /// ensure thread-safety statically (we implement [`Send`] and not
+ /// [`Sync`]). That said
+ ///
+ /// Critically, even if this flag is used, the [`Connection`] is not
+ /// safe to use across multiple threads simultaneously. To access a
+ /// database from multiple threads, you should either create multiple
+ /// connections, one for each thread (if you have very many threads,
+ /// wrapping the `rusqlite::Connection` in a mutex is also reasonable).
+ ///
+ /// This is both because of the additional per-connection state stored
+ /// by `rusqlite` (for example, the prepared statement cache), and
+ /// because not all of SQLites functions are fully thread safe, even in
+ /// serialized/`SQLITE_OPEN_FULLMUTEX` mode.
+ ///
+ /// All that said, it's fairly harmless to enable this flag with
+ /// `rusqlite`, it will just slow things down while providing no
+ /// benefit.
+ const SQLITE_OPEN_FULL_MUTEX = ffi::SQLITE_OPEN_FULLMUTEX;
+ /// The database is opened with shared cache enabled.
+ ///
+ /// This is frequently useful for in-memory connections, but note that
+ /// broadly speaking it's discouraged by SQLite itself, which states
+ /// "Any use of shared cache is discouraged" in the official
+ /// [documentation](https://www.sqlite.org/c3ref/enable_shared_cache.html).
+ const SQLITE_OPEN_SHARED_CACHE = 0x0002_0000;
+ /// The database is opened shared cache disabled.
+ const SQLITE_OPEN_PRIVATE_CACHE = 0x0004_0000;
+ /// The database filename is not allowed to be a symbolic link. (3.31.0)
+ const SQLITE_OPEN_NOFOLLOW = 0x0100_0000;
+ /// Extended result codes. (3.37.0)
+ const SQLITE_OPEN_EXRESCODE = 0x0200_0000;
+ }
+}
+
+impl Default for OpenFlags {
+ #[inline]
+ fn default() -> OpenFlags {
+ // Note: update the `Connection::open` and top-level `OpenFlags` docs if
+ // you change these.
+ OpenFlags::SQLITE_OPEN_READ_WRITE
+ | OpenFlags::SQLITE_OPEN_CREATE
+ | OpenFlags::SQLITE_OPEN_NO_MUTEX
+ | OpenFlags::SQLITE_OPEN_URI
+ }
+}
+
+bitflags::bitflags! {
+ /// Prepare flags. See
+ /// [sqlite3_prepare_v3](https://sqlite.org/c3ref/c_prepare_normalize.html) for details.
+ #[derive(Clone, Copy, Debug, Default, Eq, Hash, PartialEq)]
+ #[repr(C)]
+ pub struct PrepFlags: ::std::os::raw::c_uint {
+ /// A hint to the query planner that the prepared statement will be retained for a long time and probably reused many times.
+ const SQLITE_PREPARE_PERSISTENT = 0x01;
+ /// Causes the SQL compiler to return an error (error code SQLITE_ERROR) if the statement uses any virtual tables.
+ const SQLITE_PREPARE_NO_VTAB = 0x04;
+ }
+}
+
+/// rusqlite's check for a safe SQLite threading mode requires SQLite 3.7.0 or
+/// later. If you are running against a SQLite older than that, rusqlite
+/// attempts to ensure safety by performing configuration and initialization of
+/// SQLite itself the first time you
+/// attempt to open a connection. By default, rusqlite panics if that
+/// initialization fails, since that could mean SQLite has been initialized in
+/// single-thread mode.
+///
+/// If you are encountering that panic _and_ can ensure that SQLite has been
+/// initialized in either multi-thread or serialized mode, call this function
+/// prior to attempting to open a connection and rusqlite's initialization
+/// process will by skipped.
+///
+/// # Safety
+///
+/// This function is unsafe because if you call it and SQLite has actually been
+/// configured to run in single-thread mode,
+/// you may encounter memory errors or data corruption or any number of terrible
+/// things that should not be possible when you're using Rust.
+pub unsafe fn bypass_sqlite_initialization() {
+ BYPASS_SQLITE_INIT.store(true, Ordering::Relaxed);
+}
+
+/// Allows interrupting a long-running computation.
+pub struct InterruptHandle {
+ db_lock: Arc<Mutex<*mut ffi::sqlite3>>,
+}
+
+unsafe impl Send for InterruptHandle {}
+unsafe impl Sync for InterruptHandle {}
+
+impl InterruptHandle {
+ /// Interrupt the query currently executing on another thread. This will
+ /// cause that query to fail with a `SQLITE3_INTERRUPT` error.
+ pub fn interrupt(&self) {
+ let db_handle = self.db_lock.lock().unwrap();
+ if !db_handle.is_null() {
+ unsafe { ffi::sqlite3_interrupt(*db_handle) }
+ }
+ }
+}
+
+#[cfg(doctest)]
+doc_comment::doctest!("../README.md");
+
+#[cfg(test)]
+mod test {
+ use super::*;
+ use crate::ffi;
+ use fallible_iterator::FallibleIterator;
+ use std::error::Error as StdError;
+ use std::fmt;
+
+ // this function is never called, but is still type checked; in
+ // particular, calls with specific instantiations will require
+ // that those types are `Send`.
+ #[allow(
+ dead_code,
+ unconditional_recursion,
+ clippy::extra_unused_type_parameters
+ )]
+ fn ensure_send<T: Send>() {
+ ensure_send::<Connection>();
+ ensure_send::<InterruptHandle>();
+ }
+
+ #[allow(
+ dead_code,
+ unconditional_recursion,
+ clippy::extra_unused_type_parameters
+ )]
+ fn ensure_sync<T: Sync>() {
+ ensure_sync::<InterruptHandle>();
+ }
+
+ fn checked_memory_handle() -> Connection {
+ Connection::open_in_memory().unwrap()
+ }
+
+ #[test]
+ fn test_concurrent_transactions_busy_commit() -> Result<()> {
+ use std::time::Duration;
+ let tmp = tempfile::tempdir().unwrap();
+ let path = tmp.path().join("transactions.db3");
+
+ Connection::open(&path)?.execute_batch(
+ "
+ BEGIN; CREATE TABLE foo(x INTEGER);
+ INSERT INTO foo VALUES(42); END;",
+ )?;
+
+ let mut db1 = Connection::open_with_flags(&path, OpenFlags::SQLITE_OPEN_READ_WRITE)?;
+ let mut db2 = Connection::open_with_flags(&path, OpenFlags::SQLITE_OPEN_READ_ONLY)?;
+
+ db1.busy_timeout(Duration::from_millis(0))?;
+ db2.busy_timeout(Duration::from_millis(0))?;
+
+ {
+ let tx1 = db1.transaction()?;
+ let tx2 = db2.transaction()?;
+
+ // SELECT first makes sqlite lock with a shared lock
+ tx1.query_row("SELECT x FROM foo LIMIT 1", [], |_| Ok(()))?;
+ tx2.query_row("SELECT x FROM foo LIMIT 1", [], |_| Ok(()))?;
+
+ tx1.execute("INSERT INTO foo VALUES(?1)", [1])?;
+ let _ = tx2.execute("INSERT INTO foo VALUES(?1)", [2]);
+
+ let _ = tx1.commit();
+ let _ = tx2.commit();
+ }
+
+ let _ = db1
+ .transaction()
+ .expect("commit should have closed transaction");
+ let _ = db2
+ .transaction()
+ .expect("commit should have closed transaction");
+ Ok(())
+ }
+
+ #[test]
+ fn test_persistence() -> Result<()> {
+ let temp_dir = tempfile::tempdir().unwrap();
+ let path = temp_dir.path().join("test.db3");
+
+ {
+ let db = Connection::open(&path)?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER);
+ INSERT INTO foo VALUES(42);
+ END;";
+ db.execute_batch(sql)?;
+ }
+
+ let path_string = path.to_str().unwrap();
+ let db = Connection::open(path_string)?;
+ let the_answer: i64 = db.one_column("SELECT x FROM foo")?;
+
+ assert_eq!(42i64, the_answer);
+ Ok(())
+ }
+
+ #[test]
+ fn test_open() {
+ Connection::open_in_memory().unwrap();
+
+ let db = checked_memory_handle();
+ db.close().unwrap();
+ }
+
+ #[test]
+ fn test_path() -> Result<()> {
+ let tmp = tempfile::tempdir().unwrap();
+ let db = Connection::open("")?;
+ assert_eq!(Some(""), db.path());
+ let db = Connection::open_in_memory()?;
+ assert_eq!(Some(""), db.path());
+ let db = Connection::open("file:dummy.db?mode=memory&cache=shared")?;
+ assert_eq!(Some(""), db.path());
+ let path = tmp.path().join("file.db");
+ let db = Connection::open(path)?;
+ assert!(db.path().map(|p| p.ends_with("file.db")).unwrap_or(false));
+ Ok(())
+ }
+
+ #[test]
+ fn test_open_failure() {
+ let filename = "no_such_file.db";
+ let result = Connection::open_with_flags(filename, OpenFlags::SQLITE_OPEN_READ_ONLY);
+ let err = result.unwrap_err();
+ if let Error::SqliteFailure(e, Some(msg)) = err {
+ assert_eq!(ErrorCode::CannotOpen, e.code);
+ assert_eq!(ffi::SQLITE_CANTOPEN, e.extended_code);
+ assert!(
+ msg.contains(filename),
+ "error message '{msg}' does not contain '{filename}'"
+ );
+ } else {
+ panic!("SqliteFailure expected");
+ }
+ }
+
+ #[cfg(unix)]
+ #[test]
+ fn test_invalid_unicode_file_names() -> Result<()> {
+ use std::ffi::OsStr;
+ use std::fs::File;
+ use std::os::unix::ffi::OsStrExt;
+ let temp_dir = tempfile::tempdir().unwrap();
+
+ let path = temp_dir.path();
+ if File::create(path.join(OsStr::from_bytes(&[0xFE]))).is_err() {
+ // Skip test, filesystem doesn't support invalid Unicode
+ return Ok(());
+ }
+ let db_path = path.join(OsStr::from_bytes(&[0xFF]));
+ {
+ let db = Connection::open(&db_path)?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER);
+ INSERT INTO foo VALUES(42);
+ END;";
+ db.execute_batch(sql)?;
+ }
+
+ let db = Connection::open(&db_path)?;
+ let the_answer: i64 = db.one_column("SELECT x FROM foo")?;
+
+ assert_eq!(42i64, the_answer);
+ Ok(())
+ }
+
+ #[test]
+ fn test_close_retry() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ // force the DB to be busy by preparing a statement; this must be done at the
+ // FFI level to allow us to call .close() without dropping the prepared
+ // statement first.
+ let raw_stmt = {
+ use super::str_to_cstring;
+ use std::os::raw::c_int;
+ use std::ptr;
+
+ let raw_db = db.db.borrow_mut().db;
+ let sql = "SELECT 1";
+ let mut raw_stmt: *mut ffi::sqlite3_stmt = ptr::null_mut();
+ let cstring = str_to_cstring(sql)?;
+ let rc = unsafe {
+ ffi::sqlite3_prepare_v2(
+ raw_db,
+ cstring.as_ptr(),
+ (sql.len() + 1) as c_int,
+ &mut raw_stmt,
+ ptr::null_mut(),
+ )
+ };
+ assert_eq!(rc, ffi::SQLITE_OK);
+ raw_stmt
+ };
+
+ // now that we have an open statement, trying (and retrying) to close should
+ // fail.
+ let (db, _) = db.close().unwrap_err();
+ let (db, _) = db.close().unwrap_err();
+ let (db, _) = db.close().unwrap_err();
+
+ // finalize the open statement so a final close will succeed
+ assert_eq!(ffi::SQLITE_OK, unsafe { ffi::sqlite3_finalize(raw_stmt) });
+
+ db.close().unwrap();
+ Ok(())
+ }
+
+ #[test]
+ fn test_open_with_flags() {
+ for bad_flags in &[
+ OpenFlags::empty(),
+ OpenFlags::SQLITE_OPEN_READ_ONLY | OpenFlags::SQLITE_OPEN_READ_WRITE,
+ OpenFlags::SQLITE_OPEN_READ_ONLY | OpenFlags::SQLITE_OPEN_CREATE,
+ ] {
+ Connection::open_in_memory_with_flags(*bad_flags).unwrap_err();
+ }
+ }
+
+ #[test]
+ fn test_execute_batch() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER);
+ INSERT INTO foo VALUES(1);
+ INSERT INTO foo VALUES(2);
+ INSERT INTO foo VALUES(3);
+ INSERT INTO foo VALUES(4);
+ END;";
+ db.execute_batch(sql)?;
+
+ db.execute_batch("UPDATE foo SET x = 3 WHERE x < 3")?;
+
+ db.execute_batch("INVALID SQL").unwrap_err();
+ Ok(())
+ }
+
+ #[test]
+ fn test_execute() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(x INTEGER)")?;
+
+ assert_eq!(1, db.execute("INSERT INTO foo(x) VALUES (?1)", [1i32])?);
+ assert_eq!(1, db.execute("INSERT INTO foo(x) VALUES (?1)", [2i32])?);
+
+ assert_eq!(3i32, db.one_column::<i32>("SELECT SUM(x) FROM foo")?);
+ Ok(())
+ }
+
+ #[test]
+ #[cfg(feature = "extra_check")]
+ fn test_execute_select() {
+ let db = checked_memory_handle();
+ let err = db.execute("SELECT 1 WHERE 1 < ?1", [1i32]).unwrap_err();
+ assert_eq!(
+ err,
+ Error::ExecuteReturnedResults,
+ "Unexpected error: {err}"
+ );
+ }
+
+ #[test]
+ #[cfg(feature = "extra_check")]
+ fn test_execute_multiple() {
+ let db = checked_memory_handle();
+ let err = db
+ .execute(
+ "CREATE TABLE foo(x INTEGER); CREATE TABLE foo(x INTEGER)",
+ [],
+ )
+ .unwrap_err();
+ match err {
+ Error::MultipleStatement => (),
+ _ => panic!("Unexpected error: {err}"),
+ }
+ }
+
+ #[test]
+ fn test_prepare_column_names() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(x INTEGER);")?;
+
+ let stmt = db.prepare("SELECT * FROM foo")?;
+ assert_eq!(stmt.column_count(), 1);
+ assert_eq!(stmt.column_names(), vec!["x"]);
+
+ let stmt = db.prepare("SELECT x AS a, x AS b FROM foo")?;
+ assert_eq!(stmt.column_count(), 2);
+ assert_eq!(stmt.column_names(), vec!["a", "b"]);
+ Ok(())
+ }
+
+ #[test]
+ fn test_prepare_execute() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(x INTEGER);")?;
+
+ let mut insert_stmt = db.prepare("INSERT INTO foo(x) VALUES(?1)")?;
+ assert_eq!(insert_stmt.execute([1i32])?, 1);
+ assert_eq!(insert_stmt.execute([2i32])?, 1);
+ assert_eq!(insert_stmt.execute([3i32])?, 1);
+
+ assert_eq!(insert_stmt.execute(["hello"])?, 1);
+ assert_eq!(insert_stmt.execute(["goodbye"])?, 1);
+ assert_eq!(insert_stmt.execute([types::Null])?, 1);
+
+ let mut update_stmt = db.prepare("UPDATE foo SET x=?1 WHERE x<?2")?;
+ assert_eq!(update_stmt.execute([3i32, 3i32])?, 2);
+ assert_eq!(update_stmt.execute([3i32, 3i32])?, 0);
+ assert_eq!(update_stmt.execute([8i32, 8i32])?, 3);
+ Ok(())
+ }
+
+ #[test]
+ fn test_prepare_query() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(x INTEGER);")?;
+
+ let mut insert_stmt = db.prepare("INSERT INTO foo(x) VALUES(?1)")?;
+ assert_eq!(insert_stmt.execute([1i32])?, 1);
+ assert_eq!(insert_stmt.execute([2i32])?, 1);
+ assert_eq!(insert_stmt.execute([3i32])?, 1);
+
+ let mut query = db.prepare("SELECT x FROM foo WHERE x < ?1 ORDER BY x DESC")?;
+ {
+ let mut rows = query.query([4i32])?;
+ let mut v = Vec::<i32>::new();
+
+ while let Some(row) = rows.next()? {
+ v.push(row.get(0)?);
+ }
+
+ assert_eq!(v, [3i32, 2, 1]);
+ }
+
+ {
+ let mut rows = query.query([3i32])?;
+ let mut v = Vec::<i32>::new();
+
+ while let Some(row) = rows.next()? {
+ v.push(row.get(0)?);
+ }
+
+ assert_eq!(v, [2i32, 1]);
+ }
+ Ok(())
+ }
+
+ #[test]
+ fn test_query_map() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER, y TEXT);
+ INSERT INTO foo VALUES(4, \"hello\");
+ INSERT INTO foo VALUES(3, \", \");
+ INSERT INTO foo VALUES(2, \"world\");
+ INSERT INTO foo VALUES(1, \"!\");
+ END;";
+ db.execute_batch(sql)?;
+
+ let mut query = db.prepare("SELECT x, y FROM foo ORDER BY x DESC")?;
+ let results: Result<Vec<String>> = query.query([])?.map(|row| row.get(1)).collect();
+
+ assert_eq!(results?.concat(), "hello, world!");
+ Ok(())
+ }
+
+ #[test]
+ fn test_query_row() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER);
+ INSERT INTO foo VALUES(1);
+ INSERT INTO foo VALUES(2);
+ INSERT INTO foo VALUES(3);
+ INSERT INTO foo VALUES(4);
+ END;";
+ db.execute_batch(sql)?;
+
+ assert_eq!(10i64, db.one_column::<i64>("SELECT SUM(x) FROM foo")?);
+
+ let result: Result<i64> = db.one_column("SELECT x FROM foo WHERE x > 5");
+ match result.unwrap_err() {
+ Error::QueryReturnedNoRows => (),
+ err => panic!("Unexpected error {err}"),
+ }
+
+ let bad_query_result = db.query_row("NOT A PROPER QUERY; test123", [], |_| Ok(()));
+
+ bad_query_result.unwrap_err();
+ Ok(())
+ }
+
+ #[test]
+ fn test_optional() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ let result: Result<i64> = db.one_column("SELECT 1 WHERE 0 <> 0");
+ let result = result.optional();
+ match result? {
+ None => (),
+ _ => panic!("Unexpected result"),
+ }
+
+ let result: Result<i64> = db.one_column("SELECT 1 WHERE 0 == 0");
+ let result = result.optional();
+ match result? {
+ Some(1) => (),
+ _ => panic!("Unexpected result"),
+ }
+
+ let bad_query_result: Result<i64> = db.one_column("NOT A PROPER QUERY");
+ let bad_query_result = bad_query_result.optional();
+ bad_query_result.unwrap_err();
+ Ok(())
+ }
+
+ #[test]
+ fn test_pragma_query_row() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ assert_eq!("memory", db.one_column::<String>("PRAGMA journal_mode")?);
+ let mode = db.one_column::<String>("PRAGMA journal_mode=off")?;
+ if cfg!(features = "bundled") {
+ assert_eq!(mode, "off");
+ } else {
+ // Note: system SQLite on macOS defaults to "off" rather than
+ // "memory" for the journal mode (which cannot be changed for
+ // in-memory connections). This seems like it's *probably* legal
+ // according to the docs below, so we relax this test when not
+ // bundling:
+ //
+ // From https://www.sqlite.org/pragma.html#pragma_journal_mode
+ // > Note that the journal_mode for an in-memory database is either
+ // > MEMORY or OFF and can not be changed to a different value. An
+ // > attempt to change the journal_mode of an in-memory database to
+ // > any setting other than MEMORY or OFF is ignored.
+ assert!(mode == "memory" || mode == "off", "Got mode {mode:?}");
+ }
+
+ Ok(())
+ }
+
+ #[test]
+ fn test_prepare_failures() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(x INTEGER);")?;
+
+ let err = db.prepare("SELECT * FROM does_not_exist").unwrap_err();
+ assert!(format!("{err}").contains("does_not_exist"));
+ Ok(())
+ }
+
+ #[test]
+ fn test_last_insert_rowid() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(x INTEGER PRIMARY KEY)")?;
+ db.execute_batch("INSERT INTO foo DEFAULT VALUES")?;
+
+ assert_eq!(db.last_insert_rowid(), 1);
+
+ let mut stmt = db.prepare("INSERT INTO foo DEFAULT VALUES")?;
+ for _ in 0i32..9 {
+ stmt.execute([])?;
+ }
+ assert_eq!(db.last_insert_rowid(), 10);
+ Ok(())
+ }
+
+ #[test]
+ fn test_is_autocommit() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ assert!(
+ db.is_autocommit(),
+ "autocommit expected to be active by default"
+ );
+ Ok(())
+ }
+
+ #[test]
+ fn test_is_busy() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ assert!(!db.is_busy());
+ let mut stmt = db.prepare("PRAGMA schema_version")?;
+ assert!(!db.is_busy());
+ {
+ let mut rows = stmt.query([])?;
+ assert!(!db.is_busy());
+ let row = rows.next()?;
+ assert!(db.is_busy());
+ assert!(row.is_some());
+ }
+ assert!(!db.is_busy());
+ Ok(())
+ }
+
+ #[test]
+ fn test_statement_debugging() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let query = "SELECT 12345";
+ let stmt = db.prepare(query)?;
+
+ assert!(format!("{stmt:?}").contains(query));
+ Ok(())
+ }
+
+ #[test]
+ fn test_notnull_constraint_error() -> Result<()> {
+ // extended error codes for constraints were added in SQLite 3.7.16; if we're
+ // running on our bundled version, we know the extended error code exists.
+ fn check_extended_code(extended_code: c_int) {
+ assert_eq!(extended_code, ffi::SQLITE_CONSTRAINT_NOTNULL);
+ }
+
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(x NOT NULL)")?;
+
+ let result = db.execute("INSERT INTO foo (x) VALUES (NULL)", []);
+
+ match result.unwrap_err() {
+ Error::SqliteFailure(err, _) => {
+ assert_eq!(err.code, ErrorCode::ConstraintViolation);
+ check_extended_code(err.extended_code);
+ }
+ err => panic!("Unexpected error {err}"),
+ }
+ Ok(())
+ }
+
+ #[test]
+ fn test_version_string() {
+ let n = version_number();
+ let major = n / 1_000_000;
+ let minor = (n % 1_000_000) / 1_000;
+ let patch = n % 1_000;
+
+ assert!(version().contains(&format!("{major}.{minor}.{patch}")));
+ }
+
+ #[test]
+ #[cfg(feature = "functions")]
+ fn test_interrupt() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ let interrupt_handle = db.get_interrupt_handle();
+
+ db.create_scalar_function(
+ "interrupt",
+ 0,
+ functions::FunctionFlags::default(),
+ move |_| {
+ interrupt_handle.interrupt();
+ Ok(0)
+ },
+ )?;
+
+ let mut stmt =
+ db.prepare("SELECT interrupt() FROM (SELECT 1 UNION SELECT 2 UNION SELECT 3)")?;
+
+ let result: Result<Vec<i32>> = stmt.query([])?.map(|r| r.get(0)).collect();
+
+ assert_eq!(
+ result.unwrap_err().sqlite_error_code(),
+ Some(ErrorCode::OperationInterrupted)
+ );
+ Ok(())
+ }
+
+ #[test]
+ fn test_interrupt_close() {
+ let db = checked_memory_handle();
+ let handle = db.get_interrupt_handle();
+ handle.interrupt();
+ db.close().unwrap();
+ handle.interrupt();
+
+ // Look at it's internals to see if we cleared it out properly.
+ let db_guard = handle.db_lock.lock().unwrap();
+ assert!(db_guard.is_null());
+ // It would be nice to test that we properly handle close/interrupt
+ // running at the same time, but it seems impossible to do with any
+ // degree of reliability.
+ }
+
+ #[test]
+ fn test_get_raw() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(i, x);")?;
+ let vals = ["foobar", "1234", "qwerty"];
+ let mut insert_stmt = db.prepare("INSERT INTO foo(i, x) VALUES(?1, ?2)")?;
+ for (i, v) in vals.iter().enumerate() {
+ let i_to_insert = i as i64;
+ assert_eq!(insert_stmt.execute(params![i_to_insert, v])?, 1);
+ }
+
+ let mut query = db.prepare("SELECT i, x FROM foo")?;
+ let mut rows = query.query([])?;
+
+ while let Some(row) = rows.next()? {
+ let i = row.get_ref(0)?.as_i64()?;
+ let expect = vals[i as usize];
+ let x = row.get_ref("x")?.as_str()?;
+ assert_eq!(x, expect);
+ }
+
+ let mut query = db.prepare("SELECT x FROM foo")?;
+ let rows = query.query_map([], |row| {
+ let x = row.get_ref(0)?.as_str()?; // check From<FromSqlError> for Error
+ Ok(x[..].to_owned())
+ })?;
+
+ for (i, row) in rows.enumerate() {
+ assert_eq!(row?, vals[i]);
+ }
+ Ok(())
+ }
+
+ #[test]
+ fn test_from_handle() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let handle = unsafe { db.handle() };
+ {
+ let db = unsafe { Connection::from_handle(handle) }?;
+ db.execute_batch("PRAGMA VACUUM")?;
+ }
+ db.close().unwrap();
+ Ok(())
+ }
+
+ #[test]
+ fn test_from_handle_owned() -> Result<()> {
+ let mut handle: *mut ffi::sqlite3 = std::ptr::null_mut();
+ let r = unsafe { ffi::sqlite3_open(":memory:\0".as_ptr() as *const i8, &mut handle) };
+ assert_eq!(r, ffi::SQLITE_OK);
+ let db = unsafe { Connection::from_handle_owned(handle) }?;
+ db.execute_batch("PRAGMA VACUUM")?;
+ Ok(())
+ }
+
+ mod query_and_then_tests {
+
+ use super::*;
+
+ #[derive(Debug)]
+ enum CustomError {
+ SomeError,
+ Sqlite(Error),
+ }
+
+ impl fmt::Display for CustomError {
+ fn fmt(&self, f: &mut fmt::Formatter<'_>) -> Result<(), fmt::Error> {
+ match *self {
+ CustomError::SomeError => write!(f, "my custom error"),
+ CustomError::Sqlite(ref se) => write!(f, "my custom error: {se}"),
+ }
+ }
+ }
+
+ impl StdError for CustomError {
+ fn description(&self) -> &str {
+ "my custom error"
+ }
+
+ fn cause(&self) -> Option<&dyn StdError> {
+ match *self {
+ CustomError::SomeError => None,
+ CustomError::Sqlite(ref se) => Some(se),
+ }
+ }
+ }
+
+ impl From<Error> for CustomError {
+ fn from(se: Error) -> CustomError {
+ CustomError::Sqlite(se)
+ }
+ }
+
+ type CustomResult<T> = Result<T, CustomError>;
+
+ #[test]
+ fn test_query_and_then() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER, y TEXT);
+ INSERT INTO foo VALUES(4, \"hello\");
+ INSERT INTO foo VALUES(3, \", \");
+ INSERT INTO foo VALUES(2, \"world\");
+ INSERT INTO foo VALUES(1, \"!\");
+ END;";
+ db.execute_batch(sql)?;
+
+ let mut query = db.prepare("SELECT x, y FROM foo ORDER BY x DESC")?;
+ let results: Result<Vec<String>> =
+ query.query_and_then([], |row| row.get(1))?.collect();
+
+ assert_eq!(results?.concat(), "hello, world!");
+ Ok(())
+ }
+
+ #[test]
+ fn test_query_and_then_fails() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER, y TEXT);
+ INSERT INTO foo VALUES(4, \"hello\");
+ INSERT INTO foo VALUES(3, \", \");
+ INSERT INTO foo VALUES(2, \"world\");
+ INSERT INTO foo VALUES(1, \"!\");
+ END;";
+ db.execute_batch(sql)?;
+
+ let mut query = db.prepare("SELECT x, y FROM foo ORDER BY x DESC")?;
+ let bad_type: Result<Vec<f64>> = query.query_and_then([], |row| row.get(1))?.collect();
+
+ match bad_type.unwrap_err() {
+ Error::InvalidColumnType(..) => (),
+ err => panic!("Unexpected error {err}"),
+ }
+
+ let bad_idx: Result<Vec<String>> =
+ query.query_and_then([], |row| row.get(3))?.collect();
+
+ match bad_idx.unwrap_err() {
+ Error::InvalidColumnIndex(_) => (),
+ err => panic!("Unexpected error {err}"),
+ }
+ Ok(())
+ }
+
+ #[test]
+ fn test_query_and_then_custom_error() -> CustomResult<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER, y TEXT);
+ INSERT INTO foo VALUES(4, \"hello\");
+ INSERT INTO foo VALUES(3, \", \");
+ INSERT INTO foo VALUES(2, \"world\");
+ INSERT INTO foo VALUES(1, \"!\");
+ END;";
+ db.execute_batch(sql)?;
+
+ let mut query = db.prepare("SELECT x, y FROM foo ORDER BY x DESC")?;
+ let results: CustomResult<Vec<String>> = query
+ .query_and_then([], |row| row.get(1).map_err(CustomError::Sqlite))?
+ .collect();
+
+ assert_eq!(results?.concat(), "hello, world!");
+ Ok(())
+ }
+
+ #[test]
+ fn test_query_and_then_custom_error_fails() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER, y TEXT);
+ INSERT INTO foo VALUES(4, \"hello\");
+ INSERT INTO foo VALUES(3, \", \");
+ INSERT INTO foo VALUES(2, \"world\");
+ INSERT INTO foo VALUES(1, \"!\");
+ END;";
+ db.execute_batch(sql)?;
+
+ let mut query = db.prepare("SELECT x, y FROM foo ORDER BY x DESC")?;
+ let bad_type: CustomResult<Vec<f64>> = query
+ .query_and_then([], |row| row.get(1).map_err(CustomError::Sqlite))?
+ .collect();
+
+ match bad_type.unwrap_err() {
+ CustomError::Sqlite(Error::InvalidColumnType(..)) => (),
+ err => panic!("Unexpected error {err}"),
+ }
+
+ let bad_idx: CustomResult<Vec<String>> = query
+ .query_and_then([], |row| row.get(3).map_err(CustomError::Sqlite))?
+ .collect();
+
+ match bad_idx.unwrap_err() {
+ CustomError::Sqlite(Error::InvalidColumnIndex(_)) => (),
+ err => panic!("Unexpected error {err}"),
+ }
+
+ let non_sqlite_err: CustomResult<Vec<String>> = query
+ .query_and_then([], |_| Err(CustomError::SomeError))?
+ .collect();
+
+ match non_sqlite_err.unwrap_err() {
+ CustomError::SomeError => (),
+ err => panic!("Unexpected error {err}"),
+ }
+ Ok(())
+ }
+
+ #[test]
+ fn test_query_row_and_then_custom_error() -> CustomResult<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER, y TEXT);
+ INSERT INTO foo VALUES(4, \"hello\");
+ END;";
+ db.execute_batch(sql)?;
+
+ let query = "SELECT x, y FROM foo ORDER BY x DESC";
+ let results: CustomResult<String> =
+ db.query_row_and_then(query, [], |row| row.get(1).map_err(CustomError::Sqlite));
+
+ assert_eq!(results?, "hello");
+ Ok(())
+ }
+
+ #[test]
+ fn test_query_row_and_then_custom_error_fails() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER, y TEXT);
+ INSERT INTO foo VALUES(4, \"hello\");
+ END;";
+ db.execute_batch(sql)?;
+
+ let query = "SELECT x, y FROM foo ORDER BY x DESC";
+ let bad_type: CustomResult<f64> =
+ db.query_row_and_then(query, [], |row| row.get(1).map_err(CustomError::Sqlite));
+
+ match bad_type.unwrap_err() {
+ CustomError::Sqlite(Error::InvalidColumnType(..)) => (),
+ err => panic!("Unexpected error {err}"),
+ }
+
+ let bad_idx: CustomResult<String> =
+ db.query_row_and_then(query, [], |row| row.get(3).map_err(CustomError::Sqlite));
+
+ match bad_idx.unwrap_err() {
+ CustomError::Sqlite(Error::InvalidColumnIndex(_)) => (),
+ err => panic!("Unexpected error {err}"),
+ }
+
+ let non_sqlite_err: CustomResult<String> =
+ db.query_row_and_then(query, [], |_| Err(CustomError::SomeError));
+
+ match non_sqlite_err.unwrap_err() {
+ CustomError::SomeError => (),
+ err => panic!("Unexpected error {err}"),
+ }
+ Ok(())
+ }
+ }
+
+ #[test]
+ fn test_dynamic() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER, y TEXT);
+ INSERT INTO foo VALUES(4, \"hello\");
+ END;";
+ db.execute_batch(sql)?;
+
+ db.query_row("SELECT * FROM foo", [], |r| {
+ assert_eq!(2, r.as_ref().column_count());
+ Ok(())
+ })
+ }
+ #[test]
+ fn test_dyn_box() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(x INTEGER);")?;
+ let b: Box<dyn ToSql> = Box::new(5);
+ db.execute("INSERT INTO foo VALUES(?1)", [b])?;
+ db.query_row("SELECT x FROM foo", [], |r| {
+ assert_eq!(5, r.get_unwrap::<_, i32>(0));
+ Ok(())
+ })
+ }
+
+ #[test]
+ fn test_params() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.query_row(
+ "SELECT
+ ?1, ?2, ?3, ?4, ?5, ?6, ?7, ?8, ?9, ?10,
+ ?11, ?12, ?13, ?14, ?15, ?16, ?17, ?18, ?19, ?20,
+ ?21, ?22, ?23, ?24, ?25, ?26, ?27, ?28, ?29, ?30,
+ ?31, ?32, ?33, ?34;",
+ params![
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+ 1, 1, 1, 1, 1, 1,
+ ],
+ |r| {
+ assert_eq!(1, r.get_unwrap::<_, i32>(0));
+ Ok(())
+ },
+ )
+ }
+
+ #[test]
+ #[cfg(not(feature = "extra_check"))]
+ fn test_alter_table() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE x(t);")?;
+ // `execute_batch` should be used but `execute` should also work
+ db.execute("ALTER TABLE x RENAME TO y;", [])?;
+ Ok(())
+ }
+
+ #[test]
+ fn test_batch() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = r"
+ CREATE TABLE tbl1 (col);
+ CREATE TABLE tbl2 (col);
+ ";
+ let batch = Batch::new(&db, sql);
+ for stmt in batch {
+ let mut stmt = stmt?;
+ stmt.execute([])?;
+ }
+ Ok(())
+ }
+
+ #[test]
+ #[cfg(feature = "modern_sqlite")]
+ fn test_returning() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(x INTEGER PRIMARY KEY)")?;
+ let row_id = db.one_column::<i64>("INSERT INTO foo DEFAULT VALUES RETURNING ROWID")?;
+ assert_eq!(row_id, 1);
+ Ok(())
+ }
+
+ #[test]
+ fn test_cache_flush() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.cache_flush()
+ }
+
+ #[test]
+ fn db_readonly() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ assert!(!db.is_readonly(MAIN_DB)?);
+ Ok(())
+ }
+
+ #[test]
+ #[cfg(feature = "rusqlite-macros")]
+ fn prepare_and_bind() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let name = "Lisa";
+ let age = 8;
+ let mut stmt = prepare_and_bind!(db, "SELECT $name, $age;");
+ let (v1, v2) = stmt
+ .raw_query()
+ .next()
+ .and_then(|o| o.ok_or(Error::QueryReturnedNoRows))
+ .and_then(|r| Ok((r.get::<_, String>(0)?, r.get::<_, i64>(1)?)))?;
+ assert_eq!((v1.as_str(), v2), (name, age));
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/limits.rs b/third_party/rust/rusqlite/src/limits.rs
new file mode 100644
index 0000000000..d0694e304c
--- /dev/null
+++ b/third_party/rust/rusqlite/src/limits.rs
@@ -0,0 +1,163 @@
+//! Run-Time Limits
+
+use crate::{ffi, Connection};
+use std::os::raw::c_int;
+
+/// Run-Time limit categories, for use with [`Connection::limit`] and
+/// [`Connection::set_limit`].
+///
+/// See the official documentation for more information:
+/// - <https://www.sqlite.org/c3ref/c_limit_attached.html>
+/// - <https://www.sqlite.org/limits.html>
+#[repr(i32)]
+#[non_exhaustive]
+#[allow(clippy::upper_case_acronyms, non_camel_case_types)]
+#[cfg_attr(docsrs, doc(cfg(feature = "limits")))]
+pub enum Limit {
+ /// The maximum size of any string or BLOB or table row, in bytes.
+ SQLITE_LIMIT_LENGTH = ffi::SQLITE_LIMIT_LENGTH,
+ /// The maximum length of an SQL statement, in bytes.
+ SQLITE_LIMIT_SQL_LENGTH = ffi::SQLITE_LIMIT_SQL_LENGTH,
+ /// The maximum number of columns in a table definition or in the result set
+ /// of a SELECT or the maximum number of columns in an index or in an
+ /// ORDER BY or GROUP BY clause.
+ SQLITE_LIMIT_COLUMN = ffi::SQLITE_LIMIT_COLUMN,
+ /// The maximum depth of the parse tree on any expression.
+ SQLITE_LIMIT_EXPR_DEPTH = ffi::SQLITE_LIMIT_EXPR_DEPTH,
+ /// The maximum number of terms in a compound SELECT statement.
+ SQLITE_LIMIT_COMPOUND_SELECT = ffi::SQLITE_LIMIT_COMPOUND_SELECT,
+ /// The maximum number of instructions in a virtual machine program used to
+ /// implement an SQL statement.
+ SQLITE_LIMIT_VDBE_OP = ffi::SQLITE_LIMIT_VDBE_OP,
+ /// The maximum number of arguments on a function.
+ SQLITE_LIMIT_FUNCTION_ARG = ffi::SQLITE_LIMIT_FUNCTION_ARG,
+ /// The maximum number of attached databases.
+ SQLITE_LIMIT_ATTACHED = ffi::SQLITE_LIMIT_ATTACHED,
+ /// The maximum length of the pattern argument to the LIKE or GLOB
+ /// operators.
+ SQLITE_LIMIT_LIKE_PATTERN_LENGTH = ffi::SQLITE_LIMIT_LIKE_PATTERN_LENGTH,
+ /// The maximum index number of any parameter in an SQL statement.
+ SQLITE_LIMIT_VARIABLE_NUMBER = ffi::SQLITE_LIMIT_VARIABLE_NUMBER,
+ /// The maximum depth of recursion for triggers.
+ SQLITE_LIMIT_TRIGGER_DEPTH = ffi::SQLITE_LIMIT_TRIGGER_DEPTH,
+ /// The maximum number of auxiliary worker threads that a single prepared
+ /// statement may start.
+ SQLITE_LIMIT_WORKER_THREADS = ffi::SQLITE_LIMIT_WORKER_THREADS,
+}
+
+impl Connection {
+ /// Returns the current value of a [`Limit`].
+ #[inline]
+ #[cfg_attr(docsrs, doc(cfg(feature = "limits")))]
+ pub fn limit(&self, limit: Limit) -> i32 {
+ let c = self.db.borrow();
+ unsafe { ffi::sqlite3_limit(c.db(), limit as c_int, -1) }
+ }
+
+ /// Changes the [`Limit`] to `new_val`, returning the prior
+ /// value of the limit.
+ #[inline]
+ #[cfg_attr(docsrs, doc(cfg(feature = "limits")))]
+ pub fn set_limit(&self, limit: Limit, new_val: i32) -> i32 {
+ let c = self.db.borrow_mut();
+ unsafe { ffi::sqlite3_limit(c.db(), limit as c_int, new_val) }
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use super::*;
+ use crate::{Connection, Result};
+
+ #[test]
+ fn test_limit_values() {
+ assert_eq!(Limit::SQLITE_LIMIT_LENGTH as i32, ffi::SQLITE_LIMIT_LENGTH,);
+ assert_eq!(
+ Limit::SQLITE_LIMIT_SQL_LENGTH as i32,
+ ffi::SQLITE_LIMIT_SQL_LENGTH,
+ );
+ assert_eq!(Limit::SQLITE_LIMIT_COLUMN as i32, ffi::SQLITE_LIMIT_COLUMN,);
+ assert_eq!(
+ Limit::SQLITE_LIMIT_EXPR_DEPTH as i32,
+ ffi::SQLITE_LIMIT_EXPR_DEPTH,
+ );
+ assert_eq!(
+ Limit::SQLITE_LIMIT_COMPOUND_SELECT as i32,
+ ffi::SQLITE_LIMIT_COMPOUND_SELECT,
+ );
+ assert_eq!(
+ Limit::SQLITE_LIMIT_VDBE_OP as i32,
+ ffi::SQLITE_LIMIT_VDBE_OP,
+ );
+ assert_eq!(
+ Limit::SQLITE_LIMIT_FUNCTION_ARG as i32,
+ ffi::SQLITE_LIMIT_FUNCTION_ARG,
+ );
+ assert_eq!(
+ Limit::SQLITE_LIMIT_ATTACHED as i32,
+ ffi::SQLITE_LIMIT_ATTACHED,
+ );
+ assert_eq!(
+ Limit::SQLITE_LIMIT_LIKE_PATTERN_LENGTH as i32,
+ ffi::SQLITE_LIMIT_LIKE_PATTERN_LENGTH,
+ );
+ assert_eq!(
+ Limit::SQLITE_LIMIT_VARIABLE_NUMBER as i32,
+ ffi::SQLITE_LIMIT_VARIABLE_NUMBER,
+ );
+ #[cfg(feature = "bundled")]
+ assert_eq!(
+ Limit::SQLITE_LIMIT_TRIGGER_DEPTH as i32,
+ ffi::SQLITE_LIMIT_TRIGGER_DEPTH,
+ );
+ #[cfg(feature = "bundled")]
+ assert_eq!(
+ Limit::SQLITE_LIMIT_WORKER_THREADS as i32,
+ ffi::SQLITE_LIMIT_WORKER_THREADS,
+ );
+ }
+
+ #[test]
+ fn test_limit() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.set_limit(Limit::SQLITE_LIMIT_LENGTH, 1024);
+ assert_eq!(1024, db.limit(Limit::SQLITE_LIMIT_LENGTH));
+
+ db.set_limit(Limit::SQLITE_LIMIT_SQL_LENGTH, 1024);
+ assert_eq!(1024, db.limit(Limit::SQLITE_LIMIT_SQL_LENGTH));
+
+ db.set_limit(Limit::SQLITE_LIMIT_COLUMN, 64);
+ assert_eq!(64, db.limit(Limit::SQLITE_LIMIT_COLUMN));
+
+ db.set_limit(Limit::SQLITE_LIMIT_EXPR_DEPTH, 256);
+ assert_eq!(256, db.limit(Limit::SQLITE_LIMIT_EXPR_DEPTH));
+
+ db.set_limit(Limit::SQLITE_LIMIT_COMPOUND_SELECT, 32);
+ assert_eq!(32, db.limit(Limit::SQLITE_LIMIT_COMPOUND_SELECT));
+
+ db.set_limit(Limit::SQLITE_LIMIT_FUNCTION_ARG, 32);
+ assert_eq!(32, db.limit(Limit::SQLITE_LIMIT_FUNCTION_ARG));
+
+ db.set_limit(Limit::SQLITE_LIMIT_ATTACHED, 2);
+ assert_eq!(2, db.limit(Limit::SQLITE_LIMIT_ATTACHED));
+
+ db.set_limit(Limit::SQLITE_LIMIT_LIKE_PATTERN_LENGTH, 128);
+ assert_eq!(128, db.limit(Limit::SQLITE_LIMIT_LIKE_PATTERN_LENGTH));
+
+ db.set_limit(Limit::SQLITE_LIMIT_VARIABLE_NUMBER, 99);
+ assert_eq!(99, db.limit(Limit::SQLITE_LIMIT_VARIABLE_NUMBER));
+
+ // SQLITE_LIMIT_TRIGGER_DEPTH was added in SQLite 3.6.18.
+ if crate::version_number() >= 3_006_018 {
+ db.set_limit(Limit::SQLITE_LIMIT_TRIGGER_DEPTH, 32);
+ assert_eq!(32, db.limit(Limit::SQLITE_LIMIT_TRIGGER_DEPTH));
+ }
+
+ // SQLITE_LIMIT_WORKER_THREADS was added in SQLite 3.8.7.
+ if crate::version_number() >= 3_008_007 {
+ db.set_limit(Limit::SQLITE_LIMIT_WORKER_THREADS, 2);
+ assert_eq!(2, db.limit(Limit::SQLITE_LIMIT_WORKER_THREADS));
+ }
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/load_extension_guard.rs b/third_party/rust/rusqlite/src/load_extension_guard.rs
new file mode 100644
index 0000000000..deed3b4bdd
--- /dev/null
+++ b/third_party/rust/rusqlite/src/load_extension_guard.rs
@@ -0,0 +1,46 @@
+use crate::{Connection, Result};
+
+/// RAII guard temporarily enabling SQLite extensions to be loaded.
+///
+/// ## Example
+///
+/// ```rust,no_run
+/// # use rusqlite::{Connection, Result, LoadExtensionGuard};
+/// # use std::path::{Path};
+/// fn load_my_extension(conn: &Connection) -> Result<()> {
+/// unsafe {
+/// let _guard = LoadExtensionGuard::new(conn)?;
+/// conn.load_extension("trusted/sqlite/extension", None)
+/// }
+/// }
+/// ```
+#[cfg_attr(docsrs, doc(cfg(feature = "load_extension")))]
+pub struct LoadExtensionGuard<'conn> {
+ conn: &'conn Connection,
+}
+
+impl LoadExtensionGuard<'_> {
+ /// Attempt to enable loading extensions. Loading extensions will be
+ /// disabled when this guard goes out of scope. Cannot be meaningfully
+ /// nested.
+ ///
+ /// # Safety
+ ///
+ /// You must not run untrusted queries while extension loading is enabled.
+ ///
+ /// See the safety comment on [`Connection::load_extension_enable`] for more
+ /// details.
+ #[inline]
+ pub unsafe fn new(conn: &Connection) -> Result<LoadExtensionGuard<'_>> {
+ conn.load_extension_enable()
+ .map(|_| LoadExtensionGuard { conn })
+ }
+}
+
+#[allow(unused_must_use)]
+impl Drop for LoadExtensionGuard<'_> {
+ #[inline]
+ fn drop(&mut self) {
+ self.conn.load_extension_disable();
+ }
+}
diff --git a/third_party/rust/rusqlite/src/params.rs b/third_party/rust/rusqlite/src/params.rs
new file mode 100644
index 0000000000..a4c506667c
--- /dev/null
+++ b/third_party/rust/rusqlite/src/params.rs
@@ -0,0 +1,455 @@
+use crate::{Result, Statement, ToSql};
+
+mod sealed {
+ /// This trait exists just to ensure that the only impls of `trait Params`
+ /// that are allowed are ones in this crate.
+ pub trait Sealed {}
+}
+use sealed::Sealed;
+
+/// Trait used for [sets of parameter][params] passed into SQL
+/// statements/queries.
+///
+/// [params]: https://www.sqlite.org/c3ref/bind_blob.html
+///
+/// Note: Currently, this trait can only be implemented inside this crate.
+/// Additionally, it's methods (which are `doc(hidden)`) should currently not be
+/// considered part of the stable API, although it's possible they will
+/// stabilize in the future.
+///
+/// # Passing parameters to SQLite
+///
+/// Many functions in this library let you pass parameters to SQLite. Doing this
+/// lets you avoid any risk of SQL injection, and is simpler than escaping
+/// things manually. Aside from deprecated functions and a few helpers, this is
+/// indicated by the function taking a generic argument that implements `Params`
+/// (this trait).
+///
+/// ## Positional parameters
+///
+/// For cases where you want to pass a list of parameters where the number of
+/// parameters is known at compile time, this can be done in one of the
+/// following ways:
+///
+/// - For small lists of parameters up to 16 items, they may alternatively be
+/// passed as a tuple, as in `thing.query((1, "foo"))`.
+///
+/// This is somewhat inconvenient for a single item, since you need a
+/// weird-looking trailing comma: `thing.query(("example",))`. That case is
+/// perhaps more cleanly expressed as `thing.query(["example"])`.
+///
+/// - Using the [`rusqlite::params!`](crate::params!) macro, e.g.
+/// `thing.query(rusqlite::params![1, "foo", bar])`. This is mostly useful for
+/// heterogeneous lists where the number of parameters greater than 16, or
+/// homogenous lists of parameters where the number of parameters exceeds 32.
+///
+/// - For small homogeneous lists of parameters, they can either be passed as:
+///
+/// - an array, as in `thing.query([1i32, 2, 3, 4])` or `thing.query(["foo",
+/// "bar", "baz"])`.
+///
+/// - a reference to an array of references, as in `thing.query(&["foo",
+/// "bar", "baz"])` or `thing.query(&[&1i32, &2, &3])`.
+///
+/// (Note: in this case we don't implement this for slices for coherence
+/// reasons, so it really is only for the "reference to array" types —
+/// hence why the number of parameters must be <= 32 or you need to
+/// reach for `rusqlite::params!`)
+///
+/// Unfortunately, in the current design it's not possible to allow this for
+/// references to arrays of non-references (e.g. `&[1i32, 2, 3]`). Code like
+/// this should instead either use `params!`, an array literal, a `&[&dyn
+/// ToSql]` or if none of those work, [`ParamsFromIter`].
+///
+/// - As a slice of `ToSql` trait object references, e.g. `&[&dyn ToSql]`. This
+/// is mostly useful for passing parameter lists around as arguments without
+/// having every function take a generic `P: Params`.
+///
+/// ### Example (positional)
+///
+/// ```rust,no_run
+/// # use rusqlite::{Connection, Result, params};
+/// fn update_rows(conn: &Connection) -> Result<()> {
+/// let mut stmt = conn.prepare("INSERT INTO test (a, b) VALUES (?1, ?2)")?;
+///
+/// // Using a tuple:
+/// stmt.execute((0, "foobar"))?;
+///
+/// // Using `rusqlite::params!`:
+/// stmt.execute(params![1i32, "blah"])?;
+///
+/// // array literal — non-references
+/// stmt.execute([2i32, 3i32])?;
+///
+/// // array literal — references
+/// stmt.execute(["foo", "bar"])?;
+///
+/// // Slice literal, references:
+/// stmt.execute(&[&2i32, &3i32])?;
+///
+/// // Note: The types behind the references don't have to be `Sized`
+/// stmt.execute(&["foo", "bar"])?;
+///
+/// // However, this doesn't work (see above):
+/// // stmt.execute(&[1i32, 2i32])?;
+/// Ok(())
+/// }
+/// ```
+///
+/// ## Named parameters
+///
+/// SQLite lets you name parameters using a number of conventions (":foo",
+/// "@foo", "$foo"). You can pass named parameters in to SQLite using rusqlite
+/// in a few ways:
+///
+/// - Using the [`rusqlite::named_params!`](crate::named_params!) macro, as in
+/// `stmt.execute(named_params!{ ":name": "foo", ":age": 99 })`. Similar to
+/// the `params` macro, this is most useful for heterogeneous lists of
+/// parameters, or lists where the number of parameters exceeds 32.
+///
+/// - As a slice of `&[(&str, &dyn ToSql)]`. This is what essentially all of
+/// these boil down to in the end, conceptually at least. In theory you can
+/// pass this as `stmt`.
+///
+/// - As array references, similar to the positional params. This looks like
+/// `thing.query(&[(":foo", &1i32), (":bar", &2i32)])` or
+/// `thing.query(&[(":foo", "abc"), (":bar", "def")])`.
+///
+/// Note: Unbound named parameters will be left to the value they previously
+/// were bound with, falling back to `NULL` for parameters which have never been
+/// bound.
+///
+/// ### Example (named)
+///
+/// ```rust,no_run
+/// # use rusqlite::{Connection, Result, named_params};
+/// fn insert(conn: &Connection) -> Result<()> {
+/// let mut stmt = conn.prepare("INSERT INTO test (key, value) VALUES (:key, :value)")?;
+/// // Using `rusqlite::params!`:
+/// stmt.execute(named_params! { ":key": "one", ":val": 2 })?;
+/// // Alternatively:
+/// stmt.execute(&[(":key", "three"), (":val", "four")])?;
+/// // Or:
+/// stmt.execute(&[(":key", &100), (":val", &200)])?;
+/// Ok(())
+/// }
+/// ```
+///
+/// ## No parameters
+///
+/// You can just use an empty tuple or the empty array literal to run a query
+/// that accepts no parameters.
+///
+/// ### Example (no parameters)
+///
+/// The empty tuple:
+///
+/// ```rust,no_run
+/// # use rusqlite::{Connection, Result, params};
+/// fn delete_all_users(conn: &Connection) -> Result<()> {
+/// // You may also use `()`.
+/// conn.execute("DELETE FROM users", ())?;
+/// Ok(())
+/// }
+/// ```
+///
+/// The empty array:
+///
+/// ```rust,no_run
+/// # use rusqlite::{Connection, Result, params};
+/// fn delete_all_users(conn: &Connection) -> Result<()> {
+/// // Just use an empty array (e.g. `[]`) for no params.
+/// conn.execute("DELETE FROM users", [])?;
+/// Ok(())
+/// }
+/// ```
+///
+/// ## Dynamic parameter list
+///
+/// If you have a number of parameters which is unknown at compile time (for
+/// example, building a dynamic query at runtime), you have two choices:
+///
+/// - Use a `&[&dyn ToSql]`. This is often annoying to construct if you don't
+/// already have this type on-hand.
+/// - Use the [`ParamsFromIter`] type. This essentially lets you wrap an
+/// iterator some `T: ToSql` with something that implements `Params`. The
+/// usage of this looks like `rusqlite::params_from_iter(something)`.
+///
+/// A lot of the considerations here are similar either way, so you should see
+/// the [`ParamsFromIter`] documentation for more info / examples.
+pub trait Params: Sealed {
+ // XXX not public api, might not need to expose.
+ //
+ // Binds the parameters to the statement. It is unlikely calling this
+ // explicitly will do what you want. Please use `Statement::query` or
+ // similar directly.
+ //
+ // For now, just hide the function in the docs...
+ #[doc(hidden)]
+ fn __bind_in(self, stmt: &mut Statement<'_>) -> Result<()>;
+}
+
+// Explicitly impl for empty array. Critically, for `conn.execute([])` to be
+// unambiguous, this must be the *only* implementation for an empty array.
+//
+// This sadly prevents `impl<T: ToSql, const N: usize> Params for [T; N]`, which
+// forces people to use `params![...]` or `rusqlite::params_from_iter` for long
+// homogenous lists of parameters. This is not that big of a deal, but is
+// unfortunate, especially because I mostly did it because I wanted a simple
+// syntax for no-params that didnt require importing -- the empty tuple fits
+// that nicely, but I didn't think of it until much later.
+//
+// Admittedly, if we did have the generic impl, then we *wouldn't* support the
+// empty array literal as a parameter, since the `T` there would fail to be
+// inferred. The error message here would probably be quite bad, and so on
+// further thought, probably would end up causing *more* surprises, not less.
+impl Sealed for [&(dyn ToSql + Send + Sync); 0] {}
+impl Params for [&(dyn ToSql + Send + Sync); 0] {
+ #[inline]
+ fn __bind_in(self, stmt: &mut Statement<'_>) -> Result<()> {
+ stmt.ensure_parameter_count(0)
+ }
+}
+
+impl Sealed for &[&dyn ToSql] {}
+impl Params for &[&dyn ToSql] {
+ #[inline]
+ fn __bind_in(self, stmt: &mut Statement<'_>) -> Result<()> {
+ stmt.bind_parameters(self)
+ }
+}
+
+impl Sealed for &[(&str, &dyn ToSql)] {}
+impl Params for &[(&str, &dyn ToSql)] {
+ #[inline]
+ fn __bind_in(self, stmt: &mut Statement<'_>) -> Result<()> {
+ stmt.bind_parameters_named(self)
+ }
+}
+
+// Manual impls for the empty and singleton tuple, although the rest are covered
+// by macros.
+impl Sealed for () {}
+impl Params for () {
+ #[inline]
+ fn __bind_in(self, stmt: &mut Statement<'_>) -> Result<()> {
+ stmt.ensure_parameter_count(0)
+ }
+}
+
+// I'm pretty sure you could tweak the `single_tuple_impl` to accept this.
+impl<T: ToSql> Sealed for (T,) {}
+impl<T: ToSql> Params for (T,) {
+ #[inline]
+ fn __bind_in(self, stmt: &mut Statement<'_>) -> Result<()> {
+ stmt.ensure_parameter_count(1)?;
+ stmt.raw_bind_parameter(1, self.0)?;
+ Ok(())
+ }
+}
+
+macro_rules! single_tuple_impl {
+ ($count:literal : $(($field:tt $ftype:ident)),* $(,)?) => {
+ impl<$($ftype,)*> Sealed for ($($ftype,)*) where $($ftype: ToSql,)* {}
+ impl<$($ftype,)*> Params for ($($ftype,)*) where $($ftype: ToSql,)* {
+ fn __bind_in(self, stmt: &mut Statement<'_>) -> Result<()> {
+ stmt.ensure_parameter_count($count)?;
+ $({
+ debug_assert!($field < $count);
+ stmt.raw_bind_parameter($field + 1, self.$field)?;
+ })+
+ Ok(())
+ }
+ }
+ }
+}
+
+// We use a the macro for the rest, but don't bother with trying to implement it
+// in a single invocation (it's possible to do, but my attempts were almost the
+// same amount of code as just writing it out this way, and much more dense --
+// it is a more complicated case than the TryFrom macro we have for row->tuple).
+//
+// Note that going up to 16 (rather than the 12 that the impls in the stdlib
+// usually support) is just because we did the same in the `TryFrom<Row>` impl.
+// I didn't catch that then, but there's no reason to remove it, and it seems
+// nice to be consistent here; this way putting data in the database and getting
+// data out of the database are more symmetric in a (mostly superficial) sense.
+single_tuple_impl!(2: (0 A), (1 B));
+single_tuple_impl!(3: (0 A), (1 B), (2 C));
+single_tuple_impl!(4: (0 A), (1 B), (2 C), (3 D));
+single_tuple_impl!(5: (0 A), (1 B), (2 C), (3 D), (4 E));
+single_tuple_impl!(6: (0 A), (1 B), (2 C), (3 D), (4 E), (5 F));
+single_tuple_impl!(7: (0 A), (1 B), (2 C), (3 D), (4 E), (5 F), (6 G));
+single_tuple_impl!(8: (0 A), (1 B), (2 C), (3 D), (4 E), (5 F), (6 G), (7 H));
+single_tuple_impl!(9: (0 A), (1 B), (2 C), (3 D), (4 E), (5 F), (6 G), (7 H), (8 I));
+single_tuple_impl!(10: (0 A), (1 B), (2 C), (3 D), (4 E), (5 F), (6 G), (7 H), (8 I), (9 J));
+single_tuple_impl!(11: (0 A), (1 B), (2 C), (3 D), (4 E), (5 F), (6 G), (7 H), (8 I), (9 J), (10 K));
+single_tuple_impl!(12: (0 A), (1 B), (2 C), (3 D), (4 E), (5 F), (6 G), (7 H), (8 I), (9 J), (10 K), (11 L));
+single_tuple_impl!(13: (0 A), (1 B), (2 C), (3 D), (4 E), (5 F), (6 G), (7 H), (8 I), (9 J), (10 K), (11 L), (12 M));
+single_tuple_impl!(14: (0 A), (1 B), (2 C), (3 D), (4 E), (5 F), (6 G), (7 H), (8 I), (9 J), (10 K), (11 L), (12 M), (13 N));
+single_tuple_impl!(15: (0 A), (1 B), (2 C), (3 D), (4 E), (5 F), (6 G), (7 H), (8 I), (9 J), (10 K), (11 L), (12 M), (13 N), (14 O));
+single_tuple_impl!(16: (0 A), (1 B), (2 C), (3 D), (4 E), (5 F), (6 G), (7 H), (8 I), (9 J), (10 K), (11 L), (12 M), (13 N), (14 O), (15 P));
+
+macro_rules! impl_for_array_ref {
+ ($($N:literal)+) => {$(
+ // These are already generic, and there's a shedload of them, so lets
+ // avoid the compile time hit from making them all inline for now.
+ impl<T: ToSql + ?Sized> Sealed for &[&T; $N] {}
+ impl<T: ToSql + ?Sized> Params for &[&T; $N] {
+ fn __bind_in(self, stmt: &mut Statement<'_>) -> Result<()> {
+ stmt.bind_parameters(self)
+ }
+ }
+ impl<T: ToSql + ?Sized> Sealed for &[(&str, &T); $N] {}
+ impl<T: ToSql + ?Sized> Params for &[(&str, &T); $N] {
+ fn __bind_in(self, stmt: &mut Statement<'_>) -> Result<()> {
+ stmt.bind_parameters_named(self)
+ }
+ }
+ impl<T: ToSql> Sealed for [T; $N] {}
+ impl<T: ToSql> Params for [T; $N] {
+ #[inline]
+ fn __bind_in(self, stmt: &mut Statement<'_>) -> Result<()> {
+ stmt.bind_parameters(&self)
+ }
+ }
+ )+};
+}
+
+// Following libstd/libcore's (old) lead, implement this for arrays up to `[_;
+// 32]`. Note `[_; 0]` is intentionally omitted for coherence reasons, see the
+// note above the impl of `[&dyn ToSql; 0]` for more information.
+//
+// Note that this unfortunately means we can't use const generics here, but I
+// don't really think it matters -- users who hit that can use `params!` anyway.
+impl_for_array_ref!(
+ 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
+ 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
+);
+
+/// Adapter type which allows any iterator over [`ToSql`] values to implement
+/// [`Params`].
+///
+/// This struct is created by the [`params_from_iter`] function.
+///
+/// This can be useful if you have something like an `&[String]` (of unknown
+/// length), and you want to use them with an API that wants something
+/// implementing `Params`. This way, you can avoid having to allocate storage
+/// for something like a `&[&dyn ToSql]`.
+///
+/// This essentially is only ever actually needed when dynamically generating
+/// SQL — static SQL (by definition) has the number of parameters known
+/// statically. As dynamically generating SQL is itself pretty advanced, this
+/// API is itself for advanced use cases (See "Realistic use case" in the
+/// examples).
+///
+/// # Example
+///
+/// ## Basic usage
+///
+/// ```rust,no_run
+/// use rusqlite::{params_from_iter, Connection, Result};
+/// use std::collections::BTreeSet;
+///
+/// fn query(conn: &Connection, ids: &BTreeSet<String>) -> Result<()> {
+/// assert_eq!(ids.len(), 3, "Unrealistic sample code");
+///
+/// let mut stmt = conn.prepare("SELECT * FROM users WHERE id IN (?1, ?2, ?3)")?;
+/// let _rows = stmt.query(params_from_iter(ids.iter()))?;
+///
+/// // use _rows...
+/// Ok(())
+/// }
+/// ```
+///
+/// ## Realistic use case
+///
+/// Here's how you'd use `ParamsFromIter` to call [`Statement::exists`] with a
+/// dynamic number of parameters.
+///
+/// ```rust,no_run
+/// use rusqlite::{Connection, Result};
+///
+/// pub fn any_active_users(conn: &Connection, usernames: &[String]) -> Result<bool> {
+/// if usernames.is_empty() {
+/// return Ok(false);
+/// }
+///
+/// // Note: `repeat_vars` never returns anything attacker-controlled, so
+/// // it's fine to use it in a dynamically-built SQL string.
+/// let vars = repeat_vars(usernames.len());
+///
+/// let sql = format!(
+/// // In practice this would probably be better as an `EXISTS` query.
+/// "SELECT 1 FROM user WHERE is_active AND name IN ({}) LIMIT 1",
+/// vars,
+/// );
+/// let mut stmt = conn.prepare(&sql)?;
+/// stmt.exists(rusqlite::params_from_iter(usernames))
+/// }
+///
+/// // Helper function to return a comma-separated sequence of `?`.
+/// // - `repeat_vars(0) => panic!(...)`
+/// // - `repeat_vars(1) => "?"`
+/// // - `repeat_vars(2) => "?,?"`
+/// // - `repeat_vars(3) => "?,?,?"`
+/// // - ...
+/// fn repeat_vars(count: usize) -> String {
+/// assert_ne!(count, 0);
+/// let mut s = "?,".repeat(count);
+/// // Remove trailing comma
+/// s.pop();
+/// s
+/// }
+/// ```
+///
+/// That is fairly complex, and even so would need even more work to be fully
+/// production-ready:
+///
+/// - production code should ensure `usernames` isn't so large that it will
+/// surpass [`conn.limit(Limit::SQLITE_LIMIT_VARIABLE_NUMBER)`][limits]),
+/// chunking if too large. (Note that the limits api requires rusqlite to have
+/// the "limits" feature).
+///
+/// - `repeat_vars` can be implemented in a way that avoids needing to allocate
+/// a String.
+///
+/// - Etc...
+///
+/// [limits]: crate::Connection::limit
+///
+/// This complexity reflects the fact that `ParamsFromIter` is mainly intended
+/// for advanced use cases — most of the time you should know how many
+/// parameters you have statically (and if you don't, you're either doing
+/// something tricky, or should take a moment to think about the design).
+#[derive(Clone, Debug)]
+pub struct ParamsFromIter<I>(I);
+
+/// Constructor function for a [`ParamsFromIter`]. See its documentation for
+/// more.
+#[inline]
+pub fn params_from_iter<I>(iter: I) -> ParamsFromIter<I>
+where
+ I: IntoIterator,
+ I::Item: ToSql,
+{
+ ParamsFromIter(iter)
+}
+
+impl<I> Sealed for ParamsFromIter<I>
+where
+ I: IntoIterator,
+ I::Item: ToSql,
+{
+}
+
+impl<I> Params for ParamsFromIter<I>
+where
+ I: IntoIterator,
+ I::Item: ToSql,
+{
+ #[inline]
+ fn __bind_in(self, stmt: &mut Statement<'_>) -> Result<()> {
+ stmt.bind_parameters(self.0)
+ }
+}
diff --git a/third_party/rust/rusqlite/src/pragma.rs b/third_party/rust/rusqlite/src/pragma.rs
new file mode 100644
index 0000000000..46bbde14a2
--- /dev/null
+++ b/third_party/rust/rusqlite/src/pragma.rs
@@ -0,0 +1,454 @@
+//! Pragma helpers
+
+use std::ops::Deref;
+
+use crate::error::Error;
+use crate::ffi;
+use crate::types::{ToSql, ToSqlOutput, ValueRef};
+use crate::{Connection, DatabaseName, Result, Row};
+
+pub struct Sql {
+ buf: String,
+}
+
+impl Sql {
+ pub fn new() -> Sql {
+ Sql { buf: String::new() }
+ }
+
+ pub fn push_pragma(
+ &mut self,
+ schema_name: Option<DatabaseName<'_>>,
+ pragma_name: &str,
+ ) -> Result<()> {
+ self.push_keyword("PRAGMA")?;
+ self.push_space();
+ if let Some(schema_name) = schema_name {
+ self.push_schema_name(schema_name);
+ self.push_dot();
+ }
+ self.push_keyword(pragma_name)
+ }
+
+ pub fn push_keyword(&mut self, keyword: &str) -> Result<()> {
+ if !keyword.is_empty() && is_identifier(keyword) {
+ self.buf.push_str(keyword);
+ Ok(())
+ } else {
+ Err(Error::SqliteFailure(
+ ffi::Error::new(ffi::SQLITE_MISUSE),
+ Some(format!("Invalid keyword \"{keyword}\"")),
+ ))
+ }
+ }
+
+ pub fn push_schema_name(&mut self, schema_name: DatabaseName<'_>) {
+ match schema_name {
+ DatabaseName::Main => self.buf.push_str("main"),
+ DatabaseName::Temp => self.buf.push_str("temp"),
+ DatabaseName::Attached(s) => self.push_identifier(s),
+ };
+ }
+
+ pub fn push_identifier(&mut self, s: &str) {
+ if is_identifier(s) {
+ self.buf.push_str(s);
+ } else {
+ self.wrap_and_escape(s, '"');
+ }
+ }
+
+ pub fn push_value(&mut self, value: &dyn ToSql) -> Result<()> {
+ let value = value.to_sql()?;
+ let value = match value {
+ ToSqlOutput::Borrowed(v) => v,
+ ToSqlOutput::Owned(ref v) => ValueRef::from(v),
+ #[cfg(feature = "blob")]
+ ToSqlOutput::ZeroBlob(_) => {
+ return Err(Error::SqliteFailure(
+ ffi::Error::new(ffi::SQLITE_MISUSE),
+ Some(format!("Unsupported value \"{value:?}\"")),
+ ));
+ }
+ #[cfg(feature = "array")]
+ ToSqlOutput::Array(_) => {
+ return Err(Error::SqliteFailure(
+ ffi::Error::new(ffi::SQLITE_MISUSE),
+ Some(format!("Unsupported value \"{value:?}\"")),
+ ));
+ }
+ };
+ match value {
+ ValueRef::Integer(i) => {
+ self.push_int(i);
+ }
+ ValueRef::Real(r) => {
+ self.push_real(r);
+ }
+ ValueRef::Text(s) => {
+ let s = std::str::from_utf8(s)?;
+ self.push_string_literal(s);
+ }
+ _ => {
+ return Err(Error::SqliteFailure(
+ ffi::Error::new(ffi::SQLITE_MISUSE),
+ Some(format!("Unsupported value \"{value:?}\"")),
+ ));
+ }
+ };
+ Ok(())
+ }
+
+ pub fn push_string_literal(&mut self, s: &str) {
+ self.wrap_and_escape(s, '\'');
+ }
+
+ pub fn push_int(&mut self, i: i64) {
+ self.buf.push_str(&i.to_string());
+ }
+
+ pub fn push_real(&mut self, f: f64) {
+ self.buf.push_str(&f.to_string());
+ }
+
+ pub fn push_space(&mut self) {
+ self.buf.push(' ');
+ }
+
+ pub fn push_dot(&mut self) {
+ self.buf.push('.');
+ }
+
+ pub fn push_equal_sign(&mut self) {
+ self.buf.push('=');
+ }
+
+ pub fn open_brace(&mut self) {
+ self.buf.push('(');
+ }
+
+ pub fn close_brace(&mut self) {
+ self.buf.push(')');
+ }
+
+ pub fn as_str(&self) -> &str {
+ &self.buf
+ }
+
+ fn wrap_and_escape(&mut self, s: &str, quote: char) {
+ self.buf.push(quote);
+ let chars = s.chars();
+ for ch in chars {
+ // escape `quote` by doubling it
+ if ch == quote {
+ self.buf.push(ch);
+ }
+ self.buf.push(ch);
+ }
+ self.buf.push(quote);
+ }
+}
+
+impl Deref for Sql {
+ type Target = str;
+
+ fn deref(&self) -> &str {
+ self.as_str()
+ }
+}
+
+impl Connection {
+ /// Query the current value of `pragma_name`.
+ ///
+ /// Some pragmas will return multiple rows/values which cannot be retrieved
+ /// with this method.
+ ///
+ /// Prefer [PRAGMA function](https://sqlite.org/pragma.html#pragfunc) introduced in SQLite 3.20:
+ /// `SELECT user_version FROM pragma_user_version;`
+ pub fn pragma_query_value<T, F>(
+ &self,
+ schema_name: Option<DatabaseName<'_>>,
+ pragma_name: &str,
+ f: F,
+ ) -> Result<T>
+ where
+ F: FnOnce(&Row<'_>) -> Result<T>,
+ {
+ let mut query = Sql::new();
+ query.push_pragma(schema_name, pragma_name)?;
+ self.query_row(&query, [], f)
+ }
+
+ /// Query the current rows/values of `pragma_name`.
+ ///
+ /// Prefer [PRAGMA function](https://sqlite.org/pragma.html#pragfunc) introduced in SQLite 3.20:
+ /// `SELECT * FROM pragma_collation_list;`
+ pub fn pragma_query<F>(
+ &self,
+ schema_name: Option<DatabaseName<'_>>,
+ pragma_name: &str,
+ mut f: F,
+ ) -> Result<()>
+ where
+ F: FnMut(&Row<'_>) -> Result<()>,
+ {
+ let mut query = Sql::new();
+ query.push_pragma(schema_name, pragma_name)?;
+ let mut stmt = self.prepare(&query)?;
+ let mut rows = stmt.query([])?;
+ while let Some(result_row) = rows.next()? {
+ let row = result_row;
+ f(row)?;
+ }
+ Ok(())
+ }
+
+ /// Query the current value(s) of `pragma_name` associated to
+ /// `pragma_value`.
+ ///
+ /// This method can be used with query-only pragmas which need an argument
+ /// (e.g. `table_info('one_tbl')`) or pragmas which returns value(s)
+ /// (e.g. `integrity_check`).
+ ///
+ /// Prefer [PRAGMA function](https://sqlite.org/pragma.html#pragfunc) introduced in SQLite 3.20:
+ /// `SELECT * FROM pragma_table_info(?1);`
+ pub fn pragma<F, V>(
+ &self,
+ schema_name: Option<DatabaseName<'_>>,
+ pragma_name: &str,
+ pragma_value: V,
+ mut f: F,
+ ) -> Result<()>
+ where
+ F: FnMut(&Row<'_>) -> Result<()>,
+ V: ToSql,
+ {
+ let mut sql = Sql::new();
+ sql.push_pragma(schema_name, pragma_name)?;
+ // The argument may be either in parentheses
+ // or it may be separated from the pragma name by an equal sign.
+ // The two syntaxes yield identical results.
+ sql.open_brace();
+ sql.push_value(&pragma_value)?;
+ sql.close_brace();
+ let mut stmt = self.prepare(&sql)?;
+ let mut rows = stmt.query([])?;
+ while let Some(result_row) = rows.next()? {
+ let row = result_row;
+ f(row)?;
+ }
+ Ok(())
+ }
+
+ /// Set a new value to `pragma_name`.
+ ///
+ /// Some pragmas will return the updated value which cannot be retrieved
+ /// with this method.
+ pub fn pragma_update<V>(
+ &self,
+ schema_name: Option<DatabaseName<'_>>,
+ pragma_name: &str,
+ pragma_value: V,
+ ) -> Result<()>
+ where
+ V: ToSql,
+ {
+ let mut sql = Sql::new();
+ sql.push_pragma(schema_name, pragma_name)?;
+ // The argument may be either in parentheses
+ // or it may be separated from the pragma name by an equal sign.
+ // The two syntaxes yield identical results.
+ sql.push_equal_sign();
+ sql.push_value(&pragma_value)?;
+ self.execute_batch(&sql)
+ }
+
+ /// Set a new value to `pragma_name` and return the updated value.
+ ///
+ /// Only few pragmas automatically return the updated value.
+ pub fn pragma_update_and_check<F, T, V>(
+ &self,
+ schema_name: Option<DatabaseName<'_>>,
+ pragma_name: &str,
+ pragma_value: V,
+ f: F,
+ ) -> Result<T>
+ where
+ F: FnOnce(&Row<'_>) -> Result<T>,
+ V: ToSql,
+ {
+ let mut sql = Sql::new();
+ sql.push_pragma(schema_name, pragma_name)?;
+ // The argument may be either in parentheses
+ // or it may be separated from the pragma name by an equal sign.
+ // The two syntaxes yield identical results.
+ sql.push_equal_sign();
+ sql.push_value(&pragma_value)?;
+ self.query_row(&sql, [], f)
+ }
+}
+
+fn is_identifier(s: &str) -> bool {
+ let chars = s.char_indices();
+ for (i, ch) in chars {
+ if i == 0 {
+ if !is_identifier_start(ch) {
+ return false;
+ }
+ } else if !is_identifier_continue(ch) {
+ return false;
+ }
+ }
+ true
+}
+
+fn is_identifier_start(c: char) -> bool {
+ c.is_ascii_uppercase() || c == '_' || c.is_ascii_lowercase() || c > '\x7F'
+}
+
+fn is_identifier_continue(c: char) -> bool {
+ c == '$'
+ || c.is_ascii_digit()
+ || c.is_ascii_uppercase()
+ || c == '_'
+ || c.is_ascii_lowercase()
+ || c > '\x7F'
+}
+
+#[cfg(test)]
+mod test {
+ use super::Sql;
+ use crate::pragma;
+ use crate::{Connection, DatabaseName, Result};
+
+ #[test]
+ fn pragma_query_value() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let user_version: i32 = db.pragma_query_value(None, "user_version", |row| row.get(0))?;
+ assert_eq!(0, user_version);
+ Ok(())
+ }
+
+ #[test]
+ #[cfg(feature = "modern_sqlite")]
+ fn pragma_func_query_value() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let user_version: i32 = db.one_column("SELECT user_version FROM pragma_user_version")?;
+ assert_eq!(0, user_version);
+ Ok(())
+ }
+
+ #[test]
+ fn pragma_query_no_schema() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let mut user_version = -1;
+ db.pragma_query(None, "user_version", |row| {
+ user_version = row.get(0)?;
+ Ok(())
+ })?;
+ assert_eq!(0, user_version);
+ Ok(())
+ }
+
+ #[test]
+ fn pragma_query_with_schema() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let mut user_version = -1;
+ db.pragma_query(Some(DatabaseName::Main), "user_version", |row| {
+ user_version = row.get(0)?;
+ Ok(())
+ })?;
+ assert_eq!(0, user_version);
+ Ok(())
+ }
+
+ #[test]
+ fn pragma() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let mut columns = Vec::new();
+ db.pragma(None, "table_info", "sqlite_master", |row| {
+ let column: String = row.get(1)?;
+ columns.push(column);
+ Ok(())
+ })?;
+ assert_eq!(5, columns.len());
+ Ok(())
+ }
+
+ #[test]
+ #[cfg(feature = "modern_sqlite")]
+ fn pragma_func() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let mut table_info = db.prepare("SELECT * FROM pragma_table_info(?1)")?;
+ let mut columns = Vec::new();
+ let mut rows = table_info.query(["sqlite_master"])?;
+
+ while let Some(row) = rows.next()? {
+ let column: String = row.get(1)?;
+ columns.push(column);
+ }
+ assert_eq!(5, columns.len());
+ Ok(())
+ }
+
+ #[test]
+ fn pragma_update() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.pragma_update(None, "user_version", 1)
+ }
+
+ #[test]
+ fn pragma_update_and_check() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let journal_mode: String =
+ db.pragma_update_and_check(None, "journal_mode", "OFF", |row| row.get(0))?;
+ assert!(
+ journal_mode == "off" || journal_mode == "memory",
+ "mode: {journal_mode:?}"
+ );
+ // Sanity checks to ensure the move to a generic `ToSql` wasn't breaking
+ let mode =
+ db.pragma_update_and_check(None, "journal_mode", "OFF", |row| row.get::<_, String>(0))?;
+ assert!(mode == "off" || mode == "memory", "mode: {mode:?}");
+
+ let param: &dyn crate::ToSql = &"OFF";
+ let mode =
+ db.pragma_update_and_check(None, "journal_mode", param, |row| row.get::<_, String>(0))?;
+ assert!(mode == "off" || mode == "memory", "mode: {mode:?}");
+ Ok(())
+ }
+
+ #[test]
+ fn is_identifier() {
+ assert!(pragma::is_identifier("full"));
+ assert!(pragma::is_identifier("r2d2"));
+ assert!(!pragma::is_identifier("sp ce"));
+ assert!(!pragma::is_identifier("semi;colon"));
+ }
+
+ #[test]
+ fn double_quote() {
+ let mut sql = Sql::new();
+ sql.push_schema_name(DatabaseName::Attached(r#"schema";--"#));
+ assert_eq!(r#""schema"";--""#, sql.as_str());
+ }
+
+ #[test]
+ fn wrap_and_escape() {
+ let mut sql = Sql::new();
+ sql.push_string_literal("value'; --");
+ assert_eq!("'value''; --'", sql.as_str());
+ }
+
+ #[test]
+ fn locking_mode() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let r = db.pragma_update(None, "locking_mode", "exclusive");
+ if cfg!(feature = "extra_check") {
+ r.unwrap_err();
+ } else {
+ r?;
+ }
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/raw_statement.rs b/third_party/rust/rusqlite/src/raw_statement.rs
new file mode 100644
index 0000000000..1683c7b857
--- /dev/null
+++ b/third_party/rust/rusqlite/src/raw_statement.rs
@@ -0,0 +1,240 @@
+use super::ffi;
+use super::StatementStatus;
+use crate::util::ParamIndexCache;
+use crate::util::SqliteMallocString;
+use std::ffi::CStr;
+use std::os::raw::c_int;
+use std::ptr;
+use std::sync::Arc;
+
+// Private newtype for raw sqlite3_stmts that finalize themselves when dropped.
+#[derive(Debug)]
+pub struct RawStatement {
+ ptr: *mut ffi::sqlite3_stmt,
+ tail: usize,
+ // Cached indices of named parameters, computed on the fly.
+ cache: ParamIndexCache,
+ // Cached SQL (trimmed) that we use as the key when we're in the statement
+ // cache. This is None for statements which didn't come from the statement
+ // cache.
+ //
+ // This is probably the same as `self.sql()` in most cases, but we don't
+ // care either way -- It's a better cache key as it is anyway since it's the
+ // actual source we got from rust.
+ //
+ // One example of a case where the result of `sqlite_sql` and the value in
+ // `statement_cache_key` might differ is if the statement has a `tail`.
+ statement_cache_key: Option<Arc<str>>,
+}
+
+impl RawStatement {
+ #[inline]
+ pub unsafe fn new(stmt: *mut ffi::sqlite3_stmt, tail: usize) -> RawStatement {
+ RawStatement {
+ ptr: stmt,
+ tail,
+ cache: ParamIndexCache::default(),
+ statement_cache_key: None,
+ }
+ }
+
+ #[inline]
+ pub fn is_null(&self) -> bool {
+ self.ptr.is_null()
+ }
+
+ #[inline]
+ pub(crate) fn set_statement_cache_key(&mut self, p: impl Into<Arc<str>>) {
+ self.statement_cache_key = Some(p.into());
+ }
+
+ #[inline]
+ pub(crate) fn statement_cache_key(&self) -> Option<Arc<str>> {
+ self.statement_cache_key.clone()
+ }
+
+ #[inline]
+ pub unsafe fn ptr(&self) -> *mut ffi::sqlite3_stmt {
+ self.ptr
+ }
+
+ #[inline]
+ pub fn column_count(&self) -> usize {
+ // Note: Can't cache this as it changes if the schema is altered.
+ unsafe { ffi::sqlite3_column_count(self.ptr) as usize }
+ }
+
+ #[inline]
+ pub fn column_type(&self, idx: usize) -> c_int {
+ unsafe { ffi::sqlite3_column_type(self.ptr, idx as c_int) }
+ }
+
+ #[inline]
+ #[cfg(feature = "column_decltype")]
+ pub fn column_decltype(&self, idx: usize) -> Option<&CStr> {
+ unsafe {
+ let decltype = ffi::sqlite3_column_decltype(self.ptr, idx as c_int);
+ if decltype.is_null() {
+ None
+ } else {
+ Some(CStr::from_ptr(decltype))
+ }
+ }
+ }
+
+ #[inline]
+ pub fn column_name(&self, idx: usize) -> Option<&CStr> {
+ let idx = idx as c_int;
+ if idx < 0 || idx >= self.column_count() as c_int {
+ return None;
+ }
+ unsafe {
+ let ptr = ffi::sqlite3_column_name(self.ptr, idx);
+ // If ptr is null here, it's an OOM, so there's probably nothing
+ // meaningful we can do. Just assert instead of returning None.
+ assert!(
+ !ptr.is_null(),
+ "Null pointer from sqlite3_column_name: Out of memory?"
+ );
+ Some(CStr::from_ptr(ptr))
+ }
+ }
+
+ #[inline]
+ #[cfg(not(feature = "unlock_notify"))]
+ pub fn step(&self) -> c_int {
+ unsafe { ffi::sqlite3_step(self.ptr) }
+ }
+
+ #[cfg(feature = "unlock_notify")]
+ pub fn step(&self) -> c_int {
+ use crate::unlock_notify;
+ let mut db = ptr::null_mut::<ffi::sqlite3>();
+ loop {
+ unsafe {
+ let mut rc = ffi::sqlite3_step(self.ptr);
+ // Bail out early for success and errors unrelated to locking. We
+ // still need check `is_locked` after this, but checking now lets us
+ // avoid one or two (admittedly cheap) calls into SQLite that we
+ // don't need to make.
+ if (rc & 0xff) != ffi::SQLITE_LOCKED {
+ break rc;
+ }
+ if db.is_null() {
+ db = ffi::sqlite3_db_handle(self.ptr);
+ }
+ if !unlock_notify::is_locked(db, rc) {
+ break rc;
+ }
+ rc = unlock_notify::wait_for_unlock_notify(db);
+ if rc != ffi::SQLITE_OK {
+ break rc;
+ }
+ self.reset();
+ }
+ }
+ }
+
+ #[inline]
+ pub fn reset(&self) -> c_int {
+ unsafe { ffi::sqlite3_reset(self.ptr) }
+ }
+
+ #[inline]
+ pub fn bind_parameter_count(&self) -> usize {
+ unsafe { ffi::sqlite3_bind_parameter_count(self.ptr) as usize }
+ }
+
+ #[inline]
+ pub fn bind_parameter_index(&self, name: &str) -> Option<usize> {
+ self.cache.get_or_insert_with(name, |param_cstr| {
+ let r = unsafe { ffi::sqlite3_bind_parameter_index(self.ptr, param_cstr.as_ptr()) };
+ match r {
+ 0 => None,
+ i => Some(i as usize),
+ }
+ })
+ }
+
+ #[inline]
+ pub fn bind_parameter_name(&self, index: i32) -> Option<&CStr> {
+ unsafe {
+ let name = ffi::sqlite3_bind_parameter_name(self.ptr, index);
+ if name.is_null() {
+ None
+ } else {
+ Some(CStr::from_ptr(name))
+ }
+ }
+ }
+
+ #[inline]
+ pub fn clear_bindings(&self) {
+ unsafe {
+ ffi::sqlite3_clear_bindings(self.ptr);
+ } // rc is always SQLITE_OK
+ }
+
+ #[inline]
+ pub fn sql(&self) -> Option<&CStr> {
+ if self.ptr.is_null() {
+ None
+ } else {
+ Some(unsafe { CStr::from_ptr(ffi::sqlite3_sql(self.ptr)) })
+ }
+ }
+
+ #[inline]
+ pub fn finalize(mut self) -> c_int {
+ self.finalize_()
+ }
+
+ #[inline]
+ fn finalize_(&mut self) -> c_int {
+ let r = unsafe { ffi::sqlite3_finalize(self.ptr) };
+ self.ptr = ptr::null_mut();
+ r
+ }
+
+ // does not work for PRAGMA
+ #[inline]
+ pub fn readonly(&self) -> bool {
+ unsafe { ffi::sqlite3_stmt_readonly(self.ptr) != 0 }
+ }
+
+ #[inline]
+ pub(crate) fn expanded_sql(&self) -> Option<SqliteMallocString> {
+ unsafe { SqliteMallocString::from_raw(ffi::sqlite3_expanded_sql(self.ptr)) }
+ }
+
+ #[inline]
+ pub fn get_status(&self, status: StatementStatus, reset: bool) -> i32 {
+ assert!(!self.ptr.is_null());
+ unsafe { ffi::sqlite3_stmt_status(self.ptr, status as i32, reset as i32) }
+ }
+
+ #[inline]
+ #[cfg(feature = "extra_check")]
+ pub fn has_tail(&self) -> bool {
+ self.tail != 0
+ }
+
+ #[inline]
+ pub fn tail(&self) -> usize {
+ self.tail
+ }
+
+ #[inline]
+ #[cfg(feature = "modern_sqlite")] // 3.28.0
+ pub fn is_explain(&self) -> i32 {
+ unsafe { ffi::sqlite3_stmt_isexplain(self.ptr) }
+ }
+
+ // TODO sqlite3_normalized_sql (https://sqlite.org/c3ref/expanded_sql.html) // 3.27.0 + SQLITE_ENABLE_NORMALIZE
+}
+
+impl Drop for RawStatement {
+ fn drop(&mut self) {
+ self.finalize_();
+ }
+}
diff --git a/third_party/rust/rusqlite/src/row.rs b/third_party/rust/rusqlite/src/row.rs
new file mode 100644
index 0000000000..2d2590040a
--- /dev/null
+++ b/third_party/rust/rusqlite/src/row.rs
@@ -0,0 +1,588 @@
+use fallible_iterator::FallibleIterator;
+use fallible_streaming_iterator::FallibleStreamingIterator;
+use std::convert;
+
+use super::{Error, Result, Statement};
+use crate::types::{FromSql, FromSqlError, ValueRef};
+
+/// An handle for the resulting rows of a query.
+#[must_use = "Rows is lazy and will do nothing unless consumed"]
+pub struct Rows<'stmt> {
+ pub(crate) stmt: Option<&'stmt Statement<'stmt>>,
+ row: Option<Row<'stmt>>,
+}
+
+impl<'stmt> Rows<'stmt> {
+ #[inline]
+ fn reset(&mut self) {
+ if let Some(stmt) = self.stmt.take() {
+ stmt.reset();
+ }
+ }
+
+ /// Attempt to get the next row from the query. Returns `Ok(Some(Row))` if
+ /// there is another row, `Err(...)` if there was an error
+ /// getting the next row, and `Ok(None)` if all rows have been retrieved.
+ ///
+ /// ## Note
+ ///
+ /// This interface is not compatible with Rust's `Iterator` trait, because
+ /// the lifetime of the returned row is tied to the lifetime of `self`.
+ /// This is a fallible "streaming iterator". For a more natural interface,
+ /// consider using [`query_map`](Statement::query_map) or
+ /// [`query_and_then`](Statement::query_and_then) instead, which
+ /// return types that implement `Iterator`.
+ #[allow(clippy::should_implement_trait)] // cannot implement Iterator
+ #[inline]
+ pub fn next(&mut self) -> Result<Option<&Row<'stmt>>> {
+ self.advance()?;
+ Ok((*self).get())
+ }
+
+ /// Map over this `Rows`, converting it to a [`Map`], which
+ /// implements `FallibleIterator`.
+ /// ```rust,no_run
+ /// use fallible_iterator::FallibleIterator;
+ /// # use rusqlite::{Result, Statement};
+ /// fn query(stmt: &mut Statement) -> Result<Vec<i64>> {
+ /// let rows = stmt.query([])?;
+ /// rows.map(|r| r.get(0)).collect()
+ /// }
+ /// ```
+ // FIXME Hide FallibleStreamingIterator::map
+ #[inline]
+ pub fn map<F, B>(self, f: F) -> Map<'stmt, F>
+ where
+ F: FnMut(&Row<'_>) -> Result<B>,
+ {
+ Map { rows: self, f }
+ }
+
+ /// Map over this `Rows`, converting it to a [`MappedRows`], which
+ /// implements `Iterator`.
+ #[inline]
+ pub fn mapped<F, B>(self, f: F) -> MappedRows<'stmt, F>
+ where
+ F: FnMut(&Row<'_>) -> Result<B>,
+ {
+ MappedRows { rows: self, map: f }
+ }
+
+ /// Map over this `Rows` with a fallible function, converting it to a
+ /// [`AndThenRows`], which implements `Iterator` (instead of
+ /// `FallibleStreamingIterator`).
+ #[inline]
+ pub fn and_then<F, T, E>(self, f: F) -> AndThenRows<'stmt, F>
+ where
+ F: FnMut(&Row<'_>) -> Result<T, E>,
+ {
+ AndThenRows { rows: self, map: f }
+ }
+
+ /// Give access to the underlying statement
+ #[must_use]
+ pub fn as_ref(&self) -> Option<&Statement<'stmt>> {
+ self.stmt
+ }
+}
+
+impl<'stmt> Rows<'stmt> {
+ #[inline]
+ pub(crate) fn new(stmt: &'stmt Statement<'stmt>) -> Rows<'stmt> {
+ Rows {
+ stmt: Some(stmt),
+ row: None,
+ }
+ }
+
+ #[inline]
+ pub(crate) fn get_expected_row(&mut self) -> Result<&Row<'stmt>> {
+ match self.next()? {
+ Some(row) => Ok(row),
+ None => Err(Error::QueryReturnedNoRows),
+ }
+ }
+}
+
+impl Drop for Rows<'_> {
+ #[inline]
+ fn drop(&mut self) {
+ self.reset();
+ }
+}
+
+/// `F` is used to transform the _streaming_ iterator into a _fallible_
+/// iterator.
+#[must_use = "iterators are lazy and do nothing unless consumed"]
+pub struct Map<'stmt, F> {
+ rows: Rows<'stmt>,
+ f: F,
+}
+
+impl<F, B> FallibleIterator for Map<'_, F>
+where
+ F: FnMut(&Row<'_>) -> Result<B>,
+{
+ type Error = Error;
+ type Item = B;
+
+ #[inline]
+ fn next(&mut self) -> Result<Option<B>> {
+ match self.rows.next()? {
+ Some(v) => Ok(Some((self.f)(v)?)),
+ None => Ok(None),
+ }
+ }
+}
+
+/// An iterator over the mapped resulting rows of a query.
+///
+/// `F` is used to transform the _streaming_ iterator into a _standard_
+/// iterator.
+#[must_use = "iterators are lazy and do nothing unless consumed"]
+pub struct MappedRows<'stmt, F> {
+ rows: Rows<'stmt>,
+ map: F,
+}
+
+impl<T, F> Iterator for MappedRows<'_, F>
+where
+ F: FnMut(&Row<'_>) -> Result<T>,
+{
+ type Item = Result<T>;
+
+ #[inline]
+ fn next(&mut self) -> Option<Result<T>> {
+ let map = &mut self.map;
+ self.rows
+ .next()
+ .transpose()
+ .map(|row_result| row_result.and_then(map))
+ }
+}
+
+/// An iterator over the mapped resulting rows of a query, with an Error type
+/// unifying with Error.
+#[must_use = "iterators are lazy and do nothing unless consumed"]
+pub struct AndThenRows<'stmt, F> {
+ rows: Rows<'stmt>,
+ map: F,
+}
+
+impl<T, E, F> Iterator for AndThenRows<'_, F>
+where
+ E: From<Error>,
+ F: FnMut(&Row<'_>) -> Result<T, E>,
+{
+ type Item = Result<T, E>;
+
+ #[inline]
+ fn next(&mut self) -> Option<Self::Item> {
+ let map = &mut self.map;
+ self.rows
+ .next()
+ .transpose()
+ .map(|row_result| row_result.map_err(E::from).and_then(map))
+ }
+}
+
+/// `FallibleStreamingIterator` differs from the standard library's `Iterator`
+/// in two ways:
+/// * each call to `next` (`sqlite3_step`) can fail.
+/// * returned `Row` is valid until `next` is called again or `Statement` is
+/// reset or finalized.
+///
+/// While these iterators cannot be used with Rust `for` loops, `while let`
+/// loops offer a similar level of ergonomics:
+/// ```rust,no_run
+/// # use rusqlite::{Result, Statement};
+/// fn query(stmt: &mut Statement) -> Result<()> {
+/// let mut rows = stmt.query([])?;
+/// while let Some(row) = rows.next()? {
+/// // scan columns value
+/// }
+/// Ok(())
+/// }
+/// ```
+impl<'stmt> FallibleStreamingIterator for Rows<'stmt> {
+ type Error = Error;
+ type Item = Row<'stmt>;
+
+ #[inline]
+ fn advance(&mut self) -> Result<()> {
+ if let Some(stmt) = self.stmt {
+ match stmt.step() {
+ Ok(true) => {
+ self.row = Some(Row { stmt });
+ Ok(())
+ }
+ Ok(false) => {
+ self.reset();
+ self.row = None;
+ Ok(())
+ }
+ Err(e) => {
+ self.reset();
+ self.row = None;
+ Err(e)
+ }
+ }
+ } else {
+ self.row = None;
+ Ok(())
+ }
+ }
+
+ #[inline]
+ fn get(&self) -> Option<&Row<'stmt>> {
+ self.row.as_ref()
+ }
+}
+
+/// A single result row of a query.
+pub struct Row<'stmt> {
+ pub(crate) stmt: &'stmt Statement<'stmt>,
+}
+
+impl<'stmt> Row<'stmt> {
+ /// Get the value of a particular column of the result row.
+ ///
+ /// # Panics
+ ///
+ /// Panics if calling [`row.get(idx)`](Row::get) would return an error,
+ /// including:
+ ///
+ /// * If the underlying SQLite column type is not a valid type as a source
+ /// for `T`
+ /// * If the underlying SQLite integral value is outside the range
+ /// representable by `T`
+ /// * If `idx` is outside the range of columns in the returned query
+ #[track_caller]
+ pub fn get_unwrap<I: RowIndex, T: FromSql>(&self, idx: I) -> T {
+ self.get(idx).unwrap()
+ }
+
+ /// Get the value of a particular column of the result row.
+ ///
+ /// ## Failure
+ ///
+ /// Returns an `Error::InvalidColumnType` if the underlying SQLite column
+ /// type is not a valid type as a source for `T`.
+ ///
+ /// Returns an `Error::InvalidColumnIndex` if `idx` is outside the valid
+ /// column range for this row.
+ ///
+ /// Returns an `Error::InvalidColumnName` if `idx` is not a valid column
+ /// name for this row.
+ ///
+ /// If the result type is i128 (which requires the `i128_blob` feature to be
+ /// enabled), and the underlying SQLite column is a blob whose size is not
+ /// 16 bytes, `Error::InvalidColumnType` will also be returned.
+ #[track_caller]
+ pub fn get<I: RowIndex, T: FromSql>(&self, idx: I) -> Result<T> {
+ let idx = idx.idx(self.stmt)?;
+ let value = self.stmt.value_ref(idx);
+ FromSql::column_result(value).map_err(|err| match err {
+ FromSqlError::InvalidType => Error::InvalidColumnType(
+ idx,
+ self.stmt.column_name_unwrap(idx).into(),
+ value.data_type(),
+ ),
+ FromSqlError::OutOfRange(i) => Error::IntegralValueOutOfRange(idx, i),
+ FromSqlError::Other(err) => {
+ Error::FromSqlConversionFailure(idx, value.data_type(), err)
+ }
+ FromSqlError::InvalidBlobSize { .. } => {
+ Error::FromSqlConversionFailure(idx, value.data_type(), Box::new(err))
+ }
+ })
+ }
+
+ /// Get the value of a particular column of the result row as a `ValueRef`,
+ /// allowing data to be read out of a row without copying.
+ ///
+ /// This `ValueRef` is valid only as long as this Row, which is enforced by
+ /// it's lifetime. This means that while this method is completely safe,
+ /// it can be somewhat difficult to use, and most callers will be better
+ /// served by [`get`](Row::get) or [`get_unwrap`](Row::get_unwrap).
+ ///
+ /// ## Failure
+ ///
+ /// Returns an `Error::InvalidColumnIndex` if `idx` is outside the valid
+ /// column range for this row.
+ ///
+ /// Returns an `Error::InvalidColumnName` if `idx` is not a valid column
+ /// name for this row.
+ pub fn get_ref<I: RowIndex>(&self, idx: I) -> Result<ValueRef<'_>> {
+ let idx = idx.idx(self.stmt)?;
+ // Narrowing from `ValueRef<'stmt>` (which `self.stmt.value_ref(idx)`
+ // returns) to `ValueRef<'a>` is needed because it's only valid until
+ // the next call to sqlite3_step.
+ let val_ref = self.stmt.value_ref(idx);
+ Ok(val_ref)
+ }
+
+ /// Get the value of a particular column of the result row as a `ValueRef`,
+ /// allowing data to be read out of a row without copying.
+ ///
+ /// This `ValueRef` is valid only as long as this Row, which is enforced by
+ /// it's lifetime. This means that while this method is completely safe,
+ /// it can be difficult to use, and most callers will be better served by
+ /// [`get`](Row::get) or [`get_unwrap`](Row::get_unwrap).
+ ///
+ /// # Panics
+ ///
+ /// Panics if calling [`row.get_ref(idx)`](Row::get_ref) would return an
+ /// error, including:
+ ///
+ /// * If `idx` is outside the range of columns in the returned query.
+ /// * If `idx` is not a valid column name for this row.
+ #[track_caller]
+ pub fn get_ref_unwrap<I: RowIndex>(&self, idx: I) -> ValueRef<'_> {
+ self.get_ref(idx).unwrap()
+ }
+}
+
+impl<'stmt> AsRef<Statement<'stmt>> for Row<'stmt> {
+ fn as_ref(&self) -> &Statement<'stmt> {
+ self.stmt
+ }
+}
+
+/// Debug `Row` like an ordered `Map<Result<&str>, Result<(Type, ValueRef)>>`
+/// with column name as key except that for `Type::Blob` only its size is
+/// printed (not its content).
+impl<'stmt> std::fmt::Debug for Row<'stmt> {
+ fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+ let mut dm = f.debug_map();
+ for c in 0..self.stmt.column_count() {
+ let name = self.stmt.column_name(c);
+ dm.key(&name);
+ let value = self.get_ref(c);
+ match value {
+ Ok(value) => {
+ let dt = value.data_type();
+ match value {
+ ValueRef::Null => {
+ dm.value(&(dt, ()));
+ }
+ ValueRef::Integer(i) => {
+ dm.value(&(dt, i));
+ }
+ ValueRef::Real(f) => {
+ dm.value(&(dt, f));
+ }
+ ValueRef::Text(s) => {
+ dm.value(&(dt, String::from_utf8_lossy(s)));
+ }
+ ValueRef::Blob(b) => {
+ dm.value(&(dt, b.len()));
+ }
+ }
+ }
+ Err(ref _err) => {
+ dm.value(&value);
+ }
+ }
+ }
+ dm.finish()
+ }
+}
+
+mod sealed {
+ /// This trait exists just to ensure that the only impls of `trait Params`
+ /// that are allowed are ones in this crate.
+ pub trait Sealed {}
+ impl Sealed for usize {}
+ impl Sealed for &str {}
+}
+
+/// A trait implemented by types that can index into columns of a row.
+///
+/// It is only implemented for `usize` and `&str`.
+pub trait RowIndex: sealed::Sealed {
+ /// Returns the index of the appropriate column, or `None` if no such
+ /// column exists.
+ fn idx(&self, stmt: &Statement<'_>) -> Result<usize>;
+}
+
+impl RowIndex for usize {
+ #[inline]
+ fn idx(&self, stmt: &Statement<'_>) -> Result<usize> {
+ if *self >= stmt.column_count() {
+ Err(Error::InvalidColumnIndex(*self))
+ } else {
+ Ok(*self)
+ }
+ }
+}
+
+impl RowIndex for &'_ str {
+ #[inline]
+ fn idx(&self, stmt: &Statement<'_>) -> Result<usize> {
+ stmt.column_index(self)
+ }
+}
+
+macro_rules! tuple_try_from_row {
+ ($($field:ident),*) => {
+ impl<'a, $($field,)*> convert::TryFrom<&'a Row<'a>> for ($($field,)*) where $($field: FromSql,)* {
+ type Error = crate::Error;
+
+ // we end with index += 1, which rustc warns about
+ // unused_variables and unused_mut are allowed for ()
+ #[allow(unused_assignments, unused_variables, unused_mut)]
+ fn try_from(row: &'a Row<'a>) -> Result<Self> {
+ let mut index = 0;
+ $(
+ #[allow(non_snake_case)]
+ let $field = row.get::<_, $field>(index)?;
+ index += 1;
+ )*
+ Ok(($($field,)*))
+ }
+ }
+ }
+}
+
+macro_rules! tuples_try_from_row {
+ () => {
+ // not very useful, but maybe some other macro users will find this helpful
+ tuple_try_from_row!();
+ };
+ ($first:ident $(, $remaining:ident)*) => {
+ tuple_try_from_row!($first $(, $remaining)*);
+ tuples_try_from_row!($($remaining),*);
+ };
+}
+
+tuples_try_from_row!(A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P);
+
+#[cfg(test)]
+mod tests {
+ #![allow(clippy::redundant_closure)] // false positives due to lifetime issues; clippy issue #5594
+ use crate::{Connection, Result};
+
+ #[test]
+ fn test_try_from_row_for_tuple_1() -> Result<()> {
+ use crate::ToSql;
+ use std::convert::TryFrom;
+
+ let conn = Connection::open_in_memory()?;
+ conn.execute(
+ "CREATE TABLE test (a INTEGER)",
+ crate::params_from_iter(std::iter::empty::<&dyn ToSql>()),
+ )?;
+ conn.execute("INSERT INTO test VALUES (42)", [])?;
+ let val = conn.query_row("SELECT a FROM test", [], |row| <(u32,)>::try_from(row))?;
+ assert_eq!(val, (42,));
+ let fail = conn.query_row("SELECT a FROM test", [], |row| <(u32, u32)>::try_from(row));
+ fail.unwrap_err();
+ Ok(())
+ }
+
+ #[test]
+ fn test_try_from_row_for_tuple_2() -> Result<()> {
+ use std::convert::TryFrom;
+
+ let conn = Connection::open_in_memory()?;
+ conn.execute("CREATE TABLE test (a INTEGER, b INTEGER)", [])?;
+ conn.execute("INSERT INTO test VALUES (42, 47)", [])?;
+ let val = conn.query_row("SELECT a, b FROM test", [], |row| {
+ <(u32, u32)>::try_from(row)
+ })?;
+ assert_eq!(val, (42, 47));
+ let fail = conn.query_row("SELECT a, b FROM test", [], |row| {
+ <(u32, u32, u32)>::try_from(row)
+ });
+ fail.unwrap_err();
+ Ok(())
+ }
+
+ #[test]
+ fn test_try_from_row_for_tuple_16() -> Result<()> {
+ use std::convert::TryFrom;
+
+ let create_table = "CREATE TABLE test (
+ a INTEGER,
+ b INTEGER,
+ c INTEGER,
+ d INTEGER,
+ e INTEGER,
+ f INTEGER,
+ g INTEGER,
+ h INTEGER,
+ i INTEGER,
+ j INTEGER,
+ k INTEGER,
+ l INTEGER,
+ m INTEGER,
+ n INTEGER,
+ o INTEGER,
+ p INTEGER
+ )";
+
+ let insert_values = "INSERT INTO test VALUES (
+ 0,
+ 1,
+ 2,
+ 3,
+ 4,
+ 5,
+ 6,
+ 7,
+ 8,
+ 9,
+ 10,
+ 11,
+ 12,
+ 13,
+ 14,
+ 15
+ )";
+
+ type BigTuple = (
+ u32,
+ u32,
+ u32,
+ u32,
+ u32,
+ u32,
+ u32,
+ u32,
+ u32,
+ u32,
+ u32,
+ u32,
+ u32,
+ u32,
+ u32,
+ u32,
+ );
+
+ let conn = Connection::open_in_memory()?;
+ conn.execute(create_table, [])?;
+ conn.execute(insert_values, [])?;
+ let val = conn.query_row("SELECT * FROM test", [], |row| BigTuple::try_from(row))?;
+ // Debug is not implemented for tuples of 16
+ assert_eq!(val.0, 0);
+ assert_eq!(val.1, 1);
+ assert_eq!(val.2, 2);
+ assert_eq!(val.3, 3);
+ assert_eq!(val.4, 4);
+ assert_eq!(val.5, 5);
+ assert_eq!(val.6, 6);
+ assert_eq!(val.7, 7);
+ assert_eq!(val.8, 8);
+ assert_eq!(val.9, 9);
+ assert_eq!(val.10, 10);
+ assert_eq!(val.11, 11);
+ assert_eq!(val.12, 12);
+ assert_eq!(val.13, 13);
+ assert_eq!(val.14, 14);
+ assert_eq!(val.15, 15);
+
+ // We don't test one bigger because it's unimplemented
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/serialize.rs b/third_party/rust/rusqlite/src/serialize.rs
new file mode 100644
index 0000000000..6852761b7c
--- /dev/null
+++ b/third_party/rust/rusqlite/src/serialize.rs
@@ -0,0 +1,167 @@
+//! Serialize a database.
+use std::convert::TryInto;
+use std::marker::PhantomData;
+use std::ops::Deref;
+use std::ptr::NonNull;
+
+use crate::error::error_from_handle;
+use crate::ffi;
+use crate::{Connection, DatabaseName, Result};
+
+/// Shared (SQLITE_SERIALIZE_NOCOPY) serialized database
+pub struct SharedData<'conn> {
+ phantom: PhantomData<&'conn Connection>,
+ ptr: NonNull<u8>,
+ sz: usize,
+}
+
+/// Owned serialized database
+pub struct OwnedData {
+ ptr: NonNull<u8>,
+ sz: usize,
+}
+
+impl OwnedData {
+ /// # Safety
+ ///
+ /// Caller must be certain that `ptr` is allocated by `sqlite3_malloc`.
+ pub unsafe fn from_raw_nonnull(ptr: NonNull<u8>, sz: usize) -> Self {
+ Self { ptr, sz }
+ }
+
+ fn into_raw(self) -> (*mut u8, usize) {
+ let raw = (self.ptr.as_ptr(), self.sz);
+ std::mem::forget(self);
+ raw
+ }
+}
+
+impl Drop for OwnedData {
+ fn drop(&mut self) {
+ unsafe {
+ ffi::sqlite3_free(self.ptr.as_ptr().cast());
+ }
+ }
+}
+
+/// Serialized database
+pub enum Data<'conn> {
+ /// Shared (SQLITE_SERIALIZE_NOCOPY) serialized database
+ Shared(SharedData<'conn>),
+ /// Owned serialized database
+ Owned(OwnedData),
+}
+
+impl<'conn> Deref for Data<'conn> {
+ type Target = [u8];
+
+ fn deref(&self) -> &[u8] {
+ let (ptr, sz) = match self {
+ Data::Owned(OwnedData { ptr, sz }) => (ptr.as_ptr(), *sz),
+ Data::Shared(SharedData { ptr, sz, .. }) => (ptr.as_ptr(), *sz),
+ };
+ unsafe { std::slice::from_raw_parts(ptr, sz) }
+ }
+}
+
+impl Connection {
+ /// Serialize a database.
+ pub fn serialize(&self, schema: DatabaseName) -> Result<Data> {
+ let schema = schema.as_cstring()?;
+ let mut sz = 0;
+ let mut ptr: *mut u8 = unsafe {
+ ffi::sqlite3_serialize(
+ self.handle(),
+ schema.as_ptr(),
+ &mut sz,
+ ffi::SQLITE_SERIALIZE_NOCOPY,
+ )
+ };
+ Ok(if ptr.is_null() {
+ ptr = unsafe { ffi::sqlite3_serialize(self.handle(), schema.as_ptr(), &mut sz, 0) };
+ if ptr.is_null() {
+ return Err(unsafe { error_from_handle(self.handle(), ffi::SQLITE_NOMEM) });
+ }
+ Data::Owned(OwnedData {
+ ptr: NonNull::new(ptr).unwrap(),
+ sz: sz.try_into().unwrap(),
+ })
+ } else {
+ // shared buffer
+ Data::Shared(SharedData {
+ ptr: NonNull::new(ptr).unwrap(),
+ sz: sz.try_into().unwrap(),
+ phantom: PhantomData,
+ })
+ })
+ }
+
+ /// Deserialize a database.
+ pub fn deserialize(
+ &mut self,
+ schema: DatabaseName<'_>,
+ data: OwnedData,
+ read_only: bool,
+ ) -> Result<()> {
+ let schema = schema.as_cstring()?;
+ let (data, sz) = data.into_raw();
+ let sz = sz.try_into().unwrap();
+ let flags = if read_only {
+ ffi::SQLITE_DESERIALIZE_FREEONCLOSE | ffi::SQLITE_DESERIALIZE_READONLY
+ } else {
+ ffi::SQLITE_DESERIALIZE_FREEONCLOSE | ffi::SQLITE_DESERIALIZE_RESIZEABLE
+ };
+ let rc = unsafe {
+ ffi::sqlite3_deserialize(self.handle(), schema.as_ptr(), data, sz, sz, flags)
+ };
+ if rc != ffi::SQLITE_OK {
+ // TODO sqlite3_free(data) ?
