1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/layout/generic/nsTextRunTransformations.cpp Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,933 @@
1.4 +/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*-
1.5 + * This Source Code Form is subject to the terms of the Mozilla Public
1.6 + * License, v. 2.0. If a copy of the MPL was not distributed with this
1.7 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
1.8 +
1.9 +#include "nsTextRunTransformations.h"
1.10 +
1.11 +#include "mozilla/MemoryReporting.h"
1.12 +
1.13 +#include "nsGkAtoms.h"
1.14 +#include "nsStyleConsts.h"
1.15 +#include "nsStyleContext.h"
1.16 +#include "nsUnicodeProperties.h"
1.17 +#include "nsSpecialCasingData.h"
1.18 +#include "mozilla/gfx/2D.h"
1.19 +#include "nsTextFrameUtils.h"
1.20 +#include "nsIPersistentProperties2.h"
1.21 +#include "nsNetUtil.h"
1.22 +
1.23 +// Unicode characters needing special casing treatment in tr/az languages
1.24 +#define LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE 0x0130
1.25 +#define LATIN_SMALL_LETTER_DOTLESS_I 0x0131
1.26 +
1.27 +// Greek sigma needs custom handling for the lowercase transform; for details
1.28 +// see comments under "case NS_STYLE_TEXT_TRANSFORM_LOWERCASE" within
1.29 +// nsCaseTransformTextRunFactory::RebuildTextRun(), and bug 740120.
1.30 +#define GREEK_CAPITAL_LETTER_SIGMA 0x03A3
1.31 +#define GREEK_SMALL_LETTER_FINAL_SIGMA 0x03C2
1.32 +#define GREEK_SMALL_LETTER_SIGMA 0x03C3
1.33 +
1.34 +// Custom uppercase mapping for Greek; see bug 307039 for details
1.35 +#define GREEK_LOWER_ALPHA 0x03B1
1.36 +#define GREEK_LOWER_ALPHA_TONOS 0x03AC
1.37 +#define GREEK_LOWER_ALPHA_OXIA 0x1F71
1.38 +#define GREEK_LOWER_EPSILON 0x03B5
1.39 +#define GREEK_LOWER_EPSILON_TONOS 0x03AD
1.40 +#define GREEK_LOWER_EPSILON_OXIA 0x1F73
1.41 +#define GREEK_LOWER_ETA 0x03B7
1.42 +#define GREEK_LOWER_ETA_TONOS 0x03AE
1.43 +#define GREEK_LOWER_ETA_OXIA 0x1F75
1.44 +#define GREEK_LOWER_IOTA 0x03B9
1.45 +#define GREEK_LOWER_IOTA_TONOS 0x03AF
1.46 +#define GREEK_LOWER_IOTA_OXIA 0x1F77
1.47 +#define GREEK_LOWER_IOTA_DIALYTIKA 0x03CA
1.48 +#define GREEK_LOWER_IOTA_DIALYTIKA_TONOS 0x0390
1.49 +#define GREEK_LOWER_IOTA_DIALYTIKA_OXIA 0x1FD3
1.50 +#define GREEK_LOWER_OMICRON 0x03BF
1.51 +#define GREEK_LOWER_OMICRON_TONOS 0x03CC
1.52 +#define GREEK_LOWER_OMICRON_OXIA 0x1F79
1.53 +#define GREEK_LOWER_UPSILON 0x03C5
1.54 +#define GREEK_LOWER_UPSILON_TONOS 0x03CD
1.55 +#define GREEK_LOWER_UPSILON_OXIA 0x1F7B
1.56 +#define GREEK_LOWER_UPSILON_DIALYTIKA 0x03CB
1.57 +#define GREEK_LOWER_UPSILON_DIALYTIKA_TONOS 0x03B0
1.58 +#define GREEK_LOWER_UPSILON_DIALYTIKA_OXIA 0x1FE3
1.59 +#define GREEK_LOWER_OMEGA 0x03C9
1.60 +#define GREEK_LOWER_OMEGA_TONOS 0x03CE
1.61 +#define GREEK_LOWER_OMEGA_OXIA 0x1F7D
1.62 +#define GREEK_UPPER_ALPHA 0x0391
1.63 +#define GREEK_UPPER_EPSILON 0x0395
1.64 +#define GREEK_UPPER_ETA 0x0397
1.65 +#define GREEK_UPPER_IOTA 0x0399
1.66 +#define GREEK_UPPER_IOTA_DIALYTIKA 0x03AA
1.67 +#define GREEK_UPPER_OMICRON 0x039F
1.68 +#define GREEK_UPPER_UPSILON 0x03A5
1.69 +#define GREEK_UPPER_UPSILON_DIALYTIKA 0x03AB
1.70 +#define GREEK_UPPER_OMEGA 0x03A9
1.71 +#define GREEK_UPPER_ALPHA_TONOS 0x0386
1.72 +#define GREEK_UPPER_ALPHA_OXIA 0x1FBB
1.73 +#define GREEK_UPPER_EPSILON_TONOS 0x0388
1.74 +#define GREEK_UPPER_EPSILON_OXIA 0x1FC9
1.75 +#define GREEK_UPPER_ETA_TONOS 0x0389
1.76 +#define GREEK_UPPER_ETA_OXIA 0x1FCB
1.77 +#define GREEK_UPPER_IOTA_TONOS 0x038A
1.78 +#define GREEK_UPPER_IOTA_OXIA 0x1FDB
1.79 +#define GREEK_UPPER_OMICRON_TONOS 0x038C
1.80 +#define GREEK_UPPER_OMICRON_OXIA 0x1FF9
1.81 +#define GREEK_UPPER_UPSILON_TONOS 0x038E
1.82 +#define GREEK_UPPER_UPSILON_OXIA 0x1FEB
1.83 +#define GREEK_UPPER_OMEGA_TONOS 0x038F
1.84 +#define GREEK_UPPER_OMEGA_OXIA 0x1FFB
1.85 +#define COMBINING_ACUTE_ACCENT 0x0301
1.86 +#define COMBINING_DIAERESIS 0x0308
1.87 +#define COMBINING_ACUTE_TONE_MARK 0x0341
1.88 +#define COMBINING_GREEK_DIALYTIKA_TONOS 0x0344
1.89 +
1.90 +// When doing an Uppercase transform in Greek, we need to keep track of the
1.91 +// current state while iterating through the string, to recognize and process
1.92 +// diphthongs correctly. For clarity, we define a state for each vowel and
1.93 +// each vowel with accent, although a few of these do not actually need any
1.94 +// special treatment and could be folded into kStart.
1.95 +enum GreekCasingState {
1.96 + kStart,
1.97 + kAlpha,
1.98 + kEpsilon,
1.99 + kEta,
1.100 + kIota,
1.101 + kOmicron,
1.102 + kUpsilon,
1.103 + kOmega,
1.104 + kAlphaAcc,
1.105 + kEpsilonAcc,
1.106 + kEtaAcc,
1.107 + kIotaAcc,
1.108 + kOmicronAcc,
1.109 + kUpsilonAcc,
1.110 + kOmegaAcc,
1.111 + kOmicronUpsilon,
1.112 + kDiaeresis
1.113 +};
1.114 +
1.115 +static uint32_t
1.116 +GreekUpperCase(uint32_t aCh, GreekCasingState* aState)
1.117 +{
1.118 + switch (aCh) {
1.119 + case GREEK_UPPER_ALPHA:
1.120 + case GREEK_LOWER_ALPHA:
1.121 + *aState = kAlpha;
1.122 + return GREEK_UPPER_ALPHA;
1.123 +
1.124 + case GREEK_UPPER_EPSILON:
1.125 + case GREEK_LOWER_EPSILON:
1.126 + *aState = kEpsilon;
1.127 + return GREEK_UPPER_EPSILON;
1.128 +
1.129 + case GREEK_UPPER_ETA:
1.130 + case GREEK_LOWER_ETA:
1.131 + *aState = kEta;
1.132 + return GREEK_UPPER_ETA;
1.133 +
1.134 + case GREEK_UPPER_IOTA:
1.135 + *aState = kIota;
1.136 + return GREEK_UPPER_IOTA;
1.137 +
1.138 + case GREEK_UPPER_OMICRON:
1.139 + case GREEK_LOWER_OMICRON:
1.140 + *aState = kOmicron;
1.141 + return GREEK_UPPER_OMICRON;
1.142 +
1.143 + case GREEK_UPPER_UPSILON:
1.144 + switch (*aState) {
1.145 + case kOmicron:
1.146 + *aState = kOmicronUpsilon;
1.147 + break;
1.148 + default:
1.149 + *aState = kUpsilon;
1.150 + break;
1.151 + }
1.152 + return GREEK_UPPER_UPSILON;
1.153 +
1.154 + case GREEK_UPPER_OMEGA:
1.155 + case GREEK_LOWER_OMEGA:
1.156 + *aState = kOmega;
1.157 + return GREEK_UPPER_OMEGA;
1.158 +
1.159 + // iota and upsilon may be the second vowel of a diphthong
1.160 + case GREEK_LOWER_IOTA:
1.161 + switch (*aState) {
1.162 + case kAlphaAcc:
1.163 + case kEpsilonAcc:
1.164 + case kOmicronAcc:
1.165 + case kUpsilonAcc:
1.166 + *aState = kStart;
1.167 + return GREEK_UPPER_IOTA_DIALYTIKA;
1.168 + default:
1.169 + break;
1.170 + }
1.171 + *aState = kIota;
1.172 + return GREEK_UPPER_IOTA;
1.173 +
1.174 + case GREEK_LOWER_UPSILON:
1.175 + switch (*aState) {
1.176 + case kAlphaAcc:
1.177 + case kEpsilonAcc:
1.178 + case kEtaAcc:
1.179 + case kOmicronAcc:
1.180 + *aState = kStart;
1.181 + return GREEK_UPPER_UPSILON_DIALYTIKA;
1.182 + case kOmicron:
1.183 + *aState = kOmicronUpsilon;
1.184 + break;
1.185 + default:
1.186 + *aState = kUpsilon;
1.187 + break;
1.188 + }
1.189 + return GREEK_UPPER_UPSILON;
1.190 +
1.191 + case GREEK_UPPER_IOTA_DIALYTIKA:
1.192 + case GREEK_LOWER_IOTA_DIALYTIKA:
1.193 + case GREEK_UPPER_UPSILON_DIALYTIKA:
1.194 + case GREEK_LOWER_UPSILON_DIALYTIKA:
1.195 + case COMBINING_DIAERESIS:
1.196 + *aState = kDiaeresis;
1.197 + return ToUpperCase(aCh);
1.198 +
1.199 + // remove accent if it follows a vowel or diaeresis,
1.200 + // and set appropriate state for diphthong detection
1.201 + case COMBINING_ACUTE_ACCENT:
1.202 + case COMBINING_ACUTE_TONE_MARK:
1.203 + switch (*aState) {
1.204 + case kAlpha:
1.205 + *aState = kAlphaAcc;
1.206 + return uint32_t(-1); // omit this char from result string
1.207 + case kEpsilon:
1.208 + *aState = kEpsilonAcc;
1.209 + return uint32_t(-1);
1.210 + case kEta:
1.211 + *aState = kEtaAcc;
1.212 + return uint32_t(-1);
1.213 + case kIota:
1.214 + *aState = kIotaAcc;
1.215 + return uint32_t(-1);
1.216 + case kOmicron:
1.217 + *aState = kOmicronAcc;
1.218 + return uint32_t(-1);
1.219 + case kUpsilon:
1.220 + *aState = kUpsilonAcc;
1.221 + return uint32_t(-1);
1.222 + case kOmicronUpsilon:
1.223 + *aState = kStart; // this completed a diphthong
1.224 + return uint32_t(-1);
1.225 + case kOmega:
1.226 + *aState = kOmegaAcc;
1.227 + return uint32_t(-1);
1.228 + case kDiaeresis:
1.229 + *aState = kStart;
1.230 + return uint32_t(-1);
1.231 + default:
1.232 + break;
1.233 + }
1.234 + break;
1.235 +
1.236 + // combinations with dieresis+accent just strip the accent,
1.237 + // and reset to start state (don't form diphthong with following vowel)
1.238 + case GREEK_LOWER_IOTA_DIALYTIKA_TONOS:
1.239 + case GREEK_LOWER_IOTA_DIALYTIKA_OXIA:
1.240 + *aState = kStart;
1.241 + return GREEK_UPPER_IOTA_DIALYTIKA;
1.242 +
