Astropy status update

[Thomas Robitaille]

Hi everyone,

I thought that since things have progressed a lot in the past few
months, I'd give a status update on Astropy! Thanks to the efforts of
various team members, the Astropy core package now already contains
the following major functionality:

- FITS input/output (astropy.io.fits), which is a direct port of PyFITS
- WCS transformations (astropy.wcs), which is a direct port of PyWCS
- VO table support (astropy.io.vo), which is a direct port of the VO module
- ASCII table support (astropy.io.ascii), which is a direct port of
the asciitable module
- A generic Table class (astropy.table)
- A constants module (astropy.constants) containing both SI and cgs constants
- Skeleton code for a generic n-dimensional data class
(astropy.nddata), which is intended to contain n-dimensional
pixel/voxel based datasets with units, WCS, errors, masks, etc. This
also includes a convolution function that can be used instead of the
scipy.ndimage.convolve, and deals with NaN values more gracefully.

There is a lot more functionality in the works, including spectra,
cosmology, photometry, coordinates, time representation, etc. so there
are many places you can get involved with helping out!

As was announced recently, we also have a website at
http://www.astropy.org/ and the latest docs are automatically built
and shown at http://astropy.readthedocs.org/en/latest/

From a technical point of view, Astropy also includes a robust
built-in testing framework (with almost 3000 tests to this day), and a
configuration framework (including support for configuration files,
logging, etc.).

We use a robust continuous integration system (Jenkins) that runs all
the tests on all supported Python and Numpy versions, on Debian, MacOS
10.7, and Windows, and we have centralized the results here:
https://jenkins.shiningpanda.com/astropy/ - anytime a commit is made
to the master branch, all the tests are run on all these different
platforms and configurations, so that we know immediately if things
are broken. As of yesterday, we now have automated coverage tests that
show us which parts of the code are well covered by tests, and which
are not: https://jenkins.shiningpanda.com/astropy/job/astropy-coverage/coveragepy/

We are very close to being able to release a first 'stable' version
(more in my next email), but in the mean time, you can already install
the latest development version and start using it instead of the
standalone PyFITS, PyWCS, VO, and asciitable modules! Note that if you
uninstall any of these four modules before installing astropy, then
when you install astropy, legacy virtual packages will be installed in
place so that 'import pyfits' still works for example.

Please report any issues with building or using what is already there!
To install:

git clone g...@github.com:astropy/astropy.git
cd astropy
python setup.py install

and you can optionally run the tests with:

python setup.py test

If you get any failures during the tests, please report them! (you can
ignore skipped and xfailed tests, which indicate known failures).

Please feel free to mention anything that I've missed!

Cheers,
Tom

[James Turner]

That's all great and will make it easier for more people to get involved.

Are these things like astropy.io.fits (etc.) intended to replace PyFITS
(etc.) development now? Later? Not sure yet? Ie. which is now the offical
version of PyFITS in particular?

Cheers,

James.

[Erik Tollerud]

Erik B and Mike can confirm for sure, but my understanding is that the
astropy versions are to become the only/primary versions in the
medium-term future. Exactly when "medium-term" is probably depends on
the release schedule and development rate of astropy, though.

--
Erik Tollerud