select t.*,
(
select max(sys_connect_by_path(q,';')) result from
(select no,q,rn,lead(rn) over(partition by no order by rn) rn1
from (select no,q,row_number() over(order by no,q desc) rn from test)
)
start with no = t.no and rn1 is null connect by rn1 = prior rn
) value
from (select distinct no from test) t
可以简化一下：
select no,max(sys_connect_by_path(q,';')) result from
(
select no,q,rn,lead(rn) over(partition by no order by rn) rn1
from (
select no,q,row_number() over(order by no,q desc) rn from test
)
)
connect by rn1 = prior rn
group by no
(
select max(sys_connect_by_path(q,';')) result from
(select no,q,rn,lead(rn) over(partition by no order by rn) rn1
from (select no,q,row_number() over(order by no,q desc) rn from test)
)
start with no = t.no and rn1 is null connect by rn1 = prior rn
) value
from (select distinct no from test) t
可以简化一下：
select no,max(sys_connect_by_path(q,';')) result from
(
select no,q,rn,lead(rn) over(partition by no order by rn) rn1
from (
select no,q,row_number() over(order by no,q desc) rn from test
)
)
connect by rn1 = prior rn
group by no
to UandM(无)，你的方法不行，关键是start with 的问题，你可以再研究下我这么写的用意，还是谢谢你了，大家有什么意见可以踊跃提出
select no,max(sys_connect_by_path(q,';')) result from
(
select no,q,rn,lead(rn) over(partition by no order by rn) rn1
from (
select no,q,row_number() over(order by no,q desc) rn from test
)
)
connect by rn1 = prior rn
group by noSQL>
10 /NO RESULT
---------- --------------------------------------------------------------------------------
001 ;n5
002 ;m1
003 ;t6SQL>
(
select no,q,rn,lead(rn) over(partition by no order by rn) rn1
from (
select no,q,row_number() over(order by no,q desc) rn from test
)
)
start with rn1 is null ---呵呵，确实，这句不能少
connect by rn1 = prior rn
group by no
SQL> select * from test_lead;N Q
--- --
001 n1
001 n2
001 n3
001 n4
001 n5
002 m1
003 t1
003 t2
003 t3
003 t4
003 t5
003 t612 行 已选择SQL>
SQL> select n,rtrim(q,';0') q from (
2 select n ,q,row_number()over(partition by n order by n) as r from (
3 select n0 n,q0||';'||q1||';'||q2||';'||q3||';'||q4||';'||q5 as q from (
4 select n n0,q q0,
5 lead(q,1,0)over(partition by n order by q) q1,
6 lead(q,2,0)over(partition by n order by q) q2,
7 lead(q,3,0)over(partition by n order by q) q3,
8 lead(q,4,0)over(partition by n order by q) q4,
9 lead(q,5,0)over(partition by n order by q) q5
10 from test_lead))
11 ) where r=1;N Q
--- --------------------------------------------------------------------------------
001 n1;n2;n3;n4;n5
002 m1
003 t1;t2;t3;t4;t5;t6SQL>
--如果N有50多种，你要写到多长？
(
select no,q,rn,lead(rn) over(partition by no order by rn) rn1
from (
select no,q,row_number() over(order by no,q desc) rn from test
)
)
start with rn1 is null
connect by rn1 = prior rn
group by noSQL>
11 /NO RESULT
---------- --------------------------------------------------------------------------------
001 ;n1;n2;n3;n4;n5
002 ;m1
003 ;t1;t2;t3;t4;t5;t6
select no,max(sys_connect_by_path(q,';')) result from
(
select no,q,(row_number() over(order by no,q desc) + rank() over(order by no)) rn
from test
)
connect by rn-1 = prior rn
group by no
