peczenyj / GDPR-IAB-TCFv2 / 25480792771

96%

main: 96%

push

github

perf(range-section): avoid O(n^2) hash rebuild in TO_JSON (#60)

Phase 4 / "Performance" deliverable.  Two changes:

1. Add bench/performance.pl, a Benchmark::cmpthese harness over the
   1024-string corpus that measures Parse, Parse+TO_JSON, Validate
   (fail-fast), and Validate_all.  Useful for verifying future perf
   work; numbers are reproducible to within a few percent.

2. Fix RangeSection::TO_JSON: the previous form

       %map = ( %map, map { $_ => $true } $range->[0] .. $range->[1] );

   re-copied the running %map into a list every iteration, making
   the assembly accidentally O(n^2) over the number of vendor IDs
   already accumulated.  The new form mutates per-element directly,
   restoring O(n).

   Honest disclaimer: on the current 1024-string corpus this is a
   wash (Parse+TO_JSON 326 -> 329 ops/sec, within noise).  The
   ranges in real-world TC strings are typically small enough that
   the O(n^2) tail never dominates.  Worth landing as a scalability
   guarantee against adversarial inputs (TC strings with very wide
   ranges that some CMPs do produce).

PR #37 was the original Phase 4 attempt but was authored 38 commits
behind devel and would have reverted Validator/CLI work shipped in
v0.370.  Recommend closing it in favor of this PR.

2 of 2 new or added lines in 1 file covered. (100.0%)

911 of 965 relevant lines covered (94.4%)

12091.91 hits per line