[GHC] #14917: Allow levity polymorphism in binding position

[GHC]

#14917: Allow levity polymorphism in binding position
-------------------------------------+-------------------------------------
        Reporter:  andrewthad        |                Owner:  (none)
            Type:  feature request   |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.2.2
      Resolution:                    |             Keywords:
                                     |  LevityPolymorphism
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by Ericson2314):

 > It's a bit tiresome to have to do this -- it'd be easier for the
 programmer to say that all polymorphism can be specialised in this way, as
 .NET does -- but I think that's out of reach for us.

 Once we have the monomorphizing quantifiers, I would advocate for a
 language proposal to infer runtime rep parameters and {{{TYPE r}}} where
 where {{{Type}}} is inferred today. (Someday, I want to rewrite the RTS in
 Haskell...without getting RSI from wrist typing boilerplate :D)

 > And there are only a finite number of possiblities for {{{r}}}

 Well, we'd also like to use this to statically prevent dictionary passing,
 and constraints are in generate infinitely inhabited. Also it would be
 nice to unbox aggregates in collections, maybe reusing the unboxed tuple
 and sum representations. That also creates infinite choices for {{{r}}} in
 principle.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14917#comment:20>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler