SQL - 使用列名作为值插入

Question

我正在尝试根据 select from table1 插入 into table2，但我无法获得正确的语法。来自 table1 的列名将驱动插入到 table2 的 PD_NO 列中的值，如下例所示。有人可以帮忙吗？

表1：

         (1)     (2)     (3)     (4)     (5)     (6)
| SEQ | PD_01 | PD_02 | PD_03 | PD_04 | PD_05 | PD_06 |
|-----+-------+-------+-------+-------+-------+-------|
| 632 | 10000 |   0   |  500  |   0   | 20000 |   0   |

表2：

| SEQ | PD_NO |  AMT  |
|-----+-------+-------|
| 632 |   1   | 10000 |
|-----+-------+-------|
| 632 |   3   |  500  |
|-----+-------+-------|
| 632 |   5   | 20000 |
|-----+-------+-------|

我知道如果我在另一个方向工作（将 table2 的内容插入 table1），我可以执行以下操作：

INSERT INTO table1
SELECT 
      seq,
      SUM (CASE WHEN pd_no = 1 THEN amt ELSE 0 END) p01_amt,
      SUM (CASE WHEN pd_no = 2 THEN amt ELSE 0 END) p02_amt,
      SUM (CASE WHEN pd_no = 3 THEN amt ELSE 0 END) p03_amt,
      SUM (CASE WHEN pd_no = 4 THEN amt ELSE 0 END) p04_amt,
      SUM (CASE WHEN pd_no = 5 THEN amt ELSE 0 END) p05_amt,
      SUM (CASE WHEN pd_no = 6 THEN amt ELSE 0 END) p06_amt
FROM table2;

Answer 1

insert ... select from 以 1:1 的行为基础：“select”表中的一行数据进入“insert”表的一行。您试图从源表中取出一行并将其转换为目标表中的多行。这是直接不可能的。您必须运行多个插入/选择，每个要拆分的字段一个：

INSERT INTO table2 (SEQ, PD_NO, AMT) SELECT SEQ, PD_01, AMT FROM table1
INSERT INTO table2 (SEQ, PD_NO, AMT) SELECT SEQ, PD_02, AMT FROM table1
INSERT INTO table2 (SEQ, PD_NO, AMT) SELECT SEQ, PD_03, AMT FROM table1
INSERT INTO table2 (SEQ, PD_NO, AMT) SELECT SEQ, PD_04, AMT FROM table1
etc...

Answer 2

在纯sql中可以这样完成：

INSERT INTO table2 ( SEQ , PD_NO,  AMT )
SELECT SEQ, 1 as pd_no, PD_01 FROM Table1
UNION ALL
SELECT SEQ, 2 as pd_no, PD_02 FROM Table1
UNION ALL
SELECT SEQ, 3 as pd_no, PD_03 FROM Table1
UNION ALL
SELECT SEQ, 4 as pd_no, PD_04 FROM Table1
UNION ALL
SELECT SEQ, 5 as pd_no, PD_05 FROM Table1
UNION ALL
SELECT SEQ, 6 as pd_no, PD_06 FROM Table1

有些数据库已经优化了只读取一次源表的命令（上述查询读取源表6次），例如在ORACLE中：

INSERT ALL
INTO table2 ( SEQ , PD_NO,  AMT ) VALUES ( seq, 1, PD_01 )
INTO table2 ( SEQ , PD_NO,  AMT ) VALUES ( seq, 2, PD_02 )
INTO table2 ( SEQ , PD_NO,  AMT ) VALUES ( seq, 3, PD_03 )
INTO table2 ( SEQ , PD_NO,  AMT ) VALUES ( seq, 4, PD_04 )
INTO table2 ( SEQ , PD_NO,  AMT ) VALUES ( seq, 5, PD_05 )
INTO table2 ( SEQ , PD_NO,  AMT ) VALUES ( seq, 6, PD_06 )
SELECT * FROM table1

Answer 3

CREATE TABLE 1to6 (i INT PRIMARY KEY);
INSERT INTO 1to6 (i) VALUES (1), (2), (3), (4), (5), (6);

INSERT INTO table2 (seq, pd_no, amt)
SELECT seq, i, 
 CASE(i)
 WHEN 1 THEN pd_01
 WHEN 2 THEN pd_02
 WHEN 3 THEN pd_03
 WHEN 4 THEN pd_04
 WHEN 5 THEN pd_05
 WHEN 6 THEN pd_06
 END
FROM table1 CROSS JOIN 1to6

---

重新评论：显然 Oracle 不支持多行 INSERT 语法（尽管该语法是标准 SQL-99）。您可以一次插入一行：

INSERT INTO 1to6 (i) VALUES (1);
INSERT INTO 1to6 (i) VALUES (2);
INSERT INTO 1to6 (i) VALUES (3);
INSERT INTO 1to6 (i) VALUES (4);
INSERT INTO 1to6 (i) VALUES (5);
INSERT INTO 1to6 (i) VALUES (6);

Answer 4

这是一个典型的问题，Oracle 11 提供了 UNPIVOT 子句供查询使用：

insert  into table2(seq, pd_no, amt)
select  seq, pd_no, amt
from    ( select  *
          from    table1
          unpivot (amt for pd_no in (pd_01 as 1, pd_02 as 2, pd_03 as 3, pd_04 as 4, pd_05 as 5, pd_06 as 6))
        );