✅ Deploy Preview for opentelemetry-cpp-api-docs canceled.

Codecov Report

Attention: Patch coverage is 95.08197% with 3 lines in your changes missing coverage. Please review.

Project coverage is 89.55%. Comparing base (83ac2ae) to head (5e2778c).
Report is 1 commits behind head on main.

@@            Coverage Diff             @@
##             main    #3322      +/-   ##
==========================================
+ Coverage   89.52%   89.55%   +0.03%     
==========================================
  Files         210      210              
  Lines        6526     6502      -24     
==========================================
- Hits         5842     5822      -20     
+ Misses        684      680       -4     

@ThomsonTan - Could you also include the benchmark results from before and after in the comments. A slight regression is acceptable, as it's a tradeoff for fixing collision.

@ThomsonTan - Could you also include the benchmark results from before and after in the comments. A slight regression is acceptable, as it's a tradeoff for fixing collision.

Here are the before/after benchmark result. It looks like that the wall time doubled while the CPU time reduced slightly.

• Before

./measurements_benchmark
2025-03-26T00:43:08+00:00
Running measurements_benchmark
Run on (16 X 2445.43 MHz CPU s)
CPU Caches:
  L1 Data 32 KiB (x8)
  L1 Instruction 32 KiB (x8)
  L2 Unified 512 KiB (x8)
  L3 Unified 32768 KiB (x1)
Load Average: 1.85, 3.23, 1.73
--------------------------------------------------------------
Benchmark                    Time             CPU   Iterations
--------------------------------------------------------------
BM_MeasurementsTest    2044809 ns       148990 ns         3856

• After

measurements_benchmark
2025-03-26T00:44:26+00:00
Running measurements_benchmark
Run on (16 X 2445.43 MHz CPU s)
CPU Caches:
  L1 Data 32 KiB (x8)
  L1 Instruction 32 KiB (x8)
  L2 Unified 512 KiB (x8)
  L3 Unified 32768 KiB (x1)
Load Average: 1.02, 2.64, 1.65
--------------------------------------------------------------
Benchmark                    Time             CPU   Iterations
--------------------------------------------------------------
BM_MeasurementsTest    4287068 ns        95297 ns         1000

@lalitb it seems the major perf gap is from the below original code. It doesn't sort the incoming KeyValueIterator before hashing and query, and the new code does the sort before the hash lookup.

  Aggregation *GetOrSetDefault(const opentelemetry::common::KeyValueIterable &attributes,
                               const AttributesProcessor *attributes_processor,
                               std::function<std::unique_ptr<Aggregation>()> aggregation_callback,
                               size_t hash)
  {
    auto it = hash_map_.find(hash);
    if (it != hash_map_.end())
    {
      return it->second.second.get();
    }

Found no issues on this patch (LGTM), but I don't know the metric aggregation code well enough to approve.
Deferring code review to @lalitb , will merge once approved.

@lalitb @ThomsonTan Assuming CI is clean, ok to merge ?

@lalitb @ThomsonTan Assuming CI is clean, ok to merge ?

Yes, it is Ok to merge it if CI is clean again.