fix: don't deduplicate comments with same start index

[gzzhanghao]

Q	A
Fixed Issues?	Fixes #12769
Patch: Bug Fix?	Yes
Major: Breaking Change?	No
Minor: New Feature?	No
Tests Added + Pass?	Yes
Documentation PR Link	No
Any Dependency Changes?	No
License	MIT

Input code

const template = require('@babel/template');
const generator = require('@babel/generator');
const t = require("@babel/types");

const createStatement = template.statement("/*#__PURE__*/ foo()", {
  preserveComments: true,
});

const result = t.program([createStatement(), createStatement()]);

console.log(generator.default(result).code);

Expected behavior

/*#__PURE__*/
foo();
/*#__PURE__*/
foo();

Current behavior

/*#__PURE__*/
foo();
foo();

This PR includes following changes:

• Generate comments with same start index (Fixes @babel/generator comment deduplication removes too many comments #12769)
• Clone comments in cloneNode (Moved to fix: clone comments in cloneNode #13178)

[babel-bot]

Build successful! You can test your changes in the REPL here: https://babeljs.io/repl/build/45409/

[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 70e0c2b:

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

[kaicataldo]

The example of generating multiple statements for the program body in the PR author comment seems like a more likely scenario than what we're testing here. Can we change this to match that?

[kaicataldo]

Since modifying anything with comments ends up affecting a rather large surface area, I wonder if it would be prudent to split this up into two PRs (the cloning fix in one and the deduplication fix in the other) in case we find regressions and need to revert something.

[gzzhanghao]

@kaicataldo Good point, the cloning fix was moved to #13178.

[mischnic]

I think the check that is removed here was added so that the comment in this example (which exists twice in the AST, both as leading and trailing)

if (1) {
  a();
}
// /* */
else {
  b();
}

isn't printed twice.

Have you checked that this case still behaves correctly?

[gzzhanghao]

@mischnic We already have a similar test case: babel-generator/test/fixtures/comments/try-block-line-comment/input.js

[nicolo-ribaudo]

Yeah, @mischnic's code works in the REPL for this PR! https://babeljs.io/repl/build/45409/#?browsers=chrome%2060&build=&builtIns=false&spec=false&loose=false&code_lz=JYMwBAFAjAlGDeAoMYCGEYG5EF9EHp8x8AqME_RAUwBsBnKhZMAIw2xyA&debug=false&forceAllTransforms=false&shippedProposals=false&circleciRepo=&evaluate=false&fileSize=false&timeTravel=false&sourceType=module&lineWrap=true&presets=&prettier=false&targets=&version=7.13.15%2Bpr.13169&externalPlugins=

[nicolo-ribaudo]

👍

I think that the original use case for this object is now better handled by the _printedComments WeakSet.