Bug 1316482 - use the refactored TransformText as a template function for both char16_t and uint8_t text.
With this patch, we shall only maintain one version of the TransformText logic.
MozReview-Commit-ID: JAIksFVqvqf
--- a/layout/generic/nsTextFrameUtils.cpp
+++ b/layout/generic/nsTextFrameUtils.cpp
@@ -51,34 +51,39 @@ IsSpaceOrTab(char16_t aCh)
}
static bool
IsSpaceOrTabOrSegmentBreak(char16_t aCh)
{
return IsSpaceOrTab(aCh) || IsSegmentBreak(aCh);
}
-static char16_t*
-TransformWhiteSpaces(const char16_t* aText, uint32_t aLength,
+template<class CharT>
+static CharT*
+TransformWhiteSpaces(const CharT* aText, uint32_t aLength,
uint32_t aBegin, uint32_t aEnd,
bool aHasSegmentBreak,
bool& aInWhitespace,
- char16_t* aOutput,
+ CharT* aOutput,
uint32_t& aFlags,
nsTextFrameUtils::CompressionMode aCompression,
gfxSkipChars* aSkipChars)
{
MOZ_ASSERT(aCompression == nsTextFrameUtils::COMPRESS_WHITESPACE ||
aCompression == nsTextFrameUtils::COMPRESS_WHITESPACE_NEWLINE,
"whitespaces should be skippable!!");
// Get the context preceding/following this white space range.
+ // For 8-bit text (sizeof CharT == 1), the checks here should get optimized
+ // out, and isSegmentBreakSkippable should be initialized to be 'false'.
bool isSegmentBreakSkippable =
- (aBegin > 0 && IS_ZERO_WIDTH_SPACE(aText[aBegin - 1])) ||
- (aEnd < aLength && IS_ZERO_WIDTH_SPACE(aText[aEnd]));
- if (!isSegmentBreakSkippable && aBegin > 0 && aEnd < aLength) {
+ sizeof(CharT) > 1 &&
+ ((aBegin > 0 && IS_ZERO_WIDTH_SPACE(aText[aBegin - 1])) ||
+ (aEnd < aLength && IS_ZERO_WIDTH_SPACE(aText[aEnd])));
+ if (sizeof(CharT) > 1 && !isSegmentBreakSkippable &&
+ aBegin > 0 && aEnd < aLength) {
uint32_t ucs4before;
uint32_t ucs4after;
if (aBegin > 1 &&
NS_IS_LOW_SURROGATE(aText[aBegin - 1]) &&
NS_IS_HIGH_SURROGATE(aText[aBegin - 2])) {
ucs4before = SURROGATE_TO_UCS4(aText[aBegin - 2], aText[aBegin - 1]);
} else {
ucs4before = aText[aBegin - 1];
@@ -92,17 +97,17 @@ TransformWhiteSpaces(const char16_t* aTe
}
// Discard newlines between characters that have F, W, or H
// EastAsianWidth property and neither side is Hangul.
isSegmentBreakSkippable = IsSegmentBreakSkipChar(ucs4before) &&
IsSegmentBreakSkipChar(ucs4after);
}
for (uint32_t i = aBegin; i < aEnd; ++i) {
- char16_t ch = aText[i];
+ CharT ch = aText[i];
bool keepChar = false;
bool keepTransformedWhiteSpace = false;
if (IsDiscardable(ch, &aFlags)) {
aSkipChars->SkipChar();
continue;
}
if (IsSpaceOrTab(ch)) {
if (aHasSegmentBreak) {
@@ -162,34 +167,35 @@ TransformWhiteSpaces(const char16_t* aTe
aInWhitespace = true;
} else {
MOZ_ASSERT_UNREACHABLE("Should've skipped the character!!");
}
}
return aOutput;
}
-char16_t*
-nsTextFrameUtils::TransformText(const char16_t* aText, uint32_t aLength,
- char16_t* aOutput,
+template<class CharT>
+CharT*
+nsTextFrameUtils::TransformText(const CharT* aText, uint32_t aLength,
+ CharT* aOutput,
CompressionMode aCompression,
uint8_t* aIncomingFlags,
gfxSkipChars* aSkipChars,
uint32_t* aAnalysisFlags)
{
uint32_t flags = 0;
- char16_t* outputStart = aOutput;
+ CharT* outputStart = aOutput;
bool lastCharArabic = false;
if (aCompression == COMPRESS_NONE ||
aCompression == COMPRESS_NONE_TRANSFORM_TO_SPACE) {
// Skip discardables.
