Incorrect symbol resolution in tree_shaking/no_side_effects_in_initialization?

[overlookmotel]

I am fairly sure that symbol resolution in this rule is not correct for ExportSpecifier:

oxc/crates/oxc_linter/src/rules/tree_shaking/no_side_effects_in_initialization/listener_map.rs

Lines 193 to 202 in 1d3e973

	impl<'a> ListenerMap for ExportSpecifier<'a> {
	fn report_effects(&self, options: &NodeListenerOptions) {
	let ctx = options.ctx;
	let symbol_table = ctx.symbols();
	if has_comment_about_side_effect_check(self.exported.span(), ctx) {
	let Some(name) = self.exported.identifier_name() else { return };
	let Some(symbol_id) = options.ctx.symbols().get_symbol_id_from_name(name.as_str())
	else {
	return;
	};

Background

I would like to remove the method SymbolTable::get_symbol_id_from_name, as in my opinion it's a footgun (#5456). We use it to get the SymbolId for a reference, but actually that's not what it's doing - it will give you the SymbolId of the first binding with this variable name anywhere in the AST.

They are not the same. e.g. in this case:

{
    let foo;
}
let bar = foo;

If we want to know if foo in let bar = foo; refers to a local variable, get_symbol_id_from_name("foo").is_some() says "yes it does". But that's wrong.

#5446 removes a usage of get_symbol_id_from_name and adds tests which demonstrate it was incorrect.

This rule

I'm pretty sure the usage of get_symbol_id_from_name in this rule is wrong too. Isn't the question we should be asking in this code "is the local a bound identifier?", not "is the exported a bound identifier?"

In the case of export {foo}, they're the same, but in case of export {foo as bar}, they're not.

However, I must admit this rule rather bends my brain! Can someone more familiar with this code advise?

@Boshen git "blames" you for this code. Can you help? But maybe @DonIsaac or @camc314 also may have an idea what this rule is about?