feat(minifier): minimize switch statements

[armano2]

Phase 3: Advanced Optimizations

• Switch statement optimization

changes:

• converts 1-case and 2-case switches into more compact if or if/else statements when safe.
• removes redundant empty case suffixes at the end of a switch.
• removes empty switches (preserving the discriminant if it has side effects) and redundant trailing break statements.

note:

• i left couple of todo's for later additions, that could add additional optimizations, i can remove them if needed
• all packages seem to be affected due to napi last break removal from switch cases

future improvements:

• we could remove all empty cases if there is no default or when its empty as long as test has no side effects
• in-lining of switches with constant discriminants
• merge of identical consequent cases without side effects

fixes #17544

[codspeed-hq]

Merging this PR will not alter performance

✅ 38 untouched benchmarks
⏩ 7 skipped benchmarks¹

---

_{Comparing armano2:feat/minimize-switch (5118bba) with main (6ec08e8)²}

Footnotes

1. 7 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

2. No successful run was found on main (01d7b13) during the generation of this report, so 6ec08e8 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[armano2]

i'm pretty bad at naming stuff, suggestions are welcome 😄

[camc314]

no need to change this, I'm just interested to learn.

Is it worth checking if the discrimimnant has side effects here?

If we were to assume it does have side effects, the other pass that visits the statement expression and checks if it does have side effects would remove it anyway?

I guess there's perhaps some careful reasoning here about duplicate checks on may_have_side_effects vs duplicate traversals

[armano2]

correct, we do not want to keep expressions like switch(2){} -> 2

i' will wait for others to give feedback about this

[camc314]

This looks good to me! Thank you!

I'll let the experts (sapphi-red and Boshen have a look as well)

[armano2]

this could be made more generic, and used within eg, for (; x; ) break; to check if loop could be completely in-lined,

maybe we could add something like is_statement_terminated or is_terminated to statement nodes, instead of my is_terminated_switch_case and with addition of FindNestedBreak that could be really useful

oxc/crates/oxc_minifier/tests/peephole/esbuild.rs

Line 192 in 605dbf1

test("while (x) { if (1) break; z(); }", "for (; x; ) break;");

should i implement it in more generic way?

[armano2]

@Boshen i noticed that you started doing some changes in #17994 for switches, and looks like we worked on same thing

how i should proceed with this change?

[armano2]

new optimization has been added to main, and this code is partially redundant, tho my code supports more use cases

eg, trailing empty switch cases,

oxc/crates/oxc_minifier/src/peephole/minimize_statements.rs

Lines 605 to 629 in 7242197

	if let Some(last_case) = switch_stmt.cases.last()
	&& last_case.test.is_none()
	{
	let default_idx = switch_stmt.cases.len() - 1;
	let mut first_empty_idx = default_idx;
	// Iterate backward through preceding cases
	while first_empty_idx > 0 {
	let case = &switch_stmt.cases[first_empty_idx - 1];
	// Only remove empty cases with primitive literal tests
	if case.consequent.is_empty()
	&& case.test.as_ref().is_some_and(Expression::is_literal)
	{
	first_empty_idx -= 1;
	} else {
	break;
	}
	}
	// If we found cases to remove, keep cases [0..first_empty_idx] + [default]
	if first_empty_idx < default_idx {
	let default_case = switch_stmt.cases.pop().unwrap();
	switch_stmt.cases.truncate(first_empty_idx);
	switch_stmt.cases.push(default_case);
	ctx.state.changed = true;
	}
	}