标签:style blog http io ar color os 使用 sp
本题不是一道直接的字符串拆解,
应用场景如下,表中有一个字段,是表示事件受影响的国家集合,使用逗号进行分隔,不幸的是,居然发现有些国家本身就带有逗号,这样在规范化的时候,如何准确地找到这些国家呢?
以下的代码是有一定限制的。但基本上够用。
下面的代码使用到了分析函数lag和lead还有cte,sqlserver2012及其以后的版本都支持,oracle好像10g以上就支持了。
主要思路:
字符串的分解,可以使用数字辅助表,然后cross join刷副本,然后根本分隔符出现的位置然后切豁字符串拆解到我们需要的东东。(解决方案中我使用的递归CTE来处理找到对应的位置)
现在还需要多加一步,就是对拆解的部分进行验证和去重不符合要求的那一部。
使用LAG和LEAD的好处,就是不需要再用自连接去找到对应的下一条数据了。
本题的解题原则是如何长项能连接到正确的国家,则取长项的,否则取短项的。
代码如下:
--准备示例表与数据 drop table my_countries; drop table valid_country; create table my_countries(rid int,country_name_cc varchar(200)); insert into my_countries(rid,country_name_cc) values(1,‘china,test, public of‘); insert into my_countries(rid,country_name_cc) values(2,‘us, public of,china,Evan, public of‘); create table valid_country(cid int, country_name varchar(30)); insert into valid_country(cid,country_name) values(1,‘china‘); insert into valid_country(cid,country_name) values(2,‘test, public of‘); insert into valid_country(cid,country_name) values(3,‘Evan, public of‘); insert into valid_country(cid,country_name) values(4,‘us, public of‘); insert into valid_country(cid,country_name) values(5,‘Evan‘); --select * from my_countries; --select * from valid_country;
正确的结果是:
WITH SPLIT_COUNTRY AS ( SELECT RID, 1 AS LVL, 1 AS STARTPOS, CHARINDEX(‘,‘,COUNTRY_NAME_CC+‘,‘)-1 AS ENDPOS FROM MY_COUNTRIES UNION ALL SELECT SC.RID, LVL+1 AS LVL, ENDPOS+2, CHARINDEX(‘,‘,COUNTRY_NAME_CC+‘,‘,ENDPOS+2)-1 FROM MY_COUNTRIES CC JOIN SPLIT_COUNTRY SC ON CC.RID=SC.RID WHERE CHARINDEX(‘,‘,CC.COUNTRY_NAME_CC+‘,‘,ENDPOS+2)>0 ) ,CTE_COUNTRY AS ( SELECT RID,LVL,STARTPOS,ENDPOS,LEAD(ENDPOS,1) OVER(PARTITION BY RID ORDER BY LVL) AS NEXTENDPOS FROM SPLIT_COUNTRY ) ,CTE AS ( SELECT MC.RID,SC.LVL, CASE WHEN NEXTENDPOS IS NOT NULL AND EXISTS (SELECT * FROM VALID_COUNTRY VC WHERE VC.COUNTRY_NAME = SUBSTRING(COUNTRY_NAME_CC,STARTPOS,NEXTENDPOS-STARTPOS+1)) THEN SUBSTRING(COUNTRY_NAME_CC,STARTPOS,NEXTENDPOS-STARTPOS+1) ELSE SUBSTRING(MC.COUNTRY_NAME_CC,STARTPOS,ENDPOS-STARTPOS+1) END AS COUNTRY FROM MY_COUNTRIES MC JOIN CTE_COUNTRY SC ON MC.RID=SC.RID ) ,CHECK_VALID AS ( SELECT CASE WHEN CHARINDEX(‘,‘,LAG(COUNTRY,1) OVER(PARTITION BY RID ORDER BY LVL))>0 THEN 0 ELSE 1 END AS ISVALID, * FROM CTE ) SELECT CV.RID,CV.COUNTRY,VC.CID FROM CHECK_VALID CV JOIN VALID_COUNTRY VC ON CV.COUNTRY = VC.COUNTRY_NAME AND ISVALID=1 ORDER BY RID;
另一种方案,在第一种的基础上稍加修改:
WITH SPLIT_COUNTRY AS ( SELECT RID, 1 AS LVL, 1 AS STARTPOS, CHARINDEX(‘,‘,COUNTRY_NAME_CC+‘,‘)-1 AS ENDPOS FROM MY_COUNTRIES UNION ALL SELECT SC.RID, LVL+1 AS LVL, ENDPOS+2, CHARINDEX(‘,‘,COUNTRY_NAME_CC+‘,‘,ENDPOS+2)-1 FROM MY_COUNTRIES CC JOIN SPLIT_COUNTRY SC ON CC.RID=SC.RID WHERE CHARINDEX(‘,‘,CC.COUNTRY_NAME_CC+‘,‘,ENDPOS+2)>0 ) ,CTE_COUNTRY AS ( SELECT RID,LVL,STARTPOS,ENDPOS,LEAD(ENDPOS,1) OVER(PARTITION BY RID ORDER BY LVL) AS NEXTENDPOS FROM SPLIT_COUNTRY ) ,CTE AS ( SELECT MC.RID,SC.LVL, SUBSTRING(MC.COUNTRY_NAME_CC,STARTPOS,ENDPOS-STARTPOS+1) AS COUNTRY, SUBSTRING(COUNTRY_NAME_CC,STARTPOS,NEXTENDPOS-STARTPOS+1) AS COUNTRY2 FROM MY_COUNTRIES MC JOIN CTE_COUNTRY SC ON MC.RID=SC.RID ) SELECT CTE.RID,VC.COUNTRY_NAME,VC.CID FROM CTE JOIN VALID_COUNTRY VC ON (CASE WHEN EXISTS(SELECT * FROM VALID_COUNTRY X WHERE X.COUNTRY_NAME=CTE.COUNTRY2) THEN CTE.COUNTRY2 ELSE CTE.COUNTRY END) = VC.COUNTRY_NAME ;
标签:style blog http io ar color os 使用 sp
原文地址:http://www.cnblogs.com/huaxiaoyao/p/4114761.html