Project depends (implicitly) on both python-Levenshtein and python-Levenshtein-wheels

[SnoopJ]

The changes introduced by #18 do not apply to .pre-commit-config.yaml, which still declares a dependency on python-Levenshtein. Probably the config needs to be updated to declare python-Levenshtein-wheels as well, since it is a more recent fork of the former.

This came up in the #python channel on Freenode's IRC network when a user was confused by the apparent need for gcc caused by this change. It confused a few of us for a bit and I thought I'd document the problem here.

Aside: fuzzywuzzy works perfectly well without any extension modules, and depends on this speedup as an extra. Perhaps this project could do the same? 😇

[Lucas-C]

fuzzywuzzy works perfectly well without any extension modules, and depends on this speedup as an extra. Perhaps this project could do the same?

Sounds good!
Do you want to edit your PR to put this in place?

[SnoopJ]

Do you want to edit your PR to put this in place?

As I take a look at the pre-commit documentation, this might be trickier than I thought, because it seems like there's no equivalent concept of an "extra" on their end. The only option I can think up is to make two py.test hooks.

However, as I think about it, I'm not really sure why this discrepancy was a problem for the downstream user, since it's only in the local hook, not one you're publishing for others to use. Let me ask them, and if it seems like there's a valid concern there, I'll open a separate issue.

clarification: even if I change the relevant entry in additional_dependencies to something that will definitely break when creating an environment (I used thisisnotarealmodule), I can't reproduce the error they were seeing, because the hook is local to this project. Users who follow the instructions in your README should not experience this problem (though it's good to have consistent versions anyway)