Skip redundant tag deduplication #3902

Stephan202 · 2024-10-03T22:12:19Z

Both KeyValues and Tags unconditionally maintain the invariant that
they wrap distinct key-value pairs.

Stephan202

This PR relates to #3900 and #3901, in that all attempt to reduce the cost of instrumenting reactive streams.

Stephan202 · 2024-10-03T22:12:56Z

benchmarks/build.gradle

@@ -13,15 +13,18 @@ configurations {
 dependencies {
  // Use the baseline to avoid using new APIs in the benchmarks
  compileOnly libs.reactor.perfBaseline.core
+  compileOnly libs.reactor.perfBaseline.coreMicrometer


I'm not too familiar with Gradle, so happy to receive feedback if I should have modified this file (or gradle/libs.versions.toml below) in some other way.

Yep, this line together with the versions catalog are ok. I'll comment below what needs to be adjusted.

Stephan202 · 2024-10-03T22:14:20Z

.../core/observability/micrometer/MicrometerMeterListenerConfigurationResolveTagsBenchmark.java

+	@SuppressWarnings("unused")
+	@Benchmark
+	public Tags measureThroughput() {
+		return MicrometerMeterListenerConfiguration.resolveTags(publisher, Tags.of("k", "v"));
+	}


Benchmark results:

Before

Benchmark (distinctTagCount) (totalTagCount) Mode Cnt Score Error Units MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1 1 avgt 5 159.226 ± 8.332 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1 1 avgt 5 7331.787 ± 377.812 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1 1 avgt 5 1224.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1 1 avgt 5 100.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1 1 avgt 5 161.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1 2 avgt 5 217.163 ± 3.158 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1 2 avgt 5 5972.192 ± 87.004 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1 2 avgt 5 1360.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1 2 avgt 5 81.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1 2 avgt 5 115.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1 5 avgt 5 237.565 ± 7.034 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1 5 avgt 5 4945.661 ± 145.100 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1 5 avgt 5 1232.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1 5 avgt 5 67.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1 5 avgt 5 96.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1 10 avgt 5 286.132 ± 22.135 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1 10 avgt 5 4107.322 ± 309.521 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1 10 avgt 5 1232.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1 10 avgt 5 56.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1 10 avgt 5 83.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 2 1 avgt 5 142.298 ± 3.436 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 2 1 avgt 5 8202.963 ± 198.838 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 2 1 avgt 5 1224.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 2 1 avgt 5 112.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 2 1 avgt 5 166.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 2 2 avgt 5 254.355 ± 7.278 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 2 2 avgt 5 5369.084 ± 152.575 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 2 2 avgt 5 1432.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 2 2 avgt 5 73.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 2 2 avgt 5 107.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 2 5 avgt 5 367.433 ± 5.074 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 2 5 avgt 5 3384.417 ± 46.660 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 2 5 avgt 5 1304.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 2 5 avgt 5 46.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 2 5 avgt 5 70.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 2 10 avgt 5 343.450 ± 0.914 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 2 10 avgt 5 3976.036 ± 10.634 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 2 10 avgt 5 1432.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 2 10 avgt 5 54.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 2 10 avgt 5 84.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 5 1 avgt 5 150.793 ± 6.996 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 5 1 avgt 5 7741.417 ± 353.730 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 5 1 avgt 5 1224.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 5 1 avgt 5 105.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 5 1 avgt 5 133.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 5 2 avgt 5 321.504 ± 5.429 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 5 2 avgt 5 3867.861 ± 64.989 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 5 2 avgt 5 1304.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 5 2 avgt 5 52.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 5 2 avgt 5 75.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 5 5 avgt 5 488.914 ± 8.663 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 5 5 avgt 5 3214.443 ± 57.356 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 5 5 avgt 5 1648.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 5 5 avgt 5 43.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 5 5 avgt 5 61.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 5 10 avgt 5 527.945 ± 10.297 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 5 10 avgt 5 2861.202 ± 55.557 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 5 10 avgt 5 1584.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 5 10 avgt 5 39.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 5 10 avgt 5 61.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 10 1 avgt 5 153.768 ± 6.539 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 10 1 avgt 5 7591.569 ± 319.761 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 10 1 avgt 5 1224.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 10 1 avgt 5 104.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 10 1 avgt 5 140.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 10 2 avgt 5 349.661 ± 11.149 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 10 2 avgt 5 3905.636 ± 123.250 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 10 2 avgt 5 1432.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 10 2 avgt 5 53.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 10 2 avgt 5 76.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 10 5 avgt 5 453.534 ± 28.053 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 10 5 avgt 5 3465.851 ± 210.809 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 10 5 avgt 5 1648.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 10 5 avgt 5 47.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 10 5 avgt 5 70.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 10 10 avgt 5 647.491 ± 33.038 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 10 10 avgt 5 2957.819 ± 148.431 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 10 10 avgt 5 2008.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 10 10 avgt 5 40.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 10 10 avgt 5 58.000 ms