select no,max(sys_connect_by_path(q,';')) result
from (select no,q,
(row_number() over(order by no,q desc)
+ dense_rank() over(order by no)) rn,
max(q) over(partition by no) qs
from test
)
start with q = qs
connect by rn-1 = prior rn
group by no
下面例子中使用的表来自Oracle自带的HR用户下的表，如果没有安装该用户，可以在SYS用户下运行$ORACLE_HOME/demo/schema/human_resources/hr_main.sql来创建。
少数几个例子需要访问SH用户下的表，如果没有安装该用户，可以在SYS用户下运行$ORACLE_HOME/demo/schema/sales_history/sh_main.sql来创建。
如果未指明缺省是在HR用户下运行例子。
开窗函数的的理解：
开窗函数指定了分析函数工作的数据窗口大小，这个数据窗口大小可能会随着行的变化而变化，举例如下：
over（order by salary） 按照salary排序进行累计，order by是个默认的开窗函数
over（partition by deptno）按照部门分区
over（order by salary range between 50 preceding and 150 following）
每行对应的数据窗口是之前行幅度值不超过50，之后行幅度值不超过150
over（order by salary rows between 50 preceding and 150 following）
每行对应的数据窗口是之前50行，之后150行
over（order by salary rows between unbounded preceding and unbounded following）
每行对应的数据窗口是从第一行到最后一行，等效：
over（order by salary range between unbounded preceding and unbounded following）主要参考资料：《expert one-on-one》 Tom Kyte 《Oracle9i SQL Reference》第6章
AVG
功能描述：用于计算一个组和数据窗口内表达式的平均值。
SAMPLE：下面的例子中列c_mavg计算员工表中每个员工的平均薪水报告，该平均值由当前员工和与之具有相同经理的前一个和后一个三者的平均数得来；SELECT manager_id, last_name, hire_date, salary,
AVG(salary) OVER (PARTITION BY manager_id ORDER BY hire_date
ROWS BETWEEN 1 PRECEDING AND 1 FOLLOWING) AS c_mavg
FROM employees;MANAGER_ID LAST_NAME HIRE_DATE SALARY C_MAVG
---------- ------------------------- --------- ---------- ----------
100 Kochhar 21-SEP-89 17000 17000
100 De Haan 13-JAN-93 17000 15000
100 Raphaely 07-DEC-94 11000 11966.6667
100 Kaufling 01-MAY-95 7900 10633.3333
100 Hartstein 17-FEB-96 13000 9633.33333
100 Weiss 18-JUL-96 8000 11666.6667
100 Russell 01-OCT-96 14000 11833.3333
.
.
.
CORR
功能描述：返回一对表达式的相关系数，它是如下的缩写：
COVAR_POP(expr1,expr2)/STDDEV_POP(expr1)*STDDEV_POP(expr2))
从统计上讲，相关性是变量之间关联的强度，变量之间的关联意味着在某种程度
上一个变量的值可由其它的值进行预测。通过返回一个-1~1之间的一个数, 相关
系数给出了关联的强度，0表示不相关。
SAMPLE：下例返回1998年月销售收入和月单位销售的关系的累积系数（本例在SH用户下运行）SELECT t.calendar_month_number,
CORR (SUM(s.amount_sold), SUM(s.quantity_sold))
OVER (ORDER BY t.calendar_month_number) as CUM_CORR
FROM sales s, times t
WHERE s.time_id = t.time_id AND calendar_year = 1998
GROUP BY t.calendar_month_number
ORDER BY t.calendar_month_number;CALENDAR_MONTH_NUMBER CUM_CORR
--------------------- ----------
1
2 1
3 .994309382
4 .852040875
5 .846652204
6 .871250628
7 .910029803
8 .917556399
9 .920154356
10 .86720251
11 .844864765
12 .903542662
COVAR_POP
功能描述：返回一对表达式的总体协方差。
SAMPLE：下例CUM_COVP返回定价和最小产品价格的累积总体协方差SELECT product_id, supplier_id,
COVAR_POP(list_price, min_price)
OVER (ORDER BY product_id, supplier_id) AS CUM_COVP,
COVAR_SAMP(list_price, min_price)
OVER (ORDER BY product_id, supplier_id) AS CUM_COVS
FROM product_information p
WHERE category_id = 29
ORDER BY product_id, supplier_id;PRODUCT_ID SUPPLIER_ID CUM_COVP CUM_COVS
---------- ----------- ---------- ----------
1774 103088 0
1775 103087 1473.25 2946.5
1794 103096 1702.77778 2554.16667
1825 103093 1926.25 2568.33333
2004 103086 1591.4 1989.25
2005 103086 1512.5 1815
2416 103088 1475.97959 1721.97619
.
.
