Improve the script itemization algorithm to match Unicode 8.0

Override preceding Common-s with a subsequent non-Inherited,
non-Common script.
This produces longer script runs, which automagically improves
the shaping quality (as we don't lose the context anymore),
the shaping performance (as we're typically shape a fewer runs),
and the fallback font selection (when the font supports more
than just a single language/script).

Task-number: QTBUG-29930
Change-Id: I1c55af30bd397871d7f1f6e062605517f5a7e5a1
Reviewed-by: Lars Knoll <lars.knoll@theqtcompany.com>
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@qt.io>
This commit is contained in:
Konstantin Ritt 2016-01-30 10:18:25 +04:00
parent e81f52ecc7
commit 3df159ba17
2 changed files with 31 additions and 62 deletions

View File

@ -685,10 +685,10 @@ Q_CORE_EXPORT void initCharAttributes(const ushort *string, int length,
Q_CORE_EXPORT void initScripts(const ushort *string, int length, uchar *scripts)
{
int sor = 0;
int eor = -1;
int eor = 0;
uchar script = QChar::Script_Common;
for (int i = 0; i < length; ++i) {
eor = i;
for (int i = 0; i < length; ++i, eor = i) {
uint ucs4 = string[i];
if (QChar::isHighSurrogate(ucs4) && i + 1 < length) {
ushort low = string[i + 1];
@ -700,60 +700,37 @@ Q_CORE_EXPORT void initScripts(const ushort *string, int length, uchar *scripts)
const QUnicodeTables::Properties *prop = QUnicodeTables::properties(ucs4);
if (Q_LIKELY(prop->script == script || prop->script <= QChar::Script_Inherited))
uchar nscript = prop->script;
if (Q_LIKELY(nscript == script || nscript <= QChar::Script_Common))
continue;
// inherit preceding Common-s
if (Q_UNLIKELY(script <= QChar::Script_Common)) {
// also covers a case where the base character of Common script followed
// by one or more combining marks of non-Inherited, non-Common script
script = nscript;
continue;
}
// Never break between a combining mark (gc= Mc, Mn or Me) and its base character.
// Thus, a combining mark — whatever its script property value is — should inherit
// the script property value of its base character.
static const int test = (FLAG(QChar::Mark_NonSpacing) | FLAG(QChar::Mark_SpacingCombining) | FLAG(QChar::Mark_Enclosing));
if (Q_UNLIKELY(FLAG(prop->category) & test)) {
// In cases where the base character itself has the Common script property value,
// and it is followed by one or more combining marks with a specific script property value,
// it may be even better for processing to let the base acquire the script property value
// from the first mark. This approach can be generalized by treating all the characters
// of a combining character sequence as having the script property value
// of the first non-Inherited, non-Common character in the sequence if there is one,
// and otherwise treating all the characters as having the Common script property value.
if (Q_LIKELY(script > QChar::Script_Common || prop->script <= QChar::Script_Common))
continue;
if (Q_UNLIKELY(FLAG(prop->category) & test))
continue;
script = QChar::Script(prop->script);
}
Q_ASSERT(script > QChar::Script_Common);
Q_ASSERT(sor < eor);
::memset(scripts + sor, script, (eor - sor) * sizeof(uchar));
sor = eor;
#if 0 // ### Disabled due to regressions. The font selection algorithm is not prepared for this change.
if (Q_LIKELY(script != QChar::Script_Common)) {
// override preceding Common-s
while (sor > 0 && scripts[sor - 1] == QChar::Script_Common)
--sor;
} else {
// see if we are inheriting preceding run
if (sor > 0)
script = scripts[sor - 1];
}
#endif
while (sor < eor)
scripts[sor++] = script;
script = prop->script;
script = nscript;
}
eor = length;
#if 0 // ### Disabled due to regressions. The font selection algorithm is not prepared for this change.
if (Q_LIKELY(script != QChar::Script_Common)) {
// override preceding Common-s
while (sor > 0 && scripts[sor - 1] == QChar::Script_Common)
--sor;
} else {
// see if we are inheriting preceding run
if (sor > 0)
script = scripts[sor - 1];
}
#endif
while (sor < eor)
scripts[sor++] = script;
Q_ASSERT(script >= QChar::Script_Common);
Q_ASSERT(eor == length);
::memset(scripts + sor, script, (eor - sor) * sizeof(uchar));
}
} // namespace QUnicodeTools

View File

@ -1225,29 +1225,21 @@ void tst_QTextScriptEngine::thaiWithZWJ()
QTextLayout layout(s, font);
QTextEngine *e = layout.engine();
e->itemize();
QCOMPARE(e->layoutData->items.size(), 11);
QCOMPARE(e->layoutData->items.size(), 3);
for (int item = 0; item < e->layoutData->items.size(); ++item)
e->shape(item);
QCOMPARE(e->layoutData->items[0].num_glyphs, ushort(7)); // Thai: The ZWJ and ZWNJ characters are inherited, so should be part of the thai script
QCOMPARE(e->layoutData->items[1].num_glyphs, ushort(1)); // Common: The smart quotes cannot be handled by thai, so should be a separate item
QCOMPARE(e->layoutData->items[2].num_glyphs, ushort(1)); // Thai: Thai character
QCOMPARE(e->layoutData->items[3].num_glyphs, ushort(1)); // Common: Ellipsis
QCOMPARE(e->layoutData->items[4].num_glyphs, ushort(1)); // Thai: Thai character
QCOMPARE(e->layoutData->items[5].num_glyphs, ushort(1)); // Common: Smart quote
QCOMPARE(e->layoutData->items[6].num_glyphs, ushort(1)); // Thai: Thai character
QCOMPARE(e->layoutData->items[7].num_glyphs, ushort(1)); // Common: \xA0 = non-breaking space. Could be useful to have in thai, but not currently implemented
QCOMPARE(e->layoutData->items[8].num_glyphs, ushort(1)); // Thai: Thai character
QCOMPARE(e->layoutData->items[9].num_glyphs, ushort(1)); // Japanese: Kanji for tree
QCOMPARE(e->layoutData->items[10].num_glyphs, ushort(2)); // Thai: Thai character followed by superscript "a" which is of inherited type
QCOMPARE(e->layoutData->items[0].num_glyphs, ushort(15)); // Thai, Inherited and Common
QCOMPARE(e->layoutData->items[1].num_glyphs, ushort(1)); // Japanese: Kanji for tree
QCOMPARE(e->layoutData->items[2].num_glyphs, ushort(2)); // Thai: Thai character followed by superscript "a" which is of inherited type
//A quick sanity check - check all the characters are individual clusters
unsigned short *logClusters = e->layoutData->logClustersPtr;
for (int i = 0; i < 7; i++)
for (int i = 0; i < 15; i++)
QCOMPARE(logClusters[i], ushort(i));
for (int i = 0; i < 10; i++)
QCOMPARE(logClusters[i+7], ushort(0));
for (int i = 0; i < 3; i++)
QCOMPARE(logClusters[i+15], ushort(0));
// A thai implementation could either remove the ZWJ and ZWNJ characters, or hide them.
// The current implementation hides them, so we test for that.