[Haskell-cafe] generalized newtype deriving allows the definition of otherwise undefinable functions

[Simon Peyton-Jones]

| some time ago, it was pointed out that generalized newtype deriving could be
| used to circumvent module borders. Now, I found out that generalized newtype
| deriving can even be used to define functions that would be impossible to define
| otherwise. To me, this is surprising since I thought that generalized newtype
| deriving was only intended to save the programmer from writing boilerplate
| code, not to extend expressiveness.

Yes indeed.  See http://hackage.haskell.org/trac/ghc/ticket/1496 for why this is really a bug in general. 

The trouble described there really happens when 'item' (in your iso class) in instantiated to a data type with a constructor whose fields use type functions.

Stephanie Weirich, Steve Zdancewic, Dimitrios Vytiniotis and I have been working hard on a development of the FC intermediate language, and hence of the source language, that will close this (embarrassing) loophole, and allow some new expressiveness.  Nothing written down in a form that someone other than us can make sense of, but there will be!  In brief, though, we're going to end up with kinds looking like
	* => *
as well as the existing
	* -> *
The new form means "a type-indexed function" whereas the latter means "a type-parametric function". 

John Meacham's example is also very interesting. Even if the data type doesn't use type functions, it might have invariants concerning type classes (his example is Set), and converting all the elements might destroy the invariants.  Excellent point!  There's no type-soundness issue (no run-time seg fault) but something nearly as bad.  Will have to think about that.  Probably declaring Set to have kind (* => *) will do the job.

Thanks for the thread.

Simon