+ return Err(unsafe { error_from_handle(self.handle(), rc) });
+ }
+ /* TODO
+ if let Some(mxSize) = mxSize {
+ unsafe {
+ ffi::sqlite3_file_control(
+ self.handle(),
+ schema.as_ptr(),
+ ffi::SQLITE_FCNTL_SIZE_LIMIT,
+ &mut mxSize,
+ )
+ };
+ }*/
+ Ok(())
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use super::*;
+ use crate::{Connection, DatabaseName, Result};
+
+ #[test]
+ fn serialize() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE x AS SELECT 'data'")?;
+ let data = db.serialize(DatabaseName::Main)?;
+ let Data::Owned(data) = data else {
+ panic!("expected OwnedData")
+ };
+ assert!(data.sz > 0);
+ Ok(())
+ }
+
+ #[test]
+ fn deserialize() -> Result<()> {
+ let src = Connection::open_in_memory()?;
+ src.execute_batch("CREATE TABLE x AS SELECT 'data'")?;
+ let data = src.serialize(DatabaseName::Main)?;
+ let Data::Owned(data) = data else {
+ panic!("expected OwnedData")
+ };
+
+ let mut dst = Connection::open_in_memory()?;
+ dst.deserialize(DatabaseName::Main, data, false)?;
+ dst.execute("DELETE FROM x", [])?;
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/session.rs b/third_party/rust/rusqlite/src/session.rs
new file mode 100644
index 0000000000..0169a1cc9d
--- /dev/null
+++ b/third_party/rust/rusqlite/src/session.rs
@@ -0,0 +1,933 @@
+//! [Session Extension](https://sqlite.org/sessionintro.html)
+#![allow(non_camel_case_types)]
+
+use std::ffi::CStr;
+use std::io::{Read, Write};
+use std::marker::PhantomData;
+use std::os::raw::{c_char, c_int, c_uchar, c_void};
+use std::panic::{catch_unwind, RefUnwindSafe};
+use std::ptr;
+use std::slice::{from_raw_parts, from_raw_parts_mut};
+
+use fallible_streaming_iterator::FallibleStreamingIterator;
+
+use crate::error::{check, error_from_sqlite_code};
+use crate::ffi;
+use crate::hooks::Action;
+use crate::types::ValueRef;
+use crate::{errmsg_to_string, str_to_cstring, Connection, DatabaseName, Result};
+
+// https://sqlite.org/session.html
+
+type Filter = Option<Box<dyn Fn(&str) -> bool>>;
+
+/// An instance of this object is a session that can be
+/// used to record changes to a database.
+pub struct Session<'conn> {
+ phantom: PhantomData<&'conn Connection>,
+ s: *mut ffi::sqlite3_session,
+ filter: Filter,
+}
+
+impl Session<'_> {
+ /// Create a new session object
+ #[inline]
+ pub fn new(db: &Connection) -> Result<Session<'_>> {
+ Session::new_with_name(db, DatabaseName::Main)
+ }
+
+ /// Create a new session object
+ #[inline]
+ pub fn new_with_name<'conn>(
+ db: &'conn Connection,
+ name: DatabaseName<'_>,
+ ) -> Result<Session<'conn>> {
+ let name = name.as_cstring()?;
+
+ let db = db.db.borrow_mut().db;
+
+ let mut s: *mut ffi::sqlite3_session = ptr::null_mut();
+ check(unsafe { ffi::sqlite3session_create(db, name.as_ptr(), &mut s) })?;
+
+ Ok(Session {
+ phantom: PhantomData,
+ s,
+ filter: None,
+ })
+ }
+
+ /// Set a table filter
+ pub fn table_filter<F>(&mut self, filter: Option<F>)
+ where
+ F: Fn(&str) -> bool + Send + RefUnwindSafe + 'static,
+ {
+ unsafe extern "C" fn call_boxed_closure<F>(
+ p_arg: *mut c_void,
+ tbl_str: *const c_char,
+ ) -> c_int
+ where
+ F: Fn(&str) -> bool + RefUnwindSafe,
+ {
+ use std::str;
+
+ let boxed_filter: *mut F = p_arg as *mut F;
+ let tbl_name = {
+ let c_slice = CStr::from_ptr(tbl_str).to_bytes();
+ str::from_utf8(c_slice)
+ };
+ c_int::from(
+ catch_unwind(|| (*boxed_filter)(tbl_name.expect("non-utf8 table name")))
+ .unwrap_or_default(),
+ )
+ }
+
+ match filter {
+ Some(filter) => {
+ let boxed_filter = Box::new(filter);
+ unsafe {
+ ffi::sqlite3session_table_filter(
+ self.s,
+ Some(call_boxed_closure::<F>),
+ &*boxed_filter as *const F as *mut _,
+ );
+ }
+ self.filter = Some(boxed_filter);
+ }
+ _ => {
+ unsafe { ffi::sqlite3session_table_filter(self.s, None, ptr::null_mut()) }
+ self.filter = None;
+ }
+ };
+ }
+
+ /// Attach a table. `None` means all tables.
+ pub fn attach(&mut self, table: Option<&str>) -> Result<()> {
+ let table = if let Some(table) = table {
+ Some(str_to_cstring(table)?)
+ } else {
+ None
+ };
+ let table = table.as_ref().map(|s| s.as_ptr()).unwrap_or(ptr::null());
+ check(unsafe { ffi::sqlite3session_attach(self.s, table) })
+ }
+
+ /// Generate a Changeset
+ pub fn changeset(&mut self) -> Result<Changeset> {
+ let mut n = 0;
+ let mut cs: *mut c_void = ptr::null_mut();
+ check(unsafe { ffi::sqlite3session_changeset(self.s, &mut n, &mut cs) })?;
+ Ok(Changeset { cs, n })
+ }
+
+ /// Write the set of changes represented by this session to `output`.
+ #[inline]
+ pub fn changeset_strm(&mut self, output: &mut dyn Write) -> Result<()> {
+ let output_ref = &output;
+ check(unsafe {
+ ffi::sqlite3session_changeset_strm(
+ self.s,
+ Some(x_output),
+ output_ref as *const &mut dyn Write as *mut c_void,
+ )
+ })
+ }
+
+ /// Generate a Patchset
+ #[inline]
+ pub fn patchset(&mut self) -> Result<Changeset> {
+ let mut n = 0;
+ let mut ps: *mut c_void = ptr::null_mut();
+ check(unsafe { ffi::sqlite3session_patchset(self.s, &mut n, &mut ps) })?;
+ // TODO Validate: same struct
+ Ok(Changeset { cs: ps, n })
+ }
+
+ /// Write the set of patches represented by this session to `output`.
+ #[inline]
+ pub fn patchset_strm(&mut self, output: &mut dyn Write) -> Result<()> {
+ let output_ref = &output;
+ check(unsafe {
+ ffi::sqlite3session_patchset_strm(
+ self.s,
+ Some(x_output),
+ output_ref as *const &mut dyn Write as *mut c_void,
+ )
+ })
+ }
+
+ /// Load the difference between tables.
+ pub fn diff(&mut self, from: DatabaseName<'_>, table: &str) -> Result<()> {
+ let from = from.as_cstring()?;
+ let table = str_to_cstring(table)?;
+ let table = table.as_ptr();
+ unsafe {
+ let mut errmsg = ptr::null_mut();
+ let r =
+ ffi::sqlite3session_diff(self.s, from.as_ptr(), table, &mut errmsg as *mut *mut _);
+ if r != ffi::SQLITE_OK {
+ let errmsg: *mut c_char = errmsg;
+ let message = errmsg_to_string(&*errmsg);
+ ffi::sqlite3_free(errmsg as *mut c_void);
+ return Err(error_from_sqlite_code(r, Some(message)));
+ }
+ }
+ Ok(())
+ }
+
+ /// Test if a changeset has recorded any changes
+ #[inline]
+ pub fn is_empty(&self) -> bool {
+ unsafe { ffi::sqlite3session_isempty(self.s) != 0 }
+ }
+
+ /// Query the current state of the session
+ #[inline]
+ pub fn is_enabled(&self) -> bool {
+ unsafe { ffi::sqlite3session_enable(self.s, -1) != 0 }
+ }
+
+ /// Enable or disable the recording of changes
+ #[inline]
+ pub fn set_enabled(&mut self, enabled: bool) {
+ unsafe {
+ ffi::sqlite3session_enable(self.s, c_int::from(enabled));
+ }
+ }
+
+ /// Query the current state of the indirect flag
+ #[inline]
+ pub fn is_indirect(&self) -> bool {
+ unsafe { ffi::sqlite3session_indirect(self.s, -1) != 0 }
+ }
+
+ /// Set or clear the indirect change flag
+ #[inline]
+ pub fn set_indirect(&mut self, indirect: bool) {
+ unsafe {
+ ffi::sqlite3session_indirect(self.s, c_int::from(indirect));
+ }
+ }
+}
+
+impl Drop for Session<'_> {
+ #[inline]
+ fn drop(&mut self) {
+ if self.filter.is_some() {
+ self.table_filter(None::<fn(&str) -> bool>);
+ }
+ unsafe { ffi::sqlite3session_delete(self.s) };
+ }
+}
+
+/// Invert a changeset
+#[inline]
+pub fn invert_strm(input: &mut dyn Read, output: &mut dyn Write) -> Result<()> {
+ let input_ref = &input;
+ let output_ref = &output;
+ check(unsafe {
+ ffi::sqlite3changeset_invert_strm(
+ Some(x_input),
+ input_ref as *const &mut dyn Read as *mut c_void,
+ Some(x_output),
+ output_ref as *const &mut dyn Write as *mut c_void,
+ )
+ })
+}
+
+/// Combine two changesets
+#[inline]
+pub fn concat_strm(
+ input_a: &mut dyn Read,
+ input_b: &mut dyn Read,
+ output: &mut dyn Write,
+) -> Result<()> {
+ let input_a_ref = &input_a;
+ let input_b_ref = &input_b;
+ let output_ref = &output;
+ check(unsafe {
+ ffi::sqlite3changeset_concat_strm(
+ Some(x_input),
+ input_a_ref as *const &mut dyn Read as *mut c_void,
+ Some(x_input),
+ input_b_ref as *const &mut dyn Read as *mut c_void,
+ Some(x_output),
+ output_ref as *const &mut dyn Write as *mut c_void,
+ )
+ })
+}
+
+/// Changeset or Patchset
+pub struct Changeset {
+ cs: *mut c_void,
+ n: c_int,
+}
+
+impl Changeset {
+ /// Invert a changeset
+ #[inline]
+ pub fn invert(&self) -> Result<Changeset> {
+ let mut n = 0;
+ let mut cs = ptr::null_mut();
+ check(unsafe {
+ ffi::sqlite3changeset_invert(self.n, self.cs, &mut n, &mut cs as *mut *mut _)
+ })?;
+ Ok(Changeset { cs, n })
+ }
+
+ /// Create an iterator to traverse a changeset
+ #[inline]
+ pub fn iter(&self) -> Result<ChangesetIter<'_>> {
+ let mut it = ptr::null_mut();
+ check(unsafe { ffi::sqlite3changeset_start(&mut it as *mut *mut _, self.n, self.cs) })?;
+ Ok(ChangesetIter {
+ phantom: PhantomData,
+ it,
+ item: None,
+ })
+ }
+
+ /// Concatenate two changeset objects
+ #[inline]
+ pub fn concat(a: &Changeset, b: &Changeset) -> Result<Changeset> {
+ let mut n = 0;
+ let mut cs = ptr::null_mut();
+ check(unsafe {
+ ffi::sqlite3changeset_concat(a.n, a.cs, b.n, b.cs, &mut n, &mut cs as *mut *mut _)
+ })?;
+ Ok(Changeset { cs, n })
+ }
+}
+
+impl Drop for Changeset {
+ #[inline]
+ fn drop(&mut self) {
+ unsafe {
+ ffi::sqlite3_free(self.cs);
+ }
+ }
+}
+
+/// Cursor for iterating over the elements of a changeset
+/// or patchset.
+pub struct ChangesetIter<'changeset> {
+ phantom: PhantomData<&'changeset Changeset>,
+ it: *mut ffi::sqlite3_changeset_iter,
+ item: Option<ChangesetItem>,
+}
+
+impl ChangesetIter<'_> {
+ /// Create an iterator on `input`
+ #[inline]
+ pub fn start_strm<'input>(input: &&'input mut dyn Read) -> Result<ChangesetIter<'input>> {
+ let mut it = ptr::null_mut();
+ check(unsafe {
+ ffi::sqlite3changeset_start_strm(
+ &mut it as *mut *mut _,
+ Some(x_input),
+ input as *const &mut dyn Read as *mut c_void,
+ )
+ })?;
+ Ok(ChangesetIter {
+ phantom: PhantomData,
+ it,
+ item: None,
+ })
+ }
+}
+
+impl FallibleStreamingIterator for ChangesetIter<'_> {
+ type Error = crate::error::Error;
+ type Item = ChangesetItem;
+
+ #[inline]
+ fn advance(&mut self) -> Result<()> {
+ let rc = unsafe { ffi::sqlite3changeset_next(self.it) };
+ match rc {
+ ffi::SQLITE_ROW => {
+ self.item = Some(ChangesetItem { it: self.it });
+ Ok(())
+ }
+ ffi::SQLITE_DONE => {
+ self.item = None;
+ Ok(())
+ }
+ code => Err(error_from_sqlite_code(code, None)),
+ }
+ }
+
+ #[inline]
+ fn get(&self) -> Option<&ChangesetItem> {
+ self.item.as_ref()
+ }
+}
+
+/// Operation
+pub struct Operation<'item> {
+ table_name: &'item str,
+ number_of_columns: i32,
+ code: Action,
+ indirect: bool,
+}
+
+impl Operation<'_> {
+ /// Returns the table name.
+ #[inline]
+ pub fn table_name(&self) -> &str {
+ self.table_name
+ }
+
+ /// Returns the number of columns in table
+ #[inline]
+ pub fn number_of_columns(&self) -> i32 {
+ self.number_of_columns
+ }
+
+ /// Returns the action code.
+ #[inline]
+ pub fn code(&self) -> Action {
+ self.code
+ }
+
+ /// Returns `true` for an 'indirect' change.
+ #[inline]
+ pub fn indirect(&self) -> bool {
+ self.indirect
+ }
+}
+
+impl Drop for ChangesetIter<'_> {
+ #[inline]
+ fn drop(&mut self) {
+ unsafe {
+ ffi::sqlite3changeset_finalize(self.it);
+ }
+ }
+}
+
+/// An item passed to a conflict-handler by
+/// [`Connection::apply`](Connection::apply), or an item generated by
+/// [`ChangesetIter::next`](ChangesetIter::next).
+// TODO enum ? Delete, Insert, Update, ...
+pub struct ChangesetItem {
+ it: *mut ffi::sqlite3_changeset_iter,
+}
+
+impl ChangesetItem {
+ /// Obtain conflicting row values
+ ///
+ /// May only be called with an `SQLITE_CHANGESET_DATA` or
+ /// `SQLITE_CHANGESET_CONFLICT` conflict handler callback.
+ #[inline]
+ pub fn conflict(&self, col: usize) -> Result<ValueRef<'_>> {
+ unsafe {
+ let mut p_value: *mut ffi::sqlite3_value = ptr::null_mut();
+ check(ffi::sqlite3changeset_conflict(
+ self.it,
+ col as i32,
+ &mut p_value,
+ ))?;
+ Ok(ValueRef::from_value(p_value))
+ }
+ }
+
+ /// Determine the number of foreign key constraint violations
+ ///
+ /// May only be called with an `SQLITE_CHANGESET_FOREIGN_KEY` conflict
+ /// handler callback.
+ #[inline]
+ pub fn fk_conflicts(&self) -> Result<i32> {
+ unsafe {
+ let mut p_out = 0;
+ check(ffi::sqlite3changeset_fk_conflicts(self.it, &mut p_out))?;
+ Ok(p_out)
+ }
+ }
+
+ /// Obtain new.* Values
+ ///
+ /// May only be called if the type of change is either `SQLITE_UPDATE` or
+ /// `SQLITE_INSERT`.
+ #[inline]
+ pub fn new_value(&self, col: usize) -> Result<ValueRef<'_>> {
+ unsafe {
+ let mut p_value: *mut ffi::sqlite3_value = ptr::null_mut();
+ check(ffi::sqlite3changeset_new(self.it, col as i32, &mut p_value))?;
+ Ok(ValueRef::from_value(p_value))
+ }
+ }
+
+ /// Obtain old.* Values
+ ///
+ /// May only be called if the type of change is either `SQLITE_DELETE` or
+ /// `SQLITE_UPDATE`.
+ #[inline]
+ pub fn old_value(&self, col: usize) -> Result<ValueRef<'_>> {
+ unsafe {
+ let mut p_value: *mut ffi::sqlite3_value = ptr::null_mut();
+ check(ffi::sqlite3changeset_old(self.it, col as i32, &mut p_value))?;
+ Ok(ValueRef::from_value(p_value))
+ }
+ }
+
+ /// Obtain the current operation
+ #[inline]
+ pub fn op(&self) -> Result<Operation<'_>> {
+ let mut number_of_columns = 0;
+ let mut code = 0;
+ let mut indirect = 0;
+ let tab = unsafe {
+ let mut pz_tab: *const c_char = ptr::null();
+ check(ffi::sqlite3changeset_op(
+ self.it,
+ &mut pz_tab,
+ &mut number_of_columns,
+ &mut code,
+ &mut indirect,
+ ))?;
+ CStr::from_ptr(pz_tab)
+ };
+ let table_name = tab.to_str()?;
+ Ok(Operation {
+ table_name,
+ number_of_columns,
+ code: Action::from(code),
+ indirect: indirect != 0,
+ })
+ }
+
+ /// Obtain the primary key definition of a table
+ #[inline]
+ pub fn pk(&self) -> Result<&[u8]> {
+ let mut number_of_columns = 0;
+ unsafe {
+ let mut pks: *mut c_uchar = ptr::null_mut();
+ check(ffi::sqlite3changeset_pk(
+ self.it,
+ &mut pks,
+ &mut number_of_columns,
+ ))?;
+ Ok(from_raw_parts(pks, number_of_columns as usize))
+ }
+ }
+}
+
+/// Used to combine two or more changesets or
+/// patchsets
+pub struct Changegroup {
+ cg: *mut ffi::sqlite3_changegroup,
+}
+
+impl Changegroup {
+ /// Create a new change group.
+ #[inline]
+ pub fn new() -> Result<Self> {
+ let mut cg = ptr::null_mut();
+ check(unsafe { ffi::sqlite3changegroup_new(&mut cg) })?;
+ Ok(Changegroup { cg })
+ }
+
+ /// Add a changeset
+ #[inline]
+ pub fn add(&mut self, cs: &Changeset) -> Result<()> {
+ check(unsafe { ffi::sqlite3changegroup_add(self.cg, cs.n, cs.cs) })
+ }
+
+ /// Add a changeset read from `input` to this change group.
+ #[inline]
+ pub fn add_stream(&mut self, input: &mut dyn Read) -> Result<()> {
+ let input_ref = &input;
+ check(unsafe {
+ ffi::sqlite3changegroup_add_strm(
+ self.cg,
+ Some(x_input),
+ input_ref as *const &mut dyn Read as *mut c_void,
+ )
+ })
+ }
+
+ /// Obtain a composite Changeset
+ #[inline]
+ pub fn output(&mut self) -> Result<Changeset> {
+ let mut n = 0;
+ let mut output: *mut c_void = ptr::null_mut();
+ check(unsafe { ffi::sqlite3changegroup_output(self.cg, &mut n, &mut output) })?;
+ Ok(Changeset { cs: output, n })
+ }
+
+ /// Write the combined set of changes to `output`.
+ #[inline]
+ pub fn output_strm(&mut self, output: &mut dyn Write) -> Result<()> {
+ let output_ref = &output;
+ check(unsafe {
+ ffi::sqlite3changegroup_output_strm(
+ self.cg,
+ Some(x_output),
+ output_ref as *const &mut dyn Write as *mut c_void,
+ )
+ })
+ }
+}
+
+impl Drop for Changegroup {
+ #[inline]
+ fn drop(&mut self) {
+ unsafe {
+ ffi::sqlite3changegroup_delete(self.cg);
+ }
+ }
+}
+
+impl Connection {
+ /// Apply a changeset to a database
+ pub fn apply<F, C>(&self, cs: &Changeset, filter: Option<F>, conflict: C) -> Result<()>
+ where
+ F: Fn(&str) -> bool + Send + RefUnwindSafe + 'static,
+ C: Fn(ConflictType, ChangesetItem) -> ConflictAction + Send + RefUnwindSafe + 'static,
+ {
+ let db = self.db.borrow_mut().db;
+
+ let filtered = filter.is_some();
+ let tuple = &mut (filter, conflict);
+ check(unsafe {
+ if filtered {
+ ffi::sqlite3changeset_apply(
+ db,
+ cs.n,
+ cs.cs,
+ Some(call_filter::<F, C>),
+ Some(call_conflict::<F, C>),
+ tuple as *mut (Option<F>, C) as *mut c_void,
+ )
+ } else {
+ ffi::sqlite3changeset_apply(
+ db,
+ cs.n,
+ cs.cs,
+ None,
+ Some(call_conflict::<F, C>),
+ tuple as *mut (Option<F>, C) as *mut c_void,
+ )
+ }
+ })
+ }
+
+ /// Apply a changeset to a database
+ pub fn apply_strm<F, C>(
+ &self,
+ input: &mut dyn Read,
+ filter: Option<F>,
+ conflict: C,
+ ) -> Result<()>
+ where
+ F: Fn(&str) -> bool + Send + RefUnwindSafe + 'static,
+ C: Fn(ConflictType, ChangesetItem) -> ConflictAction + Send + RefUnwindSafe + 'static,
+ {
+ let input_ref = &input;
+ let db = self.db.borrow_mut().db;
+
+ let filtered = filter.is_some();
+ let tuple = &mut (filter, conflict);
+ check(unsafe {
+ if filtered {
+ ffi::sqlite3changeset_apply_strm(
+ db,
+ Some(x_input),
+ input_ref as *const &mut dyn Read as *mut c_void,
+ Some(call_filter::<F, C>),
+ Some(call_conflict::<F, C>),
+ tuple as *mut (Option<F>, C) as *mut c_void,
+ )
+ } else {
+ ffi::sqlite3changeset_apply_strm(
+ db,
+ Some(x_input),
+ input_ref as *const &mut dyn Read as *mut c_void,
+ None,
+ Some(call_conflict::<F, C>),
+ tuple as *mut (Option<F>, C) as *mut c_void,
+ )
+ }
+ })
+ }
+}
+
+/// Constants passed to the conflict handler
+/// See [here](https://sqlite.org/session.html#SQLITE_CHANGESET_CONFLICT) for details.
+#[allow(missing_docs)]
+#[repr(i32)]
+#[derive(Debug, PartialEq, Eq)]
+#[non_exhaustive]
+#[allow(clippy::upper_case_acronyms)]
+pub enum ConflictType {
+ UNKNOWN = -1,
+ SQLITE_CHANGESET_DATA = ffi::SQLITE_CHANGESET_DATA,
+ SQLITE_CHANGESET_NOTFOUND = ffi::SQLITE_CHANGESET_NOTFOUND,
+ SQLITE_CHANGESET_CONFLICT = ffi::SQLITE_CHANGESET_CONFLICT,
+ SQLITE_CHANGESET_CONSTRAINT = ffi::SQLITE_CHANGESET_CONSTRAINT,
+ SQLITE_CHANGESET_FOREIGN_KEY = ffi::SQLITE_CHANGESET_FOREIGN_KEY,
+}
+impl From<i32> for ConflictType {
+ fn from(code: i32) -> ConflictType {
+ match code {
+ ffi::SQLITE_CHANGESET_DATA => ConflictType::SQLITE_CHANGESET_DATA,
+ ffi::SQLITE_CHANGESET_NOTFOUND => ConflictType::SQLITE_CHANGESET_NOTFOUND,
+ ffi::SQLITE_CHANGESET_CONFLICT => ConflictType::SQLITE_CHANGESET_CONFLICT,
+ ffi::SQLITE_CHANGESET_CONSTRAINT => ConflictType::SQLITE_CHANGESET_CONSTRAINT,
+ ffi::SQLITE_CHANGESET_FOREIGN_KEY => ConflictType::SQLITE_CHANGESET_FOREIGN_KEY,
+ _ => ConflictType::UNKNOWN,
+ }
+ }
+}
+
+/// Constants returned by the conflict handler
+/// See [here](https://sqlite.org/session.html#SQLITE_CHANGESET_ABORT) for details.
+#[allow(missing_docs)]
+#[repr(i32)]
+#[derive(Debug, PartialEq, Eq)]
+#[non_exhaustive]
+#[allow(clippy::upper_case_acronyms)]
+pub enum ConflictAction {
+ SQLITE_CHANGESET_OMIT = ffi::SQLITE_CHANGESET_OMIT,
+ SQLITE_CHANGESET_REPLACE = ffi::SQLITE_CHANGESET_REPLACE,
+ SQLITE_CHANGESET_ABORT = ffi::SQLITE_CHANGESET_ABORT,
+}
+
+unsafe extern "C" fn call_filter<F, C>(p_ctx: *mut c_void, tbl_str: *const c_char) -> c_int
+where
+ F: Fn(&str) -> bool + Send + RefUnwindSafe + 'static,
+ C: Fn(ConflictType, ChangesetItem) -> ConflictAction + Send + RefUnwindSafe + 'static,
+{
+ use std::str;
+
+ let tuple: *mut (Option<F>, C) = p_ctx as *mut (Option<F>, C);
+ let tbl_name = {
+ let c_slice = CStr::from_ptr(tbl_str).to_bytes();
+ str::from_utf8(c_slice)
+ };
+ match *tuple {
+ (Some(ref filter), _) => c_int::from(
+ catch_unwind(|| filter(tbl_name.expect("illegal table name"))).unwrap_or_default(),
+ ),
+ _ => unimplemented!(),
+ }
+}
+
+unsafe extern "C" fn call_conflict<F, C>(
+ p_ctx: *mut c_void,
+ e_conflict: c_int,
+ p: *mut ffi::sqlite3_changeset_iter,
+) -> c_int
+where
+ F: Fn(&str) -> bool + Send + RefUnwindSafe + 'static,
+ C: Fn(ConflictType, ChangesetItem) -> ConflictAction + Send + RefUnwindSafe + 'static,
+{
+ let tuple: *mut (Option<F>, C) = p_ctx as *mut (Option<F>, C);
+ let conflict_type = ConflictType::from(e_conflict);
+ let item = ChangesetItem { it: p };
+ if let Ok(action) = catch_unwind(|| (*tuple).1(conflict_type, item)) {
+ action as c_int
+ } else {
+ ffi::SQLITE_CHANGESET_ABORT
+ }
+}
+
+unsafe extern "C" fn x_input(p_in: *mut c_void, data: *mut c_void, len: *mut c_int) -> c_int {
+ if p_in.is_null() {
+ return ffi::SQLITE_MISUSE;
+ }
+ let bytes: &mut [u8] = from_raw_parts_mut(data as *mut u8, *len as usize);
+ let input = p_in as *mut &mut dyn Read;
+ match (*input).read(bytes) {
+ Ok(n) => {
+ *len = n as i32; // TODO Validate: n = 0 may not mean the reader will always no longer be able to
+ // produce bytes.
+ ffi::SQLITE_OK
+ }
+ Err(_) => ffi::SQLITE_IOERR_READ, // TODO check if err is a (ru)sqlite Error => propagate
+ }
+}
+
+unsafe extern "C" fn x_output(p_out: *mut c_void, data: *const c_void, len: c_int) -> c_int {
+ if p_out.is_null() {
+ return ffi::SQLITE_MISUSE;
+ }
+ // The sessions module never invokes an xOutput callback with the third
+ // parameter set to a value less than or equal to zero.
+ let bytes: &[u8] = from_raw_parts(data as *const u8, len as usize);
+ let output = p_out as *mut &mut dyn Write;
+ match (*output).write_all(bytes) {
+ Ok(_) => ffi::SQLITE_OK,
+ Err(_) => ffi::SQLITE_IOERR_WRITE, // TODO check if err is a (ru)sqlite Error => propagate
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use fallible_streaming_iterator::FallibleStreamingIterator;
+ use std::io::Read;
+ use std::sync::atomic::{AtomicBool, Ordering};
+
+ use super::{Changeset, ChangesetIter, ConflictAction, ConflictType, Session};
+ use crate::hooks::Action;
+ use crate::{Connection, Result};
+
+ fn one_changeset() -> Result<Changeset> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(t TEXT PRIMARY KEY NOT NULL);")?;
+
+ let mut session = Session::new(&db)?;
+ assert!(session.is_empty());
+
+ session.attach(None)?;
+ db.execute("INSERT INTO foo (t) VALUES (?1);", ["bar"])?;
+
+ session.changeset()
+ }
+
+ fn one_changeset_strm() -> Result<Vec<u8>> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(t TEXT PRIMARY KEY NOT NULL);")?;
+
+ let mut session = Session::new(&db)?;
+ assert!(session.is_empty());
+
+ session.attach(None)?;
+ db.execute("INSERT INTO foo (t) VALUES (?1);", ["bar"])?;
+
+ let mut output = Vec::new();
+ session.changeset_strm(&mut output)?;
+ Ok(output)
+ }
+
+ #[test]
+ fn test_changeset() -> Result<()> {
+ let changeset = one_changeset()?;
+ let mut iter = changeset.iter()?;
+ let item = iter.next()?;
+ assert!(item.is_some());
+
+ let item = item.unwrap();
+ let op = item.op()?;
+ assert_eq!("foo", op.table_name());
+ assert_eq!(1, op.number_of_columns());
+ assert_eq!(Action::SQLITE_INSERT, op.code());
+ assert!(!op.indirect());
+
+ let pk = item.pk()?;
+ assert_eq!(&[1], pk);
+
+ let new_value = item.new_value(0)?;
+ assert_eq!(Ok("bar"), new_value.as_str());
+ Ok(())
+ }
+
+ #[test]
+ fn test_changeset_strm() -> Result<()> {
+ let output = one_changeset_strm()?;
+ assert!(!output.is_empty());
+ assert_eq!(14, output.len());
+
+ let input: &mut dyn Read = &mut output.as_slice();
+ let mut iter = ChangesetIter::start_strm(&input)?;
+ let item = iter.next()?;
+ assert!(item.is_some());
+ Ok(())
+ }
+
+ #[test]
+ fn test_changeset_apply() -> Result<()> {
+ let changeset = one_changeset()?;
+
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(t TEXT PRIMARY KEY NOT NULL);")?;
+
+ static CALLED: AtomicBool = AtomicBool::new(false);
+ db.apply(
+ &changeset,
+ None::<fn(&str) -> bool>,
+ |_conflict_type, _item| {
+ CALLED.store(true, Ordering::Relaxed);
+ ConflictAction::SQLITE_CHANGESET_OMIT
+ },
+ )?;
+
+ assert!(!CALLED.load(Ordering::Relaxed));
+ let check = db.query_row("SELECT 1 FROM foo WHERE t = ?1", ["bar"], |row| {
+ row.get::<_, i32>(0)
+ })?;
+ assert_eq!(1, check);
+
+ // conflict expected when same changeset applied again on the same db
+ db.apply(
+ &changeset,
+ None::<fn(&str) -> bool>,
+ |conflict_type, item| {
+ CALLED.store(true, Ordering::Relaxed);
+ assert_eq!(ConflictType::SQLITE_CHANGESET_CONFLICT, conflict_type);
+ let conflict = item.conflict(0).unwrap();
+ assert_eq!(Ok("bar"), conflict.as_str());
+ ConflictAction::SQLITE_CHANGESET_OMIT
+ },
+ )?;
+ assert!(CALLED.load(Ordering::Relaxed));
+ Ok(())
+ }
+
+ #[test]
+ fn test_changeset_apply_strm() -> Result<()> {
+ let output = one_changeset_strm()?;
+
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(t TEXT PRIMARY KEY NOT NULL);")?;
+
+ let mut input = output.as_slice();
+ db.apply_strm(
+ &mut input,
+ None::<fn(&str) -> bool>,
+ |_conflict_type, _item| ConflictAction::SQLITE_CHANGESET_OMIT,
+ )?;
+
+ let check = db.query_row("SELECT 1 FROM foo WHERE t = ?1", ["bar"], |row| {
+ row.get::<_, i32>(0)
+ })?;
+ assert_eq!(1, check);
+ Ok(())
+ }
+
+ #[test]
+ fn test_session_empty() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(t TEXT PRIMARY KEY NOT NULL);")?;
+
+ let mut session = Session::new(&db)?;
+ assert!(session.is_empty());
+
+ session.attach(None)?;
+ db.execute("INSERT INTO foo (t) VALUES (?1);", ["bar"])?;
+
+ assert!(!session.is_empty());
+ Ok(())
+ }
+
+ #[test]
+ fn test_session_set_enabled() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ let mut session = Session::new(&db)?;
+ assert!(session.is_enabled());
+ session.set_enabled(false);
+ assert!(!session.is_enabled());
+ Ok(())
+ }
+
+ #[test]
+ fn test_session_set_indirect() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ let mut session = Session::new(&db)?;
+ assert!(!session.is_indirect());
+ session.set_indirect(true);
+ assert!(session.is_indirect());
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/statement.rs b/third_party/rust/rusqlite/src/statement.rs
new file mode 100644
index 0000000000..d39cc1fe87
--- /dev/null
+++ b/third_party/rust/rusqlite/src/statement.rs
@@ -0,0 +1,1360 @@
+use std::iter::IntoIterator;
+use std::os::raw::{c_int, c_void};
+#[cfg(feature = "array")]
+use std::rc::Rc;
+use std::slice::from_raw_parts;
+use std::{fmt, mem, ptr, str};
+
+use super::ffi;
+use super::{len_as_c_int, str_for_sqlite};
+use super::{
+ AndThenRows, Connection, Error, MappedRows, Params, RawStatement, Result, Row, Rows, ValueRef,
+};
+use crate::types::{ToSql, ToSqlOutput};
+#[cfg(feature = "array")]
+use crate::vtab::array::{free_array, ARRAY_TYPE};
+
+/// A prepared statement.
+pub struct Statement<'conn> {
+ conn: &'conn Connection,
+ pub(crate) stmt: RawStatement,
+}
+
+impl Statement<'_> {
+ /// Execute the prepared statement.
+ ///
+ /// On success, returns the number of rows that were changed or inserted or
+ /// deleted (via `sqlite3_changes`).
+ ///
+ /// ## Example
+ ///
+ /// ### Use with positional parameters
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result, params};
+ /// fn update_rows(conn: &Connection) -> Result<()> {
+ /// let mut stmt = conn.prepare("UPDATE foo SET bar = ?1 WHERE qux = ?2")?;
+ /// // For a single parameter, or a parameter where all the values have
+ /// // the same type, just passing an array is simplest.
+ /// stmt.execute([2i32])?;
+ /// // The `rusqlite::params!` macro is mostly useful when the parameters do not
+ /// // all have the same type, or if there are more than 32 parameters
+ /// // at once, but it can be used in other cases.
+ /// stmt.execute(params![1i32])?;
+ /// // However, it's not required, many cases are fine as:
+ /// stmt.execute(&[&2i32])?;
+ /// // Or even:
+ /// stmt.execute([2i32])?;
+ /// // If you really want to, this is an option as well.
+ /// stmt.execute((2i32,))?;
+ /// Ok(())
+ /// }
+ /// ```
+ ///
+ /// #### Heterogeneous positional parameters
+ ///
+ /// ```
+ /// use rusqlite::{Connection, Result};
+ /// fn store_file(conn: &Connection, path: &str, data: &[u8]) -> Result<()> {
+ /// # // no need to do it for real.
+ /// # fn sha256(_: &[u8]) -> [u8; 32] { [0; 32] }
+ /// let query = "INSERT OR REPLACE INTO files(path, hash, data) VALUES (?1, ?2, ?3)";
+ /// let mut stmt = conn.prepare_cached(query)?;
+ /// let hash: [u8; 32] = sha256(data);
+ /// // The easiest way to pass positional parameters of have several
+ /// // different types is by using a tuple.
+ /// stmt.execute((path, hash, data))?;
+ /// // Using the `params!` macro also works, and supports longer parameter lists:
+ /// stmt.execute(rusqlite::params![path, hash, data])?;
+ /// Ok(())
+ /// }
+ /// # let c = Connection::open_in_memory().unwrap();
+ /// # c.execute_batch("CREATE TABLE files(path TEXT PRIMARY KEY, hash BLOB, data BLOB)").unwrap();
+ /// # store_file(&c, "foo/bar.txt", b"bibble").unwrap();
+ /// # store_file(&c, "foo/baz.txt", b"bobble").unwrap();
+ /// ```
+ ///
+ /// ### Use with named parameters
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result, named_params};
+ /// fn insert(conn: &Connection) -> Result<()> {
+ /// let mut stmt = conn.prepare("INSERT INTO test (key, value) VALUES (:key, :value)")?;
+ /// // The `rusqlite::named_params!` macro (like `params!`) is useful for heterogeneous
+ /// // sets of parameters (where all parameters are not the same type), or for queries
+ /// // with many (more than 32) statically known parameters.
+ /// stmt.execute(named_params! { ":key": "one", ":val": 2 })?;
+ /// // However, named parameters can also be passed like:
+ /// stmt.execute(&[(":key", "three"), (":val", "four")])?;
+ /// // Or even: (note that a &T is required for the value type, currently)
+ /// stmt.execute(&[(":key", &100), (":val", &200)])?;
+ /// Ok(())
+ /// }
+ /// ```
+ ///
+ /// ### Use without parameters
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result, params};
+ /// fn delete_all(conn: &Connection) -> Result<()> {
+ /// let mut stmt = conn.prepare("DELETE FROM users")?;
+ /// stmt.execute([])?;
+ /// Ok(())
+ /// }
+ /// ```
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if binding parameters fails, the executed statement
+ /// returns rows (in which case `query` should be used instead), or the
+ /// underlying SQLite call fails.
+ #[inline]
+ pub fn execute<P: Params>(&mut self, params: P) -> Result<usize> {
+ params.__bind_in(self)?;
+ self.execute_with_bound_parameters()
+ }
+
+ /// Execute an INSERT and return the ROWID.
+ ///
+ /// # Note
+ ///
+ /// This function is a convenience wrapper around
+ /// [`execute()`](Statement::execute) intended for queries that insert a
+ /// single item. It is possible to misuse this function in a way that it
+ /// cannot detect, such as by calling it on a statement which _updates_
+ /// a single item rather than inserting one. Please don't do that.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if no row is inserted or many rows are inserted.
+ #[inline]
+ pub fn insert<P: Params>(&mut self, params: P) -> Result<i64> {
+ let changes = self.execute(params)?;
+ match changes {
+ 1 => Ok(self.conn.last_insert_rowid()),
+ _ => Err(Error::StatementChangedRows(changes)),
+ }
+ }
+
+ /// Execute the prepared statement, returning a handle to the resulting
+ /// rows.
+ ///
+ /// Due to lifetime restrictions, the rows handle returned by `query` does
+ /// not implement the `Iterator` trait. Consider using
+ /// [`query_map`](Statement::query_map) or
+ /// [`query_and_then`](Statement::query_and_then) instead, which do.
+ ///
+ /// ## Example
+ ///
+ /// ### Use without parameters
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn get_names(conn: &Connection) -> Result<Vec<String>> {
+ /// let mut stmt = conn.prepare("SELECT name FROM people")?;
+ /// let mut rows = stmt.query([])?;
+ ///
+ /// let mut names = Vec::new();
+ /// while let Some(row) = rows.next()? {
+ /// names.push(row.get(0)?);
+ /// }
+ ///
+ /// Ok(names)
+ /// }
+ /// ```
+ ///
+ /// ### Use with positional parameters
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn query(conn: &Connection, name: &str) -> Result<()> {
+ /// let mut stmt = conn.prepare("SELECT * FROM test where name = ?1")?;
+ /// let mut rows = stmt.query(rusqlite::params![name])?;
+ /// while let Some(row) = rows.next()? {
+ /// // ...
+ /// }
+ /// Ok(())
+ /// }
+ /// ```
+ ///
+ /// Or, equivalently (but without the [`crate::params!`] macro).
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn query(conn: &Connection, name: &str) -> Result<()> {
+ /// let mut stmt = conn.prepare("SELECT * FROM test where name = ?1")?;
+ /// let mut rows = stmt.query([name])?;
+ /// while let Some(row) = rows.next()? {
+ /// // ...
+ /// }
+ /// Ok(())
+ /// }
+ /// ```
+ ///
+ /// ### Use with named parameters
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn query(conn: &Connection) -> Result<()> {
+ /// let mut stmt = conn.prepare("SELECT * FROM test where name = :name")?;
+ /// let mut rows = stmt.query(&[(":name", "one")])?;
+ /// while let Some(row) = rows.next()? {
+ /// // ...
+ /// }
+ /// Ok(())
+ /// }
+ /// ```
+ ///
+ /// Note, the `named_params!` macro is provided for syntactic convenience,
+ /// and so the above example could also be written as:
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result, named_params};
+ /// fn query(conn: &Connection) -> Result<()> {
+ /// let mut stmt = conn.prepare("SELECT * FROM test where name = :name")?;
+ /// let mut rows = stmt.query(named_params! { ":name": "one" })?;
+ /// while let Some(row) = rows.next()? {
+ /// // ...
+ /// }
+ /// Ok(())
+ /// }
+ /// ```
+ ///
+ /// ## Failure
+ ///
+ /// Will return `Err` if binding parameters fails.
+ #[inline]
+ pub fn query<P: Params>(&mut self, params: P) -> Result<Rows<'_>> {
+ params.__bind_in(self)?;
+ Ok(Rows::new(self))
+ }
+
+ /// Executes the prepared statement and maps a function over the resulting
+ /// rows, returning an iterator over the mapped function results.
+ ///
+ /// `f` is used to transform the _streaming_ iterator into a _standard_
+ /// iterator.
+ ///
+ /// This is equivalent to `stmt.query(params)?.mapped(f)`.
+ ///
+ /// ## Example
+ ///
+ /// ### Use with positional params
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn get_names(conn: &Connection) -> Result<Vec<String>> {
+ /// let mut stmt = conn.prepare("SELECT name FROM people")?;
+ /// let rows = stmt.query_map([], |row| row.get(0))?;
+ ///
+ /// let mut names = Vec::new();
+ /// for name_result in rows {
+ /// names.push(name_result?);
+ /// }
+ ///
+ /// Ok(names)
+ /// }
+ /// ```
+ ///
+ /// ### Use with named params
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn get_names(conn: &Connection) -> Result<Vec<String>> {
+ /// let mut stmt = conn.prepare("SELECT name FROM people WHERE id = :id")?;
+ /// let rows = stmt.query_map(&[(":id", &"one")], |row| row.get(0))?;
+ ///
+ /// let mut names = Vec::new();
+ /// for name_result in rows {
+ /// names.push(name_result?);
+ /// }
+ ///
+ /// Ok(names)
+ /// }
+ /// ```
+ /// ## Failure
+ ///
+ /// Will return `Err` if binding parameters fails.
+ pub fn query_map<T, P, F>(&mut self, params: P, f: F) -> Result<MappedRows<'_, F>>
+ where
+ P: Params,
+ F: FnMut(&Row<'_>) -> Result<T>,
+ {
+ self.query(params).map(|rows| rows.mapped(f))
+ }
+
+ /// Executes the prepared statement and maps a function over the resulting
+ /// rows, where the function returns a `Result` with `Error` type
+ /// implementing `std::convert::From<Error>` (so errors can be unified).
+ ///
+ /// This is equivalent to `stmt.query(params)?.and_then(f)`.
+ ///
+ /// ## Example
+ ///
+ /// ### Use with named params
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// struct Person {
+ /// name: String,
+ /// };
+ ///
+ /// fn name_to_person(name: String) -> Result<Person> {
+ /// // ... check for valid name
+ /// Ok(Person { name })
+ /// }
+ ///
+ /// fn get_names(conn: &Connection) -> Result<Vec<Person>> {
+ /// let mut stmt = conn.prepare("SELECT name FROM people WHERE id = :id")?;
+ /// let rows = stmt.query_and_then(&[(":id", "one")], |row| name_to_person(row.get(0)?))?;
+ ///
+ /// let mut persons = Vec::new();
+ /// for person_result in rows {
+ /// persons.push(person_result?);
+ /// }
+ ///
+ /// Ok(persons)
+ /// }
+ /// ```
+ ///
+ /// ### Use with positional params
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn get_names(conn: &Connection) -> Result<Vec<String>> {
+ /// let mut stmt = conn.prepare("SELECT name FROM people WHERE id = ?1")?;
+ /// let rows = stmt.query_and_then(["one"], |row| row.get::<_, String>(0))?;
+ ///
+ /// let mut persons = Vec::new();
+ /// for person_result in rows {
+ /// persons.push(person_result?);
+ /// }
+ ///
+ /// Ok(persons)
+ /// }
+ /// ```
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if binding parameters fails.
+ #[inline]
+ pub fn query_and_then<T, E, P, F>(&mut self, params: P, f: F) -> Result<AndThenRows<'_, F>>
+ where
+ P: Params,
+ E: From<Error>,
+ F: FnMut(&Row<'_>) -> Result<T, E>,
+ {
+ self.query(params).map(|rows| rows.and_then(f))
+ }
+
+ /// Return `true` if a query in the SQL statement it executes returns one
+ /// or more rows and `false` if the SQL returns an empty set.
+ #[inline]
+ pub fn exists<P: Params>(&mut self, params: P) -> Result<bool> {
+ let mut rows = self.query(params)?;
+ let exists = rows.next()?.is_some();
+ Ok(exists)
+ }
+
+ /// Convenience method to execute a query that is expected to return a
+ /// single row.
+ ///
+ /// If the query returns more than one row, all rows except the first are
+ /// ignored.
+ ///
+ /// Returns `Err(QueryReturnedNoRows)` if no results are returned. If the
+ /// query truly is optional, you can call
+ /// [`.optional()`](crate::OptionalExtension::optional) on the result of
+ /// this to get a `Result<Option<T>>` (requires that the trait
+ /// `rusqlite::OptionalExtension` is imported).
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite call fails.
+ pub fn query_row<T, P, F>(&mut self, params: P, f: F) -> Result<T>
+ where
+ P: Params,
+ F: FnOnce(&Row<'_>) -> Result<T>,
+ {
+ let mut rows = self.query(params)?;
+
+ rows.get_expected_row().and_then(f)
+ }
+
+ /// Consumes the statement.
+ ///
+ /// Functionally equivalent to the `Drop` implementation, but allows
+ /// callers to see any errors that occur.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite call fails.
+ #[inline]
+ pub fn finalize(mut self) -> Result<()> {
+ self.finalize_()
+ }
+
+ /// Return the (one-based) index of an SQL parameter given its name.
+ ///
+ /// Note that the initial ":" or "$" or "@" or "?" used to specify the
+ /// parameter is included as part of the name.
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn example(conn: &Connection) -> Result<()> {
+ /// let stmt = conn.prepare("SELECT * FROM test WHERE name = :example")?;
+ /// let index = stmt.parameter_index(":example")?;
+ /// assert_eq!(index, Some(1));
+ /// Ok(())
+ /// }
+ /// ```
+ ///
+ /// # Failure
+ ///
+ /// Will return Err if `name` is invalid. Will return Ok(None) if the name
+ /// is valid but not a bound parameter of this statement.
+ #[inline]
+ pub fn parameter_index(&self, name: &str) -> Result<Option<usize>> {
+ Ok(self.stmt.bind_parameter_index(name))
+ }
+
+ /// Return the SQL parameter name given its (one-based) index (the inverse
+ /// of [`Statement::parameter_index`]).
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn example(conn: &Connection) -> Result<()> {
+ /// let stmt = conn.prepare("SELECT * FROM test WHERE name = :example")?;
+ /// let index = stmt.parameter_name(1);
+ /// assert_eq!(index, Some(":example"));
+ /// Ok(())
+ /// }
+ /// ```
+ ///
+ /// # Failure
+ ///
+ /// Will return `None` if the column index is out of bounds or if the
+ /// parameter is positional.
+ ///
+ /// # Panics
+ ///
+ /// Panics when parameter name is not valid UTF-8.
+ #[inline]
+ pub fn parameter_name(&self, index: usize) -> Option<&'_ str> {
+ self.stmt.bind_parameter_name(index as i32).map(|name| {
+ str::from_utf8(name.to_bytes()).expect("Invalid UTF-8 sequence in parameter name")
+ })
+ }
+
+ #[inline]
+ pub(crate) fn bind_parameters<P>(&mut self, params: P) -> Result<()>
+ where
+ P: IntoIterator,
+ P::Item: ToSql,
+ {
+ let expected = self.stmt.bind_parameter_count();
+ let mut index = 0;
+ for p in params {
+ index += 1; // The leftmost SQL parameter has an index of 1.
+ if index > expected {
+ break;
+ }
+ self.bind_parameter(&p, index)?;
+ }
+ if index != expected {
+ Err(Error::InvalidParameterCount(index, expected))
+ } else {
+ Ok(())
+ }
+ }
+
+ #[inline]
+ pub(crate) fn ensure_parameter_count(&self, n: usize) -> Result<()> {
+ let count = self.parameter_count();
+ if count != n {
+ Err(Error::InvalidParameterCount(n, count))
+ } else {
+ Ok(())
+ }
+ }
+
+ #[inline]
+ pub(crate) fn bind_parameters_named<T: ?Sized + ToSql>(
+ &mut self,
+ params: &[(&str, &T)],
+ ) -> Result<()> {
+ for &(name, value) in params {
+ if let Some(i) = self.parameter_index(name)? {
+ let ts: &dyn ToSql = &value;
+ self.bind_parameter(ts, i)?;
+ } else {
+ return Err(Error::InvalidParameterName(name.into()));
+ }
+ }
+ Ok(())
+ }
+
+ /// Return the number of parameters that can be bound to this statement.
+ #[inline]
+ pub fn parameter_count(&self) -> usize {
+ self.stmt.bind_parameter_count()
+ }
+
+ /// Low level API to directly bind a parameter to a given index.
+ ///
+ /// Note that the index is one-based, that is, the first parameter index is
+ /// 1 and not 0. This is consistent with the SQLite API and the values given
+ /// to parameters bound as `?NNN`.
+ ///
+ /// The valid values for `one_based_col_index` begin at `1`, and end at
+ /// [`Statement::parameter_count`], inclusive.
+ ///
+ /// # Caveats
+ ///
+ /// This should not generally be used, but is available for special cases
+ /// such as:
+ ///
+ /// - binding parameters where a gap exists.
+ /// - binding named and positional parameters in the same query.
+ /// - separating parameter binding from query execution.
+ ///
+ /// In general, statements that have had *any* parameters bound this way
+ /// should have *all* parameters bound this way, and be queried or executed
+ /// by [`Statement::raw_query`] or [`Statement::raw_execute`], other usage
+ /// is unsupported and will likely, probably in surprising ways.
+ ///
+ /// That is: Do not mix the "raw" statement functions with the rest of the
+ /// API, or the results may be surprising, and may even change in future
+ /// versions without comment.
+ ///
+ /// # Example
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// fn query(conn: &Connection) -> Result<()> {
+ /// let mut stmt = conn.prepare("SELECT * FROM test WHERE name = :name AND value > ?2")?;
+ /// let name_index = stmt.parameter_index(":name")?.expect("No such parameter");
+ /// stmt.raw_bind_parameter(name_index, "foo")?;
+ /// stmt.raw_bind_parameter(2, 100)?;
+ /// let mut rows = stmt.raw_query();
+ /// while let Some(row) = rows.next()? {
+ /// // ...
+ /// }
+ /// Ok(())
+ /// }
+ /// ```
+ #[inline]
+ pub fn raw_bind_parameter<T: ToSql>(
+ &mut self,
+ one_based_col_index: usize,
+ param: T,
+ ) -> Result<()> {
+ // This is the same as `bind_parameter` but slightly more ergonomic and
+ // correctly takes `&mut self`.
+ self.bind_parameter(&param, one_based_col_index)
+ }
+
+ /// Low level API to execute a statement given that all parameters were
+ /// bound explicitly with the [`Statement::raw_bind_parameter`] API.
+ ///
+ /// # Caveats
+ ///
+ /// Any unbound parameters will have `NULL` as their value.
+ ///
+ /// This should not generally be used outside of special cases, and
+ /// functions in the [`Statement::execute`] family should be preferred.
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the executed statement returns rows (in which case
+ /// `query` should be used instead), or the underlying SQLite call fails.
+ #[inline]
+ pub fn raw_execute(&mut self) -> Result<usize> {
+ self.execute_with_bound_parameters()
+ }
+
+ /// Low level API to get `Rows` for this query given that all parameters
+ /// were bound explicitly with the [`Statement::raw_bind_parameter`] API.
+ ///
+ /// # Caveats
+ ///
+ /// Any unbound parameters will have `NULL` as their value.
+ ///
+ /// This should not generally be used outside of special cases, and
+ /// functions in the [`Statement::query`] family should be preferred.
+ ///
+ /// Note that if the SQL does not return results, [`Statement::raw_execute`]
+ /// should be used instead.
+ #[inline]
+ pub fn raw_query(&mut self) -> Rows<'_> {
+ Rows::new(self)
+ }
+
+ // generic because many of these branches can constant fold away.
+ fn bind_parameter<P: ?Sized + ToSql>(&self, param: &P, col: usize) -> Result<()> {
+ let value = param.to_sql()?;
+
+ let ptr = unsafe { self.stmt.ptr() };
+ let value = match value {
+ ToSqlOutput::Borrowed(v) => v,
+ ToSqlOutput::Owned(ref v) => ValueRef::from(v),
+
+ #[cfg(feature = "blob")]
+ ToSqlOutput::ZeroBlob(len) => {
+ // TODO sqlite3_bind_zeroblob64 // 3.8.11
+ return self
+ .conn
+ .decode_result(unsafe { ffi::sqlite3_bind_zeroblob(ptr, col as c_int, len) });
+ }
+ #[cfg(feature = "array")]
+ ToSqlOutput::Array(a) => {
+ return self.conn.decode_result(unsafe {
+ ffi::sqlite3_bind_pointer(
+ ptr,
+ col as c_int,
+ Rc::into_raw(a) as *mut c_void,
+ ARRAY_TYPE,
+ Some(free_array),
+ )
+ });
+ }
+ };
+ self.conn.decode_result(match value {
+ ValueRef::Null => unsafe { ffi::sqlite3_bind_null(ptr, col as c_int) },
+ ValueRef::Integer(i) => unsafe { ffi::sqlite3_bind_int64(ptr, col as c_int, i) },
+ ValueRef::Real(r) => unsafe { ffi::sqlite3_bind_double(ptr, col as c_int, r) },
+ ValueRef::Text(s) => unsafe {
+ let (c_str, len, destructor) = str_for_sqlite(s)?;
+ // TODO sqlite3_bind_text64 // 3.8.7
+ ffi::sqlite3_bind_text(ptr, col as c_int, c_str, len, destructor)
+ },
+ ValueRef::Blob(b) => unsafe {
+ let length = len_as_c_int(b.len())?;
+ if length == 0 {
+ ffi::sqlite3_bind_zeroblob(ptr, col as c_int, 0)
+ } else {
+ // TODO sqlite3_bind_blob64 // 3.8.7
+ ffi::sqlite3_bind_blob(
+ ptr,
+ col as c_int,
+ b.as_ptr().cast::<c_void>(),
+ length,
+ ffi::SQLITE_TRANSIENT(),
+ )
+ }
+ },
+ })
+ }
+
+ #[inline]
+ fn execute_with_bound_parameters(&mut self) -> Result<usize> {
+ self.check_update()?;
+ let r = self.stmt.step();
+ self.stmt.reset();
+ match r {
+ ffi::SQLITE_DONE => Ok(self.conn.changes() as usize),
+ ffi::SQLITE_ROW => Err(Error::ExecuteReturnedResults),
+ _ => Err(self.conn.decode_result(r).unwrap_err()),
+ }
+ }
+
+ #[inline]
+ fn finalize_(&mut self) -> Result<()> {
+ let mut stmt = unsafe { RawStatement::new(ptr::null_mut(), 0) };
+ mem::swap(&mut stmt, &mut self.stmt);
+ self.conn.decode_result(stmt.finalize())
+ }
+
+ #[cfg(feature = "extra_check")]
+ #[inline]
+ fn check_update(&self) -> Result<()> {
+ // sqlite3_column_count works for DML but not for DDL (ie ALTER)
+ if self.column_count() > 0 && self.stmt.readonly() {
+ return Err(Error::ExecuteReturnedResults);
+ }
+ Ok(())
+ }
+
+ #[cfg(not(feature = "extra_check"))]
+ #[inline]
+ #[allow(clippy::unnecessary_wraps)]
+ fn check_update(&self) -> Result<()> {
+ Ok(())
+ }
+
+ /// Returns a string containing the SQL text of prepared statement with
+ /// bound parameters expanded.
