[GHC] #14782: typeclass polymorphism defeats bang patterns

[GHC]

#14782: typeclass polymorphism defeats bang patterns
-------------------------------------+-------------------------------------
        Reporter:  dmwit             |                Owner:  (none)
            Type:  bug               |               Status:  closed
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.2.2
      Resolution:  invalid           |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------
Changes (by simonpj):

 * status:  new => closed
 * resolution:   => invalid


Comment:

 It is a bit confusing.

 See [http://downloads.haskell.org/~ghc/master/users-
 guide/glasgow_exts.html#recursive-and-polymorphic-let-bindings the manual
 entry for semantics of let bindings]
 {{{
     let n :: Num a => a; !n = undefined in return ()

 means

     let n :: Num a => a
         n = undefined
     in n `seq` return ()

 which, after elaboration of dictionaries etc means

      let n = /\a. \(d::Num a). undefined
      in n `seq` return ()
 }}}
 So the `seq` sees a lambda and does nothing.

 ---------
 As to the "conflicts with the MR" message, consider
 {{{
   f :: forall a. Num a => a -> a
   Just f = e
 }}}
 and consult
 [https://www.haskell.org/onlinereport/haskell2010/haskellch4.html#x10-880004.5
 the Haskell report on the monomorphism restriction].  It says that the MR
 applies unless "an explicit type signature is given for every variable in
 the group that is bound by simple pattern binding"; and "Recall that ... a
 simple pattern binding is a pattern binding in which the pattern consists
 of only a single variable".  So here `f` is not bound by a simple pattern
 binding and hence falls under the MR, signature or no signature.

 It's the same for your pattern binding `x at n`.  You may say "but it's still
 simple" but that's not what the rules currently say.

 Incidentally, there's a bad typo in the Haskell 98 and Haskell 2010
 reports.  Under
 [https://www.haskell.org/onlinereport/haskell2010/haskellch4.html#x10-860004.4.3.2
 pattern bindings] both say "A simple pattern binding has form p = e"; but
 of course ''all'' pattern bindings have that form.  It should say "A
 simple pattern binding has form x = e, where x is a variable".

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14782#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler