如何在 BigQuery BI 引擎中复制 row_number

Question

我有这个在 BigQuery 中运行良好的查询

WITH
  tt AS (
  SELECT
    StationName,
    SUM(Mwh) AS Mwh
  FROM
    `datastudio.today_view_MT`
  GROUP BY
    1)
SELECT
  StationName,
  Mwh,
  ROW_NUMBER() OVER(ORDER BY Mwh DESC ) AS rnk
FROM
  tt

但截至 2020 年 8 月，BI 引擎不加速分析功能，如何使用自联接或类似的方式重写相同的查询，以便 BI 引擎加速。

我的真实用例，查询使用Data Studio的参数，所以我无法具体化结果，有三个参数，所以有很多不同的组合。

Answer 1

根据要求，这似乎可行：

WITH
  tt AS (
  SELECT
    "A" AS fixed_value,
    StationName,
    SUM(Mwh) AS Mwh
  FROM
    `datastudio.today_view_MT`
  GROUP BY
    1, 2)
SELECT
  tt.StationName,
  tt.Mwh,
  COUNT(*) AS row_number
FROM
  tt
JOIN
  tt tt2
ON
  tt2.fixed_value = tt.fixed_value
  AND tt2.Mwh <= tt.Mwh
GROUP BY
  tt.fixed_value,
  tt.Mwh,
  tt.StationName
ORDER BY
  row_number

请注意，您可能有重复的行号

Answer 2

我尝试过使用连接，但我从来没有得到正确的结果，所以这是我的开箱即用解决方案：

WITH
  tt AS (
  SELECT
    StationName,
    SUM(Mwh) AS Mwh
  FROM
    `datastudio.today_view_MT`
  GROUP BY
    1),
  structs AS (
  SELECT
    STRUCT(StationName,
      Mwh) AS station_struct
  FROM
    tt ),
  combinations AS (
  SELECT
    ARRAY_AGG(station_struct) AS stations,
    GENERATE_ARRAY(1,(
      SELECT
        COUNT(*)
      FROM
        tt)) AS numbers
  FROM
    structs )
SELECT
  station.StationName,
  station.Mwh,
  numbers[
OFFSET
  (stations_offset)] AS number
FROM
  combinations,
  UNNEST(stations) AS station
WITH
OFFSET
  AS stations_offset

现在的问题是你能在 BI 引擎中使用 generate_arrays 和 offsets 吗？