求某列按分组连续编号的最大区间差异值

来源:互联网 发布:gtp7吉他谱软件下载 编辑:程序博客网 时间:2024/04/29 09:30

求一句SQL,求连续记录的最大个数
create table #t 
(_id varchar(3), _date int) 

insert #t select '001',20100101 
union all select '001',20100102 
union all select '002',20100103 
union all select '002',20100101 
union all select '002',20100102 
union all select '002',20100104
union all select '001',20100105 
union all select '001',20100106 
union all select '001',20100107 
union all select '002',20100105 
union all select '002',20100107 
union all select '002',20100108 
union all select '003',20100101
union all select '003',20100102
union all select '003',20100104
union all select '003',20100105
union all select '003',20100106
union all select '003',20100108
union all select '003',20100109
union all select '003',20100111

要求: 求出每个ID 的最大连续记录的个数。记录有几百万, 效率要好一点。 谢谢!
结果为:001最大连续为从20100105到20100107,数量为3, 
  002最大连续为从20100101到20100104,数量为4, 
  003最大连续为从20100104到20100106,数量为3
===================
001 3
002 4
003 3
===================
如果不影响效率的话,最好把时间段也带上:
=================================
001 3 20100105 20100107
002 4 20100101 20100104
003 3 20100104 20100106
=================================
要能加条件, 比如查询连续数量大于3的,结果只有一条
======================================
002 4 20100101 20100104

 

 

--SQL2000

 

if OBJECT_ID('tempdb..#t') is not null drop table #t
go
create table #t 
(id
varchar(3), date int

insert #t select '001',20100101 
union all select '001',20100102 
union all select '002',20100103 
union all select '002',20100101 
union all select '002',20100102 
union all select '002',20100104
union all select '001',20100105 
union all select '001',20100106 
union all select '001',20100107 
union all select '002',20100105 
union all select '002',20100107 
union all select '002',20100108 
union all select '003',20100101
union all select '003',20100102
union all select '003',20100104
union all select '003',20100105
union all select '003',20100106
union all select '003',20100108
union all select '003',20100109
union all select '003',20100111

--1.
alter table #t add d int
go
declare @i int, @j int,@k int
update #t set
    d
= @j,
   
@j = case when @i = id then @j else isnull(@j,0)+1 end,
   
@i = id
--2.

if OBJECT_ID('tempdb..#') is not null drop table #
go
select *,0 c into # from #t order by d,date
go
declare @i int, @j int,@k int
update # set
    c
= @j,
   
@j = case when @i = id and @k=date-1 then @j +1 else 1 end,
   
@i = id,
   
@k=dateselect ID,MAX(c) 最大连续数量 from # group by ID

/*
ID   最大连续数量
---- -----------
001  3
002  4
003  3

(3 行受影响)
*/

 

 

--SQL2005:

 

 

--1.更新分区段的行
declare @i int, @j int,@k int
update #t set d = @j,@j = case when @i = _id then @j else isnull(@j,0)+1 end,@i = _id
--2.查询
;with t1 as
(
   
select _id,d, _date1 = _date - row_number() over (partition by d order by _date),_date
   
from #t
)
,t2
as
(
   
select _id, d, cnt = count(1),mindate=MIN(_date),maxdate= MAX(_date)
   
from t1
   
group by _id, d, _date1) select _id, maxcnt = max(cnt),
    mindate
=(select mindate from t2 where cnt=max(t.cnt) and _id=t._id),
    maxdate
=(select maxdate from t2 where cnt=max(t.cnt) and _id=t._id)
from t2 t
group by _id

/*
_id  maxcnt      mindate     maxdate
---- ----------- ----------- -----------
001  3           20100105    20100107
002  4           20100101    20100104
003  3           20100104    20100106

(3 行受影响)
*/
原贴:http://topic.csdn.net/u/20100402/17/1565fb1e-f1e0-4961-b0c4-ae41f95c550f.html?21618

原创粉丝点击