search

msmbuilder / msmbuilder-legacy

Legacy release of MSMBuilder
http://msmbuilder.org
GNU General Public License v2.0
25 stars 28 forks source link

Python3 #384

Closed rmcgibbo closed 10 years ago

rmcgibbo commented 10 years ago

This PR is pretty big:

  1. Full python2/3 compatible code, using a lot of __future__ imports, some functions from six, the python2/3 compabilitiy library (its a single file and a copy is included in mdtraj, so it's just using that one). Also, a really helpful library to find some of the issues was futurize http://python-future.org/automatic_conversion.html, which points out a lot of stuff in the code that needs to change.
  2. Remove remaining c code in src/ext. It wasn't actually being compiled (the directives to compile it in setup.py had already been removed.

2a. Then it seemed silly to have all of the code in src/python, with nothing of value besides python inside src/, so I just moved the src/python directory to MSMBuilder.

  1. Tests on travis now run automatically on both python2.7 and python3.3.
  2. I noticed there there were two copies of the hubscore reference data, in reference/hub_ref and reference/transition_path_theory/hub_ref. I deleted one.
kyleabeauchamp commented 10 years ago

Overall looks reasonable

rmcgibbo commented 10 years ago

@schwancr: look this over?

schwancr commented 10 years ago

the assign_hierarchical test fails for me

schwancr commented 10 years ago

Ok, looks fine. I made some comments that are nitpicky

schwancr commented 10 years ago

So the ZMatrix.h5 doesn't match the reference anymore. I tried fastcluster==1.1.7 to see if that was the change, and the test still didn't pass