After

Benchmark (distinctTagCount) (totalTagCount) Mode Cnt Score Error Units MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1 1 avgt 5 122.536 ± 0.860 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1 1 avgt 5 6599.459 ± 46.318 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1 1 avgt 5 848.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1 1 avgt 5 89.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1 1 avgt 5 125.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1 2 avgt 5 147.859 ± 6.440 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1 2 avgt 5 5521.359 ± 237.145 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1 2 avgt 5 856.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1 2 avgt 5 75.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1 2 avgt 5 113.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1 5 avgt 5 221.202 ± 17.544 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1 5 avgt 5 4381.561 ± 338.380 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1 5 avgt 5 1016.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1 5 avgt 5 60.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1 5 avgt 5 91.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1 10 avgt 5 316.081 ± 4.174 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1 10 avgt 5 3354.992 ± 44.047 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1 10 avgt 5 1112.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1 10 avgt 5 46.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1 10 avgt 5 62.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 2 1 avgt 5 120.735 ± 2.683 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 2 1 avgt 5 6445.464 ± 142.300 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 2 1 avgt 5 816.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 2 1 avgt 5 87.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 2 1 avgt 5 121.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 2 2 avgt 5 142.788 ± 3.936 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 2 2 avgt 5 5717.046 ± 156.415 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 2 2 avgt 5 856.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 2 2 avgt 5 78.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 2 2 avgt 5 112.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 2 5 avgt 5 337.520 ± 4.360 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 2 5 avgt 5 2780.195 ± 35.886 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 2 5 avgt 5 984.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 2 5 avgt 5 38.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 2 5 avgt 5 61.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 2 10 avgt 5 586.248 ± 15.748 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 2 10 avgt 5 1913.004 ± 50.940 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 2 10 avgt 5 1176.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 2 10 avgt 5 26.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 2 10 avgt 5 38.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 5 1 avgt 5 122.197 ± 5.106 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 5 1 avgt 5 6618.401 ± 272.695 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 5 1 avgt 5 848.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 5 1 avgt 5 90.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 5 1 avgt 5 127.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 5 2 avgt 5 253.426 ± 13.879 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 5 2 avgt 5 3342.005 ± 179.684 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 5 2 avgt 5 888.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 5 2 avgt 5 46.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 5 2 avgt 5 66.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 5 5 avgt 5 348.439 ± 11.319 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 5 5 avgt 5 2605.632 ± 83.665 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 5 5 avgt 5 952.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 5 5 avgt 5 35.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 5 5 avgt 5 52.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 5 10 avgt 5 1080.822 ± 21.818 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 5 10 avgt 5 1009.391 ± 20.206 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 5 10 avgt 5 1144.002 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 5 10 avgt 5 13.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 5 10 avgt 5 23.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 10 1 avgt 5 114.838 ± 5.161 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 10 1 avgt 5 6776.790 ± 300.038 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 10 1 avgt 5 816.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 10 1 avgt 5 92.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 10 1 avgt 5 138.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 10 2 avgt 5 254.114 ± 6.417 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 10 2 avgt 5 3452.612 ± 86.498 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 10 2 avgt 5 920.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 10 2 avgt 5 47.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 10 2 avgt 5 70.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 10 5 avgt 5 314.048 ± 1.360 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 10 5 avgt 5 3085.122 ± 13.391 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 10 5 avgt 5 1016.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 10 5 avgt 5 42.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 10 5 avgt 5 61.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 10 10 avgt 5 453.549 ± 19.844 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 10 10 avgt 5 2472.887 ± 107.505 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 10 10 avgt 5 1176.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 10 10 avgt 5 34.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 10 10 avgt 5 52.000 ms

This shows that the new code is faster (up to 55%) and allocates less memory (up to 50%) in the common case, where few or no keys are duplicated. In cases where keys are heavily duplicated (distinctTagCount < 10, totalTagCount = 10), the new logic is slower but (a) this is not a scenerio one would expect to observe in practice and (b) users should generally be able to avoid duplicates, while they do not have control over the deduplication implementation.

micrometer-metrics/micrometer#4959 was just merged yesterday. I'm really curious how that performs together with your changes.

Thanks for sharing! Nice timing, and thanks for pushing that @mstyura 💪

I re-ran the before- and after-benchmarks against the latest Micrometer snapshot.

Before

Benchmark (testCase) Mode Cnt Score Error Units MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1|1 avgt 5 232.700 ± 12.088 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1|1 avgt 5 5377.355 ± 275.858 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1|1 avgt 5 1312.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1|1 avgt 5 73.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1|1 avgt 5 88.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1|2 avgt 5 253.188 ± 5.480 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1|2 avgt 5 4971.838 ± 106.974 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1|2 avgt 5 1320.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1|2 avgt 5 67.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1|2 avgt 5 87.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1|5 avgt 5 292.085 ± 19.269 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1|5 avgt 5 4519.590 ± 292.017 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1|5 avgt 5 1384.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1|5 avgt 5 61.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1|5 avgt 5 66.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1|10 avgt 5 303.817 ± 16.276 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1|10 avgt 5 4143.854 ± 218.000 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1|10 avgt 5 1320.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1|10 avgt 5 57.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1|10 avgt 5 73.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 2|2 avgt 5 274.764 ± 17.355 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 2|2 avgt 5 5054.366 ± 312.503 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 2|2 avgt 5 1456.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 2|2 avgt 5 68.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 2|2 avgt 5 96.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 2|5 avgt 5 384.112 ± 17.198 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 2|5 avgt 5 3615.106 ± 159.562 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 2|5 avgt 5 1456.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 2|5 avgt 5 49.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 2|5 avgt 5 59.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 2|10 avgt 5 320.729 ± 11.327 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 2|10 avgt 5 4139.081 ± 144.604 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 2|10 avgt 5 1392.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 2|10 avgt 5 56.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 2|10 avgt 5 63.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 5|5 avgt 5 415.296 ± 21.861 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 5|5 avgt 5 3692.877 ± 191.056 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 5|5 avgt 5 1608.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 5|5 avgt 5 50.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 5|5 avgt 5 56.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 5|10 avgt 5 458.328 ± 14.505 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 5|10 avgt 5 3345.821 ± 104.779 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 5|10 avgt 5 1608.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 5|10 avgt 5 46.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 5|10 avgt 5 57.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 10|10 avgt 5 606.856 ± 41.112 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 10|10 avgt 5 3093.244 ± 204.835 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 10|10 avgt 5 1968.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 10|10 avgt 5 43.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 10|10 avgt 5 47.000 ms

