Oracle分析函数ROW_NUMBER()|RANK()|LAG()使用详解

Oracle分析函数RANK()|ROW_NUMBER()|LAG()使用详解

ROW_NUMBER()函数：
row_number()和rownum差不多，功能更强一点（可以在各个分组内从1开时排序），效率更高（？效率貌似更差些，实际测验时）。

ROW_NUMBER()的使用方法：

ROW_NUMBER() OVER (PARTITION BY COL1 ORDER BY COL2)

详细说明：

根据COL1分组

在分组内部根据 COL2排序

而这个值就表示每组内部排序后的顺序编号（组内连续的唯一的）

ROW_NUMBER（）语法如下： 1、row_number() over(order by column desc)先对列column按照降序，再为每条记录返回一个序列号： SELECT D.*, ROW_NUMBER() OVER(ORDER BY D.R_OPATE_NUM DESC) AS INX FROM REPORT_DATA D

ROW_NUMBER（）语法如下：

1、row_number() over(order by column desc)先对列column按照降序，再为每条记录返回一个序列号：
SELECT D.*, ROW_NUMBER() OVER(ORDER BY D.R_OPATE_NUM DESC) AS INX FROM REPORT_DATA D

[sql]
view plaincopyprint?

2、row_number() over(partition by column1 order by column2 asc) 先按照column1分组，再对分组后的数据进行以column2升序排列 select personalid,ct_smp_type,row_number() over(partition by personalid order by ct_smp_type asc) rn from neogoodsrule 语法1的具体实例：获取前100名人员的排名信息，如下

2、row_number() over(partition by column1 order by column2 asc) 先按照column1分组，再对分组后的数据进行以column2升序排列
select personalid,ct_smp_type,row_number() over(partition by personalid order by ct_smp_type asc) rn from neogoodsrule

语法1的具体实例：获取前100名人员的排名信息，如下

[sql]
view plaincopyprint?

<PRE class=sql name="code">WITH REPORT_DATA AS
(SELECT DW.DWID,DW.DWMC,JY.JYXM,JY.JH,RMPC.R_OPATE_NUM
FROM REPORT_MONTH_PERSON_COUNT RMPC,JWT_JYXX JY,T_DWXX DW
WHERE RMPC.JYID = JY.JYUSERID
AND JY.SSDW = DW.DWID
AND RMPC.R_YEAR = 2013
AND RMPC.R_MONTH = 6
AND JY.SSDW LIKE '4102%'
ORDER BY RMPC.R_OPATE_NUM DESC)
SELECT B.*
FROM (SELECT D.*, ROW_NUMBER() OVER(ORDER BY D.R_OPATE_NUM DESC) AS INX
FROM REPORT_DATA D
) B
WHERE B.INX <=100
ORDER BY B.INX </PRE><BR>
<PRE></PRE>
<BR>

<div class="dp-highlighter bg_sql"><div class="bar"><div class="tools"><strong>[sql]</strong> <a target=_blank target="_blank" class="ViewSource" title="view plain" href="http://blog.csdn.net/liqingan880102/article/details/9924257#">view plain</a><a target=_blank target="_blank" class="CopyToClipboard" title="copy" href="http://blog.csdn.net/liqingan880102/article/details/9924257#">copy</a><a target=_blank target="_blank" class="PrintSource" title="print" href="http://blog.csdn.net/liqingan880102/article/details/9924257#">print</a><a target=_blank target="_blank" class="About" title="?" href="http://blog.csdn.net/liqingan880102/article/details/9924257#">?</a></div></div><ol class="dp-sql"><li class="alt"><span><span class="keyword">WITH</span><span> REPORT_DATA </span><span class="keyword">AS</span><span>  </span></span></li><li><span>  (</span><span class="keyword">SELECT</span><span> DW.DWID,DW.DWMC,JY.JYXM,JY.JH,RMPC.R_OPATE_NUM   </span></li><li class="alt"><span>  </span><span class="keyword">FROM</span><span> REPORT_MONTH_PERSON_COUNT RMPC,JWT_JYXX JY,T_DWXX DW  </span></li><li><span>  </span><span class="keyword">WHERE</span><span> RMPC.JYID = JY.JYUSERID  </span></li><li class="alt"><span>  </span><span class="op">AND</span><span> JY.SSDW = DW.DWID  </span></li><li><span>  </span><span class="op">AND</span><span> RMPC.R_YEAR = 2013  </span></li><li class="alt"><span>  </span><span class="op">AND</span><span> RMPC.R_MONTH = 6  </span></li><li><span>  </span><span class="op">AND</span><span> JY.SSDW </span><span class="op">LIKE</span><span> </span><span class="string">'4102%'</span><span>  </span></li><li class="alt"><span>  </span><span class="keyword">ORDER</span><span> </span><span class="keyword">BY</span><span> RMPC.R_OPATE_NUM </span><span class="keyword">DESC</span><span>)  </span></li><li><span></span><span class="keyword">SELECT</span><span> B.*  </span></li><li class="alt"><span></span><span class="keyword">FROM</span><span> (</span><span class="keyword">SELECT</span><span> D.*, ROW_NUMBER() OVER(</span><span class="keyword">ORDER</span><span> </span><span class="keyword">BY</span><span> D.R_OPATE_NUM </span><span class="keyword">DESC</span><span>) </span><span class="keyword">AS</span><span> INX  </span></li><li><span>       </span><span class="keyword">FROM</span><span> REPORT_DATA D  </span></li><li class="alt"><span>       ) B  </span></li><li><span></span><span class="keyword">WHERE</span><span> B.INX <=100  </span></li><li class="alt"><span></span><span class="keyword">ORDER</span><span> </span><span class="keyword">BY</span><span> B.INX   </span></li></ol></div><pre class="sql" name="code" style="">WITH REPORT_DATA AS
(SELECT DW.DWID,DW.DWMC,JY.JYXM,JY.JH,RMPC.R_OPATE_NUM
FROM REPORT_MONTH_PERSON_COUNT RMPC,JWT_JYXX JY,T_DWXX DW
WHERE RMPC.JYID = JY.JYUSERID
AND JY.SSDW = DW.DWID
AND RMPC.R_YEAR = 2013
AND RMPC.R_MONTH = 6
AND JY.SSDW LIKE '4102%'
ORDER BY RMPC.R_OPATE_NUM DESC)
SELECT B.*
FROM (SELECT D.*, ROW_NUMBER() OVER(ORDER BY D.R_OPATE_NUM DESC) AS INX
FROM REPORT_DATA D
) B
WHERE B.INX <=100
ORDER BY B.INX

以下为通过row_number() over(...)删除重复数据的例子，仅供参考：

delete from acc_fundnav

where rowid in (select row1

from (select rowid row1,

row_number() over(partition by HOST_ID order by rowid) lev

from acc_fundnav)

where lev > 1)

RANK()：排序的时候跟派名次一样，可以并列2个第一名之后 是第3名

LAG：分组排序后 ，组内后面一条记录减前面一条记录的差，第一条可返回 NULL

BTW: EXPERT ONE ON ONE 上讲的最详细,还有很多相关特性，文档看起来比较费劲

rank()是跳跃排序，有两个第二名时接下来就是第四名（同样是在各个分组内）

dense_rank()l是连续排序，有两个第二名时仍然跟着第三名。

相比之下row_number是没有重复值的

lag（arg1,arg2,arg3):

arg1是从其他行返回的表达式

arg2是希望检索的当前行分区的偏移量。是一个正的偏移量，时一个往回检索以前的行的数目。

arg3是在arg2表示的数目超出了分组的范围时返回的值。

另见：《oracle分析函数技术详解（如 over()）》

oracle分析函数技术详解（配上开窗函数over()）

Oracle的LAG和LEAD分析函数