1.243 + case GREEK_LOWER_UPSILON_DIALYTIKA_TONOS:
1.244 + case GREEK_LOWER_UPSILON_DIALYTIKA_OXIA:
1.245 + *aState = kStart;
1.246 + return GREEK_UPPER_UPSILON_DIALYTIKA;
1.247 +
1.248 + case COMBINING_GREEK_DIALYTIKA_TONOS:
1.249 + *aState = kStart;
1.250 + return COMBINING_DIAERESIS;
1.251 +
1.252 + // strip accents from vowels, and note the vowel seen so that we can detect
1.253 + // diphthongs where diaeresis needs to be added
1.254 + case GREEK_LOWER_ALPHA_TONOS:
1.255 + case GREEK_LOWER_ALPHA_OXIA:
1.256 + case GREEK_UPPER_ALPHA_TONOS:
1.257 + case GREEK_UPPER_ALPHA_OXIA:
1.258 + *aState = kAlphaAcc;
1.259 + return GREEK_UPPER_ALPHA;
1.260 +
1.261 + case GREEK_LOWER_EPSILON_TONOS:
1.262 + case GREEK_LOWER_EPSILON_OXIA:
1.263 + case GREEK_UPPER_EPSILON_TONOS:
1.264 + case GREEK_UPPER_EPSILON_OXIA:
1.265 + *aState = kEpsilonAcc;
1.266 + return GREEK_UPPER_EPSILON;
1.267 +
1.268 + case GREEK_LOWER_ETA_TONOS:
1.269 + case GREEK_LOWER_ETA_OXIA:
1.270 + case GREEK_UPPER_ETA_TONOS:
1.271 + case GREEK_UPPER_ETA_OXIA:
1.272 + *aState = kEtaAcc;
1.273 + return GREEK_UPPER_ETA;
1.274 +
1.275 + case GREEK_LOWER_IOTA_TONOS:
1.276 + case GREEK_LOWER_IOTA_OXIA:
1.277 + case GREEK_UPPER_IOTA_TONOS:
1.278 + case GREEK_UPPER_IOTA_OXIA:
1.279 + *aState = kIotaAcc;
1.280 + return GREEK_UPPER_IOTA;
1.281 +
1.282 + case GREEK_LOWER_OMICRON_TONOS:
1.283 + case GREEK_LOWER_OMICRON_OXIA:
1.284 + case GREEK_UPPER_OMICRON_TONOS:
1.285 + case GREEK_UPPER_OMICRON_OXIA:
1.286 + *aState = kOmicronAcc;
1.287 + return GREEK_UPPER_OMICRON;
1.288 +
1.289 + case GREEK_LOWER_UPSILON_TONOS:
1.290 + case GREEK_LOWER_UPSILON_OXIA:
1.291 + case GREEK_UPPER_UPSILON_TONOS:
1.292 + case GREEK_UPPER_UPSILON_OXIA:
1.293 + switch (*aState) {
1.294 + case kOmicron:
1.295 + *aState = kStart; // this completed a diphthong
1.296 + break;
1.297 + default:
1.298 + *aState = kUpsilonAcc;
1.299 + break;
1.300 + }
1.301 + return GREEK_UPPER_UPSILON;
1.302 +
1.303 + case GREEK_LOWER_OMEGA_TONOS:
1.304 + case GREEK_LOWER_OMEGA_OXIA:
1.305 + case GREEK_UPPER_OMEGA_TONOS:
1.306 + case GREEK_UPPER_OMEGA_OXIA:
1.307 + *aState = kOmegaAcc;
1.308 + return GREEK_UPPER_OMEGA;
1.309 + }
1.310 +
1.311 + // all other characters just reset the state, and use standard mappings
1.312 + *aState = kStart;
1.313 + return ToUpperCase(aCh);
1.314 +}
1.315 +
1.316 +nsTransformedTextRun *
1.317 +nsTransformedTextRun::Create(const gfxTextRunFactory::Parameters* aParams,
1.318 + nsTransformingTextRunFactory* aFactory,
1.319 + gfxFontGroup* aFontGroup,
1.320 + const char16_t* aString, uint32_t aLength,
1.321 + const uint32_t aFlags, nsStyleContext** aStyles,
1.322 + bool aOwnsFactory)
1.323 +{
1.324 + NS_ASSERTION(!(aFlags & gfxTextRunFactory::TEXT_IS_8BIT),
1.325 + "didn't expect text to be marked as 8-bit here");
1.326 +
1.327 + void *storage = AllocateStorageForTextRun(sizeof(nsTransformedTextRun), aLength);
1.328 + if (!storage) {
1.329 + return nullptr;
1.330 + }
1.331 +
1.332 + return new (storage) nsTransformedTextRun(aParams, aFactory, aFontGroup,
1.333 + aString, aLength,
1.334 + aFlags, aStyles, aOwnsFactory);
1.335 +}
1.336 +
1.337 +void
1.338 +nsTransformedTextRun::SetCapitalization(uint32_t aStart, uint32_t aLength,
1.339 + bool* aCapitalization,
1.340 + gfxContext* aRefContext)
1.341 +{
1.342 + if (mCapitalize.IsEmpty()) {
1.343 + if (!mCapitalize.AppendElements(GetLength()))
1.344 + return;
1.345 + memset(mCapitalize.Elements(), 0, GetLength()*sizeof(bool));
1.346 + }
1.347 + memcpy(mCapitalize.Elements() + aStart, aCapitalization, aLength*sizeof(bool));
1.348 + mNeedsRebuild = true;
1.349 +}
1.350 +
1.351 +bool
1.352 +nsTransformedTextRun::SetPotentialLineBreaks(uint32_t aStart, uint32_t aLength,
1.353 + uint8_t* aBreakBefore,
1.354 + gfxContext* aRefContext)
1.355 +{
1.356 + bool changed = gfxTextRun::SetPotentialLineBreaks(aStart, aLength,
1.357 + aBreakBefore, aRefContext);
1.358 + if (changed) {
1.359 + mNeedsRebuild = true;
1.360 + }
1.361 + return changed;
1.362 +}
1.363 +
