lyx_mirror/src/frontends/qt4/GuiFontMetrics.cpp
Jean-Marc Lasgouttes c5119c97fc Add caching for the QTextLayout objects we use
The QTextLayout handling is terribly slow on Qt 4.8.7, but some
caching has been added in Qt5 that makes it much faster. For some
reason, it is not that slow with Qt 4.8.1.

Caches are introduced for the three following methods

* width(doctring), controlled by CACHE_METRICS_WIDTH. This cache already
  existed, but the code has been cleaned up

* getTextLayout, controlled by CACHE_METRICS_QTEXTLAYOUT (disabled by
  default on Qt5, which does its own caching). This is used for pos2x
  and x2pos and now for drawing of text too. The previous code used a
  trivial caching scheme of the last used QTextLayout, but now they
  are properly kept in a QCache. Moreover, the cacheEnabled() property
  is enabled for these QTextLayout object (not sure what this does).

* breakAt, controlled by CACHE_METRICS_BREAKAT. This is the only user
  of QTextLayout which did not have some kind of caching already.

For some weird reasons related to Argument-dependent look-up, the
qHash(docstring) function has to be defined in std namespace, since
lyx::docstring is actually std::basic_string<wchar_t>.

[NOTE: this version has profiling hooks, enabled by commenting out the line
  #define DISABLE_PMPROF
that should eventually be removed.]
2016-12-19 18:47:18 +01:00