+ pub fn expanded_sql(&self) -> Option<String> {
+ self.stmt
+ .expanded_sql()
+ .map(|s| s.to_string_lossy().to_string())
+ }
+
+ /// Get the value for one of the status counters for this statement.
+ #[inline]
+ pub fn get_status(&self, status: StatementStatus) -> i32 {
+ self.stmt.get_status(status, false)
+ }
+
+ /// Reset the value of one of the status counters for this statement,
+ #[inline]
+ /// returning the value it had before resetting.
+ pub fn reset_status(&self, status: StatementStatus) -> i32 {
+ self.stmt.get_status(status, true)
+ }
+
+ /// Returns 1 if the prepared statement is an EXPLAIN statement,
+ /// or 2 if the statement is an EXPLAIN QUERY PLAN,
+ /// or 0 if it is an ordinary statement or a NULL pointer.
+ #[inline]
+ #[cfg(feature = "modern_sqlite")] // 3.28.0
+ #[cfg_attr(docsrs, doc(cfg(feature = "modern_sqlite")))]
+ pub fn is_explain(&self) -> i32 {
+ self.stmt.is_explain()
+ }
+
+ /// Returns true if the statement is read only.
+ #[inline]
+ pub fn readonly(&self) -> bool {
+ self.stmt.readonly()
+ }
+
+ #[cfg(feature = "extra_check")]
+ #[inline]
+ pub(crate) fn check_no_tail(&self) -> Result<()> {
+ if self.stmt.has_tail() {
+ Err(Error::MultipleStatement)
+ } else {
+ Ok(())
+ }
+ }
+
+ #[cfg(not(feature = "extra_check"))]
+ #[inline]
+ #[allow(clippy::unnecessary_wraps)]
+ pub(crate) fn check_no_tail(&self) -> Result<()> {
+ Ok(())
+ }
+
+ /// Safety: This is unsafe, because using `sqlite3_stmt` after the
+ /// connection has closed is illegal, but `RawStatement` does not enforce
+ /// this, as it loses our protective `'conn` lifetime bound.
+ #[inline]
+ pub(crate) unsafe fn into_raw(mut self) -> RawStatement {
+ let mut stmt = RawStatement::new(ptr::null_mut(), 0);
+ mem::swap(&mut stmt, &mut self.stmt);
+ stmt
+ }
+
+ /// Reset all bindings
+ pub fn clear_bindings(&mut self) {
+ self.stmt.clear_bindings();
+ }
+}
+
+impl fmt::Debug for Statement<'_> {
+ fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+ let sql = if self.stmt.is_null() {
+ Ok("")
+ } else {
+ str::from_utf8(self.stmt.sql().unwrap().to_bytes())
+ };
+ f.debug_struct("Statement")
+ .field("conn", self.conn)
+ .field("stmt", &self.stmt)
+ .field("sql", &sql)
+ .finish()
+ }
+}
+
+impl Drop for Statement<'_> {
+ #[allow(unused_must_use)]
+ #[inline]
+ fn drop(&mut self) {
+ self.finalize_();
+ }
+}
+
+impl Statement<'_> {
+ #[inline]
+ pub(super) fn new(conn: &Connection, stmt: RawStatement) -> Statement<'_> {
+ Statement { conn, stmt }
+ }
+
+ pub(super) fn value_ref(&self, col: usize) -> ValueRef<'_> {
+ let raw = unsafe { self.stmt.ptr() };
+
+ match self.stmt.column_type(col) {
+ ffi::SQLITE_NULL => ValueRef::Null,
+ ffi::SQLITE_INTEGER => {
+ ValueRef::Integer(unsafe { ffi::sqlite3_column_int64(raw, col as c_int) })
+ }
+ ffi::SQLITE_FLOAT => {
+ ValueRef::Real(unsafe { ffi::sqlite3_column_double(raw, col as c_int) })
+ }
+ ffi::SQLITE_TEXT => {
+ let s = unsafe {
+ // Quoting from "Using SQLite" book:
+ // To avoid problems, an application should first extract the desired type using
+ // a sqlite3_column_xxx() function, and then call the
+ // appropriate sqlite3_column_bytes() function.
+ let text = ffi::sqlite3_column_text(raw, col as c_int);
+ let len = ffi::sqlite3_column_bytes(raw, col as c_int);
+ assert!(
+ !text.is_null(),
+ "unexpected SQLITE_TEXT column type with NULL data"
+ );
+ from_raw_parts(text.cast::<u8>(), len as usize)
+ };
+
+ ValueRef::Text(s)
+ }
+ ffi::SQLITE_BLOB => {
+ let (blob, len) = unsafe {
+ (
+ ffi::sqlite3_column_blob(raw, col as c_int),
+ ffi::sqlite3_column_bytes(raw, col as c_int),
+ )
+ };
+
+ assert!(
+ len >= 0,
+ "unexpected negative return from sqlite3_column_bytes"
+ );
+ if len > 0 {
+ assert!(
+ !blob.is_null(),
+ "unexpected SQLITE_BLOB column type with NULL data"
+ );
+ ValueRef::Blob(unsafe { from_raw_parts(blob.cast::<u8>(), len as usize) })
+ } else {
+ // The return value from sqlite3_column_blob() for a zero-length BLOB
+ // is a NULL pointer.
+ ValueRef::Blob(&[])
+ }
+ }
+ _ => unreachable!("sqlite3_column_type returned invalid value"),
+ }
+ }
+
+ #[inline]
+ pub(super) fn step(&self) -> Result<bool> {
+ match self.stmt.step() {
+ ffi::SQLITE_ROW => Ok(true),
+ ffi::SQLITE_DONE => Ok(false),
+ code => Err(self.conn.decode_result(code).unwrap_err()),
+ }
+ }
+
+ #[inline]
+ pub(super) fn reset(&self) -> c_int {
+ self.stmt.reset()
+ }
+}
+
+/// Prepared statement status counters.
+///
+/// See `https://www.sqlite.org/c3ref/c_stmtstatus_counter.html`
+/// for explanations of each.
+///
+/// Note that depending on your version of SQLite, all of these
+/// may not be available.
+#[repr(i32)]
+#[derive(Clone, Copy, PartialEq, Eq)]
+#[non_exhaustive]
+pub enum StatementStatus {
+ /// Equivalent to SQLITE_STMTSTATUS_FULLSCAN_STEP
+ FullscanStep = 1,
+ /// Equivalent to SQLITE_STMTSTATUS_SORT
+ Sort = 2,
+ /// Equivalent to SQLITE_STMTSTATUS_AUTOINDEX
+ AutoIndex = 3,
+ /// Equivalent to SQLITE_STMTSTATUS_VM_STEP
+ VmStep = 4,
+ /// Equivalent to SQLITE_STMTSTATUS_REPREPARE (3.20.0)
+ RePrepare = 5,
+ /// Equivalent to SQLITE_STMTSTATUS_RUN (3.20.0)
+ Run = 6,
+ /// Equivalent to SQLITE_STMTSTATUS_FILTER_MISS
+ FilterMiss = 7,
+ /// Equivalent to SQLITE_STMTSTATUS_FILTER_HIT
+ FilterHit = 8,
+ /// Equivalent to SQLITE_STMTSTATUS_MEMUSED (3.20.0)
+ MemUsed = 99,
+}
+
+#[cfg(test)]
+mod test {
+ use crate::types::ToSql;
+ use crate::{params_from_iter, Connection, Error, Result};
+
+ #[test]
+ fn test_execute_named() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(x INTEGER)")?;
+
+ assert_eq!(
+ db.execute("INSERT INTO foo(x) VALUES (:x)", &[(":x", &1i32)])?,
+ 1
+ );
+ assert_eq!(
+ db.execute("INSERT INTO foo(x) VALUES (:x)", &[(":x", &2i32)])?,
+ 1
+ );
+ assert_eq!(
+ db.execute(
+ "INSERT INTO foo(x) VALUES (:x)",
+ crate::named_params! {":x": 3i32}
+ )?,
+ 1
+ );
+
+ assert_eq!(
+ 6i32,
+ db.query_row::<i32, _, _>(
+ "SELECT SUM(x) FROM foo WHERE x > :x",
+ &[(":x", &0i32)],
+ |r| r.get(0)
+ )?
+ );
+ assert_eq!(
+ 5i32,
+ db.query_row::<i32, _, _>(
+ "SELECT SUM(x) FROM foo WHERE x > :x",
+ &[(":x", &1i32)],
+ |r| r.get(0)
+ )?
+ );
+ Ok(())
+ }
+
+ #[test]
+ fn test_stmt_execute_named() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "CREATE TABLE test (id INTEGER PRIMARY KEY NOT NULL, name TEXT NOT NULL, flag \
+ INTEGER)";
+ db.execute_batch(sql)?;
+
+ let mut stmt = db.prepare("INSERT INTO test (name) VALUES (:name)")?;
+ stmt.execute(&[(":name", &"one")])?;
+
+ let mut stmt = db.prepare("SELECT COUNT(*) FROM test WHERE name = :name")?;
+ assert_eq!(
+ 1i32,
+ stmt.query_row::<i32, _, _>(&[(":name", "one")], |r| r.get(0))?
+ );
+ Ok(())
+ }
+
+ #[test]
+ fn test_query_named() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = r#"
+ CREATE TABLE test (id INTEGER PRIMARY KEY NOT NULL, name TEXT NOT NULL, flag INTEGER);
+ INSERT INTO test(id, name) VALUES (1, "one");
+ "#;
+ db.execute_batch(sql)?;
+
+ let mut stmt = db.prepare("SELECT id FROM test where name = :name")?;
+ let mut rows = stmt.query(&[(":name", "one")])?;
+ let id: Result<i32> = rows.next()?.unwrap().get(0);
+ assert_eq!(Ok(1), id);
+ Ok(())
+ }
+
+ #[test]
+ fn test_query_map_named() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = r#"
+ CREATE TABLE test (id INTEGER PRIMARY KEY NOT NULL, name TEXT NOT NULL, flag INTEGER);
+ INSERT INTO test(id, name) VALUES (1, "one");
+ "#;
+ db.execute_batch(sql)?;
+
+ let mut stmt = db.prepare("SELECT id FROM test where name = :name")?;
+ let mut rows = stmt.query_map(&[(":name", "one")], |row| {
+ let id: Result<i32> = row.get(0);
+ id.map(|i| 2 * i)
+ })?;
+
+ let doubled_id: i32 = rows.next().unwrap()?;
+ assert_eq!(2, doubled_id);
+ Ok(())
+ }
+
+ #[test]
+ fn test_query_and_then_by_name() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = r#"
+ CREATE TABLE test (id INTEGER PRIMARY KEY NOT NULL, name TEXT NOT NULL, flag INTEGER);
+ INSERT INTO test(id, name) VALUES (1, "one");
+ INSERT INTO test(id, name) VALUES (2, "one");
+ "#;
+ db.execute_batch(sql)?;
+
+ let mut stmt = db.prepare("SELECT id FROM test where name = :name ORDER BY id ASC")?;
+ let mut rows = stmt.query_and_then(&[(":name", "one")], |row| {
+ let id: i32 = row.get(0)?;
+ if id == 1 {
+ Ok(id)
+ } else {
+ Err(Error::SqliteSingleThreadedMode)
+ }
+ })?;
+
+ // first row should be Ok
+ let doubled_id: i32 = rows.next().unwrap()?;
+ assert_eq!(1, doubled_id);
+
+ // second row should be Err
+ #[allow(clippy::match_wild_err_arm)]
+ match rows.next().unwrap() {
+ Ok(_) => panic!("invalid Ok"),
+ Err(Error::SqliteSingleThreadedMode) => (),
+ Err(_) => panic!("invalid Err"),
+ }
+ Ok(())
+ }
+
+ #[test]
+ fn test_unbound_parameters_are_null() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "CREATE TABLE test (x TEXT, y TEXT)";
+ db.execute_batch(sql)?;
+
+ let mut stmt = db.prepare("INSERT INTO test (x, y) VALUES (:x, :y)")?;
+ stmt.execute(&[(":x", &"one")])?;
+
+ let result: Option<String> = db.one_column("SELECT y FROM test WHERE x = 'one'")?;
+ assert!(result.is_none());
+ Ok(())
+ }
+
+ #[test]
+ fn test_raw_binding() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE test (name TEXT, value INTEGER)")?;
+ {
+ let mut stmt = db.prepare("INSERT INTO test (name, value) VALUES (:name, ?3)")?;
+
+ let name_idx = stmt.parameter_index(":name")?.unwrap();
+ stmt.raw_bind_parameter(name_idx, "example")?;
+ stmt.raw_bind_parameter(3, 50i32)?;
+ let n = stmt.raw_execute()?;
+ assert_eq!(n, 1);
+ }
+
+ {
+ let mut stmt = db.prepare("SELECT name, value FROM test WHERE value = ?2")?;
+ stmt.raw_bind_parameter(2, 50)?;
+ let mut rows = stmt.raw_query();
+ {
+ let row = rows.next()?.unwrap();
+ let name: String = row.get(0)?;
+ assert_eq!(name, "example");
+ let value: i32 = row.get(1)?;
+ assert_eq!(value, 50);
+ }
+ assert!(rows.next()?.is_none());
+ }
+
+ Ok(())
+ }
+
+ #[test]
+ fn test_unbound_parameters_are_reused() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "CREATE TABLE test (x TEXT, y TEXT)";
+ db.execute_batch(sql)?;
+
+ let mut stmt = db.prepare("INSERT INTO test (x, y) VALUES (:x, :y)")?;
+ stmt.execute(&[(":x", "one")])?;
+ stmt.execute(&[(":y", "two")])?;
+
+ let result: String = db.one_column("SELECT x FROM test WHERE y = 'two'")?;
+ assert_eq!(result, "one");
+ Ok(())
+ }
+
+ #[test]
+ fn test_insert() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo(x INTEGER UNIQUE)")?;
+ let mut stmt = db.prepare("INSERT OR IGNORE INTO foo (x) VALUES (?1)")?;
+ assert_eq!(stmt.insert([1i32])?, 1);
+ assert_eq!(stmt.insert([2i32])?, 2);
+ match stmt.insert([1i32]).unwrap_err() {
+ Error::StatementChangedRows(0) => (),
+ err => panic!("Unexpected error {err}"),
+ }
+ let mut multi = db.prepare("INSERT INTO foo (x) SELECT 3 UNION ALL SELECT 4")?;
+ match multi.insert([]).unwrap_err() {
+ Error::StatementChangedRows(2) => (),
+ err => panic!("Unexpected error {err}"),
+ }
+ Ok(())
+ }
+
+ #[test]
+ fn test_insert_different_tables() -> Result<()> {
+ // Test for https://github.com/rusqlite/rusqlite/issues/171
+ let db = Connection::open_in_memory()?;
+ db.execute_batch(
+ r"
+ CREATE TABLE foo(x INTEGER);
+ CREATE TABLE bar(x INTEGER);
+ ",
+ )?;
+
+ assert_eq!(db.prepare("INSERT INTO foo VALUES (10)")?.insert([])?, 1);
+ assert_eq!(db.prepare("INSERT INTO bar VALUES (10)")?.insert([])?, 1);
+ Ok(())
+ }
+
+ #[test]
+ fn test_exists() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER);
+ INSERT INTO foo VALUES(1);
+ INSERT INTO foo VALUES(2);
+ END;";
+ db.execute_batch(sql)?;
+ let mut stmt = db.prepare("SELECT 1 FROM foo WHERE x = ?1")?;
+ assert!(stmt.exists([1i32])?);
+ assert!(stmt.exists([2i32])?);
+ assert!(!stmt.exists([0i32])?);
+ Ok(())
+ }
+ #[test]
+ fn test_tuple_params() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let s = db.query_row("SELECT printf('[%s]', ?1)", ("abc",), |r| {
+ r.get::<_, String>(0)
+ })?;
+ assert_eq!(s, "[abc]");
+ let s = db.query_row(
+ "SELECT printf('%d %s %d', ?1, ?2, ?3)",
+ (1i32, "abc", 2i32),
+ |r| r.get::<_, String>(0),
+ )?;
+ assert_eq!(s, "1 abc 2");
+ let s = db.query_row(
+ "SELECT printf('%d %s %d %d', ?1, ?2, ?3, ?4)",
+ (1, "abc", 2i32, 4i64),
+ |r| r.get::<_, String>(0),
+ )?;
+ assert_eq!(s, "1 abc 2 4");
+ #[rustfmt::skip]
+ let bigtup = (
+ 0, "a", 1, "b", 2, "c", 3, "d",
+ 4, "e", 5, "f", 6, "g", 7, "h",
+ );
+ let query = "SELECT printf(
+ '%d %s | %d %s | %d %s | %d %s || %d %s | %d %s | %d %s | %d %s',
+ ?1, ?2, ?3, ?4,
+ ?5, ?6, ?7, ?8,
+ ?9, ?10, ?11, ?12,
+ ?13, ?14, ?15, ?16
+ )";
+ let s = db.query_row(query, bigtup, |r| r.get::<_, String>(0))?;
+ assert_eq!(s, "0 a | 1 b | 2 c | 3 d || 4 e | 5 f | 6 g | 7 h");
+ Ok(())
+ }
+
+ #[test]
+ fn test_query_row() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER, y INTEGER);
+ INSERT INTO foo VALUES(1, 3);
+ INSERT INTO foo VALUES(2, 4);
+ END;";
+ db.execute_batch(sql)?;
+ let mut stmt = db.prepare("SELECT y FROM foo WHERE x = ?1")?;
+ let y: Result<i64> = stmt.query_row([1i32], |r| r.get(0));
+ assert_eq!(3i64, y?);
+ Ok(())
+ }
+
+ #[test]
+ fn test_query_by_column_name() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER, y INTEGER);
+ INSERT INTO foo VALUES(1, 3);
+ END;";
+ db.execute_batch(sql)?;
+ let mut stmt = db.prepare("SELECT y FROM foo")?;
+ let y: Result<i64> = stmt.query_row([], |r| r.get("y"));
+ assert_eq!(3i64, y?);
+ Ok(())
+ }
+
+ #[test]
+ fn test_query_by_column_name_ignore_case() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let sql = "BEGIN;
+ CREATE TABLE foo(x INTEGER, y INTEGER);
+ INSERT INTO foo VALUES(1, 3);
+ END;";
+ db.execute_batch(sql)?;
+ let mut stmt = db.prepare("SELECT y as Y FROM foo")?;
+ let y: Result<i64> = stmt.query_row([], |r| r.get("y"));
+ assert_eq!(3i64, y?);
+ Ok(())
+ }
+
+ #[test]
+ fn test_expanded_sql() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let stmt = db.prepare("SELECT ?1")?;
+ stmt.bind_parameter(&1, 1)?;
+ assert_eq!(Some("SELECT 1".to_owned()), stmt.expanded_sql());
+ Ok(())
+ }
+
+ #[test]
+ fn test_bind_parameters() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ // dynamic slice:
+ db.query_row(
+ "SELECT ?1, ?2, ?3",
+ [&1u8 as &dyn ToSql, &"one", &Some("one")],
+ |row| row.get::<_, u8>(0),
+ )?;
+ // existing collection:
+ let data = vec![1, 2, 3];
+ db.query_row("SELECT ?1, ?2, ?3", params_from_iter(&data), |row| {
+ row.get::<_, u8>(0)
+ })?;
+ db.query_row(
+ "SELECT ?1, ?2, ?3",
+ params_from_iter(data.as_slice()),
+ |row| row.get::<_, u8>(0),
+ )?;
+ db.query_row("SELECT ?1, ?2, ?3", params_from_iter(data), |row| {
+ row.get::<_, u8>(0)
+ })?;
+
+ use std::collections::BTreeSet;
+ let data: BTreeSet<String> = ["one", "two", "three"]
+ .iter()
+ .map(|s| (*s).to_string())
+ .collect();
+ db.query_row("SELECT ?1, ?2, ?3", params_from_iter(&data), |row| {
+ row.get::<_, String>(0)
+ })?;
+
+ let data = [0; 3];
+ db.query_row("SELECT ?1, ?2, ?3", params_from_iter(&data), |row| {
+ row.get::<_, u8>(0)
+ })?;
+ db.query_row("SELECT ?1, ?2, ?3", params_from_iter(data.iter()), |row| {
+ row.get::<_, u8>(0)
+ })?;
+ Ok(())
+ }
+
+ #[test]
+ fn test_parameter_name() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE test (name TEXT, value INTEGER)")?;
+ let stmt = db.prepare("INSERT INTO test (name, value) VALUES (:name, ?3)")?;
+ assert_eq!(stmt.parameter_name(0), None);
+ assert_eq!(stmt.parameter_name(1), Some(":name"));
+ assert_eq!(stmt.parameter_name(2), None);
+ Ok(())
+ }
+
+ #[test]
+ fn test_empty_stmt() -> Result<()> {
+ let conn = Connection::open_in_memory()?;
+ let mut stmt = conn.prepare("")?;
+ assert_eq!(0, stmt.column_count());
+ stmt.parameter_index("test").unwrap();
+ stmt.step().unwrap_err();
+ stmt.reset();
+ stmt.execute([]).unwrap_err();
+ Ok(())
+ }
+
+ #[test]
+ fn test_comment_stmt() -> Result<()> {
+ let conn = Connection::open_in_memory()?;
+ conn.prepare("/*SELECT 1;*/")?;
+ Ok(())
+ }
+
+ #[test]
+ fn test_comment_and_sql_stmt() -> Result<()> {
+ let conn = Connection::open_in_memory()?;
+ let stmt = conn.prepare("/*...*/ SELECT 1;")?;
+ assert_eq!(1, stmt.column_count());
+ Ok(())
+ }
+
+ #[test]
+ fn test_semi_colon_stmt() -> Result<()> {
+ let conn = Connection::open_in_memory()?;
+ let stmt = conn.prepare(";")?;
+ assert_eq!(0, stmt.column_count());
+ Ok(())
+ }
+
+ #[test]
+ fn test_utf16_conversion() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ db.pragma_update(None, "encoding", "UTF-16le")?;
+ let encoding: String = db.pragma_query_value(None, "encoding", |row| row.get(0))?;
+ assert_eq!("UTF-16le", encoding);
+ db.execute_batch("CREATE TABLE foo(x TEXT)")?;
+ let expected = "テスト";
+ db.execute("INSERT INTO foo(x) VALUES (?1)", [&expected])?;
+ let actual: String = db.one_column("SELECT x FROM foo")?;
+ assert_eq!(expected, actual);
+ Ok(())
+ }
+
+ #[test]
+ fn test_nul_byte() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let expected = "a\x00b";
+ let actual: String = db.query_row("SELECT ?1", [expected], |row| row.get(0))?;
+ assert_eq!(expected, actual);
+ Ok(())
+ }
+
+ #[test]
+ #[cfg(feature = "modern_sqlite")]
+ fn is_explain() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let stmt = db.prepare("SELECT 1;")?;
+ assert_eq!(0, stmt.is_explain());
+ Ok(())
+ }
+
+ #[test]
+ fn readonly() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ let stmt = db.prepare("SELECT 1;")?;
+ assert!(stmt.readonly());
+ Ok(())
+ }
+
+ #[test]
+ #[cfg(feature = "modern_sqlite")] // SQLite >= 3.38.0
+ fn test_error_offset() -> Result<()> {
+ use crate::ffi::ErrorCode;
+ let db = Connection::open_in_memory()?;
+ let r = db.execute_batch("SELECT CURRENT_TIMESTANP;");
+ match r.unwrap_err() {
+ Error::SqlInputError { error, offset, .. } => {
+ assert_eq!(error.code, ErrorCode::Unknown);
+ assert_eq!(offset, 7);
+ }
+ err => panic!("Unexpected error {err}"),
+ }
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/trace.rs b/third_party/rust/rusqlite/src/trace.rs
new file mode 100644
index 0000000000..7317a0ca8e
--- /dev/null
+++ b/third_party/rust/rusqlite/src/trace.rs
@@ -0,0 +1,184 @@
+//! Tracing and profiling functions. Error and warning log.
+
+use std::ffi::{CStr, CString};
+use std::mem;
+use std::os::raw::{c_char, c_int, c_void};
+use std::panic::catch_unwind;
+use std::ptr;
+use std::time::Duration;
+
+use super::ffi;
+use crate::Connection;
+
+/// Set up the process-wide SQLite error logging callback.
+///
+/// # Safety
+///
+/// This function is marked unsafe for two reasons:
+///
+/// * The function is not threadsafe. No other SQLite calls may be made while
+/// `config_log` is running, and multiple threads may not call `config_log`
+/// simultaneously.
+/// * The provided `callback` itself function has two requirements:
+/// * It must not invoke any SQLite calls.
+/// * It must be threadsafe if SQLite is used in a multithreaded way.
+///
+/// cf [The Error And Warning Log](http://sqlite.org/errlog.html).
+#[cfg(not(feature = "loadable_extension"))]
+pub unsafe fn config_log(callback: Option<fn(c_int, &str)>) -> crate::Result<()> {
+ extern "C" fn log_callback(p_arg: *mut c_void, err: c_int, msg: *const c_char) {
+ let c_slice = unsafe { CStr::from_ptr(msg).to_bytes() };
+ let callback: fn(c_int, &str) = unsafe { mem::transmute(p_arg) };
+
+ let s = String::from_utf8_lossy(c_slice);
+ drop(catch_unwind(|| callback(err, &s)));
+ }
+
+ let rc = if let Some(f) = callback {
+ ffi::sqlite3_config(
+ ffi::SQLITE_CONFIG_LOG,
+ log_callback as extern "C" fn(_, _, _),
+ f as *mut c_void,
+ )
+ } else {
+ let nullptr: *mut c_void = ptr::null_mut();
+ ffi::sqlite3_config(ffi::SQLITE_CONFIG_LOG, nullptr, nullptr)
+ };
+
+ if rc == ffi::SQLITE_OK {
+ Ok(())
+ } else {
+ Err(crate::error::error_from_sqlite_code(rc, None))
+ }
+}
+
+/// Write a message into the error log established by
+/// `config_log`.
+#[inline]
+pub fn log(err_code: c_int, msg: &str) {
+ let msg = CString::new(msg).expect("SQLite log messages cannot contain embedded zeroes");
+ unsafe {
+ ffi::sqlite3_log(err_code, b"%s\0" as *const _ as *const c_char, msg.as_ptr());
+ }
+}
+
+impl Connection {
+ /// Register or clear a callback function that can be
+ /// used for tracing the execution of SQL statements.
+ ///
+ /// Prepared statement placeholders are replaced/logged with their assigned
+ /// values. There can only be a single tracer defined for each database
+ /// connection. Setting a new tracer clears the old one.
+ pub fn trace(&mut self, trace_fn: Option<fn(&str)>) {
+ unsafe extern "C" fn trace_callback(p_arg: *mut c_void, z_sql: *const c_char) {
+ let trace_fn: fn(&str) = mem::transmute(p_arg);
+ let c_slice = CStr::from_ptr(z_sql).to_bytes();
+ let s = String::from_utf8_lossy(c_slice);
+ drop(catch_unwind(|| trace_fn(&s)));
+ }
+
+ let c = self.db.borrow_mut();
+ match trace_fn {
+ Some(f) => unsafe {
+ ffi::sqlite3_trace(c.db(), Some(trace_callback), f as *mut c_void);
+ },
+ None => unsafe {
+ ffi::sqlite3_trace(c.db(), None, ptr::null_mut());
+ },
+ }
+ }
+
+ /// Register or clear a callback function that can be
+ /// used for profiling the execution of SQL statements.
+ ///
+ /// There can only be a single profiler defined for each database
+ /// connection. Setting a new profiler clears the old one.
+ pub fn profile(&mut self, profile_fn: Option<fn(&str, Duration)>) {
+ unsafe extern "C" fn profile_callback(
+ p_arg: *mut c_void,
+ z_sql: *const c_char,
+ nanoseconds: u64,
+ ) {
+ let profile_fn: fn(&str, Duration) = mem::transmute(p_arg);
+ let c_slice = CStr::from_ptr(z_sql).to_bytes();
+ let s = String::from_utf8_lossy(c_slice);
+ const NANOS_PER_SEC: u64 = 1_000_000_000;
+
+ let duration = Duration::new(
+ nanoseconds / NANOS_PER_SEC,
+ (nanoseconds % NANOS_PER_SEC) as u32,
+ );
+ drop(catch_unwind(|| profile_fn(&s, duration)));
+ }
+
+ let c = self.db.borrow_mut();
+ match profile_fn {
+ Some(f) => unsafe {
+ ffi::sqlite3_profile(c.db(), Some(profile_callback), f as *mut c_void)
+ },
+ None => unsafe { ffi::sqlite3_profile(c.db(), None, ptr::null_mut()) },
+ };
+ }
+
+ // TODO sqlite3_trace_v2 (https://sqlite.org/c3ref/trace_v2.html) // 3.14.0, #977
+}
+
+#[cfg(test)]
+mod test {
+ use lazy_static::lazy_static;
+ use std::sync::Mutex;
+ use std::time::Duration;
+
+ use crate::{Connection, Result};
+
+ #[test]
+ fn test_trace() -> Result<()> {
+ lazy_static! {
+ static ref TRACED_STMTS: Mutex<Vec<String>> = Mutex::new(Vec::new());
+ }
+ fn tracer(s: &str) {
+ let mut traced_stmts = TRACED_STMTS.lock().unwrap();
+ traced_stmts.push(s.to_owned());
+ }
+
+ let mut db = Connection::open_in_memory()?;
+ db.trace(Some(tracer));
+ {
+ let _ = db.query_row("SELECT ?1", [1i32], |_| Ok(()));
+ let _ = db.query_row("SELECT ?1", ["hello"], |_| Ok(()));
+ }
+ db.trace(None);
+ {
+ let _ = db.query_row("SELECT ?1", [2i32], |_| Ok(()));
+ let _ = db.query_row("SELECT ?1", ["goodbye"], |_| Ok(()));
+ }
+
+ let traced_stmts = TRACED_STMTS.lock().unwrap();
+ assert_eq!(traced_stmts.len(), 2);
+ assert_eq!(traced_stmts[0], "SELECT 1");
+ assert_eq!(traced_stmts[1], "SELECT 'hello'");
+ Ok(())
+ }
+
+ #[test]
+ fn test_profile() -> Result<()> {
+ lazy_static! {
+ static ref PROFILED: Mutex<Vec<(String, Duration)>> = Mutex::new(Vec::new());
+ }
+ fn profiler(s: &str, d: Duration) {
+ let mut profiled = PROFILED.lock().unwrap();
+ profiled.push((s.to_owned(), d));
+ }
+
+ let mut db = Connection::open_in_memory()?;
+ db.profile(Some(profiler));
+ db.execute_batch("PRAGMA application_id = 1")?;
+ db.profile(None);
+ db.execute_batch("PRAGMA application_id = 2")?;
+
+ let profiled = PROFILED.lock().unwrap();
+ assert_eq!(profiled.len(), 1);
+ assert_eq!(profiled[0].0, "PRAGMA application_id = 1");
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/transaction.rs b/third_party/rust/rusqlite/src/transaction.rs
new file mode 100644
index 0000000000..2e905286b9
--- /dev/null
+++ b/third_party/rust/rusqlite/src/transaction.rs
@@ -0,0 +1,778 @@
+use crate::{Connection, Result};
+use std::ops::Deref;
+
+/// Options for transaction behavior. See [BEGIN
+/// TRANSACTION](http://www.sqlite.org/lang_transaction.html) for details.
+#[derive(Copy, Clone)]
+#[non_exhaustive]
+pub enum TransactionBehavior {
+ /// DEFERRED means that the transaction does not actually start until the
+ /// database is first accessed.
+ Deferred,
+ /// IMMEDIATE cause the database connection to start a new write
+ /// immediately, without waiting for a writes statement.
+ Immediate,
+ /// EXCLUSIVE prevents other database connections from reading the database
+ /// while the transaction is underway.
+ Exclusive,
+}
+
+/// Options for how a Transaction or Savepoint should behave when it is dropped.
+#[derive(Copy, Clone, Debug, PartialEq, Eq)]
+#[non_exhaustive]
+pub enum DropBehavior {
+ /// Roll back the changes. This is the default.
+ Rollback,
+
+ /// Commit the changes.
+ Commit,
+
+ /// Do not commit or roll back changes - this will leave the transaction or
+ /// savepoint open, so should be used with care.
+ Ignore,
+
+ /// Panic. Used to enforce intentional behavior during development.
+ Panic,
+}
+
+/// Represents a transaction on a database connection.
+///
+/// ## Note
+///
+/// Transactions will roll back by default. Use `commit` method to explicitly
+/// commit the transaction, or use `set_drop_behavior` to change what happens
+/// when the transaction is dropped.
+///
+/// ## Example
+///
+/// ```rust,no_run
+/// # use rusqlite::{Connection, Result};
+/// # fn do_queries_part_1(_conn: &Connection) -> Result<()> { Ok(()) }
+/// # fn do_queries_part_2(_conn: &Connection) -> Result<()> { Ok(()) }
+/// fn perform_queries(conn: &mut Connection) -> Result<()> {
+/// let tx = conn.transaction()?;
+///
+/// do_queries_part_1(&tx)?; // tx causes rollback if this fails
+/// do_queries_part_2(&tx)?; // tx causes rollback if this fails
+///
+/// tx.commit()
+/// }
+/// ```
+#[derive(Debug)]
+pub struct Transaction<'conn> {
+ conn: &'conn Connection,
+ drop_behavior: DropBehavior,
+}
+
+/// Represents a savepoint on a database connection.
+///
+/// ## Note
+///
+/// Savepoints will roll back by default. Use `commit` method to explicitly
+/// commit the savepoint, or use `set_drop_behavior` to change what happens
+/// when the savepoint is dropped.
+///
+/// ## Example
+///
+/// ```rust,no_run
+/// # use rusqlite::{Connection, Result};
+/// # fn do_queries_part_1(_conn: &Connection) -> Result<()> { Ok(()) }
+/// # fn do_queries_part_2(_conn: &Connection) -> Result<()> { Ok(()) }
+/// fn perform_queries(conn: &mut Connection) -> Result<()> {
+/// let sp = conn.savepoint()?;
+///
+/// do_queries_part_1(&sp)?; // sp causes rollback if this fails
+/// do_queries_part_2(&sp)?; // sp causes rollback if this fails
+///
+/// sp.commit()
+/// }
+/// ```
+#[derive(Debug)]
+pub struct Savepoint<'conn> {
+ conn: &'conn Connection,
+ name: String,
+ drop_behavior: DropBehavior,
+ committed: bool,
+}
+
+impl Transaction<'_> {
+ /// Begin a new transaction. Cannot be nested; see `savepoint` for nested
+ /// transactions.
+ ///
+ /// Even though we don't mutate the connection, we take a `&mut Connection`
+ /// so as to prevent nested transactions on the same connection. For cases
+ /// where this is unacceptable, [`Transaction::new_unchecked`] is available.
+ #[inline]
+ pub fn new(conn: &mut Connection, behavior: TransactionBehavior) -> Result<Transaction<'_>> {
+ Self::new_unchecked(conn, behavior)
+ }
+
+ /// Begin a new transaction, failing if a transaction is open.
+ ///
+ /// If a transaction is already open, this will return an error. Where
+ /// possible, [`Transaction::new`] should be preferred, as it provides a
+ /// compile-time guarantee that transactions are not nested.
+ #[inline]
+ pub fn new_unchecked(
+ conn: &Connection,
+ behavior: TransactionBehavior,
+ ) -> Result<Transaction<'_>> {
+ let query = match behavior {
+ TransactionBehavior::Deferred => "BEGIN DEFERRED",
+ TransactionBehavior::Immediate => "BEGIN IMMEDIATE",
+ TransactionBehavior::Exclusive => "BEGIN EXCLUSIVE",
+ };
+ conn.execute_batch(query).map(move |()| Transaction {
+ conn,
+ drop_behavior: DropBehavior::Rollback,
+ })
+ }
+
+ /// Starts a new [savepoint](http://www.sqlite.org/lang_savepoint.html), allowing nested
+ /// transactions.
+ ///
+ /// ## Note
+ ///
+ /// Just like outer level transactions, savepoint transactions rollback by
+ /// default.
+ ///
+ /// ## Example
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// # fn perform_queries_part_1_succeeds(_conn: &Connection) -> bool { true }
+ /// fn perform_queries(conn: &mut Connection) -> Result<()> {
+ /// let mut tx = conn.transaction()?;
+ ///
+ /// {
+ /// let sp = tx.savepoint()?;
+ /// if perform_queries_part_1_succeeds(&sp) {
+ /// sp.commit()?;
+ /// }
+ /// // otherwise, sp will rollback
+ /// }
+ ///
+ /// tx.commit()
+ /// }
+ /// ```
+ #[inline]
+ pub fn savepoint(&mut self) -> Result<Savepoint<'_>> {
+ Savepoint::new_(self.conn)
+ }
+
+ /// Create a new savepoint with a custom savepoint name. See `savepoint()`.
+ #[inline]
+ pub fn savepoint_with_name<T: Into<String>>(&mut self, name: T) -> Result<Savepoint<'_>> {
+ Savepoint::with_name_(self.conn, name)
+ }
+
+ /// Get the current setting for what happens to the transaction when it is
+ /// dropped.
+ #[inline]
+ #[must_use]
+ pub fn drop_behavior(&self) -> DropBehavior {
+ self.drop_behavior
+ }
+
+ /// Configure the transaction to perform the specified action when it is
+ /// dropped.
+ #[inline]
+ pub fn set_drop_behavior(&mut self, drop_behavior: DropBehavior) {
+ self.drop_behavior = drop_behavior;
+ }
+
+ /// A convenience method which consumes and commits a transaction.
+ #[inline]
+ pub fn commit(mut self) -> Result<()> {
+ self.commit_()
+ }
+
+ #[inline]
+ fn commit_(&mut self) -> Result<()> {
+ self.conn.execute_batch("COMMIT")?;
+ Ok(())
+ }
+
+ /// A convenience method which consumes and rolls back a transaction.
+ #[inline]
+ pub fn rollback(mut self) -> Result<()> {
+ self.rollback_()
+ }
+
+ #[inline]
+ fn rollback_(&mut self) -> Result<()> {
+ self.conn.execute_batch("ROLLBACK")?;
+ Ok(())
+ }
+
+ /// Consumes the transaction, committing or rolling back according to the
+ /// current setting (see `drop_behavior`).
+ ///
+ /// Functionally equivalent to the `Drop` implementation, but allows
+ /// callers to see any errors that occur.
+ #[inline]
+ pub fn finish(mut self) -> Result<()> {
+ self.finish_()
+ }
+
+ #[inline]
+ fn finish_(&mut self) -> Result<()> {
+ if self.conn.is_autocommit() {
+ return Ok(());
+ }
+ match self.drop_behavior() {
+ DropBehavior::Commit => self.commit_().or_else(|_| self.rollback_()),
+ DropBehavior::Rollback => self.rollback_(),
+ DropBehavior::Ignore => Ok(()),
+ DropBehavior::Panic => panic!("Transaction dropped unexpectedly."),
+ }
+ }
+}
+
+impl Deref for Transaction<'_> {
+ type Target = Connection;
+
+ #[inline]
+ fn deref(&self) -> &Connection {
+ self.conn
+ }
+}
+
+#[allow(unused_must_use)]
+impl Drop for Transaction<'_> {
+ #[inline]
+ fn drop(&mut self) {
+ self.finish_();
+ }
+}
+
+impl Savepoint<'_> {
+ #[inline]
+ fn with_name_<T: Into<String>>(conn: &Connection, name: T) -> Result<Savepoint<'_>> {
+ let name = name.into();
+ conn.execute_batch(&format!("SAVEPOINT {name}"))
+ .map(|()| Savepoint {
+ conn,
+ name,
+ drop_behavior: DropBehavior::Rollback,
+ committed: false,
+ })
+ }
+
+ #[inline]
+ fn new_(conn: &Connection) -> Result<Savepoint<'_>> {
+ Savepoint::with_name_(conn, "_rusqlite_sp")
+ }
+
+ /// Begin a new savepoint. Can be nested.
+ #[inline]
+ pub fn new(conn: &mut Connection) -> Result<Savepoint<'_>> {
+ Savepoint::new_(conn)
+ }
+
+ /// Begin a new savepoint with a user-provided savepoint name.
+ #[inline]
+ pub fn with_name<T: Into<String>>(conn: &mut Connection, name: T) -> Result<Savepoint<'_>> {
+ Savepoint::with_name_(conn, name)
+ }
+
+ /// Begin a nested savepoint.
+ #[inline]
+ pub fn savepoint(&mut self) -> Result<Savepoint<'_>> {
+ Savepoint::new_(self.conn)
+ }
+
+ /// Begin a nested savepoint with a user-provided savepoint name.
+ #[inline]
+ pub fn savepoint_with_name<T: Into<String>>(&mut self, name: T) -> Result<Savepoint<'_>> {
+ Savepoint::with_name_(self.conn, name)
+ }
+
+ /// Get the current setting for what happens to the savepoint when it is
+ /// dropped.
+ #[inline]
+ #[must_use]
+ pub fn drop_behavior(&self) -> DropBehavior {
+ self.drop_behavior
+ }
+
+ /// Configure the savepoint to perform the specified action when it is
+ /// dropped.
+ #[inline]
+ pub fn set_drop_behavior(&mut self, drop_behavior: DropBehavior) {
+ self.drop_behavior = drop_behavior;
+ }
+
+ /// A convenience method which consumes and commits a savepoint.
+ #[inline]
+ pub fn commit(mut self) -> Result<()> {
+ self.commit_()
+ }
+
+ #[inline]
+ fn commit_(&mut self) -> Result<()> {
+ self.conn.execute_batch(&format!("RELEASE {}", self.name))?;
+ self.committed = true;
+ Ok(())
+ }
+
+ /// A convenience method which rolls back a savepoint.
+ ///
+ /// ## Note
+ ///
+ /// Unlike `Transaction`s, savepoints remain active after they have been
+ /// rolled back, and can be rolled back again or committed.
+ #[inline]
+ pub fn rollback(&mut self) -> Result<()> {
+ self.conn
+ .execute_batch(&format!("ROLLBACK TO {}", self.name))
+ }
+
+ /// Consumes the savepoint, committing or rolling back according to the
+ /// current setting (see `drop_behavior`).
+ ///
+ /// Functionally equivalent to the `Drop` implementation, but allows
+ /// callers to see any errors that occur.
+ #[inline]
+ pub fn finish(mut self) -> Result<()> {
+ self.finish_()
+ }
+
+ #[inline]
+ fn finish_(&mut self) -> Result<()> {
+ if self.committed {
+ return Ok(());
+ }
+ match self.drop_behavior() {
+ DropBehavior::Commit => self
+ .commit_()
+ .or_else(|_| self.rollback().and_then(|()| self.commit_())),
+ DropBehavior::Rollback => self.rollback().and_then(|()| self.commit_()),
+ DropBehavior::Ignore => Ok(()),
+ DropBehavior::Panic => panic!("Savepoint dropped unexpectedly."),
+ }
+ }
+}
+
+impl Deref for Savepoint<'_> {
+ type Target = Connection;
+
+ #[inline]
+ fn deref(&self) -> &Connection {
+ self.conn
+ }
+}
+
+#[allow(unused_must_use)]
+impl Drop for Savepoint<'_> {
+ #[inline]
+ fn drop(&mut self) {
+ self.finish_();
+ }
+}
+
+/// Transaction state of a database
+#[derive(Clone, Copy, Debug, PartialEq, Eq)]
+#[non_exhaustive]
+#[cfg(feature = "modern_sqlite")] // 3.37.0
+#[cfg_attr(docsrs, doc(cfg(feature = "modern_sqlite")))]
+pub enum TransactionState {
+ /// Equivalent to SQLITE_TXN_NONE
+ None,
+ /// Equivalent to SQLITE_TXN_READ
+ Read,
+ /// Equivalent to SQLITE_TXN_WRITE
+ Write,
+}
+
+impl Connection {
+ /// Begin a new transaction with the default behavior (DEFERRED).
+ ///
+ /// The transaction defaults to rolling back when it is dropped. If you
+ /// want the transaction to commit, you must call
+ /// [`commit`](Transaction::commit) or
+ /// [`set_drop_behavior(DropBehavior::Commit)`](Transaction::set_drop_behavior).
+ ///
+ /// ## Example
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// # fn do_queries_part_1(_conn: &Connection) -> Result<()> { Ok(()) }
+ /// # fn do_queries_part_2(_conn: &Connection) -> Result<()> { Ok(()) }
+ /// fn perform_queries(conn: &mut Connection) -> Result<()> {
+ /// let tx = conn.transaction()?;
+ ///
+ /// do_queries_part_1(&tx)?; // tx causes rollback if this fails
+ /// do_queries_part_2(&tx)?; // tx causes rollback if this fails
+ ///
+ /// tx.commit()
+ /// }
+ /// ```
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite call fails.
+ #[inline]
+ pub fn transaction(&mut self) -> Result<Transaction<'_>> {
+ Transaction::new(self, TransactionBehavior::Deferred)
+ }
+
+ /// Begin a new transaction with a specified behavior.
+ ///
+ /// See [`transaction`](Connection::transaction).
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite call fails.
+ #[inline]
+ pub fn transaction_with_behavior(
+ &mut self,
+ behavior: TransactionBehavior,
+ ) -> Result<Transaction<'_>> {
+ Transaction::new(self, behavior)
+ }
+
+ /// Begin a new transaction with the default behavior (DEFERRED).
+ ///
+ /// Attempt to open a nested transaction will result in a SQLite error.
+ /// `Connection::transaction` prevents this at compile time by taking `&mut
+ /// self`, but `Connection::unchecked_transaction()` may be used to defer
+ /// the checking until runtime.
+ ///
+ /// See [`Connection::transaction`] and [`Transaction::new_unchecked`]
+ /// (which can be used if the default transaction behavior is undesirable).
+ ///
+ /// ## Example
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// # use std::rc::Rc;
+ /// # fn do_queries_part_1(_conn: &Connection) -> Result<()> { Ok(()) }
+ /// # fn do_queries_part_2(_conn: &Connection) -> Result<()> { Ok(()) }
+ /// fn perform_queries(conn: Rc<Connection>) -> Result<()> {
+ /// let tx = conn.unchecked_transaction()?;
+ ///
+ /// do_queries_part_1(&tx)?; // tx causes rollback if this fails
+ /// do_queries_part_2(&tx)?; // tx causes rollback if this fails
+ ///
+ /// tx.commit()
+ /// }
+ /// ```
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite call fails. The specific
+ /// error returned if transactions are nested is currently unspecified.
+ pub fn unchecked_transaction(&self) -> Result<Transaction<'_>> {
+ Transaction::new_unchecked(self, TransactionBehavior::Deferred)
+ }
+
+ /// Begin a new savepoint with the default behavior (DEFERRED).
+ ///
+ /// The savepoint defaults to rolling back when it is dropped. If you want
+ /// the savepoint to commit, you must call [`commit`](Savepoint::commit) or
+ /// [`set_drop_behavior(DropBehavior::Commit)`](Savepoint::set_drop_behavior).
+ ///
+ /// ## Example
+ ///
+ /// ```rust,no_run
+ /// # use rusqlite::{Connection, Result};
+ /// # fn do_queries_part_1(_conn: &Connection) -> Result<()> { Ok(()) }
+ /// # fn do_queries_part_2(_conn: &Connection) -> Result<()> { Ok(()) }
+ /// fn perform_queries(conn: &mut Connection) -> Result<()> {
+ /// let sp = conn.savepoint()?;
+ ///
+ /// do_queries_part_1(&sp)?; // sp causes rollback if this fails
+ /// do_queries_part_2(&sp)?; // sp causes rollback if this fails
+ ///
+ /// sp.commit()
+ /// }
+ /// ```
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite call fails.
+ #[inline]
+ pub fn savepoint(&mut self) -> Result<Savepoint<'_>> {
+ Savepoint::new(self)
+ }
+
+ /// Begin a new savepoint with a specified name.
+ ///
+ /// See [`savepoint`](Connection::savepoint).
+ ///
+ /// # Failure
+ ///
+ /// Will return `Err` if the underlying SQLite call fails.
+ #[inline]
+ pub fn savepoint_with_name<T: Into<String>>(&mut self, name: T) -> Result<Savepoint<'_>> {
+ Savepoint::with_name(self, name)
+ }
+
+ /// Determine the transaction state of a database
+ #[cfg(feature = "modern_sqlite")] // 3.37.0
+ #[cfg_attr(docsrs, doc(cfg(feature = "modern_sqlite")))]
+ pub fn transaction_state(
+ &self,
+ db_name: Option<crate::DatabaseName<'_>>,
+ ) -> Result<TransactionState> {
+ self.db.borrow().txn_state(db_name)
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use super::DropBehavior;
+ use crate::{Connection, Error, Result};
+
+ fn checked_memory_handle() -> Result<Connection> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo (x INTEGER)")?;
+ Ok(db)
+ }
+
+ #[test]
+ fn test_drop() -> Result<()> {
+ let mut db = checked_memory_handle()?;
+ {
+ let tx = db.transaction()?;
+ tx.execute_batch("INSERT INTO foo VALUES(1)")?;
+ // default: rollback
+ }
+ {
+ let mut tx = db.transaction()?;
+ tx.execute_batch("INSERT INTO foo VALUES(2)")?;
+ tx.set_drop_behavior(DropBehavior::Commit)
+ }
+ {
+ let tx = db.transaction()?;
+ assert_eq!(2i32, tx.one_column::<i32>("SELECT SUM(x) FROM foo")?);
+ }
+ Ok(())
+ }
+ fn assert_nested_tx_error(e: Error) {
+ if let Error::SqliteFailure(e, Some(m)) = &e {
+ assert_eq!(e.extended_code, crate::ffi::SQLITE_ERROR);
+ // FIXME: Not ideal...
+ assert_eq!(e.code, crate::ErrorCode::Unknown);
+ assert!(m.contains("transaction"));
+ } else {
+ panic!("Unexpected error type: {e:?}");
+ }
+ }
+
+ #[test]
+ fn test_unchecked_nesting() -> Result<()> {
+ let db = checked_memory_handle()?;
+
+ {
+ let tx = db.unchecked_transaction()?;
+ let e = tx.unchecked_transaction().unwrap_err();
+ assert_nested_tx_error(e);
+ // default: rollback
+ }
+ {
+ let tx = db.unchecked_transaction()?;
+ tx.execute_batch("INSERT INTO foo VALUES(1)")?;
+ // Ensure this doesn't interfere with ongoing transaction
+ let e = tx.unchecked_transaction().unwrap_err();
+ assert_nested_tx_error(e);
+
+ tx.execute_batch("INSERT INTO foo VALUES(1)")?;
+ tx.commit()?;
+ }
+
+ assert_eq!(2i32, db.one_column::<i32>("SELECT SUM(x) FROM foo")?);
+ Ok(())
+ }
+
+ #[test]
+ fn test_explicit_rollback_commit() -> Result<()> {
+ let mut db = checked_memory_handle()?;
+ {
+ let mut tx = db.transaction()?;
+ {
+ let mut sp = tx.savepoint()?;
+ sp.execute_batch("INSERT INTO foo VALUES(1)")?;
+ sp.rollback()?;
+ sp.execute_batch("INSERT INTO foo VALUES(2)")?;
+ sp.commit()?;
+ }
+ tx.commit()?;
+ }
+ {
+ let tx = db.transaction()?;
+ tx.execute_batch("INSERT INTO foo VALUES(4)")?;
+ tx.commit()?;
+ }
+ {
+ let tx = db.transaction()?;
+ assert_eq!(6i32, tx.one_column::<i32>("SELECT SUM(x) FROM foo")?);
+ }
+ Ok(())
+ }
+
+ #[test]
+ fn test_savepoint() -> Result<()> {
+ let mut db = checked_memory_handle()?;
+ {
+ let mut tx = db.transaction()?;
+ tx.execute_batch("INSERT INTO foo VALUES(1)")?;
+ assert_current_sum(1, &tx)?;
+ tx.set_drop_behavior(DropBehavior::Commit);
+ {
+ let mut sp1 = tx.savepoint()?;
+ sp1.execute_batch("INSERT INTO foo VALUES(2)")?;
+ assert_current_sum(3, &sp1)?;
+ // will rollback sp1
+ {
+ let mut sp2 = sp1.savepoint()?;
+ sp2.execute_batch("INSERT INTO foo VALUES(4)")?;
+ assert_current_sum(7, &sp2)?;
+ // will rollback sp2
+ {
+ let sp3 = sp2.savepoint()?;
+ sp3.execute_batch("INSERT INTO foo VALUES(8)")?;
+ assert_current_sum(15, &sp3)?;
+ sp3.commit()?;
+ // committed sp3, but will be erased by sp2 rollback
+ }
+ assert_current_sum(15, &sp2)?;
+ }
+ assert_current_sum(3, &sp1)?;
+ }
+ assert_current_sum(1, &tx)?;
+ }
+ assert_current_sum(1, &db)?;
+ Ok(())
+ }
+
+ #[test]
+ fn test_ignore_drop_behavior() -> Result<()> {
+ let mut db = checked_memory_handle()?;
+
+ let mut tx = db.transaction()?;
+ {
+ let mut sp1 = tx.savepoint()?;
+ insert(1, &sp1)?;
+ sp1.rollback()?;
+ insert(2, &sp1)?;
+ {
+ let mut sp2 = sp1.savepoint()?;
+ sp2.set_drop_behavior(DropBehavior::Ignore);
+ insert(4, &sp2)?;
+ }
+ assert_current_sum(6, &sp1)?;
+ sp1.commit()?;
+ }
+ assert_current_sum(6, &tx)?;
+ Ok(())
+ }
+
+ #[test]
+ fn test_savepoint_drop_behavior_releases() -> Result<()> {
+ let mut db = checked_memory_handle()?;
+
+ {
+ let mut sp = db.savepoint()?;
+ sp.set_drop_behavior(DropBehavior::Commit);
+ }
+ assert!(db.is_autocommit());
+ {
+ let mut sp = db.savepoint()?;
+ sp.set_drop_behavior(DropBehavior::Rollback);
+ }
+ assert!(db.is_autocommit());
+
+ Ok(())
+ }
+
+ #[test]
+ fn test_savepoint_release_error() -> Result<()> {
+ let mut db = checked_memory_handle()?;
+
+ db.pragma_update(None, "foreign_keys", true)?;
+ db.execute_batch("CREATE TABLE r(n INTEGER PRIMARY KEY NOT NULL); CREATE TABLE f(n REFERENCES r(n) DEFERRABLE INITIALLY DEFERRED);")?;
+ {
+ let mut sp = db.savepoint()?;
+ sp.execute("INSERT INTO f VALUES (0)", [])?;
+ sp.set_drop_behavior(DropBehavior::Commit);
+ }
+ assert!(db.is_autocommit());
+
+ Ok(())
+ }
+
+ #[test]
+ fn test_savepoint_names() -> Result<()> {
+ let mut db = checked_memory_handle()?;
+
+ {
+ let mut sp1 = db.savepoint_with_name("my_sp")?;
+ insert(1, &sp1)?;
+ assert_current_sum(1, &sp1)?;
+ {
+ let mut sp2 = sp1.savepoint_with_name("my_sp")?;
+ sp2.set_drop_behavior(DropBehavior::Commit);
+ insert(2, &sp2)?;
+ assert_current_sum(3, &sp2)?;
+ sp2.rollback()?;
+ assert_current_sum(1, &sp2)?;
+ insert(4, &sp2)?;
+ }
+ assert_current_sum(5, &sp1)?;
+ sp1.rollback()?;
+ {
+ let mut sp2 = sp1.savepoint_with_name("my_sp")?;
+ sp2.set_drop_behavior(DropBehavior::Ignore);
+ insert(8, &sp2)?;
+ }
+ assert_current_sum(8, &sp1)?;
+ sp1.commit()?;
+ }
+ assert_current_sum(8, &db)?;
+ Ok(())
+ }
+
+ #[test]
+ fn test_rc() -> Result<()> {
+ use std::rc::Rc;
+ let mut conn = Connection::open_in_memory()?;
+ let rc_txn = Rc::new(conn.transaction()?);
+
+ // This will compile only if Transaction is Debug
+ Rc::try_unwrap(rc_txn).unwrap();
+ Ok(())
+ }
+
+ fn insert(x: i32, conn: &Connection) -> Result<usize> {
+ conn.execute("INSERT INTO foo VALUES(?1)", [x])
+ }
+
+ fn assert_current_sum(x: i32, conn: &Connection) -> Result<()> {
+ let i = conn.one_column::<i32>("SELECT SUM(x) FROM foo")?;
+ assert_eq!(x, i);
+ Ok(())
+ }
+
+ #[test]
+ #[cfg(feature = "modern_sqlite")]
+ fn txn_state() -> Result<()> {
+ use super::TransactionState;
+ use crate::DatabaseName;
+ let db = Connection::open_in_memory()?;
+ assert_eq!(
+ TransactionState::None,
+ db.transaction_state(Some(DatabaseName::Main))?