After

Benchmark (testCase) Mode Cnt Score Error Units MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1|1 avgt 5 139.095 ± 9.897 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1|1 avgt 5 6089.635 ± 423.860 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1|1 avgt 5 888.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1|1 avgt 5 82.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1|1 avgt 5 103.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1|2 avgt 5 170.519 ± 2.546 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1|2 avgt 5 5324.073 ± 79.438 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1|2 avgt 5 952.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1|2 avgt 5 73.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1|2 avgt 5 81.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1|5 avgt 5 230.257 ± 2.628 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1|5 avgt 5 4307.261 ± 49.140 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1|5 avgt 5 1040.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1|5 avgt 5 58.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1|5 avgt 5 70.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 1|10 avgt 5 278.842 ± 9.763 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 1|10 avgt 5 3912.650 ± 135.778 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 1|10 avgt 5 1144.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 1|10 avgt 5 53.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 1|10 avgt 5 66.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 2|2 avgt 5 157.001 ± 6.063 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 2|2 avgt 5 5831.547 ± 222.480 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 2|2 avgt 5 960.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 2|2 avgt 5 79.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 2|2 avgt 5 86.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 2|5 avgt 5 298.583 ± 9.338 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 2|5 avgt 5 3245.048 ± 100.476 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 2|5 avgt 5 1016.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 2|5 avgt 5 44.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 2|5 avgt 5 50.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 2|10 avgt 5 396.633 ± 14.369 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 2|10 avgt 5 2769.929 ± 99.182 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 2|10 avgt 5 1152.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 2|10 avgt 5 38.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 2|10 avgt 5 39.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 5|5 avgt 5 293.354 ± 15.722 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 5|5 avgt 5 3329.246 ± 175.336 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 5|5 avgt 5 1024.000 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 5|5 avgt 5 46.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 5|5 avgt 5 61.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 5|10 avgt 5 452.498 ± 28.634 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 5|10 avgt 5 2512.614 ± 155.733 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 5|10 avgt 5 1192.001 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 5|10 avgt 5 35.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 5|10 avgt 5 42.000 ms MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput 10|10 avgt 5 791.217 ± 16.072 ns/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate 10|10 avgt 5 1427.047 ± 28.815 MB/sec MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.alloc.rate.norm 10|10 avgt 5 1184.002 ± 0.001 B/op MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.count 10|10 avgt 5 19.000 counts MicrometerMeterListenerConfigurationResolveTagsBenchmark.measureThroughput:gc.time 10|10 avgt 5 23.000 ms

We see that in all cases less memory is allocated. The 2|10 and 10|10 cases show a small performance degradation, but in all other cases the new code is faster or on-par. Since (in my experience) in most cases there will be fewer than 10 tags, I'd say this is an over-all performance improvement.

Stephan202 · 2024-10-03T22:20:25Z

reactor-core/src/main/java/reactor/core/publisher/FluxMetrics.java

@@ -295,9 +295,8 @@ static Tags resolveTags(Publisher<?> source, Tags tags) {
 		Scannable scannable = Scannable.from(source);

 		if (scannable.isScanAvailable()) {
-			List<Tag> discoveredTags = scannable.tagsDeduplicated()


The method Scannable#tagsDeduplicated() is now unused, but I didn't drop it, as it's part of a public interface.

I don't suspect tagsDuplicated() is used outside of our codebase. In any case we should be able to deprecate it with the explanation of the observation you made that typical usages of the stream with duplicates do deduplication so it's unnecessary to de-duplicate them by design. The method is quite straightforward so it can be replicated in case anyone needs it.

Deprecating the method SGTM; indeed a GitHub search does not seem to turn up additional usages. I've proposed a @deprecated sentence that hopefully explains the situation, without claiming that no caller would find the method useful.

Stephan202 · 2024-10-03T22:20:40Z

...rc/main/java/reactor/core/observability/micrometer/MicrometerMeterListenerConfiguration.java

 		if (scannable.isScanAvailable()) {
-			List<Tag> discoveredTags = scannable.tagsDeduplicated()
-				.entrySet().stream()
-				.map(e -> Tag.of(e.getKey(), e.getValue()))
+			List<Tag> discoveredTags = scannable.tags()
+				.map(t -> Tag.of(t.getT1(), t.getT2()))
 				.collect(Collectors.toList());
 			return tags.and(discoveredTags);
 		}


This code snippet occurred in three places; all three are updated in an identical manner. In each case the code is covered by a test that validates the desired deduplication logic ("last one wins"). (For brevity the new benchmark covers one of the two non-deprecated classes.)

That's a very important observation actually and I believe it deserves to be documented (not in javadocs, but just a line comment).

