如何在一组行之后或在没有PL/SQL块的情况下有条件地递增oracle序列？

我正在尝试执行代码重写。其中之一就是这个怪物。我有一组普通的DML，突然在脚本中出现了这个PL/SQL块（请参阅当前的解决方案），这在SQL DML中看起来很奇怪

最初，我们决定使用PL/SQL块，其假设是，

“在记录组中使用序列号更新列而不使用增量，在下一组中使用增量是无法在单个SQL中实现的。”

问题：如何在一组行之后或有条件地递增oracle序列

数据设置：

CREATE TABLE TEMP_GP_SEQ
(
   COL1   NUMBER,
   COL2   NUMBER,
   COL3   NUMBER,
   COL4   NUMBER,
   COL5   NUMBER,
   COL6   NUMBER,
   COL7   VARCHAR2 (10)
);

INSERT INTO TEMP_GP_SEQ VALUES(1,10,100,NULL,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(1,10,101,NULL,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(12,10,100,1,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(1,10,100,2,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(1,10,100,3,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(12,10,100,1,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(2,10,100,NULL,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(2,10,101,NULL,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(2,10,101,1,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(2,10,101,1,NULL,NULL,NULL);

COMMIT; 

CREATE SEQUENCE SEQ_TEMP_TEST
   START WITH 1
   INCREMENT BY 1;

DECLARE
    VAL INTEGER;
BEGIN
    FOR REC IN ( SELECT
                    COL1,
                    COL3,
                    COL4
              FROM
                    TEMP_GP_SEQ
              GROUP BY
                    COL1,
                    COL3,
                    COL4
              HAVING
                    COUNT ( * ) > 1 )
    LOOP
        SELECT SEQ_TEMP_TEST.NEXTVAL INTO VAL FROM DUAL;

        UPDATE
              TEMP_GP_SEQ
        SET
              COL7   = 'M' || VAL
        WHERE
                 ( COL1 = REC.COL1 OR ( COL1 IS NULL AND REC.COL1 IS NULL ) )
              AND ( COL3 = REC.COL3 OR ( COL3 IS NULL AND REC.COL3 IS NULL ) )
              AND ( COL4 = REC.COL4 OR ( COL4 IS NULL AND REC.COL4 IS NULL ) );
    END LOOP;
END;
/

SELECT
      A.COL1,
      A.COL2,
      A.COL3,
      A.COL4,
      A.COL5,
      A.COL6,
      DECODE ( B.ID, NULL, '', 'M' )
      || SEQ
          AS COL7
FROM
      TEMP_GP_SEQ A,
      (SELECT
            COL1,
            COL3,
            COL4,
            RETSEQ SEQ,
            ROW_NUMBER ( )
                OVER ( ORDER BY
                          COL1,
                          COL3,
                          COL4 )
                ID
       FROM
            TEMP_GP_SEQ
       GROUP BY
            COL1,
            COL3,
            COL4
       HAVING
            COUNT ( * ) > 1) B
WHERE
         A.COL1 = B.COL1(+)
      AND A.COL3 = B.COL3(+)
      AND A.COL4 = B.COL4(+);

因此，我的相关记录的选择标准

SELECT
      COL1,
      COL3,
      COL4,
      COUNT ( * )
FROM
      TEMP_GP_SEQ
GROUP BY
      COL1,
      COL3,
      COL4
HAVING
      COUNT ( * ) > 1;

会给我

COL1    COL3    COL4    COUNT(*)
2       101     1       2
12      100     1       2

我需要根据上面提到的分组范围，使用创建的序列更新临时表和COL7列上的临时表。但是序列不应该为每个记录递增，它应该只为组的更改递增。比如说

SELECT 
      COL1,
      COL2,
      COL3,
      COL4,
      COL7
FROM
      TEMP_GP_SEQ;

所需输出

COL1    COL2    COL3   COL4    COL7         
1       10      100    NULL    NULL         
1       10      101    NULL    NULL         
12      10      100    1       M2           
1       10      100    2       NULL         
1       10      100    3       NULL         
12      10      100    1       M2           
2       10      100    NULL    NULL         
2       10      101    NULL    NULL         
2       10      101    1       M1           
2       10      101    1       M1           

