diff options
Diffstat (limited to 'third_party/rust/unicode-bidi/src')
| -rw-r--r-- | third_party/rust/unicode-bidi/src/char_data/mod.rs | 173 | ||||
| -rw-r--r-- | third_party/rust/unicode-bidi/src/char_data/tables.rs | 543 | ||||
| -rw-r--r-- | third_party/rust/unicode-bidi/src/data_source.rs | 46 | ||||
| -rw-r--r-- | third_party/rust/unicode-bidi/src/deprecated.rs | 92 | ||||
| -rw-r--r-- | third_party/rust/unicode-bidi/src/explicit.rs | 210 | ||||
| -rw-r--r-- | third_party/rust/unicode-bidi/src/format_chars.rs | 42 | ||||
| -rw-r--r-- | third_party/rust/unicode-bidi/src/implicit.rs | 585 | ||||
| -rw-r--r-- | third_party/rust/unicode-bidi/src/level.rs | 400 | ||||
| -rw-r--r-- | third_party/rust/unicode-bidi/src/lib.rs | 2242 | ||||
| -rw-r--r-- | third_party/rust/unicode-bidi/src/prepare.rs | 451 | ||||
| -rw-r--r-- | third_party/rust/unicode-bidi/src/utf16.rs | 791 |
11 files changed, 5575 insertions, 0 deletions
diff --git a/third_party/rust/unicode-bidi/src/char_data/mod.rs b/third_party/rust/unicode-bidi/src/char_data/mod.rs new file mode 100644 index 0000000000..4edf5b8f4c --- /dev/null +++ b/third_party/rust/unicode-bidi/src/char_data/mod.rs @@ -0,0 +1,173 @@ +// Copyright 2015 The Servo Project Developers. See the +// COPYRIGHT file at the top-level directory of this distribution. +// +// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or +// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license +// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your +// option. This file may not be copied, modified, or distributed +// except according to those terms. + +//! Accessor for `Bidi_Class` property from Unicode Character Database (UCD) + +mod tables; + +pub use self::tables::{BidiClass, UNICODE_VERSION}; +#[cfg(feature = "hardcoded-data")] +use core::char; +#[cfg(feature = "hardcoded-data")] +use core::cmp::Ordering::{Equal, Greater, Less}; + +#[cfg(feature = "hardcoded-data")] +use self::tables::bidi_class_table; +use crate::data_source::BidiMatchedOpeningBracket; +use crate::BidiClass::*; +#[cfg(feature = "hardcoded-data")] +use crate::BidiDataSource; +/// Hardcoded Bidi data that ships with the unicode-bidi crate. +/// +/// This can be enabled with the default `hardcoded-data` Cargo feature. +#[cfg(feature = "hardcoded-data")] +pub struct HardcodedBidiData; + +#[cfg(feature = "hardcoded-data")] +impl BidiDataSource for HardcodedBidiData { + fn bidi_class(&self, c: char) -> BidiClass { + bsearch_range_value_table(c, bidi_class_table) + } +} + +/// Find the `BidiClass` of a single char. +#[cfg(feature = "hardcoded-data")] +pub fn bidi_class(c: char) -> BidiClass { + bsearch_range_value_table(c, bidi_class_table) +} + +/// If this character is a bracket according to BidiBrackets.txt, +/// return the corresponding *normalized* *opening bracket* of the pair, +/// and whether or not it itself is an opening bracket. +pub(crate) fn bidi_matched_opening_bracket(c: char) -> Option<BidiMatchedOpeningBracket> { + for pair in self::tables::bidi_pairs_table { + if pair.0 == c || pair.1 == c { + let skeleton = pair.2.unwrap_or(pair.0); + return Some(BidiMatchedOpeningBracket { + opening: skeleton, + is_open: pair.0 == c, + }); + } + } + None +} + +pub fn is_rtl(bidi_class: BidiClass) -> bool { + match bidi_class { + RLE | RLO | RLI => true, + _ => false, + } +} + +#[cfg(feature = "hardcoded-data")] +fn bsearch_range_value_table(c: char, r: &'static [(char, char, BidiClass)]) -> BidiClass { + match r.binary_search_by(|&(lo, hi, _)| { + if lo <= c && c <= hi { + Equal + } else if hi < c { + Less + } else { + Greater + } + }) { + Ok(idx) => { + let (_, _, cat) = r[idx]; + cat + } + // UCD/extracted/DerivedBidiClass.txt: "All code points not explicitly listed + // for Bidi_Class have the value Left_To_Right (L)." + Err(_) => L, + } +} + +#[cfg(all(test, feature = "hardcoded-data"))] +mod tests { + use super::*; + + #[test] + fn test_ascii() { + assert_eq!(bidi_class('\u{0000}'), BN); + assert_eq!(bidi_class('\u{0040}'), ON); + assert_eq!(bidi_class('\u{0041}'), L); + assert_eq!(bidi_class('\u{0062}'), L); + assert_eq!(bidi_class('\u{007F}'), BN); + } + + #[test] + fn test_bmp() { + // Hebrew + assert_eq!(bidi_class('\u{0590}'), R); + assert_eq!(bidi_class('\u{05D0}'), R); + assert_eq!(bidi_class('\u{05D1}'), R); + assert_eq!(bidi_class('\u{05FF}'), R); + + // Arabic + assert_eq!(bidi_class('\u{0600}'), AN); + assert_eq!(bidi_class('\u{0627}'), AL); + assert_eq!(bidi_class('\u{07BF}'), AL); + + // Default R + Arabic Extras + assert_eq!(bidi_class('\u{07C0}'), R); + assert_eq!(bidi_class('\u{085F}'), R); + assert_eq!(bidi_class('\u{0860}'), AL); + assert_eq!(bidi_class('\u{0870}'), AL); + assert_eq!(bidi_class('\u{089F}'), NSM); + assert_eq!(bidi_class('\u{08A0}'), AL); + assert_eq!(bidi_class('\u{089F}'), NSM); + assert_eq!(bidi_class('\u{08FF}'), NSM); + + // Default ET + assert_eq!(bidi_class('\u{20A0}'), ET); + assert_eq!(bidi_class('\u{20CF}'), ET); + + // Arabic Presentation Forms + assert_eq!(bidi_class('\u{FB1D}'), R); + assert_eq!(bidi_class('\u{FB4F}'), R); + assert_eq!(bidi_class('\u{FB50}'), AL); + assert_eq!(bidi_class('\u{FDCF}'), ON); + assert_eq!(bidi_class('\u{FDF0}'), AL); + assert_eq!(bidi_class('\u{FDFF}'), ON); + assert_eq!(bidi_class('\u{FE70}'), AL); + assert_eq!(bidi_class('\u{FEFE}'), AL); + assert_eq!(bidi_class('\u{FEFF}'), BN); + + // noncharacters + assert_eq!(bidi_class('\u{FDD0}'), L); + assert_eq!(bidi_class('\u{FDD1}'), L); + assert_eq!(bidi_class('\u{FDEE}'), L); + assert_eq!(bidi_class('\u{FDEF}'), L); + assert_eq!(bidi_class('\u{FFFE}'), L); + assert_eq!(bidi_class('\u{FFFF}'), L); + } + + #[test] + fn test_smp() { + // Default AL + R + assert_eq!(bidi_class('\u{10800}'), R); + assert_eq!(bidi_class('\u{10FFF}'), R); + assert_eq!(bidi_class('\u{1E800}'), R); + assert_eq!(bidi_class('\u{1EDFF}'), R); + assert_eq!(bidi_class('\u{1EE00}'), AL); + assert_eq!(bidi_class('\u{1EEFF}'), AL); + assert_eq!(bidi_class('\u{1EF00}'), R); + assert_eq!(bidi_class('\u{1EFFF}'), R); + } + + #[test] + fn test_unassigned_planes() { + assert_eq!(bidi_class('\u{30000}'), L); + assert_eq!(bidi_class('\u{40000}'), L); + assert_eq!(bidi_class('\u{50000}'), L); + assert_eq!(bidi_class('\u{60000}'), L); + assert_eq!(bidi_class('\u{70000}'), L); + assert_eq!(bidi_class('\u{80000}'), L); + assert_eq!(bidi_class('\u{90000}'), L); + assert_eq!(bidi_class('\u{a0000}'), L); + } +} diff --git a/third_party/rust/unicode-bidi/src/char_data/tables.rs b/third_party/rust/unicode-bidi/src/char_data/tables.rs new file mode 100644 index 0000000000..ecdcf496d1 --- /dev/null +++ b/third_party/rust/unicode-bidi/src/char_data/tables.rs @@ -0,0 +1,543 @@ +// NOTE: +// The following code was generated by "tools/generate.py". do not edit directly + +#![allow(missing_docs, non_upper_case_globals, non_snake_case)] +#![cfg_attr(rustfmt, rustfmt_skip)] + +/// The [Unicode version](http://www.unicode.org/versions/) of data +pub const UNICODE_VERSION: (u64, u64, u64) = (15, 0, 0); + +#[allow(non_camel_case_types)] +#[derive(Clone, Copy, Debug, PartialEq, Eq)] +/// Represents values of the Unicode character property +/// [`Bidi_Class`](http://www.unicode.org/reports/tr44/#Bidi_Class), also +/// known as the *bidirectional character type*. +/// +/// * <http://www.unicode.org/reports/tr9/#Bidirectional_Character_Types> +/// * <http://www.unicode.org/reports/tr44/#Bidi_Class_Values> +pub enum BidiClass { + AL, + AN, + B, + BN, + CS, + EN, + ES, + ET, + FSI, + L, + LRE, + LRI, + LRO, + NSM, + ON, + PDF, + PDI, + R, + RLE, + RLI, + RLO, + S, + WS, +} + +#[cfg(feature = "hardcoded-data")] +use self::BidiClass::*; + +#[cfg(feature = "hardcoded-data")] +pub const bidi_class_table: &'static [(char, char, BidiClass)] = &[ + ('\u{0}', '\u{8}', BN), ('\u{9}', '\u{9}', S), ('\u{a}', '\u{a}', B), ('\u{b}', '\u{b}', S), + ('\u{c}', '\u{c}', WS), ('\u{d}', '\u{d}', B), ('\u{e}', '\u{1b}', BN), ('\u{1c}', '\u{1e}', B), + ('\u{1f}', '\u{1f}', S), ('\u{20}', '\u{20}', WS), ('\u{21}', '\u{22}', ON), ('\u{23}', + '\u{25}', ET), ('\u{26}', '\u{2a}', ON), ('\u{2b}', '\u{2b}', ES), ('\u{2c}', '\u{2c}', CS), + ('\u{2d}', '\u{2d}', ES), ('\u{2e}', '\u{2f}', CS), ('\u{30}', '\u{39}', EN), ('\u{3a}', + '\u{3a}', CS), ('\u{3b}', '\u{40}', ON), ('\u{41}', '\u{5a}', L), ('\u{5b}', '\u{60}', ON), + ('\u{61}', '\u{7a}', L), ('\u{7b}', '\u{7e}', ON), ('\u{7f}', '\u{84}', BN), ('\u{85}', + '\u{85}', B), ('\u{86}', '\u{9f}', BN), ('\u{a0}', '\u{a0}', CS), ('\u{a1}', '\u{a1}', ON), + ('\u{a2}', '\u{a5}', ET), ('\u{a6}', '\u{a9}', ON), ('\u{aa}', '\u{aa}', L), ('\u{ab}', + '\u{ac}', ON), ('\u{ad}', '\u{ad}', BN), ('\u{ae}', '\u{af}', ON), ('\u{b0}', '\u{b1}', ET), + ('\u{b2}', '\u{b3}', EN), ('\u{b4}', '\u{b4}', ON), ('\u{b5}', '\u{b5}', L), ('\u{b6}', + '\u{b8}', ON), ('\u{b9}', '\u{b9}', EN), ('\u{ba}', '\u{ba}', L), ('\u{bb}', '\u{bf}', ON), + ('\u{c0}', '\u{d6}', L), ('\u{d7}', '\u{d7}', ON), ('\u{d8}', '\u{f6}', L), ('\u{f7}', '\u{f7}', + ON), ('\u{f8}', '\u{2b8}', L), ('\u{2b9}', '\u{2ba}', ON), ('\u{2bb}', '\u{2c1}', L), + ('\u{2c2}', '\u{2cf}', ON), ('\u{2d0}', '\u{2d1}', L), ('\u{2d2}', '\u{2df}', ON), ('\u{2e0}', + '\u{2e4}', L), ('\u{2e5}', '\u{2ed}', ON), ('\u{2ee}', '\u{2ee}', L), ('\u{2ef}', '\u{2ff}', + ON), ('\u{300}', '\u{36f}', NSM), ('\u{370}', '\u{373}', L), ('\u{374}', '\u{375}', ON), + ('\u{376}', '\u{377}', L), ('\u{37a}', '\u{37d}', L), ('\u{37e}', '\u{37e}', ON), ('\u{37f}', + '\u{37f}', L), ('\u{384}', '\u{385}', ON), ('\u{386}', '\u{386}', L), ('\u{387}', '\u{387}', + ON), ('\u{388}', '\u{38a}', L), ('\u{38c}', '\u{38c}', L), ('\u{38e}', '\u{3a1}', L), + ('\u{3a3}', '\u{3f5}', L), ('\u{3f6}', '\u{3f6}', ON), ('\u{3f7}', '\u{482}', L), ('\u{483}', + '\u{489}', NSM), ('\u{48a}', '\u{52f}', L), ('\u{531}', '\u{556}', L), ('\u{559}', '\u{589}', + L), ('\u{58a}', '\u{58a}', ON), ('\u{58d}', '\u{58e}', ON), ('\u{58f}', '\u{58f}', ET), + ('\u{590}', '\u{590}', R), ('\u{591}', '\u{5bd}', NSM), ('\u{5be}', '\u{5be}', R), ('\u{5bf}', + '\u{5bf}', NSM), ('\u{5c0}', '\u{5c0}', R), ('\u{5c1}', '\u{5c2}', NSM), ('\u{5c3}', '\u{5c3}', + R), ('\u{5c4}', '\u{5c5}', NSM), ('\u{5c6}', '\u{5c6}', R), ('\u{5c7}', '\u{5c7}', NSM), + ('\u{5c8}', '\u{5ff}', R), ('\u{600}', '\u{605}', AN), ('\u{606}', '\u{607}', ON), ('\u{608}', + '\u{608}', AL), ('\u{609}', '\u{60a}', ET), ('\u{60b}', '\u{60b}', AL), ('\u{60c}', '\u{60c}', + CS), ('\u{60d}', '\u{60d}', AL), ('\u{60e}', '\u{60f}', ON), ('\u{610}', '\u{61a}', NSM), + ('\u{61b}', '\u{64a}', AL), ('\u{64b}', '\u{65f}', NSM), ('\u{660}', '\u{669}', AN), ('\u{66a}', + '\u{66a}', ET), ('\u{66b}', '\u{66c}', AN), ('\u{66d}', '\u{66f}', AL), ('\u{670}', '\u{670}', + NSM), ('\u{671}', '\u{6d5}', AL), ('\u{6d6}', '\u{6dc}', NSM), ('\u{6dd}', '\u{6dd}', AN), + ('\u{6de}', '\u{6de}', ON), ('\u{6df}', '\u{6e4}', NSM), ('\u{6e5}', '\u{6e6}', AL), ('\u{6e7}', + '\u{6e8}', NSM), ('\u{6e9}', '\u{6e9}', ON), ('\u{6ea}', '\u{6ed}', NSM), ('\u{6ee}', '\u{6ef}', + AL), ('\u{6f0}', '\u{6f9}', EN), ('\u{6fa}', '\u{710}', AL), ('\u{711}', '\u{711}', NSM), + ('\u{712}', '\u{72f}', AL), ('\u{730}', '\u{74a}', NSM), ('\u{74b}', '\u{7a5}', AL), ('\u{7a6}', + '\u{7b0}', NSM), ('\u{7b1}', '\u{7bf}', AL), ('\u{7c0}', '\u{7ea}', R), ('\u{7eb}', '\u{7f3}', + NSM), ('\u{7f4}', '\u{7f5}', R), ('\u{7f6}', '\u{7f9}', ON), ('\u{7fa}', '\u{7fc}', R), + ('\u{7fd}', '\u{7fd}', NSM), ('\u{7fe}', '\u{815}', R), ('\u{816}', '\u{819}', NSM), ('\u{81a}', + '\u{81a}', R), ('\u{81b}', '\u{823}', NSM), ('\u{824}', '\u{824}', R), ('\u{825}', '\u{827}', + NSM), ('\u{828}', '\u{828}', R), ('\u{829}', '\u{82d}', NSM), ('\u{82e}', '\u{858}', R), + ('\u{859}', '\u{85b}', NSM), ('\u{85c}', '\u{85f}', R), ('\u{860}', '\u{86a}', AL), ('\u{86b}', + '\u{86f}', R), ('\u{870}', '\u{88e}', AL), ('\u{88f}', '\u{88f}', R), ('\u{890}', '\u{891}', + AN), ('\u{892}', '\u{897}', R), ('\u{898}', '\u{89f}', NSM), ('\u{8a0}', '\u{8c9}', AL), + ('\u{8ca}', '\u{8e1}', NSM), ('\u{8e2}', '\u{8e2}', AN), ('\u{8e3}', '\u{902}', NSM), + ('\u{903}', '\u{939}', L), ('\u{93a}', '\u{93a}', NSM), ('\u{93b}', '\u{93b}', L), ('\u{93c}', + '\u{93c}', NSM), ('\u{93d}', '\u{940}', L), ('\u{941}', '\u{948}', NSM), ('\u{949}', '\u{94c}', + L), ('\u{94d}', '\u{94d}', NSM), ('\u{94e}', '\u{950}', L), ('\u{951}', '\u{957}', NSM), + ('\u{958}', '\u{961}', L), ('\u{962}', '\u{963}', NSM), ('\u{964}', '\u{980}', L), ('\u{981}', + '\u{981}', NSM), ('\u{982}', '\u{983}', L), ('\u{985}', '\u{98c}', L), ('\u{98f}', '\u{990}', + L), ('\u{993}', '\u{9a8}', L), ('\u{9aa}', '\u{9b0}', L), ('\u{9b2}', '\u{9b2}', L), ('\u{9b6}', + '\u{9b9}', L), ('\u{9bc}', '\u{9bc}', NSM), ('\u{9bd}', '\u{9c0}', L), ('\u{9c1}', '\u{9c4}', + NSM), ('\u{9c7}', '\u{9c8}', L), ('\u{9cb}', '\u{9cc}', L), ('\u{9cd}', '\u{9cd}', NSM), + ('\u{9ce}', '\u{9ce}', L), ('\u{9d7}', '\u{9d7}', L), ('\u{9dc}', '\u{9dd}', L), ('\u{9df}', + '\u{9e1}', L), ('\u{9e2}', '\u{9e3}', NSM), ('\u{9e6}', '\u{9f1}', L), ('\u{9f2}', '\u{9f3}', + ET), ('\u{9f4}', '\u{9fa}', L), ('\u{9fb}', '\u{9fb}', ET), ('\u{9fc}', '\u{9fd}', L), + ('\u{9fe}', '\u{9fe}', NSM), ('\u{a01}', '\u{a02}', NSM), ('\u{a03}', '\u{a03}', L), ('\u{a05}', + '\u{a0a}', L), ('\u{a0f}', '\u{a10}', L), ('\u{a13}', '\u{a28}', L), ('\u{a2a}', '\u{a30}', L), + ('\u{a32}', '\u{a33}', L), ('\u{a35}', '\u{a36}', L), ('\u{a38}', '\u{a39}', L), ('\u{a3c}', + '\u{a3c}', NSM), ('\u{a3e}', '\u{a40}', L), ('\u{a41}', '\u{a42}', NSM), ('\u{a47}', '\u{a48}', + NSM), ('\u{a4b}', '\u{a4d}', NSM), ('\u{a51}', '\u{a51}', NSM), ('\u{a59}', '\u{a5c}', L), + ('\u{a5e}', '\u{a5e}', L), ('\u{a66}', '\u{a6f}', L), ('\u{a70}', '\u{a71}', NSM), ('\u{a72}', + '\u{a74}', L), ('\u{a75}', '\u{a75}', NSM), ('\u{a76}', '\u{a76}', L), ('\u{a81}', '\u{a82}', + NSM), ('\u{a83}', '\u{a83}', L), ('\u{a85}', '\u{a8d}', L), ('\u{a8f}', '\u{a91}', L), + ('\u{a93}', '\u{aa8}', L), ('\u{aaa}', '\u{ab0}', L), ('\u{ab2}', '\u{ab3}', L), ('\u{ab5}', + '\u{ab9}', L), ('\u{abc}', '\u{abc}', NSM), ('\u{abd}', '\u{ac0}', L), ('\u{ac1}', '\u{ac5}', + NSM), ('\u{ac7}', '\u{ac8}', NSM), ('\u{ac9}', '\u{ac9}', L), ('\u{acb}', '\u{acc}', L), + ('\u{acd}', '\u{acd}', NSM), ('\u{ad0}', '\u{ad0}', L), ('\u{ae0}', '\u{ae1}', L), ('\u{ae2}', + '\u{ae3}', NSM), ('\u{ae6}', '\u{af0}', L), ('\u{af1}', '\u{af1}', ET), ('\u{af9}', '\u{af9}', + L), ('\u{afa}', '\u{aff}', NSM), ('\u{b01}', '\u{b01}', NSM), ('\u{b02}', '\u{b03}', L), + ('\u{b05}', '\u{b0c}', L), ('\u{b0f}', '\u{b10}', L), ('\u{b13}', '\u{b28}', L), ('\u{b2a}', + '\u{b30}', L), ('\u{b32}', '\u{b33}', L), ('\u{b35}', '\u{b39}', L), ('\u{b3c}', '\u{b3c}', + NSM), ('\u{b3d}', '\u{b3e}', L), ('\u{b3f}', '\u{b3f}', NSM), ('\u{b40}', '\u{b40}', L), + ('\u{b41}', '\u{b44}', NSM), ('\u{b47}', '\u{b48}', L), ('\u{b4b}', '\u{b4c}', L), ('\u{b4d}', + '\u{b4d}', NSM), ('\u{b55}', '\u{b56}', NSM), ('\u{b57}', '\u{b57}', L), ('\u{b5c}', '\u{b5d}', + L), ('\u{b5f}', '\u{b61}', L), ('\u{b62}', '\u{b63}', NSM), ('\u{b66}', '\u{b77}', L), + ('\u{b82}', '\u{b82}', NSM), ('\u{b83}', '\u{b83}', L), ('\u{b85}', '\u{b8a}', L), ('\u{b8e}', + '\u{b90}', L), ('\u{b92}', '\u{b95}', L), ('\u{b99}', '\u{b9a}', L), ('\u{b9c}', '\u{b9c}', L), + ('\u{b9e}', '\u{b9f}', L), ('\u{ba3}', '\u{ba4}', L), ('\u{ba8}', '\u{baa}', L), ('\u{bae}', + '\u{bb9}', L), ('\u{bbe}', '\u{bbf}', L), ('\u{bc0}', '\u{bc0}', NSM), ('\u{bc1}', '\u{bc2}', + L), ('\u{bc6}', '\u{bc8}', L), ('\u{bca}', '\u{bcc}', L), ('\u{bcd}', '\u{bcd}', NSM), + ('\u{bd0}', '\u{bd0}', L), ('\u{bd7}', '\u{bd7}', L), ('\u{be6}', '\u{bf2}', L), ('\u{bf3}', + '\u{bf8}', ON), ('\u{bf9}', '\u{bf9}', ET), ('\u{bfa}', '\u{bfa}', ON), ('\u{c00}', '\u{c00}', + NSM), ('\u{c01}', '\u{c03}', L), ('\u{c04}', '\u{c04}', NSM), ('\u{c05}', '\u{c0c}', L), + ('\u{c0e}', '\u{c10}', L), ('\u{c12}', '\u{c28}', L), ('\u{c2a}', '\u{c39}', L), ('\u{c3c}', + '\u{c3c}', NSM), ('\u{c3d}', '\u{c3d}', L), ('\u{c3e}', '\u{c40}', NSM), ('\u{c41}', '\u{c44}', + L), ('\u{c46}', '\u{c48}', NSM), ('\u{c4a}', '\u{c4d}', NSM), ('\u{c55}', '\u{c56}', NSM), + ('\u{c58}', '\u{c5a}', L), ('\u{c5d}', '\u{c5d}', L), ('\u{c60}', '\u{c61}', L), ('\u{c62}', + '\u{c63}', NSM), ('\u{c66}', '\u{c6f}', L), ('\u{c77}', '\u{c77}', L), ('\u{c78}', '\u{c7e}', + ON), ('\u{c7f}', '\u{c80}', L), ('\u{c81}', '\u{c81}', NSM), ('\u{c82}', '\u{c8c}', L), + ('\u{c8e}', '\u{c90}', L), ('\u{c92}', '\u{ca8}', L), ('\u{caa}', '\u{cb3}', L), ('\u{cb5}', + '\u{cb9}', L), ('\u{cbc}', '\u{cbc}', NSM), ('\u{cbd}', '\u{cc4}', L), ('\u{cc6}', '\u{cc8}', + L), ('\u{cca}', '\u{ccb}', L), ('\u{ccc}', '\u{ccd}', NSM), ('\u{cd5}', '\u{cd6}', L), + ('\u{cdd}', '\u{cde}', L), ('\u{ce0}', '\u{ce1}', L), ('\u{ce2}', '\u{ce3}', NSM), ('\u{ce6}', + '\u{cef}', L), ('\u{cf1}', '\u{cf3}', L), ('\u{d00}', '\u{d01}', NSM), ('\u{d02}', '\u{d0c}', + L), ('\u{d0e}', '\u{d10}', L), ('\u{d12}', '\u{d3a}', L), ('\u{d3b}', '\u{d3c}', NSM), + ('\u{d3d}', '\u{d40}', L), ('\u{d41}', '\u{d44}', NSM), ('\u{d46}', '\u{d48}', L), ('\u{d4a}', + '\u{d4c}', L), ('\u{d4d}', '\u{d4d}', NSM), ('\u{d4e}', '\u{d4f}', L), ('\u{d54}', '\u{d61}', + L), ('\u{d62}', '\u{d63}', NSM), ('\u{d66}', '\u{d7f}', L), ('\u{d81}', '\u{d81}', NSM), + ('\u{d82}', '\u{d83}', L), ('\u{d85}', '\u{d96}', L), ('\u{d9a}', '\u{db1}', L), ('\u{db3}', + '\u{dbb}', L), ('\u{dbd}', '\u{dbd}', L), ('\u{dc0}', '\u{dc6}', L), ('\u{dca}', '\u{dca}', + NSM), ('\u{dcf}', '\u{dd1}', L), ('\u{dd2}', '\u{dd4}', NSM), ('\u{dd6}', '\u{dd6}', NSM), + ('\u{dd8}', '\u{ddf}', L), ('\u{de6}', '\u{def}', L), ('\u{df2}', '\u{df4}', L), ('\u{e01}', + '\u{e30}', L), ('\u{e31}', '\u{e31}', NSM), ('\u{e32}', '\u{e33}', L), ('\u{e34}', '\u{e3a}', + NSM), ('\u{e3f}', '\u{e3f}', ET), ('\u{e40}', '\u{e46}', L), ('\u{e47}', '\u{e4e}', NSM), + ('\u{e4f}', '\u{e5b}', L), ('\u{e81}', '\u{e82}', L), ('\u{e84}', '\u{e84}', L), ('\u{e86}', + '\u{e8a}', L), ('\u{e8c}', '\u{ea3}', L), ('\u{ea5}', '\u{ea5}', L), ('\u{ea7}', '\u{eb0}', L), + ('\u{eb1}', '\u{eb1}', NSM), ('\u{eb2}', '\u{eb3}', L), ('\u{eb4}', '\u{ebc}', NSM), ('\u{ebd}', + '\u{ebd}', L), ('\u{ec0}', '\u{ec4}', L), ('\u{ec6}', '\u{ec6}', L), ('\u{ec8}', '\u{ece}', + NSM), ('\u{ed0}', '\u{ed9}', L), ('\u{edc}', '\u{edf}', L), ('\u{f00}', '\u{f17}', L), + ('\u{f18}', '\u{f19}', NSM), ('\u{f1a}', '\u{f34}', L), ('\u{f35}', '\u{f35}', NSM), ('\u{f36}', + '\u{f36}', L), ('\u{f37}', '\u{f37}', NSM), ('\u{f38}', '\u{f38}', L), ('\u{f39}', '\u{f39}', + NSM), ('\u{f3a}', '\u{f3d}', ON), ('\u{f3e}', '\u{f47}', L), ('\u{f49}', '\u{f6c}', L), + ('\u{f71}', '\u{f7e}', NSM), ('\u{f7f}', '\u{f7f}', L), ('\u{f80}', '\u{f84}', NSM), ('\u{f85}', + '\u{f85}', L), ('\u{f86}', '\u{f87}', NSM), ('\u{f88}', '\u{f8c}', L), ('\u{f8d}', '\u{f97}', + NSM), ('\u{f99}', '\u{fbc}', NSM), ('\u{fbe}', '\u{fc5}', L), ('\u{fc6}', '\u{fc6}', NSM), + ('\u{fc7}', '\u{fcc}', L), ('\u{fce}', '\u{fda}', L), ('\u{1000}', '\u{102c}', L), ('\u{102d}', + '\u{1030}', NSM), ('\u{1031}', '\u{1031}', L), ('\u{1032}', '\u{1037}', NSM), ('\u{1038}', + '\u{1038}', L), ('\u{1039}', '\u{103a}', NSM), ('\u{103b}', '\u{103c}', L), ('\u{103d}', + '\u{103e}', NSM), ('\u{103f}', '\u{1057}', L), ('\u{1058}', '\u{1059}', NSM), ('\u{105a}', + '\u{105d}', L), ('\u{105e}', '\u{1060}', NSM), ('\u{1061}', '\u{1070}', L), ('\u{1071}', + '\u{1074}', NSM), ('\u{1075}', '\u{1081}', L), ('\u{1082}', '\u{1082}', NSM), ('\u{1083}', + '\u{1084}', L), ('\u{1085}', '\u{1086}', NSM), ('\u{1087}', '\u{108c}', L), ('\u{108d}', + '\u{108d}', NSM), ('\u{108e}', '\u{109c}', L), ('\u{109d}', '\u{109d}', NSM), ('\u{109e}', + '\u{10c5}', L), ('\u{10c7}', '\u{10c7}', L), ('\u{10cd}', '\u{10cd}', L), ('\u{10d0}', + '\u{1248}', L), ('\u{124a}', '\u{124d}', L), ('\u{1250}', '\u{1256}', L), ('\u{1258}', + '\u{1258}', L), ('\u{125a}', '\u{125d}', L), ('\u{1260}', '\u{1288}', L), ('\u{128a}', + '\u{128d}', L), ('\u{1290}', '\u{12b0}', L), ('\u{12b2}', '\u{12b5}', L), ('\u{12b8}', + '\u{12be}', L), ('\u{12c0}', '\u{12c0}', L), ('\u{12c2}', '\u{12c5}', L), ('\u{12c8}', + '\u{12d6}', L), ('\u{12d8}', '\u{1310}', L), ('\u{1312}', '\u{1315}', L), ('\u{1318}', + '\u{135a}', L), ('\u{135d}', '\u{135f}', NSM), ('\u{1360}', '\u{137c}', L), ('\u{1380}', + '\u{138f}', L), ('\u{1390}', '\u{1399}', ON), ('\u{13a0}', '\u{13f5}', L), ('\u{13f8}', + '\u{13fd}', L), ('\u{1400}', '\u{1400}', ON), ('\u{1401}', '\u{167f}', L), ('\u{1680}', + '\u{1680}', WS), ('\u{1681}', '\u{169a}', L), ('\u{169b}', '\u{169c}', ON), ('\u{16a0}', + '\u{16f8}', L), ('\u{1700}', '\u{1711}', L), ('\u{1712}', '\u{1714}', NSM), ('\u{1715}', + '\u{1715}', L), ('\u{171f}', '\u{1731}', L), ('\u{1732}', '\u{1733}', NSM), ('\u{1734}', + '\u{1736}', L), ('\u{1740}', '\u{1751}', L), ('\u{1752}', '\u{1753}', NSM), ('\u{1760}', + '\u{176c}', L), ('\u{176e}', '\u{1770}', L), ('\u{1772}', '\u{1773}', NSM), ('\u{1780}', + '\u{17b3}', L), ('\u{17b4}', '\u{17b5}', NSM), ('\u{17b6}', '\u{17b6}', L), ('\u{17b7}', + '\u{17bd}', NSM), ('\u{17be}', '\u{17c5}', L), ('\u{17c6}', '\u{17c6}', NSM), ('\u{17c7}', + '\u{17c8}', L), ('\u{17c9}', '\u{17d3}', NSM), ('\u{17d4}', '\u{17da}', L), ('\u{17db}', + '\u{17db}', ET), ('\u{17dc}', '\u{17dc}', L), ('\u{17dd}', '\u{17dd}', NSM), ('\u{17e0}', + '\u{17e9}', L), ('\u{17f0}', '\u{17f9}', ON), ('\u{1800}', '\u{180a}', ON), ('\u{180b}', + '\u{180d}', NSM), ('\u{180e}', '\u{180e}', BN), ('\u{180f}', '\u{180f}', NSM), ('\u{1810}', + '\u{1819}', L), ('\u{1820}', '\u{1878}', L), ('\u{1880}', '\u{1884}', L), ('\u{1885}', + '\u{1886}', NSM), ('\u{1887}', '\u{18a8}', L), ('\u{18a9}', '\u{18a9}', NSM), ('\u{18aa}', + '\u{18aa}', L), ('\u{18b0}', '\u{18f5}', L), ('\u{1900}', '\u{191e}', L), ('\u{1920}', + '\u{1922}', NSM), ('\u{1923}', '\u{1926}', L), ('\u{1927}', '\u{1928}', NSM), ('\u{1929}', + '\u{192b}', L), ('\u{1930}', '\u{1931}', L), ('\u{1932}', '\u{1932}', NSM), ('\u{1933}', + '\u{1938}', L), ('\u{1939}', '\u{193b}', NSM), ('\u{1940}', '\u{1940}', ON), ('\u{1944}', + '\u{1945}', ON), ('\u{1946}', '\u{196d}', L), ('\u{1970}', '\u{1974}', L), ('\u{1980}', + '\u{19ab}', L), ('\u{19b0}', '\u{19c9}', L), ('\u{19d0}', '\u{19da}', L), ('\u{19de}', + '\u{19ff}', ON), ('\u{1a00}', '\u{1a16}', L), ('\u{1a17}', '\u{1a18}', NSM), ('\u{1a19}', + '\u{1a1a}', L), ('\u{1a1b}', '\u{1a1b}', NSM), ('\u{1a1e}', '\u{1a55}', L), ('\u{1a56}', + '\u{1a56}', NSM), ('\u{1a57}', '\u{1a57}', L), ('\u{1a58}', '\u{1a5e}', NSM), ('\u{1a60}', + '\u{1a60}', NSM), ('\u{1a61}', '\u{1a61}', L), ('\u{1a62}', '\u{1a62}', NSM), ('\u{1a63}', + '\u{1a64}', L), ('\u{1a65}', '\u{1a6c}', NSM), ('\u{1a6d}', '\u{1a72}', L), ('\u{1a73}', + '\u{1a7c}', NSM), ('\u{1a7f}', '\u{1a7f}', NSM), ('\u{1a80}', '\u{1a89}', L), ('\u{1a90}', + '\u{1a99}', L), ('\u{1aa0}', '\u{1aad}', L), ('\u{1ab0}', '\u{1ace}', NSM), ('\u{1b00}', + '\u{1b03}', NSM), ('\u{1b04}', '\u{1b33}', L), ('\u{1b34}', '\u{1b34}', NSM), ('\u{1b35}', + '\u{1b35}', L), ('\u{1b36}', '\u{1b3a}', NSM), ('\u{1b3b}', '\u{1b3b}', L), ('\u{1b3c}', + '\u{1b3c}', NSM), ('\u{1b3d}', '\u{1b41}', L), ('\u{1b42}', '\u{1b42}', NSM), ('\u{1b43}', + '\u{1b4c}', L), ('\u{1b50}', '\u{1b6a}', L), ('\u{1b6b}', '\u{1b73}', NSM), ('\u{1b74}', + '\u{1b7e}', L), ('\u{1b80}', '\u{1b81}', NSM), ('\u{1b82}', '\u{1ba1}', L), ('\u{1ba2}', + '\u{1ba5}', NSM), ('\u{1ba6}', '\u{1ba7}', L), ('\u{1ba8}', '\u{1ba9}', NSM), ('\u{1baa}', + '\u{1baa}', L), ('\u{1bab}', '\u{1bad}', NSM), ('\u{1bae}', '\u{1be5}', L), ('\u{1be6}', + '\u{1be6}', NSM), ('\u{1be7}', '\u{1be7}', L), ('\u{1be8}', '\u{1be9}', NSM), ('\u{1bea}', + '\u{1bec}', L), ('\u{1bed}', '\u{1bed}', NSM), ('\u{1bee}', '\u{1bee}', L), ('\u{1bef}', + '\u{1bf1}', NSM), ('\u{1bf2}', '\u{1bf3}', L), ('\u{1bfc}', '\u{1c2b}', L), ('\u{1c2c}', + '\u{1c33}', NSM), ('\u{1c34}', '\u{1c35}', L), ('\u{1c36}', '\u{1c37}', NSM), ('\u{1c3b}', + '\u{1c49}', L), ('\u{1c4d}', '\u{1c88}', L), ('\u{1c90}', '\u{1cba}', L), ('\u{1cbd}', + '\u{1cc7}', L), ('\u{1cd0}', '\u{1cd2}', NSM), ('\u{1cd3}', '\u{1cd3}', L), ('\u{1cd4}', + '\u{1ce0}', NSM), ('\u{1ce1}', '\u{1ce1}', L), ('\u{1ce2}', '\u{1ce8}', NSM), ('\u{1ce9}', + '\u{1cec}', L), ('\u{1ced}', '\u{1ced}', NSM), ('\u{1cee}', '\u{1cf3}', L), ('\u{1cf4}', + '\u{1cf4}', NSM), ('\u{1cf5}', '\u{1cf7}', L), ('\u{1cf8}', '\u{1cf9}', NSM), ('\u{1cfa}', + '\u{1cfa}', L), ('\u{1d00}', '\u{1dbf}', L), ('\u{1dc0}', '\u{1dff}', NSM), ('\u{1e00}', + '\u{1f15}', L), ('\u{1f18}', '\u{1f1d}', L), ('\u{1f20}', '\u{1f45}', L), ('\u{1f48}', + '\u{1f4d}', L), ('\u{1f50}', '\u{1f57}', L), ('\u{1f59}', '\u{1f59}', L), ('\u{1f5b}', + '\u{1f5b}', L), ('\u{1f5d}', '\u{1f5d}', L), ('\u{1f5f}', '\u{1f7d}', L), ('\u{1f80}', + '\u{1fb4}', L), ('\u{1fb6}', '\u{1fbc}', L), ('\u{1fbd}', '\u{1fbd}', ON), ('\u{1fbe}', + '\u{1fbe}', L), ('\u{1fbf}', '\u{1fc1}', ON), ('\u{1fc2}', '\u{1fc4}', L), ('\u{1fc6}', + '\u{1fcc}', L), ('\u{1fcd}', '\u{1fcf}', ON), ('\u{1fd0}', '\u{1fd3}', L), ('\u{1fd6}', + '\u{1fdb}', L), ('\u{1fdd}', '\u{1fdf}', ON), ('\u{1fe0}', '\u{1fec}', L), ('\u{1fed}', + '\u{1fef}', ON), ('\u{1ff2}', '\u{1ff4}', L), ('\u{1ff6}', '\u{1ffc}', L), ('\u{1ffd}', + '\u{1ffe}', ON), ('\u{2000}', '\u{200a}', WS), ('\u{200b}', '\u{200d}', BN), ('\u{200e}', + '\u{200e}', L), ('\u{200f}', '\u{200f}', R), ('\u{2010}', '\u{2027}', ON), ('\u{2028}', + '\u{2028}', WS), ('\u{2029}', '\u{2029}', B), ('\u{202a}', '\u{202a}', LRE), ('\u{202b}', + '\u{202b}', RLE), ('\u{202c}', '\u{202c}', PDF), ('\u{202d}', '\u{202d}', LRO), ('\u{202e}', + '\u{202e}', RLO), ('\u{202f}', '\u{202f}', CS), ('\u{2030}', '\u{2034}', ET), ('\u{2035}', + '\u{2043}', ON), ('\u{2044}', '\u{2044}', CS), ('\u{2045}', '\u{205e}', ON), ('\u{205f}', + '\u{205f}', WS), ('\u{2060}', '\u{2064}', BN), ('\u{2066}', '\u{2066}', LRI), ('\u{2067}', + '\u{2067}', RLI), ('\u{2068}', '\u{2068}', FSI), ('\u{2069}', '\u{2069}', PDI), ('\u{206a}', + '\u{206f}', BN), ('\u{2070}', '\u{2070}', EN), ('\u{2071}', '\u{2071}', L), ('\u{2074}', + '\u{2079}', EN), ('\u{207a}', '\u{207b}', ES), ('\u{207c}', '\u{207