[Python-Dev] multi-version parallel-installed modules (original) (raw)

Michael Foord fuzzyman at voidspace.org.uk
Wed Aug 18 14:50:20 CEST 2010

On 18/08/2010 15:34, Neal Becker wrote:

Is there any proposal to accommodate having parallel-installed multiple versions of modules?

I have client code in multiple projects using version x.y of a C-compiled module A. I want to test a new version x.z of module A, but all client software needs to be recompiled against the new version. If I just install the module, all the other client software breaks. I know I could test using virtualenv, but there would be a lot of modules to install into virtualenv to run the tests, so this would be cumbersome. I'd prefer to have multiple version co-exist so I could update projects to the new version at my convenience. How does this situation happen? I have lots of c++ code using pyublas, which allows c++ code written to the boost::ublas interface to operate on numpy vectors/matrixes. pyublas is built against boost libs. pyublas installs a module, whose purpose is to register conversions. When I update boost libs, I have to rebuild pyublas and install the updated module. Then rebuild my client software modules. If pyublas is built against a different boost version than my client modules, the conversions will fail.

I believe that setuptools / distribute already has a mechanism for supporting this. It basically does it with sys.path hackery.

As far as I know there are no plans to include this in distutils2 - but Tarek can correct me if I am wrong.

All the best,

Michael Foord

Python-Dev mailing list Python-Dev at python.org http://mail.python.org/mailman/listinfo/python-dev Unsubscribe: http://mail.python.org/mailman/options/python-dev/fuzzyman%40voidspace.org.uk

-- http://www.ironpythoninaction.com/

More information about the Python-Dev mailing list