+ );
+ assert_eq!(TransactionState::None, db.transaction_state(None)?);
+ db.execute_batch("BEGIN")?;
+ assert_eq!(TransactionState::None, db.transaction_state(None)?);
+ let _: i32 = db.pragma_query_value(None, "user_version", |row| row.get(0))?;
+ assert_eq!(TransactionState::Read, db.transaction_state(None)?);
+ db.pragma_update(None, "user_version", 1)?;
+ assert_eq!(TransactionState::Write, db.transaction_state(None)?);
+ db.execute_batch("ROLLBACK")?;
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/types/chrono.rs b/third_party/rust/rusqlite/src/types/chrono.rs
new file mode 100644
index 0000000000..6b50e01099
--- /dev/null
+++ b/third_party/rust/rusqlite/src/types/chrono.rs
@@ -0,0 +1,319 @@
+//! Convert most of the [Time Strings](http://sqlite.org/lang_datefunc.html) to chrono types.
+
+use chrono::{DateTime, FixedOffset, Local, NaiveDate, NaiveDateTime, NaiveTime, TimeZone, Utc};
+
+use crate::types::{FromSql, FromSqlError, FromSqlResult, ToSql, ToSqlOutput, ValueRef};
+use crate::Result;
+
+/// ISO 8601 calendar date without timezone => "YYYY-MM-DD"
+impl ToSql for NaiveDate {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ let date_str = self.format("%F").to_string();
+ Ok(ToSqlOutput::from(date_str))
+ }
+}
+
+/// "YYYY-MM-DD" => ISO 8601 calendar date without timezone.
+impl FromSql for NaiveDate {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ value
+ .as_str()
+ .and_then(|s| match NaiveDate::parse_from_str(s, "%F") {
+ Ok(dt) => Ok(dt),
+ Err(err) => Err(FromSqlError::Other(Box::new(err))),
+ })
+ }
+}
+
+/// ISO 8601 time without timezone => "HH:MM:SS.SSS"
+impl ToSql for NaiveTime {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ let date_str = self.format("%T%.f").to_string();
+ Ok(ToSqlOutput::from(date_str))
+ }
+}
+
+/// "HH:MM"/"HH:MM:SS"/"HH:MM:SS.SSS" => ISO 8601 time without timezone.
+impl FromSql for NaiveTime {
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ value.as_str().and_then(|s| {
+ let fmt = match s.len() {
+ 5 => "%H:%M",
+ 8 => "%T",
+ _ => "%T%.f",
+ };
+ match NaiveTime::parse_from_str(s, fmt) {
+ Ok(dt) => Ok(dt),
+ Err(err) => Err(FromSqlError::Other(Box::new(err))),
+ }
+ })
+ }
+}
+
+/// ISO 8601 combined date and time without timezone =>
+/// "YYYY-MM-DD HH:MM:SS.SSS"
+impl ToSql for NaiveDateTime {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ let date_str = self.format("%F %T%.f").to_string();
+ Ok(ToSqlOutput::from(date_str))
+ }
+}
+
+/// "YYYY-MM-DD HH:MM:SS"/"YYYY-MM-DD HH:MM:SS.SSS" => ISO 8601 combined date
+/// and time without timezone. ("YYYY-MM-DDTHH:MM:SS"/"YYYY-MM-DDTHH:MM:SS.SSS"
+/// also supported)
+impl FromSql for NaiveDateTime {
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ value.as_str().and_then(|s| {
+ let fmt = if s.len() >= 11 && s.as_bytes()[10] == b'T' {
+ "%FT%T%.f"
+ } else {
+ "%F %T%.f"
+ };
+
+ match NaiveDateTime::parse_from_str(s, fmt) {
+ Ok(dt) => Ok(dt),
+ Err(err) => Err(FromSqlError::Other(Box::new(err))),
+ }
+ })
+ }
+}
+
+/// UTC time => UTC RFC3339 timestamp
+/// ("YYYY-MM-DD HH:MM:SS.SSS+00:00").
+impl ToSql for DateTime<Utc> {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ let date_str = self.format("%F %T%.f%:z").to_string();
+ Ok(ToSqlOutput::from(date_str))
+ }
+}
+
+/// Local time => UTC RFC3339 timestamp
+/// ("YYYY-MM-DD HH:MM:SS.SSS+00:00").
+impl ToSql for DateTime<Local> {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ let date_str = self.with_timezone(&Utc).format("%F %T%.f%:z").to_string();
+ Ok(ToSqlOutput::from(date_str))
+ }
+}
+
+/// Date and time with time zone => RFC3339 timestamp
+/// ("YYYY-MM-DD HH:MM:SS.SSS[+-]HH:MM").
+impl ToSql for DateTime<FixedOffset> {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ let date_str = self.format("%F %T%.f%:z").to_string();
+ Ok(ToSqlOutput::from(date_str))
+ }
+}
+
+/// RFC3339 ("YYYY-MM-DD HH:MM:SS.SSS[+-]HH:MM") into `DateTime<Utc>`.
+impl FromSql for DateTime<Utc> {
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ {
+ // Try to parse value as rfc3339 first.
+ let s = value.as_str()?;
+
+ let fmt = if s.len() >= 11 && s.as_bytes()[10] == b'T' {
+ "%FT%T%.f%#z"
+ } else {
+ "%F %T%.f%#z"
+ };
+
+ if let Ok(dt) = DateTime::parse_from_str(s, fmt) {
+ return Ok(dt.with_timezone(&Utc));
+ }
+ }
+
+ // Couldn't parse as rfc3339 - fall back to NaiveDateTime.
+ NaiveDateTime::column_result(value).map(|dt| Utc.from_utc_datetime(&dt))
+ }
+}
+
+/// RFC3339 ("YYYY-MM-DD HH:MM:SS.SSS[+-]HH:MM") into `DateTime<Local>`.
+impl FromSql for DateTime<Local> {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ let utc_dt = DateTime::<Utc>::column_result(value)?;
+ Ok(utc_dt.with_timezone(&Local))
+ }
+}
+
+/// RFC3339 ("YYYY-MM-DD HH:MM:SS.SSS[+-]HH:MM") into `DateTime<FixedOffset>`.
+impl FromSql for DateTime<FixedOffset> {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ let s = String::column_result(value)?;
+ Self::parse_from_rfc3339(s.as_str())
+ .or_else(|_| Self::parse_from_str(s.as_str(), "%F %T%.f%:z"))
+ .map_err(|e| FromSqlError::Other(Box::new(e)))
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use crate::{
+ types::{FromSql, ValueRef},
+ Connection, Result,
+ };
+ use chrono::{
+ DateTime, Duration, FixedOffset, Local, NaiveDate, NaiveDateTime, NaiveTime, TimeZone, Utc,
+ };
+
+ fn checked_memory_handle() -> Result<Connection> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo (t TEXT, i INTEGER, f FLOAT, b BLOB)")?;
+ Ok(db)
+ }
+
+ #[test]
+ fn test_naive_date() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let date = NaiveDate::from_ymd_opt(2016, 2, 23).unwrap();
+ db.execute("INSERT INTO foo (t) VALUES (?1)", [date])?;
+
+ let s: String = db.one_column("SELECT t FROM foo")?;
+ assert_eq!("2016-02-23", s);
+ let t: NaiveDate = db.one_column("SELECT t FROM foo")?;
+ assert_eq!(date, t);
+ Ok(())
+ }
+
+ #[test]
+ fn test_naive_time() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let time = NaiveTime::from_hms_opt(23, 56, 4).unwrap();
+ db.execute("INSERT INTO foo (t) VALUES (?1)", [time])?;
+
+ let s: String = db.one_column("SELECT t FROM foo")?;
+ assert_eq!("23:56:04", s);
+ let v: NaiveTime = db.one_column("SELECT t FROM foo")?;
+ assert_eq!(time, v);
+ Ok(())
+ }
+
+ #[test]
+ fn test_naive_date_time() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let date = NaiveDate::from_ymd_opt(2016, 2, 23).unwrap();
+ let time = NaiveTime::from_hms_opt(23, 56, 4).unwrap();
+ let dt = NaiveDateTime::new(date, time);
+
+ db.execute("INSERT INTO foo (t) VALUES (?1)", [dt])?;
+
+ let s: String = db.one_column("SELECT t FROM foo")?;
+ assert_eq!("2016-02-23 23:56:04", s);
+ let v: NaiveDateTime = db.one_column("SELECT t FROM foo")?;
+ assert_eq!(dt, v);
+
+ db.execute("UPDATE foo set b = datetime(t)", [])?; // "YYYY-MM-DD HH:MM:SS"
+ let hms: NaiveDateTime = db.one_column("SELECT b FROM foo")?;
+ assert_eq!(dt, hms);
+ Ok(())
+ }
+
+ #[test]
+ fn test_date_time_utc() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let date = NaiveDate::from_ymd_opt(2016, 2, 23).unwrap();
+ let time = NaiveTime::from_hms_milli_opt(23, 56, 4, 789).unwrap();
+ let dt = NaiveDateTime::new(date, time);
+ let utc = Utc.from_utc_datetime(&dt);
+
+ db.execute("INSERT INTO foo (t) VALUES (?1)", [utc])?;
+
+ let s: String = db.one_column("SELECT t FROM foo")?;
+ assert_eq!("2016-02-23 23:56:04.789+00:00", s);
+
+ let v1: DateTime<Utc> = db.one_column("SELECT t FROM foo")?;
+ assert_eq!(utc, v1);
+
+ let v2: DateTime<Utc> = db.one_column("SELECT '2016-02-23 23:56:04.789'")?;
+ assert_eq!(utc, v2);
+
+ let v3: DateTime<Utc> = db.one_column("SELECT '2016-02-23 23:56:04'")?;
+ assert_eq!(utc - Duration::milliseconds(789), v3);
+
+ let v4: DateTime<Utc> = db.one_column("SELECT '2016-02-23 23:56:04.789+00:00'")?;
+ assert_eq!(utc, v4);
+ Ok(())
+ }
+
+ #[test]
+ fn test_date_time_local() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let date = NaiveDate::from_ymd_opt(2016, 2, 23).unwrap();
+ let time = NaiveTime::from_hms_milli_opt(23, 56, 4, 789).unwrap();
+ let dt = NaiveDateTime::new(date, time);
+ let local = Local.from_local_datetime(&dt).single().unwrap();
+
+ db.execute("INSERT INTO foo (t) VALUES (?1)", [local])?;
+
+ // Stored string should be in UTC
+ let s: String = db.one_column("SELECT t FROM foo")?;
+ assert!(s.ends_with("+00:00"));
+
+ let v: DateTime<Local> = db.one_column("SELECT t FROM foo")?;
+ assert_eq!(local, v);
+ Ok(())
+ }
+
+ #[test]
+ fn test_date_time_fixed() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let time = DateTime::parse_from_rfc3339("2020-04-07T11:23:45+04:00").unwrap();
+
+ db.execute("INSERT INTO foo (t) VALUES (?1)", [time])?;
+
+ // Stored string should preserve timezone offset
+ let s: String = db.one_column("SELECT t FROM foo")?;
+ assert!(s.ends_with("+04:00"));
+
+ let v: DateTime<FixedOffset> = db.one_column("SELECT t FROM foo")?;
+ assert_eq!(time.offset(), v.offset());
+ assert_eq!(time, v);
+ Ok(())
+ }
+
+ #[test]
+ fn test_sqlite_functions() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let result: Result<NaiveTime> = db.one_column("SELECT CURRENT_TIME");
+ result.unwrap();
+ let result: Result<NaiveDate> = db.one_column("SELECT CURRENT_DATE");
+ result.unwrap();
+ let result: Result<NaiveDateTime> = db.one_column("SELECT CURRENT_TIMESTAMP");
+ result.unwrap();
+ let result: Result<DateTime<Utc>> = db.one_column("SELECT CURRENT_TIMESTAMP");
+ result.unwrap();
+ Ok(())
+ }
+
+ #[test]
+ fn test_naive_date_time_param() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let result: Result<bool> = db.query_row("SELECT 1 WHERE ?1 BETWEEN datetime('now', '-1 minute') AND datetime('now', '+1 minute')", [Utc::now().naive_utc()], |r| r.get(0));
+ result.unwrap();
+ Ok(())
+ }
+
+ #[test]
+ fn test_date_time_param() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let result: Result<bool> = db.query_row("SELECT 1 WHERE ?1 BETWEEN datetime('now', '-1 minute') AND datetime('now', '+1 minute')", [Utc::now()], |r| r.get(0));
+ result.unwrap();
+ Ok(())
+ }
+
+ #[test]
+ fn test_lenient_parse_timezone() {
+ DateTime::<Utc>::column_result(ValueRef::Text(b"1970-01-01T00:00:00Z")).unwrap();
+ DateTime::<Utc>::column_result(ValueRef::Text(b"1970-01-01T00:00:00+00")).unwrap();
+ }
+}
diff --git a/third_party/rust/rusqlite/src/types/from_sql.rs b/third_party/rust/rusqlite/src/types/from_sql.rs
new file mode 100644
index 0000000000..acbb99ef30
--- /dev/null
+++ b/third_party/rust/rusqlite/src/types/from_sql.rs
@@ -0,0 +1,366 @@
+use super::{Value, ValueRef};
+use std::convert::TryInto;
+use std::error::Error;
+use std::fmt;
+
+/// Enum listing possible errors from [`FromSql`] trait.
+#[derive(Debug)]
+#[non_exhaustive]
+pub enum FromSqlError {
+ /// Error when an SQLite value is requested, but the type of the result
+ /// cannot be converted to the requested Rust type.
+ InvalidType,
+
+ /// Error when the i64 value returned by SQLite cannot be stored into the
+ /// requested type.
+ OutOfRange(i64),
+
+ /// Error when the blob result returned by SQLite cannot be stored into the
+ /// requested type due to a size mismatch.
+ InvalidBlobSize {
+ /// The expected size of the blob.
+ expected_size: usize,
+ /// The actual size of the blob that was returned.
+ blob_size: usize,
+ },
+
+ /// An error case available for implementors of the [`FromSql`] trait.
+ Other(Box<dyn Error + Send + Sync + 'static>),
+}
+
+impl PartialEq for FromSqlError {
+ fn eq(&self, other: &FromSqlError) -> bool {
+ match (self, other) {
+ (FromSqlError::InvalidType, FromSqlError::InvalidType) => true,
+ (FromSqlError::OutOfRange(n1), FromSqlError::OutOfRange(n2)) => n1 == n2,
+ (
+ FromSqlError::InvalidBlobSize {
+ expected_size: es1,
+ blob_size: bs1,
+ },
+ FromSqlError::InvalidBlobSize {
+ expected_size: es2,
+ blob_size: bs2,
+ },
+ ) => es1 == es2 && bs1 == bs2,
+ (..) => false,
+ }
+ }
+}
+
+impl fmt::Display for FromSqlError {
+ fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+ match *self {
+ FromSqlError::InvalidType => write!(f, "Invalid type"),
+ FromSqlError::OutOfRange(i) => write!(f, "Value {i} out of range"),
+ FromSqlError::InvalidBlobSize {
+ expected_size,
+ blob_size,
+ } => {
+ write!(
+ f,
+ "Cannot read {expected_size} byte value out of {blob_size} byte blob"
+ )
+ }
+ FromSqlError::Other(ref err) => err.fmt(f),
+ }
+ }
+}
+
+impl Error for FromSqlError {
+ fn source(&self) -> Option<&(dyn Error + 'static)> {
+ if let FromSqlError::Other(ref err) = self {
+ Some(&**err)
+ } else {
+ None
+ }
+ }
+}
+
+/// Result type for implementors of the [`FromSql`] trait.
+pub type FromSqlResult<T> = Result<T, FromSqlError>;
+
+/// A trait for types that can be created from a SQLite value.
+pub trait FromSql: Sized {
+ /// Converts SQLite value into Rust value.
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self>;
+}
+
+macro_rules! from_sql_integral(
+ ($t:ident) => (
+ impl FromSql for $t {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ let i = i64::column_result(value)?;
+ i.try_into().map_err(|_| FromSqlError::OutOfRange(i))
+ }
+ }
+ );
+ (non_zero $nz:ty, $z:ty) => (
+ impl FromSql for $nz {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ let i = <$z>::column_result(value)?;
+ <$nz>::new(i).ok_or(FromSqlError::OutOfRange(0))
+ }
+ }
+ )
+);
+
+from_sql_integral!(i8);
+from_sql_integral!(i16);
+from_sql_integral!(i32);
+// from_sql_integral!(i64); // Not needed because the native type is i64.
+from_sql_integral!(isize);
+from_sql_integral!(u8);
+from_sql_integral!(u16);
+from_sql_integral!(u32);
+from_sql_integral!(u64);
+from_sql_integral!(usize);
+
+from_sql_integral!(non_zero std::num::NonZeroIsize, isize);
+from_sql_integral!(non_zero std::num::NonZeroI8, i8);
+from_sql_integral!(non_zero std::num::NonZeroI16, i16);
+from_sql_integral!(non_zero std::num::NonZeroI32, i32);
+from_sql_integral!(non_zero std::num::NonZeroI64, i64);
+#[cfg(feature = "i128_blob")]
+#[cfg_attr(docsrs, doc(cfg(feature = "i128_blob")))]
+from_sql_integral!(non_zero std::num::NonZeroI128, i128);
+
+from_sql_integral!(non_zero std::num::NonZeroUsize, usize);
+from_sql_integral!(non_zero std::num::NonZeroU8, u8);
+from_sql_integral!(non_zero std::num::NonZeroU16, u16);
+from_sql_integral!(non_zero std::num::NonZeroU32, u32);
+from_sql_integral!(non_zero std::num::NonZeroU64, u64);
+// std::num::NonZeroU128 is not supported since u128 isn't either
+
+impl FromSql for i64 {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ value.as_i64()
+ }
+}
+
+impl FromSql for f32 {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ match value {
+ ValueRef::Integer(i) => Ok(i as f32),
+ ValueRef::Real(f) => Ok(f as f32),
+ _ => Err(FromSqlError::InvalidType),
+ }
+ }
+}
+
+impl FromSql for f64 {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ match value {
+ ValueRef::Integer(i) => Ok(i as f64),
+ ValueRef::Real(f) => Ok(f),
+ _ => Err(FromSqlError::InvalidType),
+ }
+ }
+}
+
+impl FromSql for bool {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ i64::column_result(value).map(|i| i != 0)
+ }
+}
+
+impl FromSql for String {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ value.as_str().map(ToString::to_string)
+ }
+}
+
+impl FromSql for Box<str> {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ value.as_str().map(Into::into)
+ }
+}
+
+impl FromSql for std::rc::Rc<str> {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ value.as_str().map(Into::into)
+ }
+}
+
+impl FromSql for std::sync::Arc<str> {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ value.as_str().map(Into::into)
+ }
+}
+
+impl FromSql for Vec<u8> {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ value.as_blob().map(<[u8]>::to_vec)
+ }
+}
+
+impl<const N: usize> FromSql for [u8; N] {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ let slice = value.as_blob()?;
+ slice.try_into().map_err(|_| FromSqlError::InvalidBlobSize {
+ expected_size: N,
+ blob_size: slice.len(),
+ })
+ }
+}
+
+#[cfg(feature = "i128_blob")]
+#[cfg_attr(docsrs, doc(cfg(feature = "i128_blob")))]
+impl FromSql for i128 {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ let bytes = <[u8; 16]>::column_result(value)?;
+ Ok(i128::from_be_bytes(bytes) ^ (1_i128 << 127))
+ }
+}
+
+#[cfg(feature = "uuid")]
+#[cfg_attr(docsrs, doc(cfg(feature = "uuid")))]
+impl FromSql for uuid::Uuid {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ let bytes = <[u8; 16]>::column_result(value)?;
+ Ok(uuid::Uuid::from_u128(u128::from_be_bytes(bytes)))
+ }
+}
+
+impl<T: FromSql> FromSql for Option<T> {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ match value {
+ ValueRef::Null => Ok(None),
+ _ => FromSql::column_result(value).map(Some),
+ }
+ }
+}
+
+impl FromSql for Value {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ Ok(value.into())
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use super::FromSql;
+ use crate::{Connection, Error, Result};
+
+ #[test]
+ fn test_integral_ranges() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ fn check_ranges<T>(db: &Connection, out_of_range: &[i64], in_range: &[i64])
+ where
+ T: Into<i64> + FromSql + std::fmt::Debug,
+ {
+ for n in out_of_range {
+ let err = db
+ .query_row("SELECT ?1", [n], |r| r.get::<_, T>(0))
+ .unwrap_err();
+ match err {
+ Error::IntegralValueOutOfRange(_, value) => assert_eq!(*n, value),
+ _ => panic!("unexpected error: {err}"),
+ }
+ }
+ for n in in_range {
+ assert_eq!(
+ *n,
+ db.query_row("SELECT ?1", [n], |r| r.get::<_, T>(0))
+ .unwrap()
+ .into()
+ );
+ }
+ }
+
+ check_ranges::<i8>(&db, &[-129, 128], &[-128, 0, 1, 127]);
+ check_ranges::<i16>(&db, &[-32769, 32768], &[-32768, -1, 0, 1, 32767]);
+ check_ranges::<i32>(
+ &db,
+ &[-2_147_483_649, 2_147_483_648],
+ &[-2_147_483_648, -1, 0, 1, 2_147_483_647],
+ );
+ check_ranges::<u8>(&db, &[-2, -1, 256], &[0, 1, 255]);
+ check_ranges::<u16>(&db, &[-2, -1, 65536], &[0, 1, 65535]);
+ check_ranges::<u32>(&db, &[-2, -1, 4_294_967_296], &[0, 1, 4_294_967_295]);
+ Ok(())
+ }
+
+ #[test]
+ fn test_nonzero_ranges() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ macro_rules! check_ranges {
+ ($nz:ty, $out_of_range:expr, $in_range:expr) => {
+ for &n in $out_of_range {
+ assert_eq!(
+ db.query_row("SELECT ?1", [n], |r| r.get::<_, $nz>(0)),
+ Err(Error::IntegralValueOutOfRange(0, n)),
+ "{}",
+ std::any::type_name::<$nz>()
+ );
+ }
+ for &n in $in_range {
+ let non_zero = <$nz>::new(n).unwrap();
+ assert_eq!(
+ Ok(non_zero),
+ db.query_row("SELECT ?1", [non_zero], |r| r.get::<_, $nz>(0))
+ );
+ }
+ };
+ }
+
+ check_ranges!(std::num::NonZeroI8, &[0, -129, 128], &[-128, 1, 127]);
+ check_ranges!(
+ std::num::NonZeroI16,
+ &[0, -32769, 32768],
+ &[-32768, -1, 1, 32767]
+ );
+ check_ranges!(
+ std::num::NonZeroI32,
+ &[0, -2_147_483_649, 2_147_483_648],
+ &[-2_147_483_648, -1, 1, 2_147_483_647]
+ );
+ check_ranges!(
+ std::num::NonZeroI64,
+ &[0],
+ &[-2_147_483_648, -1, 1, 2_147_483_647, i64::MAX, i64::MIN]
+ );
+ check_ranges!(
+ std::num::NonZeroIsize,
+ &[0],
+ &[-2_147_483_648, -1, 1, 2_147_483_647]
+ );
+ check_ranges!(std::num::NonZeroU8, &[0, -2, -1, 256], &[1, 255]);
+ check_ranges!(std::num::NonZeroU16, &[0, -2, -1, 65536], &[1, 65535]);
+ check_ranges!(
+ std::num::NonZeroU32,
+ &[0, -2, -1, 4_294_967_296],
+ &[1, 4_294_967_295]
+ );
+ check_ranges!(
+ std::num::NonZeroU64,
+ &[0, -2, -1, -4_294_967_296],
+ &[1, 4_294_967_295, i64::MAX as u64]
+ );
+ check_ranges!(
+ std::num::NonZeroUsize,
+ &[0, -2, -1, -4_294_967_296],
+ &[1, 4_294_967_295]
+ );
+
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/types/mod.rs b/third_party/rust/rusqlite/src/types/mod.rs
new file mode 100644
index 0000000000..4dbc19dd95
--- /dev/null
+++ b/third_party/rust/rusqlite/src/types/mod.rs
@@ -0,0 +1,447 @@
+//! Traits dealing with SQLite data types.
+//!
+//! SQLite uses a [dynamic type system](https://www.sqlite.org/datatype3.html). Implementations of
+//! the [`ToSql`] and [`FromSql`] traits are provided for the basic types that
+//! SQLite provides methods for:
+//!
+//! * Strings (`String` and `&str`)
+//! * Blobs (`Vec<u8>` and `&[u8]`)
+//! * Numbers
+//!
+//! The number situation is a little complicated due to the fact that all
+//! numbers in SQLite are stored as `INTEGER` (`i64`) or `REAL` (`f64`).
+//!
+//! [`ToSql`] and [`FromSql`] are implemented for all primitive number types.
+//! [`FromSql`] has different behaviour depending on the SQL and Rust types, and
+//! the value.
+//!
+//! * `INTEGER` to integer: returns an
+//! [`Error::IntegralValueOutOfRange`](crate::Error::IntegralValueOutOfRange)
+//! error if the value does not fit in the Rust type.
+//! * `REAL` to integer: always returns an
+//! [`Error::InvalidColumnType`](crate::Error::InvalidColumnType) error.
+//! * `INTEGER` to float: casts using `as` operator. Never fails.
+//! * `REAL` to float: casts using `as` operator. Never fails.
+//!
+//! [`ToSql`] always succeeds except when storing a `u64` or `usize` value that
+//! cannot fit in an `INTEGER` (`i64`). Also note that SQLite ignores column
+//! types, so if you store an `i64` in a column with type `REAL` it will be
+//! stored as an `INTEGER`, not a `REAL` (unless the column is part of a
+//! [STRICT table](https://www.sqlite.org/stricttables.html)).
+//!
+//! If the `time` feature is enabled, implementations are
+//! provided for `time::OffsetDateTime` that use the RFC 3339 date/time format,
+//! `"%Y-%m-%dT%H:%M:%S.%fZ"`, to store time values as strings. These values
+//! can be parsed by SQLite's builtin
+//! [datetime](https://www.sqlite.org/lang_datefunc.html) functions. If you
+//! want different storage for datetimes, you can use a newtype.
+#![cfg_attr(
+ feature = "time",
+ doc = r##"
+For example, to store datetimes as `i64`s counting the number of seconds since
+the Unix epoch:
+
+```
+use rusqlite::types::{FromSql, FromSqlError, FromSqlResult, ToSql, ToSqlOutput, ValueRef};
+use rusqlite::Result;
+
+pub struct DateTimeSql(pub time::OffsetDateTime);
+
+impl FromSql for DateTimeSql {
+ fn column_result(value: ValueRef) -> FromSqlResult<Self> {
+ i64::column_result(value).and_then(|as_i64| {
+ time::OffsetDateTime::from_unix_timestamp(as_i64)
+ .map(|odt| DateTimeSql(odt))
+ .map_err(|err| FromSqlError::Other(Box::new(err)))
+ })
+ }
+}
+
+impl ToSql for DateTimeSql {
+ fn to_sql(&self) -> Result<ToSqlOutput> {
+ Ok(self.0.unix_timestamp().into())
+ }
+}
+```
+
+"##
+)]
+//! [`ToSql`] and [`FromSql`] are also implemented for `Option<T>` where `T`
+//! implements [`ToSql`] or [`FromSql`] for the cases where you want to know if
+//! a value was NULL (which gets translated to `None`).
+
+pub use self::from_sql::{FromSql, FromSqlError, FromSqlResult};
+pub use self::to_sql::{ToSql, ToSqlOutput};
+pub use self::value::Value;
+pub use self::value_ref::ValueRef;
+
+use std::fmt;
+
+#[cfg(feature = "chrono")]
+#[cfg_attr(docsrs, doc(cfg(feature = "chrono")))]
+mod chrono;
+mod from_sql;
+#[cfg(feature = "serde_json")]
+#[cfg_attr(docsrs, doc(cfg(feature = "serde_json")))]
+mod serde_json;
+#[cfg(feature = "time")]
+#[cfg_attr(docsrs, doc(cfg(feature = "time")))]
+mod time;
+mod to_sql;
+#[cfg(feature = "url")]
+#[cfg_attr(docsrs, doc(cfg(feature = "url")))]
+mod url;
+mod value;
+mod value_ref;
+
+/// Empty struct that can be used to fill in a query parameter as `NULL`.
+///
+/// ## Example
+///
+/// ```rust,no_run
+/// # use rusqlite::{Connection, Result};
+/// # use rusqlite::types::{Null};
+///
+/// fn insert_null(conn: &Connection) -> Result<usize> {
+/// conn.execute("INSERT INTO people (name) VALUES (?1)", [Null])
+/// }
+/// ```
+#[derive(Copy, Clone)]
+pub struct Null;
+
+/// SQLite data types.
+/// See [Fundamental Datatypes](https://sqlite.org/c3ref/c_blob.html).
+#[derive(Clone, Debug, PartialEq, Eq)]
+pub enum Type {
+ /// NULL
+ Null,
+ /// 64-bit signed integer
+ Integer,
+ /// 64-bit IEEE floating point number
+ Real,
+ /// String
+ Text,
+ /// BLOB
+ Blob,
+}
+
+impl fmt::Display for Type {
+ fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+ match *self {
+ Type::Null => f.pad("Null"),
+ Type::Integer => f.pad("Integer"),
+ Type::Real => f.pad("Real"),
+ Type::Text => f.pad("Text"),
+ Type::Blob => f.pad("Blob"),
+ }
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use super::Value;
+ use crate::{params, Connection, Error, Result, Statement};
+ use std::os::raw::{c_double, c_int};
+
+ fn checked_memory_handle() -> Result<Connection> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo (b BLOB, t TEXT, i INTEGER, f FLOAT, n)")?;
+ Ok(db)
+ }
+
+ #[test]
+ fn test_blob() -> Result<()> {
+ let db = checked_memory_handle()?;
+
+ let v1234 = vec![1u8, 2, 3, 4];
+ db.execute("INSERT INTO foo(b) VALUES (?1)", [&v1234])?;
+
+ let v: Vec<u8> = db.one_column("SELECT b FROM foo")?;
+ assert_eq!(v, v1234);
+ Ok(())
+ }
+
+ #[test]
+ fn test_empty_blob() -> Result<()> {
+ let db = checked_memory_handle()?;
+
+ let empty = vec![];
+ db.execute("INSERT INTO foo(b) VALUES (?1)", [&empty])?;
+
+ let v: Vec<u8> = db.one_column("SELECT b FROM foo")?;
+ assert_eq!(v, empty);
+ Ok(())
+ }
+
+ #[test]
+ fn test_str() -> Result<()> {
+ let db = checked_memory_handle()?;
+
+ let s = "hello, world!";
+ db.execute("INSERT INTO foo(t) VALUES (?1)", [&s])?;
+
+ let from: String = db.one_column("SELECT t FROM foo")?;
+ assert_eq!(from, s);
+ Ok(())
+ }
+
+ #[test]
+ fn test_string() -> Result<()> {
+ let db = checked_memory_handle()?;
+
+ let s = "hello, world!";
+ db.execute("INSERT INTO foo(t) VALUES (?1)", [s.to_owned()])?;
+
+ let from: String = db.one_column("SELECT t FROM foo")?;
+ assert_eq!(from, s);
+ Ok(())
+ }
+
+ #[test]
+ fn test_value() -> Result<()> {
+ let db = checked_memory_handle()?;
+
+ db.execute("INSERT INTO foo(i) VALUES (?1)", [Value::Integer(10)])?;
+
+ assert_eq!(10i64, db.one_column::<i64>("SELECT i FROM foo")?);
+ Ok(())
+ }
+
+ #[test]
+ fn test_option() -> Result<()> {
+ let db = checked_memory_handle()?;
+
+ let s = "hello, world!";
+ let b = Some(vec![1u8, 2, 3, 4]);
+
+ db.execute("INSERT INTO foo(t) VALUES (?1)", [Some(s)])?;
+ db.execute("INSERT INTO foo(b) VALUES (?1)", [&b])?;
+
+ let mut stmt = db.prepare("SELECT t, b FROM foo ORDER BY ROWID ASC")?;
+ let mut rows = stmt.query([])?;
+
+ {
+ let row1 = rows.next()?.unwrap();
+ let s1: Option<String> = row1.get_unwrap(0);
+ let b1: Option<Vec<u8>> = row1.get_unwrap(1);
+ assert_eq!(s, s1.unwrap());
+ assert!(b1.is_none());
+ }
+
+ {
+ let row2 = rows.next()?.unwrap();
+ let s2: Option<String> = row2.get_unwrap(0);
+ let b2: Option<Vec<u8>> = row2.get_unwrap(1);
+ assert!(s2.is_none());
+ assert_eq!(b, b2);
+ }
+ Ok(())
+ }
+
+ #[test]
+ #[allow(clippy::cognitive_complexity)]
+ fn test_mismatched_types() -> Result<()> {
+ fn is_invalid_column_type(err: Error) -> bool {
+ matches!(err, Error::InvalidColumnType(..))
+ }
+
+ let db = checked_memory_handle()?;
+
+ db.execute(
+ "INSERT INTO foo(b, t, i, f) VALUES (X'0102', 'text', 1, 1.5)",
+ [],
+ )?;
+
+ let mut stmt = db.prepare("SELECT b, t, i, f, n FROM foo")?;
+ let mut rows = stmt.query([])?;
+
+ let row = rows.next()?.unwrap();
+
+ // check the correct types come back as expected
+ assert_eq!(vec![1, 2], row.get::<_, Vec<u8>>(0)?);
+ assert_eq!("text", row.get::<_, String>(1)?);
+ assert_eq!(1, row.get::<_, c_int>(2)?);
+ assert!((1.5 - row.get::<_, c_double>(3)?).abs() < f64::EPSILON);
+ assert_eq!(row.get::<_, Option<c_int>>(4)?, None);
+ assert_eq!(row.get::<_, Option<c_double>>(4)?, None);
+ assert_eq!(row.get::<_, Option<String>>(4)?, None);
+
+ // check some invalid types
+
+ // 0 is actually a blob (Vec<u8>)
+ assert!(is_invalid_column_type(row.get::<_, c_int>(0).unwrap_err()));
+ assert!(is_invalid_column_type(row.get::<_, c_int>(0).unwrap_err()));
+ assert!(is_invalid_column_type(row.get::<_, i64>(0).err().unwrap()));
+ assert!(is_invalid_column_type(
+ row.get::<_, c_double>(0).unwrap_err()
+ ));
+ assert!(is_invalid_column_type(row.get::<_, String>(0).unwrap_err()));
+ #[cfg(feature = "time")]
+ assert!(is_invalid_column_type(
+ row.get::<_, time::OffsetDateTime>(0).unwrap_err()
+ ));
+ assert!(is_invalid_column_type(
+ row.get::<_, Option<c_int>>(0).unwrap_err()
+ ));
+
+ // 1 is actually a text (String)
+ assert!(is_invalid_column_type(row.get::<_, c_int>(1).unwrap_err()));
+ assert!(is_invalid_column_type(row.get::<_, i64>(1).err().unwrap()));
+ assert!(is_invalid_column_type(
+ row.get::<_, c_double>(1).unwrap_err()
+ ));
+ assert!(is_invalid_column_type(
+ row.get::<_, Vec<u8>>(1).unwrap_err()
+ ));
+ assert!(is_invalid_column_type(
+ row.get::<_, Option<c_int>>(1).unwrap_err()
+ ));
+
+ // 2 is actually an integer
+ assert!(is_invalid_column_type(row.get::<_, String>(2).unwrap_err()));
+ assert!(is_invalid_column_type(
+ row.get::<_, Vec<u8>>(2).unwrap_err()
+ ));
+ assert!(is_invalid_column_type(
+ row.get::<_, Option<String>>(2).unwrap_err()
+ ));
+
+ // 3 is actually a float (c_double)
+ assert!(is_invalid_column_type(row.get::<_, c_int>(3).unwrap_err()));
+ assert!(is_invalid_column_type(row.get::<_, i64>(3).err().unwrap()));
+ assert!(is_invalid_column_type(row.get::<_, String>(3).unwrap_err()));
+ assert!(is_invalid_column_type(
+ row.get::<_, Vec<u8>>(3).unwrap_err()
+ ));
+ assert!(is_invalid_column_type(
+ row.get::<_, Option<c_int>>(3).unwrap_err()
+ ));
+
+ // 4 is actually NULL
+ assert!(is_invalid_column_type(row.get::<_, c_int>(4).unwrap_err()));
+ assert!(is_invalid_column_type(row.get::<_, i64>(4).err().unwrap()));
+ assert!(is_invalid_column_type(
+ row.get::<_, c_double>(4).unwrap_err()
+ ));
+ assert!(is_invalid_column_type(row.get::<_, String>(4).unwrap_err()));
+ assert!(is_invalid_column_type(
+ row.get::<_, Vec<u8>>(4).unwrap_err()
+ ));
+ #[cfg(feature = "time")]
+ assert!(is_invalid_column_type(
+ row.get::<_, time::OffsetDateTime>(4).unwrap_err()
+ ));
+ Ok(())
+ }
+
+ #[test]
+ fn test_dynamic_type() -> Result<()> {
+ use super::Value;
+ let db = checked_memory_handle()?;
+
+ db.execute(
+ "INSERT INTO foo(b, t, i, f) VALUES (X'0102', 'text', 1, 1.5)",
+ [],
+ )?;
+
+ let mut stmt = db.prepare("SELECT b, t, i, f, n FROM foo")?;
+ let mut rows = stmt.query([])?;
+
+ let row = rows.next()?.unwrap();
+ assert_eq!(Value::Blob(vec![1, 2]), row.get::<_, Value>(0)?);
+ assert_eq!(Value::Text(String::from("text")), row.get::<_, Value>(1)?);
+ assert_eq!(Value::Integer(1), row.get::<_, Value>(2)?);
+ match row.get::<_, Value>(3)? {
+ Value::Real(val) => assert!((1.5 - val).abs() < f64::EPSILON),
+ x => panic!("Invalid Value {x:?}"),
+ }
+ assert_eq!(Value::Null, row.get::<_, Value>(4)?);
+ Ok(())
+ }
+
+ macro_rules! test_conversion {
+ ($db_etc:ident, $insert_value:expr, $get_type:ty,expect $expected_value:expr) => {
+ $db_etc.insert_statement.execute(params![$insert_value])?;
+ let res = $db_etc
+ .query_statement
+ .query_row([], |row| row.get::<_, $get_type>(0));
+ assert_eq!(res?, $expected_value);
+ $db_etc.delete_statement.execute([])?;
+ };
+ ($db_etc:ident, $insert_value:expr, $get_type:ty,expect_from_sql_error) => {
+ $db_etc.insert_statement.execute(params![$insert_value])?;
+ let res = $db_etc
+ .query_statement
+ .query_row([], |row| row.get::<_, $get_type>(0));
+ res.unwrap_err();
+ $db_etc.delete_statement.execute([])?;
+ };
+ ($db_etc:ident, $insert_value:expr, $get_type:ty,expect_to_sql_error) => {
+ $db_etc
+ .insert_statement
+ .execute(params![$insert_value])
+ .unwrap_err();
+ };
+ }
+
+ #[test]
+ fn test_numeric_conversions() -> Result<()> {
+ #![allow(clippy::float_cmp)]
+
+ // Test what happens when we store an f32 and retrieve an i32 etc.
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo (x)")?;
+
+ // SQLite actually ignores the column types, so we just need to test
+ // different numeric values.
+
+ struct DbEtc<'conn> {
+ insert_statement: Statement<'conn>,
+ query_statement: Statement<'conn>,
+ delete_statement: Statement<'conn>,
+ }
+
+ let mut db_etc = DbEtc {
+ insert_statement: db.prepare("INSERT INTO foo VALUES (?1)")?,
+ query_statement: db.prepare("SELECT x FROM foo")?,
+ delete_statement: db.prepare("DELETE FROM foo")?,
+ };
+
+ // Basic non-converting test.
+ test_conversion!(db_etc, 0u8, u8, expect 0u8);
+
+ // In-range integral conversions.
+ test_conversion!(db_etc, 100u8, i8, expect 100i8);
+ test_conversion!(db_etc, 200u8, u8, expect 200u8);
+ test_conversion!(db_etc, 100u16, i8, expect 100i8);
+ test_conversion!(db_etc, 200u16, u8, expect 200u8);
+ test_conversion!(db_etc, u32::MAX, u64, expect u32::MAX as u64);
+ test_conversion!(db_etc, i64::MIN, i64, expect i64::MIN);
+ test_conversion!(db_etc, i64::MAX, i64, expect i64::MAX);
+ test_conversion!(db_etc, i64::MAX, u64, expect i64::MAX as u64);
+ test_conversion!(db_etc, 100usize, usize, expect 100usize);
+ test_conversion!(db_etc, 100u64, u64, expect 100u64);
+ test_conversion!(db_etc, i64::MAX as u64, u64, expect i64::MAX as u64);
+
+ // Out-of-range integral conversions.
+ test_conversion!(db_etc, 200u8, i8, expect_from_sql_error);
+ test_conversion!(db_etc, 400u16, i8, expect_from_sql_error);
+ test_conversion!(db_etc, 400u16, u8, expect_from_sql_error);
+ test_conversion!(db_etc, -1i8, u8, expect_from_sql_error);
+ test_conversion!(db_etc, i64::MIN, u64, expect_from_sql_error);
+ test_conversion!(db_etc, u64::MAX, i64, expect_to_sql_error);
+ test_conversion!(db_etc, u64::MAX, u64, expect_to_sql_error);
+ test_conversion!(db_etc, i64::MAX as u64 + 1, u64, expect_to_sql_error);
+
+ // FromSql integer to float, always works.
+ test_conversion!(db_etc, i64::MIN, f32, expect i64::MIN as f32);
+ test_conversion!(db_etc, i64::MAX, f32, expect i64::MAX as f32);
+ test_conversion!(db_etc, i64::MIN, f64, expect i64::MIN as f64);
+ test_conversion!(db_etc, i64::MAX, f64, expect i64::MAX as f64);
+
+ // FromSql float to int conversion, never works even if the actual value
+ // is an integer.
+ test_conversion!(db_etc, 0f64, i64, expect_from_sql_error);
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/types/serde_json.rs b/third_party/rust/rusqlite/src/types/serde_json.rs
new file mode 100644
index 0000000000..6e38ba36ba
--- /dev/null
+++ b/third_party/rust/rusqlite/src/types/serde_json.rs
@@ -0,0 +1,135 @@
+//! [`ToSql`] and [`FromSql`] implementation for JSON `Value`.
+
+use serde_json::{Number, Value};
+
+use crate::types::{FromSql, FromSqlError, FromSqlResult, ToSql, ToSqlOutput, ValueRef};
+use crate::{Error, Result};
+
+/// Serialize JSON `Value` to text:
+///
+///
+/// | JSON | SQLite |
+/// |----------|---------|
+/// | Null | NULL |
+/// | Bool | 'true' / 'false' |
+/// | Number | INT or REAL except u64 |
+/// | _ | TEXT |
+impl ToSql for Value {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ match self {
+ Value::Null => Ok(ToSqlOutput::Borrowed(ValueRef::Null)),
+ Value::Number(n) if n.is_i64() => Ok(ToSqlOutput::from(n.as_i64().unwrap())),
+ Value::Number(n) if n.is_f64() => Ok(ToSqlOutput::from(n.as_f64().unwrap())),
+ _ => serde_json::to_string(self)
+ .map(ToSqlOutput::from)
+ .map_err(|err| Error::ToSqlConversionFailure(err.into())),
+ }
+ }
+}
+
+/// Deserialize SQLite value to JSON `Value`:
+///
+/// | SQLite | JSON |
+/// |----------|---------|
+/// | NULL | Null |
+/// | 'null' | Null |
+/// | 'true' | Bool |
+/// | 1 | Number |
+/// | 0.1 | Number |
+/// | '"text"' | String |
+/// | 'text' | _Error_ |
+/// | '[0, 1]' | Array |
+/// | '{"x": 1}' | Object |
+impl FromSql for Value {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ match value {
+ ValueRef::Text(s) => serde_json::from_slice(s), // KO for b"text"
+ ValueRef::Blob(b) => serde_json::from_slice(b),
+ ValueRef::Integer(i) => Ok(Value::Number(Number::from(i))),
+ ValueRef::Real(f) => {
+ match Number::from_f64(f) {
+ Some(n) => Ok(Value::Number(n)),
+ _ => return Err(FromSqlError::InvalidType), // FIXME
+ }
+ }
+ ValueRef::Null => Ok(Value::Null),
+ }
+ .map_err(|err| FromSqlError::Other(Box::new(err)))
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use crate::types::ToSql;
+ use crate::{Connection, Result};
+ use serde_json::{Number, Value};
+
+ fn checked_memory_handle() -> Result<Connection> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo (t TEXT, b BLOB)")?;
+ Ok(db)
+ }
+
+ #[test]
+ fn test_json_value() -> Result<()> {
+ let db = checked_memory_handle()?;
+
+ let json = r#"{"foo": 13, "bar": "baz"}"#;
+ let data: Value = serde_json::from_str(json).unwrap();
+ db.execute(
+ "INSERT INTO foo (t, b) VALUES (?1, ?2)",
+ [&data as &dyn ToSql, &json.as_bytes()],
+ )?;
+
+ let t: Value = db.one_column("SELECT t FROM foo")?;
+ assert_eq!(data, t);
+ let b: Value = db.one_column("SELECT b FROM foo")?;
+ assert_eq!(data, b);
+ Ok(())
+ }
+
+ #[test]
+ fn test_to_sql() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ let v: Option<String> = db.query_row("SELECT ?", [Value::Null], |r| r.get(0))?;
+ assert_eq!(None, v);
+ let v: String = db.query_row("SELECT ?", [Value::Bool(true)], |r| r.get(0))?;
+ assert_eq!("true", v);
+ let v: i64 = db.query_row("SELECT ?", [Value::Number(Number::from(1))], |r| r.get(0))?;
+ assert_eq!(1, v);
+ let v: f64 = db.query_row(
+ "SELECT ?",
+ [Value::Number(Number::from_f64(0.1).unwrap())],
+ |r| r.get(0),
+ )?;
+ assert_eq!(0.1, v);
+ let v: String =
+ db.query_row("SELECT ?", [Value::String("text".to_owned())], |r| r.get(0))?;
+ assert_eq!("\"text\"", v);
+ Ok(())
+ }
+
+ #[test]
+ fn test_from_sql() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+
+ let v: Value = db.one_column("SELECT NULL")?;
+ assert_eq!(Value::Null, v);
+ let v: Value = db.one_column("SELECT 'null'")?;
+ assert_eq!(Value::Null, v);
+ let v: Value = db.one_column("SELECT 'true'")?;
+ assert_eq!(Value::Bool(true), v);
+ let v: Value = db.one_column("SELECT 1")?;
+ assert_eq!(Value::Number(Number::from(1)), v);
+ let v: Value = db.one_column("SELECT 0.1")?;
+ assert_eq!(Value::Number(Number::from_f64(0.1).unwrap()), v);
+ let v: Value = db.one_column("SELECT '\"text\"'")?;
+ assert_eq!(Value::String("text".to_owned()), v);
+ let v: Result<Value> = db.one_column("SELECT 'text'");
+ assert!(v.is_err());
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/types/time.rs b/third_party/rust/rusqlite/src/types/time.rs
new file mode 100644
index 0000000000..265d6c6bd9
--- /dev/null
+++ b/third_party/rust/rusqlite/src/types/time.rs
@@ -0,0 +1,426 @@
+//! Convert formats 1-10 in [Time Values](https://sqlite.org/lang_datefunc.html#time_values) to time types.
+//! [`ToSql`] and [`FromSql`] implementation for [`time::OffsetDateTime`].
+//! [`ToSql`] and [`FromSql`] implementation for [`time::PrimitiveDateTime`].
+//! [`ToSql`] and [`FromSql`] implementation for [`time::Date`].
+//! [`ToSql`] and [`FromSql`] implementation for [`time::Time`].
+//! Time Strings in:
+//! - Format 2: "YYYY-MM-DD HH:MM"
+//! - Format 5: "YYYY-MM-DDTHH:MM"
+//! - Format 8: "HH:MM"
+//! without an explicit second value will assume 0 seconds.
+//! Time String that contain an optional timezone without an explicit date are unsupported.
+//! All other assumptions described in [Time Values](https://sqlite.org/lang_datefunc.html#time_values) section are unsupported.
+
+use crate::types::{FromSql, FromSqlError, FromSqlResult, ToSql, ToSqlOutput, ValueRef};
+use crate::{Error, Result};
+use time::format_description::FormatItem;
+use time::macros::format_description;
+use time::{Date, OffsetDateTime, PrimitiveDateTime, Time};
+
+const OFFSET_DATE_TIME_ENCODING: &[FormatItem<'_>] = format_description!(
+ version = 2,
+ "[year]-[month]-[day] [hour]:[minute]:[second].[subsecond][offset_hour sign:mandatory]:[offset_minute]"
+);
+const PRIMITIVE_DATE_TIME_ENCODING: &[FormatItem<'_>] = format_description!(
+ version = 2,
+ "[year]-[month]-[day] [hour]:[minute]:[second].[subsecond]"
+);
+const TIME_ENCODING: &[FormatItem<'_>] =
+ format_description!(version = 2, "[hour]:[minute]:[second].[subsecond]");
+
+const DATE_FORMAT: &[FormatItem<'_>] = format_description!(version = 2, "[year]-[month]-[day]");
+const TIME_FORMAT: &[FormatItem<'_>] = format_description!(
+ version = 2,
+ "[hour]:[minute][optional [:[second][optional [.[subsecond]]]]]"
+);
+const PRIMITIVE_DATE_TIME_FORMAT: &[FormatItem<'_>] = format_description!(
+ version = 2,
+ "[year]-[month]-[day][first [ ][T]][hour]:[minute][optional [:[second][optional [.[subsecond]]]]]"
+);
+const UTC_DATE_TIME_FORMAT: &[FormatItem<'_>] = format_description!(
+ version = 2,
+ "[year]-[month]-[day][first [ ][T]][hour]:[minute][optional [:[second][optional [.[subsecond]]]]][optional [Z]]"
+);
+const OFFSET_DATE_TIME_FORMAT: &[FormatItem<'_>] = format_description!(
+ version = 2,
+ "[year]-[month]-[day][first [ ][T]][hour]:[minute][optional [:[second][optional [.[subsecond]]]]][offset_hour sign:mandatory]:[offset_minute]"
+);
+const LEGACY_DATE_TIME_FORMAT: &[FormatItem<'_>] = format_description!(
+ version = 2,
+ "[year]-[month]-[day] [hour]:[minute]:[second]:[subsecond] [offset_hour sign:mandatory]:[offset_minute]"
+);
+
+/// OffsetDatetime => RFC3339 format ("YYYY-MM-DD HH:MM:SS.SSS[+-]HH:MM")
+impl ToSql for OffsetDateTime {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ let time_string = self
+ .format(&OFFSET_DATE_TIME_ENCODING)
+ .map_err(|err| Error::ToSqlConversionFailure(err.into()))?;
+ Ok(ToSqlOutput::from(time_string))
+ }
+}
+
+// Supports parsing formats 2-7 from https://www.sqlite.org/lang_datefunc.html
+// Formats 2-7 without a timezone assumes UTC
+impl FromSql for OffsetDateTime {
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ value.as_str().and_then(|s| {
+ if let Some(b' ') = s.as_bytes().get(23) {
+ // legacy
+ return OffsetDateTime::parse(s, &LEGACY_DATE_TIME_FORMAT)
+ .map_err(|err| FromSqlError::Other(Box::new(err)));
+ }
+ if s[8..].contains('+') || s[8..].contains('-') {
+ // Formats 2-7 with timezone
+ return OffsetDateTime::parse(s, &OFFSET_DATE_TIME_FORMAT)
+ .map_err(|err| FromSqlError::Other(Box::new(err)));
+ }
+ // Formats 2-7 without timezone
+ PrimitiveDateTime::parse(s, &UTC_DATE_TIME_FORMAT)
+ .map(|p| p.assume_utc())
+ .map_err(|err| FromSqlError::Other(Box::new(err)))
+ })
+ }
+}
+
+/// ISO 8601 calendar date without timezone => "YYYY-MM-DD"
+impl ToSql for Date {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ let date_str = self
+ .format(&DATE_FORMAT)
+ .map_err(|err| Error::ToSqlConversionFailure(err.into()))?;
+ Ok(ToSqlOutput::from(date_str))
+ }
+}
+
+/// "YYYY-MM-DD" => ISO 8601 calendar date without timezone.
+impl FromSql for Date {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ value.as_str().and_then(|s| {
+ Date::parse(s, &DATE_FORMAT).map_err(|err| FromSqlError::Other(err.into()))
+ })
+ }
+}
+
+/// ISO 8601 time without timezone => "HH:MM:SS.SSS"
+impl ToSql for Time {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ let time_str = self
+ .format(&TIME_ENCODING)
+ .map_err(|err| Error::ToSqlConversionFailure(err.into()))?;
+ Ok(ToSqlOutput::from(time_str))
+ }
+}
+
+/// "HH:MM"/"HH:MM:SS"/"HH:MM:SS.SSS" => ISO 8601 time without timezone.
