PIVOT 子句

适用于： Databricks SQL Databricks Runtime

通过将指定列列表的唯一值轮换到单独的列中，转换 table_reference 的行。

语法

table_reference PIVOT ( { aggregate_expression [ [ AS ] agg_column_alias ] } [, ...]
    FOR column_list IN ( expression_list ) )

column_list
 { column_name |
   ( column_name [, ...] ) }

expression_list
 { expression [ AS ] [ column_alias ] |
   { ( expression [, ...] ) [ AS ] [ column_alias] } [, ...] ) }

参数

• table_reference

  标识 PIVOT 操作的主题。

• aggregate_expression

  任何类型的表达式，其中的所有列引用 table_reference 都是聚合函数的参数。

• agg_column_alias

  聚合结果的可选别名。 如果未指定别名，PIVOT 会根据 aggregate_expression 生成别名。

• column_list

  要轮换的列集。

  • column_name

    来自 table_reference 的列。

• expression_list

  将 column_list 中的值映射到列别名。

  • expression

    一种文本表达式，它与相应的 column_name 都使用一个最不常见的类型。

    每个元组中表达式的数量必须与 column_list 中 column_names 的数量匹配。

  • column_alias

    一个可选别名，用于指定生成的列的名称。 如果未指定别名，PIVOT 会根据 expression 生成别名。

结果

采用以下形式的临时表：

• table_reference 的中间结果集中尚未在任何 aggregate_expression 或 column_list 中指定的所有列。

  这些列是分组列。

• 对于每个 expression 元组和 aggregate_expression 组合，PIVOT 都会生成一列。 类型为 aggregate_expression 的类型。

  如果只有一个 aggregate_expression，则使用 column_alias 命名该列。 否则，使用 column_alias_agg_column_alias 命名。

  每个单元格中的值是使用 FILTER ( WHERE column_list IN (expression, ...) 的 aggregation_expression 的结果。

示例

-- A very basic PIVOT
-- Given a table with sales by quarter, return a table that returns sales across quarters per year.
> CREATE TEMP VIEW sales(year, quarter, region, sales) AS
   VALUES (2018, 1, 'east', 100),
          (2018, 2, 'east',  20),
          (2018, 3, 'east',  40),
          (2018, 4, 'east',  40),
          (2019, 1, 'east', 120),
          (2019, 2, 'east', 110),
          (2019, 3, 'east',  80),
          (2019, 4, 'east',  60),
          (2018, 1, 'west', 105),
          (2018, 2, 'west',  25),
          (2018, 3, 'west',  45),
          (2018, 4, 'west',  45),
          (2019, 1, 'west', 125),
          (2019, 2, 'west', 115),
          (2019, 3, 'west',  85),
          (2019, 4, 'west',  65);

> SELECT year, region, q1, q2, q3, q4
  FROM sales
  PIVOT (sum(sales) AS sales
    FOR quarter
    IN (1 AS q1, 2 AS q2, 3 AS q3, 4 AS q4));
 2018  east  100  20  40  40
 2019  east  120  110  80  60
 2018  west  105  25  45  45
 2019  west  125  115  85  65

-- The same query written without PIVOT
> SELECT year, region,
         sum(sales) FILTER(WHERE quarter = 1) AS q1,
         sum(sales) FILTER(WHERE quarter = 2) AS q2,
         sum(sales) FILTER(WHERE quarter = 3) AS q2,
         sum(sales) FILTER(WHERE quarter = 4) AS q4
  FROM sales
  GROUP BY year, region;
 2018  east  100  20  40  40
 2019  east  120  110  80  60
 2018  west  105  25  45  45
 2019  west  125  115  85  65

-- Also PIVOT on region
> SELECT year, q1_east, q1_west, q2_east, q2_west, q3_east, q3_west, q4_east, q4_west
    FROM sales
    PIVOT (sum(sales) AS sales
      FOR (quarter, region)
      IN ((1, 'east') AS q1_east, (1, 'west') AS q1_west, (2, 'east') AS q2_east, (2, 'west') AS q2_west,
          (3, 'east') AS q3_east, (3, 'west') AS q3_west, (4, 'east') AS q4_east, (4, 'west') AS q4_west));
 2018  100  105  20  25  40  45  40  45
 2019  120  125  110  115  80  85  60  65

-- The same query written without PIVOT
> SELECT year,
    sum(sales) FILTER(WHERE (quarter, region) = (1, 'east')) AS q1_east,
    sum(sales) FILTER(WHERE (quarter, region) = (1, 'west')) AS q1_west,
    sum(sales) FILTER(WHERE (quarter, region) = (2, 'east')) AS q2_east,
    sum(sales) FILTER(WHERE (quarter, region) = (2, 'west')) AS q2_west,
    sum(sales) FILTER(WHERE (quarter, region) = (3, 'east')) AS q3_east,
    sum(sales) FILTER(WHERE (quarter, region) = (3, 'west')) AS q3_west,
    sum(sales) FILTER(WHERE (quarter, region) = (4, 'east')) AS q4_east,
    sum(sales) FILTER(WHERE (quarter, region) = (4, 'west')) AS q4_west
    FROM sales
    GROUP BY year, region;
 2018  100  105  20  25  40  45  40  45
 2019  120  125  110  115  80  85  60  65

-- To aggregate across regions the column must be removed from the input.
> SELECT year, q1, q2, q3, q4
  FROM (SELECT year, quarter, sales FROM sales) AS s
  PIVOT (sum(sales) AS sales
    FOR quarter
    IN (1 AS q1, 2 AS q2, 3 AS q3, 4 AS q4));
  2018  205  45  85  85
  2019  245  225  165  125

-- The same query without PIVOT
> SELECT year,
    sum(sales) FILTER(WHERE quarter = 1) AS q1,
    sum(sales) FILTER(WHERE quarter = 2) AS q2,
    sum(sales) FILTER(WHERE quarter = 3) AS q3,
    sum(sales) FILTER(WHERE quarter = 4) AS q4
    FROM sales
    GROUP BY year;

-- A PIVOT with multiple aggregations
> SELECT year, q1_total, q1_avg, q2_total, q2_avg, q3_total, q3_avg, q4_total, q4_avg
    FROM (SELECT year, quarter, sales FROM sales) AS s
    PIVOT (sum(sales) AS total, avg(sales) AS avg
      FOR quarter
      IN (1 AS q1, 2 AS q2, 3 AS q3, 4 AS q4));
 2018  205  102.5  45  22.5  85  42.5  85  42.5
 2019  245  122.5  225  112.5  165  82.5  125  62.5

-- The same query without PIVOT
> SELECT year,
         sum(sales) FILTER(WHERE quarter = 1) AS q1_total,
         avg(sales) FILTER(WHERE quarter = 1) AS q1_avg,
         sum(sales) FILTER(WHERE quarter = 2) AS q2_total,
         avg(sales) FILTER(WHERE quarter = 2) AS q2_avg,
         sum(sales) FILTER(WHERE quarter = 3) AS q3_total,
         avg(sales) FILTER(WHERE quarter = 3) AS q3_avg,
         sum(sales) FILTER(WHERE quarter = 4) AS q4_total,
         avg(sales) FILTER(WHERE quarter = 4) AS q4_avg
    FROM sales
    GROUP BY year;

> CREATE TEMP VIEW person (id, name, age, class, address) AS
    VALUES (100, 'John', 30, 1, 'Street 1'),
           (200, 'Mary', NULL, 1, 'Street 2'),
           (300, 'Mike', 80, 3, 'Street 3'),
           (400, 'Dan', 50, 4, 'Street 4');
 2018  205  102.5  45  22.5  85  42.5  85  42.5
 2019  245  122.5  225  112.5  165  82.5  125  62.5

相关文章

• SELECT
• 聚合函数
• UNPIVOT 子句