Subject: Re: integrating bidi (was Re: version 1.0)
From: Paul Rohr (paul@abisource.com)
Date: Wed Feb 07 2001 - 13:22:03 CST
At 08:56 PM 2/6/01 -0000, Tomas Frydrych wrote:
>> 1. How well localized are the bidi changes?
>
>The bidi changes are localized pretty much to /src/text/, some of
>the interface (couple of toolbar buttons, about four extra
>checkboxes in the Options dialogue), and couple of extra
>properties. The bigest changes are to the the run classes and to
>fp_Line, where the bidi algorithm itself lives.
Gotcha. Sounds good.
>> Adding a well-localized set of #ifdef ABI_BIDI changes to help speed up
work
>> on the feature makes sense. Sprinkling random bidi dust all over the tree
>> is less appealing. (Think a dozen or two ifdefs, rather than an order of
>> magnitude more.)
>
>As the patch stands at the moment, all the BIDI stuff is handled via
>conditional complilation by specifying
>ABI_OPT_BIDI_ENABLED=1, (the #define used in the sources is
>BIDI_ENABLED); there is though much more of it than two dozen.
>The one thing that would make this number immediately smaller
>would be if we could agree to change the signature of the
>findPointCoords functions to that required by bidi; this is
>responsible for the worst and most confusing #ifdef's.
That ifdef reduction sounds very desirable.
>I am not advocating that we immediately make the BiDi the default
>mode, but checking it into the tree would (1) make my life much
>easier; (2) would make people aware which bits of the code are bidi-
>sensitive when they are making changes and designing new
>features. (3) would encourage other people to work on it -- it is
>really not very well possible for more than one person to be
>developing a feature that only exists as a patch; there have been at
>least two people who indicated they would like to work on this, but
>all I can tell them at the moment is to wait until it makes its way
>into the CVS.
Yes, yes, and yes.
Unless one of the usual refactoring experts is ready to weigh in (soon) with
concrete suggestions about how Tomas' code could be factored even more
cleanly, it sounds like it's time to find ways to get this into the tree.
Can anyone think of other reasons to wait?
Paul
This archive was generated by hypermail 2b25 : Wed Feb 07 2001 - 13:14:25 CST