diff options
Diffstat (limited to 'vendor/ucd-trie/src/lib.rs')
-rw-r--r-- | vendor/ucd-trie/src/lib.rs | 117 |
1 files changed, 117 insertions, 0 deletions
diff --git a/vendor/ucd-trie/src/lib.rs b/vendor/ucd-trie/src/lib.rs new file mode 100644 index 000000000..2c921d53b --- /dev/null +++ b/vendor/ucd-trie/src/lib.rs @@ -0,0 +1,117 @@ +/*! +The ucd-trie crate provides a compressed trie set specifically tailored for +Unicode codepoints. The principle use case for such a trie is to represent +properties defined by Unicode that correspond to sets of Unicode codepoints. +(These properties are formally called boolean properties or "single valued" +properties. See +[UTR#23 S3.3](http://www.unicode.org/reports/tr23/#PropertyTypeDefinitions) +for more details.) + +This crate has two principle types: `TrieSetOwned` and `TrieSetSlice`, +corresponding to a similar split as there is between `Vec<T>` and `&[T]`. +`TrieSetOwned` is the only way to construct a trie from a set of Unicode +codepoints. + +The intended use of this library is to embed a static instance of +`TrieSetSlice` into your source code, and then use its methods as defined in +this crate to test membership. (The `ucd-generate` tool can likely generate +this code for you.) + +Finally, while this crate uses the standard library by default, it provides +`no_std` functionality by disabling the `std` feature. When `no_std` is +enabled, then `TrieSetOwned` is not provided. Instead, only `TrieSetSlice` is +provided, which means `no_std` crates can still embed tries into their code. +*/ + +#![deny(missing_docs)] +#![cfg_attr(not(feature = "std"), no_std)] + +use core::fmt; + +#[cfg(feature = "std")] +pub use crate::owned::{Error, Result, TrieSetOwned}; + +#[cfg(test)] +#[allow(dead_code)] +mod general_category; +#[cfg(feature = "std")] +mod owned; + +const CHUNK_SIZE: usize = 64; + +/// A type alias for `TrieSetSlice<'static>`. +pub type TrieSet = TrieSetSlice<'static>; + +/// A borrowed trie set. +#[derive(Clone, Copy)] +pub struct TrieSetSlice<'a> { + /// first tree, one level + #[doc(hidden)] + pub tree1_level1: &'a [u64], + /// second tree, first level + #[doc(hidden)] + pub tree2_level1: &'a [u8], + /// second tree, second level + #[doc(hidden)] + pub tree2_level2: &'a [u64], + /// third tree, first level + #[doc(hidden)] + pub tree3_level1: &'a [u8], + /// third tree, second level + #[doc(hidden)] + pub tree3_level2: &'a [u8], + /// third tree, third level + #[doc(hidden)] + pub tree3_level3: &'a [u64], +} + +impl<'a> fmt::Debug for TrieSetSlice<'a> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + write!(f, "TrieSetSlice(...)") + } +} + +impl<'a> TrieSetSlice<'a> { + /// Returns true if and only if the given Unicode scalar value is in this + /// set. + pub fn contains_char(&self, c: char) -> bool { + self.contains(c as usize) + } + + /// Returns true if and only if the given codepoint is in this set. + /// + /// If the given value exceeds the codepoint range (i.e., it's greater + /// than `0x10FFFF`), then this returns false. + pub fn contains_u32(&self, cp: u32) -> bool { + if cp > 0x10FFFF { + return false; + } + self.contains(cp as usize) + } + + #[inline(always)] + fn contains(&self, cp: usize) -> bool { + if cp < 0x800 { + self.chunk_contains(cp, self.tree1_level1[cp >> 6]) + } else if cp < 0x10000 { + let leaf = match self.tree2_level1.get((cp >> 6) - 0x20) { + None => return false, + Some(&leaf) => leaf, + }; + self.chunk_contains(cp, self.tree2_level2[leaf as usize]) + } else { + let child = match self.tree3_level1.get((cp >> 12) - 0x10) { + None => return false, + Some(&child) => child, + }; + let i = ((child as usize) * CHUNK_SIZE) + ((cp >> 6) & 0b111111); + let leaf = self.tree3_level2[i]; + self.chunk_contains(cp, self.tree3_level3[leaf as usize]) + } + } + + #[inline(always)] + fn chunk_contains(&self, cp: usize, chunk: u64) -> bool { + ((chunk >> (cp & 0b111111)) & 1) == 1 + } +} |