COVAR_SAMP
功能描述：返回一对表达式的样本协方差
SAMPLE：下例CUM_COVS返回定价和最小产品价格的累积样本协方差SELECT product_id, supplier_id,
COVAR_POP(list_price, min_price)
OVER (ORDER BY product_id, supplier_id) AS CUM_COVP,
COVAR_SAMP(list_price, min_price)
OVER (ORDER BY product_id, supplier_id) AS CUM_COVS
FROM product_information p
WHERE category_id = 29
ORDER BY product_id, supplier_id;PRODUCT_ID SUPPLIER_ID CUM_COVP CUM_COVS
---------- ----------- ---------- ----------
1774 103088 0
1775 103087 1473.25 2946.5
1794 103096 1702.77778 2554.16667
1825 103093 1926.25 2568.33333
2004 103086 1591.4 1989.25
2005 103086 1512.5 1815
2416 103088 1475.97959 1721.97619
.
.
功能描述：对一组内发生的事情进行累积计数，如果指定*或一些非空常数，count将对所有行计数，如果指定一个表达式，count返回表达式非空赋值的计数，当有相同值出现时，这些相等的值都会被纳入被计算的值；可以使用DISTINCT来记录去掉一组中完全相同的数据后出现的行数。
SAMPLE：下面例子中计算每个员工在按薪水排序中当前行附近薪水在[n-50,n+150]之间的行数，n表示当前行的薪水
例如，Philtanker的薪水2200，排在他之前的行中薪水大于等于2200-50的有1行，排在他之后的行中薪水小于等于2200＋150的行没有，所以count计数值cnt3为2（包括自己当前行）；cnt2值相当于小于等于当前行的SALARY值的所有行数SELECT last_name, salary, COUNT(*) OVER () AS cnt1,
COUNT(*) OVER (ORDER BY salary) AS cnt2,
COUNT(*) OVER (ORDER BY salary RANGE BETWEEN 50 PRECEDING
AND 150 FOLLOWING) AS cnt3 FROM employees;LAST_NAME SALARY CNT1 CNT2 CNT3
------------------------- ---------- ---------- ---------- ----------
Olson 2100 107 1 3
Markle 2200 107 3 2
Philtanker 2200 107 3 2
Landry 2400 107 5 8
Gee 2400 107 5 8
Colmenares 2500 107 11 10
Patel 2500 107 11 10
.
.
CUME_DIST
功能描述：计算一行在组中的相对位置，CUME_DIST总是返回大于0、小于或等于1的数，该数表示该行在N行中的位置。例如，在一个3行的组中，返回的累计分布值为1/3、2/3、3/3
SAMPLE：下例中计算每个工种的员工按薪水排序依次累积出现的分布百分比SELECT job_id, last_name, salary, CUME_DIST()
OVER (PARTITION BY job_id ORDER BY salary) AS cume_dist
FROM employees WHERE job_id LIKE 'PU%';JOB_ID LAST_NAME SALARY CUME_DIST
---------- ------------------------- ---------- ----------
PU_CLERK Colmenares 2500 .2
PU_CLERK Himuro 2600 .4
PU_CLERK Tobias 2800 .6
PU_CLERK Baida 2900 .8
PU_CLERK Khoo 3100 1
PU_MAN Raphaely 11000 1
DENSE_RANK
功能描述：根据ORDER BY子句中表达式的值，从查询返回的每一行，计算它们与其它行的相对位置。组内的数据按ORDER BY子句排序，然后给每一行赋一个号，从而形成一个序列，该序列从1开始，往后累加。每次ORDER BY表达式的值发生变化时，该序列也随之增加。有同样值的行得到同样的数字序号（认为null时相等的）。密集的序列返回的时没有间隔的数
SAMPLE：下例中计算每个员工按部门分区再按薪水排序，依次出现的序列号（注意与RANK函数的区别）SELECT d.department_id , e.last_name, e.salary, DENSE_RANK()
OVER (PARTITION BY e.department_id ORDER BY e.salary) as drank
FROM employees e, departments d
WHERE e.department_id = d.department_id
AND d.department_id IN ('60', '90'); DEPARTMENT_ID LAST_NAME SALARY DRANK
------------- ------------------------- ---------- ----------
60 Lorentz 4200 1
60 Austin 4800 2
60 Pataballa 4800 2
60 Ernst 6000 3
60 Hunold 9000 4
90 Kochhar 17000 1
90 De Haan 17000 1
90 King 24000 2
FIRST
功能描述：从DENSE_RANK返回的集合中取出排在最前面的一个值的行（可能多行，因为值可能相等），因此完整的语法需要在开始处加上一个集合函数以从中取出记录