1.364 +size_t
1.365 +nsTransformedTextRun::SizeOfExcludingThis(mozilla::MallocSizeOf aMallocSizeOf)
1.366 +{
1.367 + size_t total = gfxTextRun::SizeOfExcludingThis(aMallocSizeOf);
1.368 + total += mStyles.SizeOfExcludingThis(aMallocSizeOf);
1.369 + total += mCapitalize.SizeOfExcludingThis(aMallocSizeOf);
1.370 + if (mOwnsFactory) {
1.371 + total += aMallocSizeOf(mFactory);
1.372 + }
1.373 + return total;
1.374 +}
1.375 +
1.376 +size_t
1.377 +nsTransformedTextRun::SizeOfIncludingThis(mozilla::MallocSizeOf aMallocSizeOf)
1.378 +{
1.379 + return aMallocSizeOf(this) + SizeOfExcludingThis(aMallocSizeOf);
1.380 +}
1.381 +
1.382 +nsTransformedTextRun*
1.383 +nsTransformingTextRunFactory::MakeTextRun(const char16_t* aString, uint32_t aLength,
1.384 + const gfxTextRunFactory::Parameters* aParams,
1.385 + gfxFontGroup* aFontGroup, uint32_t aFlags,
1.386 + nsStyleContext** aStyles, bool aOwnsFactory)
1.387 +{
1.388 + return nsTransformedTextRun::Create(aParams, this, aFontGroup,
1.389 + aString, aLength, aFlags, aStyles, aOwnsFactory);
1.390 +}
1.391 +
1.392 +nsTransformedTextRun*
1.393 +nsTransformingTextRunFactory::MakeTextRun(const uint8_t* aString, uint32_t aLength,
1.394 + const gfxTextRunFactory::Parameters* aParams,
1.395 + gfxFontGroup* aFontGroup, uint32_t aFlags,
1.396 + nsStyleContext** aStyles, bool aOwnsFactory)
1.397 +{
1.398 + // We'll only have a Unicode code path to minimize the amount of code needed
1.399 + // for these rarely used features
1.400 + NS_ConvertASCIItoUTF16 unicodeString(reinterpret_cast<const char*>(aString), aLength);
1.401 + return MakeTextRun(unicodeString.get(), aLength, aParams, aFontGroup,
1.402 + aFlags & ~(gfxFontGroup::TEXT_IS_PERSISTENT | gfxFontGroup::TEXT_IS_8BIT),
1.403 + aStyles, aOwnsFactory);
1.404 +}
1.405 +
1.406 +void
1.407 +MergeCharactersInTextRun(gfxTextRun* aDest, gfxTextRun* aSrc,
1.408 + const bool* aCharsToMerge, const bool* aDeletedChars)
1.409 +{
1.410 + aDest->ResetGlyphRuns();
1.411 +
1.412 + gfxTextRun::GlyphRunIterator iter(aSrc, 0, aSrc->GetLength());
1.413 + uint32_t offset = 0;
1.414 + nsAutoTArray<gfxTextRun::DetailedGlyph,2> glyphs;
1.415 + while (iter.NextRun()) {
1.416 + gfxTextRun::GlyphRun* run = iter.GetGlyphRun();
1.417 + nsresult rv = aDest->AddGlyphRun(run->mFont, run->mMatchType,
1.418 + offset, false);
1.419 + if (NS_FAILED(rv))
1.420 + return;
1.421 +
1.422 + bool anyMissing = false;
1.423 + uint32_t mergeRunStart = iter.GetStringStart();
1.424 + const gfxTextRun::CompressedGlyph *srcGlyphs = aSrc->GetCharacterGlyphs();
1.425 + gfxTextRun::CompressedGlyph mergedGlyph = srcGlyphs[mergeRunStart];
1.426 + uint32_t stringEnd = iter.GetStringEnd();
1.427 + for (uint32_t k = iter.GetStringStart(); k < stringEnd; ++k) {
1.428 + const gfxTextRun::CompressedGlyph g = srcGlyphs[k];
1.429 + if (g.IsSimpleGlyph()) {
1.430 + if (!anyMissing) {
1.431 + gfxTextRun::DetailedGlyph details;
1.432 + details.mGlyphID = g.GetSimpleGlyph();
1.433 + details.mAdvance = g.GetSimpleAdvance();
1.434 + details.mXOffset = 0;
1.435 + details.mYOffset = 0;
1.436 + glyphs.AppendElement(details);
1.437 + }
1.438 + } else {
1.439 + if (g.IsMissing()) {
1.440 + anyMissing = true;
1.441 + glyphs.Clear();
1.442 + }
1.443 + if (g.GetGlyphCount() > 0) {
1.444 + glyphs.AppendElements(aSrc->GetDetailedGlyphs(k), g.GetGlyphCount());
1.445 + }
1.446 + }
1.447 +
1.448 + if (k + 1 < iter.GetStringEnd() && aCharsToMerge[k + 1]) {
1.449 + // next char is supposed to merge with current, so loop without
1.450 + // writing current merged glyph to the destination
1.451 + continue;
1.452 + }
1.453 +
1.454 + // If the start of the merge run is actually a character that should
1.455 + // have been merged with the previous character (this can happen
1.456 + // if there's a font change in the middle of a case-mapped character,
1.457 + // that decomposed into a sequence of base+diacritics, for example),
1.458 + // just discard the entire merge run. See comment at start of this
1.459 + // function.
1.460 + NS_WARN_IF_FALSE(!aCharsToMerge[mergeRunStart],
1.461 + "unable to merge across a glyph run boundary, "
1.462 + "glyph(s) discarded");
1.463 + if (!aCharsToMerge[mergeRunStart]) {
1.464 + if (anyMissing) {
1.465 + mergedGlyph.SetMissing(glyphs.Length());
1.466 + } else {
1.467 + mergedGlyph.SetComplex(mergedGlyph.IsClusterStart(),
1.468 + mergedGlyph.IsLigatureGroupStart(),
1.469 + glyphs.Length());
1.470 + }
1.471 + aDest->SetGlyphs(offset, mergedGlyph, glyphs.Elements());
1.472 + ++offset;
1.473 +
1.474 + while (offset < aDest->GetLength() && aDeletedChars[offset]) {
1.475 + aDest->SetGlyphs(offset++, gfxTextRun::CompressedGlyph(), nullptr);
1.476 + }
1.477 + }
1.478 +
1.479 + glyphs.Clear();
1.480 + anyMissing = false;
1.481 + mergeRunStart = k + 1;
1.482 + if (mergeRunStart < stringEnd) {
1.483 + mergedGlyph = srcGlyphs[mergeRunStart];
1.484 + }
1.485 + }
1.486 + NS_ASSERTION(glyphs.Length() == 0,
1.487 + "Leftover glyphs, don't request merging of the last character with its next!");
1.488 + }
1.489 + NS_ASSERTION(offset == aDest->GetLength(), "Bad offset calculations");
1.490 +}
1.491 +
1.492 +gfxTextRunFactory::Parameters
1.493 +GetParametersForInner(nsTransformedTextRun* aTextRun, uint32_t* aFlags,
1.494 + gfxContext* aRefContext)
1.495 +{
1.496 + gfxTextRunFactory::Parameters params =
1.497 + { aRefContext, nullptr, nullptr,
1.498 + nullptr, 0, aTextRun->GetAppUnitsPerDevUnit()
1.499 + };
1.500 + *aFlags = aTextRun->GetFlags() & ~gfxFontGroup::TEXT_IS_PERSISTENT;
1.501 + return params;
1.502 +}
1.503 +
1.504 +void
1.505 +nsFontVariantTextRunFactory::RebuildTextRun(nsTransformedTextRun* aTextRun,
1.506 + gfxContext* aRefContext)
1.507 +{
1.508 + gfxFontGroup* fontGroup = aTextRun->GetFontGroup();
1.509 + gfxFontStyle fontStyle = *fontGroup->GetStyle();
1.510 + fontStyle.size *= 0.8;
1.511 + nsRefPtr<gfxFontGroup> smallFont = fontGroup->Copy(&fontStyle);
1.512 + if (!smallFont)
1.513 + return;
1.514 +
1.515 + uint32_t flags;
1.516 + gfxTextRunFactory::Parameters innerParams =
1.517 + GetParametersForInner(aTextRun, &flags, aRefContext);
1.518 +
1.519 + uint32_t length = aTextRun->GetLength();
1.520 + const char16_t* str = aTextRun->mString.BeginReading();
1.521 + nsRefPtr<nsStyleContext>* styles = aTextRun->mStyles.Elements();
1.522 + // Create a textrun so we can check cluster-start properties
1.523 + nsAutoPtr<gfxTextRun> inner(fontGroup->MakeTextRun(str, length, &innerParams, flags));
1.524 + if (!inner.get())
1.525 + return;
1.526 +
1.527 + nsCaseTransformTextRunFactory uppercaseFactory(nullptr, true);
1.528 +
1.529 + aTextRun->ResetGlyphRuns();
1.530 +
1.531 + uint32_t runStart = 0;
1.532 + nsAutoTArray<nsStyleContext*,50> styleArray;
1.533 + nsAutoTArray<uint8_t,50> canBreakBeforeArray;
1.534 +
1.535 + enum RunCaseState {
1.536 + kUpperOrCaseless, // will be untouched by font-variant:small-caps
1.537 + kLowercase, // will be uppercased and reduced
1.538 + kSpecialUpper // specials: don't shrink, but apply uppercase mapping
1.539 + };
1.540 + RunCaseState runCase = kUpperOrCaseless;
1.541 +
1.542 + // Note that this loop runs from 0 to length *inclusive*, so the last
1.543 + // iteration is in effect beyond the end of the input text, to give a
1.544 + // chance to finish the last casing run we've found.
1.545 + // The last iteration, when i==length, must not attempt to look at the
1.546 + // character position [i] or the style data for styles[i], as this would
1.547 + // be beyond the valid length of the textrun or its style array.
1.548 + for (uint32_t i = 0; i <= length; ++i) {
1.549 + RunCaseState chCase = kUpperOrCaseless;
1.550 + // Unless we're at the end, figure out what treatment the current
1.551 + // character will need.
1.552 + if (i < length) {
1.553 + nsStyleContext* styleContext = styles[i];
1.554 + // Characters that aren't the start of a cluster are ignored here. They
1.555 + // get added to whatever lowercase/non-lowercase run we're in.
1.556 + if (!inner->IsClusterStart(i)) {
1.557 + chCase = runCase;
1.558 + } else {
1.559 + if (styleContext->StyleFont()->mFont.variant == NS_STYLE_FONT_VARIANT_SMALL_CAPS) {
1.560 + uint32_t ch = str[i];
1.561 + if (NS_IS_HIGH_SURROGATE(ch) && i < length - 1 && NS_IS_LOW_SURROGATE(str[i + 1])) {
1.562 + ch = SURROGATE_TO_UCS4(ch, str[i + 1]);
1.563 + }
1.564 + uint32_t ch2 = ToUpperCase(ch);
1.565 + if (ch != ch2 || mozilla::unicode::SpecialUpper(ch)) {
1.566 + chCase = kLowercase;
1.567 + } else if (styleContext->StyleFont()->mLanguage == nsGkAtoms::el) {
1.568 + // In Greek, check for characters that will be modified by the
1.569 + // GreekUpperCase mapping - this catches accented capitals where
1.570 + // the accent is to be removed (bug 307039). These are handled by
1.571 + // a transformed child run using the full-size font.
1.572 + GreekCasingState state = kStart; // don't need exact context here
1.573 + ch2 = GreekUpperCase(ch, &state);
1.574 + if (ch != ch2) {
1.575 + chCase = kSpecialUpper;
1.576 + }
1.577 + }
1.578 + } else {
1.579 + // Don't transform the character! I.e., pretend that it's not lowercase
1.580 + }
1.581 + }
1.582 + }
1.583 +
1.584 + // At the end of the text, or when the current character needs different
1.585 + // casing treatment from the current run, finish the run-in-progress
1.586 + // and prepare to accumulate a new run.
1.587 + // Note that we do not look at any source data for offset [i] here,
1.588 + // as that would be invalid in the case where i==length.
1.589 + if ((i == length || runCase != chCase) && runStart < i) {
1.590 + nsAutoPtr<nsTransformedTextRun> transformedChild;
1.591 + nsAutoPtr<gfxTextRun> cachedChild;
1.592 + gfxTextRun* child;
1.593 +
1.594 + switch (runCase) {
1.595 + case kUpperOrCaseless:
1.596 + cachedChild =
1.597 + fontGroup->MakeTextRun(str + runStart, i - runStart, &innerParams,
1.598 + flags);
1.599 + child = cachedChild.get();
1.600 + break;
1.601 + case kLowercase:
1.602 + transformedChild =
1.603 + uppercaseFactory.MakeTextRun(str + runStart, i - runStart,
1.604 + &innerParams, smallFont, flags,
1.605 + styleArray.Elements(), false);
1.606 + child = transformedChild;
1.607 + break;
1.608 + case kSpecialUpper:
1.609 + transformedChild =
1.610 + uppercaseFactory.MakeTextRun(str + runStart, i - runStart,
1.611 + &innerParams, fontGroup, flags,
1.612 + styleArray.Elements(), false);
1.613 + child = transformedChild;
1.614 + break;
1.615 + }
1.616 + if (!child)
1.617 + return;
1.618 + // Copy potential linebreaks into child so they're preserved
1.619 + // (and also child will be shaped appropriately)
1.620 + NS_ASSERTION(canBreakBeforeArray.Length() == i - runStart,
1.621 + "lost some break-before values?");
1.622 + child->SetPotentialLineBreaks(0, canBreakBeforeArray.Length(),
1.623 + canBreakBeforeArray.Elements(), aRefContext);
1.624 + if (transformedChild) {
1.625 + transformedChild->FinishSettingProperties(aRefContext);
1.626 + }
1.627 + aTextRun->CopyGlyphDataFrom(child, 0, child->GetLength(), runStart);
1.628 +
1.629 + runStart = i;
1.630 + styleArray.Clear();
1.631 + canBreakBeforeArray.Clear();
1.632 + }
1.633 +
1.634 + if (i < length) {
1.635 + runCase = chCase;
1.636 + styleArray.AppendElement(styles[i]);
1.637 + canBreakBeforeArray.AppendElement(aTextRun->CanBreakLineBefore(i));
1.638 + }
1.639 + }
1.640 +}
1.641 +
1.642 +void
1.643 +nsCaseTransformTextRunFactory::RebuildTextRun(nsTransformedTextRun* aTextRun,
1.644 + gfxContext* aRefContext)
1.645 +{
1.646 + uint32_t length = aTextRun->GetLength();
1.647 + const char16_t* str = aTextRun->mString.BeginReading();
1.648 + nsRefPtr<nsStyleContext>* styles = aTextRun->mStyles.Elements();
1.649 +
1.650 + nsAutoString convertedString;
1.651 + nsAutoTArray<bool,50> charsToMergeArray;
1.652 + nsAutoTArray<bool,50> deletedCharsArray;
1.653 + nsAutoTArray<nsStyleContext*,50> styleArray;
1.654 + nsAutoTArray<uint8_t,50> canBreakBeforeArray;
1.655 + bool mergeNeeded = false;
1.656 +
1.657 + // Some languages have special casing conventions that differ from the
1.658 + // default Unicode mappings.
1.659 + // The enum values here are named for well-known exemplar languages that
1.660 + // exhibit the behavior in question; multiple lang tags may map to the
1.661 + // same setting here, if the behavior is shared by other languages.
1.662 + enum {
1.663 + eNone, // default non-lang-specific behavior
1.664 + eTurkish, // preserve dotted/dotless-i distinction in uppercase
1.665 + eDutch, // treat "ij" digraph as a unit for capitalization
1.666 + eGreek // strip accent when uppercasing Greek vowels
1.667 + } languageSpecificCasing = eNone;
1.668 +
1.669 + const nsIAtom* lang = nullptr;
1.670 + bool capitalizeDutchIJ = false;
1.671 + bool prevIsLetter = false;
1.672 + uint32_t sigmaIndex = uint32_t(-1);
1.673 + nsIUGenCategory::nsUGenCategory cat;
1.674 + GreekCasingState greekState = kStart;
1.675 + uint32_t i;
1.676 + for (i = 0; i < length; ++i) {
1.677 + uint32_t ch = str[i];
1.678 + nsStyleContext* styleContext = styles[i];
1.679 +
1.680 + uint8_t style = mAllUppercase ? NS_STYLE_TEXT_TRANSFORM_UPPERCASE
1.681 + : styleContext->StyleText()->mTextTransform;
1.682 + int extraChars = 0;
1.683 + const mozilla::unicode::MultiCharMapping *mcm;
1.684 +
1.685 + if (NS_IS_HIGH_SURROGATE(ch) && i < length - 1 && NS_IS_LOW_SURROGATE(str[i + 1])) {
1.686 + ch = SURROGATE_TO_UCS4(ch, str[i + 1]);
1.687 + }
1.688 +
1.689 + if (lang != styleContext->StyleFont()->mLanguage) {
1.690 + lang = styleContext->StyleFont()->mLanguage;
1.691 + if (lang == nsGkAtoms::tr || lang == nsGkAtoms::az ||
1.692 + lang == nsGkAtoms::ba || lang == nsGkAtoms::crh ||
1.693 + lang == nsGkAtoms::tt) {
1.694 + languageSpecificCasing = eTurkish;
1.695 + } else if (lang == nsGkAtoms::nl) {
1.696 + languageSpecificCasing = eDutch;
1.697 + } else if (lang == nsGkAtoms::el) {
1.698 + languageSpecificCasing = eGreek;
1.699 + greekState = kStart;
1.700 + } else {
1.701 + languageSpecificCasing = eNone;
1.702 + }
1.703 + }
1.704 +
1.705 + switch (style) {
1.706 + case NS_STYLE_TEXT_TRANSFORM_LOWERCASE:
1.707 + if (languageSpecificCasing == eTurkish) {
1.708 + if (ch == 'I') {
1.709 + ch = LATIN_SMALL_LETTER_DOTLESS_I;
1.710 + prevIsLetter = true;
1.711 + sigmaIndex = uint32_t(-1);
1.712 + break;
1.713 + }
1.714 + if (ch == LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE) {
1.715 + ch = 'i';
1.716 + prevIsLetter = true;
1.717 + sigmaIndex = uint32_t(-1);
1.718 + break;
1.719 + }
1.720 + }
1.721 +
1.722 + // Special lowercasing behavior for Greek Sigma: note that this is listed
1.723 + // as context-sensitive in Unicode's SpecialCasing.txt, but is *not* a
1.724 + // language-specific mapping; it applies regardless of the language of
1.725 + // the element.
1.726 + //
1.727 + // The lowercase mapping for CAPITAL SIGMA should be to SMALL SIGMA (i.e.
1.728 + // the non-final form) whenever there is a following letter, or when the
1.729 + // CAPITAL SIGMA occurs in isolation (neither preceded nor followed by a
1.730 + // LETTER); and to FINAL SIGMA when it is preceded by another letter but
1.731 + // not followed by one.
1.732 + //
1.733 + // To implement the context-sensitive nature of this mapping, we keep
1.734 + // track of whether the previous character was a letter. If not, CAPITAL
1.735 + // SIGMA will map directly to SMALL SIGMA. If the previous character
1.736 + // was a letter, CAPITAL SIGMA maps to FINAL SIGMA and we record the
1.737 + // position in the converted string; if we then encounter another letter,
1.738 + // that FINAL SIGMA is replaced with a standard SMALL SIGMA.
1.739 +
1.740 + cat = mozilla::unicode::GetGenCategory(ch);
1.741 +
1.742 + // If sigmaIndex is not -1, it marks where we have provisionally mapped
1.743 + // a CAPITAL SIGMA to FINAL SIGMA; if we now find another letter, we
1.744 + // need to change it to SMALL SIGMA.
1.745 + if (sigmaIndex != uint32_t(-1)) {
1.746 + if (cat == nsIUGenCategory::kLetter) {
1.747 + convertedString.SetCharAt(GREEK_SMALL_LETTER_SIGMA, sigmaIndex);
1.748 + }
1.749 + }
1.750 +
1.751 + if (ch == GREEK_CAPITAL_LETTER_SIGMA) {
1.752 + // If preceding char was a letter, map to FINAL instead of SMALL,
1.753 + // and note where it occurred by setting sigmaIndex; we'll change it
1.754 + // to standard SMALL SIGMA later if another letter follows
1.755 + if (prevIsLetter) {
1.756 + ch = GREEK_SMALL_LETTER_FINAL_SIGMA;
1.757 + sigmaIndex = convertedString.Length();
1.758 + } else {
1.759 + // CAPITAL SIGMA not preceded by a letter is unconditionally mapped
1.760 + // to SMALL SIGMA
1.761 + ch = GREEK_SMALL_LETTER_SIGMA;
1.762 + sigmaIndex = uint32_t(-1);
1.763 + }
1.764 + prevIsLetter = true;
1.765 + break;
1.766 + }
1.767 +
1.768 + // ignore diacritics for the purpose of contextual sigma mapping;
1.769 + // otherwise, reset prevIsLetter appropriately and clear the
1.770 + // sigmaIndex marker
1.771 + if (cat != nsIUGenCategory::kMark) {
1.772 + prevIsLetter = (cat == nsIUGenCategory::kLetter);
1.773 + sigmaIndex = uint32_t(-1);
1.774 + }
1.775 +
1.776 + mcm = mozilla::unicode::SpecialLower(ch);
1.777 + if (mcm) {
1.778 + int j = 0;
1.779 + while (j < 2 && mcm->mMappedChars[j + 1]) {
1.780 + convertedString.Append(mcm->mMappedChars[j]);
1.781 + ++extraChars;
1.782 + ++j;
1.783 + }
1.784 + ch = mcm->mMappedChars[j];
1.785 + break;
1.786 + }
1.787 +
1.788 + ch = ToLowerCase(ch);
1.789 + break;
1.790 +
1.791 + case NS_STYLE_TEXT_TRANSFORM_UPPERCASE:
1.792 + if (languageSpecificCasing == eTurkish && ch == 'i') {
1.793 + ch = LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE;
1.794 + break;
1.795 + }
1.796 +
1.797 + if (languageSpecificCasing == eGreek) {
1.798 + ch = GreekUpperCase(ch, &greekState);
1.799 + break;
1.800 + }
1.801 +
1.802 + mcm = mozilla::unicode::SpecialUpper(ch);
1.803 + if (mcm) {
1.804 + int j = 0;
1.805 + while (j < 2 && mcm->mMappedChars[j + 1]) {
1.806 + convertedString.Append(mcm->mMappedChars[j]);
1.807 + ++extraChars;
1.808 + ++j;
1.809 + }
1.810 + ch = mcm->mMappedChars[j];
1.811 + break;
1.812 + }
1.813 +
1.814 + ch = ToUpperCase(ch);
1.815 + break;
1.816 +
1.817 + case NS_STYLE_TEXT_TRANSFORM_CAPITALIZE:
1.818 + if (capitalizeDutchIJ && ch == 'j') {
1.819 + ch = 'J';
1.820 + capitalizeDutchIJ = false;
1.821 + break;
1.822 + }
1.823 + capitalizeDutchIJ = false;
1.824 + if (i < aTextRun->mCapitalize.Length() && aTextRun->mCapitalize[i]) {
1.825 + if (languageSpecificCasing == eTurkish && ch == 'i') {
1.826 + ch = LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE;
1.827 + break;
1.828 + }
1.829 + if (languageSpecificCasing == eDutch && ch == 'i') {
1.830 + ch = 'I';
1.831 + capitalizeDutchIJ = true;
1.832 + break;
1.833 + }
1.834 +
1.835 + mcm = mozilla::unicode::SpecialTitle(ch);
1.836 + if (mcm) {
1.837 + int j = 0;
1.838 + while (j < 2 && mcm->mMappedChars[j + 1]) {
1.839 + convertedString.Append(mcm->mMappedChars[j]);
1.840 + ++extraChars;
1.841 + ++j;
1.842 + }
1.843 + ch = mcm->mMappedChars[j];
1.844 + break;
1.845 + }
1.846 +
1.847 + ch = ToTitleCase(ch);
1.848 + }
1.849 + break;
1.850 +
1.851 + case NS_STYLE_TEXT_TRANSFORM_FULLWIDTH:
1.852 + ch = mozilla::unicode::GetFullWidth(ch);
1.853 + break;
1.854 +
1.855 + default:
1.856 + break;
1.857 + }
1.858 +
1.859 + if (ch == uint32_t(-1)) {
1.860 + deletedCharsArray.AppendElement(true);
1.861 + mergeNeeded = true;
1.862 + } else {
1.863 + deletedCharsArray.AppendElement(false);
1.864 + charsToMergeArray.AppendElement(false);
1.865 + styleArray.AppendElement(styleContext);
1.866 + canBreakBeforeArray.AppendElement(aTextRun->CanBreakLineBefore(i));
1.867 +
1.868 + if (IS_IN_BMP(ch)) {
1.869 + convertedString.Append(ch);
1.870 + } else {
1.871 + convertedString.Append(H_SURROGATE(ch));
1.872 + convertedString.Append(L_SURROGATE(ch));
1.873 + ++i;
1.874 + deletedCharsArray.AppendElement(true); // not exactly deleted, but the
1.875 + // trailing surrogate is skipped
1.876 + ++extraChars;
1.877 + }
1.878 +
1.879 + while (extraChars-- > 0) {
1.880 + mergeNeeded = true;
1.881 + charsToMergeArray.AppendElement(true);
1.882 + styleArray.AppendElement(styleContext);
1.883 + canBreakBeforeArray.AppendElement(false);
1.884 + }
1.885 + }
1.886 + }
1.887 +
1.888 + uint32_t flags;
1.889 + gfxTextRunFactory::Parameters innerParams =
1.890 + GetParametersForInner(aTextRun, &flags, aRefContext);
1.891 + gfxFontGroup* fontGroup = aTextRun->GetFontGroup();
1.892 +
1.893 + nsAutoPtr<nsTransformedTextRun> transformedChild;
1.894 + nsAutoPtr<gfxTextRun> cachedChild;
1.895 + gfxTextRun* child;
1.896 +
1.897 + if (mInnerTransformingTextRunFactory) {
1.898 + transformedChild = mInnerTransformingTextRunFactory->MakeTextRun(
1.899 + convertedString.BeginReading(), convertedString.Length(),
1.900 + &innerParams, fontGroup, flags, styleArray.Elements(), false);
1.901 + child = transformedChild.get();
1.902 + } else {
1.903 + cachedChild = fontGroup->MakeTextRun(
1.904 + convertedString.BeginReading(), convertedString.Length(),
1.905 + &innerParams, flags);
1.906 + child = cachedChild.get();
1.907 + }
1.908 + if (!child)
1.909 + return;
1.910 + // Copy potential linebreaks into child so they're preserved
1.911 + // (and also child will be shaped appropriately)
1.912 + NS_ASSERTION(convertedString.Length() == canBreakBeforeArray.Length(),
1.913 + "Dropped characters or break-before values somewhere!");
1.914 + child->SetPotentialLineBreaks(0, canBreakBeforeArray.Length(),
1.915 + canBreakBeforeArray.Elements(), aRefContext);
1.916 + if (transformedChild) {
1.917 + transformedChild->FinishSettingProperties(aRefContext);
1.918 + }
1.919 +
1.920 + if (mergeNeeded) {
1.921 + // Now merge multiple characters into one multi-glyph character as required
1.922 + // and deal with skipping deleted accent chars
1.923 + NS_ASSERTION(charsToMergeArray.Length() == child->GetLength(),
1.924 + "source length mismatch");
1.925 + NS_ASSERTION(deletedCharsArray.Length() == aTextRun->GetLength(),
1.926 + "destination length mismatch");
1.927 + MergeCharactersInTextRun(aTextRun, child, charsToMergeArray.Elements(),
1.928 + deletedCharsArray.Elements());
1.929 + } else {
1.930 + // No merging to do, so just copy; this produces a more optimized textrun.
1.931 + // We can't steal the data because the child may be cached and stealing
1.932 + // the data would break the cache.
1.933 + aTextRun->ResetGlyphRuns();
1.934 + aTextRun->CopyGlyphDataFrom(child, 0, child->GetLength(), 0);
1.935 + }
1.936 +}
