Low allocation OTLP metrics marshaler

[laurit]

Continuation of #6410

Benchmark                                                                 Mode  Cnt     Score    Error   Units
MetricsRequestMarshalerBenchmark.marshalStateful                          avgt   10     1.013 ±  0.021   us/op
MetricsRequestMarshalerBenchmark.marshalStateful:gc.alloc.rate            avgt   10  3080.843 ± 62.442  MB/sec
MetricsRequestMarshalerBenchmark.marshalStateful:gc.alloc.rate.norm       avgt   10  3272.006 ±  0.001    B/op
MetricsRequestMarshalerBenchmark.marshalStateful:gc.count                 avgt   10    92.000           counts
MetricsRequestMarshalerBenchmark.marshalStateful:gc.time                  avgt   10    32.000               ms
MetricsRequestMarshalerBenchmark.marshalStatefulJson                      avgt   10     5.096 ±  0.079   us/op
MetricsRequestMarshalerBenchmark.marshalStatefulJson:gc.alloc.rate        avgt   10  1489.729 ± 23.126  MB/sec
MetricsRequestMarshalerBenchmark.marshalStatefulJson:gc.alloc.rate.norm   avgt   10  7960.030 ±  0.001    B/op
MetricsRequestMarshalerBenchmark.marshalStatefulJson:gc.count             avgt   10    45.000           counts
MetricsRequestMarshalerBenchmark.marshalStatefulJson:gc.time              avgt   10    17.000               ms
MetricsRequestMarshalerBenchmark.marshalStateless                         avgt   10     1.206 ±  0.016   us/op
MetricsRequestMarshalerBenchmark.marshalStateless:gc.alloc.rate           avgt   10    18.991 ±  0.248  MB/sec
MetricsRequestMarshalerBenchmark.marshalStateless:gc.alloc.rate.norm      avgt   10    24.007 ±  0.001    B/op
MetricsRequestMarshalerBenchmark.marshalStateless:gc.count                avgt   10       ≈ 0           counts
MetricsRequestMarshalerBenchmark.marshalStatelessJson                     avgt   10     4.292 ±  0.041   us/op
MetricsRequestMarshalerBenchmark.marshalStatelessJson:gc.alloc.rate       avgt   10   847.853 ±  8.212  MB/sec
MetricsRequestMarshalerBenchmark.marshalStatelessJson:gc.alloc.rate.norm  avgt   10  3816.025 ±  0.001    B/op
MetricsRequestMarshalerBenchmark.marshalStatelessJson:gc.count            avgt   10    25.000           counts
MetricsRequestMarshalerBenchmark.marshalStatelessJson:gc.time             avgt   10    11.000               ms

[codecov]

Codecov Report

Attention: Patch coverage is 96.95431% with 12 lines in your changes are missing coverage. Please review.

Project coverage is 91.26%. Comparing base (b5699ad) to head (113e663).
Report is 6 commits behind head on main.

Files	Patch %	Lines
.../metrics/LowAllocationMetricsRequestMarshaler.java	76.47%	3 Missing and 1 partial ⚠️
...nternal/otlp/metrics/MetricStatelessMarshaler.java	94.36%	2 Missing and 2 partials ⚠️
...ernal/otlp/metrics/ExemplarStatelessMarshaler.java	94.73%	0 Missing and 2 partials ⚠️
...orter/internal/otlp/metrics/ExemplarMarshaler.java	80.00%	0 Missing and 1 partial ⚠️
...nternal/otlp/metrics/NumberDataPointMarshaler.java	80.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #6422      +/-   ##
============================================
+ Coverage     90.85%   91.26%   +0.41%     
- Complexity     6028     6107      +79     
============================================
  Files           651      667      +16     
  Lines         17729    18114     +385     
  Branches       1777     1792      +15     
============================================
+ Hits          16107    16532     +425     
+ Misses         1105     1061      -44     
- Partials        517      521       +4     

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

[jack-berg]

Left a few straight forward comments, but looks good. Thanks!

[jack-berg]

What's the didfference between the sizeRepeatedMessageWithContext overloads? One accepts a MarshalerContext.Key and another doesn't.

[laurit]

The ones with key use forEach to traverse Collection or Map. Key is used to cache the Consumer that is passed to forEach. The ones without key use a counted loop.