See Scrap more boilerplate: reflection, zips, and generalised casts by Ralf
Lämmel and Simon Peyton Jones
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.114.6907&rep=rep1&type=pdf
for
why gcast1 and dataCast1 exist as they do.
On Fri, Jul 17, 2015 at 8:58 PM, Ryan Scott
It does seem unnecessary. The Data instance for Proxy is manually written in Data.Data:
proxyConstr :: ConstrproxyConstr = mkConstr proxyDataType "Proxy" [] PrefixproxyDataType :: DataTypeproxyDataType = mkDataType "Data.Proxy.Proxy" [proxyConstr]
instance (Data t) => Data (Proxy t) where gfoldl _ z Proxy = z Proxy toConstr Proxy = proxyConstr gunfold _ z c = case constrIndex c of 1 -> z Proxy _ -> error "Data.Data.gunfold(Proxy)" dataTypeOf _ = proxyDataType dataCast1 f = gcast1 f
If dataCast1 weren't defined to be gcast1, then that instance would only require a Typeable t constraint, which would make it polykinded. To be honest, I'm not sure why gcast1 requires a Data t constraint (I don't use Data that often, so hopefully someone more knowledgeable than me can answer this).
_______________________________________________ Libraries mailing list Libraries@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/libraries