Hello,

I got a question about what the form of types in Haskell are allowed. Specifically, what are the requirements of the context of a type? The paper

    Typing Haskell in Haskell

required the type to be in the form P => t, where P to be in head-normal form or in other words shouldn't be something like Num Bool.

Now what about multi-parameter classes? The paper

    Type classes: an exploration of the design space

listed several potential choices but didn't say exactly what.

The Haskell 2010 report covered single-parameter type classes only.

Given the following definition.

class Collects ce e | ce -> e where
    empty :: ce
    insert :: e -> ce -> ce
   
insert2 c = insert True (insert False c)

The function insert2 has the following type

insert2 :: Collects ce Bool => ce -> ce

This shows that the definition of head-normal forms for multi-parameter class differs from that for single-parameter classes. Is there any definite specification about the context of Haskell types?

Thank you,
Sheng