Transact-SQL统计某字段中的值第一次出现后的2小时内出现的次数

Posted lonelyxmas

tags:

篇首语：本文由小常识网(cha138.com)小编为大家整理，主要介绍了Transact-SQL统计某字段中的值第一次出现后的2小时内出现的次数相关的知识，希望对你有一定的参考价值。

原文:【Transact-SQL】统计某字段中的值第一次出现后的2小时内出现的次数

 

table1
name               createdate
a              2011-03-01 10:00:00
a              2011-03-01 11:00:00
a              2011-03-01 14:00:00
b              2011-03-01 13:00:00
b              2011-03-01 13:20:00
b              2011-03-01 14:00:00
 
查询结果为
name             createdate                    count
a             2011-03-01 10:00:00               2
a             2011-03-01 14:00:00               1
b             2011-03-01 13:00:00               3

就相当于是统计name字段中的值在第一次出现后的2小时内，总共出现了几次？

 

这个是网上的解答：

  1. declare @table1 table(name nvarchar,createdate smalldatetime)
  2.  
  3. insert into @table1
  4. select ‘a‘,‘2011-03-01 10:00:00‘
  5. union all select ‘a‘,‘2011-03-01 11:00:00‘
  6. union all select ‘a‘,‘2011-03-01 14:00:00‘
  7. union all select ‘b‘,‘2011-03-01 13:00:00‘
  8. union all select ‘b‘,‘2011-03-01 13:20:00‘
  9. union all select ‘b‘,‘2011-03-01 14:00:00‘
  10.  
  11. select name,
  12.        createdate,
  13.        (select count(createdate)
  14.         from @table1 b 
  15.         where a.name=b.name and 
  16.               a.createdate<=b.createdate and 
  17.               dateadd(hh,2,a.createdate) >= b.createdate
  18.        ) as count
  19.  
  20. from @table1 a 
  21. where not exists 
  22.          (select 1 from 
  23.           @table1 b 
  24.           where a.name=b.name and 
  25.                 a.createdate>b.createdate and 
  26.                 a.createdate<dateadd(hh,2,b.createdate)) 
  27. group by name,createdate


但是这个解答其实是有问题的，当把临时表中的第3条数据的createdate改为‘2011-03-01 12:00:00‘，那么显示的结果是：

name        createdate                           count
a                2011-03-01 10:00:00        3
b                2011-03-01 13:00:00        3

在其中没有包括createdate为‘2011-03-01 12:00:00‘的记录，因为这个时间到为‘2011-03-01 10:00:00‘是超过2个小时了，也就是说为‘2011-03-01 10:00:00‘是第一个出现时间，到为‘2011-03-01 11:59:59‘为止，接下来应该是从‘2011-03-01 12:00:00‘开始的下个区间了，而这里显然是有问题的。

以下是我写的解法，虽然效率不是太高，但是能解决这个问题：

  1. declare @table1 table(name nvarchar,createdate smalldatetime)
  2.  
  3. insert into @table1
  4. select ‘a‘,‘2011-03-01 10:00:00‘
  5. union all select ‘a‘,‘2011-03-01 11:00:00‘
  6. union all select ‘a‘,‘2011-03-01 12:00:00‘
  7. union all select ‘b‘,‘2011-03-01 13:10:00‘
  8. union all select ‘b‘,‘2011-03-01 13:20:00‘
  9. union all select ‘b‘,‘2011-03-01 14:30:00‘
  10. union all select ‘b‘,‘2011-03-01 15:15:00‘
  11. union all select ‘b‘,‘2011-03-01 16:00:00‘
  12. union all select ‘b‘,‘2011-03-01 17:00:00‘
  13.  
  14.  
  15.   
  16. ;with aa  --按照name分区，同时按照createdate排序编号
  17. as
  18. (
  19. select name,
  20.        createdate,
  21.        ROW_NUMBER() over(partition by name
  22.                              order by createdate) as k1
  23. from @table1 
  24. ),
  25.  
  26. r
  27. as
  28. (
  29. select v.name,
  30.        starts=v.k1,  --区间开始的编号
  31.        
  32.        ends=isnull(
  33.                min(case when v.k1<a.k1 
  34.                              and DATEADD(hour,2,v.createdate) <= a.createdate 
  35.                              then a.k1 
  36.                          else null 
  37.                    end)-1,
  38.  
  39.                max(case when v.k1<a.k1 
  40.                              then a.k1 
  41.                         else v.k1 
  42.                    end)
  43.              ),     --区间结尾的编号      
  44.                              
  45.            isnull(
  46.                min(case when v.k1<a.k1 
  47.                              and DATEADD(hour,2,v.createdate) <= a.createdate 
  48.                              then a.k1 
  49.                          else null 
  50.                    end)-1,
  51.  
  52.                max(case when v.k1<a.k1 
  53.                              then a.k1 
  54.                         else v.k1 
  55.                    end)
  56.              ) - v.k1 as diff     --区间结尾编号与区间开始编号之间的差值
  57.                               
  58. from aa v  
  59. inner join aa a 
  60.         on v.name = a.name  --只关联name相等的
  61. group by v.name,
  62.          v.k1
  63. having isnull(
  64.                min(case when v.k1<a.k1 
  65.                              and DATEADD(hour,2,v.createdate) <= a.createdate 
  66.                              then a.k1 
  67.                          else null 
  68.                    end)-1,
  69.  
  70.                max(case when v.k1<a.k1 
  71.                              then a.k1 
  72.                         else v.k1 
  73.                    end)
  74.              )  >=v.k1   
  75.         and            
  76.         isnull(
  77.                max(case when v.k1>a.k1 and 
  78.                              DATEADD(hour,-2,v.createdate) >= a.createdate
  79.                              then v.k1 - 1 
  80.                         else null 
  81.                    end) + 1,
  82.  
  83.                min(case when v.k1>a.k1 
  84.                              then a.k1 
  85.                         else v.k1 
  86.                    end)
  87.               ) = v.k1
  88. )
  89.  
  90. --select * from r
  91.  
  92.  
  93. select aa.name,
  94.        aa.createdate,
  95.        diff + 1
  96. from r
  97. inner join aa 
  98.         on aa.name = r.name
  99.            and aa.k1 =r.starts
  100. where not exists
  101. 		(select 1
  102. 		 from r rr
  103. 		 where rr.name = r.name and
  104. 		       rr.starts <> r.starts and
  105. 		       rr.starts < r.starts and
  106. 		       (rr.ends = r.ends or
  107. 		        rr.ends = r.starts)
  108. 		)

