Merge remote-tracking branch 'elastic/main' into faster-field-caps

Author: original-brownbear

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

@@ -0,0 +1,78 @@
+/*
+ * Copyright Elasticsearch B.V. and/or licensed to Elasticsearch B.V. under one
+ * or more contributor license agreements. Licensed under the "Elastic License
+ * 2.0", the "GNU Affero General Public License v3.0 only", and the "Server Side
+ * Public License v 1"; you may not use this file except in compliance with, at
+ * your election, the "Elastic License 2.0", the "GNU Affero General Public
+ * License v3.0 only", or the "Server Side Public License, v 1".
+ */
+
+package org.elasticsearch.benchmark.vector;
+
+import org.apache.lucene.index.VectorSimilarityFunction;
+import org.elasticsearch.common.logging.LogConfigurator;
+import org.elasticsearch.index.codec.vectors.OptimizedScalarQuantizer;
+import org.openjdk.jmh.annotations.Benchmark;
+import org.openjdk.jmh.annotations.BenchmarkMode;
+import org.openjdk.jmh.annotations.Fork;
+import org.openjdk.jmh.annotations.Level;
+import org.openjdk.jmh.annotations.Measurement;
+import org.openjdk.jmh.annotations.Mode;
+import org.openjdk.jmh.annotations.OutputTimeUnit;
+import org.openjdk.jmh.annotations.Param;
+import org.openjdk.jmh.annotations.Scope;
+import org.openjdk.jmh.annotations.Setup;
+import org.openjdk.jmh.annotations.State;
+import org.openjdk.jmh.annotations.Warmup;
+
+import java.util.concurrent.ThreadLocalRandom;
+import java.util.concurrent.TimeUnit;
+
+@BenchmarkMode(Mode.Throughput)
+@OutputTimeUnit(TimeUnit.MILLISECONDS)
+@State(Scope.Benchmark)
+@Warmup(iterations = 3, time = 1)
+@Measurement(iterations = 5, time = 1)
+@Fork(value = 3)
+public class OptimizedScalarQuantizerBenchmark {
+    static {
+        LogConfigurator.configureESLogging(); // native access requires logging to be initialized
+    }
+    @Param({ "384", "702", "1024" })
+    int dims;
+
+    float[] vector;
+    float[] centroid;
+    byte[] destination;
+
+    @Param({ "1", "4", "7" })
+    byte bits;
+
+    OptimizedScalarQuantizer osq = new OptimizedScalarQuantizer(VectorSimilarityFunction.DOT_PRODUCT);
+
+    @Setup(Level.Iteration)
+    public void init() {
+        ThreadLocalRandom random = ThreadLocalRandom.current();
+        // random byte arrays for binary methods
+        destination = new byte[dims];
+        vector = new float[dims];
+        centroid = new float[dims];
+        for (int i = 0; i < dims; ++i) {
+            vector[i] = random.nextFloat();
+            centroid[i] = random.nextFloat();
+        }
+    }
+
+    @Benchmark
+    public byte[] scalar() {
+        osq.scalarQuantize(vector, destination, bits, centroid);
+        return destination;
+    }
+
+    @Benchmark
+    @Fork(jvmArgsPrepend = { "--add-modules=jdk.incubator.vector" })
+    public byte[] vector() {
+        osq.scalarQuantize(vector, destination, bits, centroid);
+        return destination;
+    }
+}

docs/changelog/125570.yaml

@@ -0,0 +1,5 @@
+pr: 125570
+summary: ES|QL random sampling
+area: Machine Learning
+type: feature
+issues: []

docs/changelog/126990.yaml

@@ -0,0 +1,7 @@
+pr: 126990
+summary: "Fix: consider case sensitiveness differences in Windows/Unix-like filesystems\
+  \ for files entitlements"
+area: Infra/Core
+type: bug
+issues:
+ - 127047

docs/changelog/127118.yaml

@@ -0,0 +1,5 @@
+pr: 127118
+summary: Panama vector accelerated optimized scalar quantization
+area: Vector Search
+type: enhancement
+issues: []

docs/changelog/127267.yaml

@@ -0,0 +1,5 @@
+pr: 127267
+summary: Bypass competitive iteration in single filter bucket case
+area: "Aggregations"
+type: bug
+issues: [127262]

libs/entitlement/src/main/java/org/elasticsearch/entitlement/runtime/policy/CaseInsensitiveComparison.java

