Eric Berquist

I have noticed that the MolSSI cookiecutter (and perhaps the original) places tests inside the top-level module, rather than alongside it as we have now. I support that change.

Yes, that looks good too. The interface should be via class methods that appear to use 2, 3, and 4 indices, so that we can start with storing things in...

Something like methods ```python def bond(self, i: int, j: int) -> float: ... def angle(self, i: int, j: int, k: int) -> float: ... ``` for a user interface. Ignore...

I agree with your comments above about using dictionaries; even if not directly comparing results from different gauge origins, it can be important to know which method has been used...

Yes, that's what I mean. It seems as though the size of the attribute list could grow very quickly if everything is placed directly inside of it. Does this go...

I'll think about the structure once 1.5.2 is out, but I can comment on the first part. In practice, one is interested only in GIAO for the gauge problem because...

#425 and #1002 parsed NMR shielding tensors for ORCA. All other parsers are missing implementations; we need - [ ] ADF - [x] DALTON - [ ] GAMESS (US) -...

@oliver-s-lee would you be able to run an NMR chemical shielding calculation with Turbomole? I can try coming up with an example input if you can't or don't have time.

At work we use [`towncrier`](https://pypi.org/project/towncrier/) to generate the changelog at each version bump from fragment files that get added during each PR.

In some ways, towncrier solves a problem that doesn't exist. It's a tool that just adds lines to a text file almost verbatim once a release is minted. This problem,...