On Fri, Jan 1, 2010 at 6:28 AM, Dag Sverre Seljebotn <da...@student.matnat.uio.no> wrote: > Happy new year everyone! > > David Cournapeau from the NumPy project has (with good reason, IMO) > grown fed up with distutils, and is launching an effort to get the > scientific Python community off the distutils habit. (This has of course > given rise to some controversy, but David appears to have given it a lot > of thinking over the course of many years). > > It's been a long-standing issue that the scientific Python community has > more complicated needs to build systems than the Python web development > community, which currently drives the distutils/setuptools/distribute > projects. Also end-users rather than sysadmins typically installs > software and so on. > > The full discussion is here: > > http://thread.gmane.org/gmane.comp.python.numeric.general/35276 > > As a summary, David propose: > > a) An easy way to get off the distutils for building without loosing > the ability to very easily make a distributable Python package. David > has started "toydist" to accomplish this: Declarative metadata rather > than a setup.py. > b) Tools to create RPMs, DEBs, OS X installers, Windows installers etc. > from Python packages. > c) A website a la PyPI for scientific Python projects (but done > right!) (dubbed SciPI, but perhaps jokingly), with automatic build > farms/validation on package upload etc.
Thanks for bringing this discussion to our attention on sage-devel. Toydist, SciPi -- It's good these names are temporary. Above, when David says "get people off of distutils", it's important to emphasize that he means "get people off of distutils, setuptools, and distribute by doing something new, like is doing with HASKELL and R". Anyway, that thread you link to is really interesting, since it's all about the problem of making it easy for end users to install Python-based scientific software. This is one of the problems Sage also addresses, though our goals are much more focused. It's an incredibly important problem to solve, in order to compete with MATLAB. So +1 to any discussion or work that contributes to solving this problem. It's an interesting discussion and worth being aware of. > I think these developments are worth keeping track of (and support, at > least morally) for the Sage project as well -- imagine being able to > upload a simple Python package to the SciPI website, and have SPKGs > generated and tested automatically (in addition to DEBs, Windows > installers etc.). > > This could make the typical "SciPy stack" developers automatically or > more easily provide Sage SPKGs for lots of scientific software (such as > mpi4py, PyTables, etc.) > > Pierre Raybaut of the Python(x,y) project has already voiced his support > of the project and says he might in the future convert the Python(x,y) > package distribution system to what David proposes. > I spent some time a year ago learning about Python(x,y)'s package distribution system from Pierre, since some of us wanted to reuse something from it for Sage's windows porting work. It's mostly just some Windows program that records mouse strokes as Pierre installs standard Python windows binaries, then plays back those recorded actions. He's just automating somewhat installation of what the numpy/scipy/etc., developers release (not to minimize this -- he provides a very valuable service!!). If those developers switch to toydist, he'll have no choice but to switch... -- William -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
