[Haskell-cafe] Re: Cyclic data declarations

Michal D. michal.dobrogost at gmail.com
Sun Aug 2 21:06:27 EDT 2009

>   newtype StmtRec = StmtRec (Stmt [StmtRec])

That's pretty much were I threw in the towel last night. Except I had
a bunch of places where I had to add the extra constructor statements.
I wish there was a solution that didn't require these... they really
butcher pattern matching clarity.

> will do. More generally, you can use
>   newtype Fix f = In { out :: f (Fix f) }
> and define
>   type StmtRec = Fix ([] `O` Stmt)
> where  O  denotes composition of functors
>   newtype O f g a = O (f (g a))

Thanks for that! This provoked some thought on my part about what
exactly is going on. I think I could solve this if I added some way to
identify that a type parameter is actually referring to the whole
type. Say we had a reserved word "fixpoint" for this. Then we'd have
something like:

data Stmt x = SIf x x

then when we actually go to use it, it would be referred to as the type:

"Stmt [fixpoint]"

Which would get treated exactly like the data declaration:

data Stmt = SIf [Stmt] [Stmt]

I'll need to add the newtype declaration for the code but I'd be
interested if anyone had further thoughts on this topic. I have an
implementation of both approaches on a toy parser, but I doubt
anyone's interested in seeing that.


More information about the Haskell-Cafe mailing list