e}', ON), ('\u{207f}', + '\u{207f}', L), ('\u{2080}', '\u{2089}', EN), ('\u{208a}', '\u{208b}', ES), ('\u{208c}', + '\u{208e}', ON), ('\u{2090}', '\u{209c}', L), ('\u{20a0}', '\u{20cf}', ET), ('\u{20d0}', + '\u{20f0}', NSM), ('\u{2100}', '\u{2101}', ON), ('\u{2102}', '\u{2102}', L), ('\u{2103}', + '\u{2106}', ON), ('\u{2107}', '\u{2107}', L), ('\u{2108}', '\u{2109}', ON), ('\u{210a}', + '\u{2113}', L), ('\u{2114}', '\u{2114}', ON), ('\u{2115}', '\u{2115}', L), ('\u{2116}', + '\u{2118}', ON), ('\u{2119}', '\u{211d}', L), ('\u{211e}', '\u{2123}', ON), ('\u{2124}', + '\u{2124}', L), ('\u{2125}', '\u{2125}', ON), ('\u{2126}', '\u{2126}', L), ('\u{2127}', + '\u{2127}', ON), ('\u{2128}', '\u{2128}', L), ('\u{2129}', '\u{2129}', ON), ('\u{212a}', + '\u{212d}', L), ('\u{212e}', '\u{212e}', ET), ('\u{212f}', '\u{2139}', L), ('\u{213a}', + '\u{213b}', ON), ('\u{213c}', '\u{213f}', L), ('\u{2140}', '\u{2144}', ON), ('\u{2145}', + '\u{2149}', L), ('\u{214a}', '\u{214d}', ON), ('\u{214e}', '\u{214f}', L), ('\u{2150}', + '\u{215f}', ON), ('\u{2160}', '\u{2188}', L), ('\u{2189}', '\u{218b}', ON), ('\u{2190}', + '\u{2211}', ON), ('\u{2212}', '\u{2212}', ES), ('\u{2213}', '\u{2213}', ET), ('\u{2214}', + '\u{2335}', ON), ('\u{2336}', '\u{237a}', L), ('\u{237b}', '\u{2394}', ON), ('\u{2395}', + '\u{2395}', L), ('\u{2396}', '\u{2426}', ON), ('\u{2440}', '\u{244a}', ON), ('\u{2460}', + '\u{2487}', ON), ('\u{2488}', '\u{249b}', EN), ('\u{249c}', '\u{24e9}', L), ('\u{24ea}', + '\u{26ab}', ON), ('\u{26ac}', '\u{26ac}', L), ('\u{26ad}', '\u{27ff}', ON), ('\u{2800}', + '\u{28ff}', L), ('\u{2900}', '\u{2b73}', ON), ('\u{2b76}', '\u{2b95}', ON), ('\u{2b97}', + '\u{2bff}', ON), ('\u{2c00}', '\u{2ce4}', L), ('\u{2ce5}', '\u{2cea}', ON), ('\u{2ceb}', + '\u{2cee}', L), ('\u{2cef}', '\u{2cf1}', NSM), ('\u{2cf2}', '\u{2cf3}', L), ('\u{2cf9}', + '\u{2cff}', ON), ('\u{2d00}', '\u{2d25}', L), ('\u{2d27}', '\u{2d27}', L), ('\u{2d2d}', + '\u{2d2d}', L), ('\u{2d30}', '\u{2d67}', L), ('\u{2d6f}', '\u{2d70}', L), ('\u{2d7f}', + '\u{2d7f}', NSM), ('\u{2d80}', '\u{2d96}', L), ('\u{2da0}', '\u{2da6}', L), ('\u{2da8}', + '\u{2dae}', L), ('\u{2db0}', '\u{2db6}', L), ('\u{2db8}', '\u{2dbe}', L), ('\u{2dc0}', + '\u{2dc6}', L), ('\u{2dc8}', '\u{2dce}', L), ('\u{2dd0}', '\u{2dd6}', L), ('\u{2dd8}', + '\u{2dde}', L), ('\u{2de0}', '\u{2dff}', NSM), ('\u{2e00}', '\u{2e5d}', ON), ('\u{2e80}', + '\u{2e99}', ON), ('\u{2e9b}', '\u{2ef3}', ON), ('\u{2f00}', '\u{2fd5}', ON), ('\u{2ff0}', + '\u{2ffb}', ON), ('\u{3000}', '\u{3000}', WS), ('\u{3001}', '\u{3004}', ON), ('\u{3005}', + '\u{3007}', L), ('\u{3008}', '\u{3020}', ON), ('\u{3021}', '\u{3029}', L), ('\u{302a}', + '\u{302d}', NSM), ('\u{302e}', '\u{302f}', L), ('\u{3030}', '\u{3030}', ON), ('\u{3031}', + '\u{3035}', L), ('\u{3036}', '\u{3037}', ON), ('\u{3038}', '\u{303c}', L), ('\u{303d}', + '\u{303f}', ON), ('\u{3041}', '\u{3096}', L), ('\u{3099}', '\u{309a}', NSM), ('\u{309b}', + '\u{309c}', ON), ('\u{309d}', '\u{309f}', L), ('\u{30a0}', '\u{30a0}', ON), ('\u{30a1}', + '\u{30fa}', L), ('\u{30fb}', '\u{30fb}', ON), ('\u{30fc}', '\u{30ff}', L), ('\u{3105}', + '\u{312f}', L), ('\u{3131}', '\u{318e}', L), ('\u{3190}', '\u{31bf}', L), ('\u{31c0}', + '\u{31e3}', ON), ('\u{31f0}', '\u{321c}', L), ('\u{321d}', '\u{321e}', ON), ('\u{3220}', + '\u{324f}', L), ('\u{3250}', '\u{325f}', ON), ('\u{3260}', '\u{327b}', L), ('\u{327c}', + '\u{327e}', ON), ('\u{327f}', '\u{32b0}', L), ('\u{32b1}', '\u{32bf}', ON), ('\u{32c0}', + '\u{32cb}', L), ('\u{32cc}', '\u{32cf}', ON), ('\u{32d0}', '\u{3376}', L), ('\u{3377}', + '\u{337a}', ON), ('\u{337b}', '\u{33dd}', L), ('\u{33de}', '\u{33df}', ON), ('\u{33e0}', + '\u{33fe}', L), ('\u{33ff}', '\u{33ff}', ON), ('\u{3400}', '\u{4dbf}', L), ('\u{4dc0}', + '\u{4dff}', ON), ('\u{4e00}', '\u{a48c}', L), ('\u{a490}', '\u{a4c6}', ON), ('\u{a4d0}', + '\u{a60c}', L), ('\u{a60d}', '\u{a60f}', ON), ('\u{a610}', '\u{a62b}', L), ('\u{a640}', + '\u{a66e}', L), ('\u{a66f}', '\u{a672}', NSM), ('\u{a673}', '\u{a673}', ON), ('\u{a674}', + '\u{a67d}', NSM), ('\u{a67e}', '\u{a67f}', ON), ('\u{a680}', '\u{a69d}', L), ('\u{a69e}', + '\u{a69f}', NSM), ('\u{a6a0}', '\u{a6ef}', L), ('\u{a6f0}', '\u{a6f1}', NSM), ('\u{a6f2}', + '\u{a6f7}', L), ('\u{a700}', '\u{a721}', ON), ('\u{a722}', '\u{a787}', L), ('\u{a788}', + '\u{a788}', ON), ('\u{a789}', '\u{a7ca}', L), ('\u{a7d0}', '\u{a7d1}', L), ('\u{a7d3}', + '\u{a7d3}', L), ('\u{a7d5}', '\u{a7d9}', L), ('\u{a7f2}', '\u{a801}', L), ('\u{a802}', + '\u{a802}', NSM), ('\u{a803}', '\u{a805}', L), ('\u{a806}', '\u{a806}', NSM), ('\u{a807}', + '\u{a80a}', L), ('\u{a80b}', '\u{a80b}', NSM), ('\u{a80c}', '\u{a824}', L), ('\u{a825}', + '\u{a826}', NSM), ('\u{a827}', '\u{a827}', L), ('\u{a828}', '\u{a82b}', ON), ('\u{a82c}', + '\u{a82c}', NSM), ('\u{a830}', '\u{a837}', L), ('\u{a838}', '\u{a839}', ET), ('\u{a840}', + '\u{a873}', L), ('\u{a874}', '\u{a877}', ON), ('\u{a880}', '\u{a8c3}', L), ('\u{a8c4}', + '\u{a8c5}', NSM), ('\u{a8ce}', '\u{a8d9}', L), ('\u{a8e0}', '\u{a8f1}', NSM), ('\u{a8f2}', + '\u{a8fe}', L), ('\u{a8ff}', '\u{a8ff}', NSM), ('\u{a900}', '\u{a925}', L), ('\u{a926}', + '\u{a92d}', NSM), ('\u{a92e}', '\u{a946}', L), ('\u{a947}', '\u{a951}', NSM), ('\u{a952}', + '\u{a953}', L), ('\u{a95f}', '\u{a97c}', L), ('\u{a980}', '\u{a982}', NSM), ('\u{a983}', + '\u{a9b2}', L), ('\u{a9b3}', '\u{a9b3}', NSM), ('\u{a9b4}', '\u{a9b5}', L), ('\u{a9b6}', + '\u{a9b9}', NSM), ('\u{a9ba}', '\u{a9bb}', L), ('\u{a9bc}', '\u{a9bd}', NSM), ('\u{a9be}', + '\u{a9cd}', L), ('\u{a9cf}', '\u{a9d9}', L), ('\u{a9de}', '\u{a9e4}', L), ('\u{a9e5}', + '\u{a9e5}', NSM), ('\u{a9e6}', '\u{a9fe}', L), ('\u{aa00}', '\u{aa28}', L), ('\u{aa29}', + '\u{aa2e}', NSM), ('\u{aa2f}', '\u{aa30}', L), ('\u{aa31}', '\u{aa32}', NSM), ('\u{aa33}', + '\u{aa34}', L), ('\u{aa35}', '\u{aa36}', NSM), ('\u{aa40}', '\u{aa42}', L), ('\u{aa43}', + '\u{aa43}', NSM), ('\u{aa44}', '\u{aa4b}', L), ('\u{aa4c}', '\u{aa4c}', NSM), ('\u{aa4d}', + '\u{aa4d}', L), ('\u{aa50}', '\u{aa59}', L), ('\u{aa5c}', '\u{aa7b}', L), ('\u{aa7c}', + '\u{aa7c}', NSM), ('\u{aa7d}', '\u{aaaf}', L), ('\u{aab0}', '\u{aab0}', NSM), ('\u{aab1}', + '\u{aab1}', L), ('\u{aab2}', '\u{aab4}', NSM), ('\u{aab5}', '\u{aab6}', L), ('\u{aab7}', + '\u{aab8}', NSM), ('\u{aab9}', '\u{aabd}', L), ('\u{aabe}', '\u{aabf}', NSM), ('\u{aac0}', + '\u{aac0}', L), ('\u{aac1}', '\u{aac1}', NSM), ('\u{aac2}', '\u{aac2}', L), ('\u{aadb}', + '\u{aaeb}', L), ('\u{aaec}', '\u{aaed}', NSM), ('\u{aaee}', '\u{aaf5}', L), ('\u{aaf6}', + '\u{aaf6}', NSM), ('\u{ab01}', '\u{ab06}', L), ('\u{ab09}', '\u{ab0e}', L), ('\u{ab11}', + '\u{ab16}', L), ('\u{ab20}', '\u{ab26}', L), ('\u{ab28}', '\u{ab2e}', L), ('\u{ab30}', + '\u{ab69}', L), ('\u{ab6a}', '\u{ab6b}', ON), ('\u{ab70}', '\u{abe4}', L), ('\u{abe5}', + '\u{abe5}', NSM), ('\u{abe6}', '\u{abe7}', L), ('\u{abe8}', '\u{abe8}', NSM), ('\u{abe9}', + '\u{abec}', L), ('\u{abed}', '\u{abed}', NSM), ('\u{abf0}', '\u{abf9}', L), ('\u{ac00}', + '\u{d7a3}', L), ('\u{d7b0}', '\u{d7c6}', L), ('\u{d7cb}', '\u{d7fb}', L), ('\u{e000}', + '\u{fa6d}', L), ('\u{fa70}', '\u{fad9}', L), ('\u{fb00}', '\u{fb06}', L), ('\u{fb13}', + '\u{fb17}', L), ('\u{fb1d}', '\u{fb1d}', R), ('\u{fb1e}', '\u{fb1e}', NSM), ('\u{fb1f}', + '\u{fb28}', R), ('\u{fb29}', '\u{fb29}', ES), ('\u{fb2a}', '\u{fb4f}', R), ('\u{fb50}', + '\u{fd3d}', AL), ('\u{fd3e}', '\u{fd4f}', ON), ('\u{fd50}', '\u{fdce}', AL), ('\u{fdcf}', + '\u{fdcf}', ON), ('\u{fdf0}', '\u{fdfc}', AL), ('\u{fdfd}', '\u{fdff}', ON), ('\u{fe00}', + '\u{fe0f}', NSM), ('\u{fe10}', '\u{fe19}', ON), ('\u{fe20}', '\u{fe2f}', NSM), ('\u{fe30}', + '\u{fe4f}', ON), ('\u{fe50}', '\u{fe50}', CS), ('\u{fe51}', '\u{fe51}', ON), ('\u{fe52}', + '\u{fe52}', CS), ('\u{fe54}', '\u{fe54}', ON), ('\u{fe55}', '\u{fe55}', CS), ('\u{fe56}', + '\u{fe5e}', ON), ('\u{fe5f}', '\u{fe5f}', ET), ('\u{fe60}', '\u{fe61}', ON), ('\u{fe62}', + '\u{fe63}', ES), ('\u{fe64}', '\u{fe66}', ON), ('\u{fe68}', '\u{fe68}', ON), ('\u{fe69}', + '\u{fe6a}', ET), ('\u{fe6b}', '\u{fe6b}', ON), ('\u{fe70}', '\u{fefe}', AL), ('\u{feff}', + '\u{feff}', BN), ('\u{ff01}', '\u{ff02}', ON), ('\u{ff03}', '\u{ff05}', ET), ('\u{ff06}', + '\u{ff0a}', ON), ('\u{ff0b}', '\u{ff0b}', ES), ('\u{ff0c}', '\u{ff0c}', CS), ('\u{ff0d}', + '\u{ff0d}', ES), ('\u{ff0e}', '\u{ff0f}', CS), ('\u{ff10}', '\u{ff19}', EN), ('\u{ff1a}', + '\u{ff1a}', CS), ('\u{ff1b}', '\u{ff20}', ON), ('\u{ff21}', '\u{ff3a}', L), ('\u{ff3b}', + '\u{ff40}', ON), ('\u{ff41}', '\u{ff5a}', L), ('\u{ff5b}', '\u{ff65}', ON), ('\u{ff66}', + '\u{ffbe}', L), ('\u{ffc2}', '\u{ffc7}', L), ('\u{ffca}', '\u{ffcf}', L), ('\u{ffd2}', + '\u{ffd7}', L), ('\u{ffda}', '\u{ffdc}', L), ('\u{ffe0}', '\u{ffe1}', ET), ('\u{ffe2}', + '\u{ffe4}', ON), ('\u{ffe5}', '\u{ffe6}', ET), ('\u{ffe8}', '\u{ffee}', ON), ('\u{fff9}', + '\u{fffd}', ON), ('\u{10000}', '\u{1000b}', L), ('\u{1000d}', '\u{10026}', L), ('\u{10028}', + '\u{1003a}', L), ('\u{1003c}', '\u{1003d}', L), ('\u{1003f}', '\u{1004d}', L), ('\u{10050}', + '\u{1005d}', L), ('\u{10080}', '\u{100fa}', L), ('\u{10100}', '\u{10100}', L), ('\u{10101}', + '\u{10101}', ON), ('\u{10102}', '\u{10102}', L), ('\u{10107}', '\u{10133}', L), ('\u{10137}', + '\u{1013f}', L), ('\u{10140}', '\u{1018c}', ON), ('\u{1018d}', '\u{1018e}', L), ('\u{10190}', + '\u{1019c}', ON), ('\u{101a0}', '\u{101a0}', ON), ('\u{101d0}', '\u{101fc}', L), ('\u{101fd}', + '\u{101fd}', NSM), ('\u{10280}', '\u{1029c}', L), ('\u{102a0}', '\u{102d0}', L), ('\u{102e0}', + '\u{102e0}', NSM), ('\u{102e1}', '\u{102fb}', EN), ('\u{10300}', '\u{10323}', L), ('\u{1032d}', + '\u{1034a}', L), ('\u{10350}', '\u{10375}', L), ('\u{10376}', '\u{1037a}', NSM), ('\u{10380}', + '\u{1039d}', L), ('\u{1039f}', '\u{103c3}', L), ('\u{103c8}', '\u{103d5}', L), ('\u{10400}', + '\u{1049d}', L), ('\u{104a0}', '\u{104a9}', L), ('\u{104b0}', '\u{104d3}', L), ('\u{104d8}', + '\u{104fb}', L), ('\u{10500}', '\u{10527}', L), ('\u{10530}', '\u{10563}', L), ('\u{1056f}', + '\u{1057a}', L), ('\u{1057c}', '\u{1058a}', L), ('\u{1058c}', '\u{10592}', L), ('\u{10594}', + '\u{10595}', L), ('\u{10597}', '\u{105a1}', L), ('\u{105a3}', '\u{105b1}', L), ('\u{105b3}', + '\u{105b9}', L), ('\u{105bb}', '\u{105bc}', L), ('\u{10600}', '\u{10736}', L), ('\u{10740}', + '\u{10755}', L), ('\u{10760}', '\u{10767}', L), ('\u{10780}', '\u{10785}', L), ('\u{10787}', + '\u{107b0}', L), ('\u{107b2}', '\u{107ba}', L), ('\u{10800}', '\u{1091e}', R), ('\u{1091f}', + '\u{1091f}', ON), ('\u{10920}', '\u{10a00}', R), ('\u{10a01}', '\u{10a03}', NSM), ('\u{10a04}', + '\u{10a04}', R), ('\u{10a05}', '\u{10a06}', NSM), ('\u{10a07}', '\u{10a0b}', R), ('\u{10a0c}', + '\u{10a0f}', NSM), ('\u{10a10}', '\u{10a37}', R), ('\u{10a38}', '\u{10a3a}', NSM), ('\u{10a3b}', + '\u{10a3e}', R), ('\u{10a3f}', '\u{10a3f}', NSM), ('\u{10a40}', '\u{10ae4}', R), ('\u{10ae5}', + '\u{10ae6}', NSM), ('\u{10ae7}', '\u{10b38}', R), ('\u{10b39}', '\u{10b3f}', ON), ('\u{10b40}', + '\u{10cff}', R), ('\u{10d00}', '\u{10d23}', AL), ('\u{10d24}', '\u{10d27}', NSM), ('\u{10d28}', + '\u{10d2f}', R), ('\u{10d30}', '\u{10d39}', AN), ('\u{10d3a}', '\u{10e5f}', R), ('\u{10e60}', + '\u{10e7e}', AN), ('\u{10e7f}', '\u{10eaa}', R), ('\u{10eab}', '\u{10eac}', NSM), ('\u{10ead}', + '\u{10efc}', R), ('\u{10efd}', '\u{10eff}', NSM), ('\u{10f00}', '\u{10f2f}', R), ('\u{10f30}', + '\u{10f45}', AL), ('\u{10f46}', '\u{10f50}', NSM), ('\u{10f51}', '\u{10f59}', AL), ('\u{10f5a}', + '\u{10f81}', R), ('\u{10f82}', '\u{10f85}', NSM), ('\u{10f86}', '\u{10fff}', R), ('\u{11000}', + '\u{11000}', L), ('\u{11001}', '\u{11001}', NSM), ('\u{11002}', '\u{11037}', L), ('\u{11038}', + '\u{11046}', NSM), ('\u{11047}', '\u{1104d}', L), ('\u{11052}', '\u{11065}', ON), ('\u{11066}', + '\u{1106f}', L), ('\u{11070}', '\u{11070}', NSM), ('\u{11071}', '\u{11072}', L), ('\u{11073}', + '\u{11074}', NSM), ('\u{11075}', '\u{11075}', L), ('\u{1107f}', '\u{11081}', NSM), ('\u{11082}', + '\u{110b2}', L), ('\u{110b3}', '\u{110b6}', NSM), ('\u{110b7}', '\u{110b8}', L), ('\u{110b9}', + '\u{110ba}', NSM), ('\u{110bb}', '\u{110c1}', L), ('\u{110c2}', '\u{110c2}', NSM), ('\u{110cd}', + '\u{110cd}', L), ('\u{110d0}', '\u{110e8}', L), ('\u{110f0}', '\u{110f9}', L), ('\u{11100}', + '\u{11102}', NSM), ('\u{11103}', '\u{11126}', L), ('\u{11127}', '\u{1112b}', NSM), ('\u{1112c}', + '\u{1112c}', L), ('\u{1112d}', '\u{11134}', NSM), ('\u{11136}', '\u{11147}', L), ('\u{11150}', + '\u{11172}', L), ('\u{11173}', '\u{11173}', NSM), ('\u{11174}', '\u{11176}', L), ('\u{11180}', + '\u{11181}', NSM), ('\u{11182}', '\u{111b5}', L), ('\u{111b6}', '\u{111be}', NSM), ('\u{111bf}', + '\u{111c8}', L), ('\u{111c9}', '\u{111cc}', NSM), ('\u{111cd}', '\u{111ce}', L), ('\u{111cf}', + '\u{111cf}', NSM), ('\u{111d0}', '\u{111df}', L), ('\u{111e1}', '\u{111f4}', L), ('\u{11200}', + '\u{11211}', L), ('\u{11213}', '\u{1122e}', L), ('\u{1122f}', '\u{11231}', NSM), ('\u{11232}', + '\u{11233}', L), ('\u{11234}', '\u{11234}', NSM), ('\u{11235}', '\u{11235}', L), ('\u{11236}', + '\u{11237}', NSM), ('\u{11238}', '\u{1123d}', L), ('\u{1123e}', '\u{1123e}', NSM), ('\u{1123f}', + '\u{11240}', L), ('\u{11241}', '\u{11241}', NSM), ('\u{11280}', '\u{11286}', L), ('\u{11288}', + '\u{11288}', L), ('\u{1128a}', '\u{1128d}', L), ('\u{1128f}', '\u{1129d}', L), ('\u{1129f}', + '\u{112a9}', L), ('\u{112b0}', '\u{112de}', L), ('\u{112df}', '\u{112df}', NSM), ('\u{112e0}', + '\u{112e2}', L), ('\u{112e3}', '\u{112ea}', NSM), ('\u{112f0}', '\u{112f9}', L), ('\u{11300}', + '\u{11301}', NSM), ('\u{11302}', '\u{11303}', L), ('\u{11305}', '\u{1130c}', L), ('\u{1130f}', + '\u{11310}', L), ('\u{11313}', '\u{11328}', L), ('\u{1132a}', '\u{11330}', L), ('\u{11332}', + '\u{11333}', L), ('\u{11335}', '\u{11339}', L), ('\u{1133b}', '\u{1133c}', NSM), ('\u{1133d}', + '\u{1133f}', L), ('\u{11340}', '\u{11340}', NSM), ('\u{11341}', '\u{11344}', L), ('\u{11347}', + '\u{11348}', L), ('\u{1134b}', '\u{1134d}', L), ('\u{11350}', '\u{11350}', L), ('\u{11357}', + '\u{11357}', L), ('\u{1135d}', '\u{11363}', L), ('\u{11366}', '\u{1136c}', NSM), ('\u{11370}', + '\u{11374}', NSM), ('\u{11400}', '\u{11437}', L), ('\u{11438}', '\u{1143f}', NSM), ('\u{11440}', + '\u{11441}', L), ('\u{11442}', '\u{11444}', NSM), ('\u{11445}', '\u{11445}', L), ('\u{11446}', + '\u{11446}', NSM), ('\u{11447}', '\u{1145b}', L), ('\u{1145d}', '\u{1145d}', L), ('\u{1145e}', + '\u{1145e}', NSM), ('\u{1145f}', '\u{11461}', L), ('\u{11480}', '\u{114b2}', L), ('\u{114b3}', + '\u{114b8}', NSM), ('\u{114b9}', '\u{114b9}', L), ('\u{114ba}', '\u{114ba}', NSM), ('\u{114bb}', + '\u{114be}', L), ('\u{114bf}', '\u{114c0}', NSM), ('\u{114c1}', '\u{114c1}', L), ('\u{114c2}', + '\u{114c3}', NSM), ('\u{114c4}', '\u{114c7}', L), ('\u{114d0}', '\u{114d9}', L), ('\u{11580}', + '\u{115b1}', L), ('\u{115b2}', '\u{115b5}', NSM), ('\u{115b8}', '\u{115bb}', L), ('\u{115bc}', + '\u{115bd}', NSM), ('\u{115be}', '\u{115be}', L), ('\u{115bf}', '\u{115c0}', NSM), ('\u{115c1}', + '\u{115db}', L), ('\u{115dc}', '\u{115dd}', NSM), ('\u{11600}', '\u{11632}', L), ('\u{11633}', + '\u{1163a}', NSM), ('\u{1163b}', '\u{1163c}', L), ('\u{1163d}', '\u{1163d}', NSM), ('\u{1163e}', + '\u{1163e}', L), ('\u{1163f}', '\u{11640}', NSM), ('\u{11641}', '\u{11644}', L), ('\u{11650}', + '\u{11659}', L), ('\u{11660}', '\u{1166c}', ON), ('\u{11680}', '\u{116aa}', L), ('\u{116ab}', + '\u{116ab}', NSM), ('\u{116ac}', '\u{116ac}', L), ('\u{116ad}', '\u{116ad}', NSM), ('\u{116ae}', + '\u{116af}', L), ('\u{116b0}', '\u{116b5}', NSM), ('\u{116b6}', '\u{116b6}', L), ('\u{116b7}', + '\u{116b7}', NSM), ('\u{116b8}', '\u{116b9}', L), ('\u{116c0}', '\u{116c9}', L), ('\u{11700}', + '\u{1171a}', L), ('\u{1171d}', '\u{1171f}', NSM), ('\u{11720}', '\u{11721}', L), ('\u{11722}', + '\u{11725}', NSM), ('\u{11726}', '\u{11726}', L), ('\u{11727}', '\u{1172b}', NSM), ('\u{11730}', + '\u{11746}', L), ('\u{11800}', '\u{1182e}', L), ('\u{1182f}', '\u{11837}', NSM), ('\u{11838}', + '\u{11838}', L), ('\u{11839}', '\u{1183a}', NSM), ('\u{1183b}', '\u{1183b}', L), ('\u{118a0}', + '\u{118f2}', L), ('\u{118ff}', '\u{11906}', L), ('\u{11909}', '\u{11909}', L), ('\u{1190c}', + '\u{11913}', L), ('\u{11915}', '\u{11916}', L), ('\u{11918}', '\u{11935}', L), ('\u{11937}', + '\u{11938}', L), ('\u{1193b}', '\u{1193c}', NSM), ('\u{1193d}', '\u{1193d}', L), ('\u{1193e}', + '\u{1193e}', NSM), ('\u{1193f}', '\u{11942}', L), ('\u{11943}', '\u{11943}', NSM), ('\u{11944}', + '\u{11946}', L), ('\u{11950}', '\u{11959}', L), ('\u{119a0}', '\u{119a7}', L), ('\u{119aa}', + '\u{119d3}', L), ('\u{119d4}', '\u{119d7}', NSM), ('\u{119da}', '\u{119db}', NSM), ('\u{119dc}', + '\u{119df}', L), ('\u{119e0}', '\u{119e0}', NSM), ('\u{119e1}', '\u{119e4}', L), ('\u{11a00}', + '\u{11a00}', L), ('\u{11a01}', '\u{11a06}', NSM), ('\u{11a07}', '\u{11a08}', L), ('\u{11a09}', + '\u{11a0a}', NSM), ('\u{11a0b}', '\u{11a32}', L), ('\u{11a33}', '\u{11a38}', NSM), ('\u{11a39}', + '\u{11a3a}', L), ('\u{11a3b}', '\u{11a3e}', NSM), ('\u{11a3f}', '\u{11a46}', L), ('\u{11a47}', + '\u{11a47}', NSM), ('\u{11a50}', '\u{11a50}', L), ('\u{11a51}', '\u{11a56}', NSM), ('\u{11a57}', + '\u{11a58}', L), ('\u{11a59}', '\u{11a5b}', NSM), ('\u{11a5c}', '\u{11a89}', L), ('\u{11a8a}', + '\u{11a96}', NSM), ('\u{11a97}', '\u{11a97}', L), ('\u{11a98}', '\u{11a99}', NSM), ('\u{11a9a}', + '\u{11aa2}', L), ('\u{11ab0}', '\u{11af8}', L), ('\u{11b00}', '\u{11b09}', L), ('\u{11c00}', + '\u{11c08}', L), ('\u{11c0a}', '\u{11c2f}', L), ('\u{11c30}', '\u{11c36}', NSM), ('\u{11c38}', + '\u{11c3d}', NSM), ('\u{11c3e}', '\u{11c45}', L), ('\u{11c50}', '\u{11c6c}', L), ('\u{11c70}', + '\u{11c8f}', L), ('\u{11c92}', '\u{11ca7}', NSM), ('\u{11ca9}', '\u{11ca9}', L), ('\u{11caa}', + '\u{11cb0}', NSM), ('\u{11cb1}', '\u{11cb1}', L), ('\u{11cb2}', '\u{11cb3}', NSM), ('\u{11cb4}', + '\u{11cb4}', L), ('\u{11cb5}', '\u{11cb6}', NSM), ('\u{11d00}', '\u{11d06}', L), ('\u{11d08}', + '\u{11d09}', L), ('\u{11d0b}', '\u{11d30}', L), ('\u{11d31}', '\u{11d36}', NSM), ('\u{11d3a}', + '\u{11d3a}', NSM), ('\u{11d3c}', '\u{11d3d}', NSM), ('\u{11d3f}', '\u{11d45}', NSM), + ('\u{11d46}', '\u{11d46}', L), ('\u{11d47}', '\u{11d47}', NSM), ('\u{11d50}', '\u{11d59}', L), + ('\u{11d60}', '\u{11d65}', L), ('\u{11d67}', '\u{11d68}', L), ('\u{11d6a}', '\u{11d8e}', L), + ('\u{11d90}', '\u{11d91}', NSM), ('\u{11d93}', '\u{11d94}', L), ('\u{11d95}', '\u{11d95}', NSM), + ('\u{11d96}', '\u{11d96}', L), ('\u{11d97}', '\u{11d97}', NSM), ('\u{11d98}', '\u{11d98}', L), + ('\u{11da0}', '\u{11da9}', L), ('\u{11ee0}', '\u{11ef2}', L), ('\u{11ef3}', '\u{11ef4}', NSM), + ('\u{11ef5}', '\u{11ef8}', L), ('\u{11f00}', '\u{11f01}', NSM), ('\u{11f02}', '\u{11f10}', L), + ('\u{11f12}', '\u{11f35}', L), ('\u{11f36}', '\u{11f3a}', NSM), ('\u{11f3e}', '\u{11f3f}', L), + ('\u{11f40}', '\u{11f40}', NSM), ('\u{11f41}', '\u{11f41}', L), ('\u{11f42}', '\u{11f42}', NSM), + ('\u{11f43}', '\u{11f59}', L), ('\u{11fb0}', '\u{11fb0}', L), ('\u{11fc0}', '\u{11fd4}', L), + ('\u{11fd5}', '\u{11fdc}', ON), ('\u{11fdd}', '\u{11fe0}', ET), ('\u{11fe1}', '\u{11ff1}', ON), + ('\u{11fff}', '\u{12399}', L), ('\u{12400}', '\u{1246e}', L), ('\u{12470}', '\u{12474}', L), + ('\u{12480}', '\u{12543}', L), ('\u{12f90}', '\u{12ff2}', L), ('\u{13000}', '\u{1343f}', L), + ('\u{13440}', '\u{13440}', NSM), ('\u{13441}', '\u{13446}', L), ('\u{13447}', '\u{13455}', NSM), + ('\u{14400}', '\u{14646}', L), ('\u{16800}', '\u{16a38}', L), ('\u{16a40}', '\u{16a5e}', L), + ('\u{16a60}', '\u{16a69}', L), ('\u{16a6e}', '\u{16abe}', L), ('\u{16ac0}', '\u{16ac9}', L), + ('\u{16ad0}', '\u{16aed}', L), ('\u{16af0}', '\u{16af4}', NSM), ('\u{16af5}', '\u{16af5}', L), + ('\u{16b00}', '\u{16b2f}', L), ('\u{16b30}', '\u{16b36}', NSM), ('\u{16b37}', '\u{16b45}', L), + ('\u{16b50}', '\u{16b59}', L), ('\u{16b5b}', '\u{16b61}', L), ('\u{16b63}', '\u{16b77}', L), + ('\u{16b7d}', '\u{16b8f}', L), ('\u{16e40}', '\u{16e9a}', L), ('\u{16f00}', '\u{16f4a}', L), + ('\u{16f4f}', '\u{16f4f}', NSM), ('\u{16f50}', '\u{16f87}', L), ('\u{16f8f}', '\u{16f92}', NSM), + ('\u{16f93}', '\u{16f9f}', L), ('\u{16fe0}', '\u{16fe1}', L), ('\u{16fe2}', '\u{16fe2}', ON), + ('\u{16fe3}', '\u{16fe3}', L), ('\u{16fe4}', '\u{16fe4}', NSM), ('\u{16ff0}', '\u{16ff1}', L), + ('\u{17000}', '\u{187f7}', L), ('\u{18800}', '\u{18cd5}', L), ('\u{18d00}', '\u{18d08}', L), + ('\u{1aff0}', '\u{1aff3}', L), ('\u{1aff5}', '\u{1affb}', L), ('\u{1affd}', '\u{1affe}', L), + ('\u{1b000}', '\u{1b122}', L), ('\u{1b132}', '\u{1b132}', L), ('\u{1b150}', '\u{1b152}', L), + ('\u{1b155}', '\u{1b155}', L), ('\u{1b164}', '\u{1b167}', L), ('\u{1b170}', '\u{1b2fb}', L), + ('\u{1bc00}', '\u{1bc6a}', L), ('\u{1bc70}', '\u{1bc7c}', L), ('\u{1bc80}', '\u{1bc88}', L), + ('\u{1bc90}', '\u{1bc99}', L), ('\u{1bc9c}', '\u{1bc9c}', L), ('\u{1bc9d}', '\u{1bc9e}', NSM), + ('\u{1bc9f}', '\u{1bc9f}', L), ('\u{1bca0}', '\u{1bca3}', BN), ('\u{1cf00}', '\u{1cf2d}', NSM), + ('\u{1cf30}', '\u{1cf46}', NSM), ('\u{1cf50}', '\u{1cfc3}', L), ('\u{1d000}', '\u{1d0f5}', L), + ('\u{1d100}', '\u{1d126}', L), ('\u{1d129}', '\u{1d166}', L), ('\u{1d167}', '\u{1d169}', NSM), + ('\u{1d16a}', '\u{1d172}', L), ('\u{1d173}', '\u{1d17a}', BN), ('\u{1d17b}', '\u{1d182}', NSM), + ('\u{1d183}', '\u{1d184}', L), ('\u{1d185}', '\u{1d18b}', NSM), ('\u{1d18c}', '\u{1d1a9}', L), + ('\u{1d1aa}', '\u{1d1ad}', NSM), ('\u{1d1ae}', '\u{1d1e8}', L), ('\u{1d1e9}', '\u{1d1ea}', ON), + ('\u{1d200}', '\u{1d241}', ON), ('\u{1d242}', '\u{1d244}', NSM), ('\u{1d245}', '\u{1d245}', ON), + ('\u{1d2c0}', '\u{1d2d3}', L), ('\u{1d2e0}', '\u{1d2f3}', L), ('\u{1d300}', '\u{1d356}', ON), + ('\u{1d360}', '\u{1d378}', L), ('\u{1d400}', '\u{1d454}', L), ('\u{1d456}', '\u{1d49c}', L), + ('\u{1d49e}', '\u{1d49f}', L), ('\u{1d4a2}', '\u{1d4a2}', L), ('\u{1d4a5}', '\u{1d4a6}', L), + ('\u{1d4a9}', '\u{1d4ac}', L), ('\u{1d4ae}', '\u{1d4b9}', L), ('\u{1d4bb}', '\u{1d4bb}', L), + ('\u{1d4bd}', '\u{1d4c3}', L), ('\u{1d4c5}', '\u{1d505}', L), ('\u{1d507}', '\u{1d50a}', L), + ('\u{1d50d}', '\u{1d514}', L), ('\u{1d516}', '\u{1d51c}', L), ('\u{1d51e}', '\u{1d539}', L), + ('\u{1d53b}', '\u{1d53e}', L), ('\u{1d540}', '\u{1d544}', L), ('\u{1d546}', '\u{1d546}', L), + ('\u{1d54a}', '\u{1d550}', L), ('\u{1d552}', '\u{1d6a5}', L), ('\u{1d6a8}', '\u{1d6da}', L), + ('\u{1d6db}', '\u{1d6db}', ON), ('\u{1d6dc}', '\u{1d714}', L), ('\u{1d715}', '\u{1d715}', ON), + ('\u{1d716}', '\u{1d74e}', L), ('\u{1d74f}', '\u{1d74f}', ON), ('\u{1d750}', '\u{1d788}', L), + ('\u{1d789}', '\u{1d789}', ON), ('\u{1d78a}', '\u{1d7c2}', L), ('\u{1d7c3}', '\u{1d7c3}', ON), + ('\u{1d7c4}', '\u{1d7cb}', L), ('\u{1d7ce}', '\u{1d7ff}', EN), ('\u{1d800}', '\u{1d9ff}', L), + ('\u{1da00}', '\u{1da36}', NSM), ('\u{1da37}', '\u{1da3a}', L), ('\u{1da3b}', '\u{1da6c}', NSM), + ('\u{1da6d}', '\u{1da74}', L), ('\u{1da75}', '\u{1da75}', NSM), ('\u{1da76}', '\u{1da83}', L), + ('\u{1da84}', '\u{1da84}', NSM), ('\u{1da85}', '\u{1da8b}', L), ('\u{1da9b}', '\u{1da9f}', NSM), + ('\u{1daa1}', '\u{1daaf}', NSM), ('\u{1df00}', '\u{1df1e}', L), ('\u{1df25}', '\u{1df2a}', L), + ('\u{1e000}', '\u{1e006}', NSM), ('\u{1e008}', '\u{1e018}', NSM), ('\u{1e01b}', '\u{1e021}', + NSM), ('\u{1e023}', '\u{1e024}', NSM), ('\u{1e026}', '\u{1e02a}', NSM), ('\u{1e030}', + '\u{1e06d}', L), ('\u{1e08f}', '\u{1e08f}', NSM), ('\u{1e100}', '\u{1e12c}', L), ('\u{1e130}', + '\u{1e136}', NSM), ('\u{1e137}', '\u{1e13d}', L), ('\u{1e140}', '\u{1e149}', L), ('\u{1e14e}', + '\u{1e14f}', L), ('\u{1e290}', '\u{1e2ad}', L), ('\u{1e2ae}', '\u{1e2ae}', NSM), ('\u{1e2c0}', + '\u{1e2eb}', L), ('\u{1e2ec}', '\u{1e2ef}', NSM), ('\u{1e2f0}', '\u{1e2f9}', L), ('\u{1e2ff}', + '\u{1e2ff}', ET), ('\u{1e4d0}', '\u{1e4eb}', L), ('\u{1e4ec}', '\u{1e4ef}', NSM), ('\u{1e4f0}', + '\u{1e4f9}', L), ('\u{1e7e0}', '\u{1e7e6}', L), ('\u{1e7e8}', '\u{1e7eb}', L), ('\u{1e7ed}', + '\u{1e7ee}', L), ('\u{1e7f0}', '\u{1e7fe}', L), ('\u{1e800}', '\u{1e8cf}', R), ('\u{1e8d0}', + '\u{1e8d6}', NSM), ('\u{1e8d7}', '\u{1e943}', R), ('\u{1e944}', '\u{1e94a}', NSM), ('\u{1e94b}', + '\u{1ec70}', R), ('\u{1ec71}', '\u{1ecb4}', AL), ('\u{1ecb5}', '\u{1ed00}', R), ('\u{1ed01}', + '\u{1ed3d}', AL), ('\u{1ed3e}', '\u{1edff}', R), ('\u{1ee00}', '\u{1eeef}', AL), ('\u{1eef0}', + '\u{1eef1}', ON), ('\u{1eef2}', '\u{1eeff}', AL), ('\u{1ef00}', '\u{1efff}', R), ('\u{1f000}', + '\u{1f02b}', ON), ('\u{1f030}', '\u{1f093}', ON), ('\u{1f0a0}', '\u{1f0ae}', ON), ('\u{1f0b1}', + '\u{1f0bf}', ON), ('\u{1f0c1}', '\u{1f0cf}', ON), ('\u{1f0d1}', '\u{1f0f5}', ON), ('\u{1f100}', + '\u{1f10a}', EN), ('\u{1f10b}', '\u{1f10f}', ON), ('\u{1f110}', '\u{1f12e}', L), ('\u{1f12f}', + '\u{1f12f}', ON), ('\u{1f130}', '\u{1f169}', L), ('\u{1f16a}', '\u{1f16f}', ON), ('\u{1f170}', + '\u{1f1ac}', L), ('\u{1f1ad}', '\u{1f1ad}', ON), ('\u{1f1e6}', '\u{1f202}', L), ('\u{1f210}', + '\u{1f23b}', L), ('\u{1f240}', '\u{1f248}', L), ('\u{1f250}', '\u{1f251}', L), ('\u{1f260}', + '\u{1f265}', ON), ('\u{1f300}', '\u{1f6d7}', ON), ('\u{1f6dc}', '\u{1f6ec}', ON), ('\u{1f6f0}', + '\u{1f6fc}', ON), ('\u{1f700}', '\u{1f776}', ON), ('\u{1f77b}', '\u{1f7d9}', ON), ('\u{1f7e0}', + '\u{1f7eb}', ON), ('\u{1f7f0}', '\u{1f7f0}', ON), ('\u{1f800}', '\u{1f80b}', ON), ('\u{1f810}', + '\u{1f847}', ON), ('\u{1f850}', '\u{1f859}', ON), ('\u{1f860}', '\u{1f887}', ON), ('\u{1f890}', + '\u{1f8ad}', ON), ('\u{1f8b0}', '\u{1f8b1}', ON), ('\u{1f900}', '\u{1fa53}', ON), ('\u{1fa60}', + '\u{1fa6d}', ON), ('\u{1fa70}', '\u{1fa7c}', ON), ('\u{1fa80}', '\u{1fa88}', ON), ('\u{1fa90}', + '\u{1fabd}', ON), ('\u{1fabf}', '\u{1fac5}', ON), ('\u{1face}', '\u{1fadb}', ON), ('\u{1fae0}', + '\u{1fae8}', ON), ('\u{1faf0}', '\u{1faf8}', ON), ('\u{1fb00}', '\u{1fb92}', ON), ('\u{1fb94}', + '\u{1fbca}', ON), ('\u{1fbf0}', '\u{1fbf9}', EN), ('\u{20000}', '\u{2a6df}', L), ('\u{2a700}', + '\u{2b739}', L), ('\u{2b740}', '\u{2b81d}', L), ('\u{2b820}', '\u{2cea1}', L), ('\u{2ceb0}', + '\u{2ebe0}', L), ('\u{2f800}', '\u{2fa1d}', L), ('\u{30000}', '\u{3134a}', L), ('\u{31350}', + '\u{323af}', L), ('\u{e0001}', '\u{e0001}', BN), ('\u{e0020}', '\u{e007f}', BN), ('\u{e0100}', + '\u{e01ef}', NSM), ('\u{f0000}', '\u{ffffd}', L), ('\u{100000}', '\u{10fffd}', L) +]; + +pub