 不过发现我的这个解法也是有问题的，最大的问题在与不能准确的确定上限在那里，还得继续考虑问题的解法。

下面这个也是有问题的：

  1. declare @table1 table(name nvarchar,createdate smalldatetime)
  2.  
  3. insert into @table1
  4. select ‘a‘,‘2011-03-01 10:00:00‘
  5. union all select ‘a‘,‘2011-03-01 11:00:00‘
  6. union all select ‘a‘,‘2011-03-01 12:00:00‘
  7.  
  8. --union all select ‘b‘,‘2011-03-01 13:10:00‘
  9. --union all select ‘b‘,‘2011-03-01 13:20:00‘
  10. --union all select ‘b‘,‘2011-03-01 14:30:00‘
  11. --union all select ‘b‘,‘2011-03-01 15:15:00‘
  12. --union all select ‘b‘,‘2011-03-01 16:00:00‘
  13. --union all select ‘b‘,‘2011-03-01 17:15:00‘
  14.  
  15.  
  16.   
  17. ;with a  --按照name分区，同时按照createdate排序编号
  18. as
  19. (
  20. select name,
  21.        createdate,
  22.        ROW_NUMBER() over(partition by name
  23.                              order by createdate) as k1
  24. from @table1 
  25. ),
  26.  
  27. c
  28. as
  29. (
  30. select a1.name,
  31.        a1.createdate,
  32.        a1.k1,
  33.        MIN(a2.createdate) as nextCreatedate,
  34.        MIN(a2.k1) as nextK1
  35. from a a1
  36. inner join a a2
  37.         on a1.name = a2.name
  38.            and a2.createdate < DATEADD(hour,2,a1.createdate)
  39.            and a2.createdate >= a1.createdate
  40.            and a1.k1 <= a2.k1
  41. group by a1.name,
  42.          a1.createdate,
  43.          a1.k1
  44. ),
  45.  
  46. w
  47. as
  48. (
  49. select name,
  50.        createdate,
  51.        k1
  52.        
  53.        --null,
  54.        --null,
  55.        --null
  56. from a
  57. where k1 = 1
  58.  
  59. union all
  60.  
  61. select c.name,
  62.        c.nextCreatedate,
  63.        c.nextK1
  64.        
  65. from W
  66. inner join a
  67.         on a.name = w.name
  68.            and dateadd(hour,2,w.createdate) <= a.createdate
  69.            and w.k1 <= a.k1
  70.            and w.createdate <> ‘2011-03-01 12:00:00‘ 
  71. inner join c
  72.         on w.name = c.name 
  73.            and w.createdate = c.createdate
  74.            and w.k1 = c.k1
  75. where w.k1 <=3
  76. )
  77.  
  78. SELECT * 
  79. FROM w


下面的解法是正确的，不过用的是T-SQL，不是纯sql了：

  1. declare @table1 table(name nvarchar(100),createdate smalldatetime)
  2.  
  3. declare @table2 table(name nvarchar(100),createdate smalldatetime,rnum bigint)
  4.  
  5. declare @temp table(name nvarchar(100),createdate smalldatetime,rnum bigint)
  6.  
  7. declare @i int = 1;
  8.  
  9. insert into @table1
  10. select ‘a‘,‘2011-03-01 10:00:00‘
  11. union all select ‘a‘,‘2011-03-01 11:00:00‘
  12. union all select ‘a‘,‘2011-03-01 12:00:00‘
  13.  
  14. union all select ‘b‘,‘2011-03-01 13:10:00‘
  15. union all select ‘b‘,‘2011-03-01 13:20:00‘
  16. union all select ‘b‘,‘2011-03-01 14:30:00‘
  17. union all select ‘b‘,‘2011-03-01 15:15:00‘
  18. union all select ‘b‘,‘2011-03-01 16:00:00‘
  19. union all select ‘b‘,‘2011-03-01 17:16:00‘
  20. union all select ‘b‘,‘2011-03-01 17:15:00‘
  21.  
  22.   
  23. ;with a  --按照name分区，同时按照createdate排序编号
  24. as
  25. (
  26. select name,
  27.        createdate,
  28.        ROW_NUMBER() over(partition by name
  29.                              order by createdate) as k1
  30. from @table1 
  31. )
  32.  
  33. insert into @table2
  34. select * from a
  35.  
  36. insert into @temp
  37. select name,
  38.        createdate,
  39.        rnum
  40. from @table2
  41. where rnum = 1
  42.  
  43. --select * from @temp
  44.  
  45.  
  46. while @i <= (select MAX(rnum) from @table2)
  47. begin
  48.     insert into @temp
  49.     
  50. 	select t2.name,
  51. 	       min(t2.createdate),
  52. 	       @i +1	
  53. 	from @temp t1
  54. 	inner join @table2 t2        
  55. 	        on t1.name = t2.name
  56. 	           and t2.createdate >= dateadd(hour,2,t1.createdate)	           
  57. 	where t1.rnum = @i     
  58. 	group by t2.name
  59. 	            
  60. 	set @i = @i + 1	
  61. end
  62.  
  63. ;with r
  64. as
  65. (
  66. select name,
  67.        createdate
  68. from @temp
  69. group by name,
  70.          createdate
  71. )
  72.  
  73. select r.name,
  74.        r.createdate,
  75.        COUNT(1)
  76. from r
  77. inner join @table1 t
  78.         on t.name = r.name
  79.            and t.createdate >= r.createdate
  80.            and t.createdate <DATEADD(HOUR,2,r.createdate)
  81. group by r.name,
  82.          r.createdate

 

其实这个问题是个递归问题，由上一个找到下一个，但是得构造一下：

  1. declare @table1 table(name nvarchar,createdate smalldatetime)
  2.  
  3. insert into @table1
  4. select ‘a‘,‘2011-03-01 10:00:00‘
  5. union all select ‘a‘,‘2011-03-01 11:00:00‘
  6. union all select ‘a‘,‘2011-03-01 12:00:00‘
  7. union all select ‘a‘,‘2011-03-01 12:20:00‘
  8.  
  9. union all select ‘b‘,‘2011-03-01 13:10:00‘
  10. union all select ‘b‘,‘2011-03-01 13:20:00‘
  11. union all select ‘b‘,‘2011-03-01 14:30:00‘
  12. union all select ‘b‘,‘2011-03-01 15:15:00‘
  13. union all select ‘b‘,‘2011-03-01 16:00:00‘
  14. union all select ‘b‘,‘2011-03-01 17:20:00‘
  15. union all select ‘b‘,‘2011-03-01 17:15:00‘
  16. union all select ‘b‘,‘2011-03-01 19:16:00‘
  17. union all select ‘b‘,‘2011-03-01 17:15:00‘
  18.  
  19.   
  20. ;with a  --按照name分区，同时按照createdate排序编号
  21. as
  22. (
  23. select name,
  24.        createdate,
  25.        ROW_NUMBER() over(partition by name
  26.                              order by createdate) as k1
  27. from @table1 
  28. ),
  29.  
  30. c   --对于每个时间,找到大于这个时间2小时的时间中最小那个时间
  31. as
  32. (
  33. select a1.name,
  34.        a1.createdate,
  35.        a1.k1,
  36.        MIN(a2.createdate) as nextCreatedate,
  37.        MIN(a2.k1) as nextK1
  38. from a a1
  39. inner join a a2
  40.         on a1.name = a2.name
  41.            and a2.createdate >= DATEADD(hour,2,a1.createdate)
  42.            
  43. group by a1.name,
  44.          a1.createdate,
  45.          a1.k1
  46.  
  47. union all
  48.  
  49. select a.name,null,null,a.createdate,1  --构造递归运行时需要的层级
  50. from a
  51. where k1 = 1
  52. ),
  53.  
  54. w   --递归查询
  55. as
  56. (
  57. select c.name,
  58.        c.createdate,
  59.        c.k1,
  60.        c.nextCreatedate,
  61.        c.nextK1,
  62.        1 as lev
  63. from c
  64. where createdate is null
  65.       and k1 is null
  66.  
  67. union all
  68.  
  69. select c.name,
  70.        c.createdate,
  71.        c.k1,
  72.        c.nextCreatedate,
  73.        c.nextK1,
  74.        lev + 1
  75. from W
  76. inner join c
  77.         on w.name = c.name
  78.            and w.nextCreatedate = c.createdate
  79.  
  80. )
  81.  
  82. SELECT distinct name,
  83.        nextCreatedate,
  84.        nextK1 
  85. FROM w



 

 

 

技术图片 技术图片
不想长大啊
发布了416 篇原创文章 · 获赞 135 · 访问量 94万+
他的留言板 关注







以上是关于Transact-SQL统计某字段中的值第一次出现后的2小时内出现的次数的主要内容，如果未能解决你的问题，请参考以下文章

mysql特定字符出现次数统计

oracle 批量替换某字段中第一位的值

Oracle -- 多个count统计一张表中多个字段等于某值

MYSQL下统计一个字段中出现字符串的种类和个数

mysql当查询某字段结果为空并赋值

Oracle通过一个字段的值将一条记录拆分为多条记录