Use set in parser scope

[JLHwung]

Q	A
Fixed Issues?	Parsing let/const declarations costs O(n^2) time
License	MIT

This PR is similar to #13406: We use Set for checking duplicate bindings in parser scope.

Benchmark result (baseline is 7.14.4, current is this PR)

Length-1 let declarations

$ node --predictable ./many-let-declarations/1-length-binding.mjs

baseline 256 length-1 let bindings: 1772 ops/sec ±21.58% (0.564ms)
baseline 512 length-1 let bindings: 733 ops/sec ±0.68% (1.365ms)
baseline 1024 length-1 let bindings: 232 ops/sec ±0.78% (4.312ms)
baseline 2048 length-1 let bindings: 66.35 ops/sec ±1.49% (15ms)

current 256 length-1 let bindings: 2592 ops/sec ±26.52% (0.386ms)
current 512 length-1 let bindings: 1497 ops/sec ±1.32% (0.668ms)
current 1024 length-1 let bindings: 728 ops/sec ±0.85% (1.374ms)
current 2048 length-1 let bindings: 323 ops/sec ±1.32% (3.093ms)

[babel-bot]

Build successful! You can test your changes in the REPL here: https://babeljs.io/repl/build/46611/

[codesandbox-ci]

This pull request is automatically built and testable in CodeSandbox.

To see build info of the built libraries, click here or the icon next to each commit SHA.

Latest deployment of this branch, based on commit 82abbaa:

Sandbox	Source
babel-repl-custom-plugin	Configuration
babel-plugin-multi-config	Configuration

[nicolo-ribaudo]

Thanks for going through all these complexity issues.

[motiz88]

Just in case this isn't already on your radar @JLHwung, I recently noticed something similar in @babel/traverse's scope tracking - e.g. referencePaths should be a Set in

babel/packages/babel-traverse/src/scope/binding.ts

Lines 94 to 101 in b281fe3

	reference(path: NodePath) {
	if (this.referencePaths.indexOf(path) !== -1) {
	return;
	}
	this.referenced = true;
	this.references++;
	this.referencePaths.push(path);
	}

(Of course, there it's technically an API change so may not be straightforward to just change, but maybe we can add a secondary set to speed up checking for duplicates while still maintaining the array for compatibility.)

[JLHwung]

@motiz88 Thanks. That's a good catch!

A secondary set may introduce consistent issues due to multiple truth. We could provide referencePaths as an accessor for referencePathSet and migrate built-in plugins first.