Better callable subtyping for optional parameters

[yangdanny97]

If a function has N required params & additional optional params, it cannot be used in a place that expects a function with N params.

To give a concrete example: Int.toString cannot be directly used with Option.map because of the optional radix argument, so it has to be wrapped in another function.

https://rescript-lang.org/try?version=v12.1.0&module=esmodule&code=DYUwLgBGIM6QvBAFADwFwQPYAcwEtMA7AHj0LAD4BKCeCiAbwCgIIUBaCgeVwMIDoAtgENsSAJLl+YTAGUwAJzIBzKkwC+TJqEjQ4AJlrJ0WXkVLlqtes1YduZgSLEprESWGlzFK1FTXqQA

Tested in sandbox v12.1

[zth]

@cometkim you have some context why this isn't a good thing to allow IIRC?

[cristianoc]

It could be allowed (this way round, not the other way).
The question is mostly of language design: whether it would hide real mistakes, or harmless.

[yangdanny97]

FWIW, Python and TypeScript both allow this

[cometkim]

@cometkim you have some context why this isn't a good thing to allow IIRC?

There is a common accidents by binding index (the second param of the callback) as radix argument.

We can just pre-define monomorphic version of API for convenience.

[cometkim]

So... even in JS/TS project.

Providing radix argument is often recommended by additional linter.

• https://eslint.org/docs/latest/rules/radix
• https://biomejs.dev/linter/rules/use-parse-int-radix/