Oracle SQL精妙SQL语句讲解

 

 

 

 

-- 行列转换 行转列
DROP   TABLE  t_change_lc;
CREATE   TABLE  t_change_lc (card_code  VARCHAR2 ( 3 ), q  NUMBER , bal  NUMBER );

INSERT   INTO  t_change_lc 
SELECT   ' 001 '  card_code, ROWNUM q, trunc(dbms_random.VALUE  *   100 ) bal  FROM  dual CONNECT  BY  ROWNUM  <=   4
UNION  
SELECT   ' 002 '  card_code, ROWNUM q, trunc(dbms_random.VALUE  *   100 ) bal  FROM  dual CONNECT  BY  ROWNUM  <=   4 ;

SELECT   *   FROM  t_change_lc;

SELECT  a.card_code,
SUM (decode(a.q,  1 , a.bal,  0 )) q1,
SUM (decode(a.q,  2 , a.bal,  0 )) q2,
SUM (decode(a.q,  3 , a.bal,  0 )) q3,
SUM (decode(a.q,  4 , a.bal,  0 )) q4
FROM  t_change_lc a
GROUP   BY  a.card_code
ORDER   BY   1 ;

-- 行列转换 列转行
DROP   TABLE  t_change_cl;
CREATE   TABLE  t_change_cl  AS  
SELECT  a.card_code,
SUM (decode(a.q,  1 , a.bal,  0 )) q1,
SUM (decode(a.q,  2 , a.bal,  0 )) q2,
SUM (decode(a.q,  3 , a.bal,  0 )) q3,
SUM (decode(a.q,  4 , a.bal,  0 )) q4
FROM  t_change_lc a
GROUP   BY  a.card_code
ORDER   BY   1 ;

SELECT   *   FROM  t_change_cl;

SELECT  t.card_code,
t.rn q,
decode(t.rn,  1 , t.q1,  2 , t.q2,  3 , t.q3,  4 , t.q4) bal
FROM  ( SELECT  a. * , b.rn
FROM  t_change_cl a,
( SELECT  ROWNUM rn  FROM  dual CONNECT  BY  ROWNUM  <=   4 ) b) t
ORDER   BY   1 ,  2 ;

-- 行列转换 行转列 合并
DROP   TABLE  t_change_lc_comma;
CREATE   TABLE  t_change_lc_comma  AS   SELECT  card_code, ' quarter_ ' || q  AS  q  FROM  t_change_lc; 

SELECT   *   FROM  t_change_lc_comma;

SELECT  t1.card_code, substr( MAX (sys_connect_by_path(t1.q,  ' ; ' )),  2 ) q
FROM  ( SELECT  a.card_code,
a.q,
row_number()  over (PARTITION  BY  a.card_code  ORDER   BY  a.q) rn
FROM  t_change_lc_comma a) t1
START  WITH  t1.rn  =   1
CONNECT  BY  t1.card_code  =  PRIOR t1.card_code
AND  t1.rn  -   1   =  PRIOR t1.rn
GROUP   BY  t1.card_code;

-- 行列转换 列转行 分割
DROP   TABLE  t_change_cl_comma;
CREATE   TABLE  t_change_cl_comma  AS
SELECT  t1.card_code, substr( MAX (sys_connect_by_path(t1.q,  ' ; ' )),  2 ) q
FROM  ( SELECT  a.card_code,
a.q,
row_number()  over (PARTITION  BY  a.card_code  ORDER   BY  a.q) rn
FROM  t_change_lc_comma a) t1
START  WITH  t1.rn  =   1
CONNECT  BY  t1.card_code  =  PRIOR t1.card_code
AND  t1.rn  -   1   =  PRIOR t1.rn
GROUP   BY  t1.card_code;

SELECT   *   FROM  t_change_cl_comma;