COL7在这四行中更新，前缀为M，M后的数字来自序列。在这里，只有当分组标准发生变化并且在整个组中保持不变时，数字才会发生变化（序列递增）

挑战在于任何列中都可能存在空值。因此，在分组时，应考虑空值。因此使用的是NULL。（为了确保安全，NVL被忽略）

现有解决方案：

CREATE TABLE TEMP_GP_SEQ
(
   COL1   NUMBER,
   COL2   NUMBER,
   COL3   NUMBER,
   COL4   NUMBER,
   COL5   NUMBER,
   COL6   NUMBER,
   COL7   VARCHAR2 (10)
);

INSERT INTO TEMP_GP_SEQ VALUES(1,10,100,NULL,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(1,10,101,NULL,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(12,10,100,1,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(1,10,100,2,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(1,10,100,3,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(12,10,100,1,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(2,10,100,NULL,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(2,10,101,NULL,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(2,10,101,1,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(2,10,101,1,NULL,NULL,NULL);

COMMIT; 

CREATE SEQUENCE SEQ_TEMP_TEST
   START WITH 1
   INCREMENT BY 1;

DECLARE
    VAL INTEGER;
BEGIN
    FOR REC IN ( SELECT
                    COL1,
                    COL3,
                    COL4
              FROM
                    TEMP_GP_SEQ
              GROUP BY
                    COL1,
                    COL3,
                    COL4
              HAVING
                    COUNT ( * ) > 1 )
    LOOP
        SELECT SEQ_TEMP_TEST.NEXTVAL INTO VAL FROM DUAL;

        UPDATE
              TEMP_GP_SEQ
        SET
              COL7   = 'M' || VAL
        WHERE
                 ( COL1 = REC.COL1 OR ( COL1 IS NULL AND REC.COL1 IS NULL ) )
              AND ( COL3 = REC.COL3 OR ( COL3 IS NULL AND REC.COL3 IS NULL ) )
              AND ( COL4 = REC.COL4 OR ( COL4 IS NULL AND REC.COL4 IS NULL ) );
    END LOOP;
END;
/

SELECT
      A.COL1,
      A.COL2,
      A.COL3,
      A.COL4,
      A.COL5,
      A.COL6,
      DECODE ( B.ID, NULL, '', 'M' )
      || SEQ
          AS COL7
FROM
      TEMP_GP_SEQ A,
      (SELECT
            COL1,
            COL3,
            COL4,
            RETSEQ SEQ,
            ROW_NUMBER ( )
                OVER ( ORDER BY
                          COL1,
                          COL3,
                          COL4 )
                ID
       FROM
            TEMP_GP_SEQ
       GROUP BY
            COL1,
            COL3,
            COL4
       HAVING
            COUNT ( * ) > 1) B
WHERE
         A.COL1 = B.COL1(+)
      AND A.COL3 = B.COL3(+)
      AND A.COL4 = B.COL4(+);

真的有可能将其重构为普通SQL而不是PL/SQL块吗？ 如果您需要任何澄清，请告诉我

---

任何想尝试的人都可以看到小提琴

如果我正确理解您的问题，此查询将满足您的要求，那么您只需要使用它来更新您的行但是我没有使用

更新

来更改行，也没有使用fiddle，因为（我不知道为什么）它不允许我创建函数

为了运行select语句，我需要创建一个函数来返回序列，因为oracle不允许我在sql语句中使用它（至少在我的10.2.x版本中不允许）

首先，我创建了这个函数：

create or replace function retSeq return number
as
   n number;
begin
   select SEQ_TEMP_TEST.nextval into n from dual;
   return n;
end;

