Manipulating foreign file paths

[hasufell]

In GitLab by @Ericson2314 on Jun 22, 2022, 13:33

I not infrequently want to manipulate Windows paths on Unix, and Unix paths on windows; this sort of thing comes up e.g. when planning tasks for a heterogeneous build farm to perform.

filepath has always confused me as to whether this sort of thing is supported. From the way the modules are CPP'd I think the answer is "yes", but the docs in e.g. https://hackage.haskell.org/package/filepath-2.0.0.3/candidate/docs/System-AbstractFilePath-Windows.html still have "On Windows..." "On Unix...." which is very confusing. I think we should CPP the docs to make this clear?

Also, is it good that AbstractFilePath = PosixFilePath on Unix? IMO this is certainly not wrong, but it is a foot-gun. It is akin to saying Int = Int64 on x86_64. Might we consider making AbstractFilePath newtype instead?

---

Sorry for bringing this stuff up at the 11th hour, but I figure it is much easier to deal with now than post release. We have a limited window to address this and any other would-be breaking change "regrets" from the old way of doing things, but that would require a filepath-3 or AbtractFilePathTake2, neither of which is very appealing!

[hasufell]

In GitLab by @maerwald on Jun 22, 2022, 17:43

I answered your question here: https://gitlab.haskell.org/haskell/filepath/-/issues/116#note_438455

Second part of the answer. So yes, AFPP is designed exactly for this.

[hasufell]

In GitLab by @Ericson2314 on Jun 22, 2022, 23:57

@maerwald Very glad to hear that the newtypes are in place!

Can we add more CPP for the docs then?

[hasufell]

In GitLab by @Ericson2314 on Jun 23, 2022, 24:22

I started a draft MR in https://gitlab.haskell.org/haskell/filepath/-/merge_requests/111 to try to be completely clear about what sort of docs CPP I had in mind. :)

[hasufell]

In GitLab by @Ericson2314 on Aug 29, 2022, 21:21

mentioned in commit ghc/packages/filepath@cca0dc3b6f5ac0aadcc0b26537ebd246eaee042b