Lag和Lead函数可以在一次查询中取出同一字段的前N行的数据和后N行的值。这种操作可以使用对相同表的表连接来实现,不过使用LAG和LEAD有更高的效率。
lead 和lag 的语法类似以下以lag为例进行讲解! lag(exp_str,offset,defval) over() offset 是exp_str字段的偏移量 比如说 offset 为2 则 拿exp_str的第一行和第三行对比,第二行和第四行,依次类推,offset的默认值为1!
defval是当该函数无值可用的情况下返回的值。Lead函数的用法类似。 SCOTT@yangdb> set pagesize 10000
SCOTT@yangdb> select ename,job,sal ,lag(sal) over(order by sal) last_sal from emp;
ENAME JOB SAL LAST_SAL
---------- --------- ---------- ----------
SMITH CLERK 800
--此时没有设置default 值 则为空值 JAMES CLERK 950 800
ADAMS CLERK 1100 950
WARD SALESMAN 1250 1100
MARTIN SALESMAN 1250 1250
MILLER CLERK 1300 1250
TURNER SALESMAN 1500 1300
ALLEN SALESMAN 1600 1500
CLARK MANAGER 2450 1600
BLAKE MANAGER 2850 2450
JONES MANAGER 2975 2850
SCOTT ANALYST 3000 2975
FORD ANALYST 3000 3000
KING PRESIDENT 5000 3000
14 rows selected.
设置了default 值之后 第一行对应的值 为500
SCOTT@yangdb> select ename,job,sal ,lag(sal,1,500) over(order by sal) last_sal from emp;
ENAME JOB SAL LAST_SAL
---------- --------- ---------- ----------
SMITH CLERK 800
500 JAMES CLERK 950 800
ADAMS CLERK 1100 950
WARD SALESMAN 1250 1100
MARTIN SALESMAN 1250 1250
MILLER CLERK 1300 1250
TURNER SALESMAN 1500 1300
ALLEN SALESMAN 1600 1500
CLARK MANAGER 2450 1600
BLAKE MANAGER 2850 2450
JONES MANAGER 2975 2850
SCOTT ANALYST 3000 2975
FORD ANALYST 3000 3000
KING PRESIDENT 5000 3000
14 rows selected.
指定offset的值为2时
SCOTT@yangdb> select ename,job,sal ,lag(sal,2) over(order by sal) last_sal from emp;
ENAME JOB SAL LAST_SAL
---------- --------- ---------- ----------
SMITH CLERK 800
JAMES CLERK 950
ADAMS CLERK 1100 800
WARD SALESMAN 1250 950
MARTIN SALESMAN 1250 1100
MILLER CLERK 1300 1250
TURNER SALESMAN 1500 1250
ALLEN SALESMAN 1600 1300
CLARK MANAGER 2450 1500
BLAKE MANAGER 2850 1600
JONES MANAGER 2975 2450
SCOTT ANALYST 3000 2850
FORD ANALYST 3000 2975
KING PRESIDENT 5000 3000
14 rows selected.
offset的值为3
SCOTT@yangdb> select ename,job,sal ,lag(sal,3) over(order by sal) last_sal from emp;
ENAME JOB SAL LAST_SAL
---------- --------- ---------- ----------
SMITH CLERK 800
JAMES CLERK 950
ADAMS CLERK 1100
WARD SALESMAN 1250 800
MARTIN SALESMAN 1250 950
MILLER CLERK 1300 1100
TURNER SALESMAN 1500 1250
ALLEN SALESMAN 1600 1250
CLARK MANAGER 2450 1300
BLAKE MANAGER 2850 1500
JONES MANAGER 2975 1600
SCOTT ANALYST 3000 2450
FORD ANALYST 3000 2850
KING PRESIDENT 5000 2975
14 rows selected.
使用lead分析函数
SCOTT@yangdb> select ename,job,sal ,lead(sal) over(order by sal) last_sal from emp;
ENAME JOB SAL LAST_SAL
---------- --------- ---------- ----------
SMITH CLERK 800 950
JAMES CLERK 950 1100
ADAMS CLERK 1100 1250
WARD SALESMAN 1250 1250
MARTIN SALESMAN 1250 1300
MILLER CLERK 1300 1500
TURNER SALESMAN 1500 1600
ALLEN SALESMAN 1600 2450
CLARK MANAGER 2450 2850
BLAKE MANAGER 2850 2975
JONES MANAGER 2975 3000
SCOTT ANALYST 3000 3000
FORD ANALYST 3000 5000
KING PRESIDENT 5000
14 rows selected.