Check, I added a comment about the "last one wins" strategy. Let me know if you instead meant that the code snippets should reference each other and/or all point to the same benchmark class.

chemicL

Thanks for the PR. It looks very promising. I left a few comments.

chemicL · 2024-10-04T09:23:36Z

...rc/main/java/reactor/core/observability/micrometer/MicrometerMeterListenerConfiguration.java

 		if (scannable.isScanAvailable()) {
-			List<Tag> discoveredTags = scannable.tagsDeduplicated()
-				.entrySet().stream()
-				.map(e -> Tag.of(e.getKey(), e.getValue()))
+			List<Tag> discoveredTags = scannable.tags()
+				.map(t -> Tag.of(t.getT1(), t.getT2()))
 				.collect(Collectors.toList());
 			return tags.and(discoveredTags);
 		}


That's a very important observation actually and I believe it deserves to be documented (not in javadocs, but just a line comment).

chemicL · 2024-10-04T09:32:49Z

reactor-core/src/main/java/reactor/core/publisher/FluxMetrics.java

@@ -295,9 +295,8 @@ static Tags resolveTags(Publisher<?> source, Tags tags) {
 		Scannable scannable = Scannable.from(source);

 		if (scannable.isScanAvailable()) {
-			List<Tag> discoveredTags = scannable.tagsDeduplicated()


I don't suspect tagsDuplicated() is used outside of our codebase. In any case we should be able to deprecate it with the explanation of the observation you made that typical usages of the stream with duplicates do deduplication so it's unnecessary to de-duplicate them by design. The method is quite straightforward so it can be replicated in case anyone needs it.

chemicL · 2024-10-04T10:03:53Z

.../core/observability/micrometer/MicrometerMeterListenerConfigurationResolveTagsBenchmark.java

+	@Param({"1", "2", "5", "10"})
+	private int distinctTagCount;
+
+	@Param({"1", "2", "5", "10"})
+	private int totalTagCount;


Consider that this results in an outer product, which means a full matrix of combinations while you only need half of it. The runs for distinct = 10, 5, 2, 1 with total = 1exercise the same dataset. The same for distinct = 10, 5, 2, total = 2, and so on. Perhaps a String that you can split in @Setup would be useful, e.g.

@Param({"1|1", "1|2", "1|5", "1|10", "2|2", "2|5", ...})

Nice idea; applied!

chemicL · 2024-10-04T10:12:31Z

.../core/observability/micrometer/MicrometerMeterListenerConfigurationResolveTagsBenchmark.java

+	@SuppressWarnings("unused")
+	@Benchmark
+	public Tags measureThroughput() {
+		return MicrometerMeterListenerConfiguration.resolveTags(publisher, Tags.of("k", "v"));
+	}


micrometer-metrics/micrometer#4959 was just merged yesterday. I'm really curious how that performs together with your changes.

chemicL · 2024-10-04T10:15:39Z

gradle/libs.versions.toml

@@ -9,6 +9,7 @@
 # Baselines, should be updated on every release
 baseline-core-api = "3.6.10"
 baselinePerfCore = "3.6.10"
+baselinePerfCoreMicrometer = "1.1.9"


Is there any reason why you chose 1.1.9 instead of 1.1.10?

Hah, I took the most recent version from my local Maven repo; should have checked Maven Central instead, of course.

Stephan202

Thanks for the review @chemicL! I added a commit addressing the feedback.

Stephan202 · 2024-10-04T18:20:54Z

.../core/observability/micrometer/MicrometerMeterListenerConfigurationResolveTagsBenchmark.java

+	@Param({"1", "2", "5", "10"})
+	private int distinctTagCount;
+
+	@Param({"1", "2", "5", "10"})
+	private int totalTagCount;


Nice idea; applied!

Stephan202 · 2024-10-04T18:21:58Z

.../core/observability/micrometer/MicrometerMeterListenerConfigurationResolveTagsBenchmark.java

+	@SuppressWarnings("unused")
+	@Benchmark
+	public Tags measureThroughput() {
+		return MicrometerMeterListenerConfiguration.resolveTags(publisher, Tags.of("k", "v"));
+	}


Thanks for sharing! Nice timing, and thanks for pushing that @mstyura 💪

Stephan202 · 2024-10-04T18:22:58Z

gradle/libs.versions.toml

@@ -9,6 +9,7 @@
 # Baselines, should be updated on every release
 baseline-core-api = "3.6.10"
 baselinePerfCore = "3.6.10"
+baselinePerfCoreMicrometer = "1.1.9"


Hah, I took the most recent version from my local Maven repo; should have checked Maven Central instead, of course.

