How to manipulate AST and generate code with comments

[shahryarjb]

Hello friends, Sorry for this question. I am using OXC project in my Rust app and have a problem.

As you see in the below code, I generate AST and change it after that with Codegen, I create a new code again, but the problem is; it cleans all the comments of file. how can prevent it not to do this?

    let source_type = SourceType::default();
    let parser = Parser::new(allocator, file_content, source_type).with_options(ParseOptions {
        parse_regular_expression: true,
        ..ParseOptions::default()
    });
    
     parsed = parser.parse();

    // This place I change the AST

    let codegen = Codegen::new();
    let generated_code = codegen.build(&parsed.program).code;

    Ok(generated_code)

I even used this, but still it cleans all the comment inside the code

Example js code:

let js_content = r#"
    // Change heading:
    import { foo } from "module-name";
    // Change heading:
    console.log("Duplicate import test");
"#;

Example rust code:

    let codegen = Codegen::new().with_options(CodegenOptions {
        comments: true,
        ..CodegenOptions::default()
    });
    let generated_code = codegen.build(&parsed.program).code;

Thank in advance

[RohinBhargava]

I am also running to this issue.

I believe it has to do with this code block:

oxc/crates/oxc_codegen/src/comment.rs

Lines 73 to 108 in 8eccdec

	pub(crate) fn get_statement_comments(
	&mut self,
	start: u32,
	) -> Option<(Vec<Comment>, Vec<Comment>)> {
	let comments = self.comments.remove(&start)?;
	let mut leading_comments = vec![];
	let mut unused_comments = vec![];
	for comment in comments {
	if self.is_leading_comments(&comment) {
	leading_comments.push(comment);
	continue;
	}
	if comment.is_legal(self.source_text) {
	match &self.options.legal_comments {
	LegalComment::None if self.options.comments => {
	leading_comments.push(comment);
	continue;
	}
	LegalComment::Inline => {
	leading_comments.push(comment);
	continue;
	}
	LegalComment::Eof | LegalComment::Linked(_) | LegalComment::External => {
	self.legal_comments.push(comment);
	continue;
	}
	LegalComment::None => {}
	}
	}
	unused_comments.push(comment);
	}
	Some((leading_comments, unused_comments))
	}

Here, we can see that in get_statement_comments, there is a notion of unused_comments which are tracked and repushed into the comments member property. It looks like when print_comments is used downstream, it takes the leading_comments and legal_comments and prints those individual comments, but I did not see where the repushed unused comments are printed out in the program. There are many possible places to suggest a fix, but I would recommend a config that allows us to indiscriminately add comments (make all comments relevant comments), so that print_comments can use the full set.

As a workaround, I've made each of my comments pseudo "legal comments" (https://modernjs.dev/en/configure/app/output/legal-comments.html) because I find it acceptable to do so, and they are preserved in my output. e.g.

let js_content = r#"
    //! Change heading:
    import { foo } from "module-name";
    //! Change heading:
    console.log("Duplicate import test");
"#;

I would expect that we can preserve source comments (and ideally newlines) as first party nodes in the AST/writer. Am happy to contribute back code if there is agreement, let me know!