Support eliminating a variant with a record of functions with non-fixed codomain

[mrkgnao]

I'm not sure if there's something fundamental preventing the inclusion of a function of type

vmap 
  :: Short (NumCols p - 1) 
  => R (f :->: g) p -> V f p -> V g p

in sculls.

One might then use a record of functions like

funcs :: R (I :->: I) (Row0 .& "int" .= Int .& "string" .= String)
funcs = mkR 
  .* (A (\(I x) -> I (2 * x)) :: (I :->: I) "int" Int)
  .* (A (\(I x) -> I (reverse x)) :: (I :->: I) "string" String)

to eliminate a variant like V I (Row0 .& "int" .= Int .& "string" .= String).

[nfrisby]

That's a very unhelpful error message, isn't it? Though I suspect the fundamental problem here is that the `.=` operator only supports the `I` field structure, not others such as `I :->: I`. ``` (.=) :: L l -> t -> I l t (.=) = const I ``` Your example needs an operator of type `L l -> (t -> t) -> (I :->: I) l t` instead. `const coerce` implements that signature. I'll see if the error message can be improved. Thanks again!

…

On Sat, Oct 21, 2017, 23:47 Soham Chowdhury ***@***.***> wrote: Consider the following type/term. funcs :: R (I :->: I) (Row0 .& "int" .= Int .& "string" .= String) funcs = mkR .* #int .= (\x -> 2 * x) .* #string .= reverse This fails: • Couldn't match type ‘(Row0 .& (l0 .= ())) .& (l1 .= ())’ with ‘(Row0 .& ("int" .= Int)) .& ("string" .= String)’ Expected type: R (I :->: I) ((Row0 .& ("int" .= Int)) .& ("string" .= String)) Actual type: R (I :->: I) ((Row0 .& (l0 .= ())) .& (l1 .= ())) NB: ‘.&’ is a type function, and may not be injective The type variables ‘l0’, ‘l1’ are ambiguous • In the expression: mkR .* #int .= (\ x -> 2 * x) .* #string .= reverse In an equation for ‘funcs’: funcs = mkR .* #int .= (\ x -> 2 * x) .* #string .= reverse | 82 | funcs = mkR .* #int .= (\x -> 2 * x) .* #string .= reverse | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#12>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABDg00RzxlSrYVPgD70atIGdiF8BceCBks5suuUAgaJpZM4QB1zz> .

[mrkgnao]

@nfrisby: apologies for the older version of the issue report. I was actually using (.=) from lens, which had magically polymorph-ed itself into a confusingly sensible-looking type. :)

I've updated the issue with what I was really trying to do.