然后我做了select语句。我认为把代码改成这个会使它更难理解。但问题是要用一个查询解决问题，我几乎做到了（必须创建函数）。所以，不要害怕：

SELECT s1.col1, s1.col2, s1.col3, s1.col4, s1.col5, s1.col6, 
       decode(s1.id,null,'','M')
           || (SELECT retseq seq
                 FROM (SELECT   col1, col3, col4,
                                ROW_NUMBER () OVER (ORDER BY col1, col3, col4) ID
                         FROM temp_gp_seq
                        GROUP BY col1, col3, col4
                       HAVING COUNT (*) > 1)
                WHERE ID = s1.ID)
           as col7 
  FROM (SELECT a.*, b.ID
          FROM temp_gp_seq a,
               (SELECT col1, col3, col4,
                       ROW_NUMBER () OVER (ORDER BY col1, col3, col4) ID
                  FROM (SELECT   col1, col3, col4, COUNT (*) ct
                            FROM temp_gp_seq
                        GROUP BY col1, col3, col4
                          HAVING COUNT (*) > 1)) b
         WHERE a.col1 = b.col1(+) 
           AND a.col3 = b.col3(+)
           AND a.col4 = b.col4(+)) s1

结果将是（在第一次运行时，由于顺序原因）

由OP跟进：

CREATE TABLE TEMP_GP_SEQ
(
   COL1   NUMBER,
   COL2   NUMBER,
   COL3   NUMBER,
   COL4   NUMBER,
   COL5   NUMBER,
   COL6   NUMBER,
   COL7   VARCHAR2 (10)
);

INSERT INTO TEMP_GP_SEQ VALUES(1,10,100,NULL,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(1,10,101,NULL,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(12,10,100,1,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(1,10,100,2,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(1,10,100,3,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(12,10,100,1,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(2,10,100,NULL,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(2,10,101,NULL,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(2,10,101,1,NULL,NULL,NULL);
INSERT INTO TEMP_GP_SEQ VALUES(2,10,101,1,NULL,NULL,NULL);

COMMIT; 

CREATE SEQUENCE SEQ_TEMP_TEST
   START WITH 1
   INCREMENT BY 1;

DECLARE
    VAL INTEGER;
BEGIN
    FOR REC IN ( SELECT
                    COL1,
                    COL3,
                    COL4
              FROM
                    TEMP_GP_SEQ
              GROUP BY
                    COL1,
                    COL3,
                    COL4
              HAVING
                    COUNT ( * ) > 1 )
    LOOP
        SELECT SEQ_TEMP_TEST.NEXTVAL INTO VAL FROM DUAL;

        UPDATE
              TEMP_GP_SEQ
        SET
              COL7   = 'M' || VAL
        WHERE
                 ( COL1 = REC.COL1 OR ( COL1 IS NULL AND REC.COL1 IS NULL ) )
              AND ( COL3 = REC.COL3 OR ( COL3 IS NULL AND REC.COL3 IS NULL ) )
              AND ( COL4 = REC.COL4 OR ( COL4 IS NULL AND REC.COL4 IS NULL ) );
    END LOOP;
END;
/

SELECT
      A.COL1,
      A.COL2,
      A.COL3,
      A.COL4,
      A.COL5,
      A.COL6,
      DECODE ( B.ID, NULL, '', 'M' )
      || SEQ
          AS COL7
FROM
      TEMP_GP_SEQ A,
      (SELECT
            COL1,
            COL3,
            COL4,
            RETSEQ SEQ,
            ROW_NUMBER ( )
                OVER ( ORDER BY
                          COL1,
                          COL3,
                          COL4 )
                ID
       FROM
            TEMP_GP_SEQ
       GROUP BY
            COL1,
            COL3,
            COL4
       HAVING
            COUNT ( * ) > 1) B
WHERE
         A.COL1 = B.COL1(+)
      AND A.COL3 = B.COL3(+)
      AND A.COL4 = B.COL4(+);

---

PS:删除了不必要的子查询，并将窗口函数组合在一个查询中。

我相信任何pl/sql都可以转换为sql。我经常使用NVL来避免空值，也许这会有所帮助？我也相信这一点，因此发布了以下内容：）。只有当我

1000%确定虚拟值永远不会出现在列上时，才能使用
NVL
。为了避免这种情况，我使用了
IS NULL
。ROWNUM在oracle中也是一个有用的转换结构。仍然在看你的SQL代码，我的眼睛都睁大了。这个序列是必要的吗？或者你可以考虑不使用它？绝对必要。这正是我所做的，我检查并简化了你的查询，但逻辑上完全正确。：）见下文up@realspirituals是的，那太好了，我是一步一步地解决问题的，我没有尽可能地去做。感谢您的跟进：）