在 Oracle 中，如何将数据透视到多个列中？

Question

我有以下测试 table，我需要得到以下结果 table。我试过一个枢轴，但它似乎只在你有一个描述符和一个值时才起作用。在这种情况下，我们只有一个值。有些 site_id 可能只有 1 个值，但有些最多可以有 3 个其他 site_id_alt。有超过 2800 个独特的 site_id，因此将它们手动编码为枢轴函数是一种选择。

我在想某种分区可能会起作用，但不知道从哪里开始。

drop table site_id_cd_test;
create table site_id_cd_test
 (
  site_id_alt VARCHAR(30),
  site_id VARCHAR(30)
);

insert into site_id_cd_test values('1A1-071', '1A1-071');
insert into site_id_cd_test values('1A1-071O', '1A1-071');
insert into site_id_cd_test values('030256', '1A1-071');
insert into site_id_cd_test values('04268384', '1A1-071');
insert into site_id_cd_test values('04268383', '1A1-072');


drop table site_id_cd_result;
create table site_id_cd_result
 (
  site_id VARCHAR(30),
  site_id_alt_01 VARCHAR(30),
  site_id_alt_02 VARCHAR(30),
  site_id_alt_03 VARCHAR(30)
);

insert into site_id_cd_result values('1A1-071', '1A1-071O', '030256', '04268384' );
insert into site_id_cd_result values('1A1-072', '04268383', NULL, NULL);

编辑9_19_19：

在@MT0 的反馈之后，我们完成了大约 90% 的工作（它很好地以行号为中心）。当 site_id 与 site_id_alt 匹配时，它会将它们添加为新行（如下面的示例 table 所示）。理想情况下，我们希望匹配的 site_id 和 site_id_alt 与其他值位于同一行。

下面例子的总结tables:

site_id_cd_result 是结果 table 当运行行号旋转

site_id_cd_result_02 是使用 site_id <> site_id_alt

时的结果 table

drop table site_id_cd_result;

create table site_id_cd_result
 (
  site_id VARCHAR(30),
  site_id_alt_01 VARCHAR(30),
  site_id_alt_02 VARCHAR(30),
  site_id_alt_03 VARCHAR(30),
 site_id_alt_04 VARCHAR(30)

);

insert into site_id_cd_result values('1A1-071', '1A1-071', NULL, NULL, NULL);
insert into site_id_cd_result values('1A1-071', NULL, '040777', '04253626', '1A1-071O');
insert into site_id_cd_result values('1A1-072', '04268383', '123546', NULL, NULL);


drop table site_id_cd_result_02;

create table site_id_cd_result_02
 (
  site_id VARCHAR(30),
  site_id_alt_01 VARCHAR(30),
  site_id_alt_02 VARCHAR(30),
  site_id_alt_03 VARCHAR(30),
 site_id_alt_04 VARCHAR(30)

);

insert into site_id_cd_result_02 values('1A1-071', '040777', '04253626', '1A1-071O', NULL);
insert into site_id_cd_result_02 values('1A1-072', '04268383', '123546', NULL, NULL);

Answer 1

好吧，这有点笨拙，但它的优点是工作：

WITH cteAgg AS (SELECT SITE_ID,
                       LISTAGG(SITE_ID_ALT, '|') WITHIN GROUP (ORDER BY SITE_ID_ALT) ALTS
                  FROM SITE_ID_CD_TEST
                  GROUP BY SITE_ID)
SELECT SITE_ID,
       REGEXP_SUBSTR(ALTS, '[^|]+', 1, 1) AS ALT1,
       REGEXP_SUBSTR(ALTS, '[^|]+', 1, 2) AS ALT2,
       REGEXP_SUBSTR(ALTS, '[^|]+', 1, 3) AS ALT3,
       REGEXP_SUBSTR(ALTS, '[^|]+', 1, 4) AS ALT4,
       REGEXP_SUBSTR(ALTS, '[^|]+', 1, 5) AS ALT5,
       REGEXP_SUBSTR(ALTS, '[^|]+', 1, 6) AS ALT6,
       REGEXP_SUBSTR(ALTS, '[^|]+', 1, 7) AS ALT7,
       REGEXP_SUBSTR(ALTS, '[^|]+', 1, 8) AS ALT8,
       REGEXP_SUBSTR(ALTS, '[^|]+', 1, 9) AS ALT9,
       REGEXP_SUBSTR(ALTS, '[^|]+', 1, 10) AS ALT10
  FROM cteAgg;

基本上，您将所有替代项 LISTAGG 成一个大字符串，然后将子字符串分解为单独的字段。限制是替代项+分隔符的总长度不能超过您选择的版本中 VARCHAR2 的最大允许大小。尽管如此，我还是见过（并写过 :-) 更糟的。

dbfiddle here

Answer 2

使用 ROW_NUMBER() 分析函数为每个 SITE_ID_ALT 提供每个 SITE_ID 的列号，然后您可以 PIVOT ：

查询:

INSERT INTO site_id_cd_result ( site_id, site_id_alt_01, site_id_alt_02, site_id_alt_03 )
SELECT site_id,
       site_id_alt_01,
       site_id_alt_02,
       site_id_alt_03
FROM   (
  SELECT t.*,
         ROW_NUMBER() OVER ( PARTITION BY site_id ORDER BY ROWNUM ) AS rn
  FROM   site_id_cd_test t
  -- WHERE site_id <> site_id_alt    -- If you don't want the matching value
)
PIVOT ( MAX( site_id_alt ) FOR rn IN (
  1 AS site_id_alt_01,
  2 AS site_id_alt_02,
  3 AS site_id_alt_03
) )

结果:

SELECT * FROM site_id_cd_result;

SITE_ID | SITE_ID_ALT_01 | SITE_ID_ALT_02 | SITE_ID_ALT_03
:------ | :------------- | :------------- | :-------------
1A1-071 | 1A1-071        | 1A1-071O       | 030256        
1A1-072 | 04268383       | null           | null

db<>fiddle here

Answer 3

终于明白了！一位同事帮助了我，我们想出了在 site_id 到 select 值上使用 MAX 函数和 GROUP BY，这样只有一行数据在开头没有空值。可能有另一种方法可以做到这一点，但目前有效。这确实需要制作两张表，但我无法将它们压缩成一张。

DROP TABLE site_id_cd_result;

CREATE TABLE site_id_cd_result
 (
  site_id VARCHAR(30),
  site_id_alt_01 VARCHAR(30),
  site_id_alt_02 VARCHAR(30),
  site_id_alt_03 VARCHAR(30),
  site_id_alt_04 VARCHAR(30),
  site_id_alt_05 VARCHAR(30)
);


INSERT INTO site_id_cd_result (
site_id, site_id_alt_01, site_id_alt_02, site_id_alt_03, site_id_alt_04, site_id_alt_05)
SELECT site_id,
      site_id_alt_01,
      site_id_alt_02,
      site_id_alt_03,
      site_id_alt_04,
      site_id_alt_05
FROM (
  SELECT t.*,
        ROW_NUMBER() OVER ( PARTITION BY site_id ORDER BY ROWNUM ) AS rn
  FROM   site_id_cd t
)
PIVOT ( MAX( site_id_alt ) FOR rn IN (
  1 AS site_id_alt_01,
  2 AS site_id_alt_02,
  3 AS site_id_alt_03,
  4 AS site_id_alt_04,
  5 AS site_id_alt_05,
  6 AS site_id_alt_06
) )
;


DROP TABLE site_id_cd_result_02;

CREATE TABLE site_id_cd_result_02 AS
SELECT site_id, MAX(site_id_alt_01) AS site_id_alt_01, MAX(site_id_alt_02) AS site_id_alt_02, MAX(site_id_alt_03) AS site_id_alt_03, 
       MAX(site_id_alt_04) AS site_id_alt_04, MAX(site_id_alt_05) AS site_id_alt_05
FROM site_id_cd_result
GROUP BY site_id;

在 Oracle 中，如何将数据透视到多个列中？

In Oracle, how do I pivot data into multiple columns?

oracle

pivot

row-number

rownum