Changes needed to compile reason/refmt with bucklescript

[jchavarri]

I managed to compile refmt as a 1st class citizen BuckleScript library and run it in the browser successfully thanks to the bspacked Js_refmt_compiler artifact generated by ninja. See https://github.com/jchavarri/bs-refmt.

Some people have mentioned interest to have access to Reason AST using BuckleScript (cc @sgrove) and also my old self 😄 . I know the playground is not critical path, but having a version of BuckleScript that can compile refmt could allow the community to experiment and create more playgrounds & learning tools for Reason and BuckleScript.

These PR includes the changes needed in the compiler to make it work. I understand the Sys.ml change is not mergeable as is, maybe we could set BS_BROWSER env var when running ninja build -playground so the datasets get this change when building the compiler for browser?

https://github.com/BuckleScript/bucklescript/blob/5f7f45a97061b5ed9f249b63f66603506845d44b/scripts/ninja.js#L1778

[bobzhang]

hi @jchavarri , are you aware of the work we are doing here? https://github.com/BuckleScript/bucklescript.github.io/pull/211

[jchavarri]

Hi @bobzhang, yes I saw that progress, it looks awesome!

I'm not sure I understand the connection with this PR though, here we're adding some tags that were removed to the Obj module, these tags are needed to compile refmt with BuckleScript.

Here's a screenshot of the OCaml parsetree resulting of parsing some string on the browser, all this is done through Bucklescript, no jsoo involved (as you can see from records being compiled to JS objects):

[jchavarri]

I managed to build a bspacked version of refmt that can be compiled with the published version of bs-platform without relying on these changes. :)

[bobzhang]

@jchavarri glad to know you get it worked. I mention it because we may not support functions in Obj module in the future (currently not good either), the only one will in the obj module you can rely on is Obj.magic since it is used a lot..