dxtbx -> Python 3 about dxtbx HOT 4 CLOSED

this now exists. The regression test fails when tests that used to pass on Python 3 no longer do so. Python 2 testing remains in place as a separate build job.

from dxtbx.

I think this all sounds good, I like the sound of the separate feature branches - I'm hoping to learn a bit through this process as I'm not yet up to speed with all of the differences between python 2&3. For the python3-regression test branch, will we receive emails if this one fails or does it rely on checking the github page?

from dxtbx.

I was going to reply "no, we can't have emails because Jenkins can't distinguish between expected and unexpected test failures", but pytest actually makes it really easy to keep a list of conditionally expected test failures, so I've just committed f3daa9c and we now should have Python3 test regression emails.

(The same logic will also apply when you run tests locally.)

from dxtbx.

This work has now been completed.

• All tests pass on Python 2.7 and 3.6 in Jenkins and Travis
• 2 tests were disabled on Python 3 in 54113e8
  This is due to .pickle format files not being loadable. To fix this would require a substantial amount of time and changes in cctbx and libtbx for a use case that is unclear to me. If these files are actually used then someone will have to put the effort in, otherwise I suggest the relevant formats should be retired.
• We have achieved A+ rating on LGTM with 11 remaining warnings. This should go down to 3 very soon since
  • 1 was just fixed with f0f5b00
  • 3 will be fixed by #61
  • 4 will be fixed by #63
• There are 51 remaining flake8 warnings, they break down as follows:
  • 28x E402 module level import not at top of file. These are caused by Python2/3 cross-compatibility work and should be ignored
  • 7xF401, 3x F403, 13x F405 - these are covered by #62 and subject to the DIALS conventions in dials/dials#838
• Code coverage has slightly decreased in terms of files and classes (81%) and conditionals (44%). This is not entirely unexpected given that over the course of the work approximately 6% of dxtbx was removed (1500 lines), with the refactoring attention having been biased towards tested code rather than untested code.
• Project closed 10 day ahead of schedule.

Thanks to everyone involved.

from dxtbx.