Re: One key stroke --> two code-points

O/H Javier SOLA έγραψε:
Thanks Simos !!

Actually, we have had these additions for a while in X11.
Hi Javier,

Checking at;a=tree;f=nls/en_US.UTF-8
does not show these lines at the end. It is possible that these compose sequences were added as a patch to the distribution package.

We will  do an issue for GTK+, and use the variable meanwhile.

What file is it in GTK+? I have not been able to find it.
In GTK+ (HEAD), the relevant file is

However, your case of compose sequences is different from the existing compose sequences, that result to a single codepoint (you require to produce two codepoints).

Therefore, the type of support you are looking for is similar to compose sequences that result to letter+diacritic mark. Several languages have characters that no pre-composed letters exist, so the compose sequence produces letter+diacritic marks (more than one codepoint). Such support is missing, and there are already bug reports for them.

Bug 341341 – Compose mechanism in simple input method doesn't support decomposed forms

Bug 345254 – dead accents should at least produce combining characters

There is a shortcut when trying to solve the above cases of compose sequences, thus the solution I expect to be different from the Khmer compose sequences. Specifically, for the Latin compose sequences, such as (it's a made up example)

<dead_acute> <t> : "t́" # LETTER T WITH ACUTE

one could convert to something like    [ dead_acute, 't', 0].
We would put 0 for the resulting codepoint because we can deduce for this category of compose sequences that the actual codepoints are 't' and 'acute' (the resulting codepoints match the body of the compose sequence).

However, for the case of Khmer, the compose sequences look independent from the resulting code points. Therefore, a new table should be required.

To cut the story short, I have filed a bug report for this,
Bug 537457 – Support compose sequences that produce two+ codepoints




Simos Xenitellis wrote
O/H Javier SOLA έγραψε:

I am working on Khmer localization (KhmerOS project).

In Khmer, some of the basic vowels (which we include in the keyboard) require two code-points, so one keystroke must generate two code points.

It used to be that we could do the conversion in KBX by generating a fictious code-point (Pablo Saratxaga explained this to us a few years ago), which was later translated to two real code-points by puting the conversion in the en-US locale file. I did work at the time.

But now this seems to have stopped working. Does anybody knows how we can fix this?
These additions (pressing a single key and producing two codepoints), are located at
The specific lines appear to be

# Khmer digraphs
# A keystroke has to generate several characters, so they are defined
# in this file

<U17fb>    :   "ុះ"
<U17fc>    :   "ុំ"
<U17fd>    :   "េះ"
<U17fe>    :   "ោះ"
<U17ff>    :   "ាំ"

GTK+ based applications duplicate the Compose file in the gtk+ library, and currently the version of the Compose file that exists in gtk+ does not include those specific compose sequences.
I think these are a recent addition.
Technically, it is possible for gtk+ to include compose sequences that produce more than one code points (requires small change in the code), however these recent Khmer digraphs are the only compose sequences using the facility now.

To cut the long story short, you can bypass for now the GTK+ version of the Compose file and use the Compose file that comes with X.Org (shown above) by setting the environment variable GTK_IM_MODULE to "xim".
This should not have adverse effect to the OLPC software.

It is important that if other keyboard layouts as well require compose sequences that produce two or more codepoints (such as Serbian), to add them to the XOrg Compose file. In the next iteration of update of the GTK+, all these compose sequences can make it in.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]