Use located in GADT #932
Use located in GADT #932
Conversation
There was a problem hiding this comment.
Nice, the use of unLoc here just seems to be a historical oversight 👍
Just for reference what the inconsistency fixed by this PR is concretely:
| Input | Output before this PR | Output with this PR |
|---|---|---|
foo ::
A -> B |
foo ::
A -> B |
foo ::
A -> B |
data Foo where
Foo ::
A -> B |
data Foo where
Foo ::
A ->
B |
data Foo where
Foo ::
A -> B |
mrkkrp
force-pushed
the
gadt-cleanup
branch
2 times, most recently
from
569e790
to
daaa6c5
Compare
|
I can't figure out why CI is not working on this PR 🤔 |
|
Do you have CI enabled for PRs originating from forks? |
|
No, that would be unsafe. I vaguely remember that it should work if I force-push on your PR (which I did). Anyway, it worked magically when I pushed I also granted you write access. Use it wisely and don't change the indentation step, please. |
|
Ah sure; Fourmolu uses GitHub Actions, which lets you click a button to manually approve running a workflow. Not sure how buildkite does it.
Ha. Even though I work on Fourmolu, I actually prefer 2-space indentation. Ironic |
Is there a particular reason why
p_hsTypewas being called onunLoc quantifiedTypeinstead of using thelocatedcombinator? It seems like it was initially implemented when GADT args + result type were separate, so there wasn't a single HsType object representing the entire type to be located, but now that we havequantifiedType, it seems better to be consistent between function signatures + GADT constructors.