Firstly, is anybody working on Qt bindings for D? If so, the rest of this message is probably irrelevant and can safely be ignored. But continuing under the assumption that nobody is... :P

I'm wondering what would be involved in a project like that. Has anybody looked into hooking into C++ libraries enough to know what would be the best plan of attack? (And for Qt in particular, I hear it uses some custom pre-processor magic on top of the standard C++ one; is this right, and is this likely to be a huge problem?)

Qt Jambi (http://doc.trolltech.com/qtjambi-1.0/com/trolltech/qt/qtjambi-index.html) seems to be doing a similar thing for Java, so I presume they got over any problems with multiple inheritance. What other major problems would there be?

Any random thoughts appreciated.

afaik, most of the preprocessor magic is to create a decent signal/slot
mechanism. slots get indexed in order of appearence and an implicitly
generated function "invoke" dispatches calls with these indeces.
Qt relies on C++ linkage to access these invoke/emit (and other) functions.
to properly pass the Qt framework calls to D delegates, an intermediate
C++ object would have to be generated for each D widget, i guess.

on the other hand, the fact that trolltech laid down a lot of action to create this signal/slot thing, they should be tempted to port their whole framework to D ;)

just jeff wrote:
> Firstly, is anybody working on Qt bindings for D? If so, the rest of this message is probably irrelevant and can safely be ignored. But continuing under the assumption that nobody is... :P
> 
> I'm wondering what would be involved in a project like that. Has anybody looked into hooking into C++ libraries enough to know what would be the best plan of attack? (And for Qt in particular, I hear it uses some custom pre-processor magic on top of the standard C++ one; is this right, and is this likely to be a huge problem?)
> 
> Qt Jambi (http://doc.trolltech.com/qtjambi-1.0/com/trolltech/qt/qtjambi-index.html) seems to be doing a similar thing for Java, so I presume they got over any problems with multiple inheritance. What other major problems would there be?
> 
> Any random thoughts appreciated.

It's much, much easier to bind a C library than a C++ library.

qtc is a C binding to Qt.

 - Gregor Richards

PS: I don't know if it uses the same nasty preprocessor junk as Qt C++ ...

PPS: However, after binding it, you won't have anything like the original interface - you'll have a very OO-lacking C interface.

just jeff wrote:
> Firstly, is anybody working on Qt bindings for D? If so, the rest of this message is probably irrelevant and can safely be ignored. But continuing under the assumption that nobody is... :P
> 
> I'm wondering what would be involved in a project like that. Has anybody looked into hooking into C++ libraries enough to know what would be the best plan of attack? (And for Qt in particular, I hear it uses some custom pre-processor magic on top of the standard C++ one; is this right, and is this likely to be a huge problem?)
> 
> Qt Jambi (http://doc.trolltech.com/qtjambi-1.0/com/trolltech/qt/qtjambi-index.html) seems to be doing a similar thing for Java, so I presume they got over any problems with multiple inheritance. What other major problems would there be?
> 
> Any random thoughts appreciated.

We use Qt for a C++ project.  We have developed many wrappers for Qt classes for technical reasons.  One reason is that we hate annoying moc files. Anyway, we are on our way to having our entire GUI written using wrapper classes.  We have a somewhat complex solution to eliminate the need for moc files.  Because of this, we only have one moc file for all of our wrapper classes.  We retain most of the Qt functionality, including signals/slots, and event interception using custom C++ delegate classes.  One drawback is that we cannot override virtual methods in Qt classes very easily.  A long term goal of mine has always been to port our project to D.  But if we did, we would drop Qt and use a GUI native to D.  We would have to rewrite our wrapper classes to accommodate this, but its easier than rewriting the entire GUI.

-Craig

"just jeff" <jeffrparsons@optusnet.com.au> wrote in message news:epkgmb$2b33$1@digitaldaemon.com...
> Firstly, is anybody working on Qt bindings for D? If so, the rest of this message is probably irrelevant and can safely be ignored. But continuing under the assumption that nobody is... :P
>
> I'm wondering what would be involved in a project like that. Has anybody looked into hooking into C++ libraries enough to know what would be the best plan of attack? (And for Qt in particular, I hear it uses some custom pre-processor magic on top of the standard C++ one; is this right, and is this likely to be a huge problem?)
>
> Qt Jambi (http://doc.trolltech.com/qtjambi-1.0/com/trolltech/qt/qtjambi-index.html) seems to be doing a similar thing for Java, so I presume they got over any problems with multiple inheritance. What other major problems would there be?
>
> Any random thoughts appreciated.

Does anybody know what approach Qt Jambi has taken, and whether any of their work could be used for a D version?

Any idea how Py Qt, Qt Jambi et al did it? An OO wrapper around the "flattened" qtc? o_0

just jeff wrote:
> Any idea how Py Qt, Qt Jambi et al did it? An OO wrapper around the "flattened" qtc? o_0

I don't know about either of those ... PyQt might, but it's probably not too difficult to write C++ extensions to Python. Qt Jambi probably doesn't, I can't imagine Qt depending on a Qt binding they didn't write ...

QtC#, for example, uses QtC.

 - Gregor Richards

Gregor Richards wrote:
> just jeff wrote:
>> Any idea how Py Qt, Qt Jambi et al did it? An OO wrapper around the "flattened" qtc? o_0
> 
> I don't know about either of those ... PyQt might, but it's probably not too difficult to write C++ extensions to Python. 

PyQt uses something called SIP.
http://www.riverbankcomputing.co.uk/sip/index.php

> Qt Jambi probably doesn't, I can't imagine Qt depending on a Qt binding they didn't write ...
> 
> QtC#, for example, uses QtC.
> 
>  - Gregor Richards