+impl FromSql for Time {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ value.as_str().and_then(|s| {
+ Time::parse(s, &TIME_FORMAT).map_err(|err| FromSqlError::Other(err.into()))
+ })
+ }
+}
+
+/// ISO 8601 combined date and time without timezone => "YYYY-MM-DD HH:MM:SS.SSS"
+impl ToSql for PrimitiveDateTime {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ let date_time_str = self
+ .format(&PRIMITIVE_DATE_TIME_ENCODING)
+ .map_err(|err| Error::ToSqlConversionFailure(err.into()))?;
+ Ok(ToSqlOutput::from(date_time_str))
+ }
+}
+
+/// YYYY-MM-DD HH:MM
+/// YYYY-MM-DDTHH:MM
+/// YYYY-MM-DD HH:MM:SS
+/// YYYY-MM-DDTHH:MM:SS
+/// YYYY-MM-DD HH:MM:SS.SSS
+/// YYYY-MM-DDTHH:MM:SS.SSS
+/// => ISO 8601 combined date and time with timezone
+impl FromSql for PrimitiveDateTime {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ value.as_str().and_then(|s| {
+ PrimitiveDateTime::parse(s, &PRIMITIVE_DATE_TIME_FORMAT)
+ .map_err(|err| FromSqlError::Other(err.into()))
+ })
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use crate::{Connection, Result};
+ use time::macros::{date, datetime, time};
+ use time::{Date, OffsetDateTime, PrimitiveDateTime, Time};
+
+ fn checked_memory_handle() -> Result<Connection> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo (t TEXT, i INTEGER, f FLOAT, b BLOB)")?;
+ Ok(db)
+ }
+
+ #[test]
+ fn test_offset_date_time() -> Result<()> {
+ let db = checked_memory_handle()?;
+
+ let mut ts_vec = vec![];
+
+ let make_datetime = |secs: i128, nanos: i128| {
+ OffsetDateTime::from_unix_timestamp_nanos(1_000_000_000 * secs + nanos).unwrap()
+ };
+
+ ts_vec.push(make_datetime(10_000, 0)); //January 1, 1970 2:46:40 AM
+ ts_vec.push(make_datetime(10_000, 1000)); //January 1, 1970 2:46:40 AM (and one microsecond)
+ ts_vec.push(make_datetime(1_500_391_124, 1_000_000)); //July 18, 2017
+ ts_vec.push(make_datetime(2_000_000_000, 2_000_000)); //May 18, 2033
+ ts_vec.push(make_datetime(3_000_000_000, 999_999_999)); //January 24, 2065
+ ts_vec.push(make_datetime(10_000_000_000, 0)); //November 20, 2286
+
+ for ts in ts_vec {
+ db.execute("INSERT INTO foo(t) VALUES (?1)", [ts])?;
+
+ let from: OffsetDateTime = db.one_column("SELECT t FROM foo")?;
+
+ db.execute("DELETE FROM foo", [])?;
+
+ assert_eq!(from, ts);
+ }
+ Ok(())
+ }
+
+ #[test]
+ fn test_offset_date_time_parsing() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let tests = vec![
+ // Rfc3339
+ (
+ "2013-10-07T08:23:19.123456789Z",
+ datetime!(2013-10-07 8:23:19.123456789 UTC),
+ ),
+ (
+ "2013-10-07 08:23:19.123456789Z",
+ datetime!(2013-10-07 8:23:19.123456789 UTC),
+ ),
+ // Format 2
+ ("2013-10-07 08:23", datetime!(2013-10-07 8:23 UTC)),
+ ("2013-10-07 08:23Z", datetime!(2013-10-07 8:23 UTC)),
+ ("2013-10-07 08:23+04:00", datetime!(2013-10-07 8:23 +4)),
+ // Format 3
+ ("2013-10-07 08:23:19", datetime!(2013-10-07 8:23:19 UTC)),
+ ("2013-10-07 08:23:19Z", datetime!(2013-10-07 8:23:19 UTC)),
+ (
+ "2013-10-07 08:23:19+04:00",
+ datetime!(2013-10-07 8:23:19 +4),
+ ),
+ // Format 4
+ (
+ "2013-10-07 08:23:19.123",
+ datetime!(2013-10-07 8:23:19.123 UTC),
+ ),
+ (
+ "2013-10-07 08:23:19.123Z",
+ datetime!(2013-10-07 8:23:19.123 UTC),
+ ),
+ (
+ "2013-10-07 08:23:19.123+04:00",
+ datetime!(2013-10-07 8:23:19.123 +4),
+ ),
+ // Format 5
+ ("2013-10-07T08:23", datetime!(2013-10-07 8:23 UTC)),
+ ("2013-10-07T08:23Z", datetime!(2013-10-07 8:23 UTC)),
+ ("2013-10-07T08:23+04:00", datetime!(2013-10-07 8:23 +4)),
+ // Format 6
+ ("2013-10-07T08:23:19", datetime!(2013-10-07 8:23:19 UTC)),
+ ("2013-10-07T08:23:19Z", datetime!(2013-10-07 8:23:19 UTC)),
+ (
+ "2013-10-07T08:23:19+04:00",
+ datetime!(2013-10-07 8:23:19 +4),
+ ),
+ // Format 7
+ (
+ "2013-10-07T08:23:19.123",
+ datetime!(2013-10-07 8:23:19.123 UTC),
+ ),
+ (
+ "2013-10-07T08:23:19.123Z",
+ datetime!(2013-10-07 8:23:19.123 UTC),
+ ),
+ (
+ "2013-10-07T08:23:19.123+04:00",
+ datetime!(2013-10-07 8:23:19.123 +4),
+ ),
+ // Legacy
+ (
+ "2013-10-07 08:23:12:987 -07:00",
+ datetime!(2013-10-07 8:23:12.987 -7),
+ ),
+ ];
+
+ for (s, t) in tests {
+ let result: OffsetDateTime = db.query_row("SELECT ?1", [s], |r| r.get(0))?;
+ assert_eq!(result, t);
+ }
+ Ok(())
+ }
+
+ #[test]
+ fn test_date() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let date = date!(2016 - 02 - 23);
+ db.execute("INSERT INTO foo (t) VALUES (?1)", [date])?;
+
+ let s: String = db.one_column("SELECT t FROM foo")?;
+ assert_eq!("2016-02-23", s);
+ let t: Date = db.one_column("SELECT t FROM foo")?;
+ assert_eq!(date, t);
+ Ok(())
+ }
+
+ #[test]
+ fn test_time() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let time = time!(23:56:04.00001);
+ db.execute("INSERT INTO foo (t) VALUES (?1)", [time])?;
+
+ let s: String = db.one_column("SELECT t FROM foo")?;
+ assert_eq!("23:56:04.00001", s);
+ let v: Time = db.one_column("SELECT t FROM foo")?;
+ assert_eq!(time, v);
+ Ok(())
+ }
+
+ #[test]
+ fn test_primitive_date_time() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let dt = date!(2016 - 02 - 23).with_time(time!(23:56:04));
+
+ db.execute("INSERT INTO foo (t) VALUES (?1)", [dt])?;
+
+ let s: String = db.one_column("SELECT t FROM foo")?;
+ assert_eq!("2016-02-23 23:56:04.0", s);
+ let v: PrimitiveDateTime = db.one_column("SELECT t FROM foo")?;
+ assert_eq!(dt, v);
+
+ db.execute("UPDATE foo set b = datetime(t)", [])?; // "YYYY-MM-DD HH:MM:SS"
+ let hms: PrimitiveDateTime = db.one_column("SELECT b FROM foo")?;
+ assert_eq!(dt, hms);
+ Ok(())
+ }
+
+ #[test]
+ fn test_date_parsing() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let result: Date = db.query_row("SELECT ?1", ["2013-10-07"], |r| r.get(0))?;
+ assert_eq!(result, date!(2013 - 10 - 07));
+ Ok(())
+ }
+
+ #[test]
+ fn test_time_parsing() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let tests = vec![
+ ("08:23", time!(08:23)),
+ ("08:23:19", time!(08:23:19)),
+ ("08:23:19.111", time!(08:23:19.111)),
+ ];
+
+ for (s, t) in tests {
+ let result: Time = db.query_row("SELECT ?1", [s], |r| r.get(0))?;
+ assert_eq!(result, t);
+ }
+ Ok(())
+ }
+
+ #[test]
+ fn test_primitive_date_time_parsing() -> Result<()> {
+ let db = checked_memory_handle()?;
+
+ let tests = vec![
+ ("2013-10-07T08:23", datetime!(2013-10-07 8:23)),
+ ("2013-10-07T08:23:19", datetime!(2013-10-07 8:23:19)),
+ ("2013-10-07T08:23:19.111", datetime!(2013-10-07 8:23:19.111)),
+ ("2013-10-07 08:23", datetime!(2013-10-07 8:23)),
+ ("2013-10-07 08:23:19", datetime!(2013-10-07 8:23:19)),
+ ("2013-10-07 08:23:19.111", datetime!(2013-10-07 8:23:19.111)),
+ ];
+
+ for (s, t) in tests {
+ let result: PrimitiveDateTime = db.query_row("SELECT ?1", [s], |r| r.get(0))?;
+ assert_eq!(result, t);
+ }
+ Ok(())
+ }
+
+ #[test]
+ fn test_sqlite_functions() -> Result<()> {
+ let db = checked_memory_handle()?;
+ db.one_column::<Time>("SELECT CURRENT_TIME").unwrap();
+ db.one_column::<Date>("SELECT CURRENT_DATE").unwrap();
+ db.one_column::<PrimitiveDateTime>("SELECT CURRENT_TIMESTAMP")
+ .unwrap();
+ db.one_column::<OffsetDateTime>("SELECT CURRENT_TIMESTAMP")
+ .unwrap();
+ Ok(())
+ }
+
+ #[test]
+ fn test_time_param() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let now = OffsetDateTime::now_utc().time();
+ let result: Result<bool> = db.query_row(
+ "SELECT 1 WHERE ?1 BETWEEN time('now', '-1 minute') AND time('now', '+1 minute')",
+ [now],
+ |r| r.get(0),
+ );
+ result.unwrap();
+ Ok(())
+ }
+
+ #[test]
+ fn test_date_param() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let now = OffsetDateTime::now_utc().date();
+ let result: Result<bool> = db.query_row(
+ "SELECT 1 WHERE ?1 BETWEEN date('now', '-1 day') AND date('now', '+1 day')",
+ [now],
+ |r| r.get(0),
+ );
+ result.unwrap();
+ Ok(())
+ }
+
+ #[test]
+ fn test_primitive_date_time_param() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let now = PrimitiveDateTime::new(
+ OffsetDateTime::now_utc().date(),
+ OffsetDateTime::now_utc().time(),
+ );
+ let result: Result<bool> = db.query_row(
+ "SELECT 1 WHERE ?1 BETWEEN datetime('now', '-1 minute') AND datetime('now', '+1 minute')",
+ [now],
+ |r| r.get(0),
+ );
+ result.unwrap();
+ Ok(())
+ }
+
+ #[test]
+ fn test_offset_date_time_param() -> Result<()> {
+ let db = checked_memory_handle()?;
+ let result: Result<bool> = db.query_row(
+ "SELECT 1 WHERE ?1 BETWEEN datetime('now', '-1 minute') AND datetime('now', '+1 minute')",
+ [OffsetDateTime::now_utc()],
+ |r| r.get(0),
+ );
+ result.unwrap();
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/types/to_sql.rs b/third_party/rust/rusqlite/src/types/to_sql.rs
new file mode 100644
index 0000000000..29e63abdcc
--- /dev/null
+++ b/third_party/rust/rusqlite/src/types/to_sql.rs
@@ -0,0 +1,541 @@
+use super::{Null, Value, ValueRef};
+#[cfg(feature = "array")]
+use crate::vtab::array::Array;
+use crate::{Error, Result};
+use std::borrow::Cow;
+use std::convert::TryFrom;
+
+/// `ToSqlOutput` represents the possible output types for implementers of the
+/// [`ToSql`] trait.
+#[derive(Clone, Debug, PartialEq)]
+#[non_exhaustive]
+pub enum ToSqlOutput<'a> {
+ /// A borrowed SQLite-representable value.
+ Borrowed(ValueRef<'a>),
+
+ /// An owned SQLite-representable value.
+ Owned(Value),
+
+ /// A BLOB of the given length that is filled with
+ /// zeroes.
+ #[cfg(feature = "blob")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "blob")))]
+ ZeroBlob(i32),
+
+ /// `feature = "array"`
+ #[cfg(feature = "array")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "array")))]
+ Array(Array),
+}
+
+// Generically allow any type that can be converted into a ValueRef
+// to be converted into a ToSqlOutput as well.
+impl<'a, T: ?Sized> From<&'a T> for ToSqlOutput<'a>
+where
+ &'a T: Into<ValueRef<'a>>,
+{
+ #[inline]
+ fn from(t: &'a T) -> Self {
+ ToSqlOutput::Borrowed(t.into())
+ }
+}
+
+// We cannot also generically allow any type that can be converted
+// into a Value to be converted into a ToSqlOutput because of
+// coherence rules (https://github.com/rust-lang/rust/pull/46192),
+// so we'll manually implement it for all the types we know can
+// be converted into Values.
+macro_rules! from_value(
+ ($t:ty) => (
+ impl From<$t> for ToSqlOutput<'_> {
+ #[inline]
+ fn from(t: $t) -> Self { ToSqlOutput::Owned(t.into())}
+ }
+ );
+ (non_zero $t:ty) => (
+ impl From<$t> for ToSqlOutput<'_> {
+ #[inline]
+ fn from(t: $t) -> Self { ToSqlOutput::Owned(t.get().into())}
+ }
+ )
+);
+from_value!(String);
+from_value!(Null);
+from_value!(bool);
+from_value!(i8);
+from_value!(i16);
+from_value!(i32);
+from_value!(i64);
+from_value!(isize);
+from_value!(u8);
+from_value!(u16);
+from_value!(u32);
+from_value!(f32);
+from_value!(f64);
+from_value!(Vec<u8>);
+
+from_value!(non_zero std::num::NonZeroI8);
+from_value!(non_zero std::num::NonZeroI16);
+from_value!(non_zero std::num::NonZeroI32);
+from_value!(non_zero std::num::NonZeroI64);
+from_value!(non_zero std::num::NonZeroIsize);
+from_value!(non_zero std::num::NonZeroU8);
+from_value!(non_zero std::num::NonZeroU16);
+from_value!(non_zero std::num::NonZeroU32);
+
+// It would be nice if we could avoid the heap allocation (of the `Vec`) that
+// `i128` needs in `Into<Value>`, but it's probably fine for the moment, and not
+// worth adding another case to Value.
+#[cfg(feature = "i128_blob")]
+#[cfg_attr(docsrs, doc(cfg(feature = "i128_blob")))]
+from_value!(i128);
+
+#[cfg(feature = "i128_blob")]
+#[cfg_attr(docsrs, doc(cfg(feature = "i128_blob")))]
+from_value!(non_zero std::num::NonZeroI128);
+
+#[cfg(feature = "uuid")]
+#[cfg_attr(docsrs, doc(cfg(feature = "uuid")))]
+from_value!(uuid::Uuid);
+
+impl ToSql for ToSqlOutput<'_> {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ Ok(match *self {
+ ToSqlOutput::Borrowed(v) => ToSqlOutput::Borrowed(v),
+ ToSqlOutput::Owned(ref v) => ToSqlOutput::Borrowed(ValueRef::from(v)),
+
+ #[cfg(feature = "blob")]
+ ToSqlOutput::ZeroBlob(i) => ToSqlOutput::ZeroBlob(i),
+ #[cfg(feature = "array")]
+ ToSqlOutput::Array(ref a) => ToSqlOutput::Array(a.clone()),
+ })
+ }
+}
+
+/// A trait for types that can be converted into SQLite values. Returns
+/// [`Error::ToSqlConversionFailure`] if the conversion fails.
+pub trait ToSql {
+ /// Converts Rust value to SQLite value
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>>;
+}
+
+impl<T: ToSql + ToOwned + ?Sized> ToSql for Cow<'_, T> {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ self.as_ref().to_sql()
+ }
+}
+
+impl<T: ToSql + ?Sized> ToSql for Box<T> {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ self.as_ref().to_sql()
+ }
+}
+
+impl<T: ToSql + ?Sized> ToSql for std::rc::Rc<T> {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ self.as_ref().to_sql()
+ }
+}
+
+impl<T: ToSql + ?Sized> ToSql for std::sync::Arc<T> {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ self.as_ref().to_sql()
+ }
+}
+
+// We should be able to use a generic impl like this:
+//
+// impl<T: Copy> ToSql for T where T: Into<Value> {
+// fn to_sql(&self) -> Result<ToSqlOutput> {
+// Ok(ToSqlOutput::from((*self).into()))
+// }
+// }
+//
+// instead of the following macro, but this runs afoul of
+// https://github.com/rust-lang/rust/issues/30191 and reports conflicting
+// implementations even when there aren't any.
+
+macro_rules! to_sql_self(
+ ($t:ty) => (
+ impl ToSql for $t {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ Ok(ToSqlOutput::from(*self))
+ }
+ }
+ )
+);
+
+to_sql_self!(Null);
+to_sql_self!(bool);
+to_sql_self!(i8);
+to_sql_self!(i16);
+to_sql_self!(i32);
+to_sql_self!(i64);
+to_sql_self!(isize);
+to_sql_self!(u8);
+to_sql_self!(u16);
+to_sql_self!(u32);
+to_sql_self!(f32);
+to_sql_self!(f64);
+
+to_sql_self!(std::num::NonZeroI8);
+to_sql_self!(std::num::NonZeroI16);
+to_sql_self!(std::num::NonZeroI32);
+to_sql_self!(std::num::NonZeroI64);
+to_sql_self!(std::num::NonZeroIsize);
+to_sql_self!(std::num::NonZeroU8);
+to_sql_self!(std::num::NonZeroU16);
+to_sql_self!(std::num::NonZeroU32);
+
+#[cfg(feature = "i128_blob")]
+#[cfg_attr(docsrs, doc(cfg(feature = "i128_blob")))]
+to_sql_self!(i128);
+
+#[cfg(feature = "i128_blob")]
+#[cfg_attr(docsrs, doc(cfg(feature = "i128_blob")))]
+to_sql_self!(std::num::NonZeroI128);
+
+#[cfg(feature = "uuid")]
+#[cfg_attr(docsrs, doc(cfg(feature = "uuid")))]
+to_sql_self!(uuid::Uuid);
+
+macro_rules! to_sql_self_fallible(
+ ($t:ty) => (
+ impl ToSql for $t {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ Ok(ToSqlOutput::Owned(Value::Integer(
+ i64::try_from(*self).map_err(
+ // TODO: Include the values in the error message.
+ |err| Error::ToSqlConversionFailure(err.into())
+ )?
+ )))
+ }
+ }
+ );
+ (non_zero $t:ty) => (
+ impl ToSql for $t {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ Ok(ToSqlOutput::Owned(Value::Integer(
+ i64::try_from(self.get()).map_err(
+ // TODO: Include the values in the error message.
+ |err| Error::ToSqlConversionFailure(err.into())
+ )?
+ )))
+ }
+ }
+ )
+);
+
+// Special implementations for usize and u64 because these conversions can fail.
+to_sql_self_fallible!(u64);
+to_sql_self_fallible!(usize);
+to_sql_self_fallible!(non_zero std::num::NonZeroU64);
+to_sql_self_fallible!(non_zero std::num::NonZeroUsize);
+
+impl<T: ?Sized> ToSql for &'_ T
+where
+ T: ToSql,
+{
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ (*self).to_sql()
+ }
+}
+
+impl ToSql for String {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ Ok(ToSqlOutput::from(self.as_str()))
+ }
+}
+
+impl ToSql for str {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ Ok(ToSqlOutput::from(self))
+ }
+}
+
+impl ToSql for Vec<u8> {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ Ok(ToSqlOutput::from(self.as_slice()))
+ }
+}
+
+impl<const N: usize> ToSql for [u8; N] {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ Ok(ToSqlOutput::from(&self[..]))
+ }
+}
+
+impl ToSql for [u8] {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ Ok(ToSqlOutput::from(self))
+ }
+}
+
+impl ToSql for Value {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ Ok(ToSqlOutput::from(self))
+ }
+}
+
+impl<T: ToSql> ToSql for Option<T> {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ match *self {
+ None => Ok(ToSqlOutput::from(Null)),
+ Some(ref t) => t.to_sql(),
+ }
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use super::ToSql;
+
+ fn is_to_sql<T: ToSql>() {}
+
+ #[test]
+ fn test_integral_types() {
+ is_to_sql::<i8>();
+ is_to_sql::<i16>();
+ is_to_sql::<i32>();
+ is_to_sql::<i64>();
+ is_to_sql::<isize>();
+ is_to_sql::<u8>();
+ is_to_sql::<u16>();
+ is_to_sql::<u32>();
+ is_to_sql::<u64>();
+ is_to_sql::<usize>();
+ }
+
+ #[test]
+ fn test_nonzero_types() {
+ is_to_sql::<std::num::NonZeroI8>();
+ is_to_sql::<std::num::NonZeroI16>();
+ is_to_sql::<std::num::NonZeroI32>();
+ is_to_sql::<std::num::NonZeroI64>();
+ is_to_sql::<std::num::NonZeroIsize>();
+ is_to_sql::<std::num::NonZeroU8>();
+ is_to_sql::<std::num::NonZeroU16>();
+ is_to_sql::<std::num::NonZeroU32>();
+ is_to_sql::<std::num::NonZeroU64>();
+ is_to_sql::<std::num::NonZeroUsize>();
+ }
+
+ #[test]
+ fn test_u8_array() {
+ let a: [u8; 99] = [0u8; 99];
+ let _a: &[&dyn ToSql] = crate::params![a];
+ let r = ToSql::to_sql(&a);
+
+ r.unwrap();
+ }
+
+ #[test]
+ fn test_cow_str() {
+ use std::borrow::Cow;
+ let s = "str";
+ let cow: Cow<str> = Cow::Borrowed(s);
+ let r = cow.to_sql();
+ r.unwrap();
+ let cow: Cow<str> = Cow::Owned::<str>(String::from(s));
+ let r = cow.to_sql();
+ r.unwrap();
+ // Ensure this compiles.
+ let _p: &[&dyn ToSql] = crate::params![cow];
+ }
+
+ #[test]
+ fn test_box_dyn() {
+ let s: Box<dyn ToSql> = Box::new("Hello world!");
+ let _s: &[&dyn ToSql] = crate::params![s];
+ let r = ToSql::to_sql(&s);
+
+ r.unwrap();
+ }
+
+ #[test]
+ fn test_box_deref() {
+ let s: Box<str> = "Hello world!".into();
+ let _s: &[&dyn ToSql] = crate::params![s];
+ let r = s.to_sql();
+
+ r.unwrap();
+ }
+
+ #[test]
+ fn test_box_direct() {
+ let s: Box<str> = "Hello world!".into();
+ let _s: &[&dyn ToSql] = crate::params![s];
+ let r = ToSql::to_sql(&s);
+
+ r.unwrap();
+ }
+
+ #[test]
+ fn test_cells() {
+ use std::{rc::Rc, sync::Arc};
+
+ let source_str: Box<str> = "Hello world!".into();
+
+ let s: Rc<Box<str>> = Rc::new(source_str.clone());
+ let _s: &[&dyn ToSql] = crate::params![s];
+ let r = s.to_sql();
+ r.unwrap();
+
+ let s: Arc<Box<str>> = Arc::new(source_str.clone());
+ let _s: &[&dyn ToSql] = crate::params![s];
+ let r = s.to_sql();
+ r.unwrap();
+
+ let s: Arc<str> = Arc::from(&*source_str);
+ let _s: &[&dyn ToSql] = crate::params![s];
+ let r = s.to_sql();
+ r.unwrap();
+
+ let s: Arc<dyn ToSql> = Arc::new(source_str.clone());
+ let _s: &[&dyn ToSql] = crate::params![s];
+ let r = s.to_sql();
+ r.unwrap();
+
+ let s: Rc<str> = Rc::from(&*source_str);
+ let _s: &[&dyn ToSql] = crate::params![s];
+ let r = s.to_sql();
+ r.unwrap();
+
+ let s: Rc<dyn ToSql> = Rc::new(source_str);
+ let _s: &[&dyn ToSql] = crate::params![s];
+ let r = s.to_sql();
+ r.unwrap();
+ }
+
+ #[cfg(feature = "i128_blob")]
+ #[test]
+ fn test_i128() -> crate::Result<()> {
+ use crate::Connection;
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo (i128 BLOB, desc TEXT)")?;
+ db.execute(
+ "
+ INSERT INTO foo(i128, desc) VALUES
+ (?1, 'zero'),
+ (?2, 'neg one'), (?3, 'neg two'),
+ (?4, 'pos one'), (?5, 'pos two'),
+ (?6, 'min'), (?7, 'max')",
+ [0i128, -1i128, -2i128, 1i128, 2i128, i128::MIN, i128::MAX],
+ )?;
+
+ let mut stmt = db.prepare("SELECT i128, desc FROM foo ORDER BY i128 ASC")?;
+
+ let res = stmt
+ .query_map([], |row| {
+ Ok((row.get::<_, i128>(0)?, row.get::<_, String>(1)?))
+ })?
+ .collect::<Result<Vec<_>, _>>()?;
+
+ assert_eq!(
+ res,
+ &[
+ (i128::MIN, "min".to_owned()),
+ (-2, "neg two".to_owned()),
+ (-1, "neg one".to_owned()),
+ (0, "zero".to_owned()),
+ (1, "pos one".to_owned()),
+ (2, "pos two".to_owned()),
+ (i128::MAX, "max".to_owned()),
+ ]
+ );
+ Ok(())
+ }
+
+ #[cfg(feature = "i128_blob")]
+ #[test]
+ fn test_non_zero_i128() -> crate::Result<()> {
+ use std::num::NonZeroI128;
+ macro_rules! nz {
+ ($x:expr) => {
+ NonZeroI128::new($x).unwrap()
+ };
+ }
+
+ let db = crate::Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo (i128 BLOB, desc TEXT)")?;
+ db.execute(
+ "INSERT INTO foo(i128, desc) VALUES
+ (?1, 'neg one'), (?2, 'neg two'),
+ (?3, 'pos one'), (?4, 'pos two'),
+ (?5, 'min'), (?6, 'max')",
+ [
+ nz!(-1),
+ nz!(-2),
+ nz!(1),
+ nz!(2),
+ nz!(i128::MIN),
+ nz!(i128::MAX),
+ ],
+ )?;
+ let mut stmt = db.prepare("SELECT i128, desc FROM foo ORDER BY i128 ASC")?;
+
+ let res = stmt
+ .query_map([], |row| Ok((row.get(0)?, row.get(1)?)))?
+ .collect::<Result<Vec<(NonZeroI128, String)>, _>>()?;
+
+ assert_eq!(
+ res,
+ &[
+ (nz!(i128::MIN), "min".to_owned()),
+ (nz!(-2), "neg two".to_owned()),
+ (nz!(-1), "neg one".to_owned()),
+ (nz!(1), "pos one".to_owned()),
+ (nz!(2), "pos two".to_owned()),
+ (nz!(i128::MAX), "max".to_owned()),
+ ]
+ );
+ let err = db.query_row("SELECT ?1", [0i128], |row| row.get::<_, NonZeroI128>(0));
+ assert_eq!(err, Err(crate::Error::IntegralValueOutOfRange(0, 0)));
+ Ok(())
+ }
+
+ #[cfg(feature = "uuid")]
+ #[test]
+ fn test_uuid() -> crate::Result<()> {
+ use crate::{params, Connection};
+ use uuid::Uuid;
+
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE foo (id BLOB CHECK(length(id) = 16), label TEXT);")?;
+
+ let id = Uuid::new_v4();
+
+ db.execute(
+ "INSERT INTO foo (id, label) VALUES (?1, ?2)",
+ params![id, "target"],
+ )?;
+
+ let mut stmt = db.prepare("SELECT id, label FROM foo WHERE id = ?1")?;
+
+ let mut rows = stmt.query(params![id])?;
+ let row = rows.next()?.unwrap();
+
+ let found_id: Uuid = row.get_unwrap(0);
+ let found_label: String = row.get_unwrap(1);
+
+ assert_eq!(found_id, id);
+ assert_eq!(found_label, "target");
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/types/url.rs b/third_party/rust/rusqlite/src/types/url.rs
new file mode 100644
index 0000000000..e7fc2039c1
--- /dev/null
+++ b/third_party/rust/rusqlite/src/types/url.rs
@@ -0,0 +1,82 @@
+//! [`ToSql`] and [`FromSql`] implementation for [`url::Url`].
+use crate::types::{FromSql, FromSqlError, FromSqlResult, ToSql, ToSqlOutput, ValueRef};
+use crate::Result;
+use url::Url;
+
+/// Serialize `Url` to text.
+impl ToSql for Url {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ Ok(ToSqlOutput::from(self.as_str()))
+ }
+}
+
+/// Deserialize text to `Url`.
+impl FromSql for Url {
+ #[inline]
+ fn column_result(value: ValueRef<'_>) -> FromSqlResult<Self> {
+ match value {
+ ValueRef::Text(s) => {
+ let s = std::str::from_utf8(s).map_err(|e| FromSqlError::Other(Box::new(e)))?;
+ Url::parse(s).map_err(|e| FromSqlError::Other(Box::new(e)))
+ }
+ _ => Err(FromSqlError::InvalidType),
+ }
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use crate::{params, Connection, Error, Result};
+ use url::{ParseError, Url};
+
+ fn checked_memory_handle() -> Result<Connection> {
+ let db = Connection::open_in_memory()?;
+ db.execute_batch("CREATE TABLE urls (i INTEGER, v TEXT)")?;
+ Ok(db)
+ }
+
+ fn get_url(db: &Connection, id: i64) -> Result<Url> {
+ db.query_row("SELECT v FROM urls WHERE i = ?", [id], |r| r.get(0))
+ }
+
+ #[test]
+ fn test_sql_url() -> Result<()> {
+ let db = &checked_memory_handle()?;
+
+ let url0 = Url::parse("http://www.example1.com").unwrap();
+ let url1 = Url::parse("http://www.example1.com/👌").unwrap();
+ let url2 = "http://www.example2.com/👌";
+
+ db.execute(
+ "INSERT INTO urls (i, v) VALUES (0, ?1), (1, ?2), (2, ?3), (3, ?4)",
+ // also insert a non-hex encoded url (which might be present if it was
+ // inserted separately)
+ params![url0, url1, url2, "illegal"],
+ )?;
+
+ assert_eq!(get_url(db, 0)?, url0);
+
+ assert_eq!(get_url(db, 1)?, url1);
+
+ // Should successfully read it, even though it wasn't inserted as an
+ // escaped url.
+ let out_url2: Url = get_url(db, 2)?;
+ assert_eq!(out_url2, Url::parse(url2).unwrap());
+
+ // Make sure the conversion error comes through correctly.
+ let err = get_url(db, 3).unwrap_err();
+ match err {
+ Error::FromSqlConversionFailure(_, _, e) => {
+ assert_eq!(
+ *e.downcast::<ParseError>().unwrap(),
+ ParseError::RelativeUrlWithoutBase,
+ );
+ }
+ e => {
+ panic!("Expected conversion failure, got {e}");
+ }
+ }
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/types/value.rs b/third_party/rust/rusqlite/src/types/value.rs
new file mode 100644
index 0000000000..ca3ee9f49d
--- /dev/null
+++ b/third_party/rust/rusqlite/src/types/value.rs
@@ -0,0 +1,142 @@
+use super::{Null, Type};
+
+/// Owning [dynamic type value](http://sqlite.org/datatype3.html). Value's type is typically
+/// dictated by SQLite (not by the caller).
+///
+/// See [`ValueRef`](crate::types::ValueRef) for a non-owning dynamic type
+/// value.
+#[derive(Clone, Debug, PartialEq)]
+pub enum Value {
+ /// The value is a `NULL` value.
+ Null,
+ /// The value is a signed integer.
+ Integer(i64),
+ /// The value is a floating point number.
+ Real(f64),
+ /// The value is a text string.
+ Text(String),
+ /// The value is a blob of data
+ Blob(Vec<u8>),
+}
+
+impl From<Null> for Value {
+ #[inline]
+ fn from(_: Null) -> Value {
+ Value::Null
+ }
+}
+
+impl From<bool> for Value {
+ #[inline]
+ fn from(i: bool) -> Value {
+ Value::Integer(i as i64)
+ }
+}
+
+impl From<isize> for Value {
+ #[inline]
+ fn from(i: isize) -> Value {
+ Value::Integer(i as i64)
+ }
+}
+
+#[cfg(feature = "i128_blob")]
+#[cfg_attr(docsrs, doc(cfg(feature = "i128_blob")))]
+impl From<i128> for Value {
+ #[inline]
+ fn from(i: i128) -> Value {
+ // We store these biased (e.g. with the most significant bit flipped)
+ // so that comparisons with negative numbers work properly.
+ Value::Blob(i128::to_be_bytes(i ^ (1_i128 << 127)).to_vec())
+ }
+}
+
+#[cfg(feature = "uuid")]
+#[cfg_attr(docsrs, doc(cfg(feature = "uuid")))]
+impl From<uuid::Uuid> for Value {
+ #[inline]
+ fn from(id: uuid::Uuid) -> Value {
+ Value::Blob(id.as_bytes().to_vec())
+ }
+}
+
+macro_rules! from_i64(
+ ($t:ty) => (
+ impl From<$t> for Value {
+ #[inline]
+ fn from(i: $t) -> Value {
+ Value::Integer(i64::from(i))
+ }
+ }
+ )
+);
+
+from_i64!(i8);
+from_i64!(i16);
+from_i64!(i32);
+from_i64!(u8);
+from_i64!(u16);
+from_i64!(u32);
+
+impl From<i64> for Value {
+ #[inline]
+ fn from(i: i64) -> Value {
+ Value::Integer(i)
+ }
+}
+
+impl From<f32> for Value {
+ #[inline]
+ fn from(f: f32) -> Value {
+ Value::Real(f.into())
+ }
+}
+
+impl From<f64> for Value {
+ #[inline]
+ fn from(f: f64) -> Value {
+ Value::Real(f)
+ }
+}
+
+impl From<String> for Value {
+ #[inline]
+ fn from(s: String) -> Value {
+ Value::Text(s)
+ }
+}
+
+impl From<Vec<u8>> for Value {
+ #[inline]
+ fn from(v: Vec<u8>) -> Value {
+ Value::Blob(v)
+ }
+}
+
+impl<T> From<Option<T>> for Value
+where
+ T: Into<Value>,
+{
+ #[inline]
+ fn from(v: Option<T>) -> Value {
+ match v {
+ Some(x) => x.into(),
+ None => Value::Null,
+ }
+ }
+}
+
+impl Value {
+ /// Returns SQLite fundamental datatype.
+ #[inline]
+ #[must_use]
+ pub fn data_type(&self) -> Type {
+ match *self {
+ Value::Null => Type::Null,
+ Value::Integer(_) => Type::Integer,
+ Value::Real(_) => Type::Real,
+ Value::Text(_) => Type::Text,
+ Value::Blob(_) => Type::Blob,
+ }
+ }
+}
diff --git a/third_party/rust/rusqlite/src/types/value_ref.rs b/third_party/rust/rusqlite/src/types/value_ref.rs
new file mode 100644
index 0000000000..aa062f8ba4
--- /dev/null
+++ b/third_party/rust/rusqlite/src/types/value_ref.rs
@@ -0,0 +1,258 @@
+use super::{Type, Value};
+use crate::types::{FromSqlError, FromSqlResult};
+
+/// A non-owning [dynamic type value](http://sqlite.org/datatype3.html). Typically the
+/// memory backing this value is owned by SQLite.
+///
+/// See [`Value`](Value) for an owning dynamic type value.
+#[derive(Copy, Clone, Debug, PartialEq)]
+pub enum ValueRef<'a> {
+ /// The value is a `NULL` value.
+ Null,
+ /// The value is a signed integer.
+ Integer(i64),
+ /// The value is a floating point number.
+ Real(f64),
+ /// The value is a text string.
+ Text(&'a [u8]),
+ /// The value is a blob of data
+ Blob(&'a [u8]),
+}
+
+impl ValueRef<'_> {
+ /// Returns SQLite fundamental datatype.
+ #[inline]
+ #[must_use]
+ pub fn data_type(&self) -> Type {
+ match *self {
+ ValueRef::Null => Type::Null,
+ ValueRef::Integer(_) => Type::Integer,
+ ValueRef::Real(_) => Type::Real,
+ ValueRef::Text(_) => Type::Text,
+ ValueRef::Blob(_) => Type::Blob,
+ }
+ }
+}
+
+impl<'a> ValueRef<'a> {
+ /// If `self` is case `Integer`, returns the integral value. Otherwise,
+ /// returns [`Err(Error::InvalidColumnType)`](crate::Error::InvalidColumnType).
+ #[inline]
+ pub fn as_i64(&self) -> FromSqlResult<i64> {
+ match *self {
+ ValueRef::Integer(i) => Ok(i),
+ _ => Err(FromSqlError::InvalidType),
+ }
+ }
+
+ /// If `self` is case `Null` returns None.
+ /// If `self` is case `Integer`, returns the integral value.
+ /// Otherwise returns [`Err(Error::InvalidColumnType)`](crate::Error::InvalidColumnType).
+ #[inline]
+ pub fn as_i64_or_null(&self) -> FromSqlResult<Option<i64>> {
+ match *self {
+ ValueRef::Null => Ok(None),
+ ValueRef::Integer(i) => Ok(Some(i)),
+ _ => Err(FromSqlError::InvalidType),
+ }
+ }
+
+ /// If `self` is case `Real`, returns the floating point value. Otherwise,
+ /// returns [`Err(Error::InvalidColumnType)`](crate::Error::InvalidColumnType).
+ #[inline]
+ pub fn as_f64(&self) -> FromSqlResult<f64> {
+ match *self {
+ ValueRef::Real(f) => Ok(f),
+ _ => Err(FromSqlError::InvalidType),
+ }
+ }
+
+ /// If `self` is case `Null` returns None.
+ /// If `self` is case `Real`, returns the floating point value.
+ /// Otherwise returns [`Err(Error::InvalidColumnType)`](crate::Error::InvalidColumnType).
+ #[inline]
+ pub fn as_f64_or_null(&self) -> FromSqlResult<Option<f64>> {
+ match *self {
+ ValueRef::Null => Ok(None),
+ ValueRef::Real(f) => Ok(Some(f)),
+ _ => Err(FromSqlError::InvalidType),
+ }
+ }
+
+ /// If `self` is case `Text`, returns the string value. Otherwise, returns
+ /// [`Err(Error::InvalidColumnType)`](crate::Error::InvalidColumnType).
+ #[inline]
+ pub fn as_str(&self) -> FromSqlResult<&'a str> {
+ match *self {
+ ValueRef::Text(t) => {
+ std::str::from_utf8(t).map_err(|e| FromSqlError::Other(Box::new(e)))
+ }
+ _ => Err(FromSqlError::InvalidType),
+ }
+ }
+
+ /// If `self` is case `Null` returns None.
+ /// If `self` is case `Text`, returns the string value.
+ /// Otherwise returns [`Err(Error::InvalidColumnType)`](crate::Error::InvalidColumnType).
+ #[inline]
+ pub fn as_str_or_null(&self) -> FromSqlResult<Option<&'a str>> {
+ match *self {
+ ValueRef::Null => Ok(None),
+ ValueRef::Text(t) => std::str::from_utf8(t)
+ .map_err(|e| FromSqlError::Other(Box::new(e)))
+ .map(Some),
+ _ => Err(FromSqlError::InvalidType),
+ }
+ }
+
+ /// If `self` is case `Blob`, returns the byte slice. Otherwise, returns
+ /// [`Err(Error::InvalidColumnType)`](crate::Error::InvalidColumnType).
+ #[inline]
+ pub fn as_blob(&self) -> FromSqlResult<&'a [u8]> {
+ match *self {
+ ValueRef::Blob(b) => Ok(b),
+ _ => Err(FromSqlError::InvalidType),
+ }
+ }
+
+ /// If `self` is case `Null` returns None.
+ /// If `self` is case `Blob`, returns the byte slice.
+ /// Otherwise returns [`Err(Error::InvalidColumnType)`](crate::Error::InvalidColumnType).
+ #[inline]
+ pub fn as_blob_or_null(&self) -> FromSqlResult<Option<&'a [u8]>> {
+ match *self {
+ ValueRef::Null => Ok(None),
+ ValueRef::Blob(b) => Ok(Some(b)),
+ _ => Err(FromSqlError::InvalidType),
+ }
+ }
+
+ /// Returns the byte slice that makes up this `ValueRef` if it's either
+ /// [`ValueRef::Blob`] or [`ValueRef::Text`].
+ #[inline]
+ pub fn as_bytes(&self) -> FromSqlResult<&'a [u8]> {
+ match self {
+ ValueRef::Text(s) | ValueRef::Blob(s) => Ok(s),
+ _ => Err(FromSqlError::InvalidType),
+ }
+ }
+
+ /// If `self` is case `Null` returns None.
+ /// If `self` is [`ValueRef::Blob`] or [`ValueRef::Text`] returns the byte
+ /// slice that makes up this value
+ #[inline]
+ pub fn as_bytes_or_null(&self) -> FromSqlResult<Option<&'a [u8]>> {
+ match *self {
+ ValueRef::Null => Ok(None),
+ ValueRef::Text(s) | ValueRef::Blob(s) => Ok(Some(s)),
+ _ => Err(FromSqlError::InvalidType),
+ }
+ }
+}
+
+impl From<ValueRef<'_>> for Value {
+ #[inline]
+ #[track_caller]
+ fn from(borrowed: ValueRef<'_>) -> Value {
+ match borrowed {
+ ValueRef::Null => Value::Null,
+ ValueRef::Integer(i) => Value::Integer(i),
+ ValueRef::Real(r) => Value::Real(r),
+ ValueRef::Text(s) => {
+ let s = std::str::from_utf8(s).expect("invalid UTF-8");
+ Value::Text(s.to_string())
+ }
+ ValueRef::Blob(b) => Value::Blob(b.to_vec()),
+ }
+ }
+}
+
+impl<'a> From<&'a str> for ValueRef<'a> {
+ #[inline]
+ fn from(s: &str) -> ValueRef<'_> {
+ ValueRef::Text(s.as_bytes())
+ }
+}
+
+impl<'a> From<&'a [u8]> for ValueRef<'a> {
+ #[inline]
+ fn from(s: &[u8]) -> ValueRef<'_> {
+ ValueRef::Blob(s)
+ }
+}
+
+impl<'a> From<&'a Value> for ValueRef<'a> {
+ #[inline]
+ fn from(value: &'a Value) -> ValueRef<'a> {
+ match *value {
+ Value::Null => ValueRef::Null,
+ Value::Integer(i) => ValueRef::Integer(i),
+ Value::Real(r) => ValueRef::Real(r),
+ Value::Text(ref s) => ValueRef::Text(s.as_bytes()),
+ Value::Blob(ref b) => ValueRef::Blob(b),
+ }
+ }
+}
+
+impl<'a, T> From<Option<T>> for ValueRef<'a>
+where
+ T: Into<ValueRef<'a>>,
+{
+ #[inline]
+ fn from(s: Option<T>) -> ValueRef<'a> {
+ match s {
+ Some(x) => x.into(),
+ None => ValueRef::Null,
+ }
+ }
+}
+
+#[cfg(any(feature = "functions", feature = "session", feature = "vtab"))]
+impl<'a> ValueRef<'a> {
+ pub(crate) unsafe fn from_value(value: *mut crate::ffi::sqlite3_value) -> ValueRef<'a> {
+ use crate::ffi;
+ use std::slice::from_raw_parts;
+
+ match ffi::sqlite3_value_type(value) {
+ ffi::SQLITE_NULL => ValueRef::Null,
+ ffi::SQLITE_INTEGER => ValueRef::Integer(ffi::sqlite3_value_int64(value)),
+ ffi::SQLITE_FLOAT => ValueRef::Real(ffi::sqlite3_value_double(value)),
+ ffi::SQLITE_TEXT => {
+ let text = ffi::sqlite3_value_text(value);
+ let len = ffi::sqlite3_value_bytes(value);
+ assert!(
+ !text.is_null(),
+ "unexpected SQLITE_TEXT value type with NULL data"
+ );
+ let s = from_raw_parts(text.cast::<u8>(), len as usize);
+ ValueRef::Text(s)
+ }
+ ffi::SQLITE_BLOB => {
+ let (blob, len) = (
+ ffi::sqlite3_value_blob(value),
+ ffi::sqlite3_value_bytes(value),
+ );
+
+ assert!(
+ len >= 0,
+ "unexpected negative return from sqlite3_value_bytes"
+ );
+ if len > 0 {
+ assert!(
+ !blob.is_null(),
+ "unexpected SQLITE_BLOB value type with NULL data"
+ );
+ ValueRef::Blob(from_raw_parts(blob.cast::<u8>(), len as usize))
+ } else {
+ // The return value from sqlite3_value_blob() for a zero-length BLOB
+ // is a NULL pointer.
+ ValueRef::Blob(&[])
+ }
+ }
+ _ => unreachable!("sqlite3_value_type returned invalid value"),
+ }
+ }
+
+ // TODO sqlite3_value_nochange // 3.22.0 & VTab xUpdate
+ // TODO sqlite3_value_frombind // 3.28.0
+}
diff --git a/third_party/rust/rusqlite/src/unlock_notify.rs b/third_party/rust/rusqlite/src/unlock_notify.rs
new file mode 100644
index 0000000000..065c52d72a
--- /dev/null
+++ b/third_party/rust/rusqlite/src/unlock_notify.rs
@@ -0,0 +1,117 @@
+//! [Unlock Notification](http://sqlite.org/unlock_notify.html)
+
+use std::os::raw::c_int;
+use std::os::raw::c_void;
+use std::panic::catch_unwind;
+use std::sync::{Condvar, Mutex};
+
+use crate::ffi;
+
+struct UnlockNotification {
+ cond: Condvar, // Condition variable to wait on
+ mutex: Mutex<bool>, // Mutex to protect structure
+}
+
+#[allow(clippy::mutex_atomic)]
+impl UnlockNotification {
+ fn new() -> UnlockNotification {
+ UnlockNotification {
+ cond: Condvar::new(),
+ mutex: Mutex::new(false),
+ }
+ }
+
+ fn fired(&self) {
+ let mut flag = unpoison(self.mutex.lock());
+ *flag = true;
+ self.cond.notify_one();
+ }
+
+ fn wait(&self) {
+ let mut fired = unpoison(self.mutex.lock());
+ while !*fired {
+ fired = unpoison(self.cond.wait(fired));
+ }
+ }
+}
+
+#[inline]
+fn unpoison<T>(r: Result<T, std::sync::PoisonError<T>>) -> T {
+ r.unwrap_or_else(std::sync::PoisonError::into_inner)
+}
+
+/// This function is an unlock-notify callback
+unsafe extern "C" fn unlock_notify_cb(ap_arg: *mut *mut c_void, n_arg: c_int) {
+ use std::slice::from_raw_parts;
+ let args = from_raw_parts(ap_arg as *const &UnlockNotification, n_arg as usize);
+ for un in args {
+ drop(catch_unwind(std::panic::AssertUnwindSafe(|| un.fired())));
+ }
+}
+
+pub unsafe fn is_locked(db: *mut ffi::sqlite3, rc: c_int) -> bool {
+ rc == ffi::SQLITE_LOCKED_SHAREDCACHE
+ || (rc & 0xFF) == ffi::SQLITE_LOCKED
+ && ffi::sqlite3_extended_errcode(db) == ffi::SQLITE_LOCKED_SHAREDCACHE
+}
+
+/// This function assumes that an SQLite API call (either `sqlite3_prepare_v2()`
+/// or `sqlite3_step()`) has just returned `SQLITE_LOCKED`. The argument is the
+/// associated database connection.
+///
+/// This function calls `sqlite3_unlock_notify()` to register for an
+/// unlock-notify callback, then blocks until that callback is delivered
+/// and returns `SQLITE_OK`. The caller should then retry the failed operation.
+///
+/// Or, if `sqlite3_unlock_notify()` indicates that to block would deadlock
+/// the system, then this function returns `SQLITE_LOCKED` immediately. In
+/// this case the caller should not retry the operation and should roll
+/// back the current transaction (if any).
+#[cfg(feature = "unlock_notify")]
+pub unsafe fn wait_for_unlock_notify(db: *mut ffi::sqlite3) -> c_int {
+ let un = UnlockNotification::new();
+ /* Register for an unlock-notify callback. */
+ let rc = ffi::sqlite3_unlock_notify(
+ db,
+ Some(unlock_notify_cb),
+ &un as *const UnlockNotification as *mut c_void,
+ );
+ debug_assert!(
+ rc == ffi::SQLITE_LOCKED || rc == ffi::SQLITE_LOCKED_SHAREDCACHE || rc == ffi::SQLITE_OK
+ );
+ if rc == ffi::SQLITE_OK {
+ un.wait();
+ }
+ rc
+}
+
+#[cfg(test)]
+mod test {
+ use crate::{Connection, OpenFlags, Result, Transaction, TransactionBehavior};
+ use std::sync::mpsc::sync_channel;
+ use std::thread;
+ use std::time;
+
+ #[test]
+ fn test_unlock_notify() -> Result<()> {
+ let url = "file::memory:?cache=shared";
+ let flags = OpenFlags::SQLITE_OPEN_READ_WRITE | OpenFlags::SQLITE_OPEN_URI;
+ let db1 = Connection::open_with_flags(url, flags)?;
+ db1.execute_batch("CREATE TABLE foo (x)")?;
+ let (rx, tx) = sync_channel(0);
+ let child = thread::spawn(move || {
+ let mut db2 = Connection::open_with_flags(url, flags).unwrap();
+ let tx2 = Transaction::new(&mut db2, TransactionBehavior::Immediate).unwrap();
+ tx2.execute_batch("INSERT INTO foo VALUES (42)").unwrap();
+ rx.send(1).unwrap();
+ let ten_millis = time::Duration::from_millis(10);
+ thread::sleep(ten_millis);
+ tx2.commit().unwrap();
+ });
+ assert_eq!(tx.recv().unwrap(), 1);
+ let the_answer: i64 = db1.one_column("SELECT x FROM foo")?;
+ assert_eq!(42i64, the_answer);
+ child.join().unwrap();
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/util/mod.rs b/third_party/rust/rusqlite/src/util/mod.rs
new file mode 100644
index 0000000000..a759cb98fc
--- /dev/null
+++ b/third_party/rust/rusqlite/src/util/mod.rs
@@ -0,0 +1,9 @@
+// Internal utilities
+pub(crate) mod param_cache;
+mod small_cstr;
+pub(crate) use param_cache::ParamIndexCache;
+pub(crate) use small_cstr::SmallCString;
+
+// Doesn't use any modern features or vtab stuff, but is only used by them.
+mod sqlite_string;
+pub(crate) use sqlite_string::{alloc, SqliteMallocString};
diff --git a/third_party/rust/rusqlite/src/util/param_cache.rs b/third_party/rust/rusqlite/src/util/param_cache.rs
new file mode 100644
index 0000000000..6faced98af
--- /dev/null
+++ b/third_party/rust/rusqlite/src/util/param_cache.rs
@@ -0,0 +1,60 @@
+use super::SmallCString;
+use std::cell::RefCell;
+use std::collections::BTreeMap;
+
+/// Maps parameter names to parameter indices.
+#[derive(Default, Clone, Debug)]
+// BTreeMap seems to do better here unless we want to pull in a custom hash
+// function.
+pub(crate) struct ParamIndexCache(RefCell<BTreeMap<SmallCString, usize>>);
+
+impl ParamIndexCache {
+ pub fn get_or_insert_with<F>(&self, s: &str, func: F) -> Option<usize>
+ where
+ F: FnOnce(&std::ffi::CStr) -> Option<usize>,
+ {
+ let mut cache = self.0.borrow_mut();
+ // Avoid entry API, needs allocation to test membership.
+ if let Some(v) = cache.get(s) {
+ return Some(*v);
+ }
+ // If there's an internal nul in the name it couldn't have been a
+ // parameter, so early return here is ok.
+ let name = SmallCString::new(s).ok()?;
+ let val = func(&name)?;
+ cache.insert(name, val);
+ Some(val)
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use super::*;
+ #[test]
+ fn test_cache() {
+ let p = ParamIndexCache::default();
+ let v = p.get_or_insert_with("foo", |cstr| {
+ assert_eq!(cstr.to_str().unwrap(), "foo");
+ Some(3)
+ });
+ assert_eq!(v, Some(3));
+ let v = p.get_or_insert_with("foo", |_| {
+ panic!("shouldn't be called this time");
+ });
+ assert_eq!(v, Some(3));
+ let v = p.get_or_insert_with("gar\0bage", |_| {
+ panic!("shouldn't be called here either");
+ });
+ assert_eq!(v, None);
+ let v = p.get_or_insert_with("bar", |cstr| {
+ assert_eq!(cstr.to_str().unwrap(), "bar");
+ None
+ });
+ assert_eq!(v, None);
+ let v = p.get_or_insert_with("bar", |cstr| {
+ assert_eq!(cstr.to_str().unwrap(), "bar");
+ Some(30)
+ });
+ assert_eq!(v, Some(30));
+ }
+}
diff --git a/third_party/rust/rusqlite/src/util/small_cstr.rs b/third_party/rust/rusqlite/src/util/small_cstr.rs
new file mode 100644
index 0000000000..1ec73744e0
--- /dev/null
+++ b/third_party/rust/rusqlite/src/util/small_cstr.rs
@@ -0,0 +1,170 @@
+use smallvec::{smallvec, SmallVec};
+use std::ffi::{CStr, CString, NulError};
+
+/// Similar to `std::ffi::CString`, but avoids heap allocating if the string is
+/// small enough. Also guarantees it's input is UTF-8 -- used for cases where we
+/// need to pass a NUL-terminated string to SQLite, and we have a `&str`.
+#[derive(Clone, PartialEq, Eq, PartialOrd, Ord)]
+pub(crate) struct SmallCString(SmallVec<[u8; 16]>);
+
+impl SmallCString {
+ #[inline]
+ pub fn new(s: &str) -> Result<Self, NulError> {
+ if s.as_bytes().contains(&0_u8) {
+ return Err(Self::fabricate_nul_error(s));
+ }
+ let mut buf = SmallVec::with_capacity(s.len() + 1);
+ buf.extend_from_slice(s.as_bytes());
+ buf.push(0);
+ let res = Self(buf);
+ res.debug_checks();
+ Ok(res)
+ }
+
+ #[inline]
+ pub fn as_str(&self) -> &str {
+ self.debug_checks();
+ // Constructor takes a &str so this is safe.
+ unsafe { std::str::from_utf8_unchecked(self.as_bytes_without_nul()) }
+ }
+
+ /// Get the bytes not including the NUL terminator. E.g. the bytes which
+ /// make up our `str`:
+ /// - `SmallCString::new("foo").as_bytes_without_nul() == b"foo"`
+ /// - `SmallCString::new("foo").as_bytes_with_nul() == b"foo\0"`
+ #[inline]
+ pub fn as_bytes_without_nul(&self) -> &[u8] {
+ self.debug_checks();
+ &self.0[..self.len()]
+ }
+
+ /// Get the bytes behind this str *including* the NUL terminator. This
+ /// should never return an empty slice.
+ #[inline]
+ pub fn as_bytes_with_nul(&self) -> &[u8] {
+ self.debug_checks();
+ &self.0
+ }
+
+ #[inline]
+ #[cfg(debug_assertions)]
+ fn debug_checks(&self) {
+ debug_assert_ne!(self.0.len(), 0);
+ debug_assert_eq!(self.0[self.0.len() - 1], 0);
+ let strbytes = &self.0[..(self.0.len() - 1)];
+ debug_assert!(!strbytes.contains(&0));
+ debug_assert!(std::str::from_utf8(strbytes).is_ok());
+ }
+
+ #[inline]
+ #[cfg(not(debug_assertions))]
+ fn debug_checks(&self) {}
+
+ #[inline]
+ pub fn len(&self) -> usize {
+ debug_assert_ne!(self.0.len(), 0);
+ self.0.len() - 1
+ }
+
+ #[inline]
+ #[allow(unused)] // clippy wants this function.
+ pub fn is_empty(&self) -> bool {
+ self.len() == 0
+ }
+
+ #[inline]
+ pub fn as_cstr(&self) -> &CStr {
+ let bytes = self.as_bytes_with_nul();
+ debug_assert!(CStr::from_bytes_with_nul(bytes).is_ok());
+ unsafe { CStr::from_bytes_with_nul_unchecked(bytes) }
+ }
+
+ #[cold]
+ fn fabricate_nul_error(b: &str) -> NulError {
+ CString::new(b).unwrap_err()
+ }
+}
+
+impl Default for SmallCString {
+ #[inline]
+ fn default() -> Self {
+ Self(smallvec![0])
+ }
+}
+
+impl std::fmt::Debug for SmallCString {
+ fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+ f.debug_tuple("SmallCString").field(&self.as_str()).finish()
+ }
+}
+
+impl std::ops::Deref for SmallCString {
+ type Target = CStr;
+
+ #[inline]
+ fn deref(&self) -> &CStr {
+ self.as_cstr()
+ }
+}
+
+impl PartialEq<SmallCString> for str {
+ #[inline]
+ fn eq(&self, s: &SmallCString) -> bool {
+ s.as_bytes_without_nul() == self.as_bytes()
+ }
+}
+
+impl PartialEq<str> for SmallCString {
+ #[inline]
+ fn eq(&self, s: &str) -> bool {
+ self.as_bytes_without_nul() == s.as_bytes()
+ }
+}
+
+impl std::borrow::Borrow<str> for SmallCString {
+ #[inline]
+ fn borrow(&self) -> &str {
+ self.as_str()
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use super::*;
+
+ #[test]
+ fn test_small_cstring() {
+ // We don't go through the normal machinery for default, so make sure
+ // things work.
