Re: [GHC] #14112: bang patterns on pattern synonyms? (left vs right hand sides)

15 Aug 2017

      #14112: bang patterns on pattern synonyms? (left vs right hand sides)
-------------------------------------+-------------------------------------
        Reporter:  carter            |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.2.1
      Resolution:                    |             Keywords:
                                     |  PatternSynonyms
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 carter):

 well, my first question is: how do i validate that this bang pattern?

 i've looked at the core this generates, and it looks like im only getting
 strictness when i'm matching on the `Just`
 and not when i'm using it as a constructor!

 or maybe i'm not reading the Just constructor code correclty
 {{{
 -- RHS size: {terms: 5, types: 9, coercions: 0, joins: 0/0}
 $bJust1_r1wL
   :: forall a. a -> Data.Unboxed.Maybe.R:MaybeLiftedRepa a
 [GblId, Arity=1, Caf=NoCafRefs]
 $bJust1_r1wL
   = \ (@ a_a1zx) (a1_a1xs :: a_a1zx) ->
       break<5>(a1_a1xs)
       Data.Unboxed.Maybe.MaybeSum
         @ a_a1zx
         (GHC.Prim.(#_|#)
            @ 'LiftedRep @ ('TupleRep '[]) @ a_a1zx @ (# #) a1_a1xs)
 }}}

 plus the casing encoding
 {{{
 -- RHS size: {terms: 17, types: 31, coercions: 2, joins: 0/0}
 Data.Unboxed.Maybe.$mJust
   :: forall (r :: TYPE rep) a.
      Maybe a -> (a -> r) -> (Void# -> r) -> r
 [GblId, Arity=3]
 Data.Unboxed.Maybe.$mJust
   = \ (@ (rep_a1zm :: RuntimeRep))
       (@ (r_a1zn :: TYPE rep_a1zm))
       (@ a_a1xr)
       (scrut_a1zo :: Maybe a_a1xr)
       (cont_a1zp :: a_a1xr -> r_a1zn)
       _ [Occ=Dead] ->
       break<2>(scrut_a1zo,cont_a1zp)
       case scrut_a1zo
            `cast` (Data.Unboxed.Maybe.D:R:MaybeLiftedRepa0[0] <a_a1xr>_N
                    :: (Maybe a_a1xr :: *)
                       ~R#
                       (Data.Unboxed.Maybe.R:MaybeLiftedRepa a_a1xr :: *))
       of
       { MaybeSum ds_d1zX ->
       case ds_d1zX of {
         (#_|#) a1_X1y7 -> cont_a1zp a1_X1y7;
         (#|_#) ipv_s1A5 ->
           Control.Exception.Base.patError
             @ rep_a1zm @ r_a1zn "src/Data/Unboxed/Maybe.hs:37:19-37|case"#
       }
       }
 }}}

 latter seems like the "usual" eliminator form, as far as I can tell. I am
 confused about the exception throwing element of pattern synonyms, is that
 generally case?

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14112#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler