[GHC] #8809: Prettier error messages?

[GHC]

#8809: Prettier error messages?
-------------------------------------+-------------------------------------
        Reporter:  joelteon          |                Owner:
            Type:  feature request   |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  7.9
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:  13122
 Related Tickets:                    |  Differential Rev(s):
  #8809,#10073,#10179,#12906         |
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by Rufflewind):

 TBH, I think `SDoc` is too low level to embed rich semantic information
 like expressions, locations etc.  It should stick to what it does best:
 laying out dumb text data.  Instead, I think it'd be nicer to have a
 simple high-level data type for diagnostics, which could look like:


 {{{#!hs
 data Error = Error Severity (Maybe ErrorID) ErrorMessage [(Location, Maybe
 Expression)]
 type ErrorID = String -- i.e. flag that disables the warning
                       -- it'd be nice to have them for errors too
 type ErrorMessage = [ErrMsgSegment]
 data ErrMsgSegment = EDoc SDoc | ELink Int -- link to some expression of
 interest
 }}}

 … something along those lines. There's plenty of room to bikeshed on the
 details, but it'd be great to keep it as simple as possible so as to
 minimize the effort required by downstream to interpret the diagnostics.
 In contrast, an `SDoc` does not give the consumer a lot of flexibility in
 the display of an error.

 This approach allows for a gradual transition: simply edit the legacy
 error-emitting framework to emit an `Error severity Nothing (EDoc msg)
 [(loc, Nothing)]`, which is the kind of “dumb” error that we have today.
 To take advantage of the new system, we just bypass the legacy framework
 and create `Error`s directly.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8809#comment:52>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler