[GHC] #14684: combineIdenticalAlts is only partially implemented

[GHC]

#14684: combineIdenticalAlts is only partially implemented
-------------------------------------+-------------------------------------
        Reporter:  mpickering        |                Owner:  sjakobi
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:  8.6.1
       Component:  Compiler          |              Version:  8.2.2
      Resolution:                    |             Keywords:  newcomer
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):  Phab:D4542
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by sjakobi):

 Replying to [comment:13 simonpj]:

 Expanding the `DEFAULT` case before recombining the alts is an interesting
 idea. I'm wondering though if we'd recreate alts that may have been
 excluded (by the programmer or optimization steps) based on information
 that isn't available in `CSE`. At worst we'd still just recreate the same
 `DEFAULT` alt, so I believe no pessimisation is possible.

 The design I currently have in mind involves largely imitating
 `SimplUtils.prepareAlts` to

   1. detect which constructors are impossible and should not be considered
 when expanding the `DEFAULT`.
   2. use a variant of `refineDefaultAlt` to expand the `DEFAULT` alt into
 alts using the remaining possible constructors.

 Does this sound right?

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14684#comment:14>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler