根据当前值重建sequence

大家知道oracle中的序列与表之前没有必然的关系，一个序列可以给一个表使用也可以给另一个一使用。

所以最好增加字典表及文档来保存序列与表之间的对应关系。

那最好这些都没有怎么办呢？

有一个系统中是根据触发器来调用序列，通过触发器我们可以找到表与序列之间的对应关系

SELECT NAME,
MAX(CASE referenced_type
WHEN 'TABLE' THEN
referenced_name
END) AS tbl,
MAX(CASE referenced_type
WHEN 'SEQUENCE' THEN
referenced_name
END) AS seq
FROM user_dependencies seq
WHERE TYPE = 'TRIGGER'
AND NAME IN (SELECT NAME
FROM user_source
WHERE TYPE = 'TRIGGER'
AND upper(text) LIKE upper('%.nextval%'))
AND referenced_type IN ('TABLE', 'SEQUENCE')
GROUP BY NAME
HAVING COUNT(*) = 2
/

NAME       TBL        SEQ
---------- ---------- ----------
TRG_EMP2   EMP2       SEQ_EMP2

有了这个对应关系后就可以查找表对应的主键列

WITH seq AS
(SELECT /*+ materialize */
NAME,
MAX(CASE referenced_type
WHEN 'TABLE' THEN
referenced_name
END) AS tbl,
MAX(CASE referenced_type
WHEN 'SEQUENCE' THEN
referenced_name
END) AS seq
FROM user_dependencies seq
WHERE TYPE = 'TRIGGER'
AND NAME IN (SELECT NAME
FROM user_source
WHERE TYPE = 'TRIGGER'
AND upper(text) LIKE upper('%.nextval%'))
AND referenced_type IN ('TABLE', 'SEQUENCE')
GROUP BY NAME
HAVING COUNT(*) = 2)
SELECT trg,seq, tbl, column_name
FROM (SELECT seq.seq,
seq.tbl,
seq.NAME AS trg,
ic.column_name,
COUNT(*) over(PARTITION BY ic.table_name) AS cnt
FROM seq
INNER JOIN user_constraints c
ON c.table_name = seq.tbl
INNER JOIN user_ind_columns ic
ON ic.index_name = c.index_name
INNER JOIN user_tab_cols col
ON col.table_name = ic.table_name
AND col.column_name = ic.column_name
WHERE constraint_type = 'P')
WHERE cnt = 1;

TRG        SEQ        TBL        COLUMN_NAME
---------- ---------- ---------- ---------------
TRG_EMP2   SEQ_EMP2   EMP2       EMPNO

因为使用序列的表一般都是单列主键，所以这儿没有判断主键列的个数。

对应信息都得到后就可以重建seq了

DECLARE
CURSOR cur_seq IS
WITH seq AS
(SELECT /*+ materialize */
NAME,
MAX(CASE referenced_type
WHEN 'TABLE' THEN
referenced_name
END) AS tbl,
MAX(CASE referenced_type
WHEN 'SEQUENCE' THEN
referenced_name
END) AS seq
FROM user_dependencies seq
WHERE TYPE = 'TRIGGER'
AND NAME IN (SELECT NAME
FROM user_source
WHERE TYPE = 'TRIGGER'
AND upper(text) LIKE upper('%.nextval%'))
AND referenced_type IN ('TABLE', 'SEQUENCE')
GROUP BY NAME
HAVING COUNT(*) = 2)
SELECT trg,seq, tbl, column_name
FROM (SELECT seq.seq,
seq.tbl,
seq.NAME AS trg,
ic.column_name,
COUNT(*) over(PARTITION BY ic.table_name) AS cnt
FROM seq
INNER JOIN user_constraints c
ON c.table_name = seq.tbl
INNER JOIN user_ind_columns ic
ON ic.index_name = c.index_name
INNER JOIN user_tab_cols col
ON col.table_name = ic.table_name
AND col.column_name = ic.column_name
WHERE constraint_type = 'P')
WHERE cnt = 1;
v_trg   VARCHAR2(50);
v_seq   VARCHAR2(50);
v_tbl   VARCHAR2(50);
v_col   VARCHAR2(50);
v_sql   VARCHAR2(4000);
v_maxid NUMBER;
v_nextval NUMBER;
BEGIN
OPEN cur_seq;
LOOP
FETCH cur_seq
INTO v_trg,v_seq, v_tbl, v_col;
EXIT WHEN cur_seq%NOTFOUND;
v_sql := 'select max(' || v_col || ') from ' || v_tbl;
--dbms_output.put_line(v_sql);
EXECUTE IMMEDIATE v_sql INTO v_maxid;
v_maxid := NVL(v_maxid,0) + 1;
v_sql := 'drop sequence ' || v_seq;
--dbms_output.put_line(v_sql);
EXECUTE IMMEDIATE v_sql;
v_sql := 'create sequence ' || v_seq || ' start with ' || to_char(v_maxid);
--dbms_output.put_line(v_sql);
EXECUTE IMMEDIATE v_sql;
v_sql := 'alter trigger ' || v_trg || ' compile';
--dbms_output.put_line(v_sql);
EXECUTE IMMEDIATE v_sql;
NULL;
END LOOP;
CLOSE cur_seq;
END;