[Feature] Local rewriting rule schema for attributes about ppxlib HOT 5 CLOSED

Could you describe what the the [@JSX] rewriting does? Just so that I understand whether it is something that would make sense to support in ppxlib. I think I see what it does, however if my understanding is correct then there might be performance implications.

from ppxlib.

For example ppx-tea-jsx rewrites this:

([@JSX] div ~class_:"a b" ~children:[
  ([@JSX] span ~style:[("text-weight", "bold")] ~children:["foo"] ());
  ([@JSX] br ~children:[] ());
] ())

Into this:

div [class "a b"] [
  span [style "text-weight" "bold"] [text "foo"];
  br [];
]

My ppx rewrites this:

([@JSX] div ~class_:["a"; "b"] ~children:[
  ([@JSX] span ~style:"text-weight: bold" ~children:["foo"] ())
] ())

Into this:

div ~a:[a_class ["a"; "b"]] [
  span ~a:[a_style "text-weight: bold"] [txt "foo"]
]

from ppxlib.

...Now that I think about it, it's not like this would net me a huge reduction in code size, since setting that up with Ast_traverse is pretty easy. If you think there would be performance implications then maybe it doesn't make a lot of sense to support this.

from ppxlib.

The performance implication could be manageable actually. However, this code is a bit tedious and adding this feature wouldn't be super fun :/

Looking at the rewriting, it seems to me that a more idiomatic one would be to rewrite expressions of the form:

[%JSX div ~class_:"a b" ~children:[
  span ~style:[("text-weight", "bold")] ~children:["foo"];
  br ~children:[];
]

Indeed, while you are free to rewrite the input how you wish with a ppx rewriter, the generally accepted convention is that extension points are the right choice for interpreting DSLs.

Such rewritings are well supported by ppxlib.

from ppxlib.

I agree that using extension points would be better and more idiomatic; I don't really have a choice since the Reason parser only emits the attributes.

The original Reason code actually looks like this:

<div class="a b">
  <span style=[("text-weight", "bold")]> "foo" </span>
  <br />
</div>

Which is arguably nicer to read (if you're used to JSX from ReactJS) and plays nicer with Reason's formatter and all the tooling.

Yeah, now that I think about this isn't a great idea. Reason already does it but it's probably better to encourage people to use extension points for DSLs instead of this.

As for Reason in particular it would be easier to fix it upstream or make a wrapper library for JSX rewriters to remove all the boilerplate.

from ppxlib.