SELECT  t.card_code,
substr(t.q,
instr( ' ; '   ||  t.q,  ' ; ' ,  1 , rn),
instr(t.q  ||   ' ; ' ,  ' ; ' ,  1 , rn)  -  instr( ' ; '   ||  t.q,  ' ; ' ,  1 , rn)) q
FROM  ( SELECT  a.card_code, a.q, b.rn
FROM  t_change_cl_comma a,
( SELECT  ROWNUM rn  FROM  dual CONNECT  BY  ROWNUM  <=   100 ) b
WHERE  instr( ' ; '   ||  a.q,  ' ; ' ,  1 , rn)  >   0 ) t
ORDER   BY   1 ,  2 ;


--  实现一条记录根据条件多表插入
DROP   TABLE  t_ia_src;
CREATE   TABLE  t_ia_src  AS   SELECT   ' a ' || ROWNUM c1,  ' b ' || ROWNUM c2  FROM  dual CONNECT  BY  ROWNUM <= 5 ;
DROP   TABLE  t_ia_dest_1;
CREATE   TABLE  t_ia_dest_1(flag  VARCHAR2 ( 10 ) , c  VARCHAR2 ( 10 ));
DROP   TABLE  t_ia_dest_2;
CREATE   TABLE  t_ia_dest_2(flag  VARCHAR2 ( 10 ) , c  VARCHAR2 ( 10 ));
DROP   TABLE  t_ia_dest_3;
CREATE   TABLE  t_ia_dest_3(flag  VARCHAR2 ( 10 ) , c  VARCHAR2 ( 10 ));

SELECT   *   FROM  t_ia_src; 
SELECT   *   FROM  t_ia_dest_1;
SELECT   *   FROM  t_ia_dest_2;
SELECT   *   FROM  t_ia_dest_3;

INSERT   ALL
WHEN  (c1  IN  ( ' a1 ' , ' a3 ' ))  THEN  
INTO  t_ia_dest_1(flag,c)  VALUES (flag1,c2)
WHEN  (c1  IN  ( ' a2 ' , ' a4 ' ))  THEN  
INTO  t_ia_dest_2(flag,c)  VALUES (flag2,c2)
ELSE
INTO  t_ia_dest_3(flag,c)  VALUES (flag1 || flag2,c1 || c2)
SELECT  c1,c2,  ' f1 '  flag1,  ' f2 '  flag2  FROM  t_ia_src;

--  如果存在就更新，不存在就插入用一个语句实现
DROP   TABLE  t_mg;
CREATE   TABLE  t_mg(code  VARCHAR2 ( 10 ), NAME  VARCHAR2 ( 10 ));

SELECT   *   FROM  t_mg;

MERGE  INTO  t_mg a
USING ( SELECT   ' the code '  code,  ' the name '  NAME  FROM  dual) b
ON  (a.code  =  b.code)
WHEN  MATCHED  THEN
UPDATE   SET  a.NAME  =  b.NAME
WHEN   NOT  MATCHED  THEN
INSERT  (code, NAME)  VALUES  (b.code, b.NAME);

--  抽取/删除重复记录 
DROP   TABLE  t_dup;
CREATE   TABLE  t_dup  AS   SELECT   ' code_ ' || ROWNUM code, dbms_random.string( ' z ' , 5 ) NAME  FROM  dual CONNECT  BY  ROWNUM <= 10 ; 
INSERT   INTO  t_dup  SELECT   ' code_ ' || ROWNUM code, dbms_random.string( ' z ' , 5 ) NAME  FROM  dual CONNECT  BY  ROWNUM <= 2 ; 

SELECT   *   FROM  t_dup;

SELECT   *   FROM  t_dup a  WHERE  a.ROWID  <>  ( SELECT   MIN (b.ROWID)  FROM  t_dup b  WHERE  a.code = b.code);

SELECT  b.code, b.NAME
FROM  ( SELECT  a.code,
a.NAME,
row_number()  over (PARTITION  BY  a.code  ORDER   BY  a.ROWID) rn
FROM  t_dup a) b
WHERE  b.rn  >   1 ;

