[Feature request] Extend Under (&.)

[jip]

The following is a proposal to extend Under (&.) to allow various (any, I hope) possible combinations.

Syntax

  u&.(vi`v`ul`ur) y  <-->    {{vi v^:_1 ul u ur}} y
x u&.(vi`v`ul`ur) y  <-->  x {{vi v^:_1 ul u ur}} y

where

• u - verb to process transformed argument(s)
• v - verb with assigned obverse to do/undo arguments transformation
• vi - verb to produce a left argument for v^:_1, is [: if v^:_1 is monadic
• ul - verb to produce a left argument for u, is [: if u is monadic
• ur - verb to produce a right argument for u

Model

und=: 2 : 'n@.0 n@.1^:_1 n@.2 u n@.3'

Examples

Implement standard monadic &.:

   %: <: *: 4              NB. sqrt(4^2 - 1)
3.87298
   <:&.*: 4
3.87298
   <: und ([:`*:`[:`*:) 4  NB. the resulting train is ([: *:^:_1 [: <: *:)
3.87298

Implement standard dyadic &.:

   %: (*: 3) + (*: 4)               NB. sqrt(3^2 + 4^2)
5
   3 +&.*: 4
5
   3 + und ([:`*:`(*:@[)`(*:@])) 4  NB. the resulting train is ([: *:^:_1 *:@[ + *:@])
5

Implement standard dyadic &. with semiduals:

   %: 3 + (*: 4)               NB. sqrt(3 + 4^2)
4.3589
   3 +&.(a:`*:) 4
4.3589
   3 + und ([:`*:`[`(*:@])) 4  NB. the resulting train is ([: *:^:_1 [ + *:@])
4.3589

   %: (*: 3) + 4               NB. sqrt(3^2 + 4)
3.60555
   3 +&.(*:`a:) 4
3.60555
   3 + und ([:`*:`(*:@[)`]) 4  NB. the resulting train is ([: *:^:_1 *:@[ + ])
3.60555

Implement non-standard monadic &. with dyadic inversion (say, for some J dictionary jdict):

   upd=: get__jdict :. (put__jdict~)
   NB. a monad to double values corresponding to keys y
   +: und (]`upd`[:`upd)  NB. the resulting train is (] upd^:_1 [: +: upd)

Implement non-standard dyadic &. with dyadic inversion:

   NB. a dyad which passes its left argument x to the left argument of v^:_1
   u und ([`v`ul`ur)  NB. the resulting train is ([ u^:_1 ul u ur)

Implement an extension proposed in #212:

   u und ([:`(m&{)`(m&{@[)`(m&{@]))  NB. the resulting train is ([: m&{^:_1 m&{@[ + m&{@])

Use extrinsic ul and ur not related to v at all:

   u und (vi`v`f`g)  NB. the resulting train is (vi v^:_1 f u g)

Notes

Whether the u, v, v^:_1 verbs need to be ambivalent depends on the train to compose.

[HenryHRich]

Your model is so simple, why not leave it at that? Or, since half your model is @.n, why not just write the verbs directly?

u&.v greatly aids understanding. You could write v^:_1 u v just as easily, but the dual has a meaning of its own. I don't see that with your longer form.

I vote against.