[Haskell-cafe] Polymorphic algebraic type constructors

[Ralf Laemmel]

Hi Graham,

Graham Klyne wrote:

> It's been a while since I looked at the "boilerplate" work, but 
> looking at your code I think it depends on gmapQ of the polymorphic 
> value to be converted.  Does your generic Haskell processor generate 
> this automagically?

One needs Data and Typeable instances which are automagically derived by 
GHC, indeed.
http://www.cs.vu.nl/boilerplate/
(The code I sent relies on GHC CVS as of months ago. GHC 6.2's 
Data/Generics uses some different names in a few spots.)

> Anyway, it reminds me of a private communication I received on this 
> topic, suggesting that the "problem" could be resolved by making the 
> polymorphic container type an instance of Functor, and using fmap to 
> do the conversion.  This ensures that the other constructors only need 
> to be mentioned once (in the fmap instance).

As you probably notice, "Either" would require a *bifunctor*.
Unless you fix the last argument as variation point.
And then, the bmap would take two argument functions.
One would be the id function in your instance.

Ralf