intl/lwbrk/WordBreaker.cpp

   1 /* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
   2 /* This Source Code Form is subject to the terms of the Mozilla Public
   3  * License, v. 2.0. If a copy of the MPL was not distributed with this
   4  * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
   5
   6 #include "mozilla/CheckedInt.h"
   7 #include "mozilla/intl/UnicodeProperties.h"
   8 #include "mozilla/intl/WordBreaker.h"
   9 #include "mozilla/StaticPrefs_layout.h"
  10 #include "nsComplexBreaker.h"
  11 #include "nsTArray.h"
  12 #include "nsUnicodeProperties.h"
  13
  14 #if defined(MOZ_ICU4X) && defined(JS_HAS_INTL_API)
  15 #  include "ICU4XDataProvider.h"
  16 #  include "ICU4XWordBreakIteratorUtf16.hpp"
  17 #  include "ICU4XWordSegmenter.hpp"
  18 #  include "mozilla/intl/ICU4XGeckoDataProvider.h"
  19 #  include "mozilla/StaticPrefs_intl.h"
  20 #  include "nsUnicharUtils.h"
  21 #endif
  22
  23 using mozilla::intl::Script;
  24 using mozilla::intl::UnicodeProperties;
  25 using mozilla::intl::WordBreaker;
  26 using mozilla::intl::WordRange;
  27 using mozilla::unicode::GetGenCategory;
  28
  29 #define ASCII_IS_ALPHA(c) \
  30   ((('a' <= (c)) && ((c) <= 'z')) || (('A' <= (c)) && ((c) <= 'Z')))
  31 #define ASCII_IS_DIGIT(c) (('0' <= (c)) && ((c) <= '9'))
  32 #define ASCII_IS_SPACE(c) \
  33   ((' ' == (c)) || ('\t' == (c)) || ('\r' == (c)) || ('\n' == (c)))
  34 #define IS_ALPHABETICAL_SCRIPT(c) ((c) < 0x2E80)
  35
  36 // we change the beginning of IS_HAN from 0x4e00 to 0x3400 to relfect
  37 // Unicode 3.0
  38 #define IS_HAN(c) \
  39   ((0x3400 <= (c)) && ((c) <= 0x9fff)) || ((0xf900 <= (c)) && ((c) <= 0xfaff))
  40 #define IS_KATAKANA(c) ((0x30A0 <= (c)) && ((c) <= 0x30FF))
  41 #define IS_HIRAGANA(c) ((0x3040 <= (c)) && ((c) <= 0x309F))
  42 #define IS_HALFWIDTHKATAKANA(c) ((0xFF60 <= (c)) && ((c) <= 0xFF9F))
  43
  44 // Return true if aChar belongs to a SEAsian script that is written without
  45 // word spaces, so we need to use the "complex breaker" to find possible word
  46 // boundaries. (https://en.wikipedia.org/wiki/Scriptio_continua)
  47 // (How well this works depends on the level of platform support for finding
  48 // possible line breaks - or possible word boundaries - in the particular
  49 // script. Thai, at least, works pretty well on the major desktop OSes. If
  50 // the script is not supported by the platform, we just won't find any useful
  51 // boundaries.)
  52 static bool IsScriptioContinua(char16_t aChar) {
  53   Script sc = UnicodeProperties::GetScriptCode(aChar);
  54   return sc == Script::THAI || sc == Script::MYANMAR || sc == Script::KHMER ||
  55          sc == Script::JAVANESE || sc == Script::BALINESE ||
  56          sc == Script::SUNDANESE || sc == Script::LAO;
  57 }
  58
  59 /* static */
  60 WordBreaker::WordBreakClass WordBreaker::GetClass(char16_t c) {
  61   // begin of the hack
  62
  63   if (IS_ALPHABETICAL_SCRIPT(c)) {
  64     if (IS_ASCII(c)) {
  65       if (ASCII_IS_SPACE(c)) {
  66         return kWbClassSpace;
  67       }
  68       if (ASCII_IS_ALPHA(c) || ASCII_IS_DIGIT(c) ||
  69           (c == '_' && !StaticPrefs::layout_word_select_stop_at_underscore())) {
  70         return kWbClassAlphaLetter;
  71       }
  72       return kWbClassPunct;
  73     }
  74     if (c == 0x00A0 /*NBSP*/) {
  75       return kWbClassSpace;
  76     }
  77     if (GetGenCategory(c) == nsUGenCategory::kPunctuation) {
  78       return kWbClassPunct;
  79     }
  80     if (IsScriptioContinua(c)) {
  81       return kWbClassScriptioContinua;
  82     }
  83     return kWbClassAlphaLetter;
  84   }
  85   if (IS_HAN(c)) {
  86     return kWbClassHanLetter;
  87   }
  88   if (IS_KATAKANA(c)) {
  89     return kWbClassKatakanaLetter;
  90   }
  91   if (IS_HIRAGANA(c)) {
  92     return kWbClassHiraganaLetter;
  93   }
  94   if (IS_HALFWIDTHKATAKANA(c)) {
  95     return kWbClassHWKatakanaLetter;
  96   }
  97   if (GetGenCategory(c) == nsUGenCategory::kPunctuation) {
  98     return kWbClassPunct;
  99   }
 100   if (IsScriptioContinua(c)) {
 101     return kWbClassScriptioContinua;
 102   }
 103   return kWbClassAlphaLetter;
 104 }
 105
 106 WordRange WordBreaker::FindWord(const nsAString& aText, uint32_t aPos,
 107                                 const FindWordOptions aOptions) {
 108   const CheckedInt<uint32_t> len = aText.Length();
 109   MOZ_RELEASE_ASSERT(len.isValid());
 110
 111   if (aPos >= len.value()) {
 112     return {len.value(), len.value()};
 113   }
 114
 115   WordRange range{0, len.value()};
 116
 117 #if defined(MOZ_ICU4X) && defined(JS_HAS_INTL_API)
 118   if (StaticPrefs::intl_icu4x_segmenter_enabled()) {
 119     auto result =
 120         capi::ICU4XWordSegmenter_create_auto(mozilla::intl::GetDataProvider());
 121     MOZ_ASSERT(result.is_ok);
 122     ICU4XWordSegmenter segmenter(result.ok);
 123     ICU4XWordBreakIteratorUtf16 iterator =
 124         segmenter.segment_utf16(diplomat::span<const uint16_t>(
 125             (const uint16_t*)aText.BeginReading(), aText.Length()));
 126
 127     uint32_t previousPos = 0;
 128     while (true) {
 129       const int32_t nextPos = iterator.next();
 130       if (nextPos < 0) {
 131         range.mBegin = previousPos;
 132         range.mEnd = len.value();
 133         break;
 134       }
 135       if ((uint32_t)nextPos > aPos) {
 136         range.mBegin = previousPos;
 137         range.mEnd = (uint32_t)nextPos;
 138         break;
 139       }
 140
 141       previousPos = nextPos;
 142     }
 143
 144     if (aOptions != FindWordOptions::StopAtPunctuation) {
 145       return range;
 146     }
 147
 148     for (uint32_t i = range.mBegin; i < range.mEnd; i++) {
 149       if (mozilla::IsPunctuationForWordSelect(aText[i])) {
 150         if (i > aPos) {
 151           range.mEnd = i;
 152           break;
 153         }
 154         if (i == aPos) {
 155           range.mBegin = i;
 156           range.mEnd = i + 1;
 157           break;
 158         }
 159         if (i < aPos) {
 160           range.mBegin = i + 1;
 161         }
 162       }
 163     }
 164
 165     return range;
 166   }
 167 #endif
 168
 169   WordBreakClass c = GetClass(aText[aPos]);
 170
 171   // Scan forward
 172   for (uint32_t i = aPos + 1; i < len.value(); i++) {
 173     if (c != GetClass(aText[i])) {
 174       range.mEnd = i;
 175       break;
 176     }
 177   }
 178
 179   // Scan backward
 180   for (uint32_t i = aPos; i > 0; i--) {
 181     if (c != GetClass(aText[i - 1])) {
 182       range.mBegin = i;
 183       break;
 184     }
 185   }
 186
 187   if (kWbClassScriptioContinua == c) {
 188     // we pass the whole text segment to the complex word breaker to find a
 189     // shorter answer
 190     AutoTArray<uint8_t, 256> breakBefore;
 191     breakBefore.SetLength(range.mEnd - range.mBegin);
 192     ComplexBreaker::GetBreaks(aText.BeginReading() + range.mBegin,
 193                               range.mEnd - range.mBegin,
 194                               breakBefore.Elements());
 195
 196     // Scan forward
 197     for (uint32_t i = aPos + 1; i < range.mEnd; i++) {
 198       if (breakBefore[i - range.mBegin]) {
 199         range.mEnd = i;
 200         break;
 201       }
 202     }
 203
 204     // Scan backward
 205     for (uint32_t i = aPos; i > range.mBegin; i--) {
 206       if (breakBefore[i - range.mBegin]) {
 207         range.mBegin = i;
 208         break;
 209       }
 210     }
 211   }
 212   return range;
 213 }
 214
 215 int32_t WordBreaker::Next(const char16_t* aText, uint32_t aLen, uint32_t aPos) {
 216   MOZ_ASSERT(aText);
 217
 218   if (aPos >= aLen) {
 219     return NS_WORDBREAKER_NEED_MORE_TEXT;
 220   }
 221
 222   const WordBreakClass posClass = GetClass(aText[aPos]);
 223   uint32_t nextBreakPos;
 224   for (nextBreakPos = aPos + 1; nextBreakPos < aLen; ++nextBreakPos) {
 225     if (posClass != GetClass(aText[nextBreakPos])) {
 226       break;
 227     }
 228   }
 229
 230   if (kWbClassScriptioContinua == posClass) {
 231     // We pass the whole text segment to the complex word breaker to find a
 232     // shorter answer.
 233     const char16_t* segStart = aText + aPos;
 234     const uint32_t segLen = nextBreakPos - aPos + 1;
 235     AutoTArray<uint8_t, 256> breakBefore;
 236     breakBefore.SetLength(segLen);
 237     ComplexBreaker::GetBreaks(segStart, segLen, breakBefore.Elements());
 238
 239     for (uint32_t i = aPos + 1; i < nextBreakPos; ++i) {
 240       if (breakBefore[i - aPos]) {
 241         nextBreakPos = i;
 242         break;
 243       }
 244     }
 245   }
 246
 247   MOZ_ASSERT(nextBreakPos != aPos);
 248   return nextBreakPos;
 249 }