Merge remote-tracking branch 'upstream/main' into qualifiers-in-lu-join

Author: alex-spies

.buildkite/pipelines/pull-request/performance-benchmark.yml

@@ -0,0 +1,3 @@
+steps:
+  - label: ":pipeline: TODO"
+    command: echo TODO

.buildkite/pull-requests.json

@@ -18,7 +18,7 @@
     },
     {
       "enabled": true,
-      "pipeline_slug": "elasticsearch-performance-esbench-pr",
+      "pipeline_slug": "elasticsearch-pull-request-performance-benchmark",
       "allow_org_users": true,
       "allowed_repo_permissions": [
         "admin",

.gitignore

@@ -69,7 +69,7 @@ testfixtures_shared/
 # Generated
 checkstyle_ide.xml
 x-pack/plugin/esql/src/main/generated-src/generated/
-server/src/main/resources/transport/defined/manifest.txt
+server/src/main/resources/transport/definitions/manifest.txt
 
 # JEnv
 .java-version

BUILDING.md

@@ -92,7 +92,7 @@ uses the changed dependencies. In most cases, `precommit` or `check` are good ca
 We prefer sha256 checksums as md5 and sha1 are not considered safe anymore these days. The generated entry
 will have the `origin` attribute been set to `Generated by Gradle`.
 
-> [!Tip]  
+> [!Tip]
 > A manual confirmation of the Gradle generated checksums is currently not mandatory.
 > If you want to add a level of verification you can manually confirm the checksum (e.g. by looking it up on the website of the library)
 > Please replace the content of the `origin` attribute by `official site` in that case.
@@ -186,6 +186,25 @@ dependencies {
 
 To test an unreleased development version of a third party dependency you have several options.
 
+### How do I test against java early access (ea) versions?
+
+Currently only openjdk EA builds by oracle are supported.
+To test against an early access version java version you can pass the major
+java version appended with `-ea` as a system property (e.g. -Druntime.java=26-ea) to the Gradle build:
+
+```
+./gradlew clean test -Druntime.java=26-ea
+```
+
+This will run the tests using the JDK 26 EA version and pick the latest available build of the matching JDK EA version we expose
+in our custom jdk catalogue at `https://storage.googleapis.com/elasticsearch-jdk-archive/jdks/openjdk/latest.json`.
+
+To run against a specific build number of the EA build you can pass a second system property (e.g. `-Druntime.java.build=6`):
+
+```
+./gradlew clean test -Druntime.java=26-ea -Druntime.java.build=6
+```
+
 #### How to use a Maven based third party dependency via `mavenlocal`?
 
 1. Clone the third party repository locally
@@ -229,7 +248,7 @@ In addition to snapshot builds JitPack supports building Pull Requests. Simply u
 3. Run the Gradle build as needed. Keep in mind the initial resolution might take a bit longer as this needs to be built
 by JitPack in the background before we can resolve the adhoc built dependency.
 
-> [!Note]  
+> [!Note]
 > You should only use that approach locally or on a developer branch for production dependencies as we do
 not want to ship unreleased libraries into our releases.
 
