hotfix: register "vertex" fields only in unstructured meshes

[neutrinoceros]

PR Summary

fix #2430

This is a hotfix, I'm convinced there are better ways to solve this but they would require much more work (refactoring).

PR Checklist

• pass black --check yt/
• pass isort . --check --diff
• pass flake8 yt/
• pass flynt yt/ --fail-on-change --dry-run -e yt/extern
• New features are documented, with docstrings and narrative docs
• Adds a test for any bugs fixed. Adds tests for new features.

[neutrinoceros]

added a test that fails on master and passes here

[neutrinoceros]

switching to draft since #2850 will make this a tiny bit simpler.

[neutrinoceros]

So I think I'm going to split this between the hotfix and the experimental part where I build a long exception list to avoid except Exception

[neutrinoceros]

moved the exception expansion there #2851

[matthewturk]

Hm, I guess I replied over email but it didn't get through. Anyway, I think there's a possibility we could have vertex-centered fields in some grid calculations. Does this preclude that?

[neutrinoceros]

I don't think so. My understanding is that vertex fields only make sense for data where ds.index.meshes is defined. If I'm wrong then this needs to be updated of course.
More broadly, I think this is one more case where having specialised abstract Dataset subclasses could help.

[matthewturk]

@neutrinoceros As long as this does not preclude vertex centered fields (which might be handled with some weird logic at present) for grid fields, this looks good to me and I think it can go in.

[neutrinoceros]

I don't really know how to test this, or how to demonstrate this is fine

[matthewturk]

OK, let's hold off, it may not be the case that there are any grid frontends to be worried about at the moment with this issue.

…

On Wed, Sep 16, 2020 at 9:10 AM Clément Robert ***@***.***> wrote: I don't really know how to test this, or how to demonstrate this is fine — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#2848 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAVXO4KHJDHJUJHUD3XR6TSGDBMRANCNFSM4P5FES7A> .