[GHC] #11188: Confusing "parse error in pattern" for spurious indentation.

[GHC]

#11188: Confusing "parse error in pattern" for spurious indentation.
-------------------------------------+-------------------------------------
        Reporter:  andreas.abel      |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  high              |            Milestone:
       Component:  Compiler          |              Version:  7.10.1
  (Parser)                           |
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Incorrect         |  Unknown/Multiple
  warning at compile-time            |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):  Phab:D4497
       Wiki Page:                    |
-------------------------------------+-------------------------------------
Changes (by osa1):

 * differential:   => Phab:D4497


Comment:

 I think the problem here is not that it's suggesting adding more `do`s
 (it's saying "possibly", meaning the reason may be something else), but
 it's showing only a small part of the large expression that the parser
 considers as a pattern. I submitted Phab:D4497 to improve this. Previously
 the error message was:

 {{{
 Main.hs:2:3: error:
     Parse error in pattern: putStrLn
     Possibly caused by a missing 'do'?
   |
 2 |   putStrLn "" $ do
   |   ^^^^^^^^^^^
 }}}

 this is not good enough because it doesn't show the whole part that's
 considered as a pattern. With Phab:D4497:

 {{{
 Main.hs:2:3: error:
     Parse error in pattern: putStrLn "" $ do a <- return 3 c
     Possibly caused by a missing 'do'?
     putStrLn "" $ do a <- return 3 c
   |
 2 |   putStrLn "" $ do
   |   ^^^^^^^^^^^^^^^^...
 }}}

 It's now clear that left hand side of second `<-` now covers the first
 `<-` because the whole part is shown.

 If the error message is still not good enough perhaps we can split the
 parsers for expressions and patterns (rather than using expression parser
 for patterns and then doing a check to see whether it's a valid pattern).
 In that case I think we'd still get a parse error in expression parser
 while parsing `return 3 c <- do ...` which would raise a better error
 message I think.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11188#comment:4>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler