fix: Skip buffered rows which are not joined with streamed side when …

…checking join filter results
viirya · Aug 25, 2024 · 0e6e5a2 · 0e6e5a2
1 parent cb3ec77
commit 0e6e5a2
Showing 1 changed file with 6 additions and 0 deletions.
diff --git a/datafusion/physical-plan/src/joins/sort_merge_join.rs b/datafusion/physical-plan/src/joins/sort_merge_join.rs
@@ -1460,6 +1460,12 @@ impl SMJStream {
                         // we need to output it with nulls as streamed side.
                         if matches!(self.join_type, JoinType::Full) {
                             for i in 0..pre_mask.len() {
+                                // If the buffered row is not joined with streamed side,
+                                // skip it.
+                                if buffered_indices.is_null(i) {
+                                    continue;
+                                }
+
                                 let buffered_batch = &mut self.buffered_data.batches
                                     [chunk.buffered_batch_idx.unwrap()];
                                 let buffered_index = buffered_indices.value(i);