Back-porting PRs to 0.8.4 branch

[j-wags]

We're aiming to make a 0.8.4 release simultaneous with 0.9.1. If possible, these releases will be feature-equivalent.

0.8.4 will be the last release of the openforcefield namespace.

This PR cherry picks the development that has occurred on master to the 0.8.4-rc branch. It's my first time doing a cherry pick of this scale, so I suspect that this will require a little polishing to get CI happy again.

• Fixing roundtrip from qcschema to qcschema #790
• Add an Initial ELF Implementation Which Uses RDKit #793
• Add Molecule.smirnoff_impropers and .amber_impropers #799
• Set electrostatics 1-4 scaling based on value specified in force field #802
• Add the 0.9.0 Release Notes #813
• Match file paths case-insensitive when loading ForceField data #816
• Handle box vectors set with list #827
• Remove deprecated np.float #828
• Expose ELF conformer selection through the OpenEye wrapper. #831
• Expose ELF conformer selection through the Molecule API #832
• Add Support for Multi-conformer and ELF10 WBOs #839
• Don't use existing conformers for WBO calculation during system creation #849

[mattwthompson]

I have a few failing tests from the start

• Some of the links in the README pointed to the master branch, which is now broken because permalinks do not redirect to respect the new code paths. I updated these to point to the 084-update branch.
• z_3_hydroxy_propenal.sdf wasn't in the tree, so I added it
• Some number of the recent PRs were missing (at least Expose ELF conformer selection through the OpenEye wrapper. #831, probably also a WBO change). These are my currently failing tests
  • openforcefield/tests/test_molecule.py::TestMolecule::test_apply_elf_conformer_selection
    • (missing method in toolkit wrappers? or did not mean to add in test?)
  • openforcefield/tests/test_toolkits.py::TestOpenEyeToolkitWrapper::test_assign_fractional_bond_orders (several combinations of ELF10 methods and molecules)
  • openforcefield/tests/test_toolkits.py::TestOpenEyeToolkitWrapper::test_assign_fractional_bond_orders_neutral_charge_mol
  • openforcefield/tests/test_toolkits.py::TestOpenEyeToolkitWrapper::test_assign_fractional_bond_orders_multi_conf

This was done with my current development environment with both OpenEye toolkits and The RDKit installed. I also ran without OpenEye toolkits and everything passed - effectively a false positive because of

@requires_openeye
class TestMolecule:

Running with The RDKit but no OpenEye toolkits produces the same failed tests as above.

The current deprecation warning seems accurate. Would it be worth it to also link to #819, here or in the release notes?

__main__:1: DeprecationWarning: Importing this package as `import openforcefield.XXX` and `from openforcefield import XXX` was marked for deprecation in version `0.8.3`. From version `0.9.0` onwards this package will need to be imported as `import openff.toolkit.XXX` and `from openff.toolkit import XXX`. See the `0.8.3` release notes for more information.

[j-wags]

Thanks, Matt. I'm going to keep banging on this today and will ping you when it's ready for review!

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (084-rc@520bc28). Click here to learn what that means.
The diff coverage is n/a.

[j-wags]

Don't review this yet. I think there's actually a significant test failure. https://github.com/openforcefield/openff-toolkit/runs/2036523162#step:12:50065

This is code adjacent to some of our current WBO work/bugfixes. I'm not sure that it's good to keep this as xfail. https://github.com/openforcefield/openff-toolkit/blame/master/openff/toolkit/tests/test_toolkits.py#L2930

I'll debug more tomorrow.

[mattwthompson]

Took a brief skim through the files and couldn't find anything objectionable. I'm restarting the tests; if the only failure is the stochastic one we've already discussed, LGTM

[j-wags]

Good news❓ Nothing fails any more❓❓ Hooray❓❓❓