+ assert_eq!(SmallCString::default().0, SmallCString::new("").unwrap().0);
+ assert_eq!(SmallCString::new("foo").unwrap().len(), 3);
+ assert_eq!(
+ SmallCString::new("foo").unwrap().as_bytes_with_nul(),
+ b"foo\0"
+ );
+ assert_eq!(
+ SmallCString::new("foo").unwrap().as_bytes_without_nul(),
+ b"foo",
+ );
+
+ assert_eq!(SmallCString::new("😀").unwrap().len(), 4);
+ assert_eq!(
+ SmallCString::new("😀").unwrap().0.as_slice(),
+ b"\xf0\x9f\x98\x80\0",
+ );
+ assert_eq!(
+ SmallCString::new("😀").unwrap().as_bytes_without_nul(),
+ b"\xf0\x9f\x98\x80",
+ );
+
+ assert_eq!(SmallCString::new("").unwrap().len(), 0);
+ assert!(SmallCString::new("").unwrap().is_empty());
+
+ assert_eq!(SmallCString::new("").unwrap().0.as_slice(), b"\0");
+ assert_eq!(SmallCString::new("").unwrap().as_bytes_without_nul(), b"");
+
+ SmallCString::new("\0").unwrap_err();
+ SmallCString::new("\0abc").unwrap_err();
+ SmallCString::new("abc\0").unwrap_err();
+ }
+}
diff --git a/third_party/rust/rusqlite/src/util/sqlite_string.rs b/third_party/rust/rusqlite/src/util/sqlite_string.rs
new file mode 100644
index 0000000000..1d69552c9d
--- /dev/null
+++ b/third_party/rust/rusqlite/src/util/sqlite_string.rs
@@ -0,0 +1,239 @@
+// This is used when either vtab or modern-sqlite is on. Different methods are
+// used in each feature. Avoid having to track this for each function. We will
+// still warn for anything that's not used by either, though.
+#![cfg_attr(not(feature = "vtab"), allow(dead_code))]
+use crate::ffi;
+use std::marker::PhantomData;
+use std::os::raw::{c_char, c_int};
+use std::ptr::NonNull;
+
+// Space to hold this string must be obtained
+// from an SQLite memory allocation function
+pub(crate) fn alloc(s: &str) -> *mut c_char {
+ SqliteMallocString::from_str(s).into_raw()
+}
+
+/// A string we own that's allocated on the SQLite heap. Automatically calls
+/// `sqlite3_free` when dropped, unless `into_raw` (or `into_inner`) is called
+/// on it. If constructed from a rust string, `sqlite3_malloc` is used.
+///
+/// It has identical representation to a nonnull `*mut c_char`, so you can use
+/// it transparently as one. It's nonnull, so Option<SqliteMallocString> can be
+/// used for nullable ones (it's still just one pointer).
+///
+/// Most strings shouldn't use this! Only places where the string needs to be
+/// freed with `sqlite3_free`. This includes `sqlite3_extended_sql` results,
+/// some error message pointers... Note that misuse is extremely dangerous!
+///
+/// Note that this is *not* a lossless interface. Incoming strings with internal
+/// NULs are modified, and outgoing strings which are non-UTF8 are modified.
+/// This seems unavoidable -- it tries very hard to not panic.
+#[repr(transparent)]
+pub(crate) struct SqliteMallocString {
+ ptr: NonNull<c_char>,
+ _boo: PhantomData<Box<[c_char]>>,
+}
+// This is owned data for a primitive type, and thus it's safe to implement
+// these. That said, nothing needs them, and they make things easier to misuse.
+
+// unsafe impl Send for SqliteMallocString {}
+// unsafe impl Sync for SqliteMallocString {}
+
+impl SqliteMallocString {
+ /// SAFETY: Caller must be certain that `m` a nul-terminated c string
+ /// allocated by `sqlite3_malloc`, and that SQLite expects us to free it!
+ #[inline]
+ pub(crate) unsafe fn from_raw_nonnull(ptr: NonNull<c_char>) -> Self {
+ Self {
+ ptr,
+ _boo: PhantomData,
+ }
+ }
+
+ /// SAFETY: Caller must be certain that `m` a nul-terminated c string
+ /// allocated by `sqlite3_malloc`, and that SQLite expects us to free it!
+ #[inline]
+ pub(crate) unsafe fn from_raw(ptr: *mut c_char) -> Option<Self> {
+ NonNull::new(ptr).map(|p| Self::from_raw_nonnull(p))
+ }
+
+ /// Get the pointer behind `self`. After this is called, we no longer manage
+ /// it.
+ #[inline]
+ pub(crate) fn into_inner(self) -> NonNull<c_char> {
+ let p = self.ptr;
+ std::mem::forget(self);
+ p
+ }
+
+ /// Get the pointer behind `self`. After this is called, we no longer manage
+ /// it.
+ #[inline]
+ pub(crate) fn into_raw(self) -> *mut c_char {
+ self.into_inner().as_ptr()
+ }
+
+ /// Borrow the pointer behind `self`. We still manage it when this function
+ /// returns. If you want to relinquish ownership, use `into_raw`.
+ #[inline]
+ pub(crate) fn as_ptr(&self) -> *const c_char {
+ self.ptr.as_ptr()
+ }
+
+ #[inline]
+ pub(crate) fn as_cstr(&self) -> &std::ffi::CStr {
+ unsafe { std::ffi::CStr::from_ptr(self.as_ptr()) }
+ }
+
+ #[inline]
+ pub(crate) fn to_string_lossy(&self) -> std::borrow::Cow<'_, str> {
+ self.as_cstr().to_string_lossy()
+ }
+
+ /// Convert `s` into a SQLite string.
+ ///
+ /// This should almost never be done except for cases like error messages or
+ /// other strings that SQLite frees.
+ ///
+ /// If `s` contains internal NULs, we'll replace them with
+ /// `NUL_REPLACE_CHAR`.
+ ///
+ /// Except for `debug_assert`s which may trigger during testing, this
+ /// function never panics. If we hit integer overflow or the allocation
+ /// fails, we call `handle_alloc_error` which aborts the program after
+ /// calling a global hook.
+ ///
+ /// This means it's safe to use in extern "C" functions even outside of
+ /// `catch_unwind`.
+ pub(crate) fn from_str(s: &str) -> Self {
+ let s = if s.as_bytes().contains(&0) {
+ std::borrow::Cow::Owned(make_nonnull(s))
+ } else {
+ std::borrow::Cow::Borrowed(s)
+ };
+ debug_assert!(!s.as_bytes().contains(&0));
+ let bytes: &[u8] = s.as_ref().as_bytes();
+ let src_ptr: *const c_char = bytes.as_ptr().cast();
+ let src_len = bytes.len();
+ let maybe_len_plus_1 = s.len().checked_add(1).and_then(|v| c_int::try_from(v).ok());
+ unsafe {
+ let res_ptr = maybe_len_plus_1
+ .and_then(|len_to_alloc| {
+ // `>` because we added 1.
+ debug_assert!(len_to_alloc > 0);
+ debug_assert_eq!((len_to_alloc - 1) as usize, src_len);
+ NonNull::new(ffi::sqlite3_malloc(len_to_alloc).cast::<c_char>())
+ })
+ .unwrap_or_else(|| {
+ use std::alloc::{handle_alloc_error, Layout};
+ // Report via handle_alloc_error so that it can be handled with any
+ // other allocation errors and properly diagnosed.
+ //
+ // This is safe:
+ // - `align` is never 0
+ // - `align` is always a power of 2.
+ // - `size` needs no realignment because it's guaranteed to be aligned
+ // (everything is aligned to 1)
+ // - `size` is also never zero, although this function doesn't actually require
+ // it now.
+ let len = s.len().saturating_add(1).min(isize::MAX as usize);
+ let layout = Layout::from_size_align_unchecked(len, 1);
+ // Note: This call does not return.
+ handle_alloc_error(layout);
+ });
+ let buf: *mut c_char = res_ptr.as_ptr().cast::<c_char>();
+ src_ptr.copy_to_nonoverlapping(buf, src_len);
+ buf.add(src_len).write(0);
+ debug_assert_eq!(std::ffi::CStr::from_ptr(res_ptr.as_ptr()).to_bytes(), bytes);
+ Self::from_raw_nonnull(res_ptr)
+ }
+ }
+}
+
+const NUL_REPLACE: &str = "␀";
+
+#[cold]
+fn make_nonnull(v: &str) -> String {
+ v.replace('\0', NUL_REPLACE)
+}
+
+impl Drop for SqliteMallocString {
+ #[inline]
+ fn drop(&mut self) {
+ unsafe { ffi::sqlite3_free(self.ptr.as_ptr().cast()) };
+ }
+}
+
+impl std::fmt::Debug for SqliteMallocString {
+ fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+ self.to_string_lossy().fmt(f)
+ }
+}
+
+impl std::fmt::Display for SqliteMallocString {
+ fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+ self.to_string_lossy().fmt(f)
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use super::*;
+ #[test]
+ fn test_from_str() {
+ let to_check = [
+ ("", ""),
+ ("\0", "␀"),
+ ("␀", "␀"),
+ ("\0bar", "␀bar"),
+ ("foo\0bar", "foo␀bar"),
+ ("foo\0", "foo␀"),
+ ("a\0b\0c\0\0d", "a␀b␀c␀␀d"),
+ ("foobar0123", "foobar0123"),
+ ];
+
+ for &(input, output) in &to_check {
+ let s = SqliteMallocString::from_str(input);
+ assert_eq!(s.to_string_lossy(), output);
+ assert_eq!(s.as_cstr().to_str().unwrap(), output);
+ }
+ }
+
+ // This will trigger an asan error if into_raw still freed the ptr.
+ #[test]
+ fn test_lossy() {
+ let p = SqliteMallocString::from_str("abcd").into_raw();
+ // Make invalid
+ let s = unsafe {
+ p.cast::<u8>().write(b'\xff');
+ SqliteMallocString::from_raw(p).unwrap()
+ };
+ assert_eq!(s.to_string_lossy().as_ref(), "\u{FFFD}bcd");
+ }
+
+ // This will trigger an asan error if into_raw still freed the ptr.
+ #[test]
+ fn test_into_raw() {
+ let mut v = vec![];
+ for i in 0..1000 {
+ v.push(SqliteMallocString::from_str(&i.to_string()).into_raw());
+ v.push(SqliteMallocString::from_str(&format!("abc {i} 😀")).into_raw());
+ }
+ unsafe {
+ for (i, s) in v.chunks_mut(2).enumerate() {
+ let s0 = std::mem::replace(&mut s[0], std::ptr::null_mut());
+ let s1 = std::mem::replace(&mut s[1], std::ptr::null_mut());
+ assert_eq!(
+ std::ffi::CStr::from_ptr(s0).to_str().unwrap(),
+ &i.to_string()
+ );
+ assert_eq!(
+ std::ffi::CStr::from_ptr(s1).to_str().unwrap(),
+ &format!("abc {i} 😀")
+ );
+ let _ = SqliteMallocString::from_raw(s0).unwrap();
+ let _ = SqliteMallocString::from_raw(s1).unwrap();
+ }
+ }
+ }
+}
diff --git a/third_party/rust/rusqlite/src/version.rs b/third_party/rust/rusqlite/src/version.rs
new file mode 100644
index 0000000000..44053b741d
--- /dev/null
+++ b/third_party/rust/rusqlite/src/version.rs
@@ -0,0 +1,27 @@
+use crate::ffi;
+use std::ffi::CStr;
+
+/// Returns the SQLite version as an integer; e.g., `3016002` for version
+/// 3.16.2.
+///
+/// See [`sqlite3_libversion_number()`](https://www.sqlite.org/c3ref/libversion.html).
+#[inline]
+#[must_use]
+pub fn version_number() -> i32 {
+ unsafe { ffi::sqlite3_libversion_number() }
+}
+
+/// Returns the SQLite version as a string; e.g., `"3.16.2"` for version 3.16.2.
+///
+/// See [`sqlite3_libversion()`](https://www.sqlite.org/c3ref/libversion.html).
+///
+/// # Panics
+///
+/// Panics when version is not valid UTF-8.
+#[inline]
+#[must_use]
+pub fn version() -> &'static str {
+ let cstr = unsafe { CStr::from_ptr(ffi::sqlite3_libversion()) };
+ cstr.to_str()
+ .expect("SQLite version string is not valid UTF8 ?!")
+}
diff --git a/third_party/rust/rusqlite/src/vtab/array.rs b/third_party/rust/rusqlite/src/vtab/array.rs
new file mode 100644
index 0000000000..be19a3e127
--- /dev/null
+++ b/third_party/rust/rusqlite/src/vtab/array.rs
@@ -0,0 +1,223 @@
+//! Array Virtual Table.
+//!
+//! Note: `rarray`, not `carray` is the name of the table valued function we
+//! define.
+//!
+//! Port of [carray](http://www.sqlite.org/cgi/src/finfo?name=ext/misc/carray.c)
+//! C extension: `https://www.sqlite.org/carray.html`
+//!
+//! # Example
+//!
+//! ```rust,no_run
+//! # use rusqlite::{types::Value, Connection, Result, params};
+//! # use std::rc::Rc;
+//! fn example(db: &Connection) -> Result<()> {
+//! // Note: This should be done once (usually when opening the DB).
+//! rusqlite::vtab::array::load_module(&db)?;
+//! let v = [1i64, 2, 3, 4];
+//! // Note: A `Rc<Vec<Value>>` must be used as the parameter.
+//! let values = Rc::new(v.iter().copied().map(Value::from).collect::<Vec<Value>>());
+//! let mut stmt = db.prepare("SELECT value from rarray(?1);")?;
+//! let rows = stmt.query_map([values], |row| row.get::<_, i64>(0))?;
+//! for value in rows {
+//! println!("{}", value?);
+//! }
+//! Ok(())
+//! }
+//! ```
+
+use std::default::Default;
+use std::marker::PhantomData;
+use std::os::raw::{c_char, c_int, c_void};
+use std::rc::Rc;
+
+use crate::ffi;
+use crate::types::{ToSql, ToSqlOutput, Value};
+use crate::vtab::{
+ eponymous_only_module, Context, IndexConstraintOp, IndexInfo, VTab, VTabConnection, VTabCursor,
+ Values,
+};
+use crate::{Connection, Result};
+
+// http://sqlite.org/bindptr.html
+
+pub(crate) const ARRAY_TYPE: *const c_char = (b"rarray\0" as *const u8).cast::<c_char>();
+
+pub(crate) unsafe extern "C" fn free_array(p: *mut c_void) {
+ drop(Rc::from_raw(p as *const Vec<Value>));
+}
+
+/// Array parameter / pointer
+pub type Array = Rc<Vec<Value>>;
+
+impl ToSql for Array {
+ #[inline]
+ fn to_sql(&self) -> Result<ToSqlOutput<'_>> {
+ Ok(ToSqlOutput::Array(self.clone()))
+ }
+}
+
+/// Register the "rarray" module.
+pub fn load_module(conn: &Connection) -> Result<()> {
+ let aux: Option<()> = None;
+ conn.create_module("rarray", eponymous_only_module::<ArrayTab>(), aux)
+}
+
+// Column numbers
+// const CARRAY_COLUMN_VALUE : c_int = 0;
+const CARRAY_COLUMN_POINTER: c_int = 1;
+
+/// An instance of the Array virtual table
+#[repr(C)]
+struct ArrayTab {
+ /// Base class. Must be first
+ base: ffi::sqlite3_vtab,
+}
+
+unsafe impl<'vtab> VTab<'vtab> for ArrayTab {
+ type Aux = ();
+ type Cursor = ArrayTabCursor<'vtab>;
+
+ fn connect(
+ _: &mut VTabConnection,
+ _aux: Option<&()>,
+ _args: &[&[u8]],
+ ) -> Result<(String, ArrayTab)> {
+ let vtab = ArrayTab {
+ base: ffi::sqlite3_vtab::default(),
+ };
+ Ok(("CREATE TABLE x(value,pointer hidden)".to_owned(), vtab))
+ }
+
+ fn best_index(&self, info: &mut IndexInfo) -> Result<()> {
+ // Index of the pointer= constraint
+ let mut ptr_idx = false;
+ for (constraint, mut constraint_usage) in info.constraints_and_usages() {
+ if !constraint.is_usable() {
+ continue;
+ }
+ if constraint.operator() != IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_EQ {
+ continue;
+ }
+ if let CARRAY_COLUMN_POINTER = constraint.column() {
+ ptr_idx = true;
+ constraint_usage.set_argv_index(1);
+ constraint_usage.set_omit(true);
+ }
+ }
+ if ptr_idx {
+ info.set_estimated_cost(1_f64);
+ info.set_estimated_rows(100);
+ info.set_idx_num(1);
+ } else {
+ info.set_estimated_cost(2_147_483_647_f64);
+ info.set_estimated_rows(2_147_483_647);
+ info.set_idx_num(0);
+ }
+ Ok(())
+ }
+
+ fn open(&mut self) -> Result<ArrayTabCursor<'_>> {
+ Ok(ArrayTabCursor::new())
+ }
+}
+
+/// A cursor for the Array virtual table
+#[repr(C)]
+struct ArrayTabCursor<'vtab> {
+ /// Base class. Must be first
+ base: ffi::sqlite3_vtab_cursor,
+ /// The rowid
+ row_id: i64,
+ /// Pointer to the array of values ("pointer")
+ ptr: Option<Array>,
+ phantom: PhantomData<&'vtab ArrayTab>,
+}
+
+impl ArrayTabCursor<'_> {
+ fn new<'vtab>() -> ArrayTabCursor<'vtab> {
+ ArrayTabCursor {
+ base: ffi::sqlite3_vtab_cursor::default(),
+ row_id: 0,
+ ptr: None,
+ phantom: PhantomData,
+ }
+ }
+
+ fn len(&self) -> i64 {
+ match self.ptr {
+ Some(ref a) => a.len() as i64,
+ _ => 0,
+ }
+ }
+}
+unsafe impl VTabCursor for ArrayTabCursor<'_> {
+ fn filter(&mut self, idx_num: c_int, _idx_str: Option<&str>, args: &Values<'_>) -> Result<()> {
+ if idx_num > 0 {
+ self.ptr = args.get_array(0);
+ } else {
+ self.ptr = None;
+ }
+ self.row_id = 1;
+ Ok(())
+ }
+
+ fn next(&mut self) -> Result<()> {
+ self.row_id += 1;
+ Ok(())
+ }
+
+ fn eof(&self) -> bool {
+ self.row_id > self.len()
+ }
+
+ fn column(&self, ctx: &mut Context, i: c_int) -> Result<()> {
+ match i {
+ CARRAY_COLUMN_POINTER => Ok(()),
+ _ => {
+ if let Some(ref array) = self.ptr {
+ let value = &array[(self.row_id - 1) as usize];
+ ctx.set_result(&value)
+ } else {
+ Ok(())
+ }
+ }
+ }
+ }
+
+ fn rowid(&self) -> Result<i64> {
+ Ok(self.row_id)
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use crate::types::Value;
+ use crate::vtab::array;
+ use crate::{Connection, Result};
+ use std::rc::Rc;
+
+ #[test]
+ fn test_array_module() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ array::load_module(&db)?;
+
+ let v = vec![1i64, 2, 3, 4];
+ let values: Vec<Value> = v.into_iter().map(Value::from).collect();
+ let ptr = Rc::new(values);
+ {
+ let mut stmt = db.prepare("SELECT value from rarray(?1);")?;
+
+ let rows = stmt.query_map([&ptr], |row| row.get::<_, i64>(0))?;
+ assert_eq!(2, Rc::strong_count(&ptr));
+ let mut count = 0;
+ for (i, value) in rows.enumerate() {
+ assert_eq!(i as i64, value? - 1);
+ count += 1;
+ }
+ assert_eq!(4, count);
+ }
+ assert_eq!(1, Rc::strong_count(&ptr));
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/vtab/csvtab.rs b/third_party/rust/rusqlite/src/vtab/csvtab.rs
new file mode 100644
index 0000000000..acf3cd8180
--- /dev/null
+++ b/third_party/rust/rusqlite/src/vtab/csvtab.rs
@@ -0,0 +1,387 @@
+//! CSV Virtual Table.
+//!
+//! Port of [csv](http://www.sqlite.org/cgi/src/finfo?name=ext/misc/csv.c) C
+//! extension: `https://www.sqlite.org/csv.html`
+//!
+//! # Example
+//!
+//! ```rust,no_run
+//! # use rusqlite::{Connection, Result};
+//! fn example() -> Result<()> {
+//! // Note: This should be done once (usually when opening the DB).
+//! let db = Connection::open_in_memory()?;
+//! rusqlite::vtab::csvtab::load_module(&db)?;
+//! // Assume my_csv.csv
+//! let schema = "
+//! CREATE VIRTUAL TABLE my_csv_data
+//! USING csv(filename = 'my_csv.csv')
+//! ";
+//! db.execute_batch(schema)?;
+//! // Now the `my_csv_data` (virtual) table can be queried as normal...
+//! Ok(())
+//! }
+//! ```
+use std::fs::File;
+use std::marker::PhantomData;
+use std::os::raw::c_int;
+use std::path::Path;
+use std::str;
+
+use crate::ffi;
+use crate::types::Null;
+use crate::vtab::{
+ escape_double_quote, parse_boolean, read_only_module, Context, CreateVTab, IndexInfo, VTab,
+ VTabConfig, VTabConnection, VTabCursor, VTabKind, Values,
+};
+use crate::{Connection, Error, Result};
+
+/// Register the "csv" module.
+/// ```sql
+/// CREATE VIRTUAL TABLE vtab USING csv(
+/// filename=FILENAME -- Name of file containing CSV content
+/// [, schema=SCHEMA] -- Alternative CSV schema. 'CREATE TABLE x(col1 TEXT NOT NULL, col2 INT, ...);'
+/// [, header=YES|NO] -- First row of CSV defines the names of columns if "yes". Default "no".
+/// [, columns=N] -- Assume the CSV file contains N columns.
+/// [, delimiter=C] -- CSV delimiter. Default ','.
+/// [, quote=C] -- CSV quote. Default '"'. 0 means no quote.
+/// );
+/// ```
+pub fn load_module(conn: &Connection) -> Result<()> {
+ let aux: Option<()> = None;
+ conn.create_module("csv", read_only_module::<CsvTab>(), aux)
+}
+
+/// An instance of the CSV virtual table
+#[repr(C)]
+struct CsvTab {
+ /// Base class. Must be first
+ base: ffi::sqlite3_vtab,
+ /// Name of the CSV file
+ filename: String,
+ has_headers: bool,
+ delimiter: u8,
+ quote: u8,
+ /// Offset to start of data
+ offset_first_row: csv::Position,
+}
+
+impl CsvTab {
+ fn reader(&self) -> Result<csv::Reader<File>, csv::Error> {
+ csv::ReaderBuilder::new()
+ .has_headers(self.has_headers)
+ .delimiter(self.delimiter)
+ .quote(self.quote)
+ .from_path(&self.filename)
+ }
+
+ fn parse_byte(arg: &str) -> Option<u8> {
+ if arg.len() == 1 {
+ arg.bytes().next()
+ } else {
+ None
+ }
+ }
+}
+
+unsafe impl<'vtab> VTab<'vtab> for CsvTab {
+ type Aux = ();
+ type Cursor = CsvTabCursor<'vtab>;
+
+ fn connect(
+ db: &mut VTabConnection,
+ _aux: Option<&()>,
+ args: &[&[u8]],
+ ) -> Result<(String, CsvTab)> {
+ if args.len() < 4 {
+ return Err(Error::ModuleError("no CSV file specified".to_owned()));
+ }
+
+ let mut vtab = CsvTab {
+ base: ffi::sqlite3_vtab::default(),
+ filename: "".to_owned(),
+ has_headers: false,
+ delimiter: b',',
+ quote: b'"',
+ offset_first_row: csv::Position::new(),
+ };
+ let mut schema = None;
+ let mut n_col = None;
+
+ let args = &args[3..];
+ for c_slice in args {
+ let (param, value) = super::parameter(c_slice)?;
+ match param {
+ "filename" => {
+ if !Path::new(value).exists() {
+ return Err(Error::ModuleError(format!("file '{value}' does not exist")));
+ }
+ vtab.filename = value.to_owned();
+ }
+ "schema" => {
+ schema = Some(value.to_owned());
+ }
+ "columns" => {
+ if let Ok(n) = value.parse::<u16>() {
+ if n_col.is_some() {
+ return Err(Error::ModuleError(
+ "more than one 'columns' parameter".to_owned(),
+ ));
+ } else if n == 0 {
+ return Err(Error::ModuleError(
+ "must have at least one column".to_owned(),
+ ));
+ }
+ n_col = Some(n);
+ } else {
+ return Err(Error::ModuleError(format!(
+ "unrecognized argument to 'columns': {value}"
+ )));
+ }
+ }
+ "header" => {
+ if let Some(b) = parse_boolean(value) {
+ vtab.has_headers = b;
+ } else {
+ return Err(Error::ModuleError(format!(
+ "unrecognized argument to 'header': {value}"
+ )));
+ }
+ }
+ "delimiter" => {
+ if let Some(b) = CsvTab::parse_byte(value) {
+ vtab.delimiter = b;
+ } else {
+ return Err(Error::ModuleError(format!(
+ "unrecognized argument to 'delimiter': {value}"
+ )));
+ }
+ }
+ "quote" => {
+ if let Some(b) = CsvTab::parse_byte(value) {
+ if b == b'0' {
+ vtab.quote = 0;
+ } else {
+ vtab.quote = b;
+ }
+ } else {
+ return Err(Error::ModuleError(format!(
+ "unrecognized argument to 'quote': {value}"
+ )));
+ }
+ }
+ _ => {
+ return Err(Error::ModuleError(format!(
+ "unrecognized parameter '{param}'"
+ )));
+ }
+ }
+ }
+
+ if vtab.filename.is_empty() {
+ return Err(Error::ModuleError("no CSV file specified".to_owned()));
+ }
+
+ let mut cols: Vec<String> = Vec::new();
+ if vtab.has_headers || (n_col.is_none() && schema.is_none()) {
+ let mut reader = vtab.reader()?;
+ if vtab.has_headers {
+ {
+ let headers = reader.headers()?;
+ // headers ignored if cols is not empty
+ if n_col.is_none() && schema.is_none() {
+ cols = headers
+ .into_iter()
+ .map(|header| escape_double_quote(header).into_owned())
+ .collect();
+ }
+ }
+ vtab.offset_first_row = reader.position().clone();
+ } else {
+ let mut record = csv::ByteRecord::new();
+ if reader.read_byte_record(&mut record)? {
+ for (i, _) in record.iter().enumerate() {
+ cols.push(format!("c{i}"));
+ }
+ }
+ }
+ } else if let Some(n_col) = n_col {
+ for i in 0..n_col {
+ cols.push(format!("c{i}"));
+ }
+ }
+
+ if cols.is_empty() && schema.is_none() {
+ return Err(Error::ModuleError("no column specified".to_owned()));
+ }
+
+ if schema.is_none() {
+ let mut sql = String::from("CREATE TABLE x(");
+ for (i, col) in cols.iter().enumerate() {
+ sql.push('"');
+ sql.push_str(col);
+ sql.push_str("\" TEXT");
+ if i == cols.len() - 1 {
+ sql.push_str(");");
+ } else {
+ sql.push_str(", ");
+ }
+ }
+ schema = Some(sql);
+ }
+ db.config(VTabConfig::DirectOnly)?;
+ Ok((schema.unwrap(), vtab))
+ }
+
+ // Only a forward full table scan is supported.
+ fn best_index(&self, info: &mut IndexInfo) -> Result<()> {
+ info.set_estimated_cost(1_000_000.);
+ Ok(())
+ }
+
+ fn open(&mut self) -> Result<CsvTabCursor<'_>> {
+ Ok(CsvTabCursor::new(self.reader()?))
+ }
+}
+
+impl CreateVTab<'_> for CsvTab {
+ const KIND: VTabKind = VTabKind::Default;
+}
+
+/// A cursor for the CSV virtual table
+#[repr(C)]
+struct CsvTabCursor<'vtab> {
+ /// Base class. Must be first
+ base: ffi::sqlite3_vtab_cursor,
+ /// The CSV reader object
+ reader: csv::Reader<File>,
+ /// Current cursor position used as rowid
+ row_number: usize,
+ /// Values of the current row
+ cols: csv::StringRecord,
+ eof: bool,
+ phantom: PhantomData<&'vtab CsvTab>,
+}
+
+impl CsvTabCursor<'_> {
+ fn new<'vtab>(reader: csv::Reader<File>) -> CsvTabCursor<'vtab> {
+ CsvTabCursor {
+ base: ffi::sqlite3_vtab_cursor::default(),
+ reader,
+ row_number: 0,
+ cols: csv::StringRecord::new(),
+ eof: false,
+ phantom: PhantomData,
+ }
+ }
+
+ /// Accessor to the associated virtual table.
+ fn vtab(&self) -> &CsvTab {
+ unsafe { &*(self.base.pVtab as *const CsvTab) }
+ }
+}
+
+unsafe impl VTabCursor for CsvTabCursor<'_> {
+ // Only a full table scan is supported. So `filter` simply rewinds to
+ // the beginning.
+ fn filter(
+ &mut self,
+ _idx_num: c_int,
+ _idx_str: Option<&str>,
+ _args: &Values<'_>,
+ ) -> Result<()> {
+ {
+ let offset_first_row = self.vtab().offset_first_row.clone();
+ self.reader.seek(offset_first_row)?;
+ }
+ self.row_number = 0;
+ self.next()
+ }
+
+ fn next(&mut self) -> Result<()> {
+ {
+ self.eof = self.reader.is_done();
+ if self.eof {
+ return Ok(());
+ }
+
+ self.eof = !self.reader.read_record(&mut self.cols)?;
+ }
+
+ self.row_number += 1;
+ Ok(())
+ }
+
+ fn eof(&self) -> bool {
+ self.eof
+ }
+
+ fn column(&self, ctx: &mut Context, col: c_int) -> Result<()> {
+ if col < 0 || col as usize >= self.cols.len() {
+ return Err(Error::ModuleError(format!(
+ "column index out of bounds: {col}"
+ )));
+ }
+ if self.cols.is_empty() {
+ return ctx.set_result(&Null);
+ }
+ // TODO Affinity
+ ctx.set_result(&self.cols[col as usize].to_owned())
+ }
+
+ fn rowid(&self) -> Result<i64> {
+ Ok(self.row_number as i64)
+ }
+}
+
+impl From<csv::Error> for Error {
+ #[cold]
+ fn from(err: csv::Error) -> Error {
+ Error::ModuleError(err.to_string())
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use crate::vtab::csvtab;
+ use crate::{Connection, Result};
+ use fallible_iterator::FallibleIterator;
+
+ #[test]
+ fn test_csv_module() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ csvtab::load_module(&db)?;
+ db.execute_batch("CREATE VIRTUAL TABLE vtab USING csv(filename='test.csv', header=yes)")?;
+
+ {
+ let mut s = db.prepare("SELECT rowid, * FROM vtab")?;
+ {
+ let headers = s.column_names();
+ assert_eq!(vec!["rowid", "colA", "colB", "colC"], headers);
+ }
+
+ let ids: Result<Vec<i32>> = s.query([])?.map(|row| row.get::<_, i32>(0)).collect();
+ let sum = ids?.iter().sum::<i32>();
+ assert_eq!(sum, 15);
+ }
+ db.execute_batch("DROP TABLE vtab")
+ }
+
+ #[test]
+ fn test_csv_cursor() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ csvtab::load_module(&db)?;
+ db.execute_batch("CREATE VIRTUAL TABLE vtab USING csv(filename='test.csv', header=yes)")?;
+
+ {
+ let mut s = db.prepare(
+ "SELECT v1.rowid, v1.* FROM vtab v1 NATURAL JOIN vtab v2 WHERE \
+ v1.rowid < v2.rowid",
+ )?;
+
+ let mut rows = s.query([])?;
+ let row = rows.next()?.unwrap();
+ assert_eq!(row.get_unwrap::<_, i32>(0), 2);
+ }
+ db.execute_batch("DROP TABLE vtab")
+ }
+}
diff --git a/third_party/rust/rusqlite/src/vtab/mod.rs b/third_party/rust/rusqlite/src/vtab/mod.rs
new file mode 100644
index 0000000000..7e2f5f54f2
--- /dev/null
+++ b/third_party/rust/rusqlite/src/vtab/mod.rs
@@ -0,0 +1,1333 @@
+//! Create virtual tables.
+//!
+//! Follow these steps to create your own virtual table:
+//! 1. Write implementation of [`VTab`] and [`VTabCursor`] traits.
+//! 2. Create an instance of the [`Module`] structure specialized for [`VTab`]
+//! impl. from step 1.
+//! 3. Register your [`Module`] structure using [`Connection::create_module`].
+//! 4. Run a `CREATE VIRTUAL TABLE` command that specifies the new module in the
+//! `USING` clause.
+//!
+//! (See [SQLite doc](http://sqlite.org/vtab.html))
+use std::borrow::Cow::{self, Borrowed, Owned};
+use std::marker::PhantomData;
+use std::marker::Sync;
+use std::os::raw::{c_char, c_int, c_void};
+use std::ptr;
+use std::slice;
+
+use crate::context::set_result;
+use crate::error::{error_from_sqlite_code, to_sqlite_error};
+use crate::ffi;
+pub use crate::ffi::{sqlite3_vtab, sqlite3_vtab_cursor};
+use crate::types::{FromSql, FromSqlError, ToSql, ValueRef};
+use crate::util::alloc;
+use crate::{str_to_cstring, Connection, Error, InnerConnection, Result};
+
+// let conn: Connection = ...;
+// let mod: Module = ...; // VTab builder
+// conn.create_module("module", mod);
+//
+// conn.execute("CREATE VIRTUAL TABLE foo USING module(...)");
+// \-> Module::xcreate
+// |-> let vtab: VTab = ...; // on the heap
+// \-> conn.declare_vtab("CREATE TABLE foo (...)");
+// conn = Connection::open(...);
+// \-> Module::xconnect
+// |-> let vtab: VTab = ...; // on the heap
+// \-> conn.declare_vtab("CREATE TABLE foo (...)");
+//
+// conn.close();
+// \-> vtab.xdisconnect
+// conn.execute("DROP TABLE foo");
+// \-> vtab.xDestroy
+//
+// let stmt = conn.prepare("SELECT ... FROM foo WHERE ...");
+// \-> vtab.xbestindex
+// stmt.query().next();
+// \-> vtab.xopen
+// |-> let cursor: VTabCursor = ...; // on the heap
+// |-> cursor.xfilter or xnext
+// |-> cursor.xeof
+// \-> if not eof { cursor.column or xrowid } else { cursor.xclose }
+//
+
+// db: *mut ffi::sqlite3 => VTabConnection
+// module: *const ffi::sqlite3_module => Module
+// aux: *mut c_void => Module::Aux
+// ffi::sqlite3_vtab => VTab
+// ffi::sqlite3_vtab_cursor => VTabCursor
+
+/// Virtual table kind
+pub enum VTabKind {
+ /// Non-eponymous
+ Default,
+ /// [`create`](CreateVTab::create) == [`connect`](VTab::connect)
+ ///
+ /// See [SQLite doc](https://sqlite.org/vtab.html#eponymous_virtual_tables)
+ Eponymous,
+ /// No [`create`](CreateVTab::create) / [`destroy`](CreateVTab::destroy) or
+ /// not used
+ ///
+ /// SQLite >= 3.9.0
+ ///
+ /// See [SQLite doc](https://sqlite.org/vtab.html#eponymous_only_virtual_tables)
+ EponymousOnly,
+}
+
+/// Virtual table module
+///
+/// (See [SQLite doc](https://sqlite.org/c3ref/module.html))
+#[repr(transparent)]
+pub struct Module<'vtab, T: VTab<'vtab>> {
+ base: ffi::sqlite3_module,
+ phantom: PhantomData<&'vtab T>,
+}
+
+unsafe impl<'vtab, T: VTab<'vtab>> Send for Module<'vtab, T> {}
+unsafe impl<'vtab, T: VTab<'vtab>> Sync for Module<'vtab, T> {}
+
+union ModuleZeroHack {
+ bytes: [u8; std::mem::size_of::<ffi::sqlite3_module>()],
+ module: ffi::sqlite3_module,
+}
+
+// Used as a trailing initializer for sqlite3_module -- this way we avoid having
+// the build fail if buildtime_bindgen is on. This is safe, as bindgen-generated
+// structs are allowed to be zeroed.
+const ZERO_MODULE: ffi::sqlite3_module = unsafe {
+ ModuleZeroHack {
+ bytes: [0_u8; std::mem::size_of::<ffi::sqlite3_module>()],
+ }
+ .module
+};
+
+macro_rules! module {
+ ($lt:lifetime, $vt:ty, $ct:ty, $xc:expr, $xd:expr, $xu:expr) => {
+ #[allow(clippy::needless_update)]
+ &Module {
+ base: ffi::sqlite3_module {
+ // We don't use V3
+ iVersion: 2,
+ xCreate: $xc,
+ xConnect: Some(rust_connect::<$vt>),
+ xBestIndex: Some(rust_best_index::<$vt>),
+ xDisconnect: Some(rust_disconnect::<$vt>),
+ xDestroy: $xd,
+ xOpen: Some(rust_open::<$vt>),
+ xClose: Some(rust_close::<$ct>),
+ xFilter: Some(rust_filter::<$ct>),
+ xNext: Some(rust_next::<$ct>),
+ xEof: Some(rust_eof::<$ct>),
+ xColumn: Some(rust_column::<$ct>),
+ xRowid: Some(rust_rowid::<$ct>), // FIXME optional
+ xUpdate: $xu,
+ xBegin: None,
+ xSync: None,
+ xCommit: None,
+ xRollback: None,
+ xFindFunction: None,
+ xRename: None,
+ xSavepoint: None,
+ xRelease: None,
+ xRollbackTo: None,
+ ..ZERO_MODULE
+ },
+ phantom: PhantomData::<&$lt $vt>,
+ }
+ };
+}
+
+/// Create an modifiable virtual table implementation.
+///
+/// Step 2 of [Creating New Virtual Table Implementations](https://sqlite.org/vtab.html#creating_new_virtual_table_implementations).
+#[must_use]
+pub fn update_module<'vtab, T: UpdateVTab<'vtab>>() -> &'static Module<'vtab, T> {
+ match T::KIND {
+ VTabKind::EponymousOnly => {
+ module!('vtab, T, T::Cursor, None, None, Some(rust_update::<T>))
+ }
+ VTabKind::Eponymous => {
+ module!('vtab, T, T::Cursor, Some(rust_connect::<T>), Some(rust_disconnect::<T>), Some(rust_update::<T>))
+ }
+ _ => {
+ module!('vtab, T, T::Cursor, Some(rust_create::<T>), Some(rust_destroy::<T>), Some(rust_update::<T>))
+ }
+ }
+}
+
+/// Create a read-only virtual table implementation.
+///
+/// Step 2 of [Creating New Virtual Table Implementations](https://sqlite.org/vtab.html#creating_new_virtual_table_implementations).
+#[must_use]
+pub fn read_only_module<'vtab, T: CreateVTab<'vtab>>() -> &'static Module<'vtab, T> {
+ match T::KIND {
+ VTabKind::EponymousOnly => eponymous_only_module(),
+ VTabKind::Eponymous => {
+ // A virtual table is eponymous if its xCreate method is the exact same function
+ // as the xConnect method
+ module!('vtab, T, T::Cursor, Some(rust_connect::<T>), Some(rust_disconnect::<T>), None)
+ }
+ _ => {
+ // The xConnect and xCreate methods may do the same thing, but they must be
+ // different so that the virtual table is not an eponymous virtual table.
+ module!('vtab, T, T::Cursor, Some(rust_create::<T>), Some(rust_destroy::<T>), None)
+ }
+ }
+}
+
+/// Create an eponymous only virtual table implementation.
+///
+/// Step 2 of [Creating New Virtual Table Implementations](https://sqlite.org/vtab.html#creating_new_virtual_table_implementations).
+#[must_use]
+pub fn eponymous_only_module<'vtab, T: VTab<'vtab>>() -> &'static Module<'vtab, T> {
+ // For eponymous-only virtual tables, the xCreate method is NULL
+ module!('vtab, T, T::Cursor, None, None, None)
+}
+
+/// Virtual table configuration options
+#[repr(i32)]
+#[non_exhaustive]
+#[derive(Debug, Clone, Copy, Eq, PartialEq)]
+pub enum VTabConfig {
+ /// Equivalent to SQLITE_VTAB_CONSTRAINT_SUPPORT
+ ConstraintSupport = 1,
+ /// Equivalent to SQLITE_VTAB_INNOCUOUS
+ Innocuous = 2,
+ /// Equivalent to SQLITE_VTAB_DIRECTONLY
+ DirectOnly = 3,
+ /// Equivalent to SQLITE_VTAB_USES_ALL_SCHEMAS
+ UsesAllSchemas = 4,
+}
+
+/// `feature = "vtab"`
+pub struct VTabConnection(*mut ffi::sqlite3);
+
+impl VTabConnection {
+ /// Configure various facets of the virtual table interface
+ pub fn config(&mut self, config: VTabConfig) -> Result<()> {
+ crate::error::check(unsafe { ffi::sqlite3_vtab_config(self.0, config as c_int) })
+ }
+
+ // TODO sqlite3_vtab_on_conflict (http://sqlite.org/c3ref/vtab_on_conflict.html) & xUpdate
+
+ /// Get access to the underlying SQLite database connection handle.
+ ///
+ /// # Warning
+ ///
+ /// You should not need to use this function. If you do need to, please
+ /// [open an issue on the rusqlite repository](https://github.com/rusqlite/rusqlite/issues) and describe
+ /// your use case.
+ ///
+ /// # Safety
+ ///
+ /// This function is unsafe because it gives you raw access
+ /// to the SQLite connection, and what you do with it could impact the
+ /// safety of this `Connection`.
+ pub unsafe fn handle(&mut self) -> *mut ffi::sqlite3 {
+ self.0
+ }
+}
+
+/// Eponymous-only virtual table instance trait.
+///
+/// # Safety
+///
+/// The first item in a struct implementing `VTab` must be
+/// `rusqlite::sqlite3_vtab`, and the struct must be `#[repr(C)]`.
+///
+/// ```rust,ignore
+/// #[repr(C)]
+/// struct MyTab {
+/// /// Base class. Must be first
+/// base: rusqlite::vtab::sqlite3_vtab,
+/// /* Virtual table implementations will typically add additional fields */
+/// }
+/// ```
+///
+/// (See [SQLite doc](https://sqlite.org/c3ref/vtab.html))
+pub unsafe trait VTab<'vtab>: Sized {
+ /// Client data passed to [`Connection::create_module`].
+ type Aux;
+ /// Specific cursor implementation
+ type Cursor: VTabCursor;
+
+ /// Establish a new connection to an existing virtual table.
+ ///
+ /// (See [SQLite doc](https://sqlite.org/vtab.html#the_xconnect_method))
+ fn connect(
+ db: &mut VTabConnection,
+ aux: Option<&Self::Aux>,
+ args: &[&[u8]],
+ ) -> Result<(String, Self)>;
+
+ /// Determine the best way to access the virtual table.
+ /// (See [SQLite doc](https://sqlite.org/vtab.html#the_xbestindex_method))
+ fn best_index(&self, info: &mut IndexInfo) -> Result<()>;
+
+ /// Create a new cursor used for accessing a virtual table.
+ /// (See [SQLite doc](https://sqlite.org/vtab.html#the_xopen_method))
+ fn open(&'vtab mut self) -> Result<Self::Cursor>;
+}
+
+/// Read-only virtual table instance trait.
+///
+/// (See [SQLite doc](https://sqlite.org/c3ref/vtab.html))
+pub trait CreateVTab<'vtab>: VTab<'vtab> {
+ /// For [`EponymousOnly`](VTabKind::EponymousOnly),
+ /// [`create`](CreateVTab::create) and [`destroy`](CreateVTab::destroy) are
+ /// not called
+ const KIND: VTabKind;
+ /// Create a new instance of a virtual table in response to a CREATE VIRTUAL
+ /// TABLE statement. The `db` parameter is a pointer to the SQLite
+ /// database connection that is executing the CREATE VIRTUAL TABLE
+ /// statement.
+ ///
+ /// Call [`connect`](VTab::connect) by default.
+ /// (See [SQLite doc](https://sqlite.org/vtab.html#the_xcreate_method))
+ fn create(
+ db: &mut VTabConnection,
+ aux: Option<&Self::Aux>,
+ args: &[&[u8]],
+ ) -> Result<(String, Self)> {
+ Self::connect(db, aux, args)
+ }
+
+ /// Destroy the underlying table implementation. This method undoes the work
+ /// of [`create`](CreateVTab::create).
+ ///
+ /// Do nothing by default.
+ /// (See [SQLite doc](https://sqlite.org/vtab.html#the_xdestroy_method))
+ fn destroy(&self) -> Result<()> {
+ Ok(())
+ }
+}
+
+/// Writable virtual table instance trait.
+///
+/// (See [SQLite doc](https://sqlite.org/vtab.html#xupdate))
+pub trait UpdateVTab<'vtab>: CreateVTab<'vtab> {
+ /// Delete rowid or PK
+ fn delete(&mut self, arg: ValueRef<'_>) -> Result<()>;
+ /// Insert: `args[0] == NULL: old rowid or PK, args[1]: new rowid or PK,
+ /// args[2]: ...`
+ ///
+ /// Return the new rowid.
+ // TODO Make the distinction between argv[1] == NULL and argv[1] != NULL ?
+ fn insert(&mut self, args: &Values<'_>) -> Result<i64>;
+ /// Update: `args[0] != NULL: old rowid or PK, args[1]: new row id or PK,
+ /// args[2]: ...`
+ fn update(&mut self, args: &Values<'_>) -> Result<()>;
+}
+
+/// Index constraint operator.
+/// See [Virtual Table Constraint Operator Codes](https://sqlite.org/c3ref/c_index_constraint_eq.html) for details.
+#[derive(Debug, Eq, PartialEq)]
+#[allow(non_snake_case, non_camel_case_types, missing_docs)]
+#[allow(clippy::upper_case_acronyms)]
+pub enum IndexConstraintOp {
+ SQLITE_INDEX_CONSTRAINT_EQ,
+ SQLITE_INDEX_CONSTRAINT_GT,
+ SQLITE_INDEX_CONSTRAINT_LE,
+ SQLITE_INDEX_CONSTRAINT_LT,
+ SQLITE_INDEX_CONSTRAINT_GE,
+ SQLITE_INDEX_CONSTRAINT_MATCH,
+ SQLITE_INDEX_CONSTRAINT_LIKE, // 3.10.0
+ SQLITE_INDEX_CONSTRAINT_GLOB, // 3.10.0
+ SQLITE_INDEX_CONSTRAINT_REGEXP, // 3.10.0
+ SQLITE_INDEX_CONSTRAINT_NE, // 3.21.0
+ SQLITE_INDEX_CONSTRAINT_ISNOT, // 3.21.0
+ SQLITE_INDEX_CONSTRAINT_ISNOTNULL, // 3.21.0
+ SQLITE_INDEX_CONSTRAINT_ISNULL, // 3.21.0
+ SQLITE_INDEX_CONSTRAINT_IS, // 3.21.0
+ SQLITE_INDEX_CONSTRAINT_LIMIT, // 3.38.0
+ SQLITE_INDEX_CONSTRAINT_OFFSET, // 3.38.0
+ SQLITE_INDEX_CONSTRAINT_FUNCTION(u8), // 3.25.0
+}
+
+impl From<u8> for IndexConstraintOp {
+ fn from(code: u8) -> IndexConstraintOp {
+ match code {
+ 2 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_EQ,
+ 4 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_GT,
+ 8 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_LE,
+ 16 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_LT,
+ 32 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_GE,
+ 64 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_MATCH,
+ 65 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_LIKE,
+ 66 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_GLOB,
+ 67 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_REGEXP,
+ 68 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_NE,
+ 69 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_ISNOT,
+ 70 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_ISNOTNULL,
+ 71 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_ISNULL,
+ 72 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_IS,
+ 73 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_LIMIT,
+ 74 => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_OFFSET,
+ v => IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_FUNCTION(v),
+ }
+ }
+}
+
+bitflags::bitflags! {
+ /// Virtual table scan flags
+ /// See [Function Flags](https://sqlite.org/c3ref/c_index_scan_unique.html) for details.
+ #[repr(C)]
+ pub struct IndexFlags: ::std::os::raw::c_int {
+ /// Default
+ const NONE = 0;
+ /// Scan visits at most 1 row.
+ const SQLITE_INDEX_SCAN_UNIQUE = ffi::SQLITE_INDEX_SCAN_UNIQUE;
+ }
+}
+
+/// Pass information into and receive the reply from the
+/// [`VTab::best_index`] method.
+///
+/// (See [SQLite doc](http://sqlite.org/c3ref/index_info.html))
+#[derive(Debug)]
+pub struct IndexInfo(*mut ffi::sqlite3_index_info);
+
+impl IndexInfo {
+ /// Iterate on index constraint and its associated usage.
+ #[inline]
+ pub fn constraints_and_usages(&mut self) -> IndexConstraintAndUsageIter<'_> {
+ let constraints =
+ unsafe { slice::from_raw_parts((*self.0).aConstraint, (*self.0).nConstraint as usize) };
+ let constraint_usages = unsafe {
+ slice::from_raw_parts_mut((*self.0).aConstraintUsage, (*self.0).nConstraint as usize)
+ };
+ IndexConstraintAndUsageIter {
+ iter: constraints.iter().zip(constraint_usages.iter_mut()),
+ }
+ }
+
+ /// Record WHERE clause constraints.
+ #[inline]
+ #[must_use]
+ pub fn constraints(&self) -> IndexConstraintIter<'_> {
+ let constraints =
+ unsafe { slice::from_raw_parts((*self.0).aConstraint, (*self.0).nConstraint as usize) };
+ IndexConstraintIter {
+ iter: constraints.iter(),
+ }
+ }
+
+ /// Information about the ORDER BY clause.
+ #[inline]
+ #[must_use]
+ pub fn order_bys(&self) -> OrderByIter<'_> {
+ let order_bys =
+ unsafe { slice::from_raw_parts((*self.0).aOrderBy, (*self.0).nOrderBy as usize) };
+ OrderByIter {
+ iter: order_bys.iter(),
+ }
+ }
+
+ /// Number of terms in the ORDER BY clause
+ #[inline]
+ #[must_use]
+ pub fn num_of_order_by(&self) -> usize {
+ unsafe { (*self.0).nOrderBy as usize }
+ }
+
+ /// Information about what parameters to pass to [`VTabCursor::filter`].
+ #[inline]
+ pub fn constraint_usage(&mut self, constraint_idx: usize) -> IndexConstraintUsage<'_> {
+ let constraint_usages = unsafe {
+ slice::from_raw_parts_mut((*self.0).aConstraintUsage, (*self.0).nConstraint as usize)
+ };
+ IndexConstraintUsage(&mut constraint_usages[constraint_idx])
+ }
+
+ /// Number used to identify the index
+ #[inline]
+ pub fn set_idx_num(&mut self, idx_num: c_int) {
+ unsafe {
+ (*self.0).idxNum = idx_num;
+ }
+ }
+
+ /// String used to identify the index
+ pub fn set_idx_str(&mut self, idx_str: &str) {
+ unsafe {
+ (*self.0).idxStr = alloc(idx_str);
+ (*self.0).needToFreeIdxStr = 1;
+ }
+ }
+
+ /// True if output is already ordered
+ #[inline]
+ pub fn set_order_by_consumed(&mut self, order_by_consumed: bool) {
+ unsafe {
+ (*self.0).orderByConsumed = order_by_consumed as c_int;
+ }
+ }
+
+ /// Estimated cost of using this index
+ #[inline]
+ pub fn set_estimated_cost(&mut self, estimated_ost: f64) {
+ unsafe {
+ (*self.0).estimatedCost = estimated_ost;
+ }
+ }
+
+ /// Estimated number of rows returned.
+ #[inline]
+ pub fn set_estimated_rows(&mut self, estimated_rows: i64) {
+ unsafe {
+ (*self.0).estimatedRows = estimated_rows;
+ }
+ }
+
+ /// Mask of SQLITE_INDEX_SCAN_* flags.
+ #[inline]
+ pub fn set_idx_flags(&mut self, flags: IndexFlags) {
+ unsafe { (*self.0).idxFlags = flags.bits() };
+ }
+
+ /// Mask of columns used by statement
+ #[inline]
+ pub fn col_used(&self) -> u64 {
+ unsafe { (*self.0).colUsed }
+ }
+
+ /// Determine the collation for a virtual table constraint
+ #[cfg(feature = "modern_sqlite")] // SQLite >= 3.22.0
+ #[cfg_attr(docsrs, doc(cfg(feature = "modern_sqlite")))]
+ pub fn collation(&self, constraint_idx: usize) -> Result<&str> {
+ use std::ffi::CStr;
+ let idx = constraint_idx as c_int;
+ let collation = unsafe { ffi::sqlite3_vtab_collation(self.0, idx) };
+ if collation.is_null() {
+ return Err(Error::SqliteFailure(
+ ffi::Error::new(ffi::SQLITE_MISUSE),
+ Some(format!("{constraint_idx} is out of range")),
+ ));
+ }
+ Ok(unsafe { CStr::from_ptr(collation) }.to_str()?)
+ }
+
+ /*/// Determine if a virtual table query is DISTINCT
+ #[cfg(feature = "modern_sqlite")] // SQLite >= 3.38.0
+ #[cfg_attr(docsrs, doc(cfg(feature = "modern_sqlite")))]
+ pub fn distinct(&self) -> c_int {
+ unsafe { ffi::sqlite3_vtab_distinct(self.0) }
+ }
+
+ /// Constraint values
+ #[cfg(feature = "modern_sqlite")] // SQLite >= 3.38.0
+ #[cfg_attr(docsrs, doc(cfg(feature = "modern_sqlite")))]
+ pub fn set_rhs_value(&mut self, constraint_idx: c_int, value: ValueRef) -> Result<()> {
+ // TODO ValueRef to sqlite3_value
+ crate::error::check(unsafe { ffi::sqlite3_vtab_rhs_value(self.O, constraint_idx, value) })
+ }
+
+ /// Identify and handle IN constraints
+ #[cfg(feature = "modern_sqlite")] // SQLite >= 3.38.0
+ #[cfg_attr(docsrs, doc(cfg(feature = "modern_sqlite")))]
+ pub fn set_in_constraint(&mut self, constraint_idx: c_int, b_handle: c_int) -> bool {
+ unsafe { ffi::sqlite3_vtab_in(self.0, constraint_idx, b_handle) != 0 }
+ } // TODO sqlite3_vtab_in_first / sqlite3_vtab_in_next https://sqlite.org/c3ref/vtab_in_first.html
+ */
+}
+
+/// Iterate on index constraint and its associated usage.
