Edward: Note that x = #x is a perfectly legal definition, and now you can have your lenses exactly as before. When discussing this with Simon, I actually proposed that x = #x be automatically generated by the data definitions, and then nub'd after. Not sure it's a good idea or not, but it's certainly possible. I was also of the opinion that data should produce FieldSelector classes etc, but _not_ link them to IV, specifically to avoid problems with the stab lenses. I expect that if you only wanted stab lenses, and never selectors, you could (probably) tie them up in a way that did the resolution nicely without ambiguity problems. With those two pieces I think you can still have your: foo^.bar.baz.quux Thanks, Neil On Tue, Jan 27, 2015 at 9:19 AM, Adam Gundry <adam@well-typed.com> wrote:
On 27/01/15 09:16, Simon Peyton Jones wrote:
Adam, are you willing to update the wiki page to reflect the latest state of the conversation, identifying remaining choices? That would be v helpful.
I'm on it now. It'll take a little while because I'm merging plans A and B into a single coherent story.
Adam
-- Adam Gundry, Haskell Consultant Well-Typed LLP, http://www.well-typed.com/ _______________________________________________ ghc-devs mailing list ghc-devs@haskell.org http://www.haskell.org/mailman/listinfo/ghc-devs