Re: type specs not making it in to functions
On 25-Jan-2002, Hal Daume III [EMAIL PROTECTED] wrote: consider the following definition: class C a where foo :: a - Int instance C Bool where foo _ = 5 I can then say: bar :: C a = a - Int bar (x :: a) = foo (undefined :: a) But not: bar :: C a = a - Int bar x = foo (undefined :: a) because it tries to use a new scope for the type variable a and doesn't unify it with the one in the type spec. Why not? I don't know why it isn't the case in Haskell. In Mercury we do allow type variables in function type declarations to scope over explicit type qualifications in clause bodies. One drawback with allowing this is that the type declaration may be separated from the clause by an arbitrary amount of text, e.g. bar :: C a = a - Int -- type var `a' introduced here... foo :: Int quux :: String foo = 42 quux = hello world bar x = foo (undefined :: a)-- ... and used here. which means that this feature can be used in ways that have the potential to make programs harder to read. -- Fergus Henderson [EMAIL PROTECTED] | I have always known that the pursuit The University of Melbourne | of excellence is a lethal habit WWW: http://www.cs.mu.oz.au/~fjh | -- the last words of T. S. Garp. ___ Haskell mailing list [EMAIL PROTECTED] http://www.haskell.org/mailman/listinfo/haskell
RE: type specs not making it in to functions
| I can then say: | | bar :: C a = a - Int | bar (x :: a) = foo (undefined :: a) | | But not: | | bar :: C a = a - Int | bar x = foo (undefined :: a) | | because it tries to use a new scope for the type variable a and | doesn't unify it with the one in the type spec. Why not? | | I don't know why it isn't the case in Haskell. | In Mercury we do allow type variables in function type | declarations to scope over explicit type qualifications in | clause bodies. It's a deliberate design choice, but certainly only one of taste. It seems a bit odd for a type signature (perhaps separated by a long way from the definition) should bind a type variable that scopes over the function body. Mark Shields and I are writing a paper about the scoped-type-variable design in GHC; we'll discuss this point. Simon ___ Haskell mailing list [EMAIL PROTECTED] http://www.haskell.org/mailman/listinfo/haskell
Re: type specs not making it in to functions
At 2002-01-25 14:00, Hal Daume III wrote: class D a where constMember :: Int instance D Int where constMember = 8 It seems ehre that there's no way to extract constMember for a /particular/ class, since you can't tell it what a is supposed to be. So, instead, I do: class D a where constMember :: a - Int -- should be independent of arg instance D Int where constMember _ = 8 (which brings me to my use of undefined::a). But is there a better/preferred way to do this? Unfortunately this seems to be the preferred way to do this. For instance, from the Prelude: class (RealFrac a, Floating a) = RealFloat a where floatRadix :: a - Integer floatDigits :: a - Int It's not the _best_ way to do it, in my opinion. I do this: data Type a = Type class D a where constMember :: Type a - Int instance D Int where constMember Type = 8 intConst = constMember (Type :: Type Int) See http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/*checkout*/jvm-bridge/sourc e/Haskell/Type.hs?rev=HEADcontent-type=text/plain. Why is this better? 1. It's absolutely clear to the user of the class that constMember depends only on the type. 2. It prevents the user from putting in an inappropriate back channel in the member for some instance of the class. 3. It avoids use of 'undefined', which is just plain ugly. After all, intuitively everything is defined. -- Ashley Yakeley, Seattle WA Almost empty page: http://semantic.org/ ___ Haskell mailing list [EMAIL PROTECTED] http://www.haskell.org/mailman/listinfo/haskell