[AstroPy] Deployment and packaging

Tom Aldcroft aldcroft at head.cfa.harvard.edu
Thu Jun 16 13:56:13 EDT 2011

On Thu, Jun 16, 2011 at 1:33 PM, Peter Erwin <erwin at mpe.mpg.de> wrote:
> On Jun 16, 2011, at 7:07 PM, Lisa M Winter wrote:
>> In response to the issue of ease of installation for numpy/scipy/
>> matplotlib on a Mac, this is not a trivial feat for the average user.
>> You need to install the correct versions of C, fortran, etc., in order
>> to get things to work properly.  I hadn't found any easy binary
>> installs either.  While looking for installations for my undergrad
>> research students this summer, we tried the standard installs, got
>> frustrated, and turned to the extremely easy method of installing
>> everything with Enthought (a packaged python distribution).  A decent
>> set of instructions will be crucial for the packaging of the data so
>> that standard users like myself can get things running smoothly
>> without spending days trying to get everything set up (which happened
>> the first few times I tried installing on my own).
> I'm puzzled by this, since the current version of numpy, scipy, and matplotlib
> all have standard binary installers in disk-image format for Mac OS X, none of which
> require that you have C or Fortran compilers. I've had no trouble installing these on
> my various Mac laptops.  E.g., the current release of numpy (1.6.0) has
> the following three binary installers:
> http://sourceforge.net/projects/numpy/files/NumPy/1.6.0/

The problem is the blizzard of available options and conflicting
opinions.  Developers and experts have no trouble sorting through the
options and finding a clean solution that works for them on their
computer.  But Lisa's perspective and troubles need to be taken
seriously as the reality of astronomers coming to Python.  Clear
instructions are very much needed.

- Tom A

More information about the AstroPy mailing list