[Haskell-cafe] Gluing pipes

[Matt Hellige]

On Thu, Dec 4, 2008 at 12:52 PM, Dan Piponi <dpiponi at gmail.com> wrote:
> On Thu, Dec 4, 2008 at 10:21 AM, Matt Hellige <matt at immute.net> wrote:
>>  \ f x y z -> f x z y == id ~> flip
>> It's not clear to me whether your operad class can express this (or
>> whether operads in general can express this)
>
> There exists an operad that can (at the cost of even more notation),
> but you're right that the specific operad that I implemented can't.

Makes sense. It turns out that with one more definition, we can
express this fairly easily in my approach:
  infixr 2 ~*>
  (~*>) = (.)
Now we can use functions that inspect multiple arguments, without
moving our "position" in the argument list:
  f w x y z = concat $ map show [w,x,y,z]
  ex1 = (f $. id ~> flip ~*> id ~> id ~> length ~> id) 1 2 3 [4,5,6,7]
  roll f x y z = f y z x
  ex2 = (f $. id ~> roll ~*> id ~> id ~> length ~> id) 1 [4,5,6,7] 3 2
However, I have grave doubts about the usefulness of this, and I think
it more than sacrifices whatever we gained in clarity. I maintain that
($.) and (~>) are clear and useful, but this seems to be one step too
far...

> Actually, if you look at the papers, mathematicians do have a
> perfectly good notation for this, and it'd be cool if there were a
> programming language that supported it well: drawing diagrams!

Hear, hear!

> Johannes Waldmann said:
>
>> Well, there is Combinatory Logic.
>
> But it's not known for its ease of use :-)

And we'd also have to change the definitions of all of our functions.
Ideally, we'd like the same definitions to work either way.

Anyway, thanks for all your thoughts...

Matt