From: msevior_at_physics.unimelb.edu.au
Date: Thu Mar 11 2004 - 20:23:38 EST
>
> slowly getting there: able to get some text on screen now.
>
> I am seriously contemplating removal of the fl_BlockLayout char width
> cache -- we already cache character widths in the font cache and with
> the uniscribe engine I end up with yet a third cache. In any case,
> the block is an illogical place to keep these since the block does
> not have actual character content, does not use the widths for
> anything, and inserting text into the cache requires constant
> resizing. The with cache should be associated with an individual text
> run and, preferably, its implementation should be opaque to the
> layout classes (e.g., on win32 char widths are int16, and we have to
> convert the int32s on every draw). Caching them inside the new
> GR_FERenderInfo class seems much preferrable.
>
Think through the implications of things like embedded layout structures
like footnotes and endnotes. Right now I can't think of a reason why they
would worry about where the width cache is but I hope can't either. (Apart
from removing the code that inserts the width cache.)
Oh wat about spell checking? I think the underlining code might need the
widths.
Cheers
Martin
This archive was generated by hypermail 2.1.4 : Thu Mar 11 2004 - 20:25:21 EST