Add sig lens for where clauses #2966
Conversation
| -- | Get the type string of a binding id | ||
| bindToSig :: Id -> HscEnv -> GlobalRdrEnv -> IOEnv (Env TcGblEnv TcLclEnv) String | ||
| bindToSig id hsc rdrEnv = do | ||
| env <- tcInitTidyEnv |
There was a problem hiding this comment.
is this expensive? It seems like we might do it a lot...
| | WhereBindings{..} <- localBindings | ||
| , let sigSpans = getSrcSpan <$> existedSigNames | ||
| , WhereBinding{..} <- bindings | ||
| , let idsWithoutSig = filter (\x -> getSrcSpan (idName x) `notElem` sigSpans) bindingId |
There was a problem hiding this comment.
this seems a bit odd to me: we're checking that a binding doesn't have a signature by comparing source spans. Why not just compare the ids directly?
There was a problem hiding this comment.
The two names are not exactly equal, one name is for the signature and another is for the binding, but they have the same SrcSpan. I've commented it in the definition of WhereBinidng.
There was a problem hiding this comment.
That seems surprising? I would really have expected that the names of the identifiers would be the same? Or maybe there's a different comparison method you're supposed to use there?
|
I rewrite the test to make it clear. I increasingly find that I know little about type information in Haskell, I'm confused by the concept of |
| f aa bb = (aa, ida bb) | ||
| where | ||
| ida :: b -> b | ||
| ida = id |
There was a problem hiding this comment.
It fails as you may guess...
| f = undefined | ||
| where | ||
| g :: Num a => a -> a -> a | ||
| g a b = a + b |
There was a problem hiding this comment.
Num a lost here, the same method can show constraints for top-level bindings...
There was a problem hiding this comment.
Are you saying that the lens cannot deduce Num a => here if the signature is not present?
I find that ok, constraints belong at the top level, not in local type signatures
There was a problem hiding this comment.
Sorry for ignoring this...
Are you saying that the lens cannot deduce Num a => here if the signature is not present?
Yes.
I find that ok, constraints belong at the top level, not in local type signatures
But it's not correct without constraints.
Yeah, it's very complicated and I don't understand it either. Worth talking to the GHC devs in #ghc, though, they're very helpful. |
|
Some notes from what I observed:
|
July541
force-pushed
the
local-binding-type-lens
branch
from
47f2b3b to
c9daf86
Compare
Close #2943.