Stephan202 · 2024-10-04T18:23:51Z

...rc/main/java/reactor/core/observability/micrometer/MicrometerMeterListenerConfiguration.java

 		if (scannable.isScanAvailable()) {
-			List<Tag> discoveredTags = scannable.tagsDeduplicated()
-				.entrySet().stream()
-				.map(e -> Tag.of(e.getKey(), e.getValue()))
+			List<Tag> discoveredTags = scannable.tags()
+				.map(t -> Tag.of(t.getT1(), t.getT2()))
 				.collect(Collectors.toList());
 			return tags.and(discoveredTags);
 		}


Check, I added a comment about the "last one wins" strategy. Let me know if you instead meant that the code snippets should reference each other and/or all point to the same benchmark class.

Stephan202 · 2024-10-04T18:29:15Z

reactor-core/src/main/java/reactor/core/publisher/FluxMetrics.java

@@ -295,9 +295,8 @@ static Tags resolveTags(Publisher<?> source, Tags tags) {
 		Scannable scannable = Scannable.from(source);

 		if (scannable.isScanAvailable()) {
-			List<Tag> discoveredTags = scannable.tagsDeduplicated()


Deprecating the method SGTM; indeed a GitHub search does not seem to turn up additional usages. I've proposed a @deprecated sentence that hopefully explains the situation, without claiming that no caller would find the method useful.

chemicL

Please see my comments regardng the benchmark setup and merge with latest main to get the versions aligned. I think it's reasonable to merge this proposal after the changes, thanks!

chemicL · 2024-10-17T12:14:20Z

benchmarks/build.gradle

@@ -13,15 +13,18 @@ configurations {
 dependencies {
  // Use the baseline to avoid using new APIs in the benchmarks
  compileOnly libs.reactor.perfBaseline.core
+  compileOnly libs.reactor.perfBaseline.coreMicrometer


Yep, this line together with the versions catalog are ok. I'll comment below what needs to be adjusted.

chemicL · 2024-10-17T12:16:20Z

benchmarks/build.gradle

  annotationProcessor "org.openjdk.jmh:jmh-generator-annprocess:$jmhVersion"

  current project(':reactor-core')
+  current project(':reactor-core-micrometer')


Please also modify the baseline source set and add the reactor-core-micrometer module with appropriate version. The idea here is that we can have a comparison of performance in the previously released version against the current version. That's why there's a "current" and a "baseline" source set.

Stephan202

I've rebased the branch and applied the requested change. Tnx for the review @chemicL!

Stephan202 · 2024-10-19T12:52:01Z

gradle/libs.versions.toml

@@ -9,6 +9,7 @@
 # Baselines, should be updated on every release
 baseline-core-api = "3.6.10"
 baselinePerfCore = "3.6.10"
+baselinePerfCoreMicrometer = "1.1.10"


By now 1.1.11 is out; will update.

Stephan202 · 2024-11-11T07:45:19Z

Hey @chemicL! Anything else I should do to move this PR forward? Thanks for considering 🙏

chemicL · 2024-11-21T14:52:25Z

@Stephan202 please merge in the latest main, update to the latest versions and we can merge.

Both `KeyValues` and `Tags` unconditionally maintain the invariant that they wrap distinct key-value pairs.

Stephan202 · 2024-11-21T17:06:37Z

Cool! I rebased and added a commit to bump the new dependency once more. Pre-push diff-of-diff for validation purposes:

$ diff -u <(git diff main...origin/sschroevers/skip-redundant-tag-deduplication) <(git diff main...sschroevers/skip-redundant-tag-deduplication)
--- /dev/fd/63	2024-11-21 18:01:47.805553471 +0100
+++ /dev/fd/62	2024-11-21 18:01:47.806553477 +0100
@@ -104,14 +104,14 @@
 +	}
 +}
 diff --git a/gradle/libs.versions.toml b/gradle/libs.versions.toml
-index a4139cb8e..a689711d8 100644
+index 20d092160..70f9fba45 100644
 --- a/gradle/libs.versions.toml
 +++ b/gradle/libs.versions.toml
 @@ -9,6 +9,7 @@
  # Baselines, should be updated on every release
- baseline-core-api = "3.6.11"
- baselinePerfCore = "3.6.11"
-+baselinePerfCoreMicrometer = "1.1.11"
+ baseline-core-api = "3.7.0"
+ baselinePerfCore = "3.7.0"
++baselinePerfCoreMicrometer = "1.2.0"
  baselinePerfExtra = "3.5.2"
  
