search

MobleyLab / chemper

Repository for Chemical Perception Sampling Tools
MIT License
19 stars 10 forks source link

[WIP] Try using psi4numpy's utils for jupyter notebook tests #23

Closed bannanc closed 6 years ago

bannanc commented 6 years ago

I'll start by seeing if the tests pass here, they pass locally.

bannanc commented 6 years ago

These passed locally after I pip install pytest-shutil which I'm already doing in the travis file. I think I must be missing something installed locally, still trying to figure out what it is.

bannanc commented 6 years ago

OK, I fixed the potentially more obvious problem that jupyter wasn't installed for the travis tests. I don't understand why the OE tests are failing now.

codecov-io commented 6 years ago

Codecov Report

Merging #23 into master will increase coverage by 0.93%. The diff coverage is 100%.

bannanc commented 6 years ago

OK so I'm a dummy and missed that all the OE tests were failing, there is something wrong with the license. I'm encrypting a new one right now.

bannanc commented 6 years ago

OK, it isn't the license itself, its that it can't be accessed from the directory that the jupyter notebook is running in. Everything else passes. I'm willing to bet that is the same reason #21 is failing too you just don't see the license error and it gets caught in the notebook.

bannanc commented 6 years ago

@davidlmobley @janash @dgasmith any of you used had issues accessing a license file before?

davidlmobley commented 6 years ago

I have not had that problem. Can you reproduce locally?

bannanc commented 6 years ago

I haven't tried a travis docker, but following the conda installation on a new environment on my computer works, I tried having travis make the *.py files in the example folder, I wander if there is a way to generate them as test_*.py files instead?

I haven't been able to find anything with people running into similar problems online.

davidlmobley commented 6 years ago

OK, I have zero experience on this and I don't think I'll get to dig in to it in the immediate future. Maybe get a second opinion from @nathanmlim , as (a) he can probably figure it out faster than me, and (b) may also benefit from being able to use the same technology...

bannanc commented 6 years ago

I think nbval is a better way to test notebooks, you can check that the output hasn't changed and it is more general. I merged PR #21 I'm going to close this one, but keep the branch incase something else goes wrong and we want to switch back.