From c7a7d48d657f29ba43ea87d539d05cff07dc67ad Mon Sep 17 00:00:00 2001
From: Armin Braun <me@obrown.io>
Date: Fri, 7 Jun 2024 13:27:10 +0200
Subject: [PATCH] Reduce the heap use of BKDReader instances (#13464)

We consume a lot of memory for the `indexIn` slices. If `indexIn` is of
type `MemorySegmentIndexInput` the overhead of keeping loads of slices
around just for cloning is far higher than the extra 12b per reader this
adds (the slice description alone often costs a lot).
In a number of Elasticsearch example uses with high segment counts I
investigated, this change would save up to O(GB) of heap.
---
 .../java/org/apache/lucene/util/bkd/BKDReader.java    | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/lucene/core/src/java/org/apache/lucene/util/bkd/BKDReader.java b/lucene/core/src/java/org/apache/lucene/util/bkd/BKDReader.java
index d2d326b3a15..b2b109769f5 100644
--- a/lucene/core/src/java/org/apache/lucene/util/bkd/BKDReader.java
+++ b/lucene/core/src/java/org/apache/lucene/util/bkd/BKDReader.java
@@ -43,7 +43,9 @@ public class BKDReader extends PointValues {
   final int version;
   final long minLeafBlockFP;
 
-  final IndexInput packedIndex;
+  private final long indexStartPointer;
+  private final int numIndexBytes;
+  private final IndexInput indexIn;
   // if true, the tree is a legacy balanced tree
   private final boolean isTreeBalanced;
 
@@ -95,8 +97,7 @@ public class BKDReader extends PointValues {
     pointCount = metaIn.readVLong();
     docCount = metaIn.readVInt();
 
-    int numIndexBytes = metaIn.readVInt();
-    long indexStartPointer;
+    numIndexBytes = metaIn.readVInt();
     if (version >= BKDWriter.VERSION_META_FILE) {
       minLeafBlockFP = metaIn.readLong();
       indexStartPointer = metaIn.readLong();
@@ -105,7 +106,7 @@ public class BKDReader extends PointValues {
       minLeafBlockFP = indexIn.readVLong();
       indexIn.seek(indexStartPointer);
     }
-    this.packedIndex = indexIn.slice("packedIndex", indexStartPointer, numIndexBytes);
+    this.indexIn = indexIn;
     this.in = dataIn;
     // for only one leaf, balanced and unbalanced trees can be handled the same way
     // we set it to unbalanced.
@@ -158,7 +159,7 @@ public class BKDReader extends PointValues {
   @Override
   public PointTree getPointTree() throws IOException {
     return new BKDPointTree(
-        packedIndex.clone(),
+        indexIn.slice("packedIndex", indexStartPointer, numIndexBytes),
         this.in.clone(),
         config,
         numLeaves,