Make expr_simplifier handle expressions when indexing strings, arrays and objects

[levi-nz]

Description:
This PR is a WIP that fixes all known issues in #8747.

BREAKING CHANGE:

Related issue (if exists):
Closes #8747

[changeset-bot]

⚠️ No Changeset found

Latest commit: 5458077

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[levi-nz]

Not totally sure if the changed code for array literals is correct -- the code for this is a little more complicated than strings.

[][0] works correctly for now, but not [][[]]. Will look into this later.

[levi-nz]

Another broken test case:

var _ = {0.5: 'test'}[0.5];

Currently simplifies to:

var _ = {
    0.5: 'test'
}[0.5];

Should simplify to:

var _ = 'test';

Not entirely sure if this is correct though because removing the ObjectLit could have side effects, ie if a value is a CallExpr

[levi-nz]

Only one potential issue so far, that KnownOp no longer implements Eq because of Index now taking f64 instead of i64, but seems OK so far to me.

[levi-nz]

@kdy1 The only remaining issue in terms of code changes is this line. Currently is_literal(props) does not work with the unit test below because the 0.5 is not considered a literal. I'm not sure why this fract() check exists and I'm not sure if changing this would break something.

https://github.com/levi-nz/swc/blob/454378d7eaac82bae40c5a770888e84580a1381b/crates/swc_ecma_utils/src/lib.rs#L1883

fold("({0.5: 'a'})['0.5']", "'a';");

[levi-nz]

I'm pretty sure Index(f64) can be changed back to Index(i64) as well, as IndexStr is always used for objects.

[levi-nz]

I have ran UPDATE=1 cargo test --test projects --test tsc, but not sure if all of these files should be changed, especially crates/swc/tests/tsc-references/templateStringInIndexExpressionES6.2.minified.js, which is now empty.

[kdy1]

CI failed

[levi-nz]

To do:

• Move stuff to ctx
• Handle known object properties
• Possibly add more unit tests

Will comment here again when everything is finalised

[levi-nz]

Is there a reason why stuff like [1][0] was replaced with (0, 1) before instead of just 1? The tests seem to pass fine but I want to be sure. I don't see why it should be replaced with a SeqExpr instead of just a Literal if there are no side effects.

[levi-nz]

cargo test -p swc_ecma_minifier terser_exec_tests__terser__compress__evaluate__unsafe_string_bad_index__input_js --features concurrent currently fails with a weird infinite loop bug, doesn't seem to be related to optimize_member_expr, need to investigate further

[kdy1]

(0, foo) is about making this undefined

[levi-nz]

(0, foo) is about making this undefined

Does this apply to objects too? Like should ({a: 5}).a be replaced with (0, 5) or just 5? Currently it's replaced with the latter but I want to be sure

[levi-nz]

Need to handle known object properties in simplify's optimize_member_expr. Will require a function signature change to accept obj and prop instead of expr. This PR should be ready for review tomorrow.

[kdy1]

(0, foo) is about making this undefined

Does this apply to objects too? Like should ({a: 5}).a be replaced with (0, 5) or just 5? Currently it's replaced with the latter but I want to be sure

I don't think so. this only matters for member exprs or direct eval calls