@@ -261,7 +280,7 @@ allprojects {
   ```
 4. Run the Gradle build as needed with `--write-verification-metadata` to ensure the Gradle dependency verification does not fail on your custom dependency.
 
-> [!Note]  
+> [!Note]
 > As Gradle prefers to use modules whose descriptor has been created from real meta-data rather than being generated,
 flat directory repositories cannot be used to override artifacts with real meta-data from other repositories declared in the build.
 > For example, if Gradle finds only `jmxri-1.2.1.jar` in a flat directory repository, but `jmxri-1.2.1.pom` in another repository

benchmarks/src/main/java/org/elasticsearch/benchmark/_nightly/esql/QueryPlanningBenchmark.java

@@ -92,7 +92,8 @@ public void setup() {
         var fields = 10_000;
         var mapping = LinkedHashMap.<String, EsField>newLinkedHashMap(fields);
         for (int i = 0; i < fields; i++) {
-            mapping.put("field" + i, new EsField("field-" + i, TEXT, emptyMap(), true));
+            // We're creating a standard index, so none of these fields should be marked as dimensions.
+            mapping.put("field" + i, new EsField("field-" + i, TEXT, emptyMap(), true, EsField.TimeSeriesFieldType.NONE));
         }
 
         var esIndex = new EsIndex("test", mapping, Map.of("test", IndexMode.STANDARD));

benchmarks/src/main/java/org/elasticsearch/benchmark/compute/operator/EvalBenchmark.java

@@ -212,7 +212,7 @@ private static EvalOperator.ExpressionEvaluator evaluator(String operation) {
                 FieldAttribute timestamp = new FieldAttribute(
                     Source.EMPTY,
                     "timestamp",
-                    new EsField("timestamp", DataType.DATETIME, Map.of(), true)
+                    new EsField("timestamp", DataType.DATETIME, Map.of(), true, EsField.TimeSeriesFieldType.NONE)
                 );
                 yield EvalMapper.toEvaluator(
                     FOLD_CONTEXT,
@@ -321,19 +321,35 @@ private static EvalOperator.ExpressionEvaluator evaluator(String operation) {
     }
 
     private static FieldAttribute longField() {
-        return new FieldAttribute(Source.EMPTY, "long", new EsField("long", DataType.LONG, Map.of(), true));
+        return new FieldAttribute(
+            Source.EMPTY,
+            "long",
+            new EsField("long", DataType.LONG, Map.of(), true, EsField.TimeSeriesFieldType.NONE)
+        );
     }
 
     private static FieldAttribute doubleField() {
-        return new FieldAttribute(Source.EMPTY, "double", new EsField("double", DataType.DOUBLE, Map.of(), true));
+        return new FieldAttribute(
+            Source.EMPTY,
+            "double",
+            new EsField("double", DataType.DOUBLE, Map.of(), true, EsField.TimeSeriesFieldType.NONE)
+        );
     }
 
     private static FieldAttribute intField() {
-        return new FieldAttribute(Source.EMPTY, "int", new EsField("int", DataType.INTEGER, Map.of(), true));
+        return new FieldAttribute(
+            Source.EMPTY,
+            "int",
+            new EsField("int", DataType.INTEGER, Map.of(), true, EsField.TimeSeriesFieldType.NONE)
+        );
     }
 
     private static FieldAttribute keywordField() {
-        return new FieldAttribute(Source.EMPTY, "keyword", new EsField("keyword", DataType.KEYWORD, Map.of(), true));
+        return new FieldAttribute(
+            Source.EMPTY,
+            "keyword",
+            new EsField("keyword", DataType.KEYWORD, Map.of(), true, EsField.TimeSeriesFieldType.NONE)
+        );
     }
 
     private static Configuration configuration() {

benchmarks/src/main/java/org/elasticsearch/benchmark/exponentialhistogram/ExponentialHistogramGenerationBench.java

@@ -9,6 +9,7 @@
 
 package org.elasticsearch.benchmark.exponentialhistogram;
 
+import org.elasticsearch.exponentialhistogram.ExponentialHistogramCircuitBreaker;
 import org.elasticsearch.exponentialhistogram.ExponentialHistogramGenerator;
 import org.openjdk.jmh.annotations.Benchmark;
 import org.openjdk.jmh.annotations.BenchmarkMode;
@@ -59,7 +60,7 @@ public class ExponentialHistogramGenerationBench {
     @Setup
     public void setUp() {
         random = ThreadLocalRandom.current();
-        histoGenerator = new ExponentialHistogramGenerator(bucketCount);
+        histoGenerator = ExponentialHistogramGenerator.create(bucketCount, ExponentialHistogramCircuitBreaker.noop());
 
         DoubleSupplier nextRandom = () -> distribution.equals("GAUSSIAN") ? random.nextGaussian() : random.nextDouble();
 

benchmarks/src/main/java/org/elasticsearch/benchmark/exponentialhistogram/ExponentialHistogramMergeBench.java

@@ -11,6 +11,7 @@
 
 import org.elasticsearch.exponentialhistogram.BucketIterator;
 import org.elasticsearch.exponentialhistogram.ExponentialHistogram;
+import org.elasticsearch.exponentialhistogram.ExponentialHistogramCircuitBreaker;
 import org.elasticsearch.exponentialhistogram.ExponentialHistogramGenerator;
 import org.elasticsearch.exponentialhistogram.ExponentialHistogramMerger;
 import org.openjdk.jmh.annotations.Benchmark;
@@ -56,13 +57,14 @@ public class ExponentialHistogramMergeBench {
     @Setup
     public void setUp() {
         random = ThreadLocalRandom.current();
-        histoMerger = new ExponentialHistogramMerger(bucketCount);
+        ExponentialHistogramCircuitBreaker breaker = ExponentialHistogramCircuitBreaker.noop();
+        histoMerger = ExponentialHistogramMerger.create(bucketCount, breaker);
 
-        ExponentialHistogramGenerator initial = new ExponentialHistogramGenerator(bucketCount);
+        ExponentialHistogramGenerator initialGenerator = ExponentialHistogramGenerator.create(bucketCount, breaker);
         for (int j = 0; j < bucketCount; j++) {
-            initial.add(Math.pow(1.001, j));
+            initialGenerator.add(Math.pow(1.001, j));
         }
-        ExponentialHistogram initialHisto = initial.get();
+        ExponentialHistogram initialHisto = initialGenerator.getAndClear();
         int cnt = getBucketCount(initialHisto);
         if (cnt < bucketCount) {
             throw new IllegalArgumentException("Expected bucket count to be " + bucketCount + ", but was " + cnt);
@@ -72,14 +74,14 @@ public void setUp() {
         int dataPointSize = (int) Math.round(bucketCount * mergedHistoSizeFactor);
 
         for (int i = 0; i < toMerge.length; i++) {
-            ExponentialHistogramGenerator generator = new ExponentialHistogramGenerator(dataPointSize);
+            ExponentialHistogramGenerator generator = ExponentialHistogramGenerator.create(dataPointSize, breaker);
 
             int bucketIndex = 0;
             for (int j = 0; j < dataPointSize; j++) {
                 bucketIndex += 1 + random.nextInt(bucketCount) % (Math.max(1, bucketCount / dataPointSize));
                 generator.add(Math.pow(1.001, bucketIndex));
             }
-            toMerge[i] = generator.get();
+            toMerge[i] = generator.getAndClear();
             cnt = getBucketCount(toMerge[i]);
             if (cnt < dataPointSize) {
                 throw new IllegalArgumentException("Expected bucket count to be " + dataPointSize + ", but was " + cnt);

benchmarks/src/main/java/org/elasticsearch/benchmark/vector/PackAsBinaryBenchmark.java

@@ -82,4 +82,13 @@ public void packAsBinaryLegacy(Blackhole bh) {
             bh.consume(packed);
         }
     }
+
+    @Benchmark
+    @Fork(jvmArgsPrepend = { "--add-modules=jdk.incubator.vector" })
+    public void packAsBinaryPanama(Blackhole bh) {
+        for (int i = 0; i < numVectors; i++) {
+            BQVectorUtils.packAsBinary(qVectors[i], packed);
+            bh.consume(packed);
+        }
+    }
 }

benchmarks/src/main/java/org/elasticsearch/benchmark/vector/TransposeHalfByteBenchmark.java

@@ -83,4 +83,13 @@ public void transposeHalfByteLegacy(Blackhole bh) {
             bh.consume(packed);
         }
     }
+
+    @Benchmark
+    @Fork(jvmArgsPrepend = { "--add-modules=jdk.incubator.vector" })
+    public void transposeHalfBytePanama(Blackhole bh) {
+        for (int i = 0; i < numVectors; i++) {
+            BQSpaceUtils.transposeHalfByte(qVectors[i], packed);
+            bh.consume(packed);
+        }
+    }
 }