Repository navigation
RFC: Replace language-javascript with Tree Sitter #3669
Description
PureScript relies on language-javascript package to parse FFI modules. This information is used to check whether FFI module contents match all foreign import declarations.
Unfortunately, moving to ES modules (#3613) requires extensive additions to language-javascript and even more to support all of the ES2015+ syntax. Instead of investing that time, we could leverage the excellent work that GitHub has done with Tree Sitter, an incremental parsing library with support for JavaScript, JSX, and even TypeScript.
Pros / Cons / Unknowns: Tree Sitter
- ✅Support parsing of latest JavaScript/JSX spec
- ✅Used by GitHub in production
- ✅Fast
- 🚫C library with Haskell bindings: https://github.com/tree-sitter/haskell-tree-sitter/blob/bdcd9cb715671c4b133bcc87c4a0c9905258ebf6/src/TreeSitter/Parser.hs#L11-L17
- 🚫Not on Hackage/Stackage
- ❓Error handling
- ❓Post-processing required (?)
Details: https://github.com/github/semantic/blob/master/docs/why-tree-sitter.md
Early POC: gasi#2
If there is interest to more easily enable features such as ES Modules (#3613) and allow more ES2015+ syntax in FFI files, e.g. arrow syntax, I’d be happy to take this task on.
Let me know if this change would be considered for merge 😄
Activity
hdgarrood commentedon Jun 4, 2019
I’m not super keen, for a few reasons:
purs bundle, and tree-sitter seems to provide a much lower level API to the AST than language-javascript does. We’d essentially need to rewrite the entirety ofpurs bundle. It’s not a huge amount of code, admittedly, but it’s not fantastically well tested, so I’d be a bit nervous.gasi commentedon Jun 6, 2019
That’s fair. Thanks for responding. I’ll see if I can continue adding support to
language-javascript.