SAMPLE：下面例子中DENSE_RANK按部门分区，再按佣金commission_pct排序，FIRST取出佣金最低的对应的所有行，然后前面的MAX函数从这个集合中取出薪水最低的值；LAST取出佣金最高的对应的所有行，然后前面的MIN函数从这个集合中取出薪水最高的值
SELECT last_name, department_id, salary,
MIN(salary) KEEP (DENSE_RANK FIRST ORDER BY commission_pct)
OVER (PARTITION BY department_id) "Worst",
MAX(salary) KEEP (DENSE_RANK LAST ORDER BY commission_pct)
OVER (PARTITION BY department_id) "Best"
FROM employees
WHERE department_id in (20,80)
ORDER BY department_id, salary;LAST_NAME DEPARTMENT_ID SALARY Worst Best
------------------------- ------------- ---------- ---------- ----------
Fay 20 6000 6000 13000
Hartstein 20 13000 6000 13000
Kumar 80 6100 6100 14000
Banda 80 6200 6100 14000
Johnson 80 6200 6100 14000
Ande 80 6400 6100 14000
Lee 80 6800 6100 14000
Tuvault 80 7000 6100 14000
Sewall 80 7000 6100 14000
Marvins 80 7200 6100 14000
Bates 80 7300 6100 14000
.
.
.
FIRST_VALUE
功能描述：返回组中数据窗口的第一个值。
SAMPLE：下面例子计算按部门分区按薪水排序的数据窗口的第一个值对应的名字，如果薪水的第一个值有多个，则从多个对应的名字中取缺省排序的第一个名字SELECT department_id, last_name, salary, FIRST_VALUE(last_name)
OVER (PARTITION BY department_id ORDER BY salary ASC ) AS lowest_sal
FROM employees
WHERE department_id in(20,30);DEPARTMENT_ID LAST_NAME SALARY LOWEST_SAL
------------- ------------------------- ---------- --------------
20 Fay 6000 Fay
20 Hartstein 13000 Fay
30 Colmenares 2500 Colmenares
30 Himuro 2600 Colmenares
30 Tobias 2800 Colmenares
30 Baida 2900 Colmenares
30 Khoo 3100 Colmenares
30 Raphaely 11000 Colmenares
经常看到你的回帖，感觉你比较善用ORACLE分析函数，有机会认识一下，
QQ77443484
最近也看到一个定义的SQL函数，比较方便能和GROUP BY 连用
分享一下Oracle9i 以后的活，创建函数create or replace type ListImpl as object
(
listv VARCHAR2(4000),static function ODCIAggregateInitialize(sctx IN OUT ListImpl)
return number,member function ODCIAggregateIterate(self IN OUT ListImpl,
value IN varchar) return number,member function ODCIAggregateTerminate(self IN OUT ListImpl,
returnValue OUT varchar, flags IN number)
return number,member function ODCIAggregateMerge(self IN OUT ListImpl,
ctx2 IN ListImpl) return number
);
/
show errorcreate or replace type body ListImpl is
static function ODCIAggregateInitialize(sctx IN OUT ListImpl)
return number is
begin
sctx := ListImpl(null);
return ODCIConst.Success;
end;member function ODCIAggregateIterate(self IN OUT ListImpl,
value IN varchar) return number is
begin
if (self.listv is null) then
self.listv := rtrim(value);
elsif (value is not null) then
self.listv := self.listv||','||rtrim(value);
end if;
return ODCIConst.Success;
end;member function ODCIAggregateTerminate(self IN OUT ListImpl,
returnValue OUT varchar, flags IN number) return number is
begin
returnValue := self.listv;
return ODCIConst.Success;
end;member function ODCIAggregateMerge(self IN OUT ListImpl,
ctx2 IN ListImpl) return number is
begin
if (self.listv is null) then
self.listv := ctx2.listv;
elsif (ctx2.listv is not null) then
self.listv := self.listv||','||ctx2.listv;
end if;
return ODCIConst.Success;
end;
end;
/
show errorcreate or replace function LIST(input varchar) return varchar
parallel_enable aggregate using ListImpl;
/
show error
然后，用list函数select NAME,list(CONTENT) CONTENT from "表A" group by NAME;
全靠斑竹整理了。
http://www.cnoug.org/viewthread.php?tid=38387
1.AVG
2.COUNT
3.DENSE_RANK、RANK、PERCENT_RANK
4.FIRST_VALUE、LAST_VALUE
5.LAG、LEAD
6.MAX、MIN
7.NTILE
8.ROW_NUMBER
9.SUM
我接着waterfirer(水清)的补充下其他的几个函数用法：LAST_VALUE
功能描述：返回组中数据窗口的第一个值。
用法和first_value一样，不用多说了，不过LAST_VALUE函数本身好像有问题，结果和预期不一样，至少9i是这样，大家可以去试验一下LAG、LEAD
功能描述：可以访问结果集中的其它行而不用进行自连接。它允许去处理游标，就好像游标是一个数组一样。在给定组中可参考当前行之前的行，这样就可以从组中与当前行一起选择以前的行。Offset是一个正整数，其默认值为1，若索引超出窗口的范围，就返回默认值（默认返回的是组中第一行），其相反的函数是LEAD，LEAD是返回当前行以后的行
SAMPLE：下面的例子中列prev_sal返回按hire_date排序的前1行的salary值SELECT last_name, hire_date, salary,
LAG(salary, 1, 0) OVER (ORDER BY hire_date) AS prev_sal
FROM employees
WHERE job_id = 'PU_CLERK';LAST_NAME HIRE_DATE SALARY PREV_SAL
------------------------- ---------- ---------- ----------
Khoo 18-5月 -95 3100 0
Tobias 24-7月 -97 2800 3100
Baida 24-12月-97 2900 2800
Himuro 15-11月-98 2600 2900
Colmenares 10-8月 -99 2500 2600MAX 、MIN
功能描述：在一个组中的数据窗口中查找表达式的最大值和最小值。
SAMPLE：下面例子中dept_max返回当前行所在部门的最大薪水值SELECT department_id, last_name, salary,
MAX(salary) OVER (PARTITION BY department_id) AS dept_max
FROM employees WHERE department_id in (10,20,30);DEPARTMENT_ID LAST_NAME SALARY DEPT_MAX
------------- ------------------------- ---------- ----------
10 Whalen 4400 4400
20 Hartstein 13000 13000
20 Fay 6000 13000
30 Raphaely 11000 11000
30 Khoo 3100 11000
30 Baida 2900 11000
30 Tobias 2800 11000
30 Himuro 2600 11000
30 Colmenares 2500 11000NTILE
功能描述：将一个组分为"表达式"的散列表示，例如，如果表达式=4，则给组中的每一行分配一个数（从1到4），如果组中有20行，则给前5行分配1，给下5行分配2等等。如果组的基数不能由表达式值平均分开，则对这些行进行分配时，组中就没有任何percentile的行数比其它percentile的行数超过一行，最低的percentile是那些拥有额外行的percentile。例如，若表达式=4，行数=21，则percentile=1的有5行，percentile=2的有5行等等。
SAMPLE：下例中把6行数据分为4份SELECT last_name, salary,
NTILE(4) OVER (ORDER BY salary DESC) AS quartile FROM employees
WHERE department_id = 100;LAST_NAME SALARY QUARTILE
------------------------- ---------- ----------
Greenberg 12000 1
Faviet 9000 1
Chen 8200 2
Urman 7800 2
Sciarra 7700 3
Popp 6900 4RANK
功能描述：根据ORDER BY子句中表达式的值，从查询返回的每一行，计算它们与其它行的相对位置。组内的数据按ORDER BY子句排序，然后给每一行赋一个号，从而形成一个序列，该序列从1开始，往后累加。每次ORDER BY表达式的值发生变化时，该序列也随之增加。有同样值的行得到同样的数字序号（认为null时相等的）。然而，如果两行的确得到同样的排序，则序数将随后跳跃。若两行序数为1，则没有序数2，序列将给组中的下一行分配值3，DENSE_RANK则没有任何跳跃。
SAMPLE：下例中计算每个员工按部门分区再按薪水排序，依次出现的序列号（注意与DENSE_RANK函数的区别）SELECT d.department_id , e.last_name, e.salary, RANK()
OVER (PARTITION BY e.department_id ORDER BY e.salary) as drank
FROM employees e, departments d
WHERE e.department_id = d.department_id
