Opened 9 years ago

Closed 9 years ago

Last modified 44 years ago

#430 closed bug (Invalid)

Result type signatures and lexically scoped type variables

Reported by: nobody Owned by: nobody
Priority: normal Milestone:
Component: Compiler (Type checker) Version: None
Keywords: Cc:
Operating System: Architecture:
Type of failure: Difficulty:
Test Case: Blocked By:
Blocking: Related Tickets:

Description

Dear ghc developers,

The ghc documentation (7.4.10.) states that
  "A lexically scoped type variable can be bound by: 
  [...] A result type signature"

However, actually trying to use them causes some
unexpected behavior:

import Data.Typeable

foo :: Typeable b => b
foo :: a = typeOf (undefined :: a) `seq` (undefined :: a)

bar :: forall b. Typeable b => b
bar :: a = typeOf (undefined :: a) `seq` (undefined :: a)
         
baz :: forall a. Typeable a => a
baz :: a = typeOf (undefined :: a) `seq` (undefined :: a)

All three examples give rise to basically the same
error message, namely

    Inferred type is less polymorphic than expected
      Quantified type variable `b' is mentioned in the
environment:
        Scoped type variable `a' = b (bound at:
test135.hs:4:7)
    When trying to generalise the type inferred for `foo'
      Signature type:     forall b. (Typeable b) => b
      Type to generalise: b
    In the type signature for `foo'
    When generalising the type(s) for `foo'

If I understand the documentation correctly, they
should all compile. An especially interesting case is
'baz', where the 'a' from the result type annotation
seems to shadow the a from the type signature (that
doesn't happen with pattern type annotations).

Another curiosity happens if we alpha-rename foo:

qux :: Typeable a => a
qux :: a = typeOf (undefined :: a) `seq` (undefined :: a)

The error message becomes

    All of the type variables in the constraint
`Typeable a' are already in scope
        (at least one must be universally quantified here)
    In the type signature: qux :: (Typeable a) => a

This is just my wild speculation, but does this really
mean that the 'a' in qux's signature is bound by "qux
:: a"?

ghc6.2 gives the same error messages, where of course
'bar' behaves like 'foo' and 'baz' like 'qux'.

Thanks you,

-- Thomas Jäger

Change History (1)

comment:1 Changed 9 years ago by simonpj

  • Status changed from assigned to closed
Logged In: YES 
user_id=50165

This isn't a bug; read Section 7.4.10.5 carefully, esp the stuff 
about 'rev'.  

In all your examples, your type variable is taking as a pattern 
signature, not a result signature, as the manual explains.

This might not be what you expect, but it obeys the rule that 
the scope of the type variable is the same as the scope of the 
variable in the pattern.
Note: See TracTickets for help on using tickets.