TPC-DS-OBT Q2 — obt_channel_distribution

TPC-DS-OBT benchmark · all TPC-DS-OBT queries

Count sales rows by channel

Representative SQL

Rendered for DataFusion with default parameters.

WITH wscs AS (
  SELECT
    sold_date_sk,
    sales_price
  FROM (
    SELECT
      obt.sold_date_sk AS sold_date_sk,
      obt.ext_sales_price AS sales_price
    FROM tpcds_sales_returns_obt AS obt
    WHERE
      channel = 'web'
    UNION ALL
    SELECT
      obt.sold_date_sk AS sold_date_sk,
      obt.ext_sales_price AS sales_price
    FROM tpcds_sales_returns_obt AS obt
    WHERE
      channel = 'catalog'
  )
), wswscs AS (
  SELECT
    obt.sold_date_d_week_seq AS d_week_seq,
    obt.sold_date_d_year AS d_year,
    SUM(
      CASE
        WHEN (
          obt.sold_date_d_day_name = 'Sunday'
        )
        THEN obt.sales_price
        ELSE NULL
      END
    ) AS sun_sales,
    SUM(
      CASE
        WHEN (
          obt.sold_date_d_day_name = 'Monday'
        )
        THEN obt.sales_price
        ELSE NULL
      END
    ) AS mon_sales,
    SUM(
      CASE
        WHEN (
          obt.sold_date_d_day_name = 'Tuesday'
        )
        THEN obt.sales_price
        ELSE NULL
      END
    ) AS tue_sales,
    SUM(
      CASE
        WHEN (
          obt.sold_date_d_day_name = 'Wednesday'
        )
        THEN obt.sales_price
        ELSE NULL
      END
    ) AS wed_sales,
    SUM(
      CASE
        WHEN (
          obt.sold_date_d_day_name = 'Thursday'
        )
        THEN obt.sales_price
        ELSE NULL
      END
    ) AS thu_sales,
    SUM(
      CASE
        WHEN (
          obt.sold_date_d_day_name = 'Friday'
        )
        THEN obt.sales_price
        ELSE NULL
      END
    ) AS fri_sales,
    SUM(
      CASE
        WHEN (
          obt.sold_date_d_day_name = 'Saturday'
        )
        THEN obt.sales_price
        ELSE NULL
      END
    ) AS sat_sales
  FROM tpcds_sales_returns_obt AS obt
  WHERE
    obt.sold_date_d_date_sk = obt.sold_date_sk AND channel IN ('catalog', 'web')
  GROUP BY
    d_week_seq,
    d_year
)
SELECT
  d_week_seq1,
  ROUND(CAST(sun_sales1 AS DOUBLE PRECISION) / sun_sales2, 2),
  ROUND(CAST(mon_sales1 AS DOUBLE PRECISION) / mon_sales2, 2),
  ROUND(CAST(tue_sales1 AS DOUBLE PRECISION) / tue_sales2, 2),
  ROUND(CAST(wed_sales1 AS DOUBLE PRECISION) / wed_sales2, 2),
  ROUND(CAST(thu_sales1 AS DOUBLE PRECISION) / thu_sales2, 2),
  ROUND(CAST(fri_sales1 AS DOUBLE PRECISION) / fri_sales2, 2),
  ROUND(CAST(sat_sales1 AS DOUBLE PRECISION) / sat_sales2, 2)
FROM (
  SELECT
    wswscs.d_week_seq AS d_week_seq1,
    sun_sales AS sun_sales1,
    mon_sales AS mon_sales1,
    tue_sales AS tue_sales1,
    wed_sales AS wed_sales1,
    thu_sales AS thu_sales1,
    fri_sales AS fri_sales1,
    sat_sales AS sat_sales1
  FROM wswscs AS wswscs
  WHERE
    (
      d_year = 1998
    )
) AS y, (
  SELECT
    wswscs.d_week_seq AS d_week_seq2,
    sun_sales AS sun_sales2,
    mon_sales AS mon_sales2,
    tue_sales AS tue_sales2,
    wed_sales AS wed_sales2,
    thu_sales AS thu_sales2,
    fri_sales AS fri_sales2,
    sat_sales AS sat_sales2
  FROM wswscs AS wswscs
  WHERE
    (
      d_year = 1998 + 1
    )
) AS z
WHERE
  d_week_seq1 = d_week_seq2 - 53
ORDER BY
  d_week_seq1

Representative DataFrame

Expression family (Polars, PySpark, DataFusion), default parameters.

def impl(ctx: DataFrameContext) -> Any:
        table = ctx.get_table(OBT_TABLE)
        for column, op, value in spec.filters:
            table = table.filter(_FILTER_OPS[op](ctx.col(column), ctx.lit(value)))
        aggregates = [getattr(ctx.col(column), func)().alias(alias) for alias, column, func in spec.aggregations]
        if spec.group_keys:
            result = table.group_by(*spec.group_keys).agg(*aggregates)
        else:
            result = table.select(*aggregates)
        if spec.sort_keys:
            descending = list(spec.descending) if spec.descending else [False] * len(spec.sort_keys)
            result = result.sort(list(spec.sort_keys), descending=descending)
        if spec.limit:
            result = result.limit(spec.limit)
        return result

Get the exact query for your run

The blocks above are representative. To get the precise statement a specific platform runs at a specific scale, substitute your own values:

from benchbox import TPCDSOBT

bench = TPCDSOBT(scale_factor=1.0)            # your scale factor
queries = bench.get_queries(dialect="datafusion")   # your platform's dialect; drop dialect= for native SQL
print(queries['2'])

For the DataFrame form:

from benchbox.core.query_catalog import get_dataframe_render

print(get_dataframe_render('tpcds_obt', '2', family="expression").source)  # or family="pandas"

Source

Query text: benchbox/core/tpcds_obt/queries.py