On Wed, Nov 10, 2010 at 11:59:28AM +0200, John Smith wrote:
http://hackage.haskell.org/trac/haskell-prime/wiki/TypeDirectedNameResolutio...
The problem with this is that it conflates two orthogonal features: type-directed name resolution proper (also known as ad hoc overloading), and a fancy postfix application syntax. There is no connection between these except that both are useful for accessing records in a particular way. So the proposal seems to be tailored specifically to fix some inconveniences with records. I'd much rather see a true record system for Haskell, since that would fix the namespace conflict problem in a more robust way. Plain ad hoc overloading might or might not be a sensible addition to Haskell, but please at least drop the "x .f" syntax, it's a pointless hack that makes the lexical status of "." even more difficult than it currently is. After all, one can simply define e.g. "x .$ f = f x" if postfix application is needed. Cheers, Lauri