[Uim] question about function uim_reset_context()

[James Su]

Hi,
Maybe you should read this:
http://lists.freedesktop.org/pipermail/scim/2004-December/001289.html

So, consistent reset () behaviour among all input methods are necessary.

Regards
James Su

TOKUNAGA Hiroyuki wrote:

>Hi, 
>
>On Thu, 16 Dec 2004 18:39:21 +0100
>Mike FABIAN <mfabian at suse.de> wrote:
>
>  
>
>>TOKUNAGA Hiroyuki <tkng at xem.jp> さんは書きました:
>>
>>    
>>
>>> uim-0.4.5 is now available for download at:
>>>      
>>>
>>I have a question concerning the function uim_reset_context() in
>>uim-0.4.5 svn revision 1650. The header file uim.h says
>>about this function:
>>
>>    /**
>>     * Reset input context.
>>     * Pending string might be committed and preedit string might be
>>     erased. (Not yet implemented so)*
>>     * @param uc input context to reset
>>     */
>>    void
>>    uim_reset_context(uim_context uc);
>>
>>I added a call to this function from scim-uim to the focus_out()
>>method
>>
>>    UIMInstance::focus_out ()
>>    {
>>        SCIM_DEBUG_IMENGINE(2) << "focus_out.\n";
>>
>>       // commit pending preedit strings:
>>       uim_reset_context(m_uc); 
>>
>>    }
>>
>>which is called when the input method is switched off (with
>>Shift+Space or Control+Space) or when the focus is moved to a
>>different application.
>>
>>I tried this because I wanted to commit the current preedit string
>>when the input method is switched off.  I believe this is more natural
>>behavior. And it behaves like this when uim is used via uim-xim,
>>GTK_IM_MODULE or QT_IM_MODULE. In these three cases, the current
>>preedit string is committed when the input method is switched off.
>>
>>Only when using scim-uim it is different. With scim-uim, currently the
>>preedit string disappears when switching off the input method but it
>>is remembered and appears again when the input method is switched on
>>again. It is quite confusing when a previously edited preedit string
>>suddenly appears again after one has typed Latin letters for a
>>while.
>>
>>So I wanted to make scim-uim behave like the other variants
>>of using uim and added the call to uim_reset_context(m_uc)
>>to UIMInstance::focus_out ().
>>
>>Unfortunately this doesn't fix the problem.  After applying this
>>change, the preedit string still isn't committed when switching off
>>the input method. And it still comes back when switching the input
>>method on although the old preedit string disappears as soon as one
>>continues typing. I.e. effectively the old preedit string is
>>discarded, it only briefly appears again after switching the input
>>method on again and is then immediately discarded when the first new
>>key is typed.
>>
>>Probably the reason is that uim_reset_context() isn't yet
>>completely implemented, just as the comment in uim.h says.
>>What must be added to uim_reset_context() to make it really
>>commit the pending preedit string and then erase it?
>>Currently, uim_reset_context() looks like this:
>>
>>    void
>>    uim_reset_context(uim_context uc)
>>    {
>>       UIM_EVAL_FSTRING1(uc, "(reset-handler %d)", uc->id);
>>
>>       /* delete all preedit segments */
>>       uim_release_preedit_segments(uc);
>>    }
>>
>>I tried to add something like
>>
>>    uc->commit_cb(uc->ptr, s);
>>
>>but apparently that doesn't work and I don't know how to access
>>the current preedit string to give it as an argument to commit_cb().
>>The structure uim_context_ in context.h doesn't seem
>>to contain the complete preedit string, only preedit segments.
>>
>>How does this work in uim?
>>    
>>
>
>First, what you want to do is 'commit when switch off', not 'commit when
>focus out', OK?
>
>Personally I dislike the idea of on/off mode, but I can understand that.
>I agree with you, preedit strings should be committed when input method
>is disabled. But I don't think preedit strings should 'always' be
>commited when input method is 'focused out'. That's only my opinion, you
>may not agree.
>
>In this mail, I want to write 2 things, the most ugly hack to commit
>preedit strings and about semantic of reset.
>
>The most ugly hack is that:
>
>   if(there_is_preedit)
>      rv = uim_press_key(context, UKey_Return, NULL);
>
>instead of calling reset function. In most case, this will work. You
>know, uim doesn't have explicit commit function. If we want to add the
>explicit commit function to uim, we have to extend all input methods of
>uim. Becausein uim, preedit strings != commited strings and there is no
>way to know commit strings until commit_cb is called.
>
>Below from this line is about the future of uim.
>
>* semantic of reset
>
>Determination of semancic of reset is difficult problem, so uim doesn't
>have the policy about reset function (for now). A input method may
>commit preedit strings, then may erase preedit strings. Another input
>method may hold preedit strings (do nothing). For now, most of input
>methods don't implement reset function. Obviously we should define the
>policy about reset functionality.
>
>Now I'm thinking that
> - Preedit string should be commited when reset function called
> - Focus out should not call reset function
> - Instead of caling reset function, focus in/out function of input
>method should be called. 
>
>If you agree/disagree this, please reply.
>
>
>My explanation would not be enough, but that's my best for now. (I'm
>very sleepy, now 5:00 A.M. in Japan... 睡眠不足はいい仕事の敵だ！) So if
>you have any question, please send a mail to me.
>
>
>Regards,
>
>  
>