Subject: Re: BiDi patch
From: Martin Sevior (msevior@mccubbin.ph.unimelb.edu.au)
Date: Sun Jan 14 2001 - 20:31:07 CST
Hi Tomas,
On Sat, 13 Jan 2001, Tomas Frydrych wrote:
>
> The g-types are not 'real', they are redefined by macros to our UT_
> types, and only appear in the FriBiDi headers. The reason I choose
> this approach is future maintainance; the Unicode tables from the
> FriBiDi library may need to be updated time from time and this
> way, the files can just be replaced without modification.
Ahh that makes perfect sense. Thanks.
>
> > Next I randomly selected a not too difficult word doc I have. You patch
> > fails to render this correctly. Many text runs are reversed.
>
> There are two issues here. (1) the patch currently implements only
> the the core changes, and thus works only with the ABW native
> format; the importers/exporters will have to be modified to handle
> the two new properties called 'direction' and 'dominant-direction'.
Just checking. If this property is not defined do you assume LTR?
(2)
> The last patch I made was in test mode in which capitals are
> treated as RTL characters; this was an oversight, to fix it just
> comment out #define CAPS_TEST near the top of fp_TextRun.cpp
> and recompile, hopefully you should see your doc OK. (I should
> say that only the first character of each run is tested, since runs of
> different direction cannot be coalesced, so that you almost
> certainly see even lowercase characters reversed in your document
> in the test mode.)
>
OK this makes sense. Unfortunately I can't test this again until next
Monday at the earlist. After that I should be pretty much back to my usual
rate of checking in code....
I still think we should not make this a compile time option and if we are
happy with the rendering speed and everything works we should enable bidi
by default.
Do you have some Hebrew fonts and docs for us to test this with? Just put
them up on your download site.
Cheers
Martin
This archive was generated by hypermail 2b25 : Sun Jan 14 2001 - 20:31:30 CST