const bidi_pairs_table: &'static [(char, char, Option<char>)] = &[ + ('\u{28}', '\u{29}', None), ('\u{5b}', '\u{5d}', None), ('\u{7b}', '\u{7d}', None), ('\u{f3a}', + '\u{f3b}', None), ('\u{f3c}', '\u{f3d}', None), ('\u{169b}', '\u{169c}', None), ('\u{2045}', + '\u{2046}', None), ('\u{207d}', '\u{207e}', None), ('\u{208d}', '\u{208e}', None), ('\u{2308}', + '\u{2309}', None), ('\u{230a}', '\u{230b}', None), ('\u{2329}', '\u{232a}', Some('\u{3008}')), + ('\u{2768}', '\u{2769}', None), ('\u{276a}', '\u{276b}', None), ('\u{276c}', '\u{276d}', None), + ('\u{276e}', '\u{276f}', None), ('\u{2770}', '\u{2771}', None), ('\u{2772}', '\u{2773}', None), + ('\u{2774}', '\u{2775}', None), ('\u{27c5}', '\u{27c6}', None), ('\u{27e6}', '\u{27e7}', None), + ('\u{27e8}', '\u{27e9}', None), ('\u{27ea}', '\u{27eb}', None), ('\u{27ec}', '\u{27ed}', None), + ('\u{27ee}', '\u{27ef}', None), ('\u{2983}', '\u{2984}', None), ('\u{2985}', '\u{2986}', None), + ('\u{2987}', '\u{2988}', None), ('\u{2989}', '\u{298a}', None), ('\u{298b}', '\u{298c}', None), + ('\u{298d}', '\u{2990}', None), ('\u{298f}', '\u{298e}', None), ('\u{2991}', '\u{2992}', None), + ('\u{2993}', '\u{2994}', None), ('\u{2995}', '\u{2996}', None), ('\u{2997}', '\u{2998}', None), + ('\u{29d8}', '\u{29d9}', None), ('\u{29da}', '\u{29db}', None), ('\u{29fc}', '\u{29fd}', None), + ('\u{2e22}', '\u{2e23}', None), ('\u{2e24}', '\u{2e25}', None), ('\u{2e26}', '\u{2e27}', None), + ('\u{2e28}', '\u{2e29}', None), ('\u{2e55}', '\u{2e56}', None), ('\u{2e57}', '\u{2e58}', None), + ('\u{2e59}', '\u{2e5a}', None), ('\u{2e5b}', '\u{2e5c}', None), ('\u{3008}', '\u{3009}', None), + ('\u{300a}', '\u{300b}', None), ('\u{300c}', '\u{300d}', None), ('\u{300e}', '\u{300f}', None), + ('\u{3010}', '\u{3011}', None), ('\u{3014}', '\u{3015}', None), ('\u{3016}', '\u{3017}', None), + ('\u{3018}', '\u{3019}', None), ('\u{301a}', '\u{301b}', None), ('\u{fe59}', '\u{fe5a}', None), + ('\u{fe5b}', '\u{fe5c}', None), ('\u{fe5d}', '\u{fe5e}', None), ('\u{ff08}', '\u{ff09}', None), + ('\u{ff3b}', '\u{ff3d}', None), ('\u{ff5b}', '\u{ff5d}', None), ('\u{ff5f}', '\u{ff60}', None), + ('\u{ff62}', '\u{ff63}', None) +]; + diff --git a/third_party/rust/unicode-bidi/src/data_source.rs b/third_party/rust/unicode-bidi/src/data_source.rs new file mode 100644 index 0000000000..319ad53b0c --- /dev/null +++ b/third_party/rust/unicode-bidi/src/data_source.rs @@ -0,0 +1,46 @@ +// Copyright 2015 The Servo Project Developers. See the +// COPYRIGHT file at the top-level directory of this distribution. +// +// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or +// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license +// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your +// option. This file may not be copied, modified, or distributed +// except according to those terms. + +use crate::BidiClass; + +/// This is the return value of [`BidiDataSource::bidi_matched_opening_bracket()`]. +/// +/// It represents the matching *normalized* opening bracket for a given bracket in a bracket pair, +/// and whether or not that bracket is opening. +#[derive(Debug, Copy, Clone)] +pub struct BidiMatchedOpeningBracket { + /// The corresponding opening bracket in this bracket pair, normalized + /// + /// In case of opening brackets, this will be the bracket itself, except for when the bracket + /// is not normalized, in which case it will be the normalized form. + pub opening: char, + /// Whether or not the requested bracket was an opening bracket. True for opening + pub is_open: bool, +} + +/// This trait abstracts over a data source that is able to produce the Unicode Bidi class for a given +/// character +pub trait BidiDataSource { + fn bidi_class(&self, c: char) -> BidiClass; + /// If this character is a bracket according to BidiBrackets.txt, + /// return the corresponding *normalized* *opening bracket* of the pair, + /// and whether or not it itself is an opening bracket. + /// + /// This effectively buckets brackets into equivalence classes keyed on the + /// normalized opening bracket. + /// + /// The default implementation will pull in a small amount of hardcoded data, + /// regardless of the `hardcoded-data` feature. This is in part for convenience + /// (since this data is small and changes less often), and in part so that this method can be + /// added without needing a breaking version bump. + /// Override this method in your custom data source to prevent the use of hardcoded data. + fn bidi_matched_opening_bracket(&self, c: char) -> Option<BidiMatchedOpeningBracket> { + crate::char_data::bidi_matched_opening_bracket(c) + } +} diff --git a/third_party/rust/unicode-bidi/src/deprecated.rs b/third_party/rust/unicode-bidi/src/deprecated.rs new file mode 100644 index 0000000000..74a24f5b8b --- /dev/null +++ b/third_party/rust/unicode-bidi/src/deprecated.rs @@ -0,0 +1,92 @@ +// Copyright 2015 The Servo Project Developers. See the +// COPYRIGHT file at the top-level directory of this distribution. +// +// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or +// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license +// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your +// option. This file may not be copied, modified, or distributed +// except according to those terms. + +//! This module holds deprecated assets only. + +use alloc::vec::Vec; + +use super::*; + +/// Find the level runs within a line and return them in visual order. +/// +/// NOTE: This implementation is incomplete. The algorithm needs information about the text, +/// including original `BidiClass` property of each character, to be able to perform correctly. +/// Please see [`BidiInfo::visual_runs()`](../struct.BidiInfo.html#method.visual_runs) for the +/// improved implementation. +/// +/// `line` is a range of bytes indices within `levels`. +/// +/// <http://www.unicode.org/reports/tr9/#Reordering_Resolved_Levels> +#[deprecated( + since = "0.3.0", + note = "please use `BidiInfo::visual_runs()` instead." +)] +pub fn visual_runs(line: Range<usize>, levels: &[Level]) -> Vec<LevelRun> { + assert!(line.start <= levels.len()); + assert!(line.end <= levels.len()); + + let mut runs = Vec::new(); + + // Find consecutive level runs. + let mut start = line.start; + let mut run_level = levels[start]; + let mut min_level = run_level; + let mut max_level = run_level; + + for (i, &new_level) in levels.iter().enumerate().take(line.end).skip(start + 1) { + if new_level != run_level { + // End of the previous run, start of a new one. + runs.push(start..i); + start = i; + run_level = new_level; + + min_level = cmp::min(run_level, min_level); + max_level = cmp::max(run_level, max_level); + } + } + runs.push(start..line.end); + + let run_count = runs.len(); + + // Re-order the odd runs. + // <http://www.unicode.org/reports/tr9/#L2> + + // Stop at the lowest *odd* level. + min_level = min_level.new_lowest_ge_rtl().expect("Level error"); + + while max_level >= min_level { + // Look for the start of a sequence of consecutive runs of max_level or higher. + let mut seq_start = 0; + while seq_start < run_count { + if levels[runs[seq_start].start] < max_level { + seq_start += 1; + continue; + } + + // Found the start of a sequence. Now find the end. + let mut seq_end = seq_start + 1; + while seq_end < run_count { + if levels[runs[seq_end].start] < max_level { + break; + } + seq_end += 1; + } + + // Reverse the runs within this sequence. + runs[seq_start..seq_end].reverse(); + + seq_start = seq_end; + } + max_level + .lower(1) + .expect("Lowering embedding level below zero"); + } + + runs +} diff --git a/third_party/rust/unicode-bidi/src/explicit.rs b/third_party/rust/unicode-bidi/src/explicit.rs new file mode 100644 index 0000000000..d4ad897b54 --- /dev/null +++ b/third_party/rust/unicode-bidi/src/explicit.rs @@ -0,0 +1,210 @@ +// Copyright 2015 The Servo Project Developers. See the +// COPYRIGHT file at the top-level directory of this distribution. +// +// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or +// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license +// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your +// option. This file may not be copied, modified, or distributed +// except according to those terms. + +//! 3.3.2 Explicit Levels and Directions +//! +//! <http://www.unicode.org/reports/tr9/#Explicit_Levels_and_Directions> + +use alloc::vec::Vec; + +use super::char_data::{ + is_rtl, + BidiClass::{self, *}, +}; +use super::level::Level; +use super::TextSource; + +/// Compute explicit embedding levels for one paragraph of text (X1-X8). +/// +/// `processing_classes[i]` must contain the `BidiClass` of the char at byte index `i`, +/// for each char in `text`. +#[cfg_attr(feature = "flame_it", flamer::flame)] +pub fn compute<'a, T: TextSource<'a> + ?Sized>( + text: &'a T, + para_level: Level, + original_classes: &[BidiClass], + levels: &mut [Level], + processing_classes: &mut [BidiClass], +) { + assert_eq!(text.len(), original_classes.len()); + + // <http://www.unicode.org/reports/tr9/#X1> + let mut stack = DirectionalStatusStack::new(); + stack.push(para_level, OverrideStatus::Neutral); + + let mut overflow_isolate_count = 0u32; + let mut overflow_embedding_count = 0u32; + let mut valid_isolate_count = 0u32; + + for (i, len) in text.indices_lengths() { + match original_classes[i] { + // Rules X2-X5c + RLE | LRE | RLO | LRO | RLI | LRI | FSI => { + let last_level = stack.last().level; + + // <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + levels[i] = last_level; + + // X5a-X5c: Isolate initiators get the level of the last entry on the stack. + let is_isolate = match original_classes[i] { + RLI | LRI | FSI => true, + _ => false, + }; + if is_isolate { + // Redundant due to "Retaining explicit formatting characters" step. + // levels[i] = last_level; + match stack.last().status { + OverrideStatus::RTL => processing_classes[i] = R, + OverrideStatus::LTR => processing_classes[i] = L, + _ => {} + } + } + + let new_level = if is_rtl(original_classes[i]) { + last_level.new_explicit_next_rtl() + } else { + last_level.new_explicit_next_ltr() + }; + if new_level.is_ok() && overflow_isolate_count == 0 && overflow_embedding_count == 0 + { + let new_level = new_level.unwrap(); + stack.push( + new_level, + match original_classes[i] { + RLO => OverrideStatus::RTL, + LRO => OverrideStatus::LTR, + RLI | LRI | FSI => OverrideStatus::Isolate, + _ => OverrideStatus::Neutral, + }, + ); + if is_isolate { + valid_isolate_count += 1; + } else { + // The spec doesn't explicitly mention this step, but it is necessary. + // See the reference implementations for comparison. + levels[i] = new_level; + } + } else if is_isolate { + overflow_isolate_count += 1; + } else if overflow_isolate_count == 0 { + overflow_embedding_count += 1; + } + + if !is_isolate { + // X9 + + // <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + // (PDF handled below) + processing_classes[i] = BN; + } + } + + // <http://www.unicode.org/reports/tr9/#X6a> + PDI => { + if overflow_isolate_count > 0 { + overflow_isolate_count -= 1; + } else if valid_isolate_count > 0 { + overflow_embedding_count = 0; + loop { + // Pop everything up to and including the last Isolate status. + match stack.vec.pop() { + None + | Some(Status { + status: OverrideStatus::Isolate, + .. + }) => break, + _ => continue, + } + } + valid_isolate_count -= 1; + } + let last = stack.last(); + levels[i] = last.level; + match last.status { + OverrideStatus::RTL => processing_classes[i] = R, + OverrideStatus::LTR => processing_classes[i] = L, + _ => {} + } + } + + // <http://www.unicode.org/reports/tr9/#X7> + PDF => { + if overflow_isolate_count > 0 { + // do nothing + } else if overflow_embedding_count > 0 { + overflow_embedding_count -= 1; + } else if stack.last().status != OverrideStatus::Isolate && stack.vec.len() >= 2 { + stack.vec.pop(); + } + // <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + levels[i] = stack.last().level; + // X9 part of retaining explicit formatting characters. + processing_classes[i] = BN; + } + + // Nothing. + // BN case moved down to X6, see <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + B => {} + + // <http://www.unicode.org/reports/tr9/#X6> + _ => { + let last = stack.last(); + levels[i] = last.level; + // This condition is not in the spec, but I am pretty sure that is a spec bug. + // https://www.unicode.org/L2/L2023/23014-amd-to-uax9.pdf + if original_classes[i] != BN { + match last.status { + OverrideStatus::RTL => processing_classes[i] = R, + OverrideStatus::LTR => processing_classes[i] = L, + _ => {} + } + } + } + } + + // Handle multi-byte characters. + for j in 1..len { + levels[i + j] = levels[i]; + processing_classes[i + j] = processing_classes[i]; + } + } +} + +/// Entries in the directional status stack: +struct Status { + level: Level, + status: OverrideStatus, +} + +#[derive(PartialEq)] +enum OverrideStatus { + Neutral, + RTL, + LTR, + Isolate, +} + +struct DirectionalStatusStack { + vec: Vec<Status>, +} + +impl DirectionalStatusStack { + fn new() -> Self { + DirectionalStatusStack { + vec: Vec::with_capacity(Level::max_explicit_depth() as usize + 2), + } + } + + fn push(&mut self, level: Level, status: OverrideStatus) { + self.vec.push(Status { level, status }); + } + + fn last(&self) -> &Status { + self.vec.last().unwrap() + } +} diff --git a/third_party/rust/unicode-bidi/src/format_chars.rs b/third_party/rust/unicode-bidi/src/format_chars.rs new file mode 100644 index 0000000000..5b19b5e760 --- /dev/null +++ b/third_party/rust/unicode-bidi/src/format_chars.rs @@ -0,0 +1,42 @@ +// Copyright 2017 The Servo Project Developers. See the +// COPYRIGHT file at the top-level directory of this distribution. +// +// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or +// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license +// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your +// option. This file may not be copied, modified, or distributed +// except according to those terms. + +//! Directional Formatting Characters +//! +//! <http://www.unicode.org/reports/tr9/#Directional_Formatting_Characters> + +// == Implicit == +/// ARABIC LETTER MARK +pub const ALM: char = '\u{061C}'; +/// LEFT-TO-RIGHT MARK +pub const LRM: char = '\u{200E}'; +/// RIGHT-TO-LEFT MARK +pub const RLM: char = '\u{200F}'; + +// == Explicit Isolates == +/// LEFT‑TO‑RIGHT ISOLATE +pub const LRI: char = '\u{2066}'; +/// RIGHT‑TO‑LEFT ISOLATE +pub const RLI: char = '\u{2067}'; +/// FIRST STRONG ISOLATE +pub const FSI: char = '\u{2068}'; +/// POP DIRECTIONAL ISOLATE +pub const PDI: char = '\u{2069}'; + +// == Explicit Embeddings and Overrides == +/// LEFT-TO-RIGHT EMBEDDING +pub const LRE: char = '\u{202A}'; +/// RIGHT-TO-LEFT EMBEDDING +pub const RLE: char = '\u{202B}'; +/// POP DIRECTIONAL FORMATTING +pub const PDF: char = '\u{202C}'; +/// LEFT-TO-RIGHT OVERRIDE +pub const LRO: char = '\u{202D}'; +/// RIGHT-TO-LEFT OVERRIDE +pub const RLO: char = '\u{202E}'; diff --git a/third_party/rust/unicode-bidi/src/implicit.rs b/third_party/rust/unicode-bidi/src/implicit.rs new file mode 100644 index 0000000000..0311053c0a --- /dev/null +++ b/third_party/rust/unicode-bidi/src/implicit.rs @@ -0,0 +1,585 @@ +// Copyright 2015 The Servo Project Developers. See the +// COPYRIGHT file at the top-level directory of this distribution. +// +// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or +// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license +// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your +// option. This file may not be copied, modified, or distributed +// except according to those terms. + +//! 3.3.4 - 3.3.6. Resolve implicit levels and types. + +use alloc::vec::Vec; +use core::cmp::max; + +use super::char_data::BidiClass::{self, *}; +use super::level::Level; +use super::prepare::{not_removed_by_x9, IsolatingRunSequence}; +use super::{BidiDataSource, TextSource}; + +/// 3.3.4 Resolving Weak Types +/// +/// <http://www.unicode.org/reports/tr9/#Resolving_Weak_Types> +#[cfg_attr(feature = "flame_it", flamer::flame)] +pub fn resolve_weak<'a, T: TextSource<'a> + ?Sized>( + text: &'a T, + sequence: &IsolatingRunSequence, + processing_classes: &mut [BidiClass], +) { + // Note: The spec treats these steps as individual passes that are applied one after the other + // on the entire IsolatingRunSequence at once. We instead collapse it into a single iteration, + // which is straightforward for rules that are based on the state of the current character, but not + // for rules that care about surrounding characters. To deal with them, we retain additional state + // about previous character classes that may have since been changed by later rules. + + // The previous class for the purposes of rule W4/W6, not tracking changes made after or during W4. + let mut prev_class_before_w4 = sequence.sos; + // The previous class for the purposes of rule W5. + let mut prev_class_before_w5 = sequence.sos; + // The previous class for the purposes of rule W1, not tracking changes from any other rules. + let mut prev_class_before_w1 = sequence.sos; + let mut last_strong_is_al = false; + let mut et_run_indices = Vec::new(); // for W5 + let mut bn_run_indices = Vec::new(); // for W5 + <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + + for (run_index, level_run) in sequence.runs.iter().enumerate() { + for i in &mut level_run.clone() { + if processing_classes[i] == BN { + // <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + // Keeps track of bn runs for W5 in case we see an ET. + bn_run_indices.push(i); + // BNs aren't real, skip over them. + continue; + } + + // Store the processing class of all rules before W2/W1. + // Used to keep track of the last strong character for W2. W3 is able to insert new strong + // characters, so we don't want to be misled by it. + let mut w2_processing_class = processing_classes[i]; + + // <http://www.unicode.org/reports/tr9/#W1> + // + + if processing_classes[i] == NSM { + processing_classes[i] = match prev_class_before_w1 { + RLI | LRI | FSI | PDI => ON, + _ => prev_class_before_w1, + }; + // W1 occurs before W2, update this. + w2_processing_class = processing_classes[i]; + } + + prev_class_before_w1 = processing_classes[i]; + + // <http://www.unicode.org/reports/tr9/#W2> + // <http://www.unicode.org/reports/tr9/#W3> + // + match processing_classes[i] { + EN => { + if last_strong_is_al { + // W2. If previous strong char was AL, change EN to AN. + processing_classes[i] = AN; + } + } + // W3. + AL => processing_classes[i] = R, + _ => {} + } + + // update last_strong_is_al. + match w2_processing_class { + L | R => { + last_strong_is_al = false; + } + AL => { + last_strong_is_al = true; + } + _ => {} + } + + let class_before_w456 = processing_classes[i]; + + // <http://www.unicode.org/reports/tr9/#W4> + // <http://www.unicode.org/reports/tr9/#W5> + // <http://www.unicode.org/reports/tr9/#W6> (separators only) + // (see below for W6 terminator code) + // + match processing_classes[i] { + // <http://www.unicode.org/reports/tr9/#W6> + EN => { + // W5. If a run of ETs is adjacent to an EN, change the ETs to EN. + for j in &et_run_indices { + processing_classes[*j] = EN; + } + et_run_indices.clear(); + } + + // <http://www.unicode.org/reports/tr9/#W4> + // <http://www.unicode.org/reports/tr9/#W6> + ES | CS => { + // See https://github.com/servo/unicode-bidi/issues/86 for improving this. + // We want to make sure we check the correct next character by skipping past the rest + // of this one. + if let Some((_, char_len)) = text.char_at(i) { + let mut next_class = sequence + .iter_forwards_from(i + char_len, run_index) + .map(|j| processing_classes[j]) + // <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + .find(not_removed_by_x9) + .unwrap_or(sequence.eos); + if next_class == EN && last_strong_is_al { + // Apply W2 to next_class. We know that last_strong_is_al + // has no chance of changing on this character so we can still assume its value + // will be the same by the time we get to it. + next_class = AN; + } + processing_classes[i] = + match (prev_class_before_w4, processing_classes[i], next_class) { + // W4 + (EN, ES, EN) | (EN, CS, EN) => EN, + // W4 + (AN, CS, AN) => AN, + // W6 (separators only) + (_, _, _) => ON, + }; + + // W6 + <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + // We have to do this before W5 gets its grubby hands on these characters and thinks + // they're part of an ET run. + // We check for ON to ensure that we had hit the W6 branch above, since this `ES | CS` match + // arm handles both W4 and W6. + if processing_classes[i] == ON { + for idx in sequence.iter_backwards_from(i, run_index) { + let class = &mut processing_classes[idx]; + if *class != BN { + break; + } + *class = ON; + } + for idx in sequence.iter_forwards_from(i + char_len, run_index) { + let class = &mut processing_classes[idx]; + if *class != BN { + break; + } + *class = ON; + } + } + } else { + // We're in the middle of a character, copy over work done for previous bytes + // since it's going to be the same answer. + processing_classes[i] = processing_classes[i - 1]; + } + } + // <http://www.unicode.org/reports/tr9/#W5> + ET => { + match prev_class_before_w5 { + EN => processing_classes[i] = EN, + _ => { + // <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + // If there was a BN run before this, that's now a part of this ET run. + et_run_indices.extend(&bn_run_indices); + + // In case this is followed by an EN. + et_run_indices.push(i); + } + } + } + _ => {} + } + + // Common loop iteration code + // + + // <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + // BN runs would have already continued the loop, clear them before we get to the next one. + bn_run_indices.clear(); + + // W6 above only deals with separators, so it doesn't change anything W5 cares about, + // so we still can update this after running that part of W6. + prev_class_before_w5 = processing_classes[i]; + + // <http://www.unicode.org/reports/tr9/#W6> (terminators only) + // (see above for W6 separator code) + // + if prev_class_before_w5 != ET { + // W6. If we didn't find an adjacent EN, turn any ETs into ON instead. + for j in &et_run_indices { + processing_classes[*j] = ON; + } + et_run_indices.clear(); + } + + // We stashed this before W4/5/6 could get their grubby hands on it, and it's not + // used in the W6 terminator code below so we can update it now. + prev_class_before_w4 = class_before_w456; + } + } + // Rerun this check in case we ended with a sequence of BNs (i.e., we'd never + // hit the end of the for loop above). + // W6. If we didn't find an adjacent EN, turn any ETs into ON instead. + for j in &et_run_indices { + processing_classes[*j] = ON; + } + et_run_indices.clear(); + + // W7. If the previous strong char was L, change EN to L. + let mut last_strong_is_l = sequence.sos == L; + for run in &sequence.runs { + for i in run.clone() { + match processing_classes[i] { + EN if last_strong_is_l => { + processing_classes[i] = L; + } + L => { + last_strong_is_l = true; + } + R | AL => { + last_strong_is_l = false; + } + // <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + // Already scanning past BN here. + _ => {} + } + } + } +} + +/// 3.3.5 Resolving Neutral Types +/// +/// <http://www.unicode.org/reports/tr9/#Resolving_Neutral_Types> +#[cfg_attr(feature = "flame_it", flamer::flame)] +pub fn resolve_neutral<'a, D: BidiDataSource, T: TextSource<'a> + ?Sized>( + text: &'a T, + data_source: &D, + sequence: &IsolatingRunSequence, + levels: &[Level], + original_classes: &[BidiClass], + processing_classes: &mut [BidiClass], +) { + // e = embedding direction + let e: BidiClass = levels[sequence.runs[0].start].bidi_class(); + let not_e = if e == BidiClass::L { + BidiClass::R + } else { + BidiClass::L + }; + // N0. Process bracket pairs. + + // > Identify the bracket pairs in the current isolating run sequence according to BD16. + // We use processing_classes, not original_classes, due to BD14/BD15 + let bracket_pairs = identify_bracket_pairs(text, data_source, sequence, processing_classes); + + // > For each bracket-pair element in the list of pairs of text positions + // + // Note: Rust ranges are interpreted as [start..end), be careful using `pair` directly + // for indexing as it will include the opening bracket pair but not the closing one. + for pair in bracket_pairs { + #[cfg(feature = "std")] + debug_assert!( + pair.start < processing_classes.len(), + "identify_bracket_pairs returned a range that is out of bounds!" + ); + #[cfg(feature = "std")] + debug_assert!( + pair.end < processing_classes.len(), + "identify_bracket_pairs returned a range that is out of bounds!" + ); + let mut found_e = false; + let mut found_not_e = false; + let mut class_to_set = None; + + let start_char_len = + T::char_len(text.subrange(pair.start..pair.end).chars().next().unwrap()); + // > Inspect the bidirectional types of the characters enclosed within the bracket pair. + // + // `pair` is [start, end) so we will end up processing the opening character but not the closing one. + // + for enclosed_i in sequence.iter_forwards_from(pair.start + start_char_len, pair.start_run) { + if enclosed_i >= pair.end { + #[cfg(feature = "std")] + debug_assert!