515 lines
13 KiB
C++

/**
* \file GuiFontMetrics.cpp
* This file is part of LyX, the document processor.
* Licence details can be found in the file COPYING.
*
* \author unknown
* \author John Levon
*
* Full author contact details are available in file CREDITS.
*/
#include <config.h>
#include "GuiFontMetrics.h"
#include "qt_helpers.h"
#include "Dimension.h"
#include "Language.h"
#include "LyXRC.h"
#include "insets/Inset.h"
#include "support/convert.h"
#include "support/lassert.h"
#define DISABLE_PMPROF
#include "support/pmprof.h"
#ifdef CACHE_SOME_METRICS
#include <QByteArray>
#endif
using namespace std;
using namespace lyx::support;
#ifdef CACHE_SOME_METRICS
namespace std {
/*
* Argument-dependent lookup implies that this function shall be
* declared in the namespace of its argument. But this is std
* namespace, since lyx::docstring is just std::basic_string<wchar_t>.
*/
uint qHash(lyx::docstring const & s)
{
return qHash(QByteArray(reinterpret_cast<char const *>(s.data()),
s.size() * sizeof(lyx::docstring::value_type)));
}
}
#endif
namespace lyx {
namespace frontend {
namespace {
/**
* Convert a UCS4 character into a QChar.
* This is a hack (it does only make sense for the common part of the UCS4
* and UTF16 encodings) and should not be used.
* This does only exist because of performance reasons (a real conversion
* using iconv is too slow on windows).
*
* This is no real conversion but a simple cast in reality. This is the reason
* why this works well for symbol fonts used in mathed too, even though
* these are not real ucs4 characters. These are codepoints in the
* computer modern fonts used, nothing unicode related.
* See comment in GuiPainter::text() for more explanation.
**/
inline QChar const ucs4_to_qchar(char_type const ucs4)
{
LATTEST(is_utf16(ucs4));
return QChar(static_cast<unsigned short>(ucs4));
}
} // anon namespace
/*
* Limit (strwidth|breakat)_cache_ size to 512kB of string data.
* Limit qtextlayout_cache_ size to 500 elements (we do not know the
* size of the QTextLayout objects anyway).
* Note that all these numbers are arbitrary.
*/
GuiFontMetrics::GuiFontMetrics(QFont const & font)
: font_(font), metrics_(font, 0)
#ifdef CACHE_METRICS_WIDTH
, strwidth_cache_(1 << 19)
#endif
#ifdef CACHE_METRICS_BREAKAT
, breakat_cache_(1 << 19)
#endif
#ifdef CACHE_METRICS_QTEXTLAYOUT
, qtextlayout_cache_(500)
#endif
{
}
int GuiFontMetrics::maxAscent() const
{
return metrics_.ascent();
}
int GuiFontMetrics::maxDescent() const
{
// We add 1 as the value returned by QT is different than X
// See http://doc.trolltech.com/2.3/qfontmetrics.html#200b74
return metrics_.descent() + 1;
}
int GuiFontMetrics::em() const
{
return QFontInfo(font_).pixelSize();
}
int GuiFontMetrics::lineWidth() const
{
return metrics_.lineWidth();
}
int GuiFontMetrics::underlinePos() const
{
return metrics_.underlinePos();
}
int GuiFontMetrics::strikeoutPos() const
{
return metrics_.strikeOutPos();
}
int GuiFontMetrics::lbearing(char_type c) const
{
if (!is_utf16(c))
// FIXME: QFontMetrics::leftBearing does not support the
// full unicode range. Once it does, we could use:
//return metrics_.leftBearing(toqstr(docstring(1, c)));
return 0;
return metrics_.leftBearing(ucs4_to_qchar(c));
}
namespace {
int const outOfLimitMetric = -10000;
}
int GuiFontMetrics::rbearing(char_type c) const
{
int value = rbearing_cache_.value(c, outOfLimitMetric);
if (value != outOfLimitMetric)
return value;
// Qt rbearing is from the right edge of the char's width().
if (is_utf16(c)) {
QChar sc = ucs4_to_qchar(c);
value = width(c) - metrics_.rightBearing(sc);
} else {
// FIXME: QFontMetrics::leftBearing does not support the
// full unicode range. Once it does, we could use:
// metrics_.rightBearing(toqstr(docstring(1, c)));
value = width(c);
}
rbearing_cache_.insert(c, value);
return value;
}
int GuiFontMetrics::width(docstring const & s) const
{
PROFILE_THIS_BLOCK(width)
#ifdef CACHE_METRICS_WIDTH
int * pw = strwidth_cache_[s];
if (pw)
return *pw;
// For some reason QMetrics::width returns a wrong value with Qt5
// int w = metrics_.width(toqstr(s));
PROFILE_CACHE_MISS(width)
#endif
QTextLayout tl;
tl.setText(toqstr(s));
tl.setFont(font_);
tl.beginLayout();
QTextLine line = tl.createLine();
tl.endLayout();
int w = int(line.naturalTextWidth());
#ifdef CACHE_METRICS_WIDTH
strwidth_cache_.insert(s, new int(w), s.size() * sizeof(char_type));
#endif
return w;
