Skip to content
This repository has been archived by the owner on Oct 2, 2024. It is now read-only.

Example produces incorrect results #44

Open
andygrove opened this issue Nov 19, 2023 · 0 comments
Open

Example produces incorrect results #44

andygrove opened this issue Nov 19, 2023 · 0 comments

Comments

@andygrove
Copy link
Collaborator

The tips.py runs a grouping query but the result set has duplicate keys:

0    Male     No  0.160669
1  Female    Yes  0.182150
      sex smoker   tip_pct
0  Female     No  0.156921
1    Male    Yes  0.152771

Debug logging:

Planning select sex, smoker, avg(tip/total_bill) as tip_pct from tips group by sex, smoker
Query stage #0:
RayShuffleWriterExec(stage_id=0, output_partitioning=Hash([Column { name: "sex", index: 0 }, Column { name: "smoker", index: 1 }], 2))
  AggregateExec: mode=Partial, gby=[sex@2 as sex, smoker@3 as smoker], aggr=[AVG(tips.tip / tips.total_bill)]
    ParquetExec: limit=None, partitions={1 group: [[home/andy/git/personal/ray-sql/examples/tips.parquet]]}, projection=[total_bill, tip, sex, smoker]

Query stage #1:
CoalescePartitionsExec
  ProjectionExec: expr=[sex@0 as sex, smoker@1 as smoker, AVG(tips.tip / tips.total_bill)@2 as tip_pct]
    AggregateExec: mode=FinalPartitioned, gby=[sex@0 as sex, smoker@1 as smoker], aggr=[AVG(tips.tip / tips.total_bill)]
      CoalesceBatchesExec: target_batch_size=16384
        RayShuffleReaderExec(stage_id=0, input_partitioning=Hash([Column { name: "sex", index: 0 }, Column { name: "smoker", index: 1 }], 2))

Scheduling query stage #0 with 1 input partitions and 2 output partitions
Forcing reduce stage concurrency from 2 to 1
Scheduling query stage #1 with 1 input partitions and 1 output partitions
      sex smoker   tip_pct
0    Male     No  0.160669
1  Female    Yes  0.182150
      sex smoker   tip_pct
0  Female     No  0.156921
1    Male    Yes  0.152771
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant