Question

假设我有一些长格式的表格，如下所示：

CREATE TEMP TABLE tmp (
  id int,  
  value varchar,
  id2 int,
  key int);

INSERT INTO tmp VALUES 
  (1, 87.1, 1444, 102),
  (2, 144.9, 13921, 3), 
  (3, 'A032333', 13921, 7),
  (4, 88.9, 13921, 102),
  (5, 'JDS-SJDDD', 13921, 101),
  (6, 90000, 13921, 140),
  (7, 101.1, 33113, 133),
  (8, 'KKL-KKIDD', 33113, 101),
  (9, 0, 33113, 239),
  (10, 933.1, 33113, 250);

我需要将此数据转换为宽格式，如下所示：

CREATE TEMP TABLE tmp2 (
  id2 integer,
  k3 numeric,
  k7 varchar,
  k101 varchar,
  k102 numeric,
  k133 numeric,
  k140 int,
  k239 int,
  k250 numeric);

INSERT INTO tmp2 VALUES 
  (1444, NULL, NULL, NULL, 87.1, NULL, NULL, NULL, NULL),
  (13921, 144.9, 'A032333', 'JDS-SJDDD', 88.9, NULL, 90000, NULL, NULL), 
  (33113, NULL, NULL, 'KKL-KKIDD', NULL, 101.1, NULL, 0, 933.1);

我尝试使用多个CASE WHEN：

SELECT id2, 
    CASE WHEN key = 3 THEN value END AS a3,
    CASE WHEN key = 7 THEN value END AS a7,
    CASE WHEN key = 101 THEN value END AS a101,
    CASE WHEN key = 102 THEN value END AS a102,
    CASE WHEN key = 133 THEN value END AS a133,
    CASE WHEN key = 140 THEN value END AS a140,
    CASE WHEN key = 239 THEN value END AS a239,
    CASE WHEN key = 250 THEN value END AS a250
FROM tmp;

但是，输出为id2保留了多行，而每个值仅保留一行就足够了。如何调整？我在考虑类似GROUP BY + COALESCE的问题，但是COALESCE在整个行中寻找值，因此我需要针对列返回第一个非空值。此外，这种方法似乎非常麻烦，因为我的原始数据将包含大约2000个结果列，因此用CASE WHEN指定每个列将产生大量代码。有捷径吗？如果没有，如何实现？

Answer 1

您需要group by id2并进行汇总：

SELECT id2, 
    max(CASE WHEN key = 3 THEN value END) AS a3,
    max(CASE WHEN key = 7 THEN value END) AS a7,
    max(CASE WHEN key = 101 THEN value END) AS a101,
    max(CASE WHEN key = 102 THEN value END) AS a102,
    max(CASE WHEN key = 133 THEN value END) AS a133,
    max(CASE WHEN key = 140 THEN value END) AS a140,
    max(CASE WHEN key = 239 THEN value END) AS a239,
    max(CASE WHEN key = 250 THEN value END) AS a250
FROM tmp
group by id2
order by id2

这将适用于您的示例数据。
请参见demo。
结果：

>   id2 | a3    | a7      | a101      | a102 | a133  | a140  | a239 | a250 
> ----: | :---- | :------ | :-------- | :--- | :---- | :---- | :--- | :----
>  1444 | null  | null    | null      | 87.1 | null  | null  | null | null 
> 13921 | 144.9 | A032333 | JDS-SJDDD | 88.9 | null  | 90000 | null | null 
> 33113 | null  | null    | KKL-KKIDD | null | 101.1 | null  | 0    | 933.1

Answer 2

Postgres支持filter关键字进行条件聚合，因此我建议：

SELECT id2, 
       MAX(value) FILTER (WHERE key = 3) AS a3,
       MAX(value) FILTER (WHERE key = 7) AS a7,
       MAX(value) FILTER (WHERE key = 101) AS a101,
       MAX(value) FILTER (WHERE key = 102) AS a102,
       MAX(value) FILTER (WHERE key = 133) AS a133,
       MAX(value) FILTER (WHERE key = 140) AS a140,
       MAX(value) FILTER (WHERE key = 239) AS a239,
       MAX(value) FILTER (WHERE key = 250) AS a250
FROM tmp
GROUP BY id2;

但是关键思想是GROUP BY。

PostgreSQL-从长格式到宽格式

2 个答案: