A general way to test scope resolving logic

[hyf0]

oxc/crates/oxc_semantic/tests/util/mod.rs

Lines 15 to 19 in 9aa78f6

	pub struct SemanticTester {
	allocator: Allocator,
	source_type: SourceType,
	source_text: &'static str,
	}

The current SemanticTester is great to do many kinds of assertions. Such as

oxc/crates/oxc_semantic/tests/symbols.rs

Lines 8 to 13 in 9aa78f6

	SemanticTester::js("export class Foo {}")
	.has_root_symbol("Foo")
	.contains_flags(SymbolFlags::Class | SymbolFlags::Export)
	.has_number_of_references(0)
	.is_exported()
	.test();

.

But I kind of think the SemanticTester might not be that intuitive for testing general scope resolving logic, which means weather the symbol has correct scope and don't care about the flags or others.

Let's see we have test case

log()

let a = 1
log(a)
let b = 1
log(b)
{
  let a = 2
  log(a)
  log(b)
}

we could transform it to

log#0()

let a#1 = 1
log(a#1)
let b#1 = 1;
log(b#1);
{
  let a#2 = 2;
  log(a#2)
  log(b#1)
}

the x in #x is the ScopeId we get from the Sematics.

In this way, we could easily visualize the whether the scope is correct. And with snapshot testing, this could be a general way for testing without any other kinds of assertions.

---

I get this idea from contributing to swc. https://github.com/swc-project/swc/blob/main/crates/swc_ecma_transforms_base/tests/resolver/hoisting/output.js

[Boshen]

Intresting. I think it's hard to cover all cases by manually writing test cases, the general approach should be running Test262 over mangled files ... let me think about this.

[Boshen]

We should rely and take advantage of test262 instead of manual tests.