[scribus-dev] Scribus CTL

[Khaled Hosny]

On Tue, Mar 22, 2016 at 09:07:21PM +0100, Craig Bradney wrote:
> 
> > On 22 Mar 2016, at 20:23, Khaled Hosny <khaledhosny at eglug.org> wrote:
> > 
> > On Tue, Mar 22, 2016 at 07:36:08PM +0100, Craig Bradney wrote:
> >>>> 
> >>>> Some rendering comparisons:
> >>>> 
> >>>> We need to do this side by side comparison of trunk vs host/boxes.
> >>>> 3 issues to note at first:
> >>>> - character positioning is higher in the frame in boxes
> >>>> - Not as many characters fit in the frames in boxes
> >>> 
> >>> We couldn’t see such differences in the file you sent, see the diff
> >>> images in Dawood’s reply.
> >> 
> >> At lower zoom levels I can see them, at higher ones, no. 
> > 
> > I can check again, but if it depends on zoom level so it might be
> > related to the different font rasterisation strategy we are using.
> 
> You can see it there. Host on is the left.

Interesting, will checkout and see if I can reproduce it.

> >>>> - The right aligned paragraphs have the top line missing a character/aren’t aligned
> >>> 
> >>> It was indeed that the last glyph in the line was always remove, should
> >>> be fixed now.
> >> 
> >> Now thats better.
> >> 
> >> We need some more complex tests apart from the file I created. Ideas
> >> on how we can do this?
> > 
> > We have a collection of files we use for testing here, not complex but
> > they try to test different features. I’ll try to collect and send them.
> > 
> >> Also, what about any ideas on rendering speed comparisons?
> > 
> > I have no idea how to measure that, but I worked on optimizing the rendering
> > (after profiling was callgrind) fearing that ours was too slow, to
> > discover that we are actually feel much faster than trunk (text
> > selection, for example is snappier and does not lag behind as the cursor
> > moves).
> > 
> > Currently (on my HiDPI machine) most of the rendering time (~30% of the
> > all application profile) is spent in QPainter::drawPixmap(). I believe
> > rewriting ScPainter on top of QPainter can make it much faster, instead
> > of the current code which redraws the whole canvas as a pixmap on each
> > draw event. But that is a small project of its own.
> 
> 
> You’re correct. cursor movement is a lost faster with Host.
> 
> I’m very tempted to say let’s import this.. but honestly, scared to go
> through the experience with marks/bullets/numbering. Dev disappears
> and we don’t know the code. 

I’d say that the bus factor for this code is much higher than the rest
of Scribus code base. We have been removing code more than writing new
one, so it should make the code base more easier to grasp on the long
run. It is your call of course.

> Unfortunately I need Jean to review it too and he is travelling. He
> will likely be reading this though and he is back at the end of the
> week.
> 
> Can you send me the patch (or instructions on how to create it myself)
> and I can test on another svn tree I have for this?

OK, let me do some final touching and I’ll send you the patch.

Regards,
Khaled