}
int GuiFontMetrics::width(QString const & ucs2) const
{
return width(qstring_to_ucs4(ucs2));
}
int GuiFontMetrics::signedWidth(docstring const & s) const
{
if (s.empty())
return 0;
if (s[0] == '-')
return -width(s.substr(1, s.size() - 1));
else
return width(s);
}
QTextLayout const *
GuiFontMetrics::getTextLayout(docstring const & s, bool const rtl,
double const wordspacing) const
{
PROFILE_THIS_BLOCK(getTextLayout)
QTextLayout * ptl;
#ifdef CACHE_METRICS_QTEXTLAYOUT
docstring const s_cache = s + (rtl ? "r" : "l") + convert<docstring>(wordspacing);
ptl = qtextlayout_cache_[s_cache];
if (!ptl) {
PROFILE_CACHE_MISS(getTextLayout)
#endif
ptl = new QTextLayout();
ptl->setCacheEnabled(true);
ptl->setText(toqstr(s));
QFont copy = font_;
copy.setWordSpacing(wordspacing);
ptl->setFont(copy);
// Note that both setFlags and the enums are undocumented
ptl->setFlags(rtl ? Qt::TextForceRightToLeft : Qt::TextForceLeftToRight);
ptl->beginLayout();
ptl->createLine();
ptl->endLayout();
#ifdef CACHE_METRICS_QTEXTLAYOUT
qtextlayout_cache_.insert(s_cache, ptl);
}
#endif
return ptl;
}
int GuiFontMetrics::pos2x(docstring const & s, int const pos, bool const rtl,
double const wordspacing) const
{
if (pos <= 0)
return rtl ? width(s) : 0;
QTextLayout const * tl = getTextLayout(s, rtl, wordspacing);
/* Since QString is UTF-16 and docstring is UCS-4, the offsets may
* not be the same when there are high-plan unicode characters
* (bug #10443).
*/
int const qpos = toqstr(s.substr(0, pos)).length();
return static_cast<int>(tl->lineForTextPosition(qpos).cursorToX(qpos));
}
int GuiFontMetrics::x2pos(docstring const & s, int & x, bool const rtl,
double const wordspacing) const
{
QTextLayout const * tl = getTextLayout(s, rtl, wordspacing);
int const qpos = tl->lineForTextPosition(0).xToCursor(x);
// correct x value to the actual cursor position.
x = static_cast<int>(tl->lineForTextPosition(0).cursorToX(qpos));
/* Since QString is UTF-16 and docstring is UCS-4, the offsets may
* not be the same when there are high-plan unicode characters
* (bug #10443).
*/
#if QT_VERSION < 0x040801 || QT_VERSION >= 0x050100
return qstring_to_ucs4(tl->text().left(qpos)).length();
#else
/* Due to QTBUG-25536 in 4.8.1 <= Qt < 5.1.0, the string returned
* by QString::toUcs4 (used by qstring_to_ucs4) may have wrong
* length. We work around the problem by trying all docstring
* positions until the right one is found. This is slow only if
* there are many high-plane Unicode characters. It might be
* worthwhile to implement a dichotomy search if this shows up
* under a profiler.
*/
int pos = min(qpos, static_cast<int>(s.length()));
while (pos >= 0 && toqstr(s.substr(0, pos)).length() != qpos)
--pos;
LASSERT(pos > 0 || qpos == 0, /**/);
return pos;
#endif
}
int GuiFontMetrics::countExpanders(docstring const & str) const
{
// Numbers of characters that are expanded by inter-word spacing. These
// characters are spaces, except for characters 09-0D which are treated
// specially. (From a combination of testing with the notepad found in qt's
// examples, and reading the source code.) In addition, consecutive spaces
// only count as one expander.
bool wasspace = false;
int nexp = 0;
for (char_type c : str)
if (c > 0x0d && QChar(c).isSpace()) {
if (!wasspace) {
++nexp;
wasspace = true;
}
} else
wasspace = false;
return nexp;
}
pair<int, int> *
GuiFontMetrics::breakAt_helper(docstring const & s, int const x,
bool const rtl, bool const force) const
{
QTextLayout tl;
/* Qt will not break at a leading or trailing space, and we need
* that sometimes, see http://www.lyx.org/trac/ticket/9921.
*
* To work around the problem, we enclose the string between
* zero-width characters so that the QTextLayout algorithm will
* agree to break the text at these extremal spaces.
*/
// Unicode character ZERO WIDTH NO-BREAK SPACE
QChar const zerow_nbsp(0xfeff);
QString qs = zerow_nbsp + toqstr(s) + zerow_nbsp;
#if 1
/* Use unicode override characters to enforce drawing direction
* Source: http://www.iamcal.com/understanding-bidirectional-text/
*/
if (rtl)
// Right-to-left override: forces to draw text right-to-left
qs = QChar(0x202E) + qs;
else
// Left-to-right override: forces to draw text left-to-right
qs = QChar(0x202D) + qs;
int const offset = 2;
#else
// Alternative version that breaks with Qt5 and arabic text (#10436)
// Note that both setFlags and the enums are undocumented
