[GHC] #4159: move Monad and MonadFix instances for Either from mtl to base

[GHC]

#4159: move Monad and MonadFix instances for Either from mtl to base
-------------------------+--------------------------------------------------
Reporter:  ross          |          Owner:                  
    Type:  proposal      |         Status:  closed          
Priority:  normal        |      Component:  libraries/base  
 Version:  6.12.3        |     Resolution:  fixed           
Keywords:                |             Os:  Unknown/Multiple
Testcase:                |   Architecture:  Unknown/Multiple
 Failure:  None/Unknown  |  
-------------------------+--------------------------------------------------
Changes (by ross):

  * status:  new => closed
  * resolution:  => fixed


Comment:

 There was a lengthy discussion.  Beyond the move, the change involves a
 classic trade-off: the Either monad instance becomes usable on arbitrary
 types, but the fail method becomes less useful; in particular, pattern
 match failure in this monad becomes an error.

 More than a dozen people were in favour of the change.  A popular argument
 was that Either is a general purpose type modelling categorical sum, so it
 should have the simple and familiar monad instance.

 There were two arguments against, in different directions:

  * Claus Reinke wished to retain pattern match failure.  We considered
 adding {{{fail s = Left (error s)}}} but this would please no-one: for the
 pattern matching use it would be equivalent to using the Maybe monad
 instead.

  * Iavor Diatchki opposed defining any instance of Monad Either, arguing
 that it promoted bad style.  It was pointed out that because instances in
 Haskell are effectively global, most code already has the instance.

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/4159#comment:1>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler