🔗 Linked issue

Closes #177

📝 Description

serialize() uses String.localeCompare() for sorting object keys, which produces different results depending on the system locale.

In Slovak (sk-SK), the digraph ch is treated as a distinct letter after h. So {checkIn: 'foo', destination: 'bar'} serializes as {destination:'bar',checkIn:'foo'} in Slovak, but {checkIn:'foo',destination:'bar'} in English. This causes hash() to produce different values for the same input on different machines.

The same issue affects Set sorting and the compare() fallback path.

Fix

Replace all localeCompare calls with code-point comparison (a < b ? -1 : a > b ? 1 : 0) which is deterministic across all locales:

1. compare() for strings — a.localeCompare(b) → a < b ? -1 : a > b ? 1 : 0
2. compare() fallback — String.prototype.localeCompare.call(...) → code-point compare on serialized values
3. Object.keys() sort — a.localeCompare(b) → code-point compare

Code-point comparison uses UTF-16 code unit values, which is stable and locale-independent.

Impact

Without this fix, hash() produces different values for the same input depending on the system's LANG/LC_ALL setting. This breaks caching, content deduplication, and any system that relies on hash stability across environments.

Summary by CodeRabbit

• Refactor
  • Optimized internal string comparison and object key sorting to use direct comparisons instead of locale-based calls.
  • Added caching for serialized values during mixed-type comparisons to reduce repeated work and improve performance.