[Localization] Status on SCIM

[Sayamindu Dasgupta]

Hi,
As a proof of concept, I wrote a XKB symbol file parser, and utilized
that to generate m17n-db format .mim files from our previous XKB work
which can be utilized both by SCIM and IBUS (if we decide to use IBUS
at some point). The parser is horribly slow (it takes around 3 minutes
to do a single conversion), and I believe it's my inexperience with
pyparsing which is causing most of the slowness. However, it does seem
to work, and I just did a test (not on an XO yet) with a converted
Devanagari keyboard layout, and it seemed to work perfectly.

Of course, one needs to tweak the output manually at times, but most
of the grunt work of figuring out the mappings will be done by the
script. For the Devanagari IM, the mapping did not require any manual
changes.

I'll check out fancy stuff like Alt-Gr mapping, etc tomorrow and let
you know how it works.

If anyone is interested, the tool can be checked out with

git-clone git://dev.laptop.org/users/sayamindu/xkb2scim

Invoke the script like:

python parse_xkb_symbols.py hi olpc > hi.mim

It's a quick and ugly hack, with some obvious areas of improvement so
patches are most welcome ;-)

I'll also get some feedback from the SCIM list on this.

Cheers,
Sayamindu






On Fri, Sep 5, 2008 at 7:25 PM, Jim Gettys <jg at laptop.org> wrote:
> Seems like the next step is a proof of concept....
>
> What's the basic footprint of SCIM, anyway?
>                     - Jim
>
>
> On Fri, 2008-09-05 at 19:01 +0530, Sayamindu Dasgupta wrote:
>> On Fri, Sep 5, 2008 at 6:22 PM, Jim Gettys <jg at laptop.org> wrote:
>> > Sayamindu,
>> >
>> > I still see no compelling reason to convert the existing layouts to
>> > scim, at least in the short term.  It may certainly be best for future
>> > keyboards to do it "the new way".
>> >
>> > What am I missing?
>> >                           - Jim
>> >
>>
>> If we want to keep the existing XKB based system, we would have to
>> load SCIM conditionally, which I want to avoid (and maintaining and
>> supporting two different IMs might be a bit of pain). Also, even some
>> of the existing keyboard layouts would benefit from a switch to SCIM
>> (eg: Amharic and Nepali).
>> >From what I understand (till now) - it is not a major effort to
>> convert existing XKB symbol files to .mim files, the process should be
>> relatively straightforward and easily scriptified. Of course, I don't
>> have a proof of concept yet :-)
>>
>> Thanks,
>> Sayamindu
>>
>>
>> >
>> > On Fri, 2008-09-05 at 17:48 +0530, Sayamindu Dasgupta wrote:
>> >> Hello,
>> >> During this week, I poked SCIM a bit more, especially on the topic of
>> >> migrating our existing layouts to the framework. It looks like the
>> >> m17n IMEngine would be the best way to go forward at this stage for
>> >> our existing layouts (though I need to check a bit more on the
>> >> dependencies it will be pulling in). I'll try to come up a (semi)
>> >> automate way to migrate our XKB symbol files to the m17n .mim format
>> >> by next week.
>> >> Thanks,
>> >> Sayamindu
>> >>
>> >>
>> > --
>> > Jim Gettys <jg at laptop.org>
>> > One Laptop Per Child
>> >
>> >
>>
>>
>>
> --
> Jim Gettys <jg at laptop.org>
> One Laptop Per Child
>
>



-- 
Sayamindu Dasgupta
[http://sayamindu.randomink.org/ramblings]