On Thu, 2008-01-17 at 09:41 -0500, Isaac Dupree wrote:
Therefore I propose that, though it may break code in the short term, Happy and Alex should only import things qualified as AlexSomething. It didn't make things too ugly; they still rely on sufficient basic Prelude stuff being imported without qualification by the module that uses Happy/Alex, but that seems the best arrangement to me anyway. (for example, if a module wants to redefine (&&) in a compatible way -- this has been done with type-classes, Alex will at least try to use that instead of importing (&&) unqualified from the Prelude and getting an ambiguity error)
This patch does that for Alex, and I've been using it to compile GHC; a similar patch would be a similar amount of work to make for Happy.
Thoughts? Does this seem like a good idea?
Sounds sensible to me. I've had code break by changing -agc options to happy because it started importing Haskell98 modules and my package did not depend on the haskell98 package. Duncan