--  IN/EXISTS的不同适用环境
--  t_orders.customer_id有索引
SELECT  a. *
FROM  t_employees a
WHERE  a.employee_id  IN
( SELECT  b.sales_rep_id  FROM  t_orders b  WHERE  b.customer_id  =   12 );

SELECT  a. *
FROM  t_employees a
WHERE   EXISTS  ( SELECT   1
FROM  t_orders b
WHERE  b.customer_id  =   12
AND  a.employee_id  =  b.sales_rep_id);

--  t_employees.department_id有索引
SELECT  a. *
FROM  t_employees a
WHERE  a.department_id  =   10
AND   EXISTS
( SELECT   1   FROM  t_orders b  WHERE  a.employee_id  =  b.sales_rep_id);

SELECT  a. *
FROM  t_employees a
WHERE  a.department_id  =   10
AND  a.employee_id  IN  ( SELECT  b.sales_rep_id  FROM  t_orders b);

--  FBI
DROP   TABLE  t_fbi;
CREATE   TABLE  t_fbi  AS
SELECT  ROWNUM rn, dbms_random.STRING( ' z ' , 10 ) NAME , SYSDATE  +  dbms_random.VALUE  *   10  dt  FROM  dual 
CONNECT  BY  ROWNUM  <= 10 ; 

CREATE   INDEX  idx_nonfbi  ON  t_fbi(dt);

DROP   INDEX  idx_fbi_1;
CREATE   INDEX  idx_fbi_1  ON  t_fbi(trunc(dt));

SELECT   *   FROM  t_fbi  WHERE  trunc(dt)  =  to_date( ' 2006-09-21 ' , ' yyyy-mm-dd ' ) ;

--  不建议使用
SELECT   *   FROM  t_fbi  WHERE  to_char(dt,  ' yyyy-mm-dd ' )  =   ' 2006-09-21 ' ;

--  LOOP中的COMMIT/ROLLBACK
DROP   TABLE  t_loop PURGE;
create   TABLE  t_loop  AS   SELECT   *   FROM  user_objects  WHERE   1 = 2 ;

SELECT   *   FROM  t_loop;

--  逐行提交
DECLARE
BEGIN
FOR  cur  IN  ( SELECT   *   FROM  user_objects) LOOP
INSERT   INTO  t_loop  VALUES  cur;
COMMIT ;
END  LOOP;
END ;

--  模拟批量提交
DECLARE
v_count  NUMBER ;
BEGIN
FOR  cur  IN  ( SELECT   *   FROM  user_objects) LOOP
INSERT   INTO  t_loop  VALUES  cur;
v_count : =  v_count  +   1 ;
IF  v_count  >=   100   THEN
COMMIT ;
END   IF ;
END  LOOP;
COMMIT ;
END ;

--  真正的批量提交
DECLARE
CURSOR  cur  IS
SELECT   *   FROM  user_objects;
TYPE rec  IS   TABLE   OF  user_objects % ROWTYPE;
recs rec;
BEGIN
OPEN  cur;
WHILE  (TRUE) LOOP
FETCH  cur  BULK  COLLECT
INTO  recs LIMIT  100 ;
--  forall 实现批量
FORALL i  IN   1  .. recs. COUNT
INSERT   INTO  t_loop  VALUES  recs (i);
COMMIT ;
EXIT   WHEN  cur % NOTFOUND;
END  LOOP;
CLOSE  cur;
END ;

--  悲观锁定/乐观锁定 
DROP   TABLE  t_lock PURGE;
CREATE   TABLE  t_lock  AS   SELECT   1  ID  FROM  dual;

SELECT   *   FROM  t_lock;

--  常见的实现逻辑，隐含bug
DECLARE
v_cnt  NUMBER ;
BEGIN
--  这里有并发性的bug
SELECT   MAX (ID)  INTO  v_cnt  FROM  t_lock;