uint32_t i;
for (i = 0; i < aLength; ++i) {
- char16_t ch = aText[i];
+ CharT ch = aText[i];
if (IsDiscardable(ch, &flags)) {
aSkipChars->SkipChar();
} else {
aSkipChars->KeepChar();
if (ch > ' ') {
lastCharArabic = IS_ARABIC_CHAR(ch);
} else if (aCompression == COMPRESS_NONE_TRANSFORM_TO_SPACE) {
if (ch == '\t' || ch == '\n') {
@@ -210,17 +216,17 @@ nsTextFrameUtils::TransformText(const ch
} else {
*aIncomingFlags &= ~INCOMING_ARABICCHAR;
}
*aIncomingFlags &= ~INCOMING_WHITESPACE;
} else {
bool inWhitespace = (*aIncomingFlags & INCOMING_WHITESPACE) != 0;
uint32_t i;
for (i = 0; i < aLength; ++i) {
- char16_t ch = aText[i];
+ CharT ch = aText[i];
// CSS Text 3 - 4.1. The White Space Processing Rules
// White space processing in CSS affects only the document white space
// characters: spaces (U+0020), tabs (U+0009), and segment breaks.
// Since we need the context of segment breaks and their surrounding
// white spaces to proceed the white space processing, a consecutive run
// of spaces/tabs/segment breaks is collected in a first pass loop, then
// we apply the collapsing and transformation rules to this run in a
// second pass loop.
@@ -239,17 +245,17 @@ nsTextFrameUtils::TransformText(const ch
}
// Exclude trailing discardables before checking space combining
// sequence tail.
for (; IsDiscardable(aText[j - 1], &flags); j--) {
countTrailingDiscardables++;
}
// If the last white space is followed by a combining sequence tail,
// exclude it from the range of TransformWhiteSpaces.
- if (aText[j - 1] == ' ' && j < aLength &&
+ if (sizeof(CharT) > 1 && aText[j - 1] == ' ' && j < aLength &&
IsSpaceCombiningSequenceTail(&aText[j], aLength - j)) {
keepLastSpace = true;
j--;
}
if (j > i) {
aOutput = TransformWhiteSpaces(aText, aLength, i, j, hasSegmentBreak,
inWhitespace, aOutput, flags,
aCompression, aSkipChars);
@@ -294,95 +300,38 @@ nsTextFrameUtils::TransformText(const ch
}
if (outputStart + aLength != aOutput) {
flags |= TEXT_WAS_TRANSFORMED;
}
*aAnalysisFlags = flags;
return aOutput;
}
-
-uint8_t*
+/*
+ * NOTE: This template is part of public API of nsTextFrameUtils, while
+ * its function body is not available in the header. It may stop working
+ * (fail to resolve symbol in link time) once its callsites are moved to a
+ * different translation unit (e.g. a different unified source file).
+ * Explicit instantiating this function template with `uint8_t` and `char16_t`
+ * could prevent us from the potential risk.
+ */
+template uint8_t*
nsTextFrameUtils::TransformText(const uint8_t* aText, uint32_t aLength,
uint8_t* aOutput,
CompressionMode aCompression,
uint8_t* aIncomingFlags,
gfxSkipChars* aSkipChars,
- uint32_t* aAnalysisFlags)
-{
- uint32_t flags = 0;
- uint8_t* outputStart = aOutput;
-
- if (aCompression == COMPRESS_NONE ||
- aCompression == COMPRESS_NONE_TRANSFORM_TO_SPACE) {
- // Skip discardables.
