add test files

[lilyminium]

Fixes #2939

Changes made in this Pull Request:

• add test file patterns in setup.py

I think this is why they're not included in the release. Maybe we should add a note to datafiles.py?

PR Checklist

• Tests?
• Docs?
• CHANGELOG updated?
• Issue raised/referenced?

[codecov]

Codecov Report

Merging #2940 into develop will decrease coverage by 0.01%.
The diff coverage is n/a.

@@             Coverage Diff             @@
##           develop    #2940      +/-   ##
===========================================
- Coverage    93.01%   93.00%   -0.02%     
===========================================
  Files          187      187              
  Lines        24940    24895      -45     
  Branches      3261     3261              
===========================================
- Hits         23198    23153      -45     
  Misses        1694     1694              
  Partials        48       48              

Impacted Files	Coverage Δ
auxiliary/base.py	90.75% <0.00%> (-0.57%)	⬇️
coordinates/base.py	94.85% <0.00%> (-0.27%)	⬇️
util.py	89.39% <0.00%> (-0.08%)	⬇️
transformations/__init__.py	100.00% <0.00%> (ø)
package/MDAnalysis/transformations/fit.py	100.00% <0.00%> (ø)
package/MDAnalysis/transformations/wrap.py	100.00% <0.00%> (ø)
package/MDAnalysis/transformations/rotate.py	100.00% <0.00%> (ø)
package/MDAnalysis/transformations/__init__.py	100.00% <0.00%> (ø)
package/MDAnalysis/transformations/translate.py	100.00% <0.00%> (ø)
...ge/MDAnalysis/transformations/positionaveraging.py	100.00% <0.00%> (ø)
... and 1 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b2b7bcb...baa871f. Read the comment docs.

[orbeckst]

Thank you. Did you manually try building and running the failing tests?

[lilyminium]

I feel this is most efficiently done (and avoids numerous other "transfer_to_memory() calls) if u is loaded in_memory=True to start off with but not sure if there's a reason for this sequential nature of events

[IAlibay]

If I remember correctly, in the original issue I couldn't get the code to work even if the test was isolated. It's possible I made a mistake somewhere, but definitely would be worth testing in isolation.

[lilyminium]

@orbeckst yes, I built the package from a new clone python setup.py sdist and ran tests on that. pytest -n 12 had no errors. python setup.py sdist on the directory I normally work out of includes all test files in the distribution regardless of what's in package_data for some reason. That might be how it slipped detection in the pre-release tests.

[orbeckst]

From my side this looks fine. Thank you @lilyminium

@IAlibay if you have more comments on the transfer_to_memory() then please block the PR.

[IAlibay]

I'll temporarily block this to allow for clarification (it's possible I'm just missing something really obvious here).

So the transfer to memory does somehow fix test_given_mean, but it's not exactly clear to me how/why pca_aligned._mean changes on memory transfer.

In fact, in both cases pca._mean is set to a NoneType? Unless I'm missing something obvious, it looks like transferring to memory just ends up providing a pre-aligned trajectory to test_given_mean which renders pca._mean obsolete?

I'm really hoping that I'm missing something obvious, but it looks to me like in all cases the else of this branch is never reached?

mdanalysis/package/MDAnalysis/analysis/pca.py

Lines 219 to 224 in 2c5e385

	if self._mean is None:
	self.mean = np.zeros(self._n_atoms*3)
	self._calc_mean = True
	else:
	self.mean = self._mean.positions
	self._calc_mean = False

This would explain why codecov has it as untested? https://codecov.io/gh/MDAnalysis/mdanalysis/src/develop/package/MDAnalysis/analysis/pca.py

Edit:

Obviously this entire question is based on the fact that I assume that test_given_mean should be testing what happens if mean is given and self._calc_mean is True. If that's not the intended purpose then that's fine.

[lilyminium]

As @IAlibay noted the test is a bit meaningless so I've reverted that for now and can fix in a separate PR.

[IAlibay]

lgtm

[IAlibay]

How odd, one of the travis jobs ran for over 50 mins, I've restarted it to see if it's just a bad node.

[IAlibay]

Thanks for the fix @lilyminium. Note; we'll probably also need to push this to master for 1.0.1.

[orbeckst]

I'll do the cherry-pick to master.

[orbeckst]

Fixed in 2ebb052 on master.