[immodule-qt] Status of Qt 4 port and some questions

[Daisuke Kameda]

Lars Knoll <lars at trolltech.com> wrote:

> I'm finally back from my vacations and we have done most of the work of 
> importing your code to our current Qt4 branch. I'll shortly mention the 
> changes I have done up to now and then have some questions to you.

Thank you for your work and report.


> 
> While importing the code into Qt4, I noticed, that some methods in the API 
> were not really needed and thus removed them:
> 
> I removed:
> * QApplication::postIMEvent as it's nothing else than qApp->postEvent, and I 
> couldn't see why you would need a slot for this.

I think that this code is required because of compatibility with Qt/E.
Probably, Yamaken will explain in detail.

> * QInputContext::deletionRequested(). I could remove this signal by using a 
> QPointer for QWidgetPrivate::ic. Now you can just call "delete this" in case 
> of an fatal error in the input context without dangling pointers in other 
> places.

It sounds good.



> I renamed QWidget::getInputContext() to QWidget::inputContext() as the old 
> name doesn't fit with our naming conventions.
> 
> I will probably need to change a few more method names a little to fit our 
> naming conventions and to make it possible to implement text widgets in a 
> cross platform way (ie. QLineEdit should not have any #ifdef's for X11, and 
> should be able to achieve the same results on Qt/Windows). 

It is good, I think.



> Now my questions:
> 
> I had a look at your documentation of isPreeditRelocationEnabled(). As fas as 
> I understand you this is used to not loose the preedit string when focusing a 
> different widget. In this case the preedit string is transferred over to the 
> new focus widget. If I understand you correctly, this is just a hack to not 
> loose the string, and the correct behaviour would be to keep the preedit 
> string in the first widget so one can continue when focus comes back to it. 
> If that is indeed the desired behaviour, I think I might have a way to 
> implement it in the general framework without lots of additional work for the 
> implementors of input methods. I think it might at the same time enable text 
> widgets to support reconversion.

Sorry, it is difficult for me to reply to this question in detail.
The desired behaviour is not single, I understand.
The intention of our code which you understood is also one of it.
Therefore, isPreeditRelocationEnabled() is needed as a flag.

Probably, Yamaken who also develop uim will explain in more detail.
Please give us detailed explanation, Yamaken.


> QInputContext::language() currently returns a QString. Wouldn't a QLocale 
> object be more correct?

I think so. Why isn't QLocale used in present code, Yamaken? Is an understanding 
that the code remains taking Qt3.2.x into consideration right? 


> The QInputContext::mouseHandler() method takes a parameter 'x'. Am I correct 
> that this is the current cursor position inside the preedit string?

I think that the value of x depend on the position which 
clicked the mouse instead of the current cursor position. 
So, QWidget::sendMouseEventToInputContext() is needed.

-- 
Daisuke Kameda <kaminmat at cc.rim.or.jp>