rluvaton opened a new issue, #23902:
URL: https://github.com/apache/datafusion/issues/23902
### Describe the bug
Sort should not be removed when doing addition (and other operators that
marked as kept ordering in binary expression code) on 2 sorted columns since
they can overflow and the default behavior is `wrapping` so the sort will be
broken
### To Reproduce
Have this data file:
```csv
a,b
1,1
2,3
200,10
255,10
```
```slt
# Adding two ASC-ordered unsigned columns is only monotonically increasing
while
# the sum does not overflow. `a + b` on UInt8 uses wrapping arithmetic, so
the
# last row (255 + 10) wraps to 9 and breaks the ordering. A sort on the sum
# therefore must NOT be optimized away just because both inputs are ASC.
statement ok
set datafusion.execution.target_partitions = 1;
statement ok
CREATE EXTERNAL TABLE ordered_u8 (
a TINYINT UNSIGNED NOT NULL,
b TINYINT UNSIGNED NOT NULL
)
STORED AS CSV
LOCATION 'data/uint8_overflow_order.csv'
OPTIONS ('format.has_header' 'true')
WITH ORDER (a ASC)
WITH ORDER (b ASC);
query TT
EXPLAIN SELECT (a + b) AS result FROM ordered_u8 ORDER BY result ASC;
----
logical_plan
01)Sort: result ASC NULLS LAST
02)--Projection: ordered_u8.a + ordered_u8.b AS result
03)----TableScan: ordered_u8 projection=[a, b]
physical_plan DataSourceExec: file_groups={1 group:
[[WORKSPACE_ROOT/datafusion/sqllogictest/data/uint8_overflow_order.csv]]},
projection=[a@0 + b@1 as result], file_type=csv, has_header=true
query I
SELECT (a + b) AS result FROM ordered_u8 ORDER BY result ASC;
----
2
5
210
9
statement ok
DROP TABLE ordered_u8;
# The SLT runner sets `target_partitions` to 4 instead of using the default,
so
# reset it explicitly.
statement ok
set datafusion.execution.target_partitions = 4;
```
### Expected behavior
The SortExec should not be removed as it break the ordering when overflow
### Additional context
_No response_
--
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]
