mustafasrepo commented on code in PR #11875:
URL: https://github.com/apache/datafusion/pull/11875#discussion_r1711614605
##########
datafusion/core/src/physical_optimizer/enforce_sorting.rs:
##########
@@ -1049,6 +1065,98 @@ mod tests {
Ok(())
}
+ #[tokio::test]
+ async fn test_remove_unnecessary_sort6() -> Result<()> {
+ let schema = create_test_schema()?;
+ let source = memory_exec(&schema);
+ let input = Arc::new(
+ SortExec::new(vec![sort_expr("non_nullable_col", &schema)], source)
+ .with_fetch(Some(2)),
+ );
+ let physical_plan = sort_exec(
+ vec![
+ sort_expr("non_nullable_col", &schema),
+ sort_expr("nullable_col", &schema),
+ ],
+ input,
+ );
+
+ let expected_input = [
+ "SortExec: expr=[non_nullable_col@1 ASC,nullable_col@0 ASC],
preserve_partitioning=[false]",
+ " SortExec: TopK(fetch=2), expr=[non_nullable_col@1 ASC],
preserve_partitioning=[false]",
+ " MemoryExec: partitions=1, partition_sizes=[0]",
+ ];
+ let expected_optimized = [
+ "SortExec: TopK(fetch=2), expr=[non_nullable_col@1
ASC,nullable_col@0 ASC], preserve_partitioning=[false]",
+ " MemoryExec: partitions=1, partition_sizes=[0]",
+ ];
+ assert_optimized!(expected_input, expected_optimized, physical_plan,
true);
+
+ Ok(())
+ }
+
+ #[tokio::test]
+ async fn test_remove_unnecessary_sort7() -> Result<()> {
+ let schema = create_test_schema()?;
+ let source = memory_exec(&schema);
+ let input = Arc::new(SortExec::new(
+ vec![sort_expr("non_nullable_col", &schema)],
+ source,
+ ));
+ let limit = Arc::new(LocalLimitExec::new(input, 2));
+ let physical_plan = sort_exec(
+ vec![
+ sort_expr("non_nullable_col", &schema),
+ sort_expr("nullable_col", &schema),
+ ],
+ limit,
+ );
+
+ let expected_input = [
+ "SortExec: expr=[non_nullable_col@1 ASC,nullable_col@0 ASC],
preserve_partitioning=[false]",
+ " LocalLimitExec: fetch=2",
+ " SortExec: expr=[non_nullable_col@1 ASC],
preserve_partitioning=[false]",
+ " MemoryExec: partitions=1, partition_sizes=[0]",
+ ];
+ let expected_optimized = [
+ "LocalLimitExec: fetch=2",
+ " SortExec: TopK(fetch=2), expr=[non_nullable_col@1
ASC,nullable_col@0 ASC], preserve_partitioning=[false]",
+ " MemoryExec: partitions=1, partition_sizes=[0]",
+ ];
+ assert_optimized!(expected_input, expected_optimized, physical_plan,
true);
+
+ Ok(())
+ }
+
+ #[tokio::test]
+ async fn test_do_not_pushdown_through_limit() -> Result<()> {
+ let schema = create_test_schema()?;
+ let source = memory_exec(&schema);
+ // let input = sort_exec(vec![sort_expr("non_nullable_col", &schema)],
source);
+ let input = Arc::new(SortExec::new(
+ vec![sort_expr("non_nullable_col", &schema)],
+ source,
+ ));
+ let limit = Arc::new(GlobalLimitExec::new(input, 0, Some(5))) as _;
+ let physical_plan = sort_exec(vec![sort_expr("nullable_col",
&schema)], limit);
+
+ let expected_input = [
+ "SortExec: expr=[nullable_col@0 ASC],
preserve_partitioning=[false]",
+ " GlobalLimitExec: skip=0, fetch=5",
+ " SortExec: expr=[non_nullable_col@1 ASC],
preserve_partitioning=[false]",
+ " MemoryExec: partitions=1, partition_sizes=[0]",
+ ];
+ let expected_optimized = [
Review Comment:
The reason is that at the input plan:
```
"SortExec: expr=[nullable_col@0 ASC],
preserve_partitioning=[false]",
" GlobalLimitExec: skip=0, fetch=5",
" SortExec: expr=[non_nullable_col@1 ASC],
preserve_partitioning=[false]",
" MemoryExec: partitions=1, partition_sizes=[0]",
```
The expressions in the bottom `SortExec` (which is `non_nullable_col@1 ASC`)
and top `SortExec` (which is `nullable_col@0 ASC`) are different and not
compatible. This situation might arise with following kind of query
```
SELECT * FROM (SELECT * FROM table ORDER BY a ASC LIMIT 10) ORDER BY b ASC;
```
User might want to take subset of a table with a given order. Then might
want to work on that table . Hence pushing down sort in this case would produce
different result than the original query.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