@@ -0,0 +1,31 @@
+/*
+ * Copyright Elasticsearch B.V. and/or licensed to Elasticsearch B.V. under one
+ * or more contributor license agreements. Licensed under the "Elastic License
+ * 2.0", the "GNU Affero General Public License v3.0 only", and the "Server Side
+ * Public License v 1"; you may not use this file except in compliance with, at
+ * your election, the "Elastic License 2.0", the "GNU Affero General Public
+ * License v3.0 only", or the "Server Side Public License, v 1".
+ */
+
+package org.elasticsearch.entitlement.runtime.policy;
+
+class CaseInsensitiveComparison extends FileAccessTreeComparison {
+
+    CaseInsensitiveComparison(char separatorChar) {
+        super(CaseInsensitiveComparison::caseInsensitiveCharacterComparator, separatorChar);
+    }
+
+    private static int caseInsensitiveCharacterComparator(char c1, char c2) {
+        return Character.compare(Character.toLowerCase(c1), Character.toLowerCase(c2));
+    }
+
+    @Override
+    protected boolean pathStartsWith(String pathString, String pathPrefix) {
+        return pathString.regionMatches(true, 0, pathPrefix, 0, pathPrefix.length());
+    }
+
+    @Override
+    protected boolean pathsAreEqual(String path1, String path2) {
+        return path1.equalsIgnoreCase(path2);
+    }
+}

libs/entitlement/src/main/java/org/elasticsearch/entitlement/runtime/policy/CaseSensitiveComparison.java

@@ -0,0 +1,27 @@
+/*
+ * Copyright Elasticsearch B.V. and/or licensed to Elasticsearch B.V. under one
+ * or more contributor license agreements. Licensed under the "Elastic License
+ * 2.0", the "GNU Affero General Public License v3.0 only", and the "Server Side
+ * Public License v 1"; you may not use this file except in compliance with, at
+ * your election, the "Elastic License 2.0", the "GNU Affero General Public
+ * License v3.0 only", or the "Server Side Public License, v 1".
+ */
+
+package org.elasticsearch.entitlement.runtime.policy;
+
+class CaseSensitiveComparison extends FileAccessTreeComparison {
+
+    CaseSensitiveComparison(char separatorChar) {
+        super(Character::compare, separatorChar);
+    }
+
+    @Override
+    protected boolean pathStartsWith(String pathString, String pathPrefix) {
+        return pathString.startsWith(pathPrefix);
+    }
+
+    @Override
+    protected boolean pathsAreEqual(String path1, String path2) {
+        return path1.equals(path2);
+    }
+}

libs/entitlement/src/main/java/org/elasticsearch/entitlement/runtime/policy/FileAccessTree.java

@@ -11,11 +11,13 @@
 
 import org.elasticsearch.core.Nullable;
 import org.elasticsearch.core.Strings;
+import org.elasticsearch.core.SuppressForbidden;
 import org.elasticsearch.entitlement.runtime.policy.entitlements.FilesEntitlement;
 import org.elasticsearch.entitlement.runtime.policy.entitlements.FilesEntitlement.Mode;
 import org.elasticsearch.logging.LogManager;
 import org.elasticsearch.logging.Logger;
 
+import java.io.File;
 import java.io.IOException;
 import java.io.UncheckedIOException;
 import java.nio.file.Files;
@@ -33,7 +35,6 @@
 
 import static java.util.Comparator.comparing;
 import static org.elasticsearch.core.PathUtils.getDefaultFileSystem;
-import static org.elasticsearch.entitlement.runtime.policy.FileUtils.PATH_ORDER;
 import static org.elasticsearch.entitlement.runtime.policy.PathLookup.BaseDir.CONFIG;
 import static org.elasticsearch.entitlement.runtime.policy.PathLookup.BaseDir.TEMP;
 import static org.elasticsearch.entitlement.runtime.policy.entitlements.FilesEntitlement.Mode.READ_WRITE;
@@ -70,9 +71,9 @@
  * Secondly, the path separator (whether slash or backslash) sorts after the dot character. This means, for example, if the array contains
  * {@code ["/a", "/a.xml"]} and we look up {@code "/a/b"} using normal string comparison, the binary search would land on {@code "/a.xml"},
  * which is neither an exact match nor a containing directory, and so the lookup would incorrectly report no match, even though
- * {@code "/a"} is in the array. To fix this, we define {@link FileUtils#PATH_ORDER} which sorts path separators before any other
- * character. In the example, this would cause {@code "/a/b"} to sort between {@code "/a"} and {@code "/a.xml"} so that it correctly
- * finds {@code "/a"}.
+ * {@code "/a"} is in the array. To fix this, we define {@link FileAccessTreeComparison#pathComparator()} which sorts path separators
+ * before any other character. In the example, this would cause {@code "/a/b"} to sort between {@code "/a"} and {@code "/a.xml"} so that
+ * it correctly finds {@code "/a"}.
  * </p>
  * With the paths pruned, sorted, and segregated by permission, each binary search has the following properties:
  * <ul>
@@ -118,7 +119,11 @@ public String toString() {
         }
     }
 
