Re: coding systems and input methods are non-intuitive stuff

[Kenichi Handa <handa@m17n.org>]

Very sorry for the late response.

In article <f7ccd24b0701260911k6209e858y54c948d52ba1d2d0@mail.gmail.com>, "Juanma Barranquero" <lekktu@gmail.com> writes:

> C-x b *scratch* RET
> C-x RET f latin-1 RET          ; buffer coding system = latin-1
> C-u C-\ romanian-prefix RET    ; input method: romanian-prefix
> ,s                             ; character: ş (2362, #o4472, #x93a,
>                                ;   U+015F)
> <left> M-x quail-show-key RET  ; To input `ş', type ",s"
> <right>
> C-x RET f utf-8 RET            ; buffer coding system = utf-8
>                                ; input method: the same as before
> ,s                             ; character: ş (331903, #o1210177,
>                                ;   #x5107f, U+015F)
> <left> M-x quail-show-key RET  ; ş can't be input by the current
>                                ;   input method

> Now, I understand that the buffer code for these characters is not the
> same... but it is quite weird nonetheless to input a character with
> the current input method, and afterwards be told that it "can't be
> input by the current input method".

The reason why the second quail-show-key fails is that the
input method is going to insert latin-2 character but as the
buffer-file-coding-system is changed to utf-8, the character
is translated by translation-table-for-input into
mule-unicode-0100-24ff character.  Unfortunately as the data
structure of the input method still contains only latin-2
character, the current code of quail-show-key thinks the
translated character is not in the data structure.

I'm now trying to find a good way to solve this problem.

---
Kenichi Handa
handa@m17n.org