Re: gtkim in gtk+2.0

[Steve Underwood <steveu coppice org>]

Owen Taylor wrote:

> [...]
>
> What you might be asking about is a table-driven IM which reads
> tables from disk. The only stumbling block to do doing that currently
> is that the module system expects that each module statically
> implements some number of input methods; there is no way for
> a module to list its availanle input methods at run time.
> (This also is a problem for other things, such as writing an
>  IIIMF input module)

It is possible to have a table driven IM which serves multiple locales to
different windows. The xcin software for Chinese allows different windows to
work in GB, Big5 or Unicode, dependant on the locale. The problem is of
trying to get these to play in the same window, so you can switch between,
say, Chinese and Arabic. I think it may be possible to work around this with
a little trickiness, though I have never tried. You can open mutilple
XIM instances from one window, and if you switch locales between these events
I think you should be able to get multiple locales working.

That said, a pure table driven mechanism can only work for trivial forms of
input method. Any sophisticated input method uses considerable language and
input method dependant logic, dictionaries, etc. From what I have seen, the
Japanese and Korean input methods are moving well beyond what a basic table
driven approach could do. Chinese certainly is. The xcin input software is
structured to support this kind of thing. You can load a series of simple GB,
Big5 and Unicode input methods, using a common table driven IM module. The
common IM framework can load other IM dependant modules to support more
specialised input schemes. These are the future!

Regards,
Steve