Add pyupgrade-docs to pre-commit

[verhovsky]

a follow up to #5705

since pytest supports only Python 3.6+, we can use f-strings in the documentation

[nicoddemus]

LGTM, thanks a lot for that and writing https://github.com/verhovsky/pyupgrade-docs. 👍

Could you also add that to our pre-commit configuration? Thanks!

[verhovsky]

@nicoddemus done.

thanks for writing/maintaining pytest, it's one of my favorite libraries!

[nicoddemus]

LGTM, thanks!

(Note to squash this before we merge it)

[asottile]

just making sure you saw this @verhovsky (since you're not watching the repo): verhovsky/pyupgrade-docs#3

[verhovsky]

@asottile I pinned pyupgrade's minor version to 2.10.* and released it as pyupgrade-docs v0.2.4. I hope that works for you.

@asottile I pinned pyupgrade to 2.10.0 and released it as pyupgrade-docs v0.2.5. I hope that works.

[nicoddemus]

Thanks @verhovsky!

@asottile anything else needs to be done here?

[asottile]

Thanks @verhovsky!

@asottile anything else needs to be done here?

yeah we probably shouldn't use the tool since it integrates in an incompatible and unsupported way (and likely will either break or be stuck on an old version)

[nicoddemus]

@asottile thanks!

@verhovsky thanks anyway for the PR! Perhaps once the issues (which I'm not familiar with TBH) are fixed, you can open another PR? Thanks!

[RonnyPfannschmidt]

@asottile any opinion on allowing pyupgrade to be invoked as a python function? i'm under the impression they just don't want to take a "i has no api" and i see providing a explicit minimal api as opportunity to make tooling easier

[asottile]

no I don't want to provide an api because it paints me into a corner and prevents me from adding features. the only supported api is the entrypoint

[RonnyPfannschmidt]

@asottile i was thinking along the lines of exposing a "limited" api in the sense of having only a parse_args and a fix_content(args, original_content) with a explicit note that the type/strucutre of the parsed args is open

[asottile]

no thanks, they might as well use the entrypoint at that point

[RonnyPfannschmidt]

@asottile i see, in that case it may be necessary to document that, also stdio replacement is still a pain