Prevent unnecessary heap sort when buckets needs to be ordered by key in NumericTermsAggregation (opensearch-project#17252)

Signed-off-by: Rishabh Maurya <[email protected]>

Author: rishabhmaurya

CHANGELOG.md

@@ -41,6 +41,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 - Propagate the sourceIncludes and excludes fields from fetchSourceContext to FieldsVisitor. ([#17080](https://github.com/opensearch-project/OpenSearch/pull/17080))
 - [Star Tree] [Search] Resolving Date histogram with metric aggregation using star-tree ([#16674](https://github.com/opensearch-project/OpenSearch/pull/16674))
 - [Star Tree] [Search] Extensible design to support different query and field types ([#17137](https://github.com/opensearch-project/OpenSearch/pull/17137))
+- Improve performace of NumericTermAggregation by avoiding unnecessary sorting([#17252](https://github.com/opensearch-project/OpenSearch/pull/17252))
 
 ### Dependencies
 - Bump `com.google.cloud:google-cloud-core-http` from 2.23.0 to 2.47.0 ([#16504](https://github.com/opensearch-project/OpenSearch/pull/16504))

server/src/main/java/org/opensearch/search/aggregations/bucket/terms/NumericTermsAggregator.java

@@ -64,6 +64,7 @@
 import java.io.IOException;
 import java.math.BigInteger;
 import java.util.Arrays;
+import java.util.Iterator;
 import java.util.List;
 import java.util.Map;
 import java.util.function.BiConsumer;
@@ -202,9 +203,19 @@ private InternalAggregation[] buildAggregations(long[] owningBucketOrds) throws
                 // Get the top buckets
                 B[] bucketsForOrd = buildBuckets(ordered.size());
                 topBucketsPerOrd[ordIdx] = bucketsForOrd;
-                for (int b = ordered.size() - 1; b >= 0; --b) {
-                    topBucketsPerOrd[ordIdx][b] = ordered.pop();
-                    otherDocCounts[ordIdx] -= topBucketsPerOrd[ordIdx][b].getDocCount();
+                if (isKeyOrder(order)) {
+                    for (int b = ordered.size() - 1; b >= 0; --b) {
+                        topBucketsPerOrd[ordIdx][b] = ordered.pop();
+                        otherDocCounts[ordIdx] -= topBucketsPerOrd[ordIdx][b].getDocCount();
+                    }
+                } else {
+                    // sorted buckets not needed as they will be sorted by key in buildResult() which is different from
+                    // order in priority queue ordered
+                    Iterator<B> itr = ordered.iterator();
+                    for (int b = ordered.size() - 1; b >= 0; --b) {
+                        topBucketsPerOrd[ordIdx][b] = itr.next();
+                        otherDocCounts[ordIdx] -= topBucketsPerOrd[ordIdx][b].getDocCount();
+                    }
                 }
             }