[IPython-dev] ipython0.9 and python2.4

[Fernando Perez]

On Sun, Sep 14, 2008 at 7:26 PM, Gael Varoquaux
<gael.varoquaux at normalesup.org> wrote:
> On Sun, Sep 14, 2008 at 07:15:05PM -0700, Fernando Perez wrote:
>> > Frontend code is cleaned up. This has been check in my branch. Brian was
>> > centralizing the fixes in his branch. I am not too sure what the status
>> > is there.
>
>> I just merged his code, where he'd merged from you.  I've added some
>> decorators to implement the win32 fixes more cleanly.
>
> Looks good.

Thanks.  I'll wait for Brian's report on win32 tests and will then push.

> Brian and I where toying with the idea of adding a method to the wx
> frontend. This method is for execution of a command in the interpreter,
> while displaying it also on the screen.
>
> The changelog can be seen on
> http://bazaar.launchpad.net/~gael-varoquaux/ipython/ipython-sync-frontend/revision/1149
> and you can pull from my branch if you want.
>
> I had to add this in my work at Enthought, and it was really an oversight
> from my side not to put this in the API of the line frontends. The reason
> being that this is actually not trivial to get right (finding out what
> event to call, and even more important, how to interact with the GUI
> event loop in a way that is robust under Vista), so I think it should be
> part of the API. I had kinda decided that we would have to wait for 0.10,
> as I missed the 0.9 release, but now that we are doing a 0.9.1, I am
> pushing for it to go in. It's been used fairly heavily in the work I have
> been doing at Enthought, so it's been tested in this regards. I can't add
> a unit test, however, as it relies on the GUI event loop.

I'd really prefer to keep this release as the emergency, fix-2.4
release.  This code is small, but it's not properly documented
(docstrings without args, no first line summary, no examples even if
they aren't doctests,...).

I believe you that it's been used, but the reality is that you could
easily have made a silly mistake (we all do), etc.  We're rushing a
release on a Sunday night to fix a brown-paper bug, this is not the
time to put in *any* new code outside the scope of the fixes we needed
(the test decorators were part of fixing broken tests).

Sorry, but I'm going to be very conservative here.  Let's just not
have another problem tomorrow, when none of us will have time to fix
it.  Brian and I were on the phone on this and he also seems to agree
with me, so I hope you'll understand.  I'll be super happy to pull it
in for 0.10 once you finish documenting it (for example, one version
of the execute method ignores the hidden argument, that should be
explained since it still takes it).

Cheers,

f