--  here for other operation
v_cnt : =  v_cnt  +   1 ;
INSERT   INTO  t_lock (ID)  VALUES  (v_cnt);
COMMIT ;
END ;

--  高并发环境下，安全的实现逻辑
DECLARE
v_cnt  NUMBER ;
BEGIN
--  对指定的行取得lock
SELECT  ID  INTO  v_cnt  FROM  t_lock  WHERE  ID = 1   FOR   UPDATE ;
--  在有lock的情况下继续下面的操作
SELECT   MAX (ID)  INTO  v_cnt  FROM  t_lock;

--  here for other operation
v_cnt : =  v_cnt  +   1 ;
INSERT   INTO  t_lock (ID)  VALUES  (v_cnt);
COMMIT ;  -- 提交并且释放lock
END ;

--  硬解析/软解析
DROP   TABLE  t_hard PURGE;
CREATE   TABLE  t_hard (ID  INT );

SELECT   *   FROM  t_hard;

DECLARE
sql_1  VARCHAR2 ( 200 );
BEGIN
--  hard parse
--  java中的同等语句是 Statement.execute()
FOR  i  IN   1  ..  1000  LOOP
sql_1 : =   ' insert into t_hard(id) values( '   ||  i  ||   ' ) ' ;
EXECUTE  IMMEDIATE sql_1;
END  LOOP;
COMMIT ;

--  soft parse
-- java中的同等语句是 PreparedStatement.execute()
sql_1 : =   ' insert into t_hard(id) values(:id) ' ;
FOR  i  IN   1  ..  1000  LOOP
EXECUTE  IMMEDIATE sql_1
USING i;
END  LOOP;
COMMIT ;
END ;



--  正确的分页算法 
SELECT   *
FROM  ( SELECT  a. * , ROWNUM rn
FROM  ( SELECT   *   FROM  t_employees  ORDER   BY  first_name) a
WHERE  ROWNUM  <=   500 )
WHERE  rn  >   480  ;

--  分页算法(why not this one)
SELECT  a. * , ROWNUM rn
FROM  ( SELECT   *   FROM  t_employees  ORDER   BY  first_name) a
WHERE  ROWNUM  <=   500   AND  ROWNUM  >   480 ;

--  分页算法(why not this one)
SELECT  b. *
FROM  ( SELECT  a. * , ROWNUM rn
FROM  t_employees a
WHERE  ROWNUM  <   =   500
ORDER   BY  first_name) b
WHERE  b.rn  >   480 ;

--  OLAP
--  小计合计
SELECT   CASE
WHEN  a.deptno  IS   NULL   THEN
' 合计 '
WHEN  a.deptno  IS   NOT   NULL   AND  a.empno  IS   NULL   THEN
' 小计 '
ELSE
''   ||  a.deptno
END  deptno,
a.empno,
a.ename,
SUM (a.sal) total_sal
FROM  scott.emp a
GROUP   BY   GROUPING  SETS((a.deptno),(a.deptno, a.empno, a.ename),());

--  分组排序
SELECT  a.deptno,
a.empno,
a.ename,
a.sal,
--  可跳跃的rank
rank()  over (PARTITION  BY  a.deptno  ORDER   BY  a.sal  DESC ) r1,
--  密集型rank
dense_rank()  over (PARTITION  BY  a.deptno  ORDER   BY  a.sal  DESC ) r2,
--  不分组排序
rank()  over ( ORDER   BY  sal  DESC ) r3
FROM  scott.emp a
ORDER   BY  a.deptno,a.sal  DESC ;

--  当前行数据和前/后n行的数据比较
SELECT  a.empno,
a.ename,
a.sal,
--  上面一行
lag(a.sal)  over ( ORDER   BY  a.sal  DESC ) lag_1,
--  下面三行
lead(a.sal,  3 )  over ( ORDER   BY  a.sal  DESC ) lead_3
FROM  scott.emp a
ORDER   BY  a.sal  DESC ;