fix: window function range offset should be long instead of int #733
Conversation
| val df = | ||
| Seq((1L, "1"), (1L, "1"), (2147483650L, "1"), (3L, "2"), (2L, "1"), (2147483650L, "2")) | ||
| .toDF("key", "value") | ||
|
|
||
| checkAnswer( | ||
| df.select( | ||
| $"key", | ||
| count("key").over( | ||
| sum("key").over( |
There was a problem hiding this comment.
So this test fails before this change?
There was a problem hiding this comment.
We currently don't support count so change to sum. Previously, even if the aggregate function is not supported, it still goes to the window frame code. In this PR, I return None directly if aggregate function is not supported.
| throw new IllegalArgumentException( | ||
| "Unsupported data type for window function row/range offset") |
There was a problem hiding this comment.
You will fail the query directly instead of falling back to Spark.
There was a problem hiding this comment.
changed to return None to fall back to Spark
huaxingao
force-pushed
the
range_offset
branch
3 times, most recently
from
2e03816
to
1f8a29e
Compare
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #733 +/- ##
=============================================
- Coverage 55.16% 34.16% -21.00%
- Complexity 857 881 +24
=============================================
Files 109 112 +3
Lines 10542 42949 +32407
Branches 2010 9491 +7481
=============================================
+ Hits 5815 14673 +8858
- Misses 3714 25274 +21560
- Partials 1013 3002 +1989 ☔ View full report in Codecov by Sentry. |
huaxingao
force-pushed
the
range_offset
branch
from
2777d18
to
0d84fd2
Compare
| WindowFrameUnits::Rows => { | ||
| WindowFrameBound::Preceding(ScalarValue::UInt64(None)) | ||
| } | ||
| WindowFrameUnits::Range | WindowFrameUnits::Groups => { |
| if (partitionColumnNames.toSet != orderColumnNames.toSet) { | ||
| withInfo(op, "Partitioning and sorting specifications do not match") | ||
| return false | ||
| } |
There was a problem hiding this comment.
Maybe check the partition column and order column one by one instead of a set? I'm not sure if (PARTITION BY k, v ORDER BY v, k) work.
Which issue does this PR close?
Closes #.
Rationale for this change
What changes are included in this PR?
How are these changes tested?