Fix charlist formatting issue on '\"'

[sabiwara]

Close #13363

[josevalim]

I am worried this may introduce other regressions. What if we check if the string has a “ and, if so, we use the sigil c with single quotes?

[sabiwara]

I am worried this may introduce other regressions.

This is a fair point, I actually started working on a prop-based tests to catch some other cases and indeed it did.

What if we check if the string has a “ and, if so, we use the sigil c with single quotes?

Could be a strategy, but it might be a bit more tedious to check for interpolation (not a big deal).
I was also evaluating replacing the whole String.replace pipe with a recursive function, will see how it goes and report back.

[sabiwara]

@josevalim a91b626 seems to pass everything my prop-based test throws at it, and might be safer than the replace approach, WDYT?

[josevalim]

I am honestly a bit worried because we are adding logic for handling '' => ~c"", which is a change of delimiters, into general delimiter logic.

So I would still prefer to emit ~c'' if there are ". It has an added benefit of not really changing the contents of the string.

[sabiwara]

I am honestly a bit worried because we are adding logic for handling '' => ~c"", which is a change of delimiters, into general delimiter logic.

That's a great point. Will go with your approach!

[sabiwara]

@josevalim we should be good now, and the property-based tests still pass ✅

[josevalim]

Beautiful, please backport to v1.16 as well!

[sabiwara]

Backported 677ad61
FYI had to fix a small test that had been removed on main a99a9cf