When there are no new values in DISTINCT aggregates skip the update step

duckdb · Mar 10, 2020 · 0fcf3bb · 0fcf3bb
1 parent 519a878
commit 0fcf3bb
Showing 1 changed file with 10 additions and 9 deletions.
diff --git a/src/execution/aggregate_hashtable.cpp b/src/execution/aggregate_hashtable.cpp
@@ -267,18 +267,19 @@ void SuperLargeHashTable::AddChunk(DataChunk &groups, DataChunk &payload) {
 					distinct_sel_vector[match_count++] = sel_idx;
 				}
 			}
+			if (match_count > 0) {
+				VectorCardinality distinct_cardinality(match_count, distinct_sel_vector);
+				Vector distinct_payload(distinct_cardinality);
+				Vector distinct_addresses(distinct_cardinality);
 
-			VectorCardinality distinct_cardinality(match_count, distinct_sel_vector);
-			Vector distinct_payload(distinct_cardinality);
-			Vector distinct_addresses(distinct_cardinality);
+				distinct_payload.Reference(payload.data[payload_idx]);
+				distinct_addresses.Reference(addresses);
 
-			distinct_payload.Reference(payload.data[payload_idx]);
-			distinct_addresses.Reference(addresses);
+				distinct_payload.Verify();
+				distinct_addresses.Verify();
 
-			distinct_payload.Verify();
-			distinct_addresses.Verify();
-
-			aggr.function.update(&distinct_payload, 1, distinct_addresses);
+				aggr.function.update(&distinct_payload, 1, distinct_addresses);
+			}
 			payload_idx++;
 		} else {
 			auto input_count = max((idx_t)1, (idx_t)aggr.child_count);