SCOTT@yangdb> select ename,job,sal ,lead(sal,1) over(order by sal) last_sal from emp;
ENAME JOB SAL LAST_SAL
---------- --------- ---------- ----------
SMITH CLERK 800 950
JAMES CLERK 950 1100
ADAMS CLERK 1100 1250
WARD SALESMAN 1250 1250
MARTIN SALESMAN 1250 1300
MILLER CLERK 1300 1500
TURNER SALESMAN 1500 1600
ALLEN SALESMAN 1600 2450
CLARK MANAGER 2450 2850
BLAKE MANAGER 2850 2975
JONES MANAGER 2975 3000
SCOTT ANALYST 3000 3000
FORD ANALYST 3000 5000
KING PRESIDENT 5000
14 rows selected.
SCOTT@yangdb> select ename,job,sal ,lead(sal,2) over(order by sal) last_sal from emp; ENAME JOB SAL LAST_SAL ---------- --------- ---------- ---------- SMITH CLERK 800 1100 JAMES CLERK 950 1250 ADAMS CLERK 1100 1250 WARD SALESMAN 1250 1300 MARTIN SALESMAN 1250 1500 MILLER CLERK 1300 1600 TURNER SALESMAN 1500 2450 ALLEN SALESMAN 1600 2850 CLARK MANAGER 2450 2975 BLAKE MANAGER 2850 3000 JONES MANAGER 2975 3000 SCOTT ANALYST 3000 5000 FORD ANALYST 3000 KING PRESIDENT 5000 SCOTT@yangdb> select ename,job,sal ,lead(sal,3) over(order by sal) last_sal from emp; ENAME JOB SAL LAST_SAL ---------- --------- ---------- ---------- SMITH CLERK 800 1250 JAMES CLERK 950 1250 ADAMS CLERK 1100 1300 WARD SALESMAN 1250 1500 MARTIN SALESMAN 1250 1600 MILLER CLERK 1300 2450 TURNER SALESMAN 1500 2850 ALLEN SALESMAN 1600 2975 CLARK MANAGER 2450 3000 BLAKE MANAGER 2850 3000 JONES MANAGER 2975 5000 SCOTT ANALYST 3000 FORD ANALYST 3000 KING PRESIDENT 5000 14 rows selected. lead 的offset N 是以记录的第N行和第一做对比注意末尾的 null 值! Lead和Lag函数也可以使用分组,以下是使用 job 分组的例子:
SCOTT@yangdb> select ename,job,sal ,lead(sal,1) over(partition by job order by sal) last_sal from emp;
ENAME JOB SAL LAST_SAL
---------- --------- ---------- ----------
FORD ANALYST 3000 3000
SCOTT ANALYST 3000
SMITH CLERK 800 950
JAMES CLERK 950 1100
ADAMS CLERK 1100 1300
MILLER CLERK 1300
CLARK MANAGER 2450 2850
BLAKE MANAGER 2850 2975
JONES MANAGER 2975
KING PRESIDENT 5000
MARTIN SALESMAN 1250 1250
WARD SALESMAN 1250 1500
TURNER SALESMAN 1500 1600
ALLEN SALESMAN 1600
14 rows selected.
SCOTT@yangdb> select ename,job,sal ,lag(sal,1) over(partition by job order by sal) last_sal from emp;
ENAME JOB SAL LAST_SAL
---------- --------- ---------- ----------
FORD ANALYST 3000
SCOTT ANALYST 3000 3000
SMITH CLERK 800
JAMES CLERK 950 800
ADAMS CLERK 1100 950
MILLER CLERK 1300 1100
CLARK MANAGER 2450
BLAKE MANAGER 2850 2450
JONES MANAGER 2975 2850
KING PRESIDENT 5000
MARTIN SALESMAN 1250
WARD SALESMAN 1250 1250
TURNER SALESMAN 1500 1250
ALLEN SALESMAN 1600 1500
14 rows selected.
SCOTT@yangdb>
使用分析函数的时候注意空值 或者null 给数据带来的影响,数据是否允许为空或者null
计算的时候会导致一定的差错 比如 800-null 肯定为null!这个结果是否是应用想要的结果?