+pub struct IndexConstraintAndUsageIter<'a> {
+ iter: std::iter::Zip<
+ slice::Iter<'a, ffi::sqlite3_index_constraint>,
+ slice::IterMut<'a, ffi::sqlite3_index_constraint_usage>,
+ >,
+}
+
+impl<'a> Iterator for IndexConstraintAndUsageIter<'a> {
+ type Item = (IndexConstraint<'a>, IndexConstraintUsage<'a>);
+
+ #[inline]
+ fn next(&mut self) -> Option<(IndexConstraint<'a>, IndexConstraintUsage<'a>)> {
+ self.iter
+ .next()
+ .map(|raw| (IndexConstraint(raw.0), IndexConstraintUsage(raw.1)))
+ }
+
+ #[inline]
+ fn size_hint(&self) -> (usize, Option<usize>) {
+ self.iter.size_hint()
+ }
+}
+
+/// `feature = "vtab"`
+pub struct IndexConstraintIter<'a> {
+ iter: slice::Iter<'a, ffi::sqlite3_index_constraint>,
+}
+
+impl<'a> Iterator for IndexConstraintIter<'a> {
+ type Item = IndexConstraint<'a>;
+
+ #[inline]
+ fn next(&mut self) -> Option<IndexConstraint<'a>> {
+ self.iter.next().map(IndexConstraint)
+ }
+
+ #[inline]
+ fn size_hint(&self) -> (usize, Option<usize>) {
+ self.iter.size_hint()
+ }
+}
+
+/// WHERE clause constraint.
+pub struct IndexConstraint<'a>(&'a ffi::sqlite3_index_constraint);
+
+impl IndexConstraint<'_> {
+ /// Column constrained. -1 for ROWID
+ #[inline]
+ #[must_use]
+ pub fn column(&self) -> c_int {
+ self.0.iColumn
+ }
+
+ /// Constraint operator
+ #[inline]
+ #[must_use]
+ pub fn operator(&self) -> IndexConstraintOp {
+ IndexConstraintOp::from(self.0.op)
+ }
+
+ /// True if this constraint is usable
+ #[inline]
+ #[must_use]
+ pub fn is_usable(&self) -> bool {
+ self.0.usable != 0
+ }
+}
+
+/// Information about what parameters to pass to
+/// [`VTabCursor::filter`].
+pub struct IndexConstraintUsage<'a>(&'a mut ffi::sqlite3_index_constraint_usage);
+
+impl IndexConstraintUsage<'_> {
+ /// if `argv_index` > 0, constraint is part of argv to
+ /// [`VTabCursor::filter`]
+ #[inline]
+ pub fn set_argv_index(&mut self, argv_index: c_int) {
+ self.0.argvIndex = argv_index;
+ }
+
+ /// if `omit`, do not code a test for this constraint
+ #[inline]
+ pub fn set_omit(&mut self, omit: bool) {
+ self.0.omit = omit as std::os::raw::c_uchar;
+ }
+}
+
+/// `feature = "vtab"`
+pub struct OrderByIter<'a> {
+ iter: slice::Iter<'a, ffi::sqlite3_index_orderby>,
+}
+
+impl<'a> Iterator for OrderByIter<'a> {
+ type Item = OrderBy<'a>;
+
+ #[inline]
+ fn next(&mut self) -> Option<OrderBy<'a>> {
+ self.iter.next().map(OrderBy)
+ }
+
+ #[inline]
+ fn size_hint(&self) -> (usize, Option<usize>) {
+ self.iter.size_hint()
+ }
+}
+
+/// A column of the ORDER BY clause.
+pub struct OrderBy<'a>(&'a ffi::sqlite3_index_orderby);
+
+impl OrderBy<'_> {
+ /// Column number
+ #[inline]
+ #[must_use]
+ pub fn column(&self) -> c_int {
+ self.0.iColumn
+ }
+
+ /// True for DESC. False for ASC.
+ #[inline]
+ #[must_use]
+ pub fn is_order_by_desc(&self) -> bool {
+ self.0.desc != 0
+ }
+}
+
+/// Virtual table cursor trait.
+///
+/// # Safety
+///
+/// Implementations must be like:
+/// ```rust,ignore
+/// #[repr(C)]
+/// struct MyTabCursor {
+/// /// Base class. Must be first
+/// base: rusqlite::vtab::sqlite3_vtab_cursor,
+/// /* Virtual table implementations will typically add additional fields */
+/// }
+/// ```
+///
+/// (See [SQLite doc](https://sqlite.org/c3ref/vtab_cursor.html))
+pub unsafe trait VTabCursor: Sized {
+ /// Begin a search of a virtual table.
+ /// (See [SQLite doc](https://sqlite.org/vtab.html#the_xfilter_method))
+ fn filter(&mut self, idx_num: c_int, idx_str: Option<&str>, args: &Values<'_>) -> Result<()>;
+ /// Advance cursor to the next row of a result set initiated by
+ /// [`filter`](VTabCursor::filter). (See [SQLite doc](https://sqlite.org/vtab.html#the_xnext_method))
+ fn next(&mut self) -> Result<()>;
+ /// Must return `false` if the cursor currently points to a valid row of
+ /// data, or `true` otherwise.
+ /// (See [SQLite doc](https://sqlite.org/vtab.html#the_xeof_method))
+ fn eof(&self) -> bool;
+ /// Find the value for the `i`-th column of the current row.
+ /// `i` is zero-based so the first column is numbered 0.
+ /// May return its result back to SQLite using one of the specified `ctx`.
+ /// (See [SQLite doc](https://sqlite.org/vtab.html#the_xcolumn_method))
+ fn column(&self, ctx: &mut Context, i: c_int) -> Result<()>;
+ /// Return the rowid of row that the cursor is currently pointing at.
+ /// (See [SQLite doc](https://sqlite.org/vtab.html#the_xrowid_method))
+ fn rowid(&self) -> Result<i64>;
+}
+
+/// Context is used by [`VTabCursor::column`] to specify the
+/// cell value.
+pub struct Context(*mut ffi::sqlite3_context);
+
+impl Context {
+ /// Set current cell value
+ #[inline]
+ pub fn set_result<T: ToSql>(&mut self, value: &T) -> Result<()> {
+ let t = value.to_sql()?;
+ unsafe { set_result(self.0, &t) };
+ Ok(())
+ }
+
+ // TODO sqlite3_vtab_nochange (http://sqlite.org/c3ref/vtab_nochange.html) // 3.22.0 & xColumn
+}
+
+/// Wrapper to [`VTabCursor::filter`] arguments, the values
+/// requested by [`VTab::best_index`].
+pub struct Values<'a> {
+ args: &'a [*mut ffi::sqlite3_value],
+}
+
+impl Values<'_> {
+ /// Returns the number of values.
+ #[inline]
+ #[must_use]
+ pub fn len(&self) -> usize {
+ self.args.len()
+ }
+
+ /// Returns `true` if there is no value.
+ #[inline]
+ #[must_use]
+ pub fn is_empty(&self) -> bool {
+ self.args.is_empty()
+ }
+
+ /// Returns value at `idx`
+ pub fn get<T: FromSql>(&self, idx: usize) -> Result<T> {
+ let arg = self.args[idx];
+ let value = unsafe { ValueRef::from_value(arg) };
+ FromSql::column_result(value).map_err(|err| match err {
+ FromSqlError::InvalidType => Error::InvalidFilterParameterType(idx, value.data_type()),
+ FromSqlError::Other(err) => {
+ Error::FromSqlConversionFailure(idx, value.data_type(), err)
+ }
+ FromSqlError::InvalidBlobSize { .. } => {
+ Error::FromSqlConversionFailure(idx, value.data_type(), Box::new(err))
+ }
+ FromSqlError::OutOfRange(i) => Error::IntegralValueOutOfRange(idx, i),
+ })
+ }
+
+ // `sqlite3_value_type` returns `SQLITE_NULL` for pointer.
+ // So it seems not possible to enhance `ValueRef::from_value`.
+ #[cfg(feature = "array")]
+ #[cfg_attr(docsrs, doc(cfg(feature = "array")))]
+ fn get_array(&self, idx: usize) -> Option<array::Array> {
+ use crate::types::Value;
+ let arg = self.args[idx];
+ let ptr = unsafe { ffi::sqlite3_value_pointer(arg, array::ARRAY_TYPE) };
+ if ptr.is_null() {
+ None
+ } else {
+ Some(unsafe {
+ let rc = array::Array::from_raw(ptr as *const Vec<Value>);
+ let array = rc.clone();
+ array::Array::into_raw(rc); // don't consume it
+ array
+ })
+ }
+ }
+
+ /// Turns `Values` into an iterator.
+ #[inline]
+ #[must_use]
+ pub fn iter(&self) -> ValueIter<'_> {
+ ValueIter {
+ iter: self.args.iter(),
+ }
+ }
+ // TODO sqlite3_vtab_in_first / sqlite3_vtab_in_next https://sqlite.org/c3ref/vtab_in_first.html & 3.38.0
+}
+
+impl<'a> IntoIterator for &'a Values<'a> {
+ type IntoIter = ValueIter<'a>;
+ type Item = ValueRef<'a>;
+
+ #[inline]
+ fn into_iter(self) -> ValueIter<'a> {
+ self.iter()
+ }
+}
+
+/// [`Values`] iterator.
+pub struct ValueIter<'a> {
+ iter: slice::Iter<'a, *mut ffi::sqlite3_value>,
+}
+
+impl<'a> Iterator for ValueIter<'a> {
+ type Item = ValueRef<'a>;
+
+ #[inline]
+ fn next(&mut self) -> Option<ValueRef<'a>> {
+ self.iter
+ .next()
+ .map(|&raw| unsafe { ValueRef::from_value(raw) })
+ }
+
+ #[inline]
+ fn size_hint(&self) -> (usize, Option<usize>) {
+ self.iter.size_hint()
+ }
+}
+
+impl Connection {
+ /// Register a virtual table implementation.
+ ///
+ /// Step 3 of [Creating New Virtual Table
+ /// Implementations](https://sqlite.org/vtab.html#creating_new_virtual_table_implementations).
+ #[inline]
+ pub fn create_module<'vtab, T: VTab<'vtab>>(
+ &self,
+ module_name: &str,
+ module: &'static Module<'vtab, T>,
+ aux: Option<T::Aux>,
+ ) -> Result<()> {
+ self.db.borrow_mut().create_module(module_name, module, aux)
+ }
+}
+
+impl InnerConnection {
+ fn create_module<'vtab, T: VTab<'vtab>>(
+ &mut self,
+ module_name: &str,
+ module: &'static Module<'vtab, T>,
+ aux: Option<T::Aux>,
+ ) -> Result<()> {
+ use crate::version;
+ if version::version_number() < 3_009_000 && module.base.xCreate.is_none() {
+ return Err(Error::ModuleError(format!(
+ "Eponymous-only virtual table not supported by SQLite version {}",
+ version::version()
+ )));
+ }
+ let c_name = str_to_cstring(module_name)?;
+ let r = match aux {
+ Some(aux) => {
+ let boxed_aux: *mut T::Aux = Box::into_raw(Box::new(aux));
+ unsafe {
+ ffi::sqlite3_create_module_v2(
+ self.db(),
+ c_name.as_ptr(),
+ &module.base,
+ boxed_aux.cast::<c_void>(),
+ Some(free_boxed_value::<T::Aux>),
+ )
+ }
+ }
+ None => unsafe {
+ ffi::sqlite3_create_module_v2(
+ self.db(),
+ c_name.as_ptr(),
+ &module.base,
+ ptr::null_mut(),
+ None,
+ )
+ },
+ };
+ self.decode_result(r)
+ }
+}
+
+/// Escape double-quote (`"`) character occurrences by
+/// doubling them (`""`).
+#[must_use]
+pub fn escape_double_quote(identifier: &str) -> Cow<'_, str> {
+ if identifier.contains('"') {
+ // escape quote by doubling them
+ Owned(identifier.replace('"', "\"\""))
+ } else {
+ Borrowed(identifier)
+ }
+}
+/// Dequote string
+#[must_use]
+pub fn dequote(s: &str) -> &str {
+ if s.len() < 2 {
+ return s;
+ }
+ match s.bytes().next() {
+ Some(b) if b == b'"' || b == b'\'' => match s.bytes().next_back() {
+ Some(e) if e == b => &s[1..s.len() - 1], // FIXME handle inner escaped quote(s)
+ _ => s,
+ },
+ _ => s,
+ }
+}
+/// The boolean can be one of:
+/// ```text
+/// 1 yes true on
+/// 0 no false off
+/// ```
+#[must_use]
+pub fn parse_boolean(s: &str) -> Option<bool> {
+ if s.eq_ignore_ascii_case("yes")
+ || s.eq_ignore_ascii_case("on")
+ || s.eq_ignore_ascii_case("true")
+ || s.eq("1")
+ {
+ Some(true)
+ } else if s.eq_ignore_ascii_case("no")
+ || s.eq_ignore_ascii_case("off")
+ || s.eq_ignore_ascii_case("false")
+ || s.eq("0")
+ {
+ Some(false)
+ } else {
+ None
+ }
+}
+
+/// `<param_name>=['"]?<param_value>['"]?` => `(<param_name>, <param_value>)`
+pub fn parameter(c_slice: &[u8]) -> Result<(&str, &str)> {
+ let arg = std::str::from_utf8(c_slice)?.trim();
+ let mut split = arg.split('=');
+ if let Some(key) = split.next() {
+ if let Some(value) = split.next() {
+ let param = key.trim();
+ let value = dequote(value);
+ return Ok((param, value));
+ }
+ }
+ Err(Error::ModuleError(format!("illegal argument: '{arg}'")))
+}
+
+// FIXME copy/paste from function.rs
+unsafe extern "C" fn free_boxed_value<T>(p: *mut c_void) {
+ drop(Box::from_raw(p.cast::<T>()));
+}
+
+unsafe extern "C" fn rust_create<'vtab, T>(
+ db: *mut ffi::sqlite3,
+ aux: *mut c_void,
+ argc: c_int,
+ argv: *const *const c_char,
+ pp_vtab: *mut *mut ffi::sqlite3_vtab,
+ err_msg: *mut *mut c_char,
+) -> c_int
+where
+ T: CreateVTab<'vtab>,
+{
+ use std::ffi::CStr;
+
+ let mut conn = VTabConnection(db);
+ let aux = aux.cast::<T::Aux>();
+ let args = slice::from_raw_parts(argv, argc as usize);
+ let vec = args
+ .iter()
+ .map(|&cs| CStr::from_ptr(cs).to_bytes()) // FIXME .to_str() -> Result<&str, Utf8Error>
+ .collect::<Vec<_>>();
+ match T::create(&mut conn, aux.as_ref(), &vec[..]) {
+ Ok((sql, vtab)) => match std::ffi::CString::new(sql) {
+ Ok(c_sql) => {
+ let rc = ffi::sqlite3_declare_vtab(db, c_sql.as_ptr());
+ if rc == ffi::SQLITE_OK {
+ let boxed_vtab: *mut T = Box::into_raw(Box::new(vtab));
+ *pp_vtab = boxed_vtab.cast::<ffi::sqlite3_vtab>();
+ ffi::SQLITE_OK
+ } else {
+ let err = error_from_sqlite_code(rc, None);
+ to_sqlite_error(&err, err_msg)
+ }
+ }
+ Err(err) => {
+ *err_msg = alloc(&err.to_string());
+ ffi::SQLITE_ERROR
+ }
+ },
+ Err(err) => to_sqlite_error(&err, err_msg),
+ }
+}
+
+unsafe extern "C" fn rust_connect<'vtab, T>(
+ db: *mut ffi::sqlite3,
+ aux: *mut c_void,
+ argc: c_int,
+ argv: *const *const c_char,
+ pp_vtab: *mut *mut ffi::sqlite3_vtab,
+ err_msg: *mut *mut c_char,
+) -> c_int
+where
+ T: VTab<'vtab>,
+{
+ use std::ffi::CStr;
+
+ let mut conn = VTabConnection(db);
+ let aux = aux.cast::<T::Aux>();
+ let args = slice::from_raw_parts(argv, argc as usize);
+ let vec = args
+ .iter()
+ .map(|&cs| CStr::from_ptr(cs).to_bytes()) // FIXME .to_str() -> Result<&str, Utf8Error>
+ .collect::<Vec<_>>();
+ match T::connect(&mut conn, aux.as_ref(), &vec[..]) {
+ Ok((sql, vtab)) => match std::ffi::CString::new(sql) {
+ Ok(c_sql) => {
+ let rc = ffi::sqlite3_declare_vtab(db, c_sql.as_ptr());
+ if rc == ffi::SQLITE_OK {
+ let boxed_vtab: *mut T = Box::into_raw(Box::new(vtab));
+ *pp_vtab = boxed_vtab.cast::<ffi::sqlite3_vtab>();
+ ffi::SQLITE_OK
+ } else {
+ let err = error_from_sqlite_code(rc, None);
+ to_sqlite_error(&err, err_msg)
+ }
+ }
+ Err(err) => {
+ *err_msg = alloc(&err.to_string());
+ ffi::SQLITE_ERROR
+ }
+ },
+ Err(err) => to_sqlite_error(&err, err_msg),
+ }
+}
+
+unsafe extern "C" fn rust_best_index<'vtab, T>(
+ vtab: *mut ffi::sqlite3_vtab,
+ info: *mut ffi::sqlite3_index_info,
+) -> c_int
+where
+ T: VTab<'vtab>,
+{
+ let vt = vtab.cast::<T>();
+ let mut idx_info = IndexInfo(info);
+ match (*vt).best_index(&mut idx_info) {
+ Ok(_) => ffi::SQLITE_OK,
+ Err(Error::SqliteFailure(err, s)) => {
+ if let Some(err_msg) = s {
+ set_err_msg(vtab, &err_msg);
+ }
+ err.extended_code
+ }
+ Err(err) => {
+ set_err_msg(vtab, &err.to_string());
+ ffi::SQLITE_ERROR
+ }
+ }
+}
+
+unsafe extern "C" fn rust_disconnect<'vtab, T>(vtab: *mut ffi::sqlite3_vtab) -> c_int
+where
+ T: VTab<'vtab>,
+{
+ if vtab.is_null() {
+ return ffi::SQLITE_OK;
+ }
+ let vtab = vtab.cast::<T>();
+ drop(Box::from_raw(vtab));
+ ffi::SQLITE_OK
+}
+
+unsafe extern "C" fn rust_destroy<'vtab, T>(vtab: *mut ffi::sqlite3_vtab) -> c_int
+where
+ T: CreateVTab<'vtab>,
+{
+ if vtab.is_null() {
+ return ffi::SQLITE_OK;
+ }
+ let vt = vtab.cast::<T>();
+ match (*vt).destroy() {
+ Ok(_) => {
+ drop(Box::from_raw(vt));
+ ffi::SQLITE_OK
+ }
+ Err(Error::SqliteFailure(err, s)) => {
+ if let Some(err_msg) = s {
+ set_err_msg(vtab, &err_msg);
+ }
+ err.extended_code
+ }
+ Err(err) => {
+ set_err_msg(vtab, &err.to_string());
+ ffi::SQLITE_ERROR
+ }
+ }
+}
+
+unsafe extern "C" fn rust_open<'vtab, T: 'vtab>(
+ vtab: *mut ffi::sqlite3_vtab,
+ pp_cursor: *mut *mut ffi::sqlite3_vtab_cursor,
+) -> c_int
+where
+ T: VTab<'vtab>,
+{
+ let vt = vtab.cast::<T>();
+ match (*vt).open() {
+ Ok(cursor) => {
+ let boxed_cursor: *mut T::Cursor = Box::into_raw(Box::new(cursor));
+ *pp_cursor = boxed_cursor.cast::<ffi::sqlite3_vtab_cursor>();
+ ffi::SQLITE_OK
+ }
+ Err(Error::SqliteFailure(err, s)) => {
+ if let Some(err_msg) = s {
+ set_err_msg(vtab, &err_msg);
+ }
+ err.extended_code
+ }
+ Err(err) => {
+ set_err_msg(vtab, &err.to_string());
+ ffi::SQLITE_ERROR
+ }
+ }
+}
+
+unsafe extern "C" fn rust_close<C>(cursor: *mut ffi::sqlite3_vtab_cursor) -> c_int
+where
+ C: VTabCursor,
+{
+ let cr = cursor.cast::<C>();
+ drop(Box::from_raw(cr));
+ ffi::SQLITE_OK
+}
+
+unsafe extern "C" fn rust_filter<C>(
+ cursor: *mut ffi::sqlite3_vtab_cursor,
+ idx_num: c_int,
+ idx_str: *const c_char,
+ argc: c_int,
+ argv: *mut *mut ffi::sqlite3_value,
+) -> c_int
+where
+ C: VTabCursor,
+{
+ use std::ffi::CStr;
+ use std::str;
+ let idx_name = if idx_str.is_null() {
+ None
+ } else {
+ let c_slice = CStr::from_ptr(idx_str).to_bytes();
+ Some(str::from_utf8_unchecked(c_slice))
+ };
+ let args = slice::from_raw_parts_mut(argv, argc as usize);
+ let values = Values { args };
+ let cr = cursor as *mut C;
+ cursor_error(cursor, (*cr).filter(idx_num, idx_name, &values))
+}
+
+unsafe extern "C" fn rust_next<C>(cursor: *mut ffi::sqlite3_vtab_cursor) -> c_int
+where
+ C: VTabCursor,
+{
+ let cr = cursor as *mut C;
+ cursor_error(cursor, (*cr).next())
+}
+
+unsafe extern "C" fn rust_eof<C>(cursor: *mut ffi::sqlite3_vtab_cursor) -> c_int
+where
+ C: VTabCursor,
+{
+ let cr = cursor.cast::<C>();
+ (*cr).eof() as c_int
+}
+
+unsafe extern "C" fn rust_column<C>(
+ cursor: *mut ffi::sqlite3_vtab_cursor,
+ ctx: *mut ffi::sqlite3_context,
+ i: c_int,
+) -> c_int
+where
+ C: VTabCursor,
+{
+ let cr = cursor.cast::<C>();
+ let mut ctxt = Context(ctx);
+ result_error(ctx, (*cr).column(&mut ctxt, i))
+}
+
+unsafe extern "C" fn rust_rowid<C>(
+ cursor: *mut ffi::sqlite3_vtab_cursor,
+ p_rowid: *mut ffi::sqlite3_int64,
+) -> c_int
+where
+ C: VTabCursor,
+{
+ let cr = cursor.cast::<C>();
+ match (*cr).rowid() {
+ Ok(rowid) => {
+ *p_rowid = rowid;
+ ffi::SQLITE_OK
+ }
+ err => cursor_error(cursor, err),
+ }
+}
+
+unsafe extern "C" fn rust_update<'vtab, T: 'vtab>(
+ vtab: *mut ffi::sqlite3_vtab,
+ argc: c_int,
+ argv: *mut *mut ffi::sqlite3_value,
+ p_rowid: *mut ffi::sqlite3_int64,
+) -> c_int
+where
+ T: UpdateVTab<'vtab>,
+{
+ assert!(argc >= 1);
+ let args = slice::from_raw_parts_mut(argv, argc as usize);
+ let vt = vtab.cast::<T>();
+ let r = if args.len() == 1 {
+ (*vt).delete(ValueRef::from_value(args[0]))
+ } else if ffi::sqlite3_value_type(args[0]) == ffi::SQLITE_NULL {
+ // TODO Make the distinction between argv[1] == NULL and argv[1] != NULL ?
+ let values = Values { args };
+ match (*vt).insert(&values) {
+ Ok(rowid) => {
+ *p_rowid = rowid;
+ Ok(())
+ }
+ Err(e) => Err(e),
+ }
+ } else {
+ let values = Values { args };
+ (*vt).update(&values)
+ };
+ match r {
+ Ok(_) => ffi::SQLITE_OK,
+ Err(Error::SqliteFailure(err, s)) => {
+ if let Some(err_msg) = s {
+ set_err_msg(vtab, &err_msg);
+ }
+ err.extended_code
+ }
+ Err(err) => {
+ set_err_msg(vtab, &err.to_string());
+ ffi::SQLITE_ERROR
+ }
+ }
+}
+
+/// Virtual table cursors can set an error message by assigning a string to
+/// `zErrMsg`.
+#[cold]
+unsafe fn cursor_error<T>(cursor: *mut ffi::sqlite3_vtab_cursor, result: Result<T>) -> c_int {
+ match result {
+ Ok(_) => ffi::SQLITE_OK,
+ Err(Error::SqliteFailure(err, s)) => {
+ if let Some(err_msg) = s {
+ set_err_msg((*cursor).pVtab, &err_msg);
+ }
+ err.extended_code
+ }
+ Err(err) => {
+ set_err_msg((*cursor).pVtab, &err.to_string());
+ ffi::SQLITE_ERROR
+ }
+ }
+}
+
+/// Virtual tables methods can set an error message by assigning a string to
+/// `zErrMsg`.
+#[cold]
+unsafe fn set_err_msg(vtab: *mut ffi::sqlite3_vtab, err_msg: &str) {
+ if !(*vtab).zErrMsg.is_null() {
+ ffi::sqlite3_free((*vtab).zErrMsg.cast::<c_void>());
+ }
+ (*vtab).zErrMsg = alloc(err_msg);
+}
+
+/// To raise an error, the `column` method should use this method to set the
+/// error message and return the error code.
+#[cold]
+unsafe fn result_error<T>(ctx: *mut ffi::sqlite3_context, result: Result<T>) -> c_int {
+ match result {
+ Ok(_) => ffi::SQLITE_OK,
+ Err(Error::SqliteFailure(err, s)) => {
+ match err.extended_code {
+ ffi::SQLITE_TOOBIG => {
+ ffi::sqlite3_result_error_toobig(ctx);
+ }
+ ffi::SQLITE_NOMEM => {
+ ffi::sqlite3_result_error_nomem(ctx);
+ }
+ code => {
+ ffi::sqlite3_result_error_code(ctx, code);
+ if let Some(Ok(cstr)) = s.map(|s| str_to_cstring(&s)) {
+ ffi::sqlite3_result_error(ctx, cstr.as_ptr(), -1);
+ }
+ }
+ };
+ err.extended_code
+ }
+ Err(err) => {
+ ffi::sqlite3_result_error_code(ctx, ffi::SQLITE_ERROR);
+ if let Ok(cstr) = str_to_cstring(&err.to_string()) {
+ ffi::sqlite3_result_error(ctx, cstr.as_ptr(), -1);
+ }
+ ffi::SQLITE_ERROR
+ }
+ }
+}
+
+#[cfg(feature = "array")]
+#[cfg_attr(docsrs, doc(cfg(feature = "array")))]
+pub mod array;
+#[cfg(feature = "csvtab")]
+#[cfg_attr(docsrs, doc(cfg(feature = "csvtab")))]
+pub mod csvtab;
+#[cfg(feature = "series")]
+#[cfg_attr(docsrs, doc(cfg(feature = "series")))]
+pub mod series; // SQLite >= 3.9.0
+#[cfg(all(test, feature = "modern_sqlite"))]
+mod vtablog;
+
+#[cfg(test)]
+mod test {
+ #[test]
+ fn test_dequote() {
+ assert_eq!("", super::dequote(""));
+ assert_eq!("'", super::dequote("'"));
+ assert_eq!("\"", super::dequote("\""));
+ assert_eq!("'\"", super::dequote("'\""));
+ assert_eq!("", super::dequote("''"));
+ assert_eq!("", super::dequote("\"\""));
+ assert_eq!("x", super::dequote("'x'"));
+ assert_eq!("x", super::dequote("\"x\""));
+ assert_eq!("x", super::dequote("x"));
+ }
+ #[test]
+ fn test_parse_boolean() {
+ assert_eq!(None, super::parse_boolean(""));
+ assert_eq!(Some(true), super::parse_boolean("1"));
+ assert_eq!(Some(true), super::parse_boolean("yes"));
+ assert_eq!(Some(true), super::parse_boolean("on"));
+ assert_eq!(Some(true), super::parse_boolean("true"));
+ assert_eq!(Some(false), super::parse_boolean("0"));
+ assert_eq!(Some(false), super::parse_boolean("no"));
+ assert_eq!(Some(false), super::parse_boolean("off"));
+ assert_eq!(Some(false), super::parse_boolean("false"));
+ }
+}
diff --git a/third_party/rust/rusqlite/src/vtab/series.rs b/third_party/rust/rusqlite/src/vtab/series.rs
new file mode 100644
index 0000000000..5b67758149
--- /dev/null
+++ b/third_party/rust/rusqlite/src/vtab/series.rs
@@ -0,0 +1,341 @@
+//! Generate series virtual table.
+//!
+//! Port of C [generate series
+//! "function"](http://www.sqlite.org/cgi/src/finfo?name=ext/misc/series.c):
+//! `https://www.sqlite.org/series.html`
+use std::default::Default;
+use std::marker::PhantomData;
+use std::os::raw::c_int;
+
+use crate::ffi;
+use crate::types::Type;
+use crate::vtab::{
+ eponymous_only_module, Context, IndexConstraintOp, IndexInfo, VTab, VTabConfig, VTabConnection,
+ VTabCursor, Values,
+};
+use crate::{Connection, Error, Result};
+
+/// Register the "generate_series" module.
+pub fn load_module(conn: &Connection) -> Result<()> {
+ let aux: Option<()> = None;
+ conn.create_module("generate_series", eponymous_only_module::<SeriesTab>(), aux)
+}
+
+// Column numbers
+// const SERIES_COLUMN_VALUE : c_int = 0;
+const SERIES_COLUMN_START: c_int = 1;
+const SERIES_COLUMN_STOP: c_int = 2;
+const SERIES_COLUMN_STEP: c_int = 3;
+
+bitflags::bitflags! {
+ #[derive(Clone, Copy)]
+ #[repr(C)]
+ struct QueryPlanFlags: ::std::os::raw::c_int {
+ // start = $value -- constraint exists
+ const START = 1;
+ // stop = $value -- constraint exists
+ const STOP = 2;
+ // step = $value -- constraint exists
+ const STEP = 4;
+ // output in descending order
+ const DESC = 8;
+ // output in ascending order
+ const ASC = 16;
+ // Both start and stop
+ const BOTH = QueryPlanFlags::START.bits() | QueryPlanFlags::STOP.bits();
+ }
+}
+
+/// An instance of the Series virtual table
+#[repr(C)]
+struct SeriesTab {
+ /// Base class. Must be first
+ base: ffi::sqlite3_vtab,
+}
+
+unsafe impl<'vtab> VTab<'vtab> for SeriesTab {
+ type Aux = ();
+ type Cursor = SeriesTabCursor<'vtab>;
+
+ fn connect(
+ db: &mut VTabConnection,
+ _aux: Option<&()>,
+ _args: &[&[u8]],
+ ) -> Result<(String, SeriesTab)> {
+ let vtab = SeriesTab {
+ base: ffi::sqlite3_vtab::default(),
+ };
+ db.config(VTabConfig::Innocuous)?;
+ Ok((
+ "CREATE TABLE x(value,start hidden,stop hidden,step hidden)".to_owned(),
+ vtab,
+ ))
+ }
+
+ fn best_index(&self, info: &mut IndexInfo) -> Result<()> {
+ // The query plan bitmask
+ let mut idx_num: QueryPlanFlags = QueryPlanFlags::empty();
+ // Mask of unusable constraints
+ let mut unusable_mask: QueryPlanFlags = QueryPlanFlags::empty();
+ // Constraints on start, stop, and step
+ let mut a_idx: [Option<usize>; 3] = [None, None, None];
+ for (i, constraint) in info.constraints().enumerate() {
+ if constraint.column() < SERIES_COLUMN_START {
+ continue;
+ }
+ let (i_col, i_mask) = match constraint.column() {
+ SERIES_COLUMN_START => (0, QueryPlanFlags::START),
+ SERIES_COLUMN_STOP => (1, QueryPlanFlags::STOP),
+ SERIES_COLUMN_STEP => (2, QueryPlanFlags::STEP),
+ _ => {
+ unreachable!()
+ }
+ };
+ if !constraint.is_usable() {
+ unusable_mask |= i_mask;
+ } else if constraint.operator() == IndexConstraintOp::SQLITE_INDEX_CONSTRAINT_EQ {
+ idx_num |= i_mask;
+ a_idx[i_col] = Some(i);
+ }
+ }
+ // Number of arguments that SeriesTabCursor::filter expects
+ let mut n_arg = 0;
+ for j in a_idx.iter().flatten() {
+ n_arg += 1;
+ let mut constraint_usage = info.constraint_usage(*j);
+ constraint_usage.set_argv_index(n_arg);
+ constraint_usage.set_omit(true);
+ #[cfg(all(test, feature = "modern_sqlite"))]
+ debug_assert_eq!(Ok("BINARY"), info.collation(*j));
+ }
+ if !(unusable_mask & !idx_num).is_empty() {
+ return Err(Error::SqliteFailure(
+ ffi::Error::new(ffi::SQLITE_CONSTRAINT),
+ None,
+ ));
+ }
+ if idx_num.contains(QueryPlanFlags::BOTH) {
+ // Both start= and stop= boundaries are available.
+ #[allow(clippy::bool_to_int_with_if)]
+ info.set_estimated_cost(f64::from(
+ 2 - if idx_num.contains(QueryPlanFlags::STEP) {
+ 1
+ } else {
+ 0
+ },
+ ));
+ info.set_estimated_rows(1000);
+ let order_by_consumed = {
+ let mut order_bys = info.order_bys();
+ if let Some(order_by) = order_bys.next() {
+ if order_by.column() == 0 {
+ if order_by.is_order_by_desc() {
+ idx_num |= QueryPlanFlags::DESC;
+ } else {
+ idx_num |= QueryPlanFlags::ASC;
+ }
+ true
+ } else {
+ false
+ }
+ } else {
+ false
+ }
+ };
+ if order_by_consumed {
+ info.set_order_by_consumed(true);
+ }
+ } else {
+ // If either boundary is missing, we have to generate a huge span
+ // of numbers. Make this case very expensive so that the query
+ // planner will work hard to avoid it.
+ info.set_estimated_rows(2_147_483_647);
+ }
+ info.set_idx_num(idx_num.bits());
+ Ok(())
+ }
+
+ fn open(&mut self) -> Result<SeriesTabCursor<'_>> {
+ Ok(SeriesTabCursor::new())
+ }
+}
+
+/// A cursor for the Series virtual table
+#[repr(C)]
+struct SeriesTabCursor<'vtab> {
+ /// Base class. Must be first
+ base: ffi::sqlite3_vtab_cursor,
+ /// True to count down rather than up
+ is_desc: bool,
+ /// The rowid
+ row_id: i64,
+ /// Current value ("value")
+ value: i64,
+ /// Minimum value ("start")
+ min_value: i64,
+ /// Maximum value ("stop")
+ max_value: i64,
+ /// Increment ("step")
+ step: i64,
+ phantom: PhantomData<&'vtab SeriesTab>,
+}
+
+impl SeriesTabCursor<'_> {
+ fn new<'vtab>() -> SeriesTabCursor<'vtab> {
+ SeriesTabCursor {
+ base: ffi::sqlite3_vtab_cursor::default(),
+ is_desc: false,
+ row_id: 0,
+ value: 0,
+ min_value: 0,
+ max_value: 0,
+ step: 0,
+ phantom: PhantomData,
+ }
+ }
+}
+#[allow(clippy::comparison_chain)]
+unsafe impl VTabCursor for SeriesTabCursor<'_> {
+ fn filter(&mut self, idx_num: c_int, _idx_str: Option<&str>, args: &Values<'_>) -> Result<()> {
+ let mut idx_num = QueryPlanFlags::from_bits_truncate(idx_num);
+ let mut i = 0;
+ if idx_num.contains(QueryPlanFlags::START) {
+ self.min_value = args.get::<Option<_>>(i)?.unwrap_or_default();
+ i += 1;
+ } else {
+ self.min_value = 0;
+ }
+ if idx_num.contains(QueryPlanFlags::STOP) {
+ self.max_value = args.get::<Option<_>>(i)?.unwrap_or_default();
+ i += 1;
+ } else {
+ self.max_value = 0xffff_ffff;
+ }
+ if idx_num.contains(QueryPlanFlags::STEP) {
+ self.step = args.get::<Option<_>>(i)?.unwrap_or_default();
+ if self.step == 0 {
+ self.step = 1;
+ } else if self.step < 0 {
+ self.step = -self.step;
+ if !idx_num.contains(QueryPlanFlags::ASC) {
+ idx_num |= QueryPlanFlags::DESC;
+ }
+ }
+ } else {
+ self.step = 1;
+ };
+ for arg in args.iter() {
+ if arg.data_type() == Type::Null {
+ // If any of the constraints have a NULL value, then return no rows.
+ self.min_value = 1;
+ self.max_value = 0;
+ break;
+ }
+ }
+ self.is_desc = idx_num.contains(QueryPlanFlags::DESC);
+ if self.is_desc {
+ self.value = self.max_value;
+ if self.step > 0 {
+ self.value -= (self.max_value - self.min_value) % self.step;
+ }
+ } else {
+ self.value = self.min_value;
+ }
+ self.row_id = 1;
+ Ok(())
+ }
+
+ fn next(&mut self) -> Result<()> {
+ if self.is_desc {
+ self.value -= self.step;
+ } else {
+ self.value += self.step;
+ }
+ self.row_id += 1;
+ Ok(())
+ }
+
+ fn eof(&self) -> bool {
+ if self.is_desc {
+ self.value < self.min_value
+ } else {
+ self.value > self.max_value
+ }
+ }
+
+ fn column(&self, ctx: &mut Context, i: c_int) -> Result<()> {
+ let x = match i {
+ SERIES_COLUMN_START => self.min_value,
+ SERIES_COLUMN_STOP => self.max_value,
+ SERIES_COLUMN_STEP => self.step,
+ _ => self.value,
+ };
+ ctx.set_result(&x)
+ }
+
+ fn rowid(&self) -> Result<i64> {
+ Ok(self.row_id)
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use crate::ffi;
+ use crate::vtab::series;
+ use crate::{Connection, Result};
+ use fallible_iterator::FallibleIterator;
+
+ #[test]
+ fn test_series_module() -> Result<()> {
+ let version = unsafe { ffi::sqlite3_libversion_number() };
+ if version < 3_008_012 {
+ return Ok(());
+ }
+
+ let db = Connection::open_in_memory()?;
+ series::load_module(&db)?;
+
+ let mut s = db.prepare("SELECT * FROM generate_series(0,20,5)")?;
+
+ let series = s.query_map([], |row| row.get::<_, i32>(0))?;
+
+ let mut expected = 0;
+ for value in series {
+ assert_eq!(expected, value?);
+ expected += 5;
+ }
+
+ let mut s =
+ db.prepare("SELECT * FROM generate_series WHERE start=1 AND stop=9 AND step=2")?;
+ let series: Vec<i32> = s.query([])?.map(|r| r.get(0)).collect()?;
+ assert_eq!(vec![1, 3, 5, 7, 9], series);
+ let mut s = db.prepare("SELECT * FROM generate_series LIMIT 5")?;
+ let series: Vec<i32> = s.query([])?.map(|r| r.get(0)).collect()?;
+ assert_eq!(vec![0, 1, 2, 3, 4], series);
+ let mut s = db.prepare("SELECT * FROM generate_series(0,32,5) ORDER BY value DESC")?;
+ let series: Vec<i32> = s.query([])?.map(|r| r.get(0)).collect()?;
+ assert_eq!(vec![30, 25, 20, 15, 10, 5, 0], series);
+
+ let mut s = db.prepare("SELECT * FROM generate_series(NULL)")?;
+ let series: Vec<i32> = s.query([])?.map(|r| r.get(0)).collect()?;
+ let empty = Vec::<i32>::new();
+ assert_eq!(empty, series);
+ let mut s = db.prepare("SELECT * FROM generate_series(5,NULL)")?;
+ let series: Vec<i32> = s.query([])?.map(|r| r.get(0)).collect()?;
+ assert_eq!(empty, series);
+ let mut s = db.prepare("SELECT * FROM generate_series(5,10,NULL)")?;
+ let series: Vec<i32> = s.query([])?.map(|r| r.get(0)).collect()?;
+ assert_eq!(empty, series);
+ let mut s = db.prepare("SELECT * FROM generate_series(NULL,10,2)")?;
+ let series: Vec<i32> = s.query([])?.map(|r| r.get(0)).collect()?;
+ assert_eq!(empty, series);
+ let mut s = db.prepare("SELECT * FROM generate_series(5,NULL,2)")?;
+ let series: Vec<i32> = s.query([])?.map(|r| r.get(0)).collect()?;
+ assert_eq!(empty, series);
+ let mut s = db.prepare("SELECT * FROM generate_series(NULL) ORDER BY value DESC")?;
+ let series: Vec<i32> = s.query([])?.map(|r| r.get(0)).collect()?;
+ assert_eq!(empty, series);
+
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/src/vtab/vtablog.rs b/third_party/rust/rusqlite/src/vtab/vtablog.rs
new file mode 100644
index 0000000000..e289cfae71
--- /dev/null
+++ b/third_party/rust/rusqlite/src/vtab/vtablog.rs
@@ -0,0 +1,297 @@
+//! Port of C [vtablog](http://www.sqlite.org/cgi/src/finfo?name=ext/misc/vtablog.c)
+use std::default::Default;
+use std::marker::PhantomData;
+use std::os::raw::c_int;
+use std::str::FromStr;
+use std::sync::atomic::{AtomicUsize, Ordering};
+
+use crate::vtab::{
+ update_module, Context, CreateVTab, IndexInfo, UpdateVTab, VTab, VTabConnection, VTabCursor,
+ VTabKind, Values,
+};
+use crate::{ffi, ValueRef};
+use crate::{Connection, Error, Result};
+
+/// Register the "vtablog" module.
+pub fn load_module(conn: &Connection) -> Result<()> {
+ let aux: Option<()> = None;
+ conn.create_module("vtablog", update_module::<VTabLog>(), aux)
+}
+
+/// An instance of the vtablog virtual table
+#[repr(C)]
+struct VTabLog {
+ /// Base class. Must be first
+ base: ffi::sqlite3_vtab,
+ /// Number of rows in the table
+ n_row: i64,
+ /// Instance number for this vtablog table
+ i_inst: usize,
+ /// Number of cursors created
+ n_cursor: usize,
+}
+
+impl VTabLog {
+ fn connect_create(
+ _: &mut VTabConnection,
+ _: Option<&()>,
+ args: &[&[u8]],
+ is_create: bool,
+ ) -> Result<(String, VTabLog)> {
+ static N_INST: AtomicUsize = AtomicUsize::new(1);
+ let i_inst = N_INST.fetch_add(1, Ordering::SeqCst);
+ println!(
+ "VTabLog::{}(tab={}, args={:?}):",
+ if is_create { "create" } else { "connect" },
+ i_inst,
+ args,
+ );
+ let mut schema = None;
+ let mut n_row = None;
+
+ let args = &args[3..];
+ for c_slice in args {
+ let (param, value) = super::parameter(c_slice)?;
+ match param {
+ "schema" => {
+ if schema.is_some() {
+ return Err(Error::ModuleError(format!(
+ "more than one '{param}' parameter"
+ )));
+ }
+ schema = Some(value.to_owned())
+ }
+ "rows" => {
+ if n_row.is_some() {
+ return Err(Error::ModuleError(format!(
+ "more than one '{param}' parameter"
+ )));
+ }
+ if let Ok(n) = i64::from_str(value) {
+ n_row = Some(n)
+ }
+ }
+ _ => {
+ return Err(Error::ModuleError(format!(
+ "unrecognized parameter '{param}'"
+ )));
+ }
+ }
+ }
+ if schema.is_none() {
+ return Err(Error::ModuleError("no schema defined".to_owned()));
+ }
+ let vtab = VTabLog {
+ base: ffi::sqlite3_vtab::default(),
+ n_row: n_row.unwrap_or(10),
+ i_inst,
+ n_cursor: 0,
+ };
+ Ok((schema.unwrap(), vtab))
+ }
+}
+
+impl Drop for VTabLog {
+ fn drop(&mut self) {
+ println!("VTabLog::drop({})", self.i_inst);
+ }
+}
+
+unsafe impl<'vtab> VTab<'vtab> for VTabLog {
+ type Aux = ();
+ type Cursor = VTabLogCursor<'vtab>;
+
+ fn connect(
+ db: &mut VTabConnection,
+ aux: Option<&Self::Aux>,
+ args: &[&[u8]],
+ ) -> Result<(String, Self)> {
+ VTabLog::connect_create(db, aux, args, false)
+ }
+
+ fn best_index(&self, info: &mut IndexInfo) -> Result<()> {
+ println!("VTabLog::best_index({})", self.i_inst);
+ info.set_estimated_cost(500.);
+ info.set_estimated_rows(500);
+ Ok(())
+ }
+
+ fn open(&'vtab mut self) -> Result<Self::Cursor> {
+ self.n_cursor += 1;
+ println!(
+ "VTabLog::open(tab={}, cursor={})",
+ self.i_inst, self.n_cursor
+ );
+ Ok(VTabLogCursor {
+ base: ffi::sqlite3_vtab_cursor::default(),
+ i_cursor: self.n_cursor,
+ row_id: 0,
+ phantom: PhantomData,
+ })
+ }
+}
+
+impl<'vtab> CreateVTab<'vtab> for VTabLog {
+ const KIND: VTabKind = VTabKind::Default;
+
+ fn create(
+ db: &mut VTabConnection,
+ aux: Option<&Self::Aux>,
+ args: &[&[u8]],
+ ) -> Result<(String, Self)> {
+ VTabLog::connect_create(db, aux, args, true)
+ }
+
+ fn destroy(&self) -> Result<()> {
+ println!("VTabLog::destroy({})", self.i_inst);
+ Ok(())
+ }
+}
+
+impl<'vtab> UpdateVTab<'vtab> for VTabLog {
+ fn delete(&mut self, arg: ValueRef<'_>) -> Result<()> {
+ println!("VTabLog::delete({}, {arg:?})", self.i_inst);
+ Ok(())
+ }
+
+ fn insert(&mut self, args: &Values<'_>) -> Result<i64> {
+ println!(
+ "VTabLog::insert({}, {:?})",
+ self.i_inst,
+ args.iter().collect::<Vec<ValueRef<'_>>>()
+ );
+ Ok(self.n_row)
+ }
+
+ fn update(&mut self, args: &Values<'_>) -> Result<()> {
+ println!(
+ "VTabLog::update({}, {:?})",
+ self.i_inst,
+ args.iter().collect::<Vec<ValueRef<'_>>>()
+ );
+ Ok(())
+ }
+}
+
+/// A cursor for the Series virtual table
+#[repr(C)]
+struct VTabLogCursor<'vtab> {
+ /// Base class. Must be first
+ base: ffi::sqlite3_vtab_cursor,
+ /// Cursor number
+ i_cursor: usize,
+ /// The rowid
+ row_id: i64,
+ phantom: PhantomData<&'vtab VTabLog>,
+}
+
+impl VTabLogCursor<'_> {
+ fn vtab(&self) -> &VTabLog {
+ unsafe { &*(self.base.pVtab as *const VTabLog) }
+ }
+}
+
+impl Drop for VTabLogCursor<'_> {
+ fn drop(&mut self) {
+ println!(
+ "VTabLogCursor::drop(tab={}, cursor={})",
+ self.vtab().i_inst,
+ self.i_cursor
+ );
+ }
+}
+
+unsafe impl VTabCursor for VTabLogCursor<'_> {
+ fn filter(&mut self, _: c_int, _: Option<&str>, _: &Values<'_>) -> Result<()> {
+ println!(
+ "VTabLogCursor::filter(tab={}, cursor={})",
+ self.vtab().i_inst,
+ self.i_cursor
+ );
+ self.row_id = 0;
+ Ok(())
+ }
+
+ fn next(&mut self) -> Result<()> {
+ println!(
+ "VTabLogCursor::next(tab={}, cursor={}): rowid {} -> {}",
+ self.vtab().i_inst,
+ self.i_cursor,
+ self.row_id,
+ self.row_id + 1
+ );
+ self.row_id += 1;
+ Ok(())
+ }
+
+ fn eof(&self) -> bool {
+ let eof = self.row_id >= self.vtab().n_row;
+ println!(
+ "VTabLogCursor::eof(tab={}, cursor={}): {}",
+ self.vtab().i_inst,
+ self.i_cursor,
+ eof,
+ );
+ eof
+ }
+
+ fn column(&self, ctx: &mut Context, i: c_int) -> Result<()> {
+ let value = if i < 26 {
+ format!(
+ "{}{}",
+ "abcdefghijklmnopqrstuvwyz".chars().nth(i as usize).unwrap(),
+ self.row_id
+ )
+ } else {
+ format!("{i}{}", self.row_id)
+ };
+ println!(
+ "VTabLogCursor::column(tab={}, cursor={}, i={}): {}",
+ self.vtab().i_inst,
+ self.i_cursor,
+ i,
+ value,
+ );
+ ctx.set_result(&value)
+ }
+
+ fn rowid(&self) -> Result<i64> {
+ println!(
+ "VTabLogCursor::rowid(tab={}, cursor={}): {}",
+ self.vtab().i_inst,
+ self.i_cursor,
+ self.row_id,
+ );
+ Ok(self.row_id)
+ }
+}
+
+#[cfg(test)]
+mod test {
+ use crate::{Connection, Result};
+ #[test]
+ fn test_module() -> Result<()> {
+ let db = Connection::open_in_memory()?;
+ super::load_module(&db)?;
+
+ db.execute_batch(
+ "CREATE VIRTUAL TABLE temp.log USING vtablog(
+ schema='CREATE TABLE x(a,b,c)',
+ rows=25
+ );",
+ )?;
+ let mut stmt = db.prepare("SELECT * FROM log;")?;
+ let mut rows = stmt.query([])?;
+ while rows.next()?.is_some() {}
+ db.execute("DELETE FROM log WHERE a = ?1", ["a1"])?;
+ db.execute(
+ "INSERT INTO log (a, b, c) VALUES (?1, ?2, ?3)",
+ ["a", "b", "c"],
+ )?;
+ db.execute(
+ "UPDATE log SET b = ?1, c = ?2 WHERE a = ?3",
+ ["bn", "cn", "a1"],
+ )?;
+ Ok(())
+ }
+}
diff --git a/third_party/rust/rusqlite/test.csv b/third_party/rust/rusqlite/test.csv
new file mode 100644
index 0000000000..708f93f195
--- /dev/null
+++ b/third_party/rust/rusqlite/test.csv
@@ -0,0 +1,6 @@
+"colA","colB","colC"
+1,2,3
+a,b,c
+a,"b",c
+"a","b","c .. z"
+"a","b","c,d"
diff --git a/third_party/rust/rusqlite/tests/config_log.rs b/third_party/rust/rusqlite/tests/config_log.rs
new file mode 100644
index 0000000000..0c28bdf1df
--- /dev/null
+++ b/third_party/rust/rusqlite/tests/config_log.rs
@@ -0,0 +1,34 @@
+//! This file contains unit tests for `rusqlite::trace::config_log`. This
+//! function affects SQLite process-wide and so is not safe to run as a normal
+//! #[test] in the library.
+
+#[cfg(feature = "trace")]
+fn main() {
+ use lazy_static::lazy_static;
+ use std::os::raw::c_int;
+ use std::sync::Mutex;
+
+ lazy_static! {
+ static ref LOGS_RECEIVED: Mutex<Vec<(c_int, String)>> = Mutex::new(Vec::new());
+ }
+
+ fn log_handler(err: c_int, message: &str) {
+ let mut logs_received = LOGS_RECEIVED.lock().unwrap();
+ logs_received.push((err, message.to_owned()));
+ }
+
+ use rusqlite::trace;
+
+ unsafe { trace::config_log(Some(log_handler)) }.unwrap();
+ trace::log(10, "First message from rusqlite");
+ unsafe { trace::config_log(None) }.unwrap();
+ trace::log(11, "Second message from rusqlite");
+
+ let logs_received = LOGS_RECEIVED.lock().unwrap();
+ assert_eq!(logs_received.len(), 1);
+ assert_eq!(logs_received[0].0, 10);
+ assert_eq!(logs_received[0].1, "First message from rusqlite");
+}
+
+#[cfg(not(feature = "trace"))]
+fn main() {}
diff --git a/third_party/rust/rusqlite/tests/deny_single_threaded_sqlite_config.rs b/third_party/rust/rusqlite/tests/deny_single_threaded_sqlite_config.rs
new file mode 100644
index 0000000000..aeb7eec317
--- /dev/null
+++ b/third_party/rust/rusqlite/tests/deny_single_threaded_sqlite_config.rs
@@ -0,0 +1,20 @@
+//! Ensure we reject connections when SQLite is in single-threaded mode, as it
+//! would violate safety if multiple Rust threads tried to use connections.
+
+#[cfg(not(feature = "loadable_extension"))]
+#[test]
+fn test_error_when_singlethread_mode() {
+ use rusqlite::ffi;
+ use rusqlite::Connection;
+ // put SQLite into single-threaded mode
+ unsafe {
+ // Note: macOS system SQLite seems to return an error if you attempt to
+ // reconfigure to single-threaded mode.
+ if ffi::sqlite3_config(ffi::SQLITE_CONFIG_SINGLETHREAD) != ffi::SQLITE_OK {
+ return;
+ }
+ assert_eq!(ffi::sqlite3_initialize(), ffi::SQLITE_OK);
+ }
+ let res = Connection::open_in_memory();
+ res.unwrap_err();
+}
diff --git a/third_party/rust/rusqlite/tests/vtab.rs b/third_party/rust/rusqlite/tests/vtab.rs
new file mode 100644
index 0000000000..65582064c9
--- /dev/null
+++ b/third_party/rust/rusqlite/tests/vtab.rs
@@ -0,0 +1,100 @@
+//! Ensure Virtual tables can be declared outside `rusqlite` crate.
+
+#[cfg(feature = "vtab")]
+#[test]
+fn test_dummy_module() -> rusqlite::Result<()> {
+ use rusqlite::vtab::{
+ eponymous_only_module, sqlite3_vtab, sqlite3_vtab_cursor, Context, IndexInfo, VTab,
+ VTabConnection, VTabCursor, Values,
+ };
+ use rusqlite::{version_number, Connection, Result};
+ use std::marker::PhantomData;
+ use std::os::raw::c_int;
+
+ let module = eponymous_only_module::<DummyTab>();
+
+ #[repr(C)]
+ struct DummyTab {
+ /// Base class. Must be first
+ base: sqlite3_vtab,
+ }
+
+ unsafe impl<'vtab> VTab<'vtab> for DummyTab {
+ type Aux = ();
+ type Cursor = DummyTabCursor<'vtab>;
+
+ fn connect(
+ _: &mut VTabConnection,
+ _aux: Option<&()>,
+ _args: &[&[u8]],
+ ) -> Result<(String, DummyTab)> {
+ let vtab = DummyTab {
+ base: sqlite3_vtab::default(),
+ };
+ Ok(("CREATE TABLE x(value)".to_owned(), vtab))
+ }
+
+ fn best_index(&self, info: &mut IndexInfo) -> Result<()> {
+ info.set_estimated_cost(1.);
+ Ok(())
+ }
+
+ fn open(&'vtab mut self) -> Result<DummyTabCursor<'vtab>> {
+ Ok(DummyTabCursor::default())
+ }
+ }
+
+ #[derive(Default)]
+ #[repr(C)]
+ struct DummyTabCursor<'vtab> {
+ /// Base class. Must be first
+ base: sqlite3_vtab_cursor,
+ /// The rowid
+ row_id: i64,
+ phantom: PhantomData<&'vtab DummyTab>,
+ }
+
+ unsafe impl VTabCursor for DummyTabCursor<'_> {
+ fn filter(
+ &mut self,
+ _idx_num: c_int,
+ _idx_str: Option<&str>,
+ _args: &Values<'_>,
+ ) -> Result<()> {
+ self.row_id = 1;
+ Ok(())
+ }
+
+ fn next(&mut self) -> Result<()> {
+ self.row_id += 1;
+ Ok(())
+ }
+
+ fn eof(&self) -> bool {
+ self.row_id > 1
+ }
+
+ fn column(&self, ctx: &mut Context, _: c_int) -> Result<()> {
+ ctx.set_result(&self.row_id)
+ }
+
+ fn rowid(&self) -> Result<i64> {
+ Ok(self.row_id)
+ }
+ }
+
+ let db = Connection::open_in_memory()?;
+
+ db.create_module::<DummyTab>("dummy", module, None)?;
+
+ let version = version_number();
+ if version < 3_009_000 {
+ return Ok(());
+ }
+
+ let mut s = db.prepare("SELECT * FROM dummy()")?;
+
+ let dummy = s.query_row([], |row| row.get::<_, i32>(0))?;
+ assert_eq!(1, dummy);
+ Ok(())
+}