[GHC] #11451: Inconsistent warnings for unused binders

[GHC]

#11451: Inconsistent warnings for unused binders
-------------------------------------+-------------------------------------
        Reporter:  simonpj           |                Owner:
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  7.10.3
      Resolution:                    |             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:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 All that you say is true about type function too.  In
 {{{
 type instance F a a = Int
 }}}
 the `a` is definitely not unused becuase it is repeated. In both your
 examples the type variables patently are used, and no one is suggesting
 they are reported as unused:
 {{{
 instance C a a where ...
 instance D a b => C (a,b) where...
 }}}


 But it really does seem useful an convenient to be able to write
 {{{
 type instance F (Maybe _) = Int
 }}}
 to stress that the argument to `Maybe` plays no further role in matching
 or on the RHS. Isn't it?

 Similarly would it not be reasonable to allow
 {{{
 instance C (Maybe _) where ...
 }}}
 to stress that the argument type of the `Maybe` plays no further role in
 resolving this class instance.

 So let me be more concrete.  Here's a proposal:

 * If a type variable (a) is bound in a type pattern (b) appears at exactly
 once anywhere in its scope, then it is reported as unused.

 * A "type pattern" is
   * a type on the LHS of a `type instance` or
   * a type in the head of a class `instance` (i.e. the bit after the `=>`)
   * a type in a pattern type signature e.g. `case x of Just (x :: (a,b))
 -> blah`

 Now, can you give me an example where that would be annoying?  Mabye you
 are right, but let's see.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11451#comment:11>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler