embray at stsci.edu
Thu Jun 16 13:01:59 EDT 2011
On 06/16/2011 12:50 PM, Mark Sienkiewicz wrote:
> I note that the discussion of packaging has completely displaced any
> discussion of what astropy is, what it does, or what might be included.
> I also note that the packaging mechanism does not matter if we have no
> software to package. Let me offer this:
> - astropy can be installed with "python setup.py install".
> - astropy can be installed with "pip install astropy".
> - Anything that can be installed via those methods can also be
> distributed via _any_ of the other packaging mechanisms under
> discussion, subject to someone committing to routinely perform the
> packaging work at each release.
> I suggest that we declare these three statements as "good enough for
> now" and get back to answering the question "What is astropy?"
This is a point I tried to make at some point yesterday, only not as
clearly :) I think that such straightforward installation should be the
goal here. Beyond that there's really nothing more to say until we know
what we're installing.
As for "what is astropy"--something I would like to add is that I agree
with Luigi Paioro that a single "astropy" package is undoubtedly
overkill. It would be huge and difficult to maintain. A more loosely
coupled Scikits-like approach is better. In this case "astropy" would
be a namespace package, and one would almost never `pip install astropy`
unless they really wanted *everything*.
Each sub-package would relate to astropy in so far as it follows the
same coding standards as, and have interfaces similar to every other
package in astropy. Also with as little duplication of functionality as
More information about the AstroPy