[Haskell-cafe] Unwrapping newtypes

[Tony Morris]

I think you might want -XGeneralizedNewtypeDeriving

http://haskell.org/ghc/docs/6.12.2/html/users_guide/deriving.html#id659906

On 08/09/10 22:01, Kevin Jardine wrote:
> I have a generic object that I want to wrap in various newtypes to
> better facilitate type checking.
>
> For example,
>
> newtype Blog = Blog Obj
> newtype Comment = Comment Obj
> newtype User = User Obj
>
> Unlike Obj itself, whose internal structure is hidden in a library
> module, the newtype wrappings are purely to facilitate type checking.
> It is no secret that each is just a wrapper around Obj.
>
> It is obvious how to construct the various wrapper objects. It is not
> so obvious how to extract the Obj they contain in a reasonably generic
> way however. What I want is a getObj function that works on all of
> them.
>
> Of course this could work if someone using the library wrote an
> instance for each wrapper object:
>
> instance GetObject Blog where
>     getObj (Blog obj) = obj
>
> but this is a pain in the neck to write for each newtype.
>
> I discovered that Foldable defines a handy toList function that
> extracts content from generic Foldable structures.
>
> So that I could write:
>
> toObj :: Foldable thing => thing Obj -> Obj
> toObj w = head $ toList w
>
> Slightly kludgy but it works.
>
> Even better, recent versions of GHC will allow you to automatically
> derive Foldable.
>
> Unfortunately,
>
> newtype Blog = Blog Obj deriving Foldable
>
> returns a kind error.
>
> What does work is:
>
> newtype BlogF a = Blog a deriving Foldable
> type Blog = BlogF Obj
>
> After having spent close to a day on this, I am a bit baffled that
> such a seemingly trivial problem seems so hard to do.
>
> I am wondering if I am missing something really, really obvious.
>
> Any suggestions? Or is there perhaps a more Haskelly way to place type
> constraints on a more generic type?
>
> Kevin
> _______________________________________________
> Haskell-Cafe mailing list
> Haskell-Cafe at haskell.org
> http://www.haskell.org/mailman/listinfo/haskell-cafe
>
>   

-- 
Tony Morris
http://tmorris.net/