On Tue, Dec 10, 2002 at 08:36:45PM +0200, Dekel Tsur wrote:
> Another idea is to just cache the width for the default text font.
This would reduce the speedup we get...
regards
john
--
"Saying that taste is just personal preference is a good way to prevent
disputes. The trouble is, it's not tr
On Tue, Dec 10, 2002 at 06:10:46PM +, John Levon wrote:
> > One solution (other than cleaning the cache) is to store
> > (qfont_loader::font_info const *, int) pairs instead of
> > (LyXFont, int) pairs.
>
> I'm going to do something similar indeed.
Another idea is to just cache the width for
On Tue, Dec 10, 2002 at 11:48:59AM +0100, Lars Gullik Bjønnes wrote:
> I have a feeling that the cache is a bit more complex than it needs to
> be, but I cannot quite put my finger on it...
I'm going to rework things a bit I think
> How often is metrics for whole strings wanted?
Not often. the
On Tue, Dec 10, 2002 at 11:57:08AM +0200, Dekel Tsur wrote:
> What happens when the user changes the font in the preferences dialog ?
> If I'm not mistaken, the cache will not be cleared.
Yes it will. Like I mentioned ...
> One solution (other than cleaning the cache) is to store
> (qfont_loade
John Levon <[EMAIL PROTECTED]> writes:
What is the used for?
| +typedef pair MetricPair;
| +typedef vector MetricList;
| +typedef map CharMetrics;
| +CharMetrics char_metrics;
| +
| +int singlewidth(Uchar val, LyXFont const & f)
| +{
| + CharMetrics::iterator it(char_metrics.find(val));
|
John Levon <[EMAIL PROTECTED]> writes:
| On Tue, Dec 10, 2002 at 12:41:26AM +, John Levon wrote:
|
| > Well I'm out of ideas so here's a dumb patch that caches single widths.
|
| btw, a map based approach is at least twice as slower than
| the vector one below
|
| > resize took 1.64121 seco
On Tue, Dec 10, 2002 at 03:01:36AM +, John Levon wrote:
>
> Unless somebody has major complaints I'm going to commit it ...
What happens when the user changes the font in the preferences dialog ?
If I'm not mistaken, the cache will not be cleared.
One solution (other than cleaning the cache)
On Tue, Dec 10, 2002 at 12:41:26AM +, John Levon wrote:
> int ascent(char c, LyXFont const & f)
> {
> - QRect r = metrics(f).boundingRect(c);
> + QRect const & r = metrics(f).boundingRect(c);
> return -r.top();
> }
Why not
return -metrics(f).boundingRect(c).top();
Andre'
On Tue, Dec 10, 2002 at 12:41:26AM +, John Levon wrote:
> Well I'm out of ideas so here's a dumb patch that caches single widths.
btw, a map based approach is at least twice as slower than
the vector one below
> resize took 1.64121 seconds
> resize took 1.18429 seconds
> resize took 1.3532 s
