[baltasarblanco]

Bug

console.log(Number.isInteger(5e-324), Number.isSafeInteger(5e-324), Number.isInteger(1e-300), Number.isSafeInteger(-1e-20));
// before: console.log(!0,!0,!0,!0)   -> true true true true
// after:  console.log(!1,!1,!1,!1)   -> false false false false

(oxc-minify 0.153.0, default options; same output on main at 9a9eb83.) Also through a constant (const x = 5e-324; Number.isInteger(x) → !0).

It also changes control flow:

var i = setImmediate(function () { console.log("ran"); });
if (Number.isInteger(5e-324)) clearImmediate(i);

minifies to var e=setImmediate(function(){console.log(`ran`)});clearImmediate(e);: the original prints ran, the minified prints nothing.

try_fold_number_methods (crates/oxc_ecmascript/src/constant_evaluation/call_expr.rs) tests extracted.fract().abs() < f64::EPSILON. f64::EPSILON is about 2.2e-16, so every number with a smaller fractional part counts as an integer. Number.isInteger and Number.isSafeInteger only accept an integral Number: "a finite Number value whose mathematical value is in the set of integers". That is an exact check.

Fix

is_integral = n.is_finite() && n.trunc() == n, used by both methods.

Tests

In crates/oxc_minifier/tests/peephole/replace_known_methods.rs: test_fold_number_functions_is_safe_integer gets 5e-324 and -1e-20 → false, and a new test_fold_number_functions_is_integer covers 1, 1e300 → true and 1.5, 5e-324, 1e-300, Infinity, NaN → false. Both functions fail without the change.

Checks: cargo test --all-features -p oxc_ecmascript -p oxc_minifier, cargo clippy --all-targets --all-features (warnings denied) on both crates, cargo fmt --check, cargo doc (warnings denied) on both crates, typos, cargo shear, napi/minify build-test + test; cargo minsize, cargo allocs and cargo coverage leave every snapshot unchanged.

Found with a differential fuzzer I wrote (it runs the original and the minified program under Node and compares the output). The fix, the tests and this description were written with AI assistance, and I reviewed them. The checks above and the before/after runs of the inputs were done by the AI agent in my local checkout; I did not re-run them by hand.

[Dunqing]

/monitor-oxc run

[oxc-guard]

Monitor Oxc

Commit: d83827d4e342

suite	result
Test262	❌
Build	✅
Runtime Bundles	✅
(codegen)	❌
Isolated Declarations	✅
(formatter_dcr)	✅
(mangler)	❌
(dce)	✅
(transformer)	❌
(compressor)	❌
(whitespace)	❌
(formatter)	❌
Uglify Corpus	✅
Runtime (zod)	✅
Runtime (es-toolkit)	✅
Runtime (svelte)	✅
Runtime (solid)	✅
Runtime (vue)	✅

[Copilot]

🟢 Approval recommended

The focused implementation matches JavaScript number semantics and is covered by relevant regression tests.

0 open findings

What changed in this PR

Corrects constant folding for Number.isInteger and Number.isSafeInteger to require exact integral values.

Changes:

• Replaces epsilon-based checks with finite, exact truncation comparison.
• Adds regression coverage for tiny fractions and special values.

File	Description
crates/​oxc_ecmascript/​src/​constant_evaluation/​call_expr.rs	Implements exact integral-number detection.
crates/​oxc_minifier/​tests/​peephole/​replace_known_methods.rs	Adds integer-folding regression tests.

🧠 Review effort: Balanced

---

💡 Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.

[codspeed]

Merging this PR will not alter performance

✅ 87 untouched benchmarks
⏩ 9 skipped benchmarks¹

---

_{Comparing baltasarblanco:fix/number-is-integer-tiny (d83827d) with main (6987a61)²}

Footnotes

1. 9 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 (666938c) during the generation of this report, so 6987a61 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