feat(next): keep original string on unknown variants

[lasantosr]

Summary

This PR aims to keep the original values for unknown variants, if they are generated.

This allows downstream crates to act on them, for example logging the unexpected value. This is useful for example on the ApiVersion for the webhook versioning, in order to be able to properly determine if the event comes from a newer api version.

Checklist

• using conventional commits to hightlight user-facing fixes and features

[arlyon]

Thanks for doing this. I am for it but I will let @mzeitlin11 pull the trigger

[mzeitlin11]

Thanks @lasantosr! Main initial question is both the purpose of the From<String> impl and its implementation. There should already be a FromStr impl for each enum - best if the raw string to enum variant logic doesn't get duplicated (since some of these enums are so giant, the FromStr impls are already the heaviest functions according to cargo bloat).

[lasantosr]

The purpose of From<String> is to have the infallible version taking directly the owned String instead of duplicating it.

I thought about replacing the implementation, removing FromStr instead of duplicating for enums with the unknown variant, which can't fail.

I didn't consider timing, so I can update the PR with one of the following, or any other you might consider (for enums with unknown variant only):

• removing From<String> impl
• removing FromStr impl
• updating FromStr impl to call From<String>
• updating From<String> impl to call FromStr

[arlyon]

I think it is reasonable to clone in the exceptional case of encountering an unknown string. When calling stripe apis the bottleneck will be the network (meaning the additional resource cost of a clone is negligible). So my vote is option 1.

[mzeitlin11]

I also like 1 for the reasons given above too. The FromStr impl for these types also already uses std::convert::Infallible for the error, so the inability to fail is already represented

[arlyon]

@lasantosr ping me here when you have removed the impl and I will get it released :)

[lasantosr]

@arlyon I've updated the PR removing the From<String> impl, but tests are failing for another reason (I've run them locally and they're fine)

[lasantosr]

Hey @arlyon it's been a while, ¿could you have a look at this and the other next PRs?

[mzeitlin11]

Sorry for the delay @lasantosr, if Alexander doesn't get to it by next weekend I should have more time then to review. For now, I just reran the CI jobs to see if happens again

[mzeitlin11]

Looks like the weird errors went away, now just a clippy one that if you don't see locally probably happens because of a different clippy version running in CI

[lasantosr]

Thanks @mzeitlin11!

It was by bad, it's fixed now.

[mzeitlin11]

Thanks @lasantosr!