[GHC] #6018: Injective type families

[GHC]

#6018: Injective type families
-------------------------------------+-------------------------------------
              Reporter:  lunaris     |             Owner:  jstolarek
                  Type:  feature     |            Status:  new
  request                            |         Milestone:  7.10.1
              Priority:  normal      |           Version:  7.4.1
             Component:  Compiler    |          Keywords:  TypeFamilies,
            Resolution:              |  Injective
Differential Revisions:              |  Operating System:  Unknown/Multiple
          Architecture:              |   Type of failure:  None/Unknown
  Unknown/Multiple                   |         Test Case:
            Difficulty:  Unknown     |          Blocking:
            Blocked By:              |
       Related Tickets:  #4259       |
-------------------------------------+-------------------------------------

Comment (by jstolarek):

 Replying to [comment:40 goldfire]:
 > I'm still not sold on the concrete syntax involving `result`, despite
 having introduced that idea myself. I like the `(**)` suggestion here a
 bit more: http://www.haskell.org/pipermail/ghc-devs/2014-July/005492.html

 So, to be precise, with this syntax the examples from the wiki page would
 look like this:

 {{{
 type family F a b c | F -> a b c
 type family G a b c | G -> a b
 type family H a b c | H a -> b c, H b -> a c, H c -> a b
 type family J a b c | J a -> b c, J b -> a c
 }}}
 I did not propose to use the notation like `H b -> a c` because I thought
 it might mislead people to think that: a) `H` could be partially applied;
 b) some parameters of `H` could be skipped (ie. we can apply it to `b`
 without first applying to `a`). Luckily, this is not a blocking issue for
 the rest of implementation. It should only affect 3 files in the parser,
 so it can be changed later.

 > Separately, a change to Core will be necessary to really make this work
 in all cases.
 >
 > (...)
 >
 > All that said, this use case is obscure, and the primary push for
 injective type families does not seem to require a change to Core. It is
 worth implementing without this piece. But, the idea came up in
 conversation, and I thought I should record it.

 This is beyond my knowledge - I would require some guidance here. But if
 it's possible to implement injective type families without this change
 then I would postpone it until later and possibly consider it a separate
 task.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/6018#comment:41>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler