[FriBidi] Experience of reducing FriBidi size?
Mikael Hedlund
mikael.hedlund at mecel.se
Tue Dec 7 10:51:48 PST 2010
Hi!
I am about to implement support for Arabic script in an embedded system,
where small footprint in flash and particularly RAM is important.
The system does also handle languages using the latin alphabet. The
handling and rendering of Arabic script will be pretty basic, and the plan
is
to support only the 28 basic characters including positional forms and the
ligature lam + alef, and also the Arabic-Indic numerals. I hope that's
"decent enough", or are there some other arabic characters/glyphs that I
"must" support? (I don't speak/read/write Arabic, so I am not 100% sure
here)
Support for Arabic script should preferably not add much more than 40-50
kB to my current binary, although 80 kB might be ok. However, 100 kB would
probably
be too much. Since limited memory is a big issue, I currently see two main
options (since I reckon using the ICU lib would not be acceptable from
memory point of view):
a) The preferred way: Use the FriBidi in a reduced version where I
comment/configure out or remove parts I don't need. As I understand, the
latest version fully complies with the Unicode BiDi algorithm,
including the bidi control codes, and is practically bug free. In addition
it also supports Arabic shaping. That all makes FriBidi attractive, but
currently the footprint is unfortunately too big for me.
b) Second way out: Use for instance the C reference implementation of the
Unicode Bidi algorithm (by adding the necessary wrapping) + write my own
code to handle the Arabic shaping.
For reference: when I compiled using GCC on Ubuntu, the C ref
implementation binary was about 20 kB (dynamically linked) and the FriBidi
lib was about 165 kB (dynamically linked).
1) Are there some easy steps to reduce the size of the FriBidi binary, and
does anyone have any experience how much the FriBidi binary in that case
could be reduced?
From a quick look in the code there seemed to be some fairly large
tables which I assume could be reduced, plus some Hebrew related
conversion routines to/from Unicode that are not needed in my case.
Are there any nifty defines or other config parameters I have missed,
that would reduce size?
2) If I don't succeed in reducing FriBidi enough, does anyone have a rough
clue or experience of how much time it might take to implement a basic
version of Arabic shaping based on my needs,
or have any clue/experience of how much that could add to the
footprint?
Implementing the Arabic shaping doesn't seem extremely hard given the
pretty basic shaping I need, but I am sure there are problems I don't
foresee, so I guess I shouldn't say anything until I have tried. :-)
Best regards,
Mikael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/fribidi/attachments/20101207/787d7b57/attachment.htm>
More information about the fribidi
mailing list