Pipe operator support

[ayshiff]

Closes #72

This PR adds pipe operator support (|>). It is in Draft because we have to wait for #44 to be resolved first. (As mentionned here)

[leostera]

Another thought I had is that we may want to just rewrite the pipes into SSA so we minimize the amount of fun allocations too. So we'd go from this:

let f x = x |> add 1 |> div 2 |> mult 3

To this:

f(X) ->
  Caramel@Tmp1 = add(1, X),
  Caramel@Tmp2 = div(2, Caramel@Tmp1),
  Caramel@Tmp3 = mult(3, Caramel@Tmp2).

This rewrite would be an optimization during translation that happens the moment we find a |> in there. We may need to just leave a mark and have a second compilation pass do this optimization. We'd still need the external to be able to type-check correctly, but we won't actually end up creating a caramel_runtime:pipe/2.

It also means we can take our time to work out the quirks of partial application in the presence of optional and named arguments as pointed out here: https://github.com/AbstractMachinesLab/caramel/pull/92/files#r605967813

@ayshiff would you be interested in trying this approach? 🙌🏽

[ayshiff]

Another thought I had is that we may want to just rewrite the pipes into SSA so we minimize the amount of fun allocations too. So we'd go from this:

let f x = x |> add 1 |> div 2 |> mult 3

To this:

f(X) ->
  Caramel@Tmp1 = add(1, X),
  Caramel@Tmp2 = div(2, Caramel@Tmp1),
  Caramel@Tmp3 = mult(3, Caramel@Tmp2).

This rewrite would be an optimization during translation that happens the moment we find a |> in there. We may need to just leave a mark and have a second compilation pass do this optimization. We'd still need the external to be able to type-check correctly, but we won't actually end up creating a caramel_runtime:pipe/2.

It also means we can take our time to work out the quirks of partial application in the presence of optional and named arguments as pointed out here: https://github.com/AbstractMachinesLab/caramel/pull/92/files#r605967813

@ayshiff would you be interested in trying this approach? 🙌🏽

Yes, absolutely! I am interested in trying this 👍

[ayshiff]

Hey @ostera @michallepicki 👋 I'm here to ask for some help. I tried to follow Leandro's recommendations but I'm a bit lost. I made a branch with my progress if you want to have a look 👍

https://github.com/ayshiff/caramel/tree/feature/pipe-operator-test

I also have a question about testing. To generate a unique variable name in the transformation phase, I wanted to rely on random hash generation. The issue is that I don't know how we could test this behavior.

[leostera]

Hi! 👋🏽 since I'm currently not working on and have no plans to continue working on this version of the compiler, I'll close this PR. Thanks a lot for your contribution 🙏🏽