Support mel.as in mel.obj

[jchavarri]

This was never supported so when having to introduce different identifiers on JS, we would have to resort back to deriving abstract, e.g. for the DOM props in reason-react.

There would be no need to use deriving abstract if mel.as was supported in mel.obj which is what this PR allows.

[anmonteiro]

This seems to already have been supported, e.g.:

melange/jscomp/test/external_ppx.ml

Lines 20 to 26 in c5bf086

	external ff :
	hi:int ->
	lo:(_ [@mel.as 3]) ->
	lo2:(_ [@mel.as {json|{hi:-3 }|json}]) ->
	lo3:(_ [@mel.as -1]) ->
	lo4:(_ [@mel.as {json|-3|json}]) ->
	_ = "" [@@mel.obj]

There seems to be 2 use cases for @mel.as, and they're a bit at odds with each other (or at least a bit confusing):

1. the current use case, e.g.:

external ff : hi:int -> lo:(_[@mel.as 1]) -> _ = "" [@@mel.obj]

let u = ff ~hi:4

generates:

// Generated by Melange
'use strict';


var u = {
  hi: 4,
  lo: 1
};

exports.u = u;
/* No side effect */

2. your proposed use case

external func :   foo:(string [@mel.as "bar"]) -> ...

generating bar in the object key instead of foo.

Discussion

A few ways I see forward:

1. distinguish Ptyp_any and don't attempt to rename in that case, falling back to the previous behavior
2. use a new @mel.rename_label attribute (which I find ugly)
3. .. more?

[jchavarri]

Another option is to change the positioning of the attribute. I chose to annotate the type of the labelled argument because it's imo more ergonomic, but should we annotate the whole arrow type instead? e.g.

external ff :
  ((hi:int -> ((lo:string -> _)[@mel.as "low"]))[@mel.as "high"]) = ""
[@@mel.obj ]

instead of

external ff : hi:(int[@mel.as "high"]) -> lo:(string[@mel.as "low"]) -> _ = ""
[@@mel.obj]

The first option looks way more readable in Reason syntax than in OCaml syntax:

[@mel.obj]
external ff:
  [@mel.as "high"] ((~hi: int) => [@mel.as "low"] ((~lo: string) => _));

[anmonteiro]

I'd rather keep it on the type, then. even in reason that feels very very weird to write.

[anmonteiro]

@jchavarri given your recent constraints around React 18, would you like me to finish this PR?

[jchavarri]

Yes, I'd appreciate it. Thanks!