For the "minor point", you'd like for example the following equivalence (modulo decomposing the Attr type)? toAttrs (FooD 7,BarD True,BazD 'X') == (Attr FooT Int,Attr BarT Bool,Attr BazT Char) If I understood correctly, take a look at the most recent edit of http://lpaste.net/97165 The code gets hairier, but you can now say
example = let (a,_,_) = toAttrs (FooD 7,BarD True,BazD 'X') in a
If you do not specify the "container type" of the result (the pattern
suffices in my example), there will be a type-error. Loosening that
requirement would require a mapping that includes
f a b goes to f Attr Attr where for example f = (,)
f a b c goes to f Attr Attr Attr where for example f = (,,)
f a b c d goes to f Attr Attr Attr Attr where for example f = (,,,)
and so on
But I don't know how to specify such a general one, since all (instances
of) tuple types are distinct types that are not formally related to one
another within the GHC type system.
So:
1) Have I understood your specification correctly?
2) Other experts: is there safe trickery to generally specify this map? I
think it might actually be doable in 7.4, but as far as I know, that was
accidental generally-unsound functionality that was removed in 7.6+.
HTH also.
On Fri, Dec 20, 2013 at 1:44 AM, AntC
Nicolas Frisby
writes: Disclaimer: I didn't read all of the previous emails. Does this do what you need?
I'm on GHC 7.4 at the moment, but I think this will work with later versions.
HTH.
Thank you Nicolas (and yes it works on GHC 7.6).
It does look 'cleaner' than the version I posted. But I still see plenty of pileups of (type and data) constructors.
However, perhaps it looks 'cleaner' because it doesn't work in the corner cases you could have seen discussed in the previous emails?
* There's no instance for the unit tuple `toAttrs ()` -- that's easily fixed:
instance ToAttrsR U1 where toAttrsR _ = []
* It crashes '*** Exception: Prelude.undefined' when toAttrs' argument is undefined. I tried a quick fix to the (:*:) instance, similar to my version, but that wasn't enough.
* (And a minor point: I wanted to be able to go feed toAttrs' output tuple into itself. -- that's fixable easily enough. My version for a bonus also grabbed the newtype's data constructor. -- that's also fixable, I think.)
AntC
_______________________________________________ Generics mailing list Generics@haskell.org http://www.haskell.org/mailman/listinfo/generics