V8intrinsic syntax plugin

[JLHwung]

Q	A
Fixed Issues?	Fixes #10104
Patch: Bug Fix?
Major: Breaking Change?
Minor: New Feature?	✔️
Tests Added + Pass?	✔️
Documentation PR Link
Any Dependency Changes?	No
License	MIT

Authored a new syntax plugin v8intrinsic to support V8 CallRuntime syntax

CallRuntime ::
  '%' Identifier Arguments

The original implementation in V8 parser peek() twice to check whether a modulo token is immediately followed by an identifier and a parenL token.

As in Babel lookahead() does not support multiple steps, here we implicitly track the execution branch inside parseSubscript to see if parseCallExpressionArguments is immediately called in the first round of parseExprSubscripts. If so we conclude that the modulo token is a valid V8Intrinsic marker.

It is also interesting to note that a V8Intrinsic call inside a new expression does not raise a SyntaxError in V8. i.e.

node --allow-natives-syntax -e "new %DebugPrint(null)"
// throws TypeError: (intermediate value) is not a constructor

node -e "new %DebugPrint(null)"
// throws SyntaxError: Unexpected token %

@mathiasbynens I don't think there are any V8 intrinsics supposed to work with new expression, aren't there?

If so, it should be reasonable to prevent v8intrisic from new expression, as is the behavior of this plugin. Otherwise we still need to take care on the new expression.

Action Items

• Add v8intrinsic to babel-parser typings after updated ParsePlugin type definition #10259 gets merged.
• Document updates: feat: add v8intrinsic language extensions website#2074

[babel-bot]

Build successful! You can test your changes in the REPL here: https://babeljs.io/repl/build/11486/

[nicolo-ribaudo]

I was thinking that maybe we can remove this.state.V8IntrinsicStart and put all the parsing logic in a single method: we can overwrite parseExprAtom and add something like this:

1. If eat %%,
   i. Parse an identifier
   ii. Set that identifier's type to V8IntrinsicIdentifier
   iii. If the upcoming token isn't (, throw an error
2. Else, return super.parseExprAtom.

This is similar to what we do for super() and import().

[JLHwung]

I was thinking that maybe we can remove this.state.V8IntrinsicStart and put all the parsing logic in a single method

I agree, actually my first attack is on rewriting parseIdentifier but I find out it would mess up with the token position internal state. Apparently overriding parseExprAtom is way better than my previous approach. I have pushed the rewritten version.

Can you also add a test for yield?

Great observation! The parseYield will check if a token can startsExpr, if not it ends up with an empty yield statement, which fails under this case where % can actually start an expression. I would try to figure out a way to do this in the plugin land.

[mathiasbynens]

Thanks for your work on this!

@mathiasbynens I don't think there are any V8 intrinsics supposed to work with new expression, aren't there?

That's correct. Here's how V8 implements parsing support based on the flag.

[JLHwung]

That's correct. Here's how V8 implements parsing support based on the flag.

The implementation of V8 parsing V8Intrinsic will lead to the fact that new %StringParseInt("42", 10) will not throw a syntax error, instead it throws a runtime type error.

However as I have rewritten the plugin and now it closely follow the way how V8 parse. I have added this newExpression test case to the success test suites.

[JLHwung]

I've talked to Nicolo offline a bit. Here we add { startsExpr: true } configuration to modulo token in parser/types since

• It is only used in parseYield. The impact is under control.
• There have been some plugin-defined tokens inside this file, i.e. braceBarL, braceBarR. It is acceptable to place plugin-defined token configurations here.

[jridgewell]

Awesome!

[existentialism]

👍

[jridgewell]

@nicolo-ribaudo: Is this waiting for a release, or can it be merged?

[nicolo-ribaudo]

We just released v7.5 so this is waiting for the next minor, but if it is urgent we could do it earlier than the usual time between two minor releases.

[nicolo-ribaudo]

I just realized that we should add this to the allowed callees of CallExpression