- uint32_t i;
- for (i = 0; i < aLength; ++i) {
- uint8_t ch = aText[i];
- if (IsDiscardable(ch, &flags)) {
- aSkipChars->SkipChar();
- } else {
- aSkipChars->KeepChar();
- if (aCompression == COMPRESS_NONE_TRANSFORM_TO_SPACE) {
- if (ch == '\t' || ch == '\n') {
- ch = ' ';
- flags |= TEXT_WAS_TRANSFORMED;
- }
- } else {
- // aCompression == COMPRESS_NONE
- if (ch == '\t') {
- flags |= TEXT_HAS_TAB;
- }
- }
- *aOutput++ = ch;
- }
- }
- *aIncomingFlags &= ~(INCOMING_ARABICCHAR | INCOMING_WHITESPACE);
- } else {
- bool inWhitespace = (*aIncomingFlags & INCOMING_WHITESPACE) != 0;
- uint32_t i;
- for (i = 0; i < aLength; ++i) {
- uint8_t ch = aText[i];
- bool nowInWhitespace = ch == ' ' || ch == '\t' ||
- (ch == '\n' && aCompression == COMPRESS_WHITESPACE_NEWLINE);
- if (!nowInWhitespace) {
- if (IsDiscardable(ch, &flags)) {
- aSkipChars->SkipChar();
- nowInWhitespace = inWhitespace;
- } else {
- *aOutput++ = ch;
- aSkipChars->KeepChar();
- }
- } else {
- if (inWhitespace) {
- aSkipChars->SkipChar();
- } else {
- if (ch != ' ') {
- flags |= TEXT_WAS_TRANSFORMED;
- }
- *aOutput++ = ' ';
- aSkipChars->KeepChar();
- }
- }
- inWhitespace = nowInWhitespace;
- }
- *aIncomingFlags &= ~INCOMING_ARABICCHAR;
- if (inWhitespace) {
- *aIncomingFlags |= INCOMING_WHITESPACE;
- } else {
- *aIncomingFlags &= ~INCOMING_WHITESPACE;
- }
- }
-
- if (outputStart + aLength != aOutput) {
- flags |= TEXT_WAS_TRANSFORMED;
- }
- *aAnalysisFlags = flags;
- return aOutput;
-}
+ uint32_t* aAnalysisFlags);
+template char16_t*
+nsTextFrameUtils::TransformText(const char16_t* aText, uint32_t aLength,
+ char16_t* aOutput,
+ CompressionMode aCompression,
+ uint8_t* aIncomingFlags,
+ gfxSkipChars* aSkipChars,
+ uint32_t* aAnalysisFlags);
uint32_t
nsTextFrameUtils::ComputeApproximateLengthWithWhitespaceCompression(
nsIContent *aContent, const nsStyleText *aStyleText)
{
const nsTextFragment *frag = aContent->GetText();
// This is an approximation so we don't really need anything
// too fancy here.
--- a/layout/generic/nsTextFrameUtils.h
+++ b/layout/generic/nsTextFrameUtils.h
@@ -82,16 +82,20 @@ public:
IsSpaceCombiningSequenceTail(const char16_t* aChars, int32_t aLength) {
return aLength > 0 &&
(mozilla::unicode::IsClusterExtender(aChars[0]) ||
(IsBidiControl(aChars[0]) &&
IsSpaceCombiningSequenceTail(aChars + 1, aLength - 1)
)
);
}
+ static bool
+ IsSpaceCombiningSequenceTail(const uint8_t* aChars, int32_t aLength) {
+ return false;
+ }
enum CompressionMode {
COMPRESS_NONE,
COMPRESS_WHITESPACE,
COMPRESS_WHITESPACE_NEWLINE,
COMPRESS_NONE_TRANSFORM_TO_SPACE
};
@@ -104,29 +108,23 @@ public:
* @param aCompression control what is compressed to a
* single space character: no compression, compress spaces (not followed
* by combining mark) and tabs, compress those plus newlines, or
* no compression except newlines are discarded.
* @param aIncomingFlags a flag indicating whether there was whitespace
* or an Arabic character preceding this text. We set it to indicate if
* there's an Arabic character or whitespace preceding the end of this text.
*/
- static char16_t* TransformText(const char16_t* aText, uint32_t aLength,
- char16_t* aOutput,
- CompressionMode aCompression,
- uint8_t* aIncomingFlags,
- gfxSkipChars* aSkipChars,
- uint32_t* aAnalysisFlags);
-
- static uint8_t* TransformText(const uint8_t* aText, uint32_t aLength,
- uint8_t* aOutput,
- CompressionMode aCompression,
- uint8_t* aIncomingFlags,
- gfxSkipChars* aSkipChars,
- uint32_t* aAnalysisFlags);
+ template<class CharT>
+ static CharT* TransformText(const CharT* aText, uint32_t aLength,
+ CharT* aOutput,
+ CompressionMode aCompression,
+ uint8_t* aIncomingFlags,
+ gfxSkipChars* aSkipChars,
+ uint32_t* aAnalysisFlags);
static void
AppendLineBreakOffset(nsTArray<uint32_t>* aArray, uint32_t aOffset)
{
if (aArray->Length() > 0 && (*aArray)[aArray->Length() - 1] == aOffset)
return;
aArray->AppendElement(aOffset);
}