#2442: Heuristics to improve error messages for badly referenced things
----------------------------+-----------------------------------------------
Reporter: batterseapower | Owner:
Type: proposal | Status: new
Priority: normal | Milestone:
Component: Compiler | Version: 6.9
Severity: normal | Resolution:
Keywords: | Difficulty: Unknown
Testcase: | Architecture: Unknown
Os: Unknown |
----------------------------+-----------------------------------------------
Comment (by batterseapower):
Great!
For posterity, the accompanying thread on Haskell-Cafe is at
http://www.haskell.org/pipermail/haskell-cafe/2008-July/045175.html
That thread also discusses what would be a cool improvement: searching all
installed packages for the unbound name. However, that would require an
actual index and as such is a bit of a heavier-weight solution.
As a practical matter, I've been developing GHC using a stage1 compiler
with this patch compiled into it and there is a noticable delay when
displaying unbound name error messages if there are many things in scope.
This is because I have made no attempt to optimize the code yet, and this
needs to be resolved.
Notes for me, when I get time to look at it:
* Cache global reader environment elements between unbound name error
messages (I'm guessing eltsUniqFM is fairly expensive)
* Revert to the Levenshtein distance
(http://en.wikipedia.org/wiki/Levenshtein_distance) rather than one that
attempts to account for transpositions too, as the associated algorithm is
considerably simpler. Unbox it to the hilt
* Possibly adapt the distance metric so it operates in conjunction with a
threshold distance it should bail out after
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/2442#comment:3>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
