Open scsmithr opened 1 month ago
First 1024 rows are correct, so I imagine we're just draining the ORDER BY operator flakily.
Pipeline PipelineId(8)
Partition 0
[Op] Read Emitted Elapsed (micro) Explain
[ 0] 0 1 4 RoundRobinRepartition
[ 1] 1 1 28 Project (projections = [*(@0, CAST(0.0001 TO Decimal64(18,4)))])
[ 2] 2 2541 65235 NestedLoopJoin (filter = and(>(@1, @3)), join_type = INNER)
[ 3] 2541 2541 291 Project (projections = [@2, @0, >(@1, @3)])
[ 4] 2541 2541 6120 LocalSort
[ 5] 2541 0 925 MergeSorted
Partition 1
[Op] Read Emitted Elapsed (micro) Explain
[ 0] 0 0 1 RoundRobinRepartition
[ 1] 0 0 0 Project (projections = [*(@0, CAST(0.0001 TO Decimal64(18,4)))])
[ 2] 0 0 1 NestedLoopJoin (filter = and(>(@1, @3)), join_type = INNER)
[ 3] 0 0 0 Project (projections = [@2, @0, >(@1, @3)])
[ 4] 0 0 17 LocalSort
[ 5] 0 0 2 MergeSorted
In LocalSort
Unknown if I actually fixed this
Not fixed
https://github.com/GlareDB/rayexec/actions/runs/11037391960/job/30658159171?pr=248
Different results