[Pythonmac-SIG] "Best" Mac Python under Leopard?

[David Warde-Farley]

On 23-Aug-08, at 9:25 AM, Dav Clark wrote:

> Andrew, I have
>
> I will add that if you need pyobjc, I had weird problems with pyobjc  
> 1.4 on leopard and a python.org framework build.  Stuff just didn't  
> work (wasn't receiving messages from an external library for  
> hardware interface).  I switched to the system python w/ objc 2.0  
> and everything worked fine.  I don't know of any serious issues  
> between 2.5.1 and 2.5.2 that provide a compelling argument to  
> upgrade to the Framework build either.

Dav and I were arguing this in person at the SciPy08 code sprint :)  
Opinions in the room were mixed.

First of all, if you want to build bundle .app's that are  
redistributable, it's a good idea not to use the system Python,  
especially if you want to support Tiger (10.4). There are things in  
Apple's Python (i.e. DTrace support) that I imagine won't play nicely  
on an older OS.

That said, DTrace support is a huge bonus for some people. Personally  
I don't really trust the Apple engineers to care enough about Python  
to iron out all the issues one might face.

> I know of no other way to get bleeding edge numpy, scipy, ipython,  
> etc. on OSX except to build it yourself (which is not so bad these  
> days - but does require install of some random things like  
> fortran).  But if you can try the system python with like 15 minutes  
> of setup, you may as well.  It's easy enough to switch to something  
> else later!

gfortran is actually remarkably easy to install now thanks to a well  
maintained binary (the link is on scipy.org, which is undergoing a  
much-needed overhaul very soon).  Building SciPy is indeed not too  
painful anymore, no matter what Python distribution you're using, and  
I personally feel it's worth it to check out the latest trunk from SVN  
-- SciPy is moving so fast that stable releases often don't include  
vital new features (NumPy is more stable, but there's still a lot of  
exciting stuff being added, particularly stride_tricks).

The nice thing about EPD is that it's a one-stop shop for the  
scientific computing folks, and maintained by extremely competent  
folks who have a vested interest in making this stuff work on all the  
platforms they support (not to mention that well over half of the  
Enthought guys at SciPy08 were using Macs!).

Unlike back in the Tiger days, it seems that most of the options on  
the table (System, Python.org, EPD) are quite acceptable, but YMMV.  
Certainly if you run into specific issues with any one of these  
distributions, please make a note of it on this list so that others  
can benefit. We really need to spruce up the MacPython FAQ again...

Regards,

David