SQL/Bigquery：将行组合转换为列，保留所有对答案

【问题标题】：SQL/Bigquery: Pivot combinations of rows into columns, keeping all pairsSQL/Bigquery：将行组合转换为列，保留所有对
【发布时间】：2021-08-04 18:11:43
【问题描述】：

假设我有一张这样的桌子：

| id |  brand |   fuel | mpg |
|:--:|:------:|:------:|:---:|
| 1  | ford   | diesel | 14  |
| 1  | ford   | gas    | 20  |
| 1  | toyota | diesel | 30  |
| 1  | toyota | gas    | 35  |

我希望旋转列，结果如下：

| id |  ford  | toyota | toyota_mpg | ford_mpg |
|:--:|:------:|--------|:----------:|:--------:|
| 1  | diesel | diesel | 30         | 14       |
| 1  | gas    | gas    | 35         | 20       |
| 1  | diesel | gas    | 35         | 14       |
| 1  | gas    | diesel | 30         | 20       |

到目前为止，我已经

SELECT id,
         MAX(CASE WHEN end_use = 'ford' THEN fuel ELSE NULL END) ford,
         SUM(CASE WHEN end_use = 'ford' THEN mpg ELSE NULL END) ford_mpg,

         MAX(CASE WHEN end_use = 'toyota' THEN fuel ELSE NULL END) toyota,
         SUM(CASE WHEN end_use = 'toyota' THEN mpg ELSE NULL END) toyota_mpg,
       FROM table GROUP BY id, fuel

结果如下，当燃料对齐时给我正确的结果：

| id |  ford  | toyota | toyota_mpg | ford_mpg |
|:--:|:------:|--------|:----------:|:--------:|
| 1  | diesel | diesel | 30         | 14       |
| 1  | gas    | gas    | 35         | 20       |

但我无法获得燃料组合（它们不匹配的地方）。

【问题讨论】：

标签： sql google-bigquery pivot case

【解决方案1】：

试试下面

select id, 
  t1.fuel ford, 
  t2.fuel toyota, 
  t1.mpg ford_mpg, 
  t2.mpg toyota_mpg
from data t1
join data t2
using (id)
where t1.brand < t2.brand

如果应用于您问题中的样本数据 - 输出是

【讨论】：

【解决方案2】：

您可以从大查询中探索pivot operator。虽然我不确定您在 id 级别上找出 sum/avg/max/min 的用例！

WITH
  result AS (
  SELECT
    1 AS id,
    'ford' AS brand,
    'diesel' AS fuel,
    14 AS mpg
  UNION ALL
  SELECT
    1 AS id,
    'ford' AS brand,
    'gas' AS fuel,
    20 AS mpg
  UNION ALL
  SELECT
    1 AS id,
    'toyota' AS brand,
    'diesel' AS fuel,
    30 AS mpg
  UNION ALL
  SELECT
    1 AS id,
    'toyota' AS brand,
    'gas' AS fuel,
    35 AS mpg ),
  pivot_result AS (
  SELECT
    id,
    ford,
    toyota,
    mpg_ford,
    mpg_toyota
  FROM (
    SELECT
      *
    FROM (
      SELECT
        id,
        fuel,
        brand brand_,
        brand,
        mpg
      FROM
        result ) PIVOT ( AVG(mpg) mpg FOR brand IN ('ford',
          'toyota')) ) PIVOT (MAX(fuel) FOR brand_ IN ('ford',
        'toyota')) )
SELECT
  f.id,
  f.ford,
  t.toyota,
  t.mpg_toyota,
  f.mpg_ford
FROM
  pivot_result f
INNER JOIN (
  SELECT
    id,
    toyota,
    mpg_toyota
  FROM
    pivot_result) t
ON
  t.id = f.id
WHERE
  (f.ford IS NOT NULL
    AND f.mpg_ford IS NOT NULL
    AND t.toyota IS NOT NULL
    AND t.mpg_toyota IS NOT NULL)
GROUP BY
  1,
  2,
  3,
  4,
  5

【讨论】：