Move result type from Belt to built-in.

[cristianoc]

No description provided.

[cristianoc]

@zth does this affect anything? E.g. pipe completion.

[zth]

I'd have to try it, I'm not sure really. I would assume no, and if it does it should be easy to fix.

[cknitt]

Why did the order in ocaml_re_test.js change?

[cristianoc]

Why did the order in ocaml_re_test.js change?

Because things change randomly whenever stuff in pervasive change.
Some long-standing nondeterminism in the compiler somewhere.
It's alway the order of mutually recursive definitions.

[cristianoc]

As a side comment, it seems likely that most tests mentioning Belt are either tests for Belt itself, or offer unclear value at this point (in catching future regressions). Both can be just removed.
The rest, should be easy to convert to not use belt.

[TheSpyder]

This breaks code that (unnecessarily) references the constructors as Belt.Result.Error and Belt.Result.Ok:
https://github.com/dusty-phillips/rescript-zora/blob/76e08155f24d8aa8c8bbdf7553522bbc5192039f/src/Zora.res#L41-L45

I don't mind submitting a PR to rescript-zora to fix it, but can this please be listed as a breaking change? It took me a while to track down that RC 5 was when it broke, and even longer to locate exactly why - because it wasn't listed in the changelog at all.

[anmonteiro]

Alternatively, the non-breaking change would be to alias the result type properly:

type result<'a, 'b> = result<'a, 'b> =
  | Ok('a)
  | Error('b)

[cknitt]

I agree, we should alias this properly.

@TheSpyder Would you mind submitting a PR to the rescript-compiler repo to for this?

[TheSpyder]

Ok, will do

[TheSpyder]

@cknitt #6514