( + enclosed_i == pair.end, + "If we skipped past this, the iterator is broken" + ); + break; + } + let class = processing_classes[enclosed_i]; + if class == e { + found_e = true; + } else if class == not_e { + found_not_e = true; + } else if class == BidiClass::EN || class == BidiClass::AN { + // > Within this scope, bidirectional types EN and AN are treated as R. + if e == BidiClass::L { + found_not_e = true; + } else { + found_e = true; + } + } + + // If we have found a character with the class of the embedding direction + // we can bail early. + if found_e { + break; + } + } + // > If any strong type (either L or R) matching the embedding direction is found + if found_e { + // > .. set the type for both brackets in the pair to match the embedding direction + class_to_set = Some(e); + // > Otherwise, if there is a strong type it must be opposite the embedding direction + } else if found_not_e { + // > Therefore, test for an established context with a preceding strong type by + // > checking backwards before the opening paired bracket + // > until the first strong type (L, R, or sos) is found. + // (see note above about processing_classes and character boundaries) + let mut previous_strong = sequence + .iter_backwards_from(pair.start, pair.start_run) + .map(|i| processing_classes[i]) + .find(|class| { + *class == BidiClass::L + || *class == BidiClass::R + || *class == BidiClass::EN + || *class == BidiClass::AN + }) + .unwrap_or(sequence.sos); + + // > Within this scope, bidirectional types EN and AN are treated as R. + if previous_strong == BidiClass::EN || previous_strong == BidiClass::AN { + previous_strong = BidiClass::R; + } + + // > If the preceding strong type is also opposite the embedding direction, + // > context is established, + // > so set the type for both brackets in the pair to that direction. + // AND + // > Otherwise set the type for both brackets in the pair to the embedding direction. + // > Either way it gets set to previous_strong + // + // Both branches amount to setting the type to the strong type. + class_to_set = Some(previous_strong); + } + + if let Some(class_to_set) = class_to_set { + // Update all processing classes corresponding to the start and end elements, as requested. + // We should include all bytes of the character, not the first one. + let end_char_len = + T::char_len(text.subrange(pair.end..text.len()).chars().next().unwrap()); + for class in &mut processing_classes[pair.start..pair.start + start_char_len] { + *class = class_to_set; + } + for class in &mut processing_classes[pair.end..pair.end + end_char_len] { + *class = class_to_set; + } + // <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + for idx in sequence.iter_backwards_from(pair.start, pair.start_run) { + let class = &mut processing_classes[idx]; + if *class != BN { + break; + } + *class = class_to_set; + } + // > Any number of characters that had original bidirectional character type NSM prior to the application of + // > W1 that immediately follow a paired bracket which changed to L or R under N0 should change to match the type of their preceding bracket. + + // This rule deals with sequences of NSMs, so we can just update them all at once, we don't need to worry + // about character boundaries. We do need to be careful to skip the full set of bytes for the parentheses characters. + let nsm_start = pair.start + start_char_len; + for idx in sequence.iter_forwards_from(nsm_start, pair.start_run) { + let class = original_classes[idx]; + if class == BidiClass::NSM || processing_classes[idx] == BN { + processing_classes[idx] = class_to_set; + } else { + break; + } + } + let nsm_end = pair.end + end_char_len; + for idx in sequence.iter_forwards_from(nsm_end, pair.end_run) { + let class = original_classes[idx]; + if class == BidiClass::NSM || processing_classes[idx] == BN { + processing_classes[idx] = class_to_set; + } else { + break; + } + } + } + // > Otherwise, there are no strong types within the bracket pair + // > Therefore, do not set the type for that bracket pair + } + + // N1 and N2. + // Indices of every byte in this isolating run sequence + let mut indices = sequence.runs.iter().flat_map(Clone::clone); + let mut prev_class = sequence.sos; + while let Some(mut i) = indices.next() { + // Process sequences of NI characters. + let mut ni_run = Vec::new(); + // The BN is for <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + if is_NI(processing_classes[i]) || processing_classes[i] == BN { + // Consume a run of consecutive NI characters. + ni_run.push(i); + let mut next_class; + loop { + match indices.next() { + Some(j) => { + i = j; + next_class = processing_classes[j]; + // The BN is for <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + if is_NI(next_class) || next_class == BN { + ni_run.push(i); + } else { + break; + } + } + None => { + next_class = sequence.eos; + break; + } + }; + } + // N1-N2. + // + // <http://www.unicode.org/reports/tr9/#N1> + // <http://www.unicode.org/reports/tr9/#N2> + let new_class = match (prev_class, next_class) { + (L, L) => L, + (R, R) + | (R, AN) + | (R, EN) + | (AN, R) + | (AN, AN) + | (AN, EN) + | (EN, R) + | (EN, AN) + | (EN, EN) => R, + (_, _) => e, + }; + for j in &ni_run { + processing_classes[*j] = new_class; + } + ni_run.clear(); + } + prev_class = processing_classes[i]; + } +} + +struct BracketPair { + /// The text-relative index of the opening bracket. + start: usize, + /// The text-relative index of the closing bracket. + end: usize, + /// The index of the run (in the run sequence) that the opening bracket is in. + start_run: usize, + /// The index of the run (in the run sequence) that the closing bracket is in. + end_run: usize, +} +/// 3.1.3 Identifying Bracket Pairs +/// +/// Returns all paired brackets in the source, as indices into the +/// text source. +/// +/// <https://www.unicode.org/reports/tr9/#BD16> +fn identify_bracket_pairs<'a, T: TextSource<'a> + ?Sized, D: BidiDataSource>( + text: &'a T, + data_source: &D, + run_sequence: &IsolatingRunSequence, + original_classes: &[BidiClass], +) -> Vec<BracketPair> { + let mut ret = vec![]; + let mut stack = vec![]; + + for (run_index, level_run) in run_sequence.runs.iter().enumerate() { + for (i, ch) in text.subrange(level_run.clone()).char_indices() { + let actual_index = level_run.start + i; + + // All paren characters are ON. + // From BidiBrackets.txt: + // > The Unicode property value stability policy guarantees that characters + // > which have bpt=o or bpt=c also have bc=ON and Bidi_M=Y + if original_classes[actual_index] != BidiClass::ON { + continue; + } + + if let Some(matched) = data_source.bidi_matched_opening_bracket(ch) { + if matched.is_open { + // > If an opening paired bracket is found ... + + // > ... and there is no room in the stack, + // > stop processing BD16 for the remainder of the isolating run sequence. + if stack.len() >= 63 { + break; + } + // > ... push its Bidi_Paired_Bracket property value and its text position onto the stack + stack.push((matched.opening, actual_index, run_index)) + } else { + // > If a closing paired bracket is found, do the following + + // > Declare a variable that holds a reference to the current stack element + // > and initialize it with the top element of the stack. + // AND + // > Else, if the current stack element is not at the bottom of the stack + for (stack_index, element) in stack.iter().enumerate().rev() { + // > Compare the closing paired bracket being inspected or its canonical + // > equivalent to the bracket in the current stack element. + if element.0 == matched.opening { + // > If the values match, meaning the two characters form a bracket pair, then + + // > Append the text position in the current stack element together with the + // > text position of the closing paired bracket to the list. + let pair = BracketPair { + start: element.1, + end: actual_index, + start_run: element.2, + end_run: run_index, + }; + ret.push(pair); + + // > Pop the stack through the current stack element inclusively. + stack.truncate(stack_index); + break; + } + } + } + } + } + } + // > Sort the list of pairs of text positions in ascending order based on + // > the text position of the opening paired bracket. + ret.sort_by_key(|r| r.start); + ret +} + +/// 3.3.6 Resolving Implicit Levels +/// +/// Returns the maximum embedding level in the paragraph. +/// +/// <http://www.unicode.org/reports/tr9/#Resolving_Implicit_Levels> +#[cfg_attr(feature = "flame_it", flamer::flame)] +pub fn resolve_levels(original_classes: &[BidiClass], levels: &mut [Level]) -> Level { + let mut max_level = Level::ltr(); + assert_eq!(original_classes.len(), levels.len()); + for i in 0..levels.len() { + match (levels[i].is_rtl(), original_classes[i]) { + (false, AN) | (false, EN) => levels[i].raise(2).expect("Level number error"), + (false, R) | (true, L) | (true, EN) | (true, AN) => { + levels[i].raise(1).expect("Level number error") + } + // <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> handled here + (_, _) => {} + } + max_level = max(max_level, levels[i]); + } + + max_level +} + +/// Neutral or Isolate formatting character (B, S, WS, ON, FSI, LRI, RLI, PDI) +/// +/// <http://www.unicode.org/reports/tr9/#NI> +#[allow(non_snake_case)] +fn is_NI(class: BidiClass) -> bool { + match class { + B | S | WS | ON | FSI | LRI | RLI | PDI => true, + _ => false, + } +} diff --git a/third_party/rust/unicode-bidi/src/level.rs b/third_party/rust/unicode-bidi/src/level.rs new file mode 100644 index 0000000000..ef4f6d9e40 --- /dev/null +++ b/third_party/rust/unicode-bidi/src/level.rs @@ -0,0 +1,400 @@ +// Copyright 2017 The Servo Project Developers. See the +// COPYRIGHT file at the top-level directory of this distribution. +// +// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or +// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license +// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your +// option. This file may not be copied, modified, or distributed +// except according to those terms. + +//! Bidi Embedding Level +//! +//! See [`Level`](struct.Level.html) for more details. +//! +//! <http://www.unicode.org/reports/tr9/#BD2> + +use alloc::string::{String, ToString}; +use alloc::vec::Vec; +use core::convert::{From, Into}; +use core::slice; + +use super::char_data::BidiClass; + +/// Embedding Level +/// +/// Embedding Levels are numbers between 0 and 126 (inclusive), where even values denote a +/// left-to-right (LTR) direction and odd values a right-to-left (RTL) direction. +/// +/// This struct maintains a *valid* status for level numbers, meaning that creating a new level, or +/// mutating an existing level, with the value smaller than `0` (before conversion to `u8`) or +/// larger than 125 results in an `Error`. +/// +/// <http://www.unicode.org/reports/tr9/#BD2> +#[derive(Copy, Clone, Debug, Eq, Ord, PartialEq, PartialOrd)] +#[cfg_attr(feature = "serde", derive(serde::Serialize, serde::Deserialize))] +#[repr(transparent)] +pub struct Level(u8); + +pub const LTR_LEVEL: Level = Level(0); +pub const RTL_LEVEL: Level = Level(1); + +const MAX_DEPTH: u8 = 125; +/// During explicit level resolution, embedding level can go as high as `max_depth`. +pub const MAX_EXPLICIT_DEPTH: u8 = MAX_DEPTH; +/// During implicit level resolution, embedding level can go as high as `max_depth + 1`. +pub const MAX_IMPLICIT_DEPTH: u8 = MAX_DEPTH + 1; + +/// Errors that can occur on Level creation or mutation +#[derive(Debug, PartialEq)] +pub enum Error { + /// Out-of-range (invalid) embedding level number. + OutOfRangeNumber, +} + +impl Level { + /// New LTR level with smallest number value (0). + #[inline] + pub fn ltr() -> Level { + LTR_LEVEL + } + + /// New RTL level with smallest number value (1). + #[inline] + pub fn rtl() -> Level { + RTL_LEVEL + } + + /// Maximum depth of the directional status stack during implicit resolutions. + pub fn max_implicit_depth() -> u8 { + MAX_IMPLICIT_DEPTH + } + + /// Maximum depth of the directional status stack during explicit resolutions. + pub fn max_explicit_depth() -> u8 { + MAX_EXPLICIT_DEPTH + } + + // == Inquiries == + + /// Create new level, fail if number is larger than `max_depth + 1`. + #[inline] + pub fn new(number: u8) -> Result<Level, Error> { + if number <= MAX_IMPLICIT_DEPTH { + Ok(Level(number)) + } else { + Err(Error::OutOfRangeNumber) + } + } + + /// Create new level, fail if number is larger than `max_depth`. + #[inline] + pub fn new_explicit(number: u8) -> Result<Level, Error> { + if number <= MAX_EXPLICIT_DEPTH { + Ok(Level(number)) + } else { + Err(Error::OutOfRangeNumber) + } + } + + // == Inquiries == + + /// The level number. + #[inline] + pub fn number(&self) -> u8 { + self.0 + } + + /// If this level is left-to-right. + #[inline] + pub fn is_ltr(&self) -> bool { + self.0 % 2 == 0 + } + + /// If this level is right-to-left. + #[inline] + pub fn is_rtl(&self) -> bool { + self.0 % 2 == 1 + } + + // == Mutators == + + /// Raise level by `amount`, fail if number is larger than `max_depth + 1`. + #[inline] + pub fn raise(&mut self, amount: u8) -> Result<(), Error> { + match self.0.checked_add(amount) { + Some(number) => { + if number <= MAX_IMPLICIT_DEPTH { + self.0 = number; + Ok(()) + } else { + Err(Error::OutOfRangeNumber) + } + } + None => Err(Error::OutOfRangeNumber), + } + } + + /// Raise level by `amount`, fail if number is larger than `max_depth`. + #[inline] + pub fn raise_explicit(&mut self, amount: u8) -> Result<(), Error> { + match self.0.checked_add(amount) { + Some(number) => { + if number <= MAX_EXPLICIT_DEPTH { + self.0 = number; + Ok(()) + } else { + Err(Error::OutOfRangeNumber) + } + } + None => Err(Error::OutOfRangeNumber), + } + } + + /// Lower level by `amount`, fail if number goes below zero. + #[inline] + pub fn lower(&mut self, amount: u8) -> Result<(), Error> { + match self.0.checked_sub(amount) { + Some(number) => { + self.0 = number; + Ok(()) + } + None => Err(Error::OutOfRangeNumber), + } + } + + // == Helpers == + + /// The next LTR (even) level greater than this, or fail if number is larger than `max_depth`. + #[inline] + pub fn new_explicit_next_ltr(&self) -> Result<Level, Error> { + Level::new_explicit((self.0 + 2) & !1) + } + + /// The next RTL (odd) level greater than this, or fail if number is larger than `max_depth`. + #[inline] + pub fn new_explicit_next_rtl(&self) -> Result<Level, Error> { + Level::new_explicit((self.0 + 1) | 1) + } + + /// The lowest RTL (odd) level greater than or equal to this, or fail if number is larger than + /// `max_depth + 1`. + #[inline] + pub fn new_lowest_ge_rtl(&self) -> Result<Level, Error> { + Level::new(self.0 | 1) + } + + /// Generate a character type based on a level (as specified in steps X10 and N2). + #[inline] + pub fn bidi_class(&self) -> BidiClass { + if self.is_rtl() { + BidiClass::R + } else { + BidiClass::L + } + } + + pub fn vec(v: &[u8]) -> Vec<Level> { + v.iter().map(|&x| x.into()).collect() + } + + /// Converts a byte slice to a slice of Levels + /// + /// Does _not_ check if each level is within bounds (`<=` [`MAX_IMPLICIT_DEPTH`]), + /// which is not a requirement for safety but is a requirement for correctness of the algorithm. + pub fn from_slice_unchecked(v: &[u8]) -> &[Level] { + debug_assert_eq!(core::mem::size_of::<u8>(), core::mem::size_of::<Level>()); + unsafe { + // Safety: The two arrays are the same size and layout-compatible since + // Level is `repr(transparent)` over `u8` + slice::from_raw_parts(v as *const [u8] as *const u8 as *const Level, v.len()) + } + } +} + +/// If levels has any RTL (odd) level +/// +/// This information is usually used to skip re-ordering of text when no RTL level is present +#[inline] +pub fn has_rtl(levels: &[Level]) -> bool { + levels.iter().any(|&lvl| lvl.is_rtl()) +} + +impl Into<u8> for Level { + /// Convert to the level number + #[inline] + fn into(self) -> u8 { + self.number() + } +} + +impl From<u8> for Level { + /// Create level by number + #[inline] + fn from(number: u8) -> Level { + Level::new(number).expect("Level number error") + } +} + +/// Used for matching levels in conformance tests +impl<'a> PartialEq<&'a str> for Level { + #[inline] + fn eq(&self, s: &&'a str) -> bool { + *s == "x" || *s == self.0.to_string() + } +} + +/// Used for matching levels in conformance tests +impl<'a> PartialEq<String> for Level { + #[inline] + fn eq(&self, s: &String) -> bool { + self == &s.as_str() + } +} + +#[cfg(test)] +mod tests { + use super::*; + + #[test] + fn test_new() { + assert_eq!(Level::new(0), Ok(Level(0))); + assert_eq!(Level::new(1), Ok(Level(1))); + assert_eq!(Level::new(10), Ok(Level(10))); + assert_eq!(Level::new(125), Ok(Level(125))); + assert_eq!(Level::new(126), Ok(Level(126))); + assert_eq!(Level::new(127), Err(Error::OutOfRangeNumber)); + assert_eq!(Level::new(255), Err(Error::OutOfRangeNumber)); + } + + #[test] + fn test_new_explicit() { + assert_eq!(Level::new_explicit(0), Ok(Level(0))); + assert_eq!(Level::new_explicit(1), Ok(Level(1))); + assert_eq!(Level::new_explicit(10), Ok(Level(10))); + assert_eq!(Level::new_explicit(125), Ok(Level(125))); + assert_eq!(Level::new_explicit(126), Err(Error::OutOfRangeNumber)); + assert_eq!(Level::new_explicit(255), Err(Error::OutOfRangeNumber)); + } + + #[test] + fn test_is_ltr() { + assert_eq!(Level(0).is_ltr(), true); + assert_eq!(Level(1).is_ltr(), false); + assert_eq!(Level(10).is_ltr(), true); + assert_eq!(Level(11).is_ltr(), false); + assert_eq!(Level(124).is_ltr(), true); + assert_eq!(Level(125).is_ltr(), false); + } + + #[test] + fn test_is_rtl() { + assert_eq!(Level(0).is_rtl(), false); + assert_eq!(Level(1).is_rtl(), true); + assert_eq!(Level(10).is_rtl(), false); + assert_eq!(Level(11).is_rtl(), true); + assert_eq!(Level(124).is_rtl(), false); + assert_eq!(Level(125).is_rtl(), true); + } + + #[test] + fn test_raise() { + let mut level = Level::ltr(); + assert_eq!(level.number(), 0); + assert!(level.raise(100).is_ok()); + assert_eq!(level.number(), 100); + assert!(level.raise(26).is_ok()); + assert_eq!(level.number(), 126); + assert!(level.raise(1).is_err()); // invalid! + assert!(level.raise(250).is_err()); // overflow! + assert_eq!(level.number(), 126); + } + + #[test] + fn test_raise_explicit() { + let mut level = Level::ltr(); + assert_eq!(level.number(), 0); + assert!(level.raise_explicit(100).is_ok()); + assert_eq!(level.number(), 100); + assert!(level.raise_explicit(25).is_ok()); + assert_eq!(level.number(), 125); + assert!(level.raise_explicit(1).is_err()); // invalid! + assert!(level.raise_explicit(250).is_err()); // overflow! + assert_eq!(level.number(), 125); + } + + #[test] + fn test_lower() { + let mut level = Level::rtl(); + assert_eq!(level.number(), 1); + assert!(level.lower(1).is_ok()); + assert_eq!(level.number(), 0); + assert!(level.lower(1).is_err()); // underflow! + assert!(level.lower(250).is_err()); // underflow! + assert_eq!(level.number(), 0); + } + + #[test] + fn test_has_rtl() { + assert_eq!(has_rtl(&Level::vec(&[0, 0, 0])), false); + assert_eq!(has_rtl(&Level::vec(&[0, 1, 0])), true); + assert_eq!(has_rtl(&Level::vec(&[0, 2, 0])), false); + assert_eq!(has_rtl(&Level::vec(&[0, 125, 0])), true); + assert_eq!(has_rtl(&Level::vec(&[0, 126, 0])), false); + } + + #[test] + fn test_into() { + let level = Level::rtl(); + let number: u8 = level.into(); + assert_eq!(1u8, number); + } + + #[test] + fn test_vec() { + assert_eq!( + Level::vec(&[0, 1, 125]), + vec![Level(0), Level(1), Level(125)] + ); + } + + #[test] + fn test_str_eq() { + assert_eq!(Level::vec(&[0, 1, 4, 125]), vec!["0", "1", "x", "125"]); + assert_ne!(Level::vec(&[0, 1, 4, 125]), vec!["0", "1", "5", "125"]); + } + + #[test] + fn test_string_eq() { + assert_eq!( + Level::vec(&[0, 1, 4, 125]), + vec!["0".to_string(), "1".to_string(), "x".to_string(), "125".to_string()] + ); + } +} + +#[cfg(all(feature = "serde", test))] +mod serde_tests { + use super::*; + use serde_test::{assert_tokens, Token}; + + #[test] + fn test_statics() { + assert_tokens( + &Level::ltr(), + &[Token::NewtypeStruct { name: "Level" }, Token::U8(0)], + ); + assert_tokens( + &Level::rtl(), + &[Token::NewtypeStruct { name: "Level" }, Token::U8(1)], + ); + } + + #[test] + fn test_new() { + let level = Level::new(42).unwrap(); + assert_tokens( + &level, + &[Token::NewtypeStruct { name: "Level" }, Token::U8(42)], + ); + } +} diff --git a/third_party/rust/unicode-bidi/src/lib.rs b/third_party/rust/unicode-bidi/src/lib.rs new file mode 100644 index 0000000000..1072b67fe0 --- /dev/null +++ b/third_party/rust/unicode-bidi/src/lib.rs @@ -0,0 +1,2242 @@ +// Copyright 2015 The Servo Project Developers. See the +// COPYRIGHT file at the top-level directory of this distribution. +// +// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or +// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license +// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your +// option. This file may not be copied, modified, or distributed +// except according to those terms. + +//! This crate implements the [Unicode Bidirectional Algorithm][tr9] for display of mixed +//! right-to-left and left-to-right text. It is written in safe Rust, compatible with the +//! current stable release. +//! +//! ## Example +//! +//! ```rust +//! # #[cfg(feature = "hardcoded-data")] { +//! use unicode_bidi::BidiInfo; +//! +//! // This example text is defined using `concat!` because some browsers +//! // and text editors have trouble displaying bidi strings. +//! let text = concat![ +//! "א", +//! "ב", +//! "ג", +//! "a", +//! "b", +//! "c", +//! ]; +//! +//! // Resolve embedding levels within the text. Pass `None` to detect the +//! // paragraph level automatically. +//! let bidi_info = BidiInfo::new(&text, None); +//! +//! // This paragraph has embedding level 1 because its first strong character is RTL. +//! assert_eq!(bidi_info.paragraphs.len(), 1); +//! let para = &bidi_info.paragraphs[0]; +//! assert_eq!(para.level.number(), 1); +//! assert_eq!(para.level.is_rtl(), true); +//! +//! // Re-ordering is done after wrapping each paragraph into a sequence of +//! // lines. For this example, I'll just use a single line that spans the +//! // entire paragraph. +//! let line = para.range.clone(); +//! +//! let display = bidi_info.reorder_line(para, line); +//! assert_eq!(display, concat![ +//! "a", +//! "b", +//! "c", +//! "ג", +//! "ב", +//! "א", +//! ]); +//! # } // feature = "hardcoded-data" +//! ``` +//! +//! # Features +//! +//! - `std`: Enabled by default, but can be disabled to make `unicode_bidi` +//! `#![no_std]` + `alloc` compatible. +//! - `hardcoded-data`: Enabled by default. Includes hardcoded Unicode bidi data and more convenient APIs. +//! - `serde`: Adds [`serde::Serialize`] and [`serde::Deserialize`] +//! implementations to relevant types. +//! +//! [tr9]: <http://www.unicode.org/reports/tr9/> + +#![no_std] +// We need to link to std to make doc tests work on older Rust versions +#[cfg(feature = "std")] +extern crate std; +#[macro_use] +extern crate alloc; + +pub mod data_source; +pub mod deprecated; +pub mod format_chars; +pub mod level; +pub mod utf16; + +mod char_data; +mod explicit; +mod implicit; +mod prepare; + +pub use crate::char_data::{BidiClass, UNICODE_VERSION}; +pub use crate::data_source::BidiDataSource; +pub use crate::level::{Level, LTR_LEVEL, RTL_LEVEL}; +pub use crate::prepare::LevelRun; + +#[cfg(feature = "hardcoded-data")] +pub use crate::char_data::{bidi_class, HardcodedBidiData}; + +use alloc::borrow::Cow; +use alloc::string::String; +use alloc::vec::Vec; +use core::char; +use core::cmp; +use core::iter::repeat; +use core::ops::Range; +use core::str::CharIndices; + +use crate::format_chars as chars; +use crate::BidiClass::*; + +/// Trait that abstracts over a text source for use by the bidi algorithms. +/// We implement this for str (UTF-8) and for [u16] (UTF-16, native-endian). +/// (For internal unicode-bidi use; API may be unstable.) +/// This trait is sealed and cannot be implemented for types outside this crate. +pub trait TextSource<'text>: private::Sealed { + type CharIter: Iterator<Item = char>; + type CharIndexIter: Iterator<Item = (usize, char)>; + type IndexLenIter: Iterator<Item = (usize, usize)>; + + /// Return the length of the text in code units. + #[doc(hidden)] + fn len(&self) -> usize; + + /// Get the character at a given code unit index, along with its length in code units. + /// Returns None if index is out of range, or points inside a multi-code-unit character. + /// Returns REPLACEMENT_CHARACTER for any unpaired surrogates in UTF-16. + #[doc(hidden)] + fn char_at(&self, index: usize) -> Option<(char, usize)>; + + /// Return a subrange of the text, indexed by code units. + /// (We don't implement all of the Index trait, just the minimum we use.) + #[doc(hidden)] + fn subrange(&self, range: Range<usize>) -> &Self; + + /// An iterator over the text returning Unicode characters, + /// REPLACEMENT_CHAR for invalid code units. + #[doc(hidden)] + fn chars(&'text self) -> Self::CharIter; + + /// An iterator over the text returning (index, char) tuples, + /// where index is the starting code-unit index of the character, + /// and char is its Unicode value (or REPLACEMENT_CHAR if invalid). + #[doc(hidden)] + fn char_indices(&'text self) -> Self::CharIndexIter; + + /// An iterator over the text returning (index, length) tuples, + /// where index is the starting code-unit index of the character, + /// and length is its length in code units. + #[doc(hidden)] + fn indices_lengths(&'text self) -> Self::IndexLenIter; + + /// Number of code units the given character uses. + #[doc(hidden)] + fn char_len(ch: char) -> usize; +} + +mod private { + pub trait Sealed {} + + // Implement for str and [u16] only. + impl Sealed for str {} + impl Sealed for [u16] {} +} + +#[derive(PartialEq, Debug)] +pub enum Direction { + Ltr, + Rtl, + Mixed, +} + +/// Bidi information about a single paragraph +#[derive(Clone, Debug, PartialEq)] +pub struct ParagraphInfo { + /// The paragraphs boundaries within the text, as byte indices. + /// + /// TODO: Shrink this to only include the starting index? + pub range: Range<usize>, + + /// The paragraph embedding level. + /// + /// <http://www.unicode.org/reports/tr9/#BD4> + pub