Hi, I've experience working at companies where, because of the network set up, having a long PYTHONPATH and searching it is quite a heavy task and can slow down the start up of the interpreter when there are lots of imports.
As a proof of concept I wanted to look at a map-based approach. The theory being that instead of having a linear list of paths to search with heavy disk checks, if we could encode the location of the modules in an environment variable we would be able to jump to the correct location much quicker. As a result I've written: https://github.com/michaeljones/porter Which implements an import hook (PEP 302) that can be set up to look at an environment variable with the following format: name=location[:name=location[:name=location[:...]]] This is parsed into a standard dict and used for quick lookups when possible, falling back to the standard Python import mechanism when nothing is found. The implementation is new and not production tested, but I was hoping to get some external thoughts on the idea, especially to hear any functionality that this might disrupt. I've only written basic tests to check import behaviour and am aware that I probably don't really appreciate the full consequences of this idea. Cheers, Michael -- http://mail.python.org/mailman/listinfo/python-list