AND d.department_id IN ('60', '90');DEPARTMENT_ID LAST_NAME SALARY DRANK
------------- ------------------------- ---------- ----------
60 Lorentz 4200 1
60 Austin 4800 2
60 Pataballa 4800 2
60 Ernst 6000 4
60 Hunold 9000 5
90 Kochhar 17000 1
90 De Haan 17000 1
90 King 24000 3PERCENT_RANK
功能描述：和CUME_DIST（累积分配）函数类似，对于一个组中给定的行来说，在计算那行的序号时，先减1，然后除以n-1（n为组中所有的行数）。该函数总是返回0～1（包括1）之间的数。
SAMPLE：下例中如果Khoo的salary为2900，则pr值为0.6，因为RANK函数对于等值的返回序列值是一样的SELECT department_id, last_name, salary,
PERCENT_RANK()
OVER (PARTITION BY department_id ORDER BY salary) AS pr
FROM employees
WHERE department_id < 50
ORDER BY department_id,salary;DEPARTMENT_ID LAST_NAME SALARY PR
------------- ------------------------- ---------- ----------
10 Whalen 4400 0
20 Fay 6000 0
20 Hartstein 13000 1
30 Colmenares 2500 0
30 Himuro 2600 0.2
30 Tobias 2800 0.4
30 Baida 2900 0.6
30 Khoo 3100 0.8
30 Raphaely 11000 1
40 Mavris 6500 0ROW_NUMBER
功能描述：返回有序组中一行的偏移量，从而可用于按特定标准排序的行号。
SAMPLE：下例返回每个员工再在每个部门中按员工号排序后的顺序号SELECT department_id, last_name, employee_id, ROW_NUMBER()
OVER (PARTITION BY department_id ORDER BY employee_id) AS emp_id
FROM employees
WHERE department_id < 50;DEPARTMENT_ID LAST_NAME EMPLOYEE_ID EMP_ID
------------- ------------------------- ----------- ----------
10 Whalen 200 1
20 Hartstein 201 1
20 Fay 202 2
30 Raphaely 114 1
30 Khoo 115 2
30 Baida 116 3
30 Tobias 117 4
30 Himuro 118 5
30 Colmenares 119 6
40 Mavris 203 1SUM
功能描述：该函数计算组中表达式的累积和。
SAMPLE：下例计算同一经理下员工的薪水累积值SELECT manager_id, last_name, salary,
SUM (salary) OVER (PARTITION BY manager_id ORDER BY salary
RANGE UNBOUNDED PRECEDING) l_csum
FROM employees
WHERE manager_id in (101,103,108);MANAGER_ID LAST_NAME SALARY L_CSUM
---------- ------------------------- ---------- ----------
101 Whalen 4400 4400
101 Mavris 6500 10900
101 Baer 10000 20900
101 Greenberg 12000 44900
101 Higgins 12000 44900
103 Lorentz 4200 4200
103 Austin 4800 13800
103 Pataballa 4800 13800
103 Ernst 6000 19800
108 Popp 6900 6900
108 Sciarra 7700 14600
108 Urman 7800 22400
108 Chen 8200 30600
108 Faviet 9000 39600
analytic_function::=
analytic_function ([ arguments ]) OVER (analytic_clause)analytic_clause::=
[ query_partition_clause ]
[ order_by_clause [ windowing_clause ] ]query_partition_clause::=
PARTITION BY value_expr [, value_expr]...order_by_clause::=
ORDER [SIBLINGS] BY
{ expr | position | c_alias } [ ASC | DESC ] [ NULLS FIRST | NULLS LAST ]
[, { expr | position | c_alias } [ ASC | DESC ] [ NULLS FIRST | NULLS LAST ]]...windowing_clause::=
{ ROWS | RANGE }
{ BETWEEN
{ UNBOUNDED PRECEDING
| CURRENT ROW
| value_expr { PRECEDING | FOLLOWING }
}
AND
{ UNBOUNDED FOLLOWING
| CURRENT ROW
| value_expr { PRECEDING | FOLLOWING }
}
| { UNBOUNDED PRECEDING
| CURRENT ROW
| value_expr PRECEDING
}
}