[WIP] Coverage Report

[jthorton]

Description

Adds a coverage report function which is exposed to the CLI preprocess group. I think this should possibly come after step 1 as we do not want to spend time making conformers for molecules that will latter fail?

The function labels each molecule and tries to generate an openmm system this way we can also check for charge failures as well as missing parameters.

The past molecules are moved to the specified output folder and a coverage report json file is made which should be easy to combine.

Todos

Notable points that this PR has either accomplished or will accomplish.

• [ ] Add tests
• [x] Add coverage reporter

Questions

• [ ] Do we want to include the coverage numbers for molecules that fail to build an openmm system?

Status

• [ ] Ready to go

[j-wags]

Started reviewing this today, which turned into "eh, I could just add that...", which turned into a kinda beefy commit. I really like how you did this. If tests pass I think this is good to merge in an initial form.

A few notes --

• I've added tests, at least the ones I could think of
• I re-jiggered the return signature of single_molecule_coverage so that the exception doesn't go in the parameter dictionary, but is instead passed as a separate return value. In my experience, exception handling has so many bizarre edge cases that I've learned to keep the "good" and "bad" datastreams as far apart as possible.
• I've temporarily disabled partial charge calculation on this step since it was taking forever. We could substitute in smaller test molecules and bring it back.
• I really like what you've done with the modularization here -- I was bothered by how my validate and generate python methods have filenames as inputs and folders as outputs. Handling the I/O in cli.py is a much cleaner and more testable design. I've moved this further in that direction.
• I want to test to ensure we catch errors for both (the case where a molecule has a valence term that we can't provide), and for (the case where a molecule that is covered by our valence terms, but that fails in antechamber). I can use C(=O)OBr for the former, but I can't seem to think of a molecule for the latter.

[j-wags]

Oh, and I mostly tested the exception/file handling. We should add a test with something like ethanol and really go through the resulting dict with a fine-toothed comb to ensure that every value is correct.