BUMP deepNOG to 1.1.0 (targeting PyPI) and update CI configurations

[LokiLuciferase]

Just what it says on the tin. Plus, some additional refactoring and added comments.

[LokiLuciferase]

It is my understanding that we currently determine coverage from the AppVeyor CI build - is it easily possible to use the linux build of Travis instead @VarIr?

[VarIr]

Thanks for the PR! I see no blockers here, just some minor comments.

[VarIr]

Just saw that all Travis builds failed. While the Linux builds fail due to the same issue we see in deepnog, the mac builds fail due to something wrong with prodigal. These should be investigated.

[VarIr]

About coverage: It seems to me, we actually run codecov on Travis, but not on AppVeyor? In any case, it should be simple to enable it on any platform. codecov claim, they automatically merge reports from different platforms. So perhaps just add a line "codecov" to .appveyor.yml after the tests (and pytest --cov, unless we already have it).

[VarIr]

Btw, see the deepnog .appveyor.yml for how to obtain torch on Windows. (Hasn't found its way into setup.py, yet).

[LokiLuciferase]

Just saw that all Travis builds failed. While the Linux builds fail due to the same issue we see in deepnog, the mac builds fail due to something wrong with prodigal. These should be investigated.

Yeah, it was possible that the prodigal binary compiled for linux wouldn't run under macos (and it sure as hell won't under Win10). I can look into having it compile from source under linux/macos, and ship the windows binary with the repo.

[LokiLuciferase]

• I have now added all prodigal binaries for linux, win32 and darwin architectures. I had to adapt some of the protein calling code since the pre-compiled packages don't allow passing gzipped files.
• I also changed the license of phenotrex to GPL v3 to comply with the license of prodigal.
• Finally, I added some more tests for file I/O and added a groups file for Sulfate_reducer to be able to test LOGO CV.

[codecov]

Codecov Report

Merging #21 into master will increase coverage by 11.00%.
The diff coverage is 91.80%.

@@             Coverage Diff             @@
##           master      #21       +/-   ##
===========================================
+ Coverage   71.76%   82.76%   +11.00%     
===========================================
  Files          21       21               
  Lines        1098     1120       +22     
===========================================
+ Hits          788      927      +139     
+ Misses        310      193      -117     

Impacted Files	Coverage Δ
phenotrex/util/taxonomy.py	0.00% <ø> (ø)
phenotrex/structure/records.py	70.96% <50.00%> (-4.90%)	⬇️
phenotrex/ml/shap_handler.py	76.37% <80.00%> (+0.78%)	⬆️
phenotrex/ml/prediction.py	58.97% <83.33%> (-1.64%)	⬇️
phenotrex/ml/cccv.py	92.39% <100.00%> (ø)
phenotrex/ml/feature_select.py	95.16% <100.00%> (+6.92%)	⬆️
phenotrex/transforms/annotation.py	98.66% <100.00%> (+84.63%)	⬆️
phenotrex/util/plotting.py	88.88% <100.00%> (+0.20%)	⬆️
phenotrex/transforms/__init__.py	60.00% <0.00%> (-40.00%)	⬇️
... and 5 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 02da355...c2ece98. Read the comment docs.

[VarIr]

It's good to see the coverage go up! I'd prefer to see the gene calling in a context manager, if possible, but otherwise LGTM.

[VarIr]

Btw, the Travis errors will likely go away with PyTorch 1.5. We might want to allow Travis to fail the Linux builds until then (or simply ignore the errors).

[LokiLuciferase]

Merging this now. @VarIr maybe you can upload to pypi when you find time?

[VarIr]

Thank you for the PR! I've just created a new release and uploaded to PyPI.