perf(json): use FxHashSet for cycle detection in JSON.stringify

[iammdzaidalam]

Description

JSON.stringify currently performs cycle detection using linear scans over the active stack.

This replaces those checks with an FxHashSet, while keeping the existing traversal order in a Vec.

StateRecord now stores:

• stack: Vec<JsObject>
• stack_set: FxHashSet<JsObject>

Both are updated together in serialize_json_object and serialize_json_array.

Behavior remains unchanged, only the lookup strategy is different.

Changes

• use FxHashSet for cycle detection
• keep Vec for order
• ensure both structures stay in sync during push/pop

Tests

Added tests for:

• cyclic object
• cyclic array
• nested cyclic structures

Notes

Ran local checks (fmt, check, clippy, tests).

Tried running cargo bench -p boa_benches, but the suite hit a stack overflow in v8-benches/deltablue, so not using it as validation for this change.

[github-actions]

Test262 conformance changes

Test result	main count	PR count	difference
Total	52,963	52,963	0
Passed	50,073	50,073	0
Ignored	2,072	2,072	0
Failed	818	818	0
Panics	0	0	0
Conformance	94.54%	94.54%	0.00%

Tested main commit: 055ee0958ce332f3f99af27536a7c6f9a91def27
Tested PR commit: b41109eafe7beaa5dae30222bb2957e5ec3021eb
Compare commits: 055ee09...b41109e

[codecov]

Codecov Report

❌ Patch coverage is 71.42857% with 12 lines in your changes missing coverage. Please review.
✅ Project coverage is 59.80%. Comparing base (6ddc2b4) to head (3fbccc7).
⚠️ Report is 934 commits behind head on main.

Files with missing lines	Patch %	Lines
core/engine/src/builtins/json/mod.rs	71.42%	12 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #5145       +/-   ##
===========================================
+ Coverage   47.24%   59.80%   +12.56%     
===========================================
  Files         476      582      +106     
  Lines       46892    63490    +16598     
===========================================
+ Hits        22154    37972    +15818     
- Misses      24738    25518      +780     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[iammdzaidalam]

Hi @jedel1043, small perf change for JSON cycle detection, would appreciate a review.

[zhuzhu81998]

Tried running cargo bench -p boa_benches, but the suite hit a stack overflow in v8-benches/deltablue, so not using it as validation for this change.

a simple test that actually uses JSON.stringify and measures the time would be better than nothing.
not sure if the current benches use json functions at all.

[iammdzaidalam]

Yeah, that makes sense.

If boa_benches is not exercising JSON.stringify directly, it is not a very useful signal for this change. I'll add a small focused benchmark for it.

[iammdzaidalam]

Added focused JSON.stringify benchmarks for this path:

• stringify_deep.js (deep acyclic case)
• stringify_circular.js (cycle detection path)

I tried running them with cargo bench -p boa_benches -- json/..., but the current bench harness hits a pre-existing stack overflow before the filter takes effect. It looks like all script benches are initialized first, so unrelated v8-benches entries fail before reaching these.

Left that as a separate follow-up rather than expanding this PR.

[github-actions]

Test262 conformance changes

Test result	main count	PR count	difference
Total	52,963	52,963	0
Passed	50,545	50,545	0
Ignored	1,426	1,426	0
Failed	992	992	0
Panics	2	2	0
Conformance	95.43%	95.43%	0.00%

Tested main commit: 817b6eaade662a49434c5d295667068a4875b378
Tested PR commit: 3fbccc7a630920fb70bb49988d155a8121d22403
Compare commits: 817b6ea...3fbccc7

[zhuzhu81998]

Added focused JSON.stringify benchmarks for this path:

* `stringify_deep.js` (deep acyclic case)

* `stringify_circular.js` (cycle detection path)

I tried running them with cargo bench -p boa_benches -- json/..., but the current bench harness hits a pre-existing stack overflow before the filter takes effect. It looks like all script benches are initialized first, so unrelated v8-benches entries fail before reaching these.

Left that as a separate follow-up rather than expanding this PR.

then dont use the cargo bench. run the script manually and time it (remember to add a loop though). and compare the time between pr and main.

[iammdzaidalam]

yeah, fair. will just time it manually with a loop on main vs this PR and post numbers.

[iammdzaidalam]

Ran the scripts manually through boa_cli instead of cargo bench, using the same files on both main and this branch and timing the built binary directly.

On my machine (after warmup), timings were roughly:

Deep case:

• PR: ~14.3–17.4 ms
• main: ~15.5–18.3 ms

Circular case:

• PR: ~14.5–16.8 ms
• main: ~14.4–17.1 ms

The deep acyclic case shows a small improvement on this branch, while the circular case is roughly flat and within noise. This is just a quick manual check, not a rigorous benchmark.

[iammdzaidalam]

Benchmark

Ran the focused stringify scripts manually through boa_cli --release (10 rounds each, after warmup).

Deep case (acyclic object)

Metric	main	PR	Improvement
Typical	~46.6–47.0s	~42.3–42.7s	~9–10% faster
Min	~46.6s	~42.3s	~9.2% faster
Max	~50.3s	~46.4s	~7–8% faster

Circular case (cycle detection)

Metric	main	PR	Improvement
Typical	~1.69–1.72s	~1.38–1.39s	~15–20% faster
Min	~1.69s	~1.36s	~19.5% faster
Max	~1.79s	~1.40s	~21.8% faster

Notes

• measured using the same scripts on both branches
• each run uses 10 rounds after warmup
• numbers are approximate ranges from manual runs (not a rigorous benchmark)

[iammdzaidalam]

hey @zhuzhu81998 @jedel1043 , any updates when u get a moment?

[jedel1043]

Thank you! Looks good!