Previously: Use HasCallStack and error in GHC.List and .NonEmpty · Issue #5 · haskell/core-libraries-committee · GitHub
Generally I think it works really well, but there are a couple of caveats:
- it’s tough to explain how to use the Partial class correctly; in my experience people tend not to understand where they should propagate a Partial constraint and just slap an unsafePartial on in situations where they really shouldn’t. Of course, people can write whatever they want in their own projects, and it’s easier to ensure that important libraries are at least using Partial correctly
- similarly, there’s a risk of people seeing Partial constraints and incorrectly thinking that the type checker will guarantee that any function lacking a Partial constraint is guaranteed to be total.
I suspect the main barrier to introducing Partial to Haskell would be backwards compatibility though: if an un-discharged Partial constraint doesn’t cause a compilation failure, then realistically there will be tons of functions that lack Partial constraints but should have them, which limits the utility of the constraint; if an un-discharged Partial constraint does cause a compilation failure then you’ll break basically everyone’s code and people will be very upset.