Reinteract at PyCon

[Owen Taylor]

Just wanted to briefly mention here that I'm going to be talking about
Reinteract at PyCon 2009 in Chicago at the end of March:

http://us.pycon.org/2009/conference/talks/

The talk is going to be introductory in nature - what is Reinteract,
why it is neat. But if other people with an interest in PyCon future
development and internals are going to be there, I'd love to meet up
informally to chat.

I haven't been doing much work on Reinteract recently ... I've been
busy with a project at work (http://live.gnome.org/GnomeShell) that is
taking a lot of my programming energy. But I'd like to get back to
some Reinteract hacking soon.

Some near term tasks:

- Finally finish getting Kai's replot rework committed
- Upgrade Windows build to Python 2.6 (if all the deps have been
upgraded - haven't checked)
- Look at some bugs on OS X
- Some sort of configuration for font size
- Merge my deep-mutations branch
- Do an offical release (0.5.0) and announce it
- Get my package into Fedora (passed the review process months ago,
but I never went ahead and actually built it)

Beyond that, my current interest is mostly making Reinteract more
robust without changing the user interface a lot. Is it possible to
make much of the Python standard library work correctly without weird
side effects and warnings out of the box? Things like that.

- Owen

[Kai Willadsen]

2009/1/22 Owen Taylor <owta...@gmail.com>:

> Beyond that, my current interest is mostly making Reinteract more
> robust without changing the user interface a lot. Is it possible to
> make much of the Python standard library work correctly without weird
> side effects and warnings out of the box? Things like that.

I'd be really interested in this kind of work, as I think that
reliability is a hard requirement for data analysis usage. I think
it's reasonable to expect that re-executing modified code might not
always be 100% accurate, as long as there is a guarantee that the
first run of a worksheet will always produce the same results as just
running 'python foo.rws'.

Also, while it would be great to have reinteract always work
flawlessly, it would probably suffice if it exposed more information
about when the history recording may have gone awry, so that the user
can try to workaround it if necessary.

On a related note, I think it would be great to extend the
non-reinteractive bits of reinteract to make it a better python
development tool. For example, the ability to compile & run (but not
record) could be added for python files; there could be better docs
integration; some basic profiling support could be added, etc.

--
Kai

[Michel S.]

Hi Owen,

On Jan 22, 12:46 am, Owen Taylor <owtay...@gmail.com> wrote:

>  - Get my package into Fedora (passed the review process months ago,
> but I never went ahead and actually built it)

I just rediscovered Reinteract yesterday -- I was testing the new
version of 0install, and one of the available application is an early
version of Reinteract. Then noticed the new website and built 0.4.3,
which is a big improvement over the 0install version.

I went ahead and submitted a review request before reading this thread
-- bizarrely, searching in Bugzilla did not show your request, which
is still marked NEW.

Would you like a comaintainer? I'd like to branch it for EL-5 as well
-- some of the classes at my university are taught using Python, and
Reinteract is much more beginner-friendly than IDLE.

Thanks,

--
Michel Salim

[owta...@gmail.com]

On Feb 28, 2009 6:40pm, "Michel S." <michel...@gmail.com> wrote:

> I just rediscovered Reinteract yesterday -- I was testing the ne

> version of 0install, and one of the available application is an early
> version of Reinteract. Then noticed the new website and built 0.4.3,
> which is a big improvement over the 0install version.
>
> I went ahead and submitted a review request before reading this thread
> -- bizarrely, searching in Bugzilla did not show your request, which
> is still marked NEW.

Hmm. Well, I just have to get into to the RESOLVED/FIXED state. Probably not this evening, but tomorrow definitely.

> Would you like a comaintainer? I'd like to branch it for EL-5 as well
> -- some of the classes at my university are taught using Python, and
> Reinteract is much more beginner-friendly than IDLE.

From experience a co-maintainer never hurts, so sure! I'll approve your requests in pkgdb when I get a chance.

EL-5 is going to be a problem though - the as-yet unmerged deep-mutations branch drops support for Python 2.4 ... I didn't want to spend time trying to keep the 2.4 code branches working in rewrite.py. I'm not sure that trying to support 2.4/2.5/2.6 is a winning combination. (I haven't spent any significant time thinking about Python 3.x. Probably the sensible thing is to get it to a stable "1.0" as 2.x only, then switch the devel branch to Python 3.x once that has sigificant deployment.

- Owen

[Michel S.]

On Feb 28, 7:07 pm, owtay...@gmail.com wrote:

I just took a stab, and discovered other problems with using Python
2.4; the codebase uses these other features:

- try/except/finally -- not so bad, but maintaining the patch might be
annoying
- with

Not to mention, gobject.property is not in (py)gtk 2.10

Running it uninstalled, I've not even managed to get to rewrite.py --
the application never even started.

Is there an earlier version that works on Python 2.4? If not, then
well, the Python classes are normally taught using Windows labs
anyway, so handicapping with having to support 2.4 (or even 2.5)
rather than strictly using 2.6/3.0 common features would probably be
too much of a hindrance.

Thanks,

--
Michel

[Owen Taylor]

Yeah, there's quite a bit of stuff like that, both intentional and
accidental. If someone wanted to pick up 2.4 support, I'd take the
patches, and I'd to try not to intentionally break it again while the
2.4 support was maintained. But I don't see it as a particularly
useful activity. There are lots (thousands) of RHEL 5 desktop users
internal to Red Hat, but the people who do programming (and Python
programming) slant heavily to Fedora for their desktops. I think that
can be generalized: if you are an active Python programmer, you want
something newer than Python 2.4.

> Running it uninstalled, I've not even managed to get to rewrite.py  --
> the application never even started.
>
> Is there an earlier version that works on Python 2.4? If not, then
> well, the Python classes are normally taught using Windows labs
> anyway, so handicapping with having to support 2.4 (or even 2.5)
> rather than strictly using 2.6/3.0 common features would probably be
> too much of a hindrance.

I've never personally seen Reinteract work on Python 2.4... I started
development with Python-2.5. Evan Martin contributed patches to
support Python 2.4 and GTK+-2.8, so my assumption is that when those
patches were applied (~November 2007) it did work for a while. It
definitely was no longer working by the time I did the 0.4.1 release.

- Owen