Setup.py for python bindings?

asked 2016-03-08 13:07:20 -0600

gmcgrath gravatar image

updated 2016-03-08 13:21:34 -0600

This is somewhere between a question and a feature request. We have a handful of requests for opencv to be installed on our computational cluster which we've done, however getting it to import into different python environments is a huge problem with the way the bindings are built currently. We could use the standard approach, but this will only add it to the root versions of python included in the OS and would only be available on the non build nodes with some additional RPM work. Additionally it would be unavailable to anybody that decides to spin up a virtualenv, add an anaconda environment, or attempts to evaluate the intel python variant.

If it were possible to build the python libraries separately against to current environment version of python either with a setup.py or separate cmake call this would resolve this issue and make the tool significantly easier to deploy to end users. Especially compared to the current model where python discovery is done by cmake during the initial setup so you have to rebuild the entire toolset in order to target a different python version (I've found no easy way to change the version of python discovered and issue a new 'make' command at least).

Any help would be appreciated.

edit retag flag offensive close merge delete