Representation-polymorphic HasField (fixes #22156)

This generalises the HasField class to support representation polymorphism, so that instead of

type HasField :: forall {k} . k -> Type -> Type -> Constraint

we have

type HasField :: forall {k} {r_rep} {a_rep} . k -> TYPE r_rep -> TYPE a_rep -> Constraint

This change was approved by the CLC here: https://github.com/haskell/core-libraries-committee/issues/194#issuecomment-1800412006

The required code changes were surprisingly small. However, a small wrinkle is that this change makes GHC.Records no longer inferred as Safe due to the new import of TYPE from GHC.Exts. As far as I know TYPE is not currently exported from a Trustworthy module, so I've resorted to marking GHC.Records as Trustworthy. I don't know if there is a better solution here. EDIT: this happened anyway with the ghc-internal split, and I don't think it is particularly specific to GHC.Records anyway.

added 2 commits

• 0ad47fd7 - Representation-polymorphic HasField (fixes #22156 (closed))
• 05c2e01b - Mark GHC.Records as Trustworthy

Compare with previous version

Please comment this Trustworthy business, since (as you say) it's really a wart.

changed this line in version 3 of the diff

Fair enough, I've added a note.

... and removed it again, since the ghc-internal split already made GHC.Records into a Trustworthy module, and on reflection this has very little to do with GHC.Records specifically.

As you say, surprisingly easy. Do comment the Trustworthy business though!

approved this merge request

added 1 commit

• c04fcaba - Mark GHC.Records as Trustworthy

Compare with previous version

added 34 commits

• c04fcaba...afca46a4 - 32 commits from branch master
• 0bb36d78 - Representation-polymorphic HasField (fixes #22156 (closed))
• ef383e21 - Mark GHC.Records as Trustworthy

Compare with previous version

added 25 commits

• ef383e21...0d12b987 - 23 commits from branch master
• 0187ab63 - Representation-polymorphic HasField (fixes #22156 (closed))
• 9ece3194 - Mark GHC.Records as Trustworthy

Compare with previous version

added 77 commits

• 9ece3194...707a32f5 - 76 commits from branch master
• 9c4d2575 - Representation-polymorphic HasField (fixes #22156 (closed))

Compare with previous version

I've rebased this now that !11400 (closed) has landed (which, it turns out, needed to mark GHC.Records as Trustworthy anyway). Hopefully it should now be good to merge?

changed the description

The code looks good to me. I approve with the following asterisks:

• In hasfieldrun03.hs, f1 and f2, g1 and g2 give identical results. I'd change something so that it's not completely trivial (e.g. make f2 return f1 + 1).

• There's a merge conflict and ghc-9.10 already had an rc1 released, so I'm afraid you have to adapt the updates to changelogs to ghc-9.12.

approved this merge request

added 468 commits

• 9c4d2575...b1e0c313 - 467 commits from branch master
• c4f4d054 - Representation-polymorphic HasField (fixes #22156 (closed))

Compare with previous version

added 1 commit

• 8554fb65 - Representation-polymorphic HasField (fixes #22156 (closed))

Compare with previous version

assigned to @marge-bot

I will attempt to batch this MR (!12647 (closed))...

Merged in b84b91f5

closed