[SciPy-dev] [Numpy-discussion] Thoughts on making it easier for numarray users to transition to scipy_core

Robert Kern robert.kern at gmail.com
Tue Dec 13 03:58:41 CST 2005


Andrew Straw wrote:

> One serious issue with eggs is that they are a package system outside
> the control/knowledge of package managers such as debian's apt-get.

This isn't *quite* true, but this perception is certainly the source of all the
brouhaha. Eggs themselves are not a packaging system. In fact, they were
originally not designed to distribute Python packages per se at all. Rather they
were designed for the distribution and discovery of plugins and data. Later,
they were adapted for distributing Python packages, and easy_install was written
to provide basic package management for systems that don't have any.

It is *easy_install* that is the package system that tries to cover part of the
territory of apt-get/yum/etc. But we don't need easy_install to use eggs. Nor do
we even have to use eggs in the form of zip files, either. Most of the fancy egg
features (namespace packages, discovery of metadata without importing, but not
multiple versions) will work fine as long as the egg metadata is provided
somewhere on sys.path.

At this point, the argument on debian-python and the Distutils-SIG has settled
down with a few predictable holdouts. It looks like that Debian packages for
setuptools-using Python packages will install the Python package parts regularly
and provide the appropriate metadata.

To provide a concrete example, suppose a Debian user has scipy_core and the rest
of stable scipy provided by proper Debian packages.

  /usr/lib/python2.4/site-packages/
      scipy/
      scipy-core.egg-info/
      scipy.egg-info/

He might also have some sandbox packages installed in /usr/local, too, because
they certainly won't be packaged by Debian.

  /usr/local/lib/python2.4/site-packages/
      scipy/
      scipy-sandbox-delaunay.egg-info/
      scipy-sandbox-maxent.egg-info/

Presuming that I can get the __init__.py fragments distributed appropriately,
everything will simply appear as one scipy package.

-- 
Robert Kern
robert.kern at gmail.com

"In the fields of hell where the grass grows high
 Are the graves of dreams allowed to die."
  -- Richard Harter




More information about the Scipy-dev mailing list