On Tue, Dec 15, 2009 at 4:34 PM, David Malcolm <dmalcolm(a)redhat.com> wrote:
Chatting with Colin Walters at FUDcon I got the impression that it
be easier to add Python 3 support to gir than to port all of the
hand-coded python bindings to python 3.
I don't know if anyone's looked at Python3 for PyGI; but if the goal
is to say demo driving GTK+ from Python 3, then I think patching PyGI
would be a fairly straightforward way to do it. You'd probably just
have to tweak the unicode handling in a handful of code points
effectively, whereas pygtk has a lot of custom API.
In the bigger picture, it seems likely that the pygtk/PyGI split will
continue, though I have mixed feelings about this. I feel bad at
effectively driving a fork in the python binding community, but on the
other hand ran into real limitations of the old system (stuff not
bound, stuff bound inconsistently, etc). It's kind of unfortunate
though to have two different transitions in progress at the same time
The hard question is porting pygtk+dependencies, as well as key
consumers (anaconda, system-config, etc). My take would be to focus
for a little while on making Python 3 (along with PyGI) show
noticeable benefits when developing *new* apps, and see if we can get
away with also porting pygtk->PyGI at the same time. That's a problem
which begs for an automated rewriting tool too.