[Uim] Questions on coding for uim

[TOKUNAGA Hiroyuki]

Hi,

On 18 Jun 2005 15:28:07 +0900
Park Jae-hyeon <jhpark at kias.re.kr> wrote:

> Hello,
> 
> I have bug reports on UIM.  I am using 0.4.6final1-3 on Debian sarge.
> 
> 1) UIM scheme interpreter does not have independent boolean values,
>    and predicates do not return boolean values.  For example, with
>    uim-sh, I get
> 
>        uim> #t
>        1
>        uim> #f
>        ()
>        uim> (equal? #t 1)
>        t
>        uim> (boolean? 1)
>        t
>        uim> (boolean? '())
>        t
>        uim> (not '())
>        t
> 
>    while, with guile, I get
> 
>        guile> #t
>        #t
>        guile> #f
>        #f
>        guile> (equal? #t 1)
>        #f
>        guile> (boolean? '())
>        #f
>        guile> (boolean? 1)
>        #f
>        guile> (not '())
>        #f
> 
>    As far as I know, scheme should distinguish #t from 1, and #f from
>    '().  '() should count as true, not false.  Also, a predicate such
>    as equal? should return either #t or #f, and nothing else.  I found
>    that this was already reported in the bugzilla as bugs #617 and
>    #642.

Yes, guile's implementation is right. The scheme interpreter of uim has
some problems. Now Kazuki seems writing a new scheme interpreter for
uim, so please wait a while.


> 2) Switching input method does not always work.  I run gedit with
> 
>        $ GTK_IM_MODULE=uim gedit
> 
>    with no other program using uim, and do
> 
>        $ uim-im-switcher-gtk
> 
>    After I choose another input method in uim-im-switcher-gtk's window
>    and click Okay, sometimes the input method is switched to the new
>    one, and sometimes it is not.

Maybe it is caused by the bug of uim-helper-server. Please test again
with 0.4.7. 0.4.7 will be released on June 26.


> 3) The optional third argument of delete in util.scm, defaults to =.
>    http://srfi.schemers.org/srfi-1/srfi-1.html#delete says that the
>    default comparison procedure should be equal?.

I think you are right. I'll improve.


> 4) If I click on a submit button before committing preedit in a
>    mozilla form, the preedit reappears the next time I begin input.
>    For example, at www.google.com, I write in the form,
> 
>        はやくちことば
> 
>    using anthy, but I do not commit the string by hitting RETURN yet.
>    I click on the Google Search button.  Then, I see the search
>    results.  Now, when I type a key in the form, はやくちことば
>    appears in the preedit area plus the newly typed key.  Maybe this
>    is a bug of mozilla, not of uim, in view of the fact that I have
>    the same phenomenon with kinput2 + canna.  Probably, this is
>    related with the issue of reset, focus-in, focus-out, and
>    focus-move.

Yes, this problem is related on the issue of reset. I'll investigate
later, but maybe it's difficult to fix from uim side.


> I have a question on writing scheme code for uim, related to bug 1).
> The second argument of context-update-preedit is a list of pairs of
> attributes and strings.  According to anthy-input-state-preedit, it
> seems that the list is also allowed to have #f or '() as one of its
> elements.  Is it #f or '() that can be an element of the second
> argument of context-update-preedit?

attr should be a number which defined in im.scm. i.e. preedit-none, 
preedit-underline, preedit-reverse, preedit-cursor or preedit-
separator. If there's no attribute, you should use preedit-none. On the
other hand, when the segment has no str, you should use "".


Regards,

-- 
TOKUNAGA Hiroyuki
tkng at xem.jp