[BUG] Currying in React components

[eWert-Online]

Since version melange.4.0.1-52 all react components are getting curried.
The currying of the components leads to the following warning from eslint:

React Hook React.useEffect has an unnecessary dependency: 'a'. Either exclude it or remove the dependency array. Outer scope values like 'a' aren't valid dependencies because mutating them doesn't re-render the component.

Take for example the following component:

[@react.component]
let make = (~a, ~b) => {
  React.useEffect2(
    () => {
      if (a) {
        Js.log(a);
      } else {
        Js.log(b);
      };
      None;
    },
    (a, b),
  );

  React.null;
};

In version 4.0.0-51, the following JavaScript is generated:

// Generated by Melange

import * as React from "react";

function _none_(Props) {
  var a = Props.a;
  var b = Props.b;
  React.useEffect((function () {
          if (a) {
            console.log(a);
          } else {
            console.log(b);
          }
        }), [
        a,
        b
      ]);
  return null;
}

var make = _none_;

export {
  make ,
}
/* react Not a pure module */

In version 4.0.1-52, the following JavaScript is generated:

// Generated by Melange

import * as React from "react";

function make(a) {
  return function (b) {
    React.useEffect((function () {
            if (a) {
              console.log(a);
            } else {
              console.log(b);
            }
          }), [
          a,
          b
        ]);
    return null;
  };
}

function Test(Props) {
  return make(Props.a)(Props.b);
}

const make$1 = Test;

export {
  make$1 as make,
}
/* react Not a pure module */

[anmonteiro]

this does seem slightly problematic.

i have a feeling that this isn't something we can solve in melange without ppxlib's help. thankfully, there's work to port ppxlib's AST to the OCaml 5.2 AST, which i think would cause this to get fixed.

FTR i believe this is the underlying change that's causing this issue: ocaml/ocaml#12236

[davesnx]

I haven't tested deeply, but while working on the reason-react-ppx last week, I found the generated Pexp_fun problematic here. If it's what I think it is (a not very optimal usage of the pexp_fun in the generation) I'm able to fix this in a future release of reason-react-ppx

[anmonteiro]

I found the generated Pexp_fun problematic here

what else can we generate instead?