Re: Patches for supporting Sinhala script

On Tue, 2003-09-16 at 14:36, Eric Mader wrote:
> Hello Harsha,
> 
> Thanks for sending these patches. It's great to get information about a 
> whole new Indic script! I looked them over briefly, and they seem fine, 
> though I share your concern about how the change in IndicReordering.cpp 
> will effect the other Indic scripts.

I think there is a pretty big need for regression tests for
IndicReordering.cpp. The main function is so complicated now it's
pretty much impossible to guess what a change for one script will
do to another script. 

It should be pretty straightforward to create such a framework, except
for gathering the test data.

I think I'll try setting up tests for for the copy of the code in Pango
before I apply any of the outstanding changes... we should be able
to share the test data pretty easily.

> Unfortunately, I don't think I'll have enought time to get these changes 
> into ICU 2.8, due at the end of this year; our 2.8 schedule was decided 
> a couple of months ago, and is pretty full. I've been accumulating a 
> small list of bugs for other Indic scripts, which I'm planning to 
> address in ICU 3.0 early next year. I'll probably add Sinhala at the 
> same time.
> 
> If I get a few spare moments, I'll try and test your patches with the 
> other Indic scripts to see if anything breaks.
> 
> Regards,
> Eric Mader
> IBM GCoC - San José
> 5600 Cottle Rd.  M/S 50-2 / B11
> San Jose, CA 95193
> 
> [email protected] wrote:
> 
> > 
> > 
> > 
> > Dear Sir,
> > 
> > I have made the necessary amendments to IndicClassTables.cpp and
> > LayoutEngine.cpp.
> > 
> > I have also made a small modification to saveMatra() method in
> > IndicReordering.cpp. This hack was done because in Sinhala there are
> > two-part dependent vowel signs  ( DIGA KOMBUVA (0x0DDA) & SINHALA VOWEL
> > SIGN KOMBUVA HAA DIGA AELA-PILLA (0x0DDD) ) which are composed of a matra
> > and a virama/halant sign, and when they are split and reordered the virama
> > sign was getting dropped. I am not familiar with other indic scripts and
> > what affect this will have on them so I just made this temp hack for my
> > testing purposes.
> > 
> > I was not quite sure about the script_flags so I just used the tamil script
> > flags since Tamil script is very close to Sinhala. I have tested the shaper
> > using Layoutsample and microsoft 'Iskoola pota'  (beta) opentype font.
> > 
> > If these patches are ok, I would like to update the pango code as well.  My
> > objective is to see Sinhala on Linux desktop.

If this is your objective, and in particular if you are interested in
Sinhala support for the GNOME desktop, then you might want to consider
porting your patches to Pango yourself and submitting them, instead of
waiting for Eric to first have time to apply them to ICU, then to port
them to Pango.

I don't know nearly about this area as Eric does, and don't have a whole
lot of time myself, but I'm going to have to find some time in the next
month or two to deal with a bunch of outstanding bugs / patches that 
have been submitted for Kannada, Bengali, Malayalam, etc. Your patches
look pretty trivial compared to the issues there.

Regards,
						Owen