[AstroPy] Plan to port astropy to python3?

Mark Sienkiewicz sienkiew at stsci.edu
Wed Jun 8 11:26:02 EDT 2011

Peter Erwin wrote:
> So Xavier Gnata's "numpy,scipy,matplotlib and pyfits are now working with python3"
> should perhaps be interpreted as "these packages are either working with Python 3,
> or there is active development on Python 3 versions" ... which is arguably reason enough
> to start thinking about Python 3 versions of other things.

I thought to ask because I saw yet another statement that implied that 
numpy/scipy/matplotlib were all ready for python 3, but in truth 
matplotlib isn't there yet.

The release notes for numpy 1.6 and scipy 0.9 both explicitly state that 
they DO support python 3, even though the FAQs say that they do not.  
They do work, mostly: numpy 1.6 has 1 erroring test and scipy 0.9 has 1 
error and 1 failure.

I just finished setting up a preliminary python 3 test environment 
yesterday, and I found that about half of the python packages that I 
normally provide are not available as a release package for python 3.  
This includes matplotlib, but also ipython, PIL, Pmw, even Sphinx.  (I 
wonder how they did the python 3 docs - maybe with the pre-release sphinx?)

I had one big idea as a result of all of this:  If you care about python 
3, put a statement on your web page about whether your package 
does/does-not work in python 3.  Even for packages that have python 3 
support, finding this information sometimes feels like you have to look 
in the bottom of a locked filing cabinet stuck in a disused lavatory 
with a sign on the door saying "Beware of the Leopard".

As far as I know, none of the material in astrolib is currently ready 
for python 3.  Should I put a notice on the wiki to that effect?

Mark S.

More information about the AstroPy mailing list