level: Level, +} + +impl ParagraphInfo { + /// Gets the length of the paragraph in the source text. + pub fn len(&self) -> usize { + self.range.end - self.range.start + } +} + +/// Initial bidi information of the text. +/// +/// Contains the text paragraphs and `BidiClass` of its characters. +#[derive(PartialEq, Debug)] +pub struct InitialInfo<'text> { + /// The text + pub text: &'text str, + + /// The BidiClass of the character at each byte in the text. + /// If a character is multiple bytes, its class will appear multiple times in the vector. + pub original_classes: Vec<BidiClass>, + + /// The boundaries and level of each paragraph within the text. + pub paragraphs: Vec<ParagraphInfo>, +} + +impl<'text> InitialInfo<'text> { + /// Find the paragraphs and BidiClasses in a string of text. + /// + /// <http://www.unicode.org/reports/tr9/#The_Paragraph_Level> + /// + /// Also sets the class for each First Strong Isolate initiator (FSI) to LRI or RLI if a strong + /// character is found before the matching PDI. If no strong character is found, the class will + /// remain FSI, and it's up to later stages to treat these as LRI when needed. + /// + /// The `hardcoded-data` Cargo feature (enabled by default) must be enabled to use this. + #[cfg_attr(feature = "flame_it", flamer::flame)] + #[cfg(feature = "hardcoded-data")] + pub fn new(text: &str, default_para_level: Option<Level>) -> InitialInfo<'_> { + Self::new_with_data_source(&HardcodedBidiData, text, default_para_level) + } + + /// Find the paragraphs and BidiClasses in a string of text, with a custom [`BidiDataSource`] + /// for Bidi data. If you just wish to use the hardcoded Bidi data, please use [`InitialInfo::new()`] + /// instead (enabled with tbe default `hardcoded-data` Cargo feature) + /// + /// <http://www.unicode.org/reports/tr9/#The_Paragraph_Level> + /// + /// Also sets the class for each First Strong Isolate initiator (FSI) to LRI or RLI if a strong + /// character is found before the matching PDI. If no strong character is found, the class will + /// remain FSI, and it's up to later stages to treat these as LRI when needed. + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn new_with_data_source<'a, D: BidiDataSource>( + data_source: &D, + text: &'a str, + default_para_level: Option<Level>, + ) -> InitialInfo<'a> { + InitialInfoExt::new_with_data_source(data_source, text, default_para_level).base + } +} + +/// Extended version of InitialInfo (not public API). +#[derive(PartialEq, Debug)] +struct InitialInfoExt<'text> { + /// The base InitialInfo for the text, recording its paragraphs and bidi classes. + base: InitialInfo<'text>, + + /// Parallel to base.paragraphs, records whether each paragraph is "pure LTR" that + /// requires no further bidi processing (i.e. there are no RTL characters or bidi + /// control codes present). + pure_ltr: Vec<bool>, +} + +impl<'text> InitialInfoExt<'text> { + /// Find the paragraphs and BidiClasses in a string of text, with a custom [`BidiDataSource`] + /// for Bidi data. If you just wish to use the hardcoded Bidi data, please use [`InitialInfo::new()`] + /// instead (enabled with tbe default `hardcoded-data` Cargo feature) + /// + /// <http://www.unicode.org/reports/tr9/#The_Paragraph_Level> + /// + /// Also sets the class for each First Strong Isolate initiator (FSI) to LRI or RLI if a strong + /// character is found before the matching PDI. If no strong character is found, the class will + /// remain FSI, and it's up to later stages to treat these as LRI when needed. + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn new_with_data_source<'a, D: BidiDataSource>( + data_source: &D, + text: &'a str, + default_para_level: Option<Level>, + ) -> InitialInfoExt<'a> { + let mut paragraphs = Vec::<ParagraphInfo>::new(); + let mut pure_ltr = Vec::<bool>::new(); + let (original_classes, _, _) = compute_initial_info( + data_source, + text, + default_para_level, + Some((&mut paragraphs, &mut pure_ltr)), + ); + + InitialInfoExt { + base: InitialInfo { + text, + original_classes, + paragraphs, + }, + pure_ltr, + } + } +} + +/// Implementation of initial-info computation for both BidiInfo and ParagraphBidiInfo. +/// To treat the text as (potentially) multiple paragraphs, the caller should pass the +/// pair of optional outparam arrays to receive the ParagraphInfo and pure-ltr flags +/// for each paragraph. Passing None for split_paragraphs will ignore any paragraph- +/// separator characters in the text, treating it just as a single paragraph. +/// Returns the array of BidiClass values for each code unit of the text, along with +/// the embedding level and pure-ltr flag for the *last* (or only) paragraph. +fn compute_initial_info<'a, D: BidiDataSource, T: TextSource<'a> + ?Sized>( + data_source: &D, + text: &'a T, + default_para_level: Option<Level>, + mut split_paragraphs: Option<(&mut Vec<ParagraphInfo>, &mut Vec<bool>)>, +) -> (Vec<BidiClass>, Level, bool) { + let mut original_classes = Vec::with_capacity(text.len()); + + // The stack contains the starting code unit index for each nested isolate we're inside. + let mut isolate_stack = Vec::new(); + + debug_assert!( + if let Some((ref paragraphs, ref pure_ltr)) = split_paragraphs { + paragraphs.is_empty() && pure_ltr.is_empty() + } else { + true + } + ); + + let mut para_start = 0; + let mut para_level = default_para_level; + + // Per-paragraph flag: can subsequent processing be skipped? Set to false if any + // RTL characters or bidi control characters are encountered in the paragraph. + let mut is_pure_ltr = true; + + #[cfg(feature = "flame_it")] + flame::start("compute_initial_info(): iter text.char_indices()"); + + for (i, c) in text.char_indices() { + let class = data_source.bidi_class(c); + + #[cfg(feature = "flame_it")] + flame::start("original_classes.extend()"); + + let len = T::char_len(c); + original_classes.extend(repeat(class).take(len)); + + #[cfg(feature = "flame_it")] + flame::end("original_classes.extend()"); + + match class { + B => { + if let Some((ref mut paragraphs, ref mut pure_ltr)) = split_paragraphs { + // P1. Split the text into separate paragraphs. The paragraph separator is kept + // with the previous paragraph. + let para_end = i + len; + paragraphs.push(ParagraphInfo { + range: para_start..para_end, + // P3. If no character is found in p2, set the paragraph level to zero. + level: para_level.unwrap_or(LTR_LEVEL), + }); + pure_ltr.push(is_pure_ltr); + // Reset state for the start of the next paragraph. + para_start = para_end; + // TODO: Support defaulting to direction of previous paragraph + // + // <http://www.unicode.org/reports/tr9/#HL1> + para_level = default_para_level; + is_pure_ltr = true; + isolate_stack.clear(); + } + } + + L | R | AL => { + if class != L { + is_pure_ltr = false; + } + match isolate_stack.last() { + Some(&start) => { + if original_classes[start] == FSI { + // X5c. If the first strong character between FSI and its matching + // PDI is R or AL, treat it as RLI. Otherwise, treat it as LRI. + for j in 0..T::char_len(chars::FSI) { + original_classes[start + j] = if class == L { LRI } else { RLI }; + } + } + } + + None => { + if para_level.is_none() { + // P2. Find the first character of type L, AL, or R, while skipping + // any characters between an isolate initiator and its matching + // PDI. + para_level = Some(if class != L { RTL_LEVEL } else { LTR_LEVEL }); + } + } + } + } + + AN | LRE | RLE | LRO | RLO => { + is_pure_ltr = false; + } + + RLI | LRI | FSI => { + is_pure_ltr = false; + isolate_stack.push(i); + } + + PDI => { + isolate_stack.pop(); + } + + _ => {} + } + } + + if let Some((paragraphs, pure_ltr)) = split_paragraphs { + if para_start < text.len() { + paragraphs.push(ParagraphInfo { + range: para_start..text.len(), + level: para_level.unwrap_or(LTR_LEVEL), + }); + pure_ltr.push(is_pure_ltr); + } + debug_assert_eq!(paragraphs.len(), pure_ltr.len()); + } + debug_assert_eq!(original_classes.len(), text.len()); + + #[cfg(feature = "flame_it")] + flame::end("compute_initial_info(): iter text.char_indices()"); + + ( + original_classes, + para_level.unwrap_or(LTR_LEVEL), + is_pure_ltr, + ) +} + +/// Bidi information of the text. +/// +/// The `original_classes` and `levels` vectors are indexed by byte offsets into the text. If a +/// character is multiple bytes wide, then its class and level will appear multiple times in these +/// vectors. +// TODO: Impl `struct StringProperty<T> { values: Vec<T> }` and use instead of Vec<T> +#[derive(Debug, PartialEq)] +pub struct BidiInfo<'text> { + /// The text + pub text: &'text str, + + /// The BidiClass of the character at each byte in the text. + pub original_classes: Vec<BidiClass>, + + /// The directional embedding level of each byte in the text. + pub levels: Vec<Level>, + + /// The boundaries and paragraph embedding level of each paragraph within the text. + /// + /// TODO: Use SmallVec or similar to avoid overhead when there are only one or two paragraphs? + /// Or just don't include the first paragraph, which always starts at 0? + pub paragraphs: Vec<ParagraphInfo>, +} + +impl<'text> BidiInfo<'text> { + /// Split the text into paragraphs and determine the bidi embedding levels for each paragraph. + /// + /// + /// The `hardcoded-data` Cargo feature (enabled by default) must be enabled to use this. + /// + /// TODO: In early steps, check for special cases that allow later steps to be skipped. like + /// text that is entirely LTR. See the `nsBidi` class from Gecko for comparison. + /// + /// TODO: Support auto-RTL base direction + #[cfg_attr(feature = "flame_it", flamer::flame)] + #[cfg(feature = "hardcoded-data")] + #[inline] + pub fn new(text: &str, default_para_level: Option<Level>) -> BidiInfo<'_> { + Self::new_with_data_source(&HardcodedBidiData, text, default_para_level) + } + + /// Split the text into paragraphs and determine the bidi embedding levels for each paragraph, with a custom [`BidiDataSource`] + /// for Bidi data. If you just wish to use the hardcoded Bidi data, please use [`BidiInfo::new()`] + /// instead (enabled with tbe default `hardcoded-data` Cargo feature). + /// + /// TODO: In early steps, check for special cases that allow later steps to be skipped. like + /// text that is entirely LTR. See the `nsBidi` class from Gecko for comparison. + /// + /// TODO: Support auto-RTL base direction + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn new_with_data_source<'a, D: BidiDataSource>( + data_source: &D, + text: &'a str, + default_para_level: Option<Level>, + ) -> BidiInfo<'a> { + let InitialInfoExt { base, pure_ltr, .. } = + InitialInfoExt::new_with_data_source(data_source, text, default_para_level); + + let mut levels = Vec::<Level>::with_capacity(text.len()); + let mut processing_classes = base.original_classes.clone(); + + for (para, is_pure_ltr) in base.paragraphs.iter().zip(pure_ltr.iter()) { + let text = &text[para.range.clone()]; + let original_classes = &base.original_classes[para.range.clone()]; + + compute_bidi_info_for_para( + data_source, + para, + *is_pure_ltr, + text, + original_classes, + &mut processing_classes, + &mut levels, + ); + } + + BidiInfo { + text, + original_classes: base.original_classes, + paragraphs: base.paragraphs, + levels, + } + } + + /// Produce the levels for this paragraph as needed for reordering, one level per *byte* + /// in the paragraph. The returned vector includes bytes that are not included + /// in the `line`, but will not adjust them. + /// + /// This runs [Rule L1], you can run + /// [Rule L2] by calling [`Self::reorder_visual()`]. + /// If doing so, you may prefer to use [`Self::reordered_levels_per_char()`] instead + /// to avoid non-byte indices. + /// + /// For an all-in-one reordering solution, consider using [`Self::reorder_visual()`]. + /// + /// [Rule L1]: https://www.unicode.org/reports/tr9/#L1 + /// [Rule L2]: https://www.unicode.org/reports/tr9/#L2 + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn reordered_levels(&self, para: &ParagraphInfo, line: Range<usize>) -> Vec<Level> { + assert!(line.start <= self.levels.len()); + assert!(line.end <= self.levels.len()); + + let mut levels = self.levels.clone(); + let line_classes = &self.original_classes[line.clone()]; + let line_levels = &mut levels[line.clone()]; + + reorder_levels( + line_classes, + line_levels, + self.text.subrange(line), + para.level, + ); + + levels + } + + /// Produce the levels for this paragraph as needed for reordering, one level per *character* + /// in the paragraph. The returned vector includes characters that are not included + /// in the `line`, but will not adjust them. + /// + /// This runs [Rule L1], you can run + /// [Rule L2] by calling [`Self::reorder_visual()`]. + /// If doing so, you may prefer to use [`Self::reordered_levels_per_char()`] instead + /// to avoid non-byte indices. + /// + /// For an all-in-one reordering solution, consider using [`Self::reorder_visual()`]. + /// + /// [Rule L1]: https://www.unicode.org/reports/tr9/#L1 + /// [Rule L2]: https://www.unicode.org/reports/tr9/#L2 + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn reordered_levels_per_char( + &self, + para: &ParagraphInfo, + line: Range<usize>, + ) -> Vec<Level> { + let levels = self.reordered_levels(para, line); + self.text.char_indices().map(|(i, _)| levels[i]).collect() + } + + /// Re-order a line based on resolved levels and return the line in display order. + /// + /// This does not apply [Rule L3] or [Rule L4] around combining characters or mirroring. + /// + /// [Rule L3]: https://www.unicode.org/reports/tr9/#L3 + /// [Rule L4]: https://www.unicode.org/reports/tr9/#L4 + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn reorder_line(&self, para: &ParagraphInfo, line: Range<usize>) -> Cow<'text, str> { + if !level::has_rtl(&self.levels[line.clone()]) { + return self.text[line].into(); + } + let (levels, runs) = self.visual_runs(para, line.clone()); + reorder_line(self.text, line, levels, runs) + } + + /// Reorders pre-calculated levels of a sequence of characters. + /// + /// NOTE: This is a convenience method that does not use a `Paragraph` object. It is + /// intended to be used when an application has determined the levels of the objects (character sequences) + /// and just needs to have them reordered. + /// + /// the index map will result in `indexMap[visualIndex]==logicalIndex`. + /// + /// This only runs [Rule L2](http://www.unicode.org/reports/tr9/#L2) as it does not have + /// information about the actual text. + /// + /// Furthermore, if `levels` is an array that is aligned with code units, bytes within a codepoint may be + /// reversed. You may need to fix up the map to deal with this. Alternatively, only pass in arrays where each `Level` + /// is for a single code point. + /// + /// + /// # # Example + /// ``` + /// use unicode_bidi::BidiInfo; + /// use unicode_bidi::Level; + /// + /// let l0 = Level::from(0); + /// let l1 = Level::from(1); + /// let l2 = Level::from(2); + /// + /// let levels = vec![l0, l0, l0, l0]; + /// let index_map = BidiInfo::reorder_visual(&levels); + /// assert_eq!(levels.len(), index_map.len()); + /// assert_eq!(index_map, [0, 1, 2, 3]); + /// + /// let levels: Vec<Level> = vec![l0, l0, l0, l1, l1, l1, l2, l2]; + /// let index_map = BidiInfo::reorder_visual(&levels); + /// assert_eq!(levels.len(), index_map.len()); + /// assert_eq!(index_map, [0, 1, 2, 6, 7, 5, 4, 3]); + /// ``` + #[cfg_attr(feature = "flame_it", flamer::flame)] + #[inline] + pub fn reorder_visual(levels: &[Level]) -> Vec<usize> { + reorder_visual(levels) + } + + /// Find the level runs within a line and return them in visual order. + /// + /// `line` is a range of bytes indices within `levels`. + /// + /// The first return value is a vector of levels used by the reordering algorithm, + /// i.e. the result of [Rule L1]. The second return value is a vector of level runs, + /// the result of [Rule L2], showing the visual order that each level run (a run of text with the + /// same level) should be displayed. Within each run, the display order can be checked + /// against the Level vector. + /// + /// This does not handle [Rule L3] (combining characters) or [Rule L4] (mirroring), + /// as that should be handled by the engine using this API. + /// + /// Conceptually, this is the same as running [`Self::reordered_levels()`] followed by + /// [`Self::reorder_visual()`], however it returns the result as a list of level runs instead + /// of producing a level map, since one may wish to deal with the fact that this is operating on + /// byte rather than character indices. + /// + /// <http://www.unicode.org/reports/tr9/#Reordering_Resolved_Levels> + /// + /// [Rule L1]: https://www.unicode.org/reports/tr9/#L1 + /// [Rule L2]: https://www.unicode.org/reports/tr9/#L2 + /// [Rule L3]: https://www.unicode.org/reports/tr9/#L3 + /// [Rule L4]: https://www.unicode.org/reports/tr9/#L4 + #[cfg_attr(feature = "flame_it", flamer::flame)] + #[inline] + pub fn visual_runs( + &self, + para: &ParagraphInfo, + line: Range<usize>, + ) -> (Vec<Level>, Vec<LevelRun>) { + let levels = self.reordered_levels(para, line.clone()); + visual_runs_for_line(levels, &line) + } + + /// If processed text has any computed RTL levels + /// + /// This information is usually used to skip re-ordering of text when no RTL level is present + #[inline] + pub fn has_rtl(&self) -> bool { + level::has_rtl(&self.levels) + } +} + +/// Bidi information of text treated as a single paragraph. +/// +/// The `original_classes` and `levels` vectors are indexed by byte offsets into the text. If a +/// character is multiple bytes wide, then its class and level will appear multiple times in these +/// vectors. +#[derive(Debug, PartialEq)] +pub struct ParagraphBidiInfo<'text> { + /// The text + pub text: &'text str, + + /// The BidiClass of the character at each byte in the text. + pub original_classes: Vec<BidiClass>, + + /// The directional embedding level of each byte in the text. + pub levels: Vec<Level>, + + /// The paragraph embedding level. + pub paragraph_level: Level, + + /// Whether the paragraph is purely LTR. + pub is_pure_ltr: bool, +} + +impl<'text> ParagraphBidiInfo<'text> { + /// Determine the bidi embedding level. + /// + /// + /// The `hardcoded-data` Cargo feature (enabled by default) must be enabled to use this. + /// + /// TODO: In early steps, check for special cases that allow later steps to be skipped. like + /// text that is entirely LTR. See the `nsBidi` class from Gecko for comparison. + /// + /// TODO: Support auto-RTL base direction + #[cfg_attr(feature = "flame_it", flamer::flame)] + #[cfg(feature = "hardcoded-data")] + #[inline] + pub fn new(text: &str, default_para_level: Option<Level>) -> ParagraphBidiInfo<'_> { + Self::new_with_data_source(&HardcodedBidiData, text, default_para_level) + } + + /// Determine the bidi embedding level, with a custom [`BidiDataSource`] + /// for Bidi data. If you just wish to use the hardcoded Bidi data, please use [`BidiInfo::new()`] + /// instead (enabled with tbe default `hardcoded-data` Cargo feature). + /// + /// (This is the single-paragraph equivalent of BidiInfo::new_with_data_source, + /// and should be kept in sync with it. + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn new_with_data_source<'a, D: BidiDataSource>( + data_source: &D, + text: &'a str, + default_para_level: Option<Level>, + ) -> ParagraphBidiInfo<'a> { + // Here we could create a ParagraphInitialInfo struct to parallel the one + // used by BidiInfo, but there doesn't seem any compelling reason for it. + let (original_classes, paragraph_level, is_pure_ltr) = + compute_initial_info(data_source, text, default_para_level, None); + + let mut levels = Vec::<Level>::with_capacity(text.len()); + let mut processing_classes = original_classes.clone(); + + let para_info = ParagraphInfo { + range: Range { + start: 0, + end: text.len(), + }, + level: paragraph_level, + }; + + compute_bidi_info_for_para( + data_source, + ¶_info, + is_pure_ltr, + text, + &original_classes, + &mut processing_classes, + &mut levels, + ); + + ParagraphBidiInfo { + text, + original_classes, + levels, + paragraph_level, + is_pure_ltr, + } + } + + /// Produce the levels for this paragraph as needed for reordering, one level per *byte* + /// in the paragraph. The returned vector includes bytes that are not included + /// in the `line`, but will not adjust them. + /// + /// See BidiInfo::reordered_levels for details. + /// + /// (This should be kept in sync with BidiInfo::reordered_levels.) + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn reordered_levels(&self, line: Range<usize>) -> Vec<Level> { + assert!(line.start <= self.levels.len()); + assert!(line.end <= self.levels.len()); + + let mut levels = self.levels.clone(); + let line_classes = &self.original_classes[line.clone()]; + let line_levels = &mut levels[line.clone()]; + + reorder_levels( + line_classes, + line_levels, + self.text.subrange(line), + self.paragraph_level, + ); + + levels + } + + /// Produce the levels for this paragraph as needed for reordering, one level per *character* + /// in the paragraph. The returned vector includes characters that are not included + /// in the `line`, but will not adjust them. + /// + /// See BidiInfo::reordered_levels_per_char for details. + /// + /// (This should be kept in sync with BidiInfo::reordered_levels_per_char.) + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn reordered_levels_per_char(&self, line: Range<usize>) -> Vec<Level> { + let levels = self.reordered_levels(line); + self.text.char_indices().map(|(i, _)| levels[i]).collect() + } + + /// Re-order a line based on resolved levels and return the line in display order. + /// + /// See BidiInfo::reorder_line for details. + /// + /// (This should be kept in sync with BidiInfo::reorder_line.) + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn reorder_line(&self, line: Range<usize>) -> Cow<'text, str> { + if !level::has_rtl(&self.levels[line.clone()]) { + return self.text[line].into(); + } + + let (levels, runs) = self.visual_runs(line.clone()); + + reorder_line(self.text, line, levels, runs) + } + + /// Reorders pre-calculated levels of a sequence of characters. + /// + /// See BidiInfo::reorder_visual for details. + #[cfg_attr(feature = "flame_it", flamer::flame)] + #[inline] + pub fn reorder_visual(levels: &[Level]) -> Vec<usize> { + reorder_visual(levels) + } + + /// Find the level runs within a line and return them in visual order. + /// + /// `line` is a range of bytes indices within `levels`. + /// + /// See BidiInfo::visual_runs for details. + /// + /// (This should be kept in sync with BidiInfo::visual_runs.) + #[cfg_attr(feature = "flame_it", flamer::flame)] + #[inline] + pub fn visual_runs(&self, line: Range<usize>) -> (Vec<Level>, Vec<LevelRun>) { + let levels = self.reordered_levels(line.clone()); + visual_runs_for_line(levels, &line) + } + + /// If processed text has any computed RTL levels + /// + /// This information is usually used to skip re-ordering of text when no RTL level is present + #[inline] + pub fn has_rtl(&self) -> bool { + !self.is_pure_ltr + } + + /// Return the paragraph's Direction (Ltr, Rtl, or Mixed) based on its levels. + #[inline] + pub fn direction(&self) -> Direction { + para_direction(&self.levels) + } +} + +/// Return a line of the text in display order based on resolved levels. +/// +/// `text` the full text passed to the `BidiInfo` or `ParagraphBidiInfo` for analysis +/// `line` a range of byte indices within `text` corresponding to one line +/// `levels` array of `Level` values, with `line`'s levels reordered into visual order +/// `runs` array of `LevelRun`s in visual order +/// +/// (`levels` and `runs` are the result of calling `BidiInfo::visual_runs()` or +/// `ParagraphBidiInfo::visual_runs()` for the line of interest.) +/// +/// Returns: the reordered text of the line. +/// +/// This does not apply [Rule L3] or [Rule L4] around combining characters or mirroring. +/// +/// [Rule L3]: https://www.unicode.org/reports/tr9/#L3 +/// [Rule L4]: https://www.unicode.org/reports/tr9/#L4 +fn reorder_line<'text>( + text: &'text str, + line: Range<usize>, + levels: Vec<Level>, + runs: Vec<LevelRun>, +) -> Cow<'text, str> { + // If all isolating run sequences are LTR, no reordering is needed + if runs.iter().all(|run| levels[run.start].is_ltr()) { + return text[line].into(); + } + + let mut result = String::with_capacity(line.len()); + for run in runs { + if levels[run.start].is_rtl() { + result.extend(text[run].chars().rev()); + } else { + result.push_str(&text[run]); + } + } + result.into() +} + +/// Find the level runs within a line and return them in visual order. +/// +/// `line` is a range of code-unit indices within `levels`. +/// +/// The first return value is a vector of levels used by the reordering algorithm, +/// i.e. the result of [Rule L1]. The second return value is a vector of level runs, +/// the result of [Rule L2], showing the visual order that each level run (a run of text with the +/// same level) should be displayed. Within each run, the display order can be checked +/// against the Level vector. +/// +/// This does not handle [Rule L3] (combining characters) or [Rule L4] (mirroring), +/// as that should be handled by the engine using this API. +/// +/// Conceptually, this is the same as running [`reordered_levels()`] followed by +/// [`reorder_visual()`], however it returns the result as a list of level runs instead +/// of producing a level map, since one may wish to deal with the fact that this is operating on +/// byte rather than character indices. +/// +/// <http://www.unicode.org/reports/tr9/#Reordering_Resolved_Levels> +/// +/// [Rule L1]: https://www.unicode.org/reports/tr9/#L1 +/// [Rule L2]: https://www.unicode.org/reports/tr9/#L2 +/// [Rule L3]: https://www.unicode.org/reports/tr9/#L3 +/// [Rule L4]: https://www.unicode.org/reports/tr9/#L4 +fn visual_runs_for_line(levels: Vec<Level>, line: &Range<usize>) -> (Vec<Level>, Vec<LevelRun>) { + // Find consecutive level runs. + let mut runs = Vec::new(); + let mut start = line.start; + let mut run_level = levels[start]; + let mut min_level = run_level; + let mut max_level = run_level; + + for (i, &new_level) in levels.iter().enumerate().take(line.end).skip(start + 1) { + if new_level != run_level { + // End of the previous run, start of a new one. + runs.push(start..i); + start = i; + run_level = new_level; + min_level = cmp::min(run_level, min_level); + max_level = cmp::max(run_level, max_level); + } + } + runs.push(start..line.end); + + let run_count = runs.len(); + + // Re-order the odd runs. + // <http://www.unicode.org/reports/tr9/#L2> + + // Stop at the lowest *odd* level. + min_level = min_level.new_lowest_ge_rtl().expect("Level error"); + // This loop goes through contiguous chunks of level runs that have a level + // ≥ max_level and reverses their contents, reducing max_level by 1 each time. + while max_level >= min_level { + // Look for the start of a sequence of consecutive runs of max_level or higher. + let mut seq_start = 0; + while seq_start < run_count { + if levels[runs[seq_start].start] < max_level { + seq_start += 1; + continue; + } + + // Found the start of a sequence. Now find the end. + let mut seq_end = seq_start + 1; + while seq_end < run_count { + if levels[runs[seq_end].start] < max_level { + break; + } + seq_end += 1; + } + // Reverse the runs within this sequence. + runs[seq_start..seq_end].reverse(); + + seq_start = seq_end; + } + max_level + .lower(1) + .expect("Lowering embedding level below zero"); + } + (levels, runs) +} + +/// Reorders pre-calculated levels of a sequence of characters. +/// +/// NOTE: This is a convenience method that does not use a `Paragraph` object. It is +/// intended to be used when an application has determined the levels of the objects (character sequences) +/// and just needs to have them reordered. +/// +/// the index map will result in `indexMap[visualIndex]==logicalIndex`. +/// +/// This only runs [Rule L2](http://www.unicode.org/reports/tr9/#L2) as it does not have +/// information about the actual text. +/// +/// Furthermore, if `levels` is an array that is aligned with code units, bytes within a codepoint may be +/// reversed. You may need to fix up the map to deal with this. Alternatively, only pass in arrays where each `Level` +/// is for a single code point. +fn reorder_visual(levels: &[Level]) -> Vec<usize> { + // Gets the next range of characters after start_index with a level greater + // than or equal to `max` + fn next_range(levels: &[level::Level], mut start_index: usize, max: Level) -> Range<usize> { + if levels.is_empty() || start_index >= levels.len() { + return start_index..start_index; + } + while let Some(l) = levels.get(start_index) { + if *l >= max { + break; + } + start_index += 1; + } + + if levels.get(start_index).is_none() { + // If at the end of the array, adding one will + // produce an out-of-range end element + return start_index..start_index; + } + + let mut end_index = start_index + 1; + while let Some(l) = levels.get(end_index) { + if *l < max { + return start_index..end_index; + } + end_index += 1; + } + + start_index..end_index + } + + // This implementation is similar to the L2 implementation in `visual_runs()` + // but it cannot benefit from a precalculated LevelRun vector so needs to be different. + + if levels.is_empty() { + return vec![]; + } + + // Get the min and max levels + let (mut min, mut max) = levels + .iter() + .fold((levels[0], levels[0]), |(min, max), &l| { + (cmp::min(min, l), cmp::max(max, l)) + }); + + // Initialize an index map + let mut result: Vec<usize> = (0..levels.len()).collect(); + + if min == max && min.is_ltr() { + // Everything is LTR and at the same level, do nothing + return result; + } + + // Stop at the lowest *odd* level, since everything below that + // is LTR and does not need further reordering + min = min.new_lowest_ge_rtl().expect("Level error"); + + // For each max level, take all contiguous chunks of + // levels ≥ max and reverse them + // + // We can do this check with the original levels instead of checking reorderings because all + // prior reorderings will have been for contiguous chunks of levels >> max, which will + // be a subset of these chunks anyway. + while min <= max { + let mut range = 0..0; + loop { + range = next_range(levels, range.end, max); + result[range.clone()].reverse(); + + if range.end >= levels.len() { + break; + } + } + + max.lower(1).expect("Level error"); + } + + result +} + +/// The core of BidiInfo initialization, factored out into a function that both +/// the utf-8 and utf-16 versions of BidiInfo can use. +fn compute_bidi_info_for_para<'a, D: BidiDataSource, T: TextSource<'a> + ?Sized>( + data_source: &D, + para: &ParagraphInfo, + is_pure_ltr: bool, + text: &'a T, + original_classes: &[BidiClass], + processing_classes: &mut [BidiClass], + levels: &mut Vec<Level>, +) { + let new_len = levels.len() + para.range.len(); + levels.resize(new_len, para.level); + if para.level == LTR_LEVEL && is_pure_ltr { + return; + } + + let processing_classes = &mut processing_classes[para.range.clone()]; + let levels = &mut levels[para.range.clone()]; + + explicit::compute( + text, + para.level, + original_classes, + levels, + processing_classes, + ); + + let sequences = prepare::isolating_run_sequences(para.level, original_classes, levels); + for sequence in &sequences { + implicit::resolve_weak(text, sequence, processing_classes); + implicit::resolve_neutral( + text, + data_source, + sequence, + levels, + original_classes, + processing_classes, + ); + } + implicit::resolve_levels(processing_classes, levels); + + assign_levels_to_removed_chars(para.level, original_classes, levels); +} + +/// Produce the levels for this paragraph as needed for reordering, one level per *code unit* +/// in the paragraph. The returned vector includes code units that are not included +/// in the `line`, but will not adjust them. +/// +/// This runs [Rule L1] +/// +/// [Rule L1]: https://www.unicode.org/reports/tr9/#L1 +fn reorder_levels<'a, T: TextSource<'a> + ?Sized>( + line_classes: &[BidiClass], + line_levels: &mut [Level], + line_text: &'a T, + para_level: Level, +) { + // Reset some whitespace chars to paragraph level. + // <http://www.unicode.org/reports/tr9/#L1> + let mut reset_from: Option<usize> = Some(0); + let mut reset_to: Option<usize> = None; + let mut prev_level = para_level; + for (i, c) in line_text.char_indices() { + match line_classes[i] { + // Segment separator, Paragraph separator + B | S => { + assert_eq!(reset_to, None); + reset_to = Some(i + T::char_len(c)); + if reset_from == None { + reset_from = Some(i); + } + } + // Whitespace, isolate formatting + WS | FSI | LRI | RLI | PDI => { + if reset_from == None { + reset_from = Some(i); + } + } + // <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + // same as above + set the level + RLE | LRE | RLO | LRO | PDF | BN => { + if reset_from == None { + reset_from = Some(i); + } + // also set the level to previous + line_levels[i] = prev_level; + } + _ => { + reset_from = None; + } + } + if let (Some(from), Some(to)) = (reset_from, reset_to) { + for level in &mut line_levels[from..to] { + *level = para_level; + } + reset_from = None; + reset_to = None; + } + prev_level = line_levels[i]; + } + if let Some(from) = reset_from { + for level in &mut line_levels[from..] { + *level = para_level; + } + } +} + +/// Contains a reference of `BidiInfo` and one of its `paragraphs`. +/// And it supports all operation in the `Paragraph` that needs also its +/// `BidiInfo` such as `direction`. +#[derive(Debug)] +pub struct Paragraph<'a, 'text> { + pub info: &'a BidiInfo<'text>, + pub para: &'a ParagraphInfo, +} + +impl<'a, 'text> Paragraph<'a, 'text> { + #[inline] + pub fn new(info: &'a BidiInfo<'text>, para: &'a ParagraphInfo) -> Paragraph<'a, 'text> { + Paragraph { info, para } + } + + /// Returns if the paragraph is Left direction, right direction or mixed. + #[inline] + pub fn direction(&self) -> Direction { + para_direction(&self.info.levels[self.para.range.clone()]) + } + + /// Returns the `Level` of a certain character in the paragraph. + #[inline] + pub fn level_at(&self, pos: usize) -> Level { + let actual_position = self.para.range.start + pos; + self.info.levels[actual_position] + } +} + +/// Return the directionality of the paragraph (Left, Right or Mixed) from its levels. +#[cfg_attr(feature = "flame_it", flamer::flame)] +fn para_direction(levels: &[Level]) -> Direction { + let mut ltr = false; + let mut rtl = false; + for level in levels { + if level.is_ltr() { + ltr = true; + if rtl { + return Direction::Mixed; + } + } + + if level.is_rtl() { + rtl = true; + if ltr { + return Direction::Mixed; + } + } + } + + if ltr { + return Direction::Ltr; + } + + Direction::Rtl +} + +/// Assign levels to characters removed by rule X9. +/// +/// The levels assigned to these characters are not specified by the algorithm. This function +/// assigns each one the level of the previous character, to avoid breaking level runs. +#[cfg_attr(feature = "flame_it", flamer::flame)] +fn assign_levels_to_removed_chars(para_level: Level, classes: &[BidiClass], levels: &mut [Level]) { + for i in 0..levels.len() { + if prepare::removed_by_x9(classes[i]) { + levels[i] = if i > 0 { levels[i - 1] } else { para_level }; + } + } +} + +/// Get the base direction of the text provided according to the Unicode Bidirectional Algorithm. +/// +/// See rules P2 and P3. +/// +/// The base direction is derived from the first character in the string with bidi character type +/// L, R, or AL. If the first such character has type L, Direction::Ltr is returned. If the first +/// such character has type R or AL, Direction::Rtl is returned. +/// +/// If the string does not contain any character of these types (outside of embedded isolate runs), +/// then Direction::Mixed is returned (but should be considered as meaning "neutral" or "unknown", +/// not in fact mixed directions). +/// +/// This is a lightweight function for use when only the base direction is needed and no further +/// bidi processing of the text is needed. +/// +/// If the text contains paragraph separators, this function considers only the first paragraph. +#[cfg(feature = "hardcoded-data")] +#[inline] +pub fn get_base_direction<'a, T: TextSource<'a> + ?Sized>(text: &'a T) -> Direction { + get_base_direction_with_data_source(&HardcodedBidiData, text) +} + +/// Get the base direction of the text provided according to the Unicode Bidirectional Algorithm, +/// considering the full text if the first paragraph is all-neutral. +/// +/// This is the same as get_base_direction except that it does not stop at the first block +/// separator, but just resets the embedding level and continues to look for a strongly- +/// directional character. So the result will be the base direction of the first paragraph +/// that is not purely neutral characters. +#[cfg(feature = "hardcoded-data")] +#[inline] +pub fn get_base_direction_full<'a, T: TextSource<'a> + ?Sized>(text: &'a T) -> Direction { + get_base_direction_full_with_data_source(&HardcodedBidiData, text) +} + +#[inline] +pub fn get_base_direction_with_data_source<'a, D: BidiDataSource, T: TextSource<'a> + ?Sized>( + data_source: &D, + text: &'a T, +) -> Direction { + get_base_direction_impl(data_source, text, false) +} + +#[inline] +pub fn get_base_direction_full_with_data_source< + 'a, + D: BidiDataSource, + T: TextSource<'a> + ?Sized, +>( + data_source: &D, + text: &'a T, +) -> Direction { + get_base_direction_impl(data_source, text, true) +} + +fn get_base_direction_impl<'a, D: BidiDataSource, T: TextSource<'a> + ?Sized>( + data_source: &D, + text: &'a T, + use_full_text: bool, +) -> Direction { + let mut isolate_level = 0; + for c in text.chars() { + match data_source.bidi_class(c) { + LRI | RLI | FSI => isolate_level = isolate_level + 1, + PDI if isolate_level > 0 => isolate_level = isolate_level - 1, + L if isolate_level == 0 => return Direction::Ltr, + R | AL if isolate_level == 0 => return Direction::Rtl, + B if !use_full_text => break, + B if use_full_text => isolate_level = 0, + _ => (), + } + } + // If no strong char was found, return Mixed. Normally this will be treated as Ltr by callers + // (see rule P3), but we don't map this to Ltr here so that a caller that wants to apply other + // heuristics to an all-neutral paragraph can tell the difference. + Direction::Mixed +} + +/// Implementation of TextSource for UTF-8 text (a string slice). +impl<'text> TextSource<'text> for str { + type CharIter = core::str::Chars<'text>; + type CharIndexIter = core::str::CharIndices<'text>; + type IndexLenIter = Utf8IndexLenIter<'text>; + + #[inline] + fn len(&self) -> usize { + (self as &str).len() + } + #[inline] + fn char_at(&self, index: usize) -> Option<(char, usize)> { + if let Some(slice) = self.get(index..) { + if let Some(ch) = slice.chars().next() { + return Some((ch, ch.len_utf8())); + } + } + None + } + #[inline] + fn subrange(&self, range: Range<usize>) -> &Self { + &(self as &str)[range] + } + #[inline] + fn chars(&'text self) -> Self::CharIter { + (self as &str).chars() + } + #[inline] + fn char_indices(&'text self) -> Self::CharIndexIter { + (self as &str).char_indices() + } + #[inline] + fn indices_lengths(&'text self) -> Self::IndexLenIter { + Utf8IndexLenIter::new(&self) + } + #[inline] + fn char_len(ch: char) -> usize { + ch.len_utf8() + } +} + +/// Iterator over (UTF-8) string slices returning (index, char_len) tuple. +#[derive(Debug)] +pub struct Utf8IndexLenIter<'text> { + iter: CharIndices<'text>, +} + +impl<'text> Utf8IndexLenIter<'text> { + #[inline] + pub fn new(text: &'text str) -> Self { + Utf8IndexLenIter { + iter: text.char_indices(), + } + } +} + +impl Iterator for Utf8IndexLenIter<'_> { + type Item = (usize, usize); + + #[inline] + fn next(&mut self) -> Option<Self::Item> { + if let Some((pos, ch)) = self.iter.next() { + return Some((pos, ch.len_utf8())); + } + None + } +} + +#[cfg(test)] +fn to_utf16(s: &str) -> Vec<u16> { + s.encode_utf16().collect() +} + +#[cfg(test)] +#[cfg(feature = "hardcoded-data")] +mod tests { + use super::*; + + use utf16::{ + BidiInfo as BidiInfoU16, InitialInfo as InitialInfoU16, Paragraph as ParagraphU16, + ParagraphBidiInfo as ParagraphBidiInfoU16, + }; + + #[test] + fn test_utf16_text_source() { + let text: &[u16] = + &[0x41, 0xD801, 0xDC01, 0x20, 0xD800, 0x20, 0xDFFF, 0x20, 0xDC00, 0xD800]; + assert_eq!(text.char_at(0), Some(('A', 1))); + assert_eq!(text.char_at(1), Some(('\u{10401}', 2))); + assert_eq!(text.char_at(2), None); + assert_eq!(text.char_at(3), Some((' ', 1))); + assert_eq!(text.char_at(4), Some((char::REPLACEMENT_CHARACTER, 1))); + assert_eq!(text.char_at(5), Some((' ', 1))); + assert_eq!(text.char_at(6), Some((char::REPLACEMENT_CHARACTER, 1))); + assert_eq!(text.char_at(7), Some((' ', 1))); + assert_eq!(text.char_at(8), Some((char::REPLACEMENT_CHARACTER, 1))); + assert_eq!(text.char_at(9), Some((char::REPLACEMENT_CHARACTER, 1))); + assert_eq!(text.char_at(10), None); + } + + #[test] + fn test_utf16_char_iter() { + let text: &[u16] = + &[0x41, 0xD801, 0xDC01, 0x20, 0xD800, 0x20, 0xDFFF, 0x20, 0xDC00, 0xD800]; + assert_eq!(text.len(), 10); + assert_eq!(text.chars().count(), 9); + let mut chars = text.chars(); + assert_eq!(chars.next(), Some('A')); + assert_eq!(chars.next(), Some('\u{10401}')); + assert_eq!(chars.next(), Some(' ')); + assert_eq!(chars.next(), Some('\u{FFFD}')); + assert_eq!(chars.next(), Some(' ')); + assert_eq!(chars.next(), Some('\u{FFFD}')); + assert_eq!(chars.next(), Some(' ')); + assert_eq!(chars.next(), Some('\u{FFFD}')); + assert_eq!(chars.next(), Some('\u{FFFD}')); + assert_eq!(chars.next(), None); + } + + #[test] + fn test_initial_text_info() { + let tests = vec![ + ( + // text + "a1", + // expected bidi classes per utf-8 byte + vec![L, EN], + // expected paragraph-info for utf-8 + vec![ParagraphInfo { + range: 0..2, + level: LTR_LEVEL, + }], + // expected bidi classes per utf-16 code unit + vec![L, EN], + // expected paragraph-info for utf-16 + vec![ParagraphInfo { + range: 0..2, + level: LTR_LEVEL, + }], + ), + ( + // Arabic, space, Hebrew + "\u{0639} \u{05D0}", + vec![AL, AL, WS, R, R], + vec![ParagraphInfo { + range: 0..5, + level: RTL_LEVEL, + }], + vec![AL, WS, R], + vec![ParagraphInfo { + range: 0..3, + level: RTL_LEVEL, + }], + ), + ( + // SMP characters from Kharoshthi, Cuneiform, Adlam: + "\u{10A00}\u{12000}\u{1E900}", + vec![R, R, R, R, L, L, L, L, R, R, R, R], + vec![ParagraphInfo { + range: 0..12, + level: RTL_LEVEL, + }], + vec![R, R, L, L, R, R], + vec![ParagraphInfo { + range: 0..6, + level: RTL_LEVEL, + }], + ), + ( + "a\u{2029}b", + vec![L, B, B, B, L], + vec![ + ParagraphInfo { + range: 0..4, + level: LTR_LEVEL, + }, + ParagraphInfo { + range: 4..5, + level: LTR_LEVEL, + }, + ], + vec![L, B, L], + vec![ + ParagraphInfo { + range: 0..2, + level: LTR_LEVEL, + }, + ParagraphInfo { + range: 2..3, + level: LTR_LEVEL, + }, + ], + ), + ( + "\u{2068}א\u{2069}a", // U+2068 FSI, U+2069 PDI + vec![RLI, RLI, RLI, R, R, PDI, PDI, PDI, L], + vec![ParagraphInfo { + range: 0..9, + level: LTR_LEVEL, + }], + vec![RLI, R, PDI, L], + vec![ParagraphInfo { + range: 0..4, + level: LTR_LEVEL, + }], + ), + ]; + + for t in tests { + assert_eq!( + InitialInfo::new(t.0, None), + InitialInfo { + text: t.0, + original_classes: t.1, + paragraphs: t.2, + } + ); + let text = &to_utf16(t.0); + assert_eq!( + InitialInfoU16::new(text, None), + InitialInfoU16 { + text, + original_classes: t.3, + paragraphs: t.4, + } + ); + } + } + + #[test] + #[cfg(feature = "hardcoded-data")] + fn test_process_text() { + let tests = vec![ + ( + // text + "abc123", + // base level + Some(LTR_LEVEL), + // levels + Level::vec(&[0, 0, 0, 0, 0, 0]), + // original_classes + vec![L, L, L, EN, EN, EN], + // paragraphs + vec![ParagraphInfo { + range: 0..6, + level: LTR_LEVEL, + }], + // levels_u16 + Level::vec(&[0, 0, 0, 0, 0, 0]), + // original_classes_u16 + vec![L, L, L, EN, EN, EN], + // paragraphs_u16 + vec![ParagraphInfo { + range: 0..6, + level: LTR_LEVEL, + }], + ), + ( + "abc \u{05D0}\u{05D1}\u{05D2}", + Some(LTR_LEVEL), + Level::vec(&[0, 0, 0, 0, 1, 1, 1, 1, 1, 1]), + vec![L, L, L, WS, R, R, R, R, R, R], + vec![ParagraphInfo { + range: 0..10, + level: LTR_LEVEL, + }], + Level::vec(&[0, 0, 0, 0, 1, 1, 1]), + vec![L, L, L, WS, R, R, R], + vec![ParagraphInfo { + range: 0..7, + level: LTR_LEVEL, + }], + ), + ( + "abc \u{05D0}\u{05D1}\u{05D2}", + Some(RTL_LEVEL), + Level::vec(&[2, 2, 2, 1, 1, 1, 1, 1, 1, 1]), + vec![L, L, L, WS, R, R, R, R, R, R], + vec![ParagraphInfo { + range: 0..10, + level: RTL_LEVEL, + }], + Level::vec(&[2, 2, 2, 1, 1, 1, 1]), + vec![L, L, L, WS, R, R, R], + vec![ParagraphInfo { + range: 0..7, + level: RTL_LEVEL, + }], + ), + ( + "\u{05D0}\u{05D1}\u{05D2} abc", + Some(LTR_LEVEL), + Level::vec(&[1, 1, 1, 1, 1, 1, 0, 0, 0, 0]), + vec![R, R, R, R, R, R, WS, L, L, L], + vec![ParagraphInfo { + range: 0..10, + level: LTR_LEVEL, + }], + Level::vec(&[1, 1, 1, 0, 0, 0, 0]), + vec![R, R, R, WS, L, L, L], + vec![ParagraphInfo { + range: 0..7, + level: LTR_LEVEL, + }], + ), + ( + "\u{05D0}\u{05D1}\u{05D2} abc", + None, + Level::vec(&[1, 1, 1, 1, 1, 1, 1, 2, 2, 2]), + vec![R, R, R, R, R, R, WS, L, L, L], + vec![ParagraphInfo { + range: 0..10, + level: RTL_LEVEL, + }], + Level::vec(&[1, 1, 1, 1, 2, 2, 2]), + vec![R, R, R, WS, L, L, L], + vec![ParagraphInfo { + range: 0..7, + level: RTL_LEVEL, + }], + ), + ( + "\u{063A}2\u{0638} \u{05D0}2\u{05D2}", + Some(LTR_LEVEL), + Level::vec(&[1, 1, 2, 1, 1, 1, 1, 1, 2, 1, 1]), + vec![AL, AL, EN, AL, AL, WS, R, R, EN, R, R], + vec![ParagraphInfo { + range: 0..11, + level: LTR_LEVEL, + }], + Level::vec(&[1, 2, 1, 1, 1, 2, 1]), + vec![AL, EN, AL, WS, R, EN, R], + vec![ParagraphInfo { + range: 0..7, + level: LTR_LEVEL, + }], + ), + ( + "a א.\nג", + None, + Level::vec(&[0, 0, 1, 1, 0, 0, 1, 1]), + vec![L, WS, R, R, CS, B, R, R], + vec![ + ParagraphInfo { + range: 0..6, + level: LTR_LEVEL, + }, + ParagraphInfo { + range: 6..8, + level: RTL_LEVEL, + }, + ], + Level::vec(&[0, 0, 1, 0, 0, 1]), + vec![L, WS, R, CS, B, R], + vec![ + ParagraphInfo { + range: 0..5, + level: LTR_LEVEL, + }, + ParagraphInfo { + range: 5..6, + level: RTL_LEVEL, + }, + ], + ), + // BidiTest:69635 (AL ET EN) + ( + "\u{060B}\u{20CF}\u{06F9}", + None, + Level::vec(&[1, 1, 1, 1, 1, 2, 2]), + vec![AL, AL, ET, ET, ET, EN, EN], + vec![ParagraphInfo { + range: 0..7, + level: RTL_LEVEL, + }], + Level::vec(&[1, 1, 2]), + vec![AL, ET, EN], + vec![ParagraphInfo { + range: 0..3, + level: RTL_LEVEL, + }], + ), + ]; + + for t in tests { + assert_eq!( + BidiInfo::new(t.0, t.1), + BidiInfo { + text: t.0, + levels: t.2.clone(), + original_classes: t.3.clone(), + paragraphs: t.4.clone(), + } + ); + // If it was a single paragraph, also test ParagraphBidiInfo. + if t.4.len() == 1 { + assert_eq!( + ParagraphBidiInfo::new(t.0, t.1), + ParagraphBidiInfo { + text: t.0, + original_classes: t.3, + levels: t.2.clone(), + paragraph_level: t.4[0].level, + is_pure_ltr: !level::has_rtl(&t.2), + } + ) + } + let text = &to_utf16(t.0); + assert_eq!( + BidiInfoU16::new(text, t.1), + BidiInfoU16 { + text, + levels: t.5.clone(), + original_classes: t.6.clone(), + paragraphs: t.7.clone(), + } + ); + if t.7.len() == 1 { + assert_eq!( + ParagraphBidiInfoU16::new(text, t.1), + ParagraphBidiInfoU16 { + text: text, + original_classes: t.6.clone(), + levels: t.5.clone(), + paragraph_level: t.7[0].level, + is_pure_ltr: !level::has_rtl(&t.5), + } + ) + } + } + } + + #[test] + #[cfg(feature = "hardcoded-data")] + fn test_paragraph_bidi_info() { + // Passing text that includes a paragraph break to the ParagraphBidiInfo API: + // this is a misuse of the API by the client, but our behavior is safe & + // consistent. The embedded paragraph break acts like a separator (tab) would. + let tests = vec![ + ( + "a א.\nג", + None, + // utf-8 results: + vec![L, WS, R, R, CS, B, R, R], + Level::vec(&[0, 0, 1, 1, 1, 1, 1, 1]), + // utf-16 results: + vec![L, WS, R, CS, B, R], + Level::vec(&[0, 0, 1, 1, 1, 1]), + // paragraph level; is_pure_ltr + LTR_LEVEL, + false, + ), + ( + "\u{5d1} a.\nb.", + None, + // utf-8 results: + vec![R, R, WS, L, CS, B, L, CS], + Level::vec(&[1, 1, 1, 2, 2, 2, 2, 1]), + // utf-16 results: + vec![R, WS, L, CS, B, L, CS], + Level::vec(&[1, 1, 2, 2, 2, 2, 1]), + // paragraph level; is_pure_ltr + RTL_LEVEL, + false, + ), + ( + "a א.\tג", + None, + // utf-8 results: + vec![L, WS, R, R, CS, S, R, R], + Level::vec(&[0, 0, 1, 1, 1, 1, 1, 1]), + // utf-16 results: + vec![L, WS, R, CS, S, R], + Level::vec(&[0, 0, 1, 1, 1, 1]), + // paragraph level; is_pure_ltr + LTR_LEVEL, + false, + ), + ( + "\u{5d1} a.\tb.", + None, + // utf-8 results: + vec![R, R, WS, L, CS, S, L, CS], + Level::vec(&[1, 1, 1, 2, 2, 2, 2, 1]), + // utf-16 results: + vec![R, WS, L, CS, S, L, CS], + Level::vec(&[1, 1, 2, 2, 2, 2, 1]), + // paragraph level; is_pure_ltr + RTL_LEVEL, + false, + ), + ]; + + for t in tests { + assert_eq!( + ParagraphBidiInfo::new(t.0, t.1), + ParagraphBidiInfo { + text: t.0, + original_classes: t.2, + levels: t.3, + paragraph_level: t.6, + is_pure_ltr: t.7, + } + ); + let text = &to_utf16(t.0); + assert_eq!( + ParagraphBidiInfoU16::new(text, t.1), + ParagraphBidiInfoU16 { + text: text, + original_classes: t.4, + levels: t.5, + paragraph_level: t.6, + is_pure_ltr: t.7, + } + ); + } + } + + #[test] + #[cfg(feature = "hardcoded-data")] + fn test_bidi_info_has_rtl() { + let tests = vec![ + // ASCII only + ("123", None, false), + ("123", Some(LTR_LEVEL), false), + ("123", Some(RTL_LEVEL), false), + ("abc", None, false), + ("abc", Some(LTR_LEVEL), false), + ("abc", Some(RTL_LEVEL), false), + ("abc 123", None, false), + ("abc\n123", None, false), + // With Hebrew + ("\u{05D0}\u{05D1}\u{05BC}\u{05D2}", None, true), + ("\u{05D0}\u{05D1}\u{05BC}\u{05D2}", Some(LTR_LEVEL), true), + ("\u{05D0}\u{05D1}\u{05BC}\u{05D2}", Some(RTL_LEVEL), true), + ("abc \u{05D0}\u{05D1}\u{05BC}\u{05D2}", None, true), + ("abc\n\u{05D0}\u{05D1}\u{05BC}\u{05D2}", None, true), + ("\u{05D0}\u{05D1}\u{05BC}\u{05D2} abc", None, true), + ("\u{05D0}\u{05D1}\u{05BC}\u{05D2}\nabc", None, true), + ("\u{05D0}\u{05D1}\u{05BC}\u{05D2} 123", None, true), + ("\u{05D0}\u{05D1}\u{05BC}\u{05D2}\n123", None, true), + ]; + + for t in tests { + assert_eq!(BidiInfo::new(t.0, t.1).has_rtl(), t.2); + assert_eq!(BidiInfoU16::new(&to_utf16(t.0), t.1).has_rtl(), t.2); + } + } + + #[cfg(feature = "hardcoded-data")] + fn reorder_paras(text: &str) -> Vec<Cow<'_, str>> { + let bidi_info = BidiInfo::new(text, None); + bidi_info + .paragraphs + .iter() + .map(|para| bidi_info.reorder_line(para, para.range.clone())) + .collect() + } + + #[cfg(feature = "hardcoded-data")] + fn reorder_paras_u16(text: &[u16]) -> Vec<Cow<'_, [u16]>> { + let bidi_info = BidiInfoU16::new(text, None); + bidi_info + .paragraphs + .iter() + .map(|para| bidi_info.reorder_line(para, para.range.clone())) + .collect() + } + + #[test] + #[cfg(feature = "hardcoded-data")] + fn test_reorder_line() { + let tests = vec![ + // Bidi_Class: L L L B L L L B L L L + ("abc\ndef\nghi", vec!["abc\n", "def\n", "ghi"]), + // Bidi_Class: L L EN B L L EN B L L EN + ("ab1\nde2\ngh3", vec!["ab1\n", "de2\n", "gh3"]), + // Bidi_Class: L L L B AL AL AL + ("abc\nابج", vec!["abc\n", "جبا"]), + // Bidi_Class: AL AL AL B L L L + ( + "\u{0627}\u{0628}\u{062C}\nabc", + vec!["\n\u{062C}\u{0628}\u{0627}", "abc"], + ), + ("1.-2", vec!["1.-2"]), + ("1-.2", vec!["1-.2"]), + ("abc אבג", vec!["abc גבא"]), + // Numbers being weak LTR characters, cannot reorder strong RTL + ("123 \u{05D0}\u{05D1}\u{05D2}", vec!["גבא 123"]), + ("abc\u{202A}def", vec!["abc\u{202A}def"]), + ( + "abc\u{202A}def\u{202C}ghi", + vec!["abc\u{202A}def\u{202C}ghi"], + ), + ( + "abc\u{2066}def\u{2069}ghi", + vec!["abc\u{2066}def\u{2069}ghi"], + ), + // Testing for RLE Character + ("\u{202B}abc אבג\u{202C}", vec!["\u{202b}גבא abc\u{202c}"]), + // Testing neutral characters + ("\u{05D0}בג? אבג", vec!["גבא ?גבא"]), + // Testing neutral characters with special case + ("A אבג?", vec!["A גבא?"]), + // Testing neutral characters with Implicit RTL Marker + ("A אבג?\u{200F}", vec!["A \u{200F}?גבא"]), + ("\u{05D0}בג abc", vec!["abc גבא"]), + ("abc\u{2067}.-\u{2069}ghi", vec!["abc\u{2067}-.\u{2069}ghi"]), + ( + "Hello, \u{2068}\u{202E}world\u{202C}\u{2069}!", + vec!["Hello, \u{2068}\u{202E}\u{202C}dlrow\u{2069}!"], + ), + // With mirrorable characters in RTL run + ("\u{05D0}(ב)ג.", vec![".ג)ב(א"]), + // With mirrorable characters on level boundary + ("\u{05D0}ב(גד[&ef].)gh", vec!["gh).]ef&[דג(בא"]), + ]; + + for t in tests { + assert_eq!(reorder_paras(t.0), t.1); + let expect_utf16 = t.1.iter().map(|v| to_utf16(v)).collect::<Vec<_>>(); + assert_eq!(reorder_paras_u16(&to_utf16(t.0)), expect_utf16); + } + } + + fn reordered_levels_for_paras(text: &str) -> Vec<Vec<Level>> { + let bidi_info = BidiInfo::new(text, None); + bidi_info + .paragraphs + .iter() + .map(|para| bidi_info.reordered_levels(para, para.range.clone())) + .collect() + } + + fn reordered_levels_per_char_for_paras(text: &str) -> Vec<Vec<Level>> { + let bidi_info = BidiInfo::new(text, None); + bidi_info + .paragraphs + .iter() + .map(|para| bidi_info.reordered_levels_per_char(para, para.range.clone())) + .collect() + } + + fn reordered_levels_for_paras_u16(text: &[u16]) -> Vec<Vec<Level>> { + let bidi_info = BidiInfoU16::new(text, None); + bidi_info + .paragraphs + .iter() + .map(|para| bidi_info.reordered_levels(para, para.range.clone())) + .collect() + } + + fn reordered_levels_per_char_for_paras_u16(text: &[u16]) -> Vec<Vec<Level>> { + let bidi_info = BidiInfoU16::new(text, None); + bidi_info + .paragraphs + .iter() + .map(|para| bidi_info.reordered_levels_per_char(para, para.range.clone())) + .collect() + } + + #[test] + #[cfg(feature = "hardcoded-data")] + fn test_reordered_levels() { + let tests = vec![ + // BidiTest:946 (LRI PDI) + ( + "\u{2067}\u{2069}", + vec![Level::vec(&[0, 0, 0, 0, 0, 0])], + vec![Level::vec(&[0, 0])], + vec![Level::vec(&[0, 0])], + ), + // BidiTest:69635 (AL ET EN) + ( + "\u{060B}\u{20CF}\u{06F9}", + vec![Level::vec(&[1, 1, 1, 1, 1, 2, 2])], + vec![Level::vec(&[1, 1, 2])], + vec![Level::vec(&[1, 1, 2])], + ), + ]; + + for t in tests { + assert_eq!(reordered_levels_for_paras(t.0), t.1); + assert_eq!(reordered_levels_per_char_for_paras(t.0), t.2); + let text = &to_utf16(t.0); + assert_eq!(reordered_levels_for_paras_u16(text), t.3); + assert_eq!(reordered_levels_per_char_for_paras_u16(text), t.2); + } + + let tests = vec![ + // BidiTest:291284 (AN RLI PDF R) + ( + "\u{0605}\u{2067}\u{202C}\u{0590}", + vec![&["2", "2", "0", "0", "0", "x", "x", "x", "1", "1"]], + vec![&["2", "0", "x", "1"]], + vec![&["2", "0", "x", "1"]], + ), + ]; + + for t in tests { + assert_eq!(reordered_levels_for_paras(t.0), t.1); + assert_eq!(reordered_levels_per_char_for_paras(t.0), t.2); + let text = &to_utf16(t.0); + assert_eq!(reordered_levels_for_paras_u16(text), t.3); + assert_eq!(reordered_levels_per_char_for_paras_u16(text), t.2); + } + + let text = "aa טֶ"; + let bidi_info = BidiInfo::new(text, None); + assert_eq!( + bidi_info.reordered_levels(&bidi_info.paragraphs[0], 3..7), + Level::vec(&[0, 0, 0, 1, 1, 1, 1]), + ); + + let text = &to_utf16(text); + let bidi_info = BidiInfoU16::new(text, None); + assert_eq!( + bidi_info.reordered_levels(&bidi_info.paragraphs[0], 1..4), + Level::vec(&[0, 0, 0, 1, 1]), + ); + } + + #[test] + fn test_paragraph_info_len() { + let text = "hello world"; + let bidi_info = BidiInfo::new(text, None); + assert_eq!(bidi_info.paragraphs.len(), 1); + assert_eq!(bidi_info.paragraphs[0].len(), text.len()); + + let text2 = "How are you"; + let whole_text = format!("{}\n{}", text, text2); + let bidi_info = BidiInfo::new(&whole_text, None); + assert_eq!(bidi_info.paragraphs.len(), 2); + + // The first paragraph include the paragraph separator. + // TODO: investigate if the paragraph separator character + // should not be part of any paragraph. + assert_eq!(bidi_info.paragraphs[0].len(), text.len() + 1); + assert_eq!(bidi_info.paragraphs[1].len(), text2.len()); + + let text = &to_utf16(text); + let bidi_info = BidiInfoU16::new(text, None); + assert_eq!(bidi_info.paragraphs.len(), 1); + assert_eq!(bidi_info.paragraphs[0].len(), text.len()); + + let text2 = &to_utf16(text2); + let whole_text = &to_utf16(&whole_text); + let bidi_info = BidiInfoU16::new(&whole_text, None); + assert_eq!(bidi_info.paragraphs.len(), 2); + + assert_eq!(bidi_info.paragraphs[0].len(), text.len() + 1); + assert_eq!(bidi_info.paragraphs[1].len(), text2.len()); + } + + #[test] + fn test_direction() { + let ltr_text = "hello world"; + let rtl_text = "أهلا بكم"; + let all_paragraphs = format!("{}\n{}\n{}{}", ltr_text, rtl_text, ltr_text, rtl_text); + let bidi_info = BidiInfo::new(&all_paragraphs, None); + assert_eq!(bidi_info.paragraphs.len(), 3); + let p_ltr = Paragraph::new(&bidi_info, &bidi_info.paragraphs[0]); + let p_rtl = Paragraph::new(&bidi_info, &bidi_info.paragraphs[1]); + let p_mixed = Paragraph::new(&bidi_info, &bidi_info.paragraphs[2]); + assert_eq!(p_ltr.direction(), Direction::Ltr); + assert_eq!(p_rtl.direction(), Direction::Rtl); + assert_eq!(p_mixed.direction(), Direction::Mixed); + + let all_paragraphs = &to_utf16(&all_paragraphs); + let bidi_info = BidiInfoU16::new(&all_paragraphs, None); + assert_eq!(bidi_info.paragraphs.len(), 3); + let p_ltr = ParagraphU16::new(&bidi_info, &bidi_info.paragraphs[0]); + let p_rtl = ParagraphU16::new(&bidi_info, &bidi_info.paragraphs[1]); + let p_mixed = ParagraphU16::new(&bidi_info, &bidi_info.paragraphs[2]); + assert_eq!(p_ltr.direction(), Direction::Ltr); + assert_eq!(p_rtl.direction(), Direction::Rtl); + assert_eq!(p_mixed.direction(), Direction::Mixed); + } + + #[test] + fn test_edge_cases_direction() { + // No paragraphs for empty text. + let empty = ""; + let bidi_info = BidiInfo::new(empty, Option::from(RTL_LEVEL)); + assert_eq!(bidi_info.paragraphs.len(), 0); + + let empty = &to_utf16(empty); + let bidi_info = BidiInfoU16::new(empty, Option::from(RTL_LEVEL)); + assert_eq!(bidi_info.paragraphs.len(), 0); + + let tests = vec![ + // The paragraph separator will take the value of the default direction + // which is left to right. + ("\n", None, Direction::Ltr), + // The paragraph separator will take the value of the given initial direction + // which is left to right. + ("\n", Option::from(LTR_LEVEL), Direction::Ltr), + // The paragraph separator will take the value of the given initial direction + // which is right to left. + ("\n", Option::from(RTL_LEVEL), Direction::Rtl), + ]; + + for t in tests { + let bidi_info = BidiInfo::new(t.0, t.1); + assert_eq!(bidi_info.paragraphs.len(), 1); + let p = Paragraph::new(&bidi_info, &bidi_info.paragraphs[0]); + assert_eq!(p.direction(), t.2); + let text = &to_utf16(t.0); + let bidi_info = BidiInfoU16::new(text, t.1); + let p = ParagraphU16::new(&bidi_info, &bidi_info.paragraphs[0]); + assert_eq!(p.direction(), t.2); + } + } + + #[test] + fn test_level_at() { + let ltr_text = "hello world"; + let rtl_text = "أهلا بكم"; + let all_paragraphs = format!("{}\n{}\n{}{}", ltr_text, rtl_text, ltr_text, rtl_text); + let bidi_info = BidiInfo::new(&all_paragraphs, None); + assert_eq!(bidi_info.paragraphs.len(), 3); + + let p_ltr = Paragraph::new(&bidi_info, &bidi_info.paragraphs[0]); + let p_rtl = Paragraph::new(&bidi_info, &bidi_info.paragraphs[1]); + let p_mixed = Paragraph::new(&bidi_info, &bidi_info.paragraphs[2]); + + assert_eq!(p_ltr.level_at(0), LTR_LEVEL); + assert_eq!(p_rtl.level_at(0), RTL_LEVEL); + assert_eq!(p_mixed.level_at(0), LTR_LEVEL); + assert_eq!(p_mixed.info.levels.len(), 54); + assert_eq!(p_mixed.para.range.start, 28); + assert_eq!(p_mixed.level_at(ltr_text.len()), RTL_LEVEL); + + let all_paragraphs = &to_utf16(&all_paragraphs); + let bidi_info = BidiInfoU16::new(&all_paragraphs, None); + assert_eq!(bidi_info.paragraphs.len(), 3); + + let p_ltr = ParagraphU16::new(&bidi_info, &bidi_info.paragraphs[0]); + let p_rtl = ParagraphU16::new(&bidi_info, &bidi_info.paragraphs[1]); + let p_mixed = ParagraphU16::new(&bidi_info, &bidi_info.paragraphs[2]); + + assert_eq!(p_ltr.level_at(0), LTR_LEVEL); + assert_eq!(p_rtl.level_at(0), RTL_LEVEL); + assert_eq!(p_mixed.level_at(0), LTR_LEVEL); + assert_eq!(p_mixed.info.levels.len(), 40); + assert_eq!(p_mixed.para.range.start, 21); + assert_eq!(p_mixed.level_at(ltr_text.len()), RTL_LEVEL); + } + + #[test] + fn test_get_base_direction() { + let tests = vec![ + ("", Direction::Mixed), // return Mixed if no strong character found + ("123[]-+\u{2019}\u{2060}\u{00bf}?", Direction::Mixed), + ("3.14\npi", Direction::Mixed), // only first paragraph is considered + ("[123 'abc']", Direction::Ltr), + ("[123 '\u{0628}' abc", Direction::Rtl), + ("[123 '\u{2066}abc\u{2069}'\u{0628}]", Direction::Rtl), // embedded isolate is ignored + ("[123 '\u{2066}abc\u{2068}'\u{0628}]", Direction::Mixed), + ]; + + for t in tests { + assert_eq!(get_base_direction(t.0), t.1); + let text = &to_utf16(t.0); + assert_eq!(get_base_direction(text.as_slice()), t.1); + } + } + + #[test] + fn test_get_base_direction_full() { + let tests = vec![ + ("", Direction::Mixed), // return Mixed if no strong character found + ("123[]-+\u{2019}\u{2060}\u{00bf}?", Direction::Mixed), + ("3.14\npi", Direction::Ltr), // direction taken from the second paragraph + ("3.14\n\u{05D0}", Direction::Rtl), // direction taken from the second paragraph + ("[123 'abc']", Direction::Ltr), + ("[123 '\u{0628}' abc", Direction::Rtl), + ("[123 '\u{2066}abc\u{2069}'\u{0628}]", Direction::Rtl), // embedded isolate is ignored + ("[123 '\u{2066}abc\u{2068}'\u{0628}]", Direction::Mixed), + ("[123 '\u{2066}abc\u{2068}'\n\u{0628}]", Direction::Rtl), // \n resets embedding level + ]; + + for t in tests { + assert_eq!(get_base_direction_full(t.0), t.1); + let text = &to_utf16(t.0); + assert_eq!(get_base_direction_full(text.as_slice()), t.1); + } + } +} + +#[cfg(all(feature = "serde", feature = "hardcoded-data", test))] +mod serde_tests { + use super::*; + use serde_test::{assert_tokens, Token}; + + #[test] + fn test_levels() { + let text = "abc אבג"; + let bidi_info = BidiInfo::new(text, None); + let levels = bidi_info.levels; + assert_eq!(text.as_bytes().len(), 10); + assert_eq!(levels.len(), 10); + assert_tokens( + &levels, + &[ + Token::Seq { len: Some(10) }, + Token::NewtypeStruct { name: "Level" }, + Token::U8(0), + Token::NewtypeStruct { name: "Level" }, + Token::U8(0), + Token::NewtypeStruct { name: "Level" }, + Token::U8(0), + Token::NewtypeStruct { name: "Level" }, + Token::U8(0), + Token::NewtypeStruct { name: "Level" }, + Token::U8(1), + Token::NewtypeStruct { name: "Level" }, + Token::U8(1), + Token::NewtypeStruct { name: "Level" }, + Token::U8(1), + Token::NewtypeStruct { name: "Level" }, + Token::U8(1), + Token::NewtypeStruct { name: "Level" }, + Token::U8(1), + Token::NewtypeStruct { name: "Level" }, + Token::U8(1), + Token::SeqEnd, + ], + ); + } +} diff --git a/third_party/rust/unicode-bidi/src/prepare.rs b/third_party/rust/unicode-bidi/src/prepare.rs new file mode 100644 index 0000000000..9234e1aa61 --- /dev/null +++ b/third_party/rust/unicode-bidi/src/prepare.rs @@ -0,0 +1,451 @@ +// Copyright 2015 The Servo Project Developers. See the +// COPYRIGHT file at the top-level directory of this distribution. +// +// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or +// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license +// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your +// option. This file may not be copied, modified, or distributed +// except according to those terms. + +//! 3.3.3 Preparations for Implicit Processing +//! +//! <http://www.unicode.org/reports/tr9/#Preparations_for_Implicit_Processing> + +use alloc::vec::Vec; +use core::cmp::max; +use core::ops::Range; + +use super::level::Level; +use super::BidiClass::{self, *}; + +/// A maximal substring of characters with the same embedding level. +/// +/// Represented as a range of byte indices. +pub type LevelRun = Range<usize>; + +/// Output of `isolating_run_sequences` (steps X9-X10) +#[derive(Debug, PartialEq)] +pub struct IsolatingRunSequence { + pub runs: Vec<LevelRun>, + pub sos: BidiClass, // Start-of-sequence type. + pub eos: BidiClass, // End-of-sequence type. +} + +/// Compute the set of isolating run sequences. +/// +/// An isolating run sequence is a maximal sequence of level runs such that for all level runs +/// except the last one in the sequence, the last character of the run is an isolate initiator +/// whose matching PDI is the first character of the next level run in the sequence. +/// +/// Note: This function does *not* return the sequences in order by their first characters. +#[cfg_attr(feature = "flame_it", flamer::flame)] +pub fn isolating_run_sequences( + para_level: Level, + original_classes: &[BidiClass], + levels: &[Level], +) -> Vec<IsolatingRunSequence> { + let runs = level_runs(levels, original_classes); + + // Compute the set of isolating run sequences. + // <http://www.unicode.org/reports/tr9/#BD13> + let mut sequences = Vec::with_capacity(runs.len()); + + // When we encounter an isolate initiator, we push the current sequence onto the + // stack so we can resume it after the matching PDI. + let mut stack = vec![Vec::new()]; + + for run in runs { + assert!(run.len() > 0); + assert!(!stack.is_empty()); + + let start_class = original_classes[run.start]; + // > In rule X10, [..] skip over any BNs when [..]. + // > Do the same when determining if the last character of the sequence is an isolate initiator. + // + // <https://www.unicode.org/reports/tr9/#Retaining_Explicit_Formatting_Characters> + let end_class = original_classes[run.start..run.end] + .iter() + .copied() + .rev() + .filter(not_removed_by_x9) + .next() + .unwrap_or(start_class); + + let mut sequence = if start_class == PDI && stack.len() > 1 { + // Continue a previous sequence interrupted by an isolate. + stack.pop().unwrap() + } else { + // Start a new sequence. + Vec::new() + }; + + sequence.push(run); + + if let RLI | LRI | FSI = end_class { + // Resume this sequence after the isolate. + stack.push(sequence); + } else { + // This sequence is finished. + sequences.push(sequence); + } + } + // Pop any remaning sequences off the stack. + sequences.extend(stack.into_iter().rev().filter(|seq| !seq.is_empty())); + + // Determine the `sos` and `eos` class for each sequence. + // <http://www.unicode.org/reports/tr9/#X10> + sequences + .into_iter() + .map(|sequence: Vec<LevelRun>| { + assert!(!sequence.is_empty()); + + let mut result = IsolatingRunSequence { + runs: sequence, + sos: L, + eos: L, + }; + + let start_of_seq = result.runs[0].start; + let runs_len = result.runs.len(); + let end_of_seq = result.runs[runs_len - 1].end; + + // > (not counting characters removed by X9) + let seq_level = result + .iter_forwards_from(start_of_seq, 0) + .filter(|i| not_removed_by_x9(&original_classes[*i])) + .map(|i| levels[i]) + .next() + .unwrap_or(levels[start_of_seq]); + + // XXXManishearth the spec talks of a start and end level, + // but for a given IRS the two should be equivalent, yes? + let end_level = result + .iter_backwards_from(end_of_seq, runs_len - 1) + .filter(|i| not_removed_by_x9(&original_classes[*i])) + .map(|i| levels[i]) + .next() + .unwrap_or(levels[end_of_seq - 1]); + + #[cfg(test)] + for run in result.runs.clone() { + for idx in run { + if not_removed_by_x9(&original_classes[idx]) { + assert_eq!(seq_level, levels[idx]); + } + } + } + + // Get the level of the last non-removed char before the runs. + let pred_level = match original_classes[..start_of_seq] + .iter() + .rposition(not_removed_by_x9) + { + Some(idx) => levels[idx], + None => para_level, + }; + + // Get the last non-removed character to check if it is an isolate initiator. + // The spec calls for an unmatched one, but matched isolate initiators + // will never be at the end of a level run (otherwise there would be more to the run). + // We unwrap_or(BN) because BN marks removed classes and it won't matter for the check. + let last_non_removed = original_classes[..end_of_seq] + .iter() + .copied() + .rev() + .find(not_removed_by_x9) + .unwrap_or(BN); + + // Get the level of the next non-removed char after the runs. + let succ_level = if let RLI | LRI | FSI = last_non_removed { + para_level + } else { + match original_classes[end_of_seq..] + .iter() + .position(not_removed_by_x9) + { + Some(idx) => levels[end_of_seq + idx], + None => para_level, + } + }; + + result.sos = max(seq_level, pred_level).bidi_class(); + result.eos = max(end_level, succ_level).bidi_class(); + result + }) + .collect() +} + +impl IsolatingRunSequence { + /// Given a text-relative position `pos` and an index of the level run it is in, + /// produce an iterator of all characters after and pos (`pos..`) that are in this + /// run sequence + pub(crate) fn iter_forwards_from( + &self, + pos: usize, + level_run_index: usize, + ) -> impl Iterator<Item = usize> + '_ { + let runs = &self.runs[level_run_index..]; + + // Check that it is in range + // (we can't use contains() since we want an inclusive range) + #[cfg(feature = "std")] + debug_assert!(runs[0].start <= pos && pos <= runs[0].end); + + (pos..runs[0].end).chain(runs[1..].iter().flat_map(Clone::clone)) + } + + /// Given a text-relative position `pos` and an index of the level run it is in, + /// produce an iterator of all characters before and excludingpos (`..pos`) that are in this + /// run sequence + pub(crate) fn iter_backwards_from( + &self, + pos: usize, + level_run_index: usize, + ) -> impl Iterator<Item = usize> + '_ { + let prev_runs = &self.runs[..level_run_index]; + let current = &self.runs[level_run_index]; + + // Check that it is in range + // (we can't use contains() since we want an inclusive range) + #[cfg(feature = "std")] + debug_assert!(current.start <= pos && pos <= current.end); + + (current.start..pos) + .rev() + .chain(prev_runs.iter().rev().flat_map(Clone::clone)) + } +} + +/// Finds the level runs in a paragraph. +/// +/// <http://www.unicode.org/reports/tr9/#BD7> +fn level_runs(levels: &[Level], original_classes: &[BidiClass]) -> Vec<LevelRun> { + assert_eq!(levels.len(), original_classes.len()); + + let mut runs = Vec::new(); + if levels.is_empty() { + return runs; + } + + let mut current_run_level = levels[0]; + let mut current_run_start = 0; + for i in 1..levels.len() { + if !removed_by_x9(original_classes[i]) && levels[i] != current_run_level { + // End the last run and start a new one. + runs.push(current_run_start..i); + current_run_level = levels[i]; + current_run_start = i; + } + } + runs.push(current_run_start..levels.len()); + + runs +} + +/// Should this character be ignored in steps after X9? +/// +/// <http://www.unicode.org/reports/tr9/#X9> +pub fn removed_by_x9(class: BidiClass) -> bool { + match class { + RLE | LRE | RLO | LRO | PDF | BN => true, + _ => false, + } +} + +// For use as a predicate for `position` / `rposition` +pub fn not_removed_by_x9(class: &BidiClass) -> bool { + !removed_by_x9(*class) +} + +#[cfg(test)] +mod tests { + use super::*; + + #[test] + fn test_level_runs() { + assert_eq!(level_runs(&Level::vec(&[]), &[]), &[]); + assert_eq!( + level_runs(&Level::vec(&[0, 0, 0, 1, 1, 2, 0, 0]), &[L; 8]), + &[0..3, 3..5, 5..6, 6..8] + ); + } + + // From <http://www.unicode.org/reports/tr9/#BD13> + #[rustfmt::skip] + #[test] + fn test_isolating_run_sequences() { + + // == Example 1 == + // text1·RLE·text2·PDF·RLE·text3·PDF·text4 + // index 0 1 2 3 4 5 6 7 + let classes = &[L, RLE, L, PDF, RLE, L, PDF, L]; + let levels = &[0, 1, 1, 1, 1, 1, 1, 0]; + let para_level = Level::ltr(); + let mut sequences = isolating_run_sequences(para_level, classes, &Level::vec(levels)); + sequences.sort_by(|a, b| a.runs[0].clone().cmp(b.runs[0].clone())); + assert_eq!( + sequences.iter().map(|s| s.runs.clone()).collect::<Vec<_>>(), + vec![vec![0..2], vec![2..7], vec![7..8]] + ); + + // == Example 2 == + // text1·RLI·text2·PDI·RLI·text3·PDI·text4 + // index 0 1 2 3 4 5 6 7 + let classes = &[L, RLI, L, PDI, RLI, L, PDI, L]; + let levels = &[0, 0, 1, 0, 0, 1, 0, 0]; + let para_level = Level::ltr(); + let mut sequences = isolating_run_sequences(para_level, classes, &Level::vec(levels)); + sequences.sort_by(|a, b| a.runs[0].clone().cmp(b.runs[0].clone())); + assert_eq!( + sequences.iter().map(|s| s.runs.clone()).collect::<Vec<_>>(), + vec![vec![0..2, 3..5, 6..8], vec![2..3], vec![5..6]] + ); + + // == Example 3 == + // text1·RLI·text2·LRI·text3·RLE·text4·PDF·text5·PDI·text6·PDI·text7 + // index 0 1 2 3 4 5 6 7 8 9 10 11 12 + let classes = &[L, RLI, L, LRI, L, RLE, L, PDF, L, PDI, L, PDI, L]; + let levels = &[0, 0, 1, 1, 2, 3, 3, 3, 2, 1, 1, 0, 0]; + let para_level = Level::ltr(); + let mut sequences = isolating_run_sequences(para_level, classes, &Level::vec(levels)); + sequences.sort_by(|a, b| a.runs[0].clone().cmp(b.runs[0].clone())); + assert_eq!( + sequences.iter().map(|s| s.runs.clone()).collect::<Vec<_>>(), + vec![vec![0..2, 11..13], vec![2..4, 9..11], vec![4..6], vec![6..8], vec![8..9]] + ); + } + + // From <http://www.unicode.org/reports/tr9/#X10> + #[rustfmt::skip] + #[test] + fn test_isolating_run_sequences_sos_and_eos() { + + // == Example 1 == + // text1·RLE·text2·LRE·text3·PDF·text4·PDF·RLE·text5·PDF·text6 + // index 0 1 2 3 4 5 6 7 8 9 10 11 + let classes = &[L, RLE, L, LRE, L, PDF, L, PDF, RLE, L, PDF, L]; + let levels = &[0, 1, 1, 2, 2, 2, 1, 1, 1, 1, 1, 0]; + let para_level = Level::ltr(); + let mut sequences = isolating_run_sequences(para_level, classes, &Level::vec(levels)); + sequences.sort_by(|a, b| a.runs[0].clone().cmp(b.runs[0].clone())); + + // text1 + assert_eq!( + &sequences[0], + &IsolatingRunSequence { + runs: vec![0..2], + sos: L, + eos: R, + } + ); + + // text2 + assert_eq!( + &sequences[1], + &IsolatingRunSequence { + runs: vec![2..4], + sos: R, + eos: L, + } + ); + + // text3 + assert_eq!( + &sequences[2], + &IsolatingRunSequence { + runs: vec![4..6], + sos: L, + eos: L, + } + ); + + // text4 text5 + assert_eq!( + &sequences[3], + &IsolatingRunSequence { + runs: vec![6..11], + sos: L, + eos: R, + } + ); + + // text6 + assert_eq!( + &sequences[4], + &IsolatingRunSequence { + runs: vec![11..12], + sos: R, + eos: L, + } + ); + + // == Example 2 == + // text1·RLI·text2·LRI·text3·PDI·text4·PDI·RLI·text5·PDI·text6 + // index 0 1 2 3 4 5 6 7 8 9 10 11 + let classes = &[L, RLI, L, LRI, L, PDI, L, PDI, RLI, L, PDI, L]; + let levels = &[0, 0, 1, 1, 2, 1, 1, 0, 0, 1, 0, 0]; + let para_level = Level::ltr(); + let mut sequences = isolating_run_sequences(para_level, classes, &Level::vec(levels)); + sequences.sort_by(|a, b| a.runs[0].clone().cmp(b.runs[0].clone())); + + // text1·RLI·PDI·RLI·PDI·text6 + assert_eq!( + &sequences[0], + &IsolatingRunSequence { + runs: vec![0..2, 7..9, 10..12], + sos: L, + eos: L, + } + ); + + // text2·LRI·PDI·text4 + assert_eq!( + &sequences[1], + &IsolatingRunSequence { + runs: vec![2..4, 5..7], + sos: R, + eos: R, + } + ); + + // text3 + assert_eq!( + &sequences[2], + &IsolatingRunSequence { + runs: vec![4..5], + sos: L, + eos: L, + } + ); + + // text5 + assert_eq!( + &sequences[3], + &IsolatingRunSequence { + runs: vec![9..10], + sos: R, + eos: R, + } + ); + } + + #[test] + fn test_removed_by_x9() { + let rem_classes = &[RLE, LRE, RLO, LRO, PDF, BN]; + let not_classes = &[L, RLI, AL, LRI, PDI]; + for x in rem_classes { + assert_eq!(removed_by_x9(*x), true); + } + for x in not_classes { + assert_eq!(removed_by_x9(*x), false); + } + } + + #[test] + fn test_not_removed_by_x9() { + let non_x9_classes = &[L, R, AL, EN, ES, ET, AN, CS, NSM, B, S, WS, ON, LRI, RLI, FSI, PDI]; + for x in non_x9_classes { + assert_eq!(not_removed_by_x9(&x), true); + } + } +} diff --git a/third_party/rust/unicode-bidi/src/utf16.rs b/third_party/rust/unicode-bidi/src/utf16.rs new file mode 100644 index 0000000000..dcd9baf2be --- /dev/null +++ b/third_party/rust/unicode-bidi/src/utf16.rs @@ -0,0 +1,791 @@ +// Copyright 2023 The Mozilla Foundation. See the +// COPYRIGHT file at the top-level directory of this distribution. +// +// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or +// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license +// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your +// option. This file may not be copied, modified, or distributed +// except according to those terms. + +use super::TextSource; + +use alloc::borrow::Cow; +use alloc::vec::Vec; +use core::char; +use core::ops::Range; + +use crate::{ + compute_bidi_info_for_para, compute_initial_info, level, para_direction, reorder_levels, + reorder_visual, visual_runs_for_line, +}; +use crate::{BidiClass, BidiDataSource, Direction, Level, LevelRun, ParagraphInfo}; + +#[cfg(feature = "hardcoded-data")] +use crate::HardcodedBidiData; + +/// Initial bidi information of the text (UTF-16 version). +/// +/// Contains the text paragraphs and `BidiClass` of its characters. +#[derive(PartialEq, Debug)] +pub struct InitialInfo<'text> { + /// The text + pub text: &'text [u16], + + /// The BidiClass of the character at each code unit in the text. + /// If a character is multiple code units, its class will appear multiple times in the vector. + pub original_classes: Vec<BidiClass>, + + /// The boundaries and level of each paragraph within the text. + pub paragraphs: Vec<ParagraphInfo>, +} + +impl<'text> InitialInfo<'text> { + /// Find the paragraphs and BidiClasses in a string of text. + /// + /// <http://www.unicode.org/reports/tr9/#The_Paragraph_Level> + /// + /// Also sets the class for each First Strong Isolate initiator (FSI) to LRI or RLI if a strong + /// character is found before the matching PDI. If no strong character is found, the class will + /// remain FSI, and it's up to later stages to treat these as LRI when needed. + /// + /// The `hardcoded-data` Cargo feature (enabled by default) must be enabled to use this. + #[cfg_attr(feature = "flame_it", flamer::flame)] + #[cfg(feature = "hardcoded-data")] + pub fn new(text: &[u16], default_para_level: Option<Level>) -> InitialInfo<'_> { + Self::new_with_data_source(&HardcodedBidiData, text, default_para_level) + } + + /// Find the paragraphs and BidiClasses in a string of text, with a custom [`BidiDataSource`] + /// for Bidi data. If you just wish to use the hardcoded Bidi data, please use [`InitialInfo::new()`] + /// instead (enabled with tbe default `hardcoded-data` Cargo feature) + /// + /// <http://www.unicode.org/reports/tr9/#The_Paragraph_Level> + /// + /// Also sets the class for each First Strong Isolate initiator (FSI) to LRI or RLI if a strong + /// character is found before the matching PDI. If no strong character is found, the class will + /// remain FSI, and it's up to later stages to treat these as LRI when needed. + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn new_with_data_source<'a, D: BidiDataSource>( + data_source: &D, + text: &'a [u16], + default_para_level: Option<Level>, + ) -> InitialInfo<'a> { + InitialInfoExt::new_with_data_source(data_source, text, default_para_level).base + } +} + +/// Extended version of InitialInfo (not public API). +#[derive(PartialEq, Debug)] +struct InitialInfoExt<'text> { + /// The base InitialInfo for the text, recording its paragraphs and bidi classes. + base: InitialInfo<'text>, + + /// Parallel to base.paragraphs, records whether each paragraph is "pure LTR" that + /// requires no further bidi processing (i.e. there are no RTL characters or bidi + /// control codes present). + pure_ltr: Vec<bool>, +} + +impl<'text> InitialInfoExt<'text> { + /// Find the paragraphs and BidiClasses in a string of text, with a custom [`BidiDataSource`] + /// for Bidi data. If you just wish to use the hardcoded Bidi data, please use [`InitialInfo::new()`] + /// instead (enabled with tbe default `hardcoded-data` Cargo feature) + /// + /// <http://www.unicode.org/reports/tr9/#The_Paragraph_Level> + /// + /// Also sets the class for each First Strong Isolate initiator (FSI) to LRI or RLI if a strong + /// character is found before the matching PDI. If no strong character is found, the class will + /// remain FSI, and it's up to later stages to treat these as LRI when needed. + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn new_with_data_source<'a, D: BidiDataSource>( + data_source: &D, + text: &'a [u16], + default_para_level: Option<Level>, + ) -> InitialInfoExt<'a> { + let mut paragraphs = Vec::<ParagraphInfo>::new(); + let mut pure_ltr = Vec::<bool>::new(); + let (original_classes, _, _) = compute_initial_info( + data_source, + text, + default_para_level, + Some((&mut paragraphs, &mut pure_ltr)), + ); + + InitialInfoExt { + base: InitialInfo { + text, + original_classes, + paragraphs, + }, + pure_ltr, + } + } +} + +/// Bidi information of the text (UTF-16 version). +/// +/// The `original_classes` and `levels` vectors are indexed by code unit offsets into the text. If a +/// character is multiple code units wide, then its class and level will appear multiple times in these +/// vectors. +// TODO: Impl `struct StringProperty<T> { values: Vec<T> }` and use instead of Vec<T> +#[derive(Debug, PartialEq)] +pub struct BidiInfo<'text> { + /// The text + pub text: &'text [u16], + + /// The BidiClass of the character at each byte in the text. + pub original_classes: Vec<BidiClass>, + + /// The directional embedding level of each byte in the text. + pub levels: Vec<Level>, + + /// The boundaries and paragraph embedding level of each paragraph within the text. + /// + /// TODO: Use SmallVec or similar to avoid overhead when there are only one or two paragraphs? + /// Or just don't include the first paragraph, which always starts at 0? + pub paragraphs: Vec<ParagraphInfo>, +} + +impl<'text> BidiInfo<'text> { + /// Split the text into paragraphs and determine the bidi embedding levels for each paragraph. + /// + /// + /// The `hardcoded-data` Cargo feature (enabled by default) must be enabled to use this. + /// + /// TODO: In early steps, check for special cases that allow later steps to be skipped. like + /// text that is entirely LTR. See the `nsBidi` class from Gecko for comparison. + /// + /// TODO: Support auto-RTL base direction + #[cfg_attr(feature = "flame_it", flamer::flame)] + #[cfg(feature = "hardcoded-data")] + #[inline] + pub fn new(text: &[u16], default_para_level: Option<Level>) -> BidiInfo<'_> { + Self::new_with_data_source(&HardcodedBidiData, text, default_para_level) + } + + /// Split the text into paragraphs and determine the bidi embedding levels for each paragraph, with a custom [`BidiDataSource`] + /// for Bidi data. If you just wish to use the hardcoded Bidi data, please use [`BidiInfo::new()`] + /// instead (enabled with tbe default `hardcoded-data` Cargo feature). + /// + /// TODO: In early steps, check for special cases that allow later steps to be skipped. like + /// text that is entirely LTR. See the `nsBidi` class from Gecko for comparison. + /// + /// TODO: Support auto-RTL base direction + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn new_with_data_source<'a, D: BidiDataSource>( + data_source: &D, + text: &'a [u16], + default_para_level: Option<Level>, + ) -> BidiInfo<'a> { + let InitialInfoExt { base, pure_ltr, .. } = + InitialInfoExt::new_with_data_source(data_source, text, default_para_level); + + let mut levels = Vec::<Level>::with_capacity(text.len()); + let mut processing_classes = base.original_classes.clone(); + + for (para, is_pure_ltr) in base.paragraphs.iter().zip(pure_ltr.iter()) { + let text = &text[para.range.clone()]; + let original_classes = &base.original_classes[para.range.clone()]; + + compute_bidi_info_for_para( + data_source, + para, + *is_pure_ltr, + text, + original_classes, + &mut processing_classes, + &mut levels, + ); + } + + BidiInfo { + text, + original_classes: base.original_classes, + paragraphs: base.paragraphs, + levels, + } + } + + /// Produce the levels for this paragraph as needed for reordering, one level per *byte* + /// in the paragraph. The returned vector includes bytes that are not included + /// in the `line`, but will not adjust them. + /// + /// This runs [Rule L1], you can run + /// [Rule L2] by calling [`Self::reorder_visual()`]. + /// If doing so, you may prefer to use [`Self::reordered_levels_per_char()`] instead + /// to avoid non-byte indices. + /// + /// For an all-in-one reordering solution, consider using [`Self::reorder_visual()`]. + /// + /// [Rule L1]: https://www.unicode.org/reports/tr9/#L1 + /// [Rule L2]: https://www.unicode.org/reports/tr9/#L2 + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn reordered_levels(&self, para: &ParagraphInfo, line: Range<usize>) -> Vec<Level> { + assert!(line.start <= self.levels.len()); + assert!(line.end <= self.levels.len()); + + let mut levels = self.levels.clone(); + let line_classes = &self.original_classes[line.clone()]; + let line_levels = &mut levels[line.clone()]; + let line_str: &[u16] = &self.text[line.clone()]; + + reorder_levels(line_classes, line_levels, line_str, para.level); + + levels + } + + /// Produce the levels for this paragraph as needed for reordering, one level per *character* + /// in the paragraph. The returned vector includes characters that are not included + /// in the `line`, but will not adjust them. + /// + /// This runs [Rule L1], you can run + /// [Rule L2] by calling [`Self::reorder_visual()`]. + /// If doing so, you may prefer to use [`Self::reordered_levels_per_char()`] instead + /// to avoid non-byte indices. + /// + /// For an all-in-one reordering solution, consider using [`Self::reorder_visual()`]. + /// + /// [Rule L1]: https://www.unicode.org/reports/tr9/#L1 + /// [Rule L2]: https://www.unicode.org/reports/tr9/#L2 + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn reordered_levels_per_char( + &self, + para: &ParagraphInfo, + line: Range<usize>, + ) -> Vec<Level> { + let levels = self.reordered_levels(para, line); + self.text.char_indices().map(|(i, _)| levels[i]).collect() + } + + /// Re-order a line based on resolved levels and return the line in display order. + /// + /// This does not apply [Rule L3] or [Rule L4] around combining characters or mirroring. + /// + /// [Rule L3]: https://www.unicode.org/reports/tr9/#L3 + /// [Rule L4]: https://www.unicode.org/reports/tr9/#L4 + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn reorder_line(&self, para: &ParagraphInfo, line: Range<usize>) -> Cow<'text, [u16]> { + if !level::has_rtl(&self.levels[line.clone()]) { + return self.text[line].into(); + } + let (levels, runs) = self.visual_runs(para, line.clone()); + reorder_line(self.text, line, levels, runs) + } + + /// Reorders pre-calculated levels of a sequence of characters. + /// + /// NOTE: This is a convenience method that does not use a `Paragraph` object. It is + /// intended to be used when an application has determined the levels of the objects (character sequences) + /// and just needs to have them reordered. + /// + /// the index map will result in `indexMap[visualIndex]==logicalIndex`. + /// + /// This only runs [Rule L2](http://www.unicode.org/reports/tr9/#L2) as it does not have + /// information about the actual text. + /// + /// Furthermore, if `levels` is an array that is aligned with code units, bytes within a codepoint may be + /// reversed. You may need to fix up the map to deal with this. Alternatively, only pass in arrays where each `Level` + /// is for a single code point. + /// + /// + /// # # Example + /// ``` + /// use unicode_bidi::BidiInfo; + /// use unicode_bidi::Level; + /// + /// let l0 = Level::from(0); + /// let l1 = Level::from(1); + /// let l2 = Level::from(2); + /// + /// let levels = vec![l0, l0, l0, l0]; + /// let index_map = BidiInfo::reorder_visual(&levels); + /// assert_eq!(levels.len(), index_map.len()); + /// assert_eq!(index_map, [0, 1, 2, 3]); + /// + /// let levels: Vec<Level> = vec![l0, l0, l0, l1, l1, l1, l2, l2]; + /// let index_map = BidiInfo::reorder_visual(&levels); + /// assert_eq!(levels.len(), index_map.len()); + /// assert_eq!(index_map, [0, 1, 2, 6, 7, 5, 4, 3]); + /// ``` + #[cfg_attr(feature = "flame_it", flamer::flame)] + #[inline] + pub fn reorder_visual(levels: &[Level]) -> Vec<usize> { + reorder_visual(levels) + } + + /// Find the level runs within a line and return them in visual order. + /// + /// `line` is a range of bytes indices within `levels`. + /// + /// The first return value is a vector of levels used by the reordering algorithm, + /// i.e. the result of [Rule L1]. The second return value is a vector of level runs, + /// the result of [Rule L2], showing the visual order that each level run (a run of text with the + /// same level) should be displayed. Within each run, the display order can be checked + /// against the Level vector. + /// + /// This does not handle [Rule L3] (combining characters) or [Rule L4] (mirroring), + /// as that should be handled by the engine using this API. + /// + /// Conceptually, this is the same as running [`Self::reordered_levels()`] followed by + /// [`Self::reorder_visual()`], however it returns the result as a list of level runs instead + /// of producing a level map, since one may wish to deal with the fact that this is operating on + /// byte rather than character indices. + /// + /// <http://www.unicode.org/reports/tr9/#Reordering_Resolved_Levels> + /// + /// [Rule L1]: https://www.unicode.org/reports/tr9/#L1 + /// [Rule L2]: https://www.unicode.org/reports/tr9/#L2 + /// [Rule L3]: https://www.unicode.org/reports/tr9/#L3 + /// [Rule L4]: https://www.unicode.org/reports/tr9/#L4 + #[cfg_attr(feature = "flame_it", flamer::flame)] + #[inline] + pub fn visual_runs( + &self, + para: &ParagraphInfo, + line: Range<usize>, + ) -> (Vec<Level>, Vec<LevelRun>) { + let levels = self.reordered_levels(para, line.clone()); + visual_runs_for_line(levels, &line) + } + + /// If processed text has any computed RTL levels + /// + /// This information is usually used to skip re-ordering of text when no RTL level is present + #[inline] + pub fn has_rtl(&self) -> bool { + level::has_rtl(&self.levels) + } +} + +/// Bidi information of text treated as a single paragraph. +/// +/// The `original_classes` and `levels` vectors are indexed by code unit offsets into the text. If a +/// character is multiple code units wide, then its class and level will appear multiple times in these +/// vectors. +#[derive(Debug, PartialEq)] +pub struct ParagraphBidiInfo<'text> { + /// The text + pub text: &'text [u16], + + /// The BidiClass of the character at each byte in the text. + pub original_classes: Vec<BidiClass>, + + /// The directional embedding level of each byte in the text. + pub levels: Vec<Level>, + + /// The paragraph embedding level. + pub paragraph_level: Level, + + /// Whether the paragraph is purely LTR. + pub is_pure_ltr: bool, +} + +impl<'text> ParagraphBidiInfo<'text> { + /// Determine the bidi embedding level. + /// + /// + /// The `hardcoded-data` Cargo feature (enabled by default) must be enabled to use this. + /// + /// TODO: In early steps, check for special cases that allow later steps to be skipped. like + /// text that is entirely LTR. See the `nsBidi` class from Gecko for comparison. + /// + /// TODO: Support auto-RTL base direction + #[cfg_attr(feature = "flame_it", flamer::flame)] + #[cfg(feature = "hardcoded-data")] + #[inline] + pub fn new(text: &[u16], default_para_level: Option<Level>) -> ParagraphBidiInfo<'_> { + Self::new_with_data_source(&HardcodedBidiData, text, default_para_level) + } + + /// Determine the bidi embedding level, with a custom [`BidiDataSource`] + /// for Bidi data. If you just wish to use the hardcoded Bidi data, please use [`BidiInfo::new()`] + /// instead (enabled with tbe default `hardcoded-data` Cargo feature). + /// + /// (This is the single-paragraph equivalent of BidiInfo::new_with_data_source, + /// and should be kept in sync with it. + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn new_with_data_source<'a, D: BidiDataSource>( + data_source: &D, + text: &'a [u16], + default_para_level: Option<Level>, + ) -> ParagraphBidiInfo<'a> { + // Here we could create a ParagraphInitialInfo struct to parallel the one + // used by BidiInfo, but there doesn't seem any compelling reason for it. + let (original_classes, paragraph_level, is_pure_ltr) = + compute_initial_info(data_source, text, default_para_level, None); + + let mut levels = Vec::<Level>::with_capacity(text.len()); + let mut processing_classes = original_classes.clone(); + + let para_info = ParagraphInfo { + range: Range { + start: 0, + end: text.len(), + }, + level: paragraph_level, + }; + + compute_bidi_info_for_para( + data_source, + ¶_info, + is_pure_ltr, + text, + &original_classes, + &mut processing_classes, + &mut levels, + ); + + ParagraphBidiInfo { + text, + original_classes, + levels, + paragraph_level, + is_pure_ltr, + } + } + + /// Produce the levels for this paragraph as needed for reordering, one level per *code unit* + /// in the paragraph. The returned vector includes code units that are not included + /// in the `line`, but will not adjust them. + /// + /// See BidiInfo::reordered_levels for details. + /// + /// (This should be kept in sync with BidiInfo::reordered_levels.) + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn reordered_levels(&self, line: Range<usize>) -> Vec<Level> { + assert!(line.start <= self.levels.len()); + assert!(line.end <= self.levels.len()); + + let mut levels = self.levels.clone(); + let line_classes = &self.original_classes[line.clone()]; + let line_levels = &mut levels[line.clone()]; + + reorder_levels( + line_classes, + line_levels, + self.text.subrange(line), + self.paragraph_level, + ); + + levels + } + + /// Produce the levels for this paragraph as needed for reordering, one level per *character* + /// in the paragraph. The returned vector includes characters that are not included + /// in the `line`, but will not adjust them. + /// + /// See BidiInfo::reordered_levels_per_char for details. + /// + /// (This should be kept in sync with BidiInfo::reordered_levels_per_char.) + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn reordered_levels_per_char(&self, line: Range<usize>) -> Vec<Level> { + let levels = self.reordered_levels(line); + self.text.char_indices().map(|(i, _)| levels[i]).collect() + } + + /// Re-order a line based on resolved levels and return the line in display order. + /// + /// See BidiInfo::reorder_line for details. + /// + /// (This should be kept in sync with BidiInfo::reorder_line.) + #[cfg_attr(feature = "flame_it", flamer::flame)] + pub fn reorder_line(&self, line: Range<usize>) -> Cow<'text, [u16]> { + if !level::has_rtl(&self.levels[line.clone()]) { + return self.text[line].into(); + } + let (levels, runs) = self.visual_runs(line.clone()); + reorder_line(self.text, line, levels, runs) + } + + /// Reorders pre-calculated levels of a sequence of characters. + /// + /// See BidiInfo::reorder_visual for details. + #[cfg_attr(feature = "flame_it", flamer::flame)] + #[inline] + pub fn reorder_visual(levels: &[Level]) -> Vec<usize> { + reorder_visual(levels) + } + + /// Find the level runs within a line and return them in visual order. + /// + /// `line` is a range of code-unit indices within `levels`. + /// + /// See `BidiInfo::visual_runs` for details. + /// + /// (This should be kept in sync with BidiInfo::visual_runs.) + #[cfg_attr(feature = "flame_it", flamer::flame)] + #[inline] + pub fn visual_runs(&self, line: Range<usize>) -> (Vec<Level>, Vec<LevelRun>) { + let levels = self.reordered_levels(line.clone()); + visual_runs_for_line(levels, &line) + } + + /// If processed text has any computed RTL levels + /// + /// This information is usually used to skip re-ordering of text when no RTL level is present + #[inline] + pub fn has_rtl(&self) -> bool { + !self.is_pure_ltr + } + + /// Return the paragraph's Direction (Ltr, Rtl, or Mixed) based on its levels. + #[inline] + pub fn direction(&self) -> Direction { + para_direction(&self.levels) + } +} + +/// Return a line of the text in display order based on resolved levels. +/// +/// `text` the full text passed to the `BidiInfo` or `ParagraphBidiInfo` for analysis +/// `line` a range of byte indices within `text` corresponding to one line +/// `levels` array of `Level` values, with `line`'s levels reordered into visual order +/// `runs` array of `LevelRun`s in visual order +/// +/// (`levels` and `runs` are the result of calling `BidiInfo::visual_runs()` or +/// `ParagraphBidiInfo::visual_runs()` for the line of interest.) +/// +/// Returns: the reordered text of the line. +/// +/// This does not apply [Rule L3] or [Rule L4] around combining characters or mirroring. +/// +/// [Rule L3]: https://www.unicode.org/reports/tr9/#L3 +/// [Rule L4]: https://www.unicode.org/reports/tr9/#L4 +fn reorder_line<'text>( + text: &'text [u16], + line: Range<usize>, + levels: Vec<Level>, + runs: Vec<LevelRun>, +) -> Cow<'text, [u16]> { + // If all isolating run sequences are LTR, no reordering is needed + if runs.iter().all(|run| levels[run.start].is_ltr()) { + return text[line].into(); + } + + let mut result = Vec::<u16>::with_capacity(line.len()); + for run in runs { + if levels[run.start].is_rtl() { + let mut buf = [0; 2]; + for c in text[run].chars().rev() { + result.extend(c.encode_utf16(&mut buf).iter()); + } + } else { + result.extend(text[run].iter()); + } + } + result.into() +} + +/// Contains a reference of `BidiInfo` and one of its `paragraphs`. +/// And it supports all operation in the `Paragraph` that needs also its +/// `BidiInfo` such as `direction`. +#[derive(Debug)] +pub struct Paragraph<'a, 'text> { + pub info: &'a BidiInfo<'text>, + pub para: &'a ParagraphInfo, +} + +impl<'a, 'text> Paragraph<'a, 'text> { + #[inline] + pub fn new(info: &'a BidiInfo<'text>, para: &'a ParagraphInfo) -> Paragraph<'a, 'text> { + Paragraph { info, para } + } + + /// Returns if the paragraph is Left direction, right direction or mixed. + #[inline] + pub fn direction(&self) -> Direction { + para_direction(&self.info.levels[self.para.range.clone()]) + } + + /// Returns the `Level` of a certain character in the paragraph. + #[inline] + pub fn level_at(&self, pos: usize) -> Level { + let actual_position = self.para.range.start + pos; + self.info.levels[actual_position] + } +} + +/// Implementation of TextSource for UTF-16 text in a [u16] array. +/// Note that there could be unpaired surrogates present! + +// Convenience functions to check whether a UTF16 code unit is a surrogate. +#[inline] +fn is_high_surrogate(code: u16) -> bool { + (code & 0xFC00) == 0xD800 +} +#[inline] +fn is_low_surrogate(code: u16) -> bool { + (code & 0xFC00) == 0xDC00 +} + +impl<'text> TextSource<'text> for [u16] { + type CharIter = Utf16CharIter<'text>; + type CharIndexIter = Utf16CharIndexIter<'text>; + type IndexLenIter = Utf16IndexLenIter<'text>; + + #[inline] + fn len(&self) -> usize { + (self as &[u16]).len() + } + fn char_at(&self, index: usize) -> Option<(char, usize)> { + if index >= self.len() { + return None; + } + // Get the indicated code unit and try simply converting it to a char; + // this will fail if it is half of a surrogate pair. + let c = self[index]; + if let Some(ch) = char::from_u32(c.into()) { + return Some((ch, 1)); + } + // If it's a low surrogate, and was immediately preceded by a high surrogate, + // then we're in the middle of a (valid) character, and should return None. + if is_low_surrogate(c) && index > 0 && is_high_surrogate(self[index - 1]) { + return None; + } + // Otherwise, try to decode, returning REPLACEMENT_CHARACTER for errors. + if let Some(ch) = char::decode_utf16(self[index..].iter().cloned()).next() { + if let Ok(ch) = ch { + // This must be a surrogate pair, otherwise char::from_u32() above should + // have succeeded! + debug_assert!(ch.len_utf16() == 2, "BMP should have already been handled"); + return Some((ch, ch.len_utf16())); + } + } else { + debug_assert!( + false, + "Why did decode_utf16 return None when we're not at the end?" + ); + return None; + } + // Failed to decode UTF-16: we must have encountered an unpaired surrogate. + // Return REPLACEMENT_CHARACTER (not None), to continue processing the following text + // and keep indexing correct. + Some((char::REPLACEMENT_CHARACTER, 1)) + } + #[inline] + fn subrange(&self, range: Range<usize>) -> &Self { + &(self as &[u16])[range] + } + #[inline] + fn chars(&'text self) -> Self::CharIter { + Utf16CharIter::new(&self) + } + #[inline] + fn char_indices(&'text self) -> Self::CharIndexIter { + Utf16CharIndexIter::new(&self) + } + #[inline] + fn indices_lengths(&'text self) -> Self::IndexLenIter { + Utf16IndexLenIter::new(&self) + } + #[inline] + fn char_len(ch: char) -> usize { + ch.len_utf16() + } +} + +/// Iterator over UTF-16 text in a [u16] slice, returning (index, char_len) tuple. +#[derive(Debug)] +pub struct Utf16IndexLenIter<'text> { + text: &'text [u16], + cur_pos: usize, +} + +impl<'text> Utf16IndexLenIter<'text> { + #[inline] + pub fn new(text: &'text [u16]) -> Self { + Utf16IndexLenIter { text, cur_pos: 0 } + } +} + +impl Iterator for Utf16IndexLenIter<'_> { + type Item = (usize, usize); + + #[inline] + fn next(&mut self) -> Option<Self::Item> { + if let Some((_, char_len)) = self.text.char_at(self.cur_pos) { + let result = (self.cur_pos, char_len); + self.cur_pos += char_len; + return Some(result); + } + None + } +} + +/// Iterator over UTF-16 text in a [u16] slice, returning (index, char) tuple. +#[derive(Debug)] +pub struct Utf16CharIndexIter<'text> { + text: &'text [u16], + cur_pos: usize, +} + +impl<'text> Utf16CharIndexIter<'text> { + pub fn new(text: &'text [u16]) -> Self { + Utf16CharIndexIter { text, cur_pos: 0 } + } +} + +impl Iterator for Utf16CharIndexIter<'_> { + type Item = (usize, char); + + fn next(&mut self) -> Option<Self::Item> { + if let Some((ch, char_len)) = self.text.char_at(self.cur_pos) { + let result = (self.cur_pos, ch); + self.cur_pos += char_len; + return Some(result); + } + None + } +} + +/// Iterator over UTF-16 text in a [u16] slice, returning Unicode chars. +/// (Unlike the other iterators above, this also supports reverse iteration.) +#[derive(Debug)] +pub struct Utf16CharIter<'text> { + text: &'text [u16], + cur_pos: usize, + end_pos: usize, +} + +impl<'text> Utf16CharIter<'text> { + pub fn new(text: &'text [u16]) -> Self { + Utf16CharIter { + text, + cur_pos: 0, + end_pos: text.len(), + } + } +} + +impl Iterator for Utf16CharIter<'_> { + type Item = char; + + fn next(&mut self) -> Option<Self::Item> { + if let Some((ch, char_len)) = self.text.char_at(self.cur_pos) { + self.cur_pos += char_len; + return Some(ch); + } + None + } +} + +impl DoubleEndedIterator for Utf16CharIter<'_> { + fn next_back(&mut self) -> Option<Self::Item> { + if self.end_pos <= self.cur_pos { + return None; + } + self.end_pos -= 1; + if let Some(ch) = char::from_u32(self.text[self.end_pos] as u32) { + return Some(ch); + } + if self.end_pos > self.cur_pos { + if let Some((ch, char_len)) = self.text.char_at(self.end_pos - 1) { + if char_len == 2 { + self.end_pos -= 1; + return Some(ch); + } + } + } + Some(char::REPLACEMENT_CHARACTER) + } +} |