Inferred type of print using function is surprising

[ocharles]

I wanted to play around with the new functionality to insert type signatures. The following:

hello = print

surprises me, as when I add a type signature with a code action, I get:

hello :: GHC.Types.Any -> IO ()
hello = print

which is mostly certainly not what I'd expect!

[ndmitchell]

If you add an argument a to both sides, does it do better? I wonder if the monomorphism restriction is involved? I know GHC replaces completely polymorphic variables with Any at some step in the process, I'm surprised its happened by the time we get to it.

[ocharles]

Ah yea, good idea - it does indeed work if I write

hello a = print a

[ndmitchell]

Note that I don't get this bug. If I write:

hello = print

Then I get the warning:

    • Ambiguous type variable ‘a0’ arising from a use of ‘print’
      prevents the constraint ‘(Show a0)’ from being solved.
      Relevant bindings include
        hello :: a0 -> IO () (bound at C:\Neil\temp\IDE.hs:3:1)
      Probable fix: use a type annotation to specify what ‘a0’ should be.
      These potential instances exist:
        instance Show Ordering -- Defined in ‘GHC.Show’
        instance Show Integer -- Defined in ‘GHC.Show’
        instance Show a => Show (Maybe a) -- Defined in ‘GHC.Show’
        ...plus 22 others
        ...plus 12 instances involving out-of-scope types
        (use -fprint-potential-instances to see them all)
    • In the expression: print
      In an equation for ‘hello’: hello = print

And that warning prevents it from detecting an incorrect type signature. If I hover over the print it suggests Any -> IO () as the type, but that's not unreasonable as the thing has failed to type check, so the fact it displays anything is moderately surprising.

What did you do to manage to persuade it to get as far as it did? Were there any errors present?

[ocharles]

There were errors, but my editor (Emacs) happily showed me a code action on hello to add a type signature. No extra convincing required

…

On Wed, 11 Dec 2019, 8:45 pm Neil Mitchell, ***@***.***> wrote: Note that I don't get this bug. If I write: hello = print Then I get the warning: • Ambiguous type variable ‘a0’ arising from a use of ‘print’ prevents the constraint ‘(Show a0)’ from being solved. Relevant bindings include hello :: a0 -> IO () (bound at C:\Neil\temp\IDE.hs:3:1) Probable fix: use a type annotation to specify what ‘a0’ should be. These potential instances exist: instance Show Ordering -- Defined in ‘GHC.Show’ instance Show Integer -- Defined in ‘GHC.Show’ instance Show a => Show (Maybe a) -- Defined in ‘GHC.Show’ ...plus 22 others ...plus 12 instances involving out-of-scope types (use -fprint-potential-instances to see them all) • In the expression: print In an equation for ‘hello’: hello = print And that warning prevents it from detecting an incorrect type signature. If I hover over the print it suggests Any -> IO () as the type, but that's not unreasonable as the thing has failed to type check, so the fact it displays anything is moderately surprising. What did you do to manage to persuade it to get as far as it did? Were there any errors present? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <https://github.com/digital-asset/ghcide/issues/228?email_source=notifications&email_token=AAAFDDQMVGDE7GLLYOQGQCLQYFGM7A5CNFSM4JZBRRZKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEGUQJYY#issuecomment-564724963>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAFDDWIAAXDWGNEKBCUZJTQYFGM7ANCNFSM4JZBRRZA> .

[ndmitchell]

I was on GHC 8.6.5, what were you on? My understanding from the diagnostics I saw was that GHC wasn't giving the top-signature warning.

[ocharles]

This would have been GHC 8.8.1. Anything I can do to provide more info?

…

On Wed, 11 Dec 2019, 10:01 pm Neil Mitchell, ***@***.***> wrote: I was on GHC 8.6.5, what were you on? My understanding from the diagnostics I saw was that GHC wasn't giving the top-signature warning. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <https://github.com/digital-asset/ghcide/issues/228?email_source=notifications&email_token=AAAFDDXUD5UFUW4IXQP7H53QYFPL3A5CNFSM4JZBRRZKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEGUXFTA#issuecomment-564753100>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAFDDRJ55G7YHXTATCRRETQYFPL3ANCNFSM4JZBRRZA> .

[ndmitchell]

Most likely GHC 8.8 manages to figure out a top-level signature warning even in the presence of warnings. I wonder if we should explicitly omit suggesting signatures with Any in them, as its almost never what anyone wants.

[jneira]

Reproduced with hls-1.5.0 and ghc-8.10.7

[hasufell]

I wonder if we should explicitly omit suggesting signatures with Any in them, as its almost never what anyone wants.

I second this.

[develop7]

Reproduced with GHC 9.8.2 & HLS 2.8.0.0