#1241: Functional dependencies not checked.
-------------------------------------+--------------------------------------
Reporter: guest | Owner:
Type: bug | Status: new
Priority: normal | Milestone: _|_
Component: Compiler (Type checker) | Version: 6.6
Severity: normal | Resolution:
Keywords: | Difficulty: Unknown
Testcase: | Architecture: Unknown
Os: Unknown |
-------------------------------------+--------------------------------------
Comment (by simonpj):
Thanks Martin. I'd like your advice. Here is GHC's spec for instance
declarations
[http://www.haskell.org/ghc/dist/current/docs/users_guide/other-type-
extensions.html#instance-decls]
Notice (1) the Paterson Conditions (PCs), (2) the Coverage Condition (CC)
We agree that these conditions are conservative. At the moment they are
both lifted by adding -fallow-undecidable-instances. But I gather you
think that maybe the CC should *not* be lifted, is that right.
So are you saying this?
* The CC should always hold
* The PCs can be lifted; the only problem is that you may get non-
termination
The trouble is, as you know, that sometimes the CC can validly not hold.
Example:
{{{
class C a b | a->b
class D a b | a->b
instance D a b => C [a] [b]
}}}
Here the CC does not hold, but the context `(D a b)` enforces the
dependency. We call this the Weak Coverage condition in the paper.
So you could be saying: you can only lift the CC if you replace it with a
weaker, but still sufficiently strong alternative.
I'm also not sure whether lifting it altogether (as GHC now does) is
simply confusing, or whether it can be unsound or something really bad
like that.
Simon
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/1241>
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