tl.setFlags(rtl ? Qt::TextForceRightToLeft : Qt::TextForceLeftToRight);
int const offset = 1;
#endif
tl.setText(qs);
tl.setFont(font_);
QTextOption to;
to.setWrapMode(force ? QTextOption::WrapAtWordBoundaryOrAnywhere
: QTextOption::WordWrap);
tl.setTextOption(to);
tl.beginLayout();
QTextLine line = tl.createLine();
line.setLineWidth(x);
tl.createLine();
tl.endLayout();
if ((force && line.textLength() == offset) || int(line.naturalTextWidth()) > x)
return new pair<int, int>(-1, -1);
/* Since QString is UTF-16 and docstring is UCS-4, the offsets may
* not be the same when there are high-plan unicode characters
* (bug #10443).
*/
// The variable `offset' is here to account for the extra leading characters.
// The ending character zerow_nbsp has to be ignored if the line is complete.
int const qlen = line.textLength() - offset - (line.textLength() == qs.length());
#if QT_VERSION < 0x040801 || QT_VERSION >= 0x050100
int len = qstring_to_ucs4(qs.mid(offset, qlen)).length();
#else
/* Due to QTBUG-25536 in 4.8.1 <= Qt < 5.1.0, the string returned
* by QString::toUcs4 (used by qstring_to_ucs4) may have wrong
* length. We work around the problem by trying all docstring
* positions until the right one is found. This is slow only if
* there are many high-plane Unicode characters. It might be
* worthwhile to implement a dichotomy search if this shows up
* under a profiler.
*/
int len = min(qlen, static_cast<int>(s.length()));
while (len >= 0 && toqstr(s.substr(0, len)).length() != qlen)
--len;
LASSERT(len > 0 || qlen == 0, /**/);
#endif
// The -1 is here to account for the leading zerow_nbsp.
return new pair<int, int>(len, int(line.naturalTextWidth()));
}
bool GuiFontMetrics::breakAt(docstring & s, int & x, bool const rtl, bool const force) const
{
PROFILE_THIS_BLOCK(breakAt)
if (s.empty())
return false;
pair<int, int> * pp;
#ifdef CACHE_METRICS_BREAKAT
docstring const s_cache = s + convert<docstring>(x) + (rtl ? "r" : "l") + (force ? "f" : "w");
pp = breakat_cache_[s_cache];
if (!pp) {
PROFILE_CACHE_MISS(breakAt)
#endif
pp = breakAt_helper(s, x, rtl, force);
#ifdef CACHE_METRICS_BREAKAT
breakat_cache_.insert(s_cache, pp, s_cache.size() * sizeof(char_type));
}
#endif
if (pp->first == -1)
return false;
s = s.substr(0, pp->first);
x = pp->second;
#ifndef CACHE_METRICS_BREAKAT
delete pp;
#endif
return true;
}
void GuiFontMetrics::rectText(docstring const & str,
int & w, int & ascent, int & descent) const
{
static int const d = Inset::TEXT_TO_INSET_OFFSET / 2;
w = width(str) + Inset::TEXT_TO_INSET_OFFSET;
ascent = metrics_.ascent() + d;
descent = metrics_.descent() + d;
}
void GuiFontMetrics::buttonText(docstring const & str,
int & w, int & ascent, int & descent) const
{
rectText(str, w, ascent, descent);
w += Inset::TEXT_TO_INSET_OFFSET;
}
Dimension const GuiFontMetrics::defaultDimension() const
{
return Dimension(0, maxAscent(), maxDescent());
}
Dimension const GuiFontMetrics::dimension(char_type c) const
{
return Dimension(width(c), ascent(c), descent(c));
}
GuiFontMetrics::AscendDescend const GuiFontMetrics::fillMetricsCache(
char_type c) const
{
QRect r;
if (is_utf16(c))
r = metrics_.boundingRect(ucs4_to_qchar(c));
else
r = metrics_.boundingRect(toqstr(docstring(1, c)));
AscendDescend ad = { -r.top(), r.bottom() + 1};
// We could as well compute the width but this is not really
// needed for now as it is done directly in width() below.
metrics_cache_.insert(c, ad);
return ad;
}
int GuiFontMetrics::width(char_type c) const
{
int value = width_cache_.value(c, outOfLimitMetric);
if (value != outOfLimitMetric)
return value;
if (is_utf16(c))
value = metrics_.width(ucs4_to_qchar(c));
else
value = metrics_.width(toqstr(docstring(1, c)));
width_cache_.insert(c, value);
return value;
}
int GuiFontMetrics::ascent(char_type c) const
{
static AscendDescend const outOfLimitAD =
{outOfLimitMetric, outOfLimitMetric};
AscendDescend value = metrics_cache_.value(c, outOfLimitAD);
if (value.ascent != outOfLimitMetric)
return value.ascent;
value = fillMetricsCache(c);
return value.ascent;
}
int GuiFontMetrics::descent(char_type c) const
{
static AscendDescend const outOfLimitAD =
{outOfLimitMetric, outOfLimitMetric};
AscendDescend value = metrics_cache_.value(c, outOfLimitAD);
if (value.descent != outOfLimitMetric)
return value.descent;
value = fillMetricsCache(c);
return value.descent;
}
} // namespace frontend
} // namespace lyx