-    static List<ExclusivePath> buildExclusivePathList(List<ExclusiveFileEntitlement> exclusiveFileEntitlements, PathLookup pathLookup) {
+    static List<ExclusivePath> buildExclusivePathList(
+        List<ExclusiveFileEntitlement> exclusiveFileEntitlements,
+        PathLookup pathLookup,
+        FileAccessTreeComparison comparison
+    ) {
         Map<String, ExclusivePath> exclusivePaths = new HashMap<>();
         for (ExclusiveFileEntitlement efe : exclusiveFileEntitlements) {
             for (FilesEntitlement.FileData fd : efe.filesEntitlement().filesData()) {
@@ -150,15 +155,16 @@ static List<ExclusivePath> buildExclusivePathList(List<ExclusiveFileEntitlement>
                 }
             }
         }
-        return exclusivePaths.values().stream().sorted(comparing(ExclusivePath::path, PATH_ORDER)).distinct().toList();
+        return exclusivePaths.values().stream().sorted(comparing(ExclusivePath::path, comparison.pathComparator())).distinct().toList();
     }
 
-    static void validateExclusivePaths(List<ExclusivePath> exclusivePaths) {
+    static void validateExclusivePaths(List<ExclusivePath> exclusivePaths, FileAccessTreeComparison comparison) {
         if (exclusivePaths.isEmpty() == false) {
             ExclusivePath currentExclusivePath = exclusivePaths.get(0);
             for (int i = 1; i < exclusivePaths.size(); ++i) {
                 ExclusivePath nextPath = exclusivePaths.get(i);
-                if (currentExclusivePath.path().equals(nextPath.path) || isParent(currentExclusivePath.path(), nextPath.path())) {
+                if (comparison.samePath(currentExclusivePath.path(), nextPath.path)
+                    || comparison.isParent(currentExclusivePath.path(), nextPath.path())) {
                     throw new IllegalArgumentException(
                         "duplicate/overlapping exclusive paths found in files entitlements: " + currentExclusivePath + " and " + nextPath
                     );
@@ -168,9 +174,18 @@ static void validateExclusivePaths(List<ExclusivePath> exclusivePaths) {
         }
     }
 
+    @SuppressForbidden(reason = "we need the separator as a char, not a string")
+    static char separatorChar() {
+        return File.separatorChar;
+    }
+
     private static final Logger logger = LogManager.getLogger(FileAccessTree.class);
     private static final String FILE_SEPARATOR = getDefaultFileSystem().getSeparator();
+    static final FileAccessTreeComparison DEFAULT_COMPARISON = Platform.LINUX.isCurrent()
+        ? new CaseSensitiveComparison(separatorChar())
+        : new CaseInsensitiveComparison(separatorChar());
 
+    private final FileAccessTreeComparison comparison;
     /**
      * lists paths that are forbidden for this component+module because some other component has granted exclusive access to one of its
      * modules
@@ -188,19 +203,27 @@ static void validateExclusivePaths(List<ExclusivePath> exclusivePaths) {
     private static String[] buildUpdatedAndSortedExclusivePaths(
         String componentName,
         String moduleName,
-        List<ExclusivePath> exclusivePaths
+        List<ExclusivePath> exclusivePaths,
+        FileAccessTreeComparison comparison
     ) {
         List<String> updatedExclusivePaths = new ArrayList<>();
         for (ExclusivePath exclusivePath : exclusivePaths) {
             if (exclusivePath.componentName().equals(componentName) == false || exclusivePath.moduleNames().contains(moduleName) == false) {
                 updatedExclusivePaths.add(exclusivePath.path());
             }
         }
-        updatedExclusivePaths.sort(PATH_ORDER);
+        updatedExclusivePaths.sort(comparison.pathComparator());
         return updatedExclusivePaths.toArray(new String[0]);
     }
 
