Skip to content

Regression in TPC-DS Q75 #11385

@gruuya

Description

@gruuya

Describe the bug

While logical and physical planning work fine for TPC-DS Q75, as of v39.0.0 the execution throws the following error

Error: External(External(ArrowError(InvalidArgumentError("RowConverter column schema mismatch, expected Int32 got Int64"), None)))

To Reproduce

Generate TPC-DS dataset, or use the one attached (for SF=0.01): tpcds_sf0_01.zip

Register parquet files as tables

...
ctx.register_parquet(
        "catalog_sales",
        "/Users/gruuya/Splitgraph/tpcds_sf0_01/catalog_sales.parquet",
        ParquetReadOptions::default(),
    )
        .await?;
...

And run the query

let df = ctx.sql("...").await?;
df.show().await?;

It will error out with the above error.

Expected behavior

The query shouldn't error out, and should return something like

+-----------+------+------------+------------+---------------+---------------+-------------+-------------+----------------+----------------+
| prev_year | year | i_brand_id | i_class_id | i_category_id | i_manufact_id | prev_yr_cnt | curr_yr_cnt | sales_cnt_diff | sales_amt_diff |
+-----------+------+------------+------------+---------------+---------------+-------------+-------------+----------------+----------------+
| 2001      | 2002 | 8015002    | 15         | 8             | 167           | 5346        | 3418        | -1928          | -94072.53      |
| 2001      | 2002 | 10016012   | 13         | 8             | 46            | 5125        | 3338        | -1787          | -81828.03      |
+-----------+------+------------+------------+---------------+---------------+-------------+-------------+----------------+----------------+

Additional context

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions