Adding benchmark for histogram collector over point range query

lucene/CHANGES.txt

@@ -94,7 +94,7 @@ Optimizations
 ---------------------
 * GITHUB#14418: Quick exit on filter query matching no docs when rewriting knn query. (Pan Guixin)
 
-* GITHUB#14439, GITHUB#14560: Efficient Histogram Collection using multi range traversal over PointTrees (Ankit Jain)
+* GITHUB#14439, GITHUB#14560, GITHUB#14622: Efficient Histogram Collection using multi range traversal over PointTrees (Ankit Jain)
 
 * GITHUB#14268: PointInSetQuery early exit on non-matching segments. (hanbj)
 

lucene/benchmark-jmh/src/java/org/apache/lucene/benchmark/jmh/HistogramCollectorBenchmark.java

@@ -33,8 +33,10 @@
 import org.apache.lucene.sandbox.facet.plain.histograms.HistogramCollectorManager;
 import org.apache.lucene.search.IndexSearcher;
 import org.apache.lucene.search.MatchAllDocsQuery;
+import org.apache.lucene.search.PointRangeQuery;
 import org.apache.lucene.store.Directory;
 import org.apache.lucene.store.MMapDirectory;
+import org.apache.lucene.util.NumericUtils;
 import org.openjdk.jmh.annotations.Benchmark;
 import org.openjdk.jmh.annotations.BenchmarkMode;
 import org.openjdk.jmh.annotations.Fork;
@@ -122,9 +124,40 @@ public static class BenchmarkParams {
   }
 
   @Benchmark
-  public void collectHistogram(BenchmarkParams params) throws IOException {
+  public void matchAllQueryHistogram(BenchmarkParams params) throws IOException {
     IndexSearcher searcher = new IndexSearcher(reader);
     searcher.search(
         new MatchAllDocsQuery(), new HistogramCollectorManager("f", params.bucketWidth, 10000));
   }
+
+  @Benchmark
+  public void pointRangeQueryHistogram(BenchmarkParams params) throws IOException {
+    IndexSearcher searcher = new IndexSearcher(reader);
+
+    Random r = new Random(0);
+    int lowerBound = r.nextInt(params.docCount / 4, 3 * params.docCount / 4);
+    // Filter for about 1/10 of the available documents
+    int upperBound = lowerBound + params.docCount / 10;
+
+    if (params.pointEnabled) {
+      byte[] lowerPoint = new byte[Long.BYTES];
+      byte[] upperPoint = new byte[Long.BYTES];
+      NumericUtils.longToSortableBytes(lowerBound, lowerPoint, 0);
+      NumericUtils.longToSortableBytes(upperBound, upperPoint, 0);
+      final PointRangeQuery prq =
+          new PointRangeQuery("f", lowerPoint, upperPoint, 1) {
+            @Override
+            protected String toString(int dimension, byte[] value) {
+              return Long.toString(NumericUtils.sortableBytesToLong(value, 0));
+            }
+          };
+
+      // Don't need to increase the default bucket count
+      searcher.search(prq, new HistogramCollectorManager("f", params.bucketWidth));
+    } else {
+      searcher.search(
+          NumericDocValuesField.newSlowRangeQuery("f", lowerBound, upperBound),
+          new HistogramCollectorManager("f", params.bucketWidth));
+    }
+  }
 }

lucene/sandbox/src/java/org/apache/lucene/sandbox/facet/plain/histograms/PointTreeBulkCollector.java

@@ -187,8 +187,11 @@ public PointValues.Relation compare(byte[] minPackedValue, byte[] maxPackedValue
           }
         }
 
-        // Not possible to have the CELL_OUTSIDE_QUERY, as bucket lower bound is updated
-        // while finalizing the previous bucket
+        // CELL_OUTSIDE_QUERY is possible for the first bucket
+        // in case of PointRangeQuery [min, max] lower
+        if (!collector.withinLowerBound(maxPackedValue)) {
+          return PointValues.Relation.CELL_OUTSIDE_QUERY;
+        }
         if (collector.withinRange(minPackedValue) && collector.withinRange(maxPackedValue)) {
           return PointValues.Relation.CELL_INSIDE_QUERY;
         }