Random error when performing tests

[jonbrenas]

Addresses #618 .

[leehart]

Thanks @jonbrenas .

I'm not sure about the assert. I suppose the idea is that it will fail if either the number of samples or the number of SNPs is 3 or fewer. I expect there would be a ValueError raised in randint if n_snps_available was fewer than 4, so I expect n_snps will be at least 4 at the assert and so it might boil down to n_samples. Perhaps we might want a way to either ensure enough samples and SNPs at this point, or otherwise raise a clear error when that hasn't happened. But perhaps that's a different problem for a different day!

[leehart]

By the way, I don't quite understand the commit that "Drops a few unused columns", which doesn't come through as a change, presumably due to having an identical change being already merged in from master? Was that commit related to the same issue, or an artefact of something else perhaps?

[jonbrenas]

Thanks @leehart .

The assert should never be False unless the test is really poorly setup. n_samples is the sum of the sizes of the two randomly chosen sample sets and n_snps_available is the number of biallelic SNPs in the randomly chosen chromosome arm using all these samples. Both values should be significantly higher than 4.

The commits have been part of all my PRs for a while. I am not sure why. I hope they disappear once one of these PRs is merged into master.

[leehart]

Thanks @jonbrenas
CI tests failed on unrelated PCA tests. I'll try re-running.

=========================== short test summary info ============================
FAILED tests/anoph/test_pca.py::test_pca_exclude_samples[af1_sim] - ValueError: empty range in randrange(3, 3)
=========== 1 failed, 379 passed, 322 warnings in 424.39s (0:07:04) ============

[jonbrenas]

It actually fails on exactly the test that I modified because it can sometimes randomly fail on unrelated PRs (but this PR has to pass the previous tests for it to be approved, obviously, which it now does).

[leehart]

Haha, I see, of course. Thanks @jonbrenas . OK if I delete the branch from GitHub now it's been merged in?

[jonbrenas]

Done!