  # Other shared versions

Locally I do see one test failure, but it also happens on main, so let's see what CI thinks 👀

ApplyingByteBuddyPluginGradleTest > applyingByteBuddyPluginDuringGradleBuild() FAILED
    org.opentest4j.AssertionFailedError: expected: <SUCCESS> but was: <UP_TO_DATE>
        at org.junit.jupiter.api.AssertionFailureBuilder.build(AssertionFailureBuilder.java:151)
        at org.junit.jupiter.api.AssertionFailureBuilder.buildAndThrow(AssertionFailureBuilder.java:132)
        at org.junit.jupiter.api.AssertEquals.failNotEqual(AssertEquals.java:197)
        at org.junit.jupiter.api.AssertEquals.assertEquals(AssertEquals.java:182)
        at org.junit.jupiter.api.AssertEquals.assertEquals(AssertEquals.java:177)
        at org.junit.jupiter.api.Assertions.assertEquals(Assertions.java:1145)
        at reactor.tools.agent.ApplyingByteBuddyPluginGradleTest.applyingByteBuddyPluginDuringGradleBuild(ApplyingByteBuddyPluginGradleTest.java:59)

chemicL · 2024-11-25T16:09:10Z

I am unable to trigger the above test failure nor did CI observe it. Seems unrelated nevertheless.

chemicL · 2024-11-25T16:18:01Z

@Stephan202 thank you for your contribution, it has now been merged 🚢

Stephan202 requested a review from a team as a code owner October 3, 2024 22:12

Stephan202 force-pushed the sschroevers/skip-redundant-tag-deduplication branch from 91f804c to a9062ba Compare October 3, 2024 22:21

Stephan202 commented Oct 3, 2024

View reviewed changes

chemicL requested changes Oct 4, 2024

View reviewed changes

chemicL added status/need-user-input This needs user input to proceed area/performance This belongs to the performance theme labels Oct 4, 2024

Stephan202 commented Oct 4, 2024

View reviewed changes

chemicL requested changes Oct 17, 2024

View reviewed changes

Stephan202 force-pushed the sschroevers/skip-redundant-tag-deduplication branch from da9c82b to 113d13b Compare October 19, 2024 12:53

Stephan202 commented Oct 19, 2024

View reviewed changes

Stephan202 added 4 commits November 21, 2024 17:57

Skip redundant tag deduplication

8004e47

Both `KeyValues` and `Tags` unconditionally maintain the invariant that they wrap distinct key-value pairs.

Address feedback

842e7da

Add JMH baseline, bump version again

3b28862

Bump reactor-core-micrometer dependency

7df3f70

Stephan202 force-pushed the sschroevers/skip-redundant-tag-deduplication branch from 113d13b to 7df3f70 Compare November 21, 2024 17:04

chemicL removed the status/need-user-input This needs user input to proceed label Nov 25, 2024

chemicL approved these changes Nov 25, 2024

View reviewed changes

chemicL merged commit 7cc701c into reactor:main Nov 25, 2024
7 checks passed

chemicL added the area/observability label Nov 25, 2024

Stephan202 deleted the sschroevers/skip-redundant-tag-deduplication branch November 25, 2024 16:28

Stephan202 mentioned this pull request Nov 27, 2024

Optimize observability and debugging experience #3901

Merged

chemicL mentioned this pull request Dec 4, 2024

NPE while recovering the Threadlocals in Reactor micrometer-metrics/tracing#898

Open

violetagg added the type/enhancement A general enhancement label Dec 10, 2024

violetagg added this to the 3.7.1 milestone Dec 10, 2024

Skip redundant tag deduplication #3902

Skip redundant tag deduplication #3902

Conversation

Stephan202 commented Oct 3, 2024

Stephan202 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Before

After

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Before

After

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

chemicL left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Stephan202 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

chemicL left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Stephan202 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Stephan202 commented Nov 11, 2024

chemicL commented Nov 21, 2024 • edited Loading

Stephan202 commented Nov 21, 2024

chemicL commented Nov 25, 2024

chemicL commented Nov 25, 2024

chemicL commented Nov 21, 2024 •

edited

Loading