Birds-of-a-feather sessions

Proposed BoF sessions

• Topic (coordinator: Person)
  • Attendee 1
  • Attendee 2
• Testing (coordinator: )
  • Fernando Perez
  • Stefan van der Walt
  • Jarrod Millman
  • Christopher Burns
• Visualisation (coordinator: )
  • Prabhu Ramachandran
  • Gael Varoquaux
  • Stefan van der Walt
  • Jarrod Millman
  • Christopher Burns
• Documentation (coordinator: )
  • Joe Harrington
  • Gael Varoquaux
  • Stefan van der Walt
  • Michael Droettboom
  • Jarrod Millman
  • Christopher Burns
• PEP 225 - extra operators for Python (coordinator: )
  • Fernando Perez
  • Jarrod Millman
• Sprints (coordinator: Jarrod Millman)
  • Jarrod Millman
• Local Groups (coordinator: Jarrod Millman)
  • Jarrod Millman
• Data IO (coordinator:)
  • Christopher Burns

Testing

NumPy and SciPy recently moved to Nose. For IPython, Fernando Perez wrote plugins to handle doctests inside C extensions. Alan McIntyre has written directives to handle random output, and to insert NumPy into the test namespace. Sage has special ways of ensuring that the test output is always the same, including those that deal with randomisation. This year has seen increased reliance on the buildbots, but still no automated code coverage reports. Producing code coverage for Python C-extensions remains elusive.

Documentation

NumPy had its first documentation marathon, and implemented a web frontend that allows the community to edit docstrings in a wiki-like fashion. Many projects moved over to Sphinx, including IPython, matplotlib, NumPy, Sympy, etc. Michael Droettboom wrote special directives to include plots and math in the matplotlib documentation. For its reference guide, NumPy generates Sphinx-compatible docs by parsing and marking up the NumPy docstrings. Pauli Virtanen implemented an autodoc directive for NumPy.

Low-level code-wrapping and optimisation

Sage had a GSOC student working on NumPy integration with Cython. Many libraries are wrapped by ctypes, but it is not clear how to best handle platform-specific compiler padding. David Cournapeau and Zachary Pincus have posted plans for aligned allocators. At Enthough, Ilan Schnell has been working on interesting weave improvements and compiler decorators.

PEP 225 - extra operators for Python

There is currently a dicsussion taking place in the Python-dev list regarding the possibility of resurrecting PEP 225 to extend the language with additional operators. These would allow, amongst other things, direct support for matrix multiplication of numpy arrays at the syntax level. Feedback from the scientific community on this matter will be very useful to guide the decision making process of the Python-dev group.

See http://www.python.org/dev/peps/pep-0225/ for details, as well as these threads:

• http://mail.python.org/pipermail/python-dev/2008-July/081508.html
• http://mail.python.org/pipermail/python-dev/2008-July/081551.html

Coding Sprints

Coding sprints are an extremely useful practice. Over the last year there was a number of SciPy related coding sprints (NumPy, SciPy, Sage, IPython, MayaVi, NIPY, etc.). It would be extremely useful to have a discussion about what works and what doesn't.

Local Groups

At UC Berkeley, we have started developing a SciPy community. We have started having regular meetings. For years, there have been local Python groups. It would be useful to discuss the pros and cons of developing more local SciPy groups as well as share our experiences at Berkeley.

Data IO

Data io (files, arrays, images) is a recurring topic on the mailing lists, particularly for new users. There has been some work recently on improving the io in numpy and scipy (like Robert's work on npy/npz files), and discussions on image io (a possible PIL plugin). A review of the current state of io and possible paths forward, to ease common io tasks, would greatly benefit the scipy community.