-    private FileAccessTree(FilesEntitlement filesEntitlement, PathLookup pathLookup, Path componentPath, String[] sortedExclusivePaths) {
+    FileAccessTree(
+        FilesEntitlement filesEntitlement,
+        PathLookup pathLookup,
+        Path componentPath,
+        String[] sortedExclusivePaths,
+        FileAccessTreeComparison comparison
+    ) {
+        this.comparison = comparison;
         List<String> readPaths = new ArrayList<>();
         List<String> writePaths = new ArrayList<>();
         BiConsumer<Path, Mode> addPath = (path, mode) -> {
@@ -252,12 +275,12 @@ private FileAccessTree(FilesEntitlement filesEntitlement, PathLookup pathLookup,
         Path jdk = Paths.get(System.getProperty("java.home"));
         addPathAndMaybeLink.accept(jdk.resolve("conf"), Mode.READ);
 
-        readPaths.sort(PATH_ORDER);
-        writePaths.sort(PATH_ORDER);
+        readPaths.sort(comparison.pathComparator());
+        writePaths.sort(comparison.pathComparator());
 
         this.exclusivePaths = sortedExclusivePaths;
-        this.readPaths = pruneSortedPaths(readPaths).toArray(new String[0]);
-        this.writePaths = pruneSortedPaths(writePaths).toArray(new String[0]);
+        this.readPaths = pruneSortedPaths(readPaths, comparison).toArray(new String[0]);
+        this.writePaths = pruneSortedPaths(writePaths, comparison).toArray(new String[0]);
 
         logger.debug(
             () -> Strings.format(
@@ -270,14 +293,14 @@ private FileAccessTree(FilesEntitlement filesEntitlement, PathLookup pathLookup,
     }
 
     // package private for testing
-    static List<String> pruneSortedPaths(List<String> paths) {
+    static List<String> pruneSortedPaths(List<String> paths, FileAccessTreeComparison comparison) {
         List<String> prunedReadPaths = new ArrayList<>();
         if (paths.isEmpty() == false) {
             String currentPath = paths.get(0);
             prunedReadPaths.add(currentPath);
             for (int i = 1; i < paths.size(); ++i) {
                 String nextPath = paths.get(i);
-                if (currentPath.equals(nextPath) == false && isParent(currentPath, nextPath) == false) {
+                if (comparison.samePath(currentPath, nextPath) == false && comparison.isParent(currentPath, nextPath) == false) {
                     prunedReadPaths.add(nextPath);
                     currentPath = nextPath;
                 }
@@ -298,7 +321,8 @@ static FileAccessTree of(
             filesEntitlement,
             pathLookup,
             componentPath,
-            buildUpdatedAndSortedExclusivePaths(componentName, moduleName, exclusivePaths)
+            buildUpdatedAndSortedExclusivePaths(componentName, moduleName, exclusivePaths, DEFAULT_COMPARISON),
+            DEFAULT_COMPARISON
         );
     }
 
@@ -310,7 +334,7 @@ public static FileAccessTree withoutExclusivePaths(
         PathLookup pathLookup,
         @Nullable Path componentPath
     ) {
-        return new FileAccessTree(filesEntitlement, pathLookup, componentPath, new String[0]);
+        return new FileAccessTree(filesEntitlement, pathLookup, componentPath, new String[0], DEFAULT_COMPARISON);
     }
 
     public boolean canRead(Path path) {
@@ -346,24 +370,18 @@ private boolean checkPath(String path, String[] paths) {
             return false;
         }
 
-        int endx = Arrays.binarySearch(exclusivePaths, path, PATH_ORDER);
-        if (endx < -1 && isParent(exclusivePaths[-endx - 2], path) || endx >= 0) {
+        int endx = Arrays.binarySearch(exclusivePaths, path, comparison.pathComparator());
+        if (endx < -1 && comparison.isParent(exclusivePaths[-endx - 2], path) || endx >= 0) {
             return false;
         }
 
-        int ndx = Arrays.binarySearch(paths, path, PATH_ORDER);
+        int ndx = Arrays.binarySearch(paths, path, comparison.pathComparator());
         if (ndx < -1) {
-            return isParent(paths[-ndx - 2], path);
+            return comparison.isParent(paths[-ndx - 2], path);
         }
         return ndx >= 0;
     }
 
-    private static boolean isParent(String maybeParent, String path) {
-        var isParent = path.startsWith(maybeParent) && path.startsWith(FILE_SEPARATOR, maybeParent.length());
-        logger.trace(() -> Strings.format("checking isParent [%s] for [%s]: %b", maybeParent, path, isParent));
-        return isParent;
-    }
-
     @Override
     public boolean equals(Object o) {
         if (o == null || getClass() != o.getClass()) return false;