Changes between Version 5 and Version 6 of AssociatedTypes


Ignore:
Timestamp:
Apr 30, 2006 11:19:55 PM (8 years ago)
Author:
ross@…
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • AssociatedTypes

    v5 v6  
    1919== Cons == 
    2020 * Only a prototype implementation so far.  We are in the process of implementing them in GHC, but that is a significant undertaking as they have to play nicely along with other extensions (e.g., GADTs) and we are after a more lightweight type-preserving translation than the one in the POPL paper.  So, given requirement that all Haskell' features need to have been implemented in at least on Haskell system, it seems unlikely that we can meet the currently proposed Haskell' timeline. 
    21  * [http://www.haskell.org/pipermail/haskell-prime/2006-April/001466.html Non-termination problems with FDs also arise with ATs]. 
     21 * The inference algorithm [http://www.haskell.org/pipermail/haskell-prime/2006-April/001502.html may accept types with unsatisfiable constraints], deferring errors until the function is used. 
    2222 * The published version of ATs does not permit constraints on associated type synonyms in superclass contexts, which are [http://www.haskell.org/pipermail/haskell-cafe/2006-February/014392.html required by some monad transformer classes]. 
    23    Allowing such constraints may lead to [http://www.haskell.org//pipermail/haskell-cafe/2006-February/014609.html further non-termination problems]. 
     23   In conjunction with other proposed extensions, allowing such constraints may lead to [http://www.haskell.org//pipermail/haskell-cafe/2006-February/014609.html further non-termination problems].