| 1 | // Copyright 2012-2015 The Rust Project Developers. See the COPYRIGHT |
| 2 | // file at the top-level directory of this distribution and at |
| 3 | // http://rust-lang.org/COPYRIGHT. |
| 4 | // |
| 5 | // Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or |
| 6 | // http://www.apache.org/licenses/LICENSE-2.0> or the MIT license |
| 7 | // <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your |
| 8 | // option. This file may not be copied, modified, or distributed |
| 9 | // except according to those terms. |
| 10 | |
| 11 | //! Functions for computing canonical and compatible decompositions for Unicode characters. |
| 12 | use crate::lookups::{ |
| 13 | canonical_fully_decomposed, cjk_compat_variants_fully_decomposed, |
| 14 | compatibility_fully_decomposed, composition_table, |
| 15 | }; |
| 16 | |
| 17 | use core::char; |
| 18 | |
| 19 | /// Compute canonical Unicode decomposition for character. |
| 20 | /// See [Unicode Standard Annex #15](http://www.unicode.org/reports/tr15/) |
| 21 | /// for more information. |
| 22 | #[inline ] |
| 23 | pub fn decompose_canonical<F>(c: char, emit_char: F) |
| 24 | where |
| 25 | F: FnMut(char), |
| 26 | { |
| 27 | decompose(c, decompose_char:canonical_fully_decomposed, emit_char) |
| 28 | } |
| 29 | |
| 30 | /// Compute canonical or compatible Unicode decomposition for character. |
| 31 | /// See [Unicode Standard Annex #15](http://www.unicode.org/reports/tr15/) |
| 32 | /// for more information. |
| 33 | #[inline ] |
| 34 | pub fn decompose_compatible<F: FnMut(char)>(c: char, emit_char: F) { |
| 35 | let decompose_char: impl Fn(char) -> Option<&'static …> = |
| 36 | |c: char| compatibility_fully_decomposed(c).or_else(|| canonical_fully_decomposed(c)); |
| 37 | decompose(c, decompose_char, emit_char) |
| 38 | } |
| 39 | |
| 40 | /// Compute standard-variation decomposition for character. |
| 41 | /// |
| 42 | /// [Standardized Variation Sequences] are used instead of the standard canonical |
| 43 | /// decompositions, notably for CJK codepoints with singleton canonical decompositions, |
| 44 | /// to avoid losing information. See the |
| 45 | /// [Unicode Variation Sequence FAQ](http://unicode.org/faq/vs.html) and the |
| 46 | /// "Other Enhancements" section of the |
| 47 | /// [Unicode 6.3 Release Summary](https://www.unicode.org/versions/Unicode6.3.0/#Summary) |
| 48 | /// for more information. |
| 49 | #[inline ] |
| 50 | pub fn decompose_cjk_compat_variants<F>(c: char, mut emit_char: F) |
| 51 | where |
| 52 | F: FnMut(char), |
| 53 | { |
| 54 | // 7-bit ASCII never decomposes |
| 55 | if c <= ' \x7f' { |
| 56 | emit_char(c); |
| 57 | return; |
| 58 | } |
| 59 | |
| 60 | // Don't perform decomposition for Hangul |
| 61 | |
| 62 | if let Some(decomposed: &'static [char]) = cjk_compat_variants_fully_decomposed(c) { |
| 63 | for &d: char in decomposed { |
| 64 | emit_char(d); |
| 65 | } |
| 66 | return; |
| 67 | } |
| 68 | |
| 69 | // Finally bottom out. |
| 70 | emit_char(c); |
| 71 | } |
| 72 | |
| 73 | #[inline ] |
| 74 | #[allow (unsafe_code)] |
| 75 | fn decompose<D, F>(c: char, decompose_char: D, mut emit_char: F) |
| 76 | where |
| 77 | D: Fn(char) -> Option<&'static [char]>, |
| 78 | F: FnMut(char), |
| 79 | { |
| 80 | // 7-bit ASCII never decomposes |
| 81 | if c <= ' \x7f' { |
| 82 | emit_char(c); |
| 83 | return; |
| 84 | } |
| 85 | |
| 86 | // Perform decomposition for Hangul |
| 87 | if is_hangul_syllable(c) { |
| 88 | // Safety: Hangul Syllables invariant checked by is_hangul_syllable above |
| 89 | unsafe { |
| 90 | decompose_hangul(c, emit_char); |
| 91 | } |
| 92 | return; |
| 93 | } |
| 94 | |
| 95 | if let Some(decomposed) = decompose_char(c) { |
| 96 | for &d in decomposed { |
| 97 | emit_char(d); |
| 98 | } |
| 99 | return; |
| 100 | } |
| 101 | |
| 102 | // Finally bottom out. |
| 103 | emit_char(c); |
| 104 | } |
| 105 | |
| 106 | /// Compose two characters into a single character, if possible. |
| 107 | /// See [Unicode Standard Annex #15](http://www.unicode.org/reports/tr15/) |
| 108 | /// for more information. |
| 109 | pub fn compose(a: char, b: char) -> Option<char> { |
| 110 | compose_hangul(a, b).or_else(|| composition_table(c1:a, c2:b)) |
| 111 | } |
| 112 | |
| 113 | // Constants from Unicode 9.0.0 Section 3.12 Conjoining Jamo Behavior |
| 114 | // http://www.unicode.org/versions/Unicode9.0.0/ch03.pdf#M9.32468.Heading.310.Combining.Jamo.Behavior |
| 115 | const S_BASE: u32 = 0xAC00; |
| 116 | const L_BASE: u32 = 0x1100; |
| 117 | const V_BASE: u32 = 0x1161; |
| 118 | const T_BASE: u32 = 0x11A7; |
| 119 | const L_COUNT: u32 = 19; |
| 120 | const V_COUNT: u32 = 21; |
| 121 | const T_COUNT: u32 = 28; |
| 122 | const N_COUNT: u32 = V_COUNT * T_COUNT; |
| 123 | const S_COUNT: u32 = L_COUNT * N_COUNT; |
| 124 | |
| 125 | const S_LAST: u32 = S_BASE + S_COUNT - 1; |
| 126 | const L_LAST: u32 = L_BASE + L_COUNT - 1; |
| 127 | const V_LAST: u32 = V_BASE + V_COUNT - 1; |
| 128 | const T_LAST: u32 = T_BASE + T_COUNT - 1; |
| 129 | |
| 130 | // Composition only occurs for `TPart`s in `U+11A8 ..= U+11C2`, |
| 131 | // i.e. `T_BASE + 1 ..= T_LAST`. |
| 132 | const T_FIRST: u32 = T_BASE + 1; |
| 133 | |
| 134 | // Safety-usable invariant: This ensures that c is a valid Hangul Syllable character (U+AC00..U+D7AF) |
| 135 | pub(crate) fn is_hangul_syllable(c: char) -> bool { |
| 136 | // Safety: This checks the range 0xAC00 (S_BASE) to 0xD7A4 (S_BASE + S_COUNT), upholding the safety-usable invariant |
| 137 | (c as u32) >= S_BASE && (c as u32) < (S_BASE + S_COUNT) |
| 138 | } |
| 139 | |
| 140 | // Decompose a precomposed Hangul syllable |
| 141 | // Safety: `s` MUST be a valid Hangul Syllable character, between U+AC00..U+D7AF |
| 142 | #[allow (unsafe_code, unused_unsafe)] |
| 143 | #[inline (always)] |
| 144 | unsafe fn decompose_hangul<F>(s: char, mut emit_char: F) |
| 145 | where |
| 146 | F: FnMut(char), |
| 147 | { |
| 148 | // This will be at most 0x2baf, the size of the Hangul Syllables block |
| 149 | let s_index: u32 = s as u32 - S_BASE; |
| 150 | // This will be at most 0x2baf / (21 * 28), 19 |
| 151 | let l_index: u32 = s_index / N_COUNT; |
| 152 | unsafe { |
| 153 | // Safety: L_BASE (0x1100) plus at most 19 is still going to be in range for a valid Unicode code point in the BMP (< 0xD800) |
| 154 | emit_char(char::from_u32_unchecked(L_BASE + l_index)); |
| 155 | |
| 156 | // Safety: This will be at most (N_COUNT - 1) / T_COUNT = (V*T - 1) / T, which gives us an upper bound of V_COUNT = 21 |
| 157 | let v_index: u32 = (s_index % N_COUNT) / T_COUNT; |
| 158 | // Safety: V_BASE (0x1161) plus at most 21 is still going to be in range for a valid Unicode code point in the BMP (< 0xD800) |
| 159 | emit_char(char::from_u32_unchecked(V_BASE + v_index)); |
| 160 | |
| 161 | // Safety: This will be at most T_COUNT - 1 (27) |
| 162 | let t_index: u32 = s_index % T_COUNT; |
| 163 | if t_index > 0 { |
| 164 | // Safety: T_BASE (0x11A7) plus at most 27 is still going to be in range for a valid Unicode code point in the BMP (< 0xD800) |
| 165 | emit_char(char::from_u32_unchecked(T_BASE + t_index)); |
| 166 | } |
| 167 | } |
| 168 | } |
| 169 | |
| 170 | #[inline ] |
| 171 | pub(crate) fn hangul_decomposition_length(s: char) -> usize { |
| 172 | let si: u32 = s as u32 - S_BASE; |
| 173 | let ti: u32 = si % T_COUNT; |
| 174 | if ti > 0 { |
| 175 | 3 |
| 176 | } else { |
| 177 | 2 |
| 178 | } |
| 179 | } |
| 180 | |
| 181 | // Compose a pair of Hangul Jamo |
| 182 | #[allow (unsafe_code)] |
| 183 | #[inline (always)] |
| 184 | #[allow (ellipsis_inclusive_range_patterns)] |
| 185 | fn compose_hangul(a: char, b: char) -> Option<char> { |
| 186 | let (a, b) = (a as u32, b as u32); |
| 187 | match (a, b) { |
| 188 | // Compose a leading consonant and a vowel together into an LV_Syllable |
| 189 | (L_BASE..=L_LAST, V_BASE..=V_LAST) => { |
| 190 | // Safety: based on the above bounds, l_index will be less than or equal to L_COUNT (19) |
| 191 | // and v_index will be <= V_COUNT (21) |
| 192 | let l_index = a - L_BASE; |
| 193 | let v_index = b - V_BASE; |
| 194 | // Safety: This will be <= 19 * (20 * 21) + (21 * 20), which is 8400. |
| 195 | let lv_index = l_index * N_COUNT + v_index * T_COUNT; |
| 196 | // Safety: This is between 0xAC00 and 0xCCD0, which are in range for Hangul Syllables (U+AC00..U+D7AF) and also in range |
| 197 | // for BMP unicode |
| 198 | let s = S_BASE + lv_index; |
| 199 | // Safety: We've verified this is in-range |
| 200 | Some(unsafe { char::from_u32_unchecked(s) }) |
| 201 | } |
| 202 | // Compose an LV_Syllable and a trailing consonant into an LVT_Syllable |
| 203 | (S_BASE..=S_LAST, T_FIRST..=T_LAST) if (a - S_BASE) % T_COUNT == 0 => { |
| 204 | // Safety: a is between 0xAC00 and (0xAC00 + 19 * 21 * 28). b - T_BASE is between 0 and 19. |
| 205 | // Adding a number 0 to 19 to a number that is at largest 0xD7A4 will not go out of bounds to 0xD800 (where the |
| 206 | // surrogates start), so this is safe. |
| 207 | Some(unsafe { char::from_u32_unchecked(a + (b - T_BASE)) }) |
| 208 | } |
| 209 | _ => None, |
| 210 | } |
| 211 | } |
| 212 | |
| 213 | #[cfg (test)] |
| 214 | mod tests { |
| 215 | use super::compose_hangul; |
| 216 | |
| 217 | // Regression test from a bugfix where we were composing an LV_Syllable with |
| 218 | // T_BASE directly. (We should only compose an LV_Syllable with a character |
| 219 | // in the range `T_BASE + 1 ..= T_LAST`.) |
| 220 | #[test ] |
| 221 | fn test_hangul_composition() { |
| 222 | assert_eq!(compose_hangul(' \u{c8e0}' , ' \u{11a7}' ), None); |
| 223 | } |
| 224 | } |
| 225 | |