[GHC] #9948: Recommend class constraint instead of instance constraint

[GHC]

#9948: Recommend class constraint instead of instance constraint
-------------------------------------+-------------------------------------
        Reporter:  crockeea          |                   Owner:
            Type:  feature request   |                  Status:  new
        Priority:  normal            |               Milestone:
       Component:  Compiler          |                 Version:  7.8.4
      Resolution:                    |                Keywords:
Operating System:  Unknown/Multiple  |            Architecture:
 Type of failure:  None/Unknown      |  Unknown/Multiple
      Blocked By:                    |               Test Case:
 Related Tickets:                    |                Blocking:
                                     |  Differential Revisions:
-------------------------------------+-------------------------------------

Comment (by goldfire):

 Replying to [comment:4 crockeea]:
 > I'm suggesting that if `foo` is a method of class `Foo`, then GHC simply
 says `cannot deduce (Foo a)`. There's no need to match instances at all.

 However, we would, of course, still like to find an instance where
 possible. Just determining if an instance exists requires looking up
 instances for `Foo`, and then perhaps recurring on any constraints to
 those instances. At the end of a chain of instance constraints, we might
 end up with some constraint, say `Bar Int a`, that is unsatisfiable. Only
 then do we know that no instance matches the original constraint. To
 report a missing `Foo a` constraint at this point, we would need to
 somehow the need for `Bar Int a` came from a desire for `Foo a`. Keeping
 track of this is the "more work" I was talking about. Nowhere near
 impossible, but strictly harder than what happens currently.

 In any case, I think the discussion about how hard this feature is to
 implement detracts from the debate we should be having: what is really the
 desired behavior, regardless of implementation challenge. (To be clear,
 the implementation is not particularly challenging here, I think -- just
 not completely trivial.) And, regardless of implementation issues, I
 prefer the current behavior over the new one proposed in this ticket.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9948#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler