fix: prefer removing later duplicates in remove_redundant_entries

Swap the order of resolve_overlap checks so that when two orderings
are exact duplicates, the later occurrence is removed instead of the
earlier one. This preserves the declaration order of orderings when
the same ordering appears multiple times (e.g. from both Pass 1 and
Pass 2 of projected_orderings).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: adriangb

datafusion/physical-expr/src/equivalence/ordering.rs

@@ -104,18 +104,21 @@ impl OrderingEquivalenceClass {
             'outer: while idx < self.orderings.len() {
                 let mut ordering_idx = idx + 1;
                 while ordering_idx < self.orderings.len() {
-                    if let Some(remove) = self.resolve_overlap(idx, ordering_idx) {
+                    // Check the later index first so that for exact
+                    // duplicates we remove the later occurrence and
+                    // preserve the declaration order of orderings:
+                    if let Some(remove) = self.resolve_overlap(ordering_idx, idx) {
                         work = true;
                         if remove {
-                            self.orderings.remove(idx);
-                            continue 'outer;
+                            self.orderings.remove(ordering_idx);
+                            continue;
                         }
                     }
-                    if let Some(remove) = self.resolve_overlap(ordering_idx, idx) {
+                    if let Some(remove) = self.resolve_overlap(idx, ordering_idx) {
                         work = true;
                         if remove {
-                            self.orderings.remove(ordering_idx);
-                            continue;
+                            self.orderings.remove(idx);
+                            continue 'outer;
                         }
                     }
                     ordering_idx += 1;

datafusion/physical-expr/src/equivalence/properties/dependency.rs

@@ -492,7 +492,7 @@ mod tests {
 
         assert_eq!(
             out_properties.to_string(),
-            "order: [[a@0 ASC, c@2 ASC, b@1 ASC, d@3 ASC], [a@0 ASC, b@1 ASC, c@2 ASC, d@3 ASC]]"
+            "order: [[a@0 ASC, b@1 ASC, c@2 ASC, d@3 ASC], [a@0 ASC, c@2 ASC, b@1 ASC, d@3 ASC]]"
         );
 
         Ok(())

datafusion/sqllogictest/test_files/monotonic_projection_test.slt

@@ -97,7 +97,7 @@ logical_plan
 01)Sort: a_big ASC NULLS LAST, multiple_ordered_table.b ASC NULLS LAST
 02)--Projection: multiple_ordered_table.a, multiple_ordered_table.a AS a_big, multiple_ordered_table.b
 03)----TableScan: multiple_ordered_table projection=[a, b]
-physical_plan DataSourceExec: file_groups={1 group: [[WORKSPACE_ROOT/datafusion/core/tests/data/window_2.csv]]}, projection=[a, a@1 as a_big, b], output_orderings=[[a_big@1 ASC NULLS LAST], [a@0 ASC NULLS LAST, b@2 ASC NULLS LAST]], file_type=csv, has_header=true
+physical_plan DataSourceExec: file_groups={1 group: [[WORKSPACE_ROOT/datafusion/core/tests/data/window_2.csv]]}, projection=[a, a@1 as a_big, b], output_orderings=[[a@0 ASC NULLS LAST, b@2 ASC NULLS LAST], [a_big@1 ASC NULLS LAST]], file_type=csv, has_header=true
 
 query TT
 EXPLAIN
@@ -109,7 +109,7 @@ logical_plan
 01)Sort: multiple_ordered_table.a ASC NULLS LAST, multiple_ordered_table.b ASC NULLS LAST
 02)--Projection: multiple_ordered_table.a, multiple_ordered_table.a AS a_big, multiple_ordered_table.b
 03)----TableScan: multiple_ordered_table projection=[a, b]
-physical_plan DataSourceExec: file_groups={1 group: [[WORKSPACE_ROOT/datafusion/core/tests/data/window_2.csv]]}, projection=[a, a@1 as a_big, b], output_orderings=[[a_big@1 ASC NULLS LAST], [a@0 ASC NULLS LAST, b@2 ASC NULLS LAST]], file_type=csv, has_header=true
+physical_plan DataSourceExec: file_groups={1 group: [[WORKSPACE_ROOT/datafusion/core/tests/data/window_2.csv]]}, projection=[a, a@1 as a_big, b], output_orderings=[[a@0 ASC NULLS LAST, b@2 ASC NULLS LAST], [a_big@1 ASC NULLS LAST]], file_type=csv, has_header=true
 
 
 # test for cast Utf8

datafusion/sqllogictest/test_files/topk.slt

@@ -371,7 +371,7 @@ explain select number, letter, age, number as column4, letter as column5 from pa
 ----
 physical_plan
 01)SortExec: TopK(fetch=3), expr=[number@0 DESC, letter@1 ASC NULLS LAST, age@2 DESC], preserve_partitioning=[false], sort_prefix=[number@0 DESC, letter@1 ASC NULLS LAST]
-02)--DataSourceExec: file_groups={1 group: [[WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/topk/partial_sorted/1.parquet]]}, projection=[number, letter, age, number@0 as column4, letter@1 as column5], output_orderings=[[column4@3 DESC], [number@0 DESC, letter@1 ASC NULLS LAST], [number@0 DESC, column5@4 ASC NULLS LAST]], file_type=parquet, predicate=DynamicFilter [ empty ]
+02)--DataSourceExec: file_groups={1 group: [[WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/topk/partial_sorted/1.parquet]]}, projection=[number, letter, age, number@0 as column4, letter@1 as column5], output_orderings=[[number@0 DESC, letter@1 ASC NULLS LAST], [column4@3 DESC], [number@0 DESC, column5@4 ASC NULLS LAST]], file_type=parquet, predicate=DynamicFilter [ empty ]
 
 # Verify that the sort prefix is correctly computed over normalized, order-maintaining projections (number + 1, number, number + 1, age)
 query TT

datafusion/sqllogictest/test_files/window.slt

@@ -3145,7 +3145,7 @@ physical_plan
 11)--------------------BoundedWindowAggExec: wdw=[sum(annotated_data_finite2.c) PARTITION BY [annotated_data_finite2.a, annotated_data_finite2.b, annotated_data_finite2.d] ORDER BY [annotated_data_finite2.c ASC NULLS LAST] ROWS BETWEEN 2 PRECEDING AND 1 FOLLOWING: Field { "sum(annotated_data_finite2.c) PARTITION BY [annotated_data_finite2.a, annotated_data_finite2.b, annotated_data_finite2.d] ORDER BY [annotated_data_finite2.c ASC NULLS LAST] ROWS BETWEEN 2 PRECEDING AND 1 FOLLOWING": nullable Int64 }, frame: ROWS BETWEEN 2 PRECEDING AND 1 FOLLOWING, sum(annotated_data_finite2.c) PARTITION BY [annotated_data_finite2.a, annotated_data_finite2.b, annotated_data_finite2.d] ORDER BY [annotated_data_finite2.c ASC NULLS LAST] ROWS BETWEEN 5 PRECEDING AND CURRENT ROW: Field { "sum(annotated_data_finite2.c) PARTITION BY [annotated_data_finite2.a, annotated_data_finite2.b, annotated_data_finite2.d] ORDER BY [annotated_data_finite2.c ASC NULLS LAST] ROWS BETWEEN 5 PRECEDING AND CURRENT ROW": nullable Int64 }, frame: ROWS BETWEEN 5 PRECEDING AND CURRENT ROW], mode=[Sorted]
 12)----------------------SortExec: expr=[a@1 ASC NULLS LAST, b@2 ASC NULLS LAST, d@4 ASC NULLS LAST, c@3 ASC NULLS LAST], preserve_partitioning=[false]
 13)------------------------BoundedWindowAggExec: wdw=[sum(annotated_data_finite2.c) PARTITION BY [annotated_data_finite2.a, annotated_data_finite2.b] ORDER BY [annotated_data_finite2.c ASC NULLS LAST] ROWS BETWEEN 2 PRECEDING AND 1 FOLLOWING: Field { "sum(annotated_data_finite2.c) PARTITION BY [annotated_data_finite2.a, annotated_data_finite2.b] ORDER BY [annotated_data_finite2.c ASC NULLS LAST] ROWS BETWEEN 2 PRECEDING AND 1 FOLLOWING": nullable Int64 }, frame: ROWS BETWEEN 2 PRECEDING AND 1 FOLLOWING, sum(annotated_data_finite2.c) PARTITION BY [annotated_data_finite2.a, annotated_data_finite2.b] ORDER BY [annotated_data_finite2.c ASC NULLS LAST] ROWS BETWEEN 5 PRECEDING AND 5 FOLLOWING: Field { "sum(annotated_data_finite2.c) PARTITION BY [annotated_data_finite2.a, annotated_data_finite2.b] ORDER BY [annotated_data_finite2.c ASC NULLS LAST] ROWS BETWEEN 5 PRECEDING AND 5 FOLLOWING": nullable Int64 }, frame: ROWS BETWEEN 5 PRECEDING AND 5 FOLLOWING], mode=[Sorted]
-14)--------------------------DataSourceExec: file_groups={1 group: [[WORKSPACE_ROOT/datafusion/core/tests/data/window_2.csv]]}, projection=[CAST(c@3 AS Int64) as __common_expr_1, a, b, c, d], output_orderings=[[a@1 ASC NULLS LAST, b@2 ASC NULLS LAST, __common_expr_1@0 ASC NULLS LAST], [a@1 ASC NULLS LAST, b@2 ASC NULLS LAST, c@3 ASC NULLS LAST]], file_type=csv, has_header=true
+14)--------------------------DataSourceExec: file_groups={1 group: [[WORKSPACE_ROOT/datafusion/core/tests/data/window_2.csv]]}, projection=[CAST(c@3 AS Int64) as __common_expr_1, a, b, c, d], output_orderings=[[a@1 ASC NULLS LAST, b@2 ASC NULLS LAST, c@3 ASC NULLS LAST], [a@1 ASC NULLS LAST, b@2 ASC NULLS LAST, __common_expr_1@0 ASC NULLS LAST]], file_type=csv, has_header=true
 
 query IIIIIIIIIIIIIII
 SELECT a, b, c,
@@ -3346,7 +3346,7 @@ logical_plan
 physical_plan
 01)ProjectionExec: expr=[sum(annotated_data_infinite2.a) PARTITION BY [annotated_data_infinite2.a, annotated_data_infinite2.b] ORDER BY [annotated_data_infinite2.c ASC NULLS LAST] RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW@3 as sum1, sum(annotated_data_infinite2.a) PARTITION BY [annotated_data_infinite2.b, annotated_data_infinite2.a] ORDER BY [annotated_data_infinite2.c ASC NULLS LAST] RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW@5 as sum2, sum(annotated_data_infinite2.a) PARTITION BY [annotated_data_infinite2.a, annotated_data_infinite2.d] ORDER BY [annotated_data_infinite2.b ASC NULLS LAST] RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW@4 as sum3, sum(annotated_data_infinite2.a) PARTITION BY [annotated_data_infinite2.d] ORDER BY [annotated_data_infinite2.a ASC NULLS LAST] RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW@6 as sum4]
 02)--BoundedWindowAggExec: wdw=[sum(annotated_data_infinite2.a) PARTITION BY [annotated_data_infinite2.d] ORDER BY [annotated_data_infinite2.a ASC NULLS LAST] RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW: Field { "sum(annotated_data_infinite2.a) PARTITION BY [annotated_data_infinite2.d] ORDER BY [annotated_data_infinite2.a ASC NULLS LAST] RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW": nullable Int64 }, frame: RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW], mode=[Linear]
-03)----RepartitionExec: partitioning=Hash([d@2], 2), input_partitions=2, preserve_order=true, sort_exprs=__common_expr_1@0 ASC NULLS LAST, a@1 ASC NULLS LAST
+03)----RepartitionExec: partitioning=Hash([d@2], 2), input_partitions=2, preserve_order=true, sort_exprs=a@1 ASC NULLS LAST, __common_expr_1@0 ASC NULLS LAST
 04)------ProjectionExec: expr=[__common_expr_1@0 as __common_expr_1, a@1 as a, d@4 as d, sum(annotated_data_infinite2.a) PARTITION BY [annotated_data_infinite2.a, annotated_data_infinite2.b] ORDER BY [annotated_data_infinite2.c ASC NULLS LAST] RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW@5 as sum(annotated_data_infinite2.a) PARTITION BY [annotated_data_infinite2.a, annotated_data_infinite2.b] ORDER BY [annotated_data_infinite2.c ASC NULLS LAST] RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW, sum(annotated_data_infinite2.a) PARTITION BY [annotated_data_infinite2.a, annotated_data_infinite2.d] ORDER BY [annotated_data_infinite2.b ASC NULLS LAST] RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW@6 as sum(annotated_data_infinite2.a) PARTITION BY [annotated_data_infinite2.a, annotated_data_infinite2.d] ORDER BY [annotated_data_infinite2.b ASC NULLS LAST] RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW, sum(annotated_data_infinite2.a) PARTITION BY [annotated_data_infinite2.b, annotated_data_infinite2.a] ORDER BY [annotated_data_infinite2.c ASC NULLS LAST] RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW@7 as sum(annotated_data_infinite2.a) PARTITION BY [annotated_data_infinite2.b, annotated_data_infinite2.a] ORDER BY [annotated_data_infinite2.c ASC NULLS LAST] RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW]
 05)--------BoundedWindowAggExec: wdw=[sum(annotated_data_infinite2.a) PARTITION BY [annotated_data_infinite2.b, annotated_data_infinite2.a] ORDER BY [annotated_data_infinite2.c ASC NULLS LAST] RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW: Field { "sum(annotated_data_infinite2.a) PARTITION BY [annotated_data_infinite2.b, annotated_data_infinite2.a] ORDER BY [annotated_data_infinite2.c ASC NULLS LAST] RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW": nullable Int64 }, frame: RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW], mode=[Sorted]
 06)----------RepartitionExec: partitioning=Hash([b@2, a@1], 2), input_partitions=2, preserve_order=true, sort_exprs=a@1 ASC NULLS LAST, b@2 ASC NULLS LAST, c@3 ASC NULLS LAST, __common_expr_1@0 ASC NULLS LAST