缺天按天分组

Question

我有一个 sql server 表，结构和数据如下：

Created              ,  keyword
'2017-10-03 19:18:00',  'test7'
'2017-10-07 01:06:00',  'test3'
'2017-10-07 15:19:00',  'test2'
'2017-10-07 21:39:00',  'test10'
'2017-10-08 00:36:00',  'test3'
'2017-10-08 01:26:00',  'test13'
'2017-10-08 01:33:00',  'test9'
'2017-10-08 08:23:00',  'test13'
'2017-10-08 09:35:00',  'test9'
'2017-10-08 12:38:00',  'test9'
'2017-10-08 15:07:00',  'test2'
'2017-10-10 05:09:00',  'test4'

我想运行一个计算活动的查询并按天分组，还考虑没有记录任何内容的天数，并将这些天的活动显示为零。因此，我想要一个返回以下结果集的查询：

'2017-10-03', 1
'2017-10-04', 0
'2017-10-05', 0
'2017-10-06', 0
'2017-10-07', 3
'2017-10-08', 7
'2017-10-09', 0
'2017-10-03', 1

我知道如何运行查询并按天数对其进行分组，但不知道如何计算没有记录的天数。由于我是 Sql 新手，如果有人能提供一个工作示例，我将不胜感激。提前致谢

Answer 1

 declare @date int
 WITH CTE_DatesTable
AS
(
  SELECT CAST('20171003' as date) AS [date]
  UNION ALL
  SELECT   DATEADD(dd, 1, [date])
  FROM CTE_DatesTable
  WHERE DATEADD(dd, 1, [date]) <= '20171010'
)
SELECT [CalendarDate]=[date] into #DimDate   FROM CTE_DatesTable

OPTION (MAXRECURSION 0);

select * from #DimDate

这将创建一个日历表以与您当前的表连接以填补空白

Answer 2

当您没有日历表时，这是一个解决方案：

select row_number() over(order by getdate()) - 1 as n
into #nums
from sys.columns cross join sys.columns c2;

declare @t table(Created datetime, keyword varchar(100));
insert into @t values
('2017-10-03 19:18:00',  'test7'),
('2017-10-07 01:06:00',  'test3'),
('2017-10-07 15:19:00',  'test2'),
('2017-10-07 21:39:00',  'test10'),
('2017-10-08 00:36:00',  'test3'),
('2017-10-08 01:26:00',  'test13'),
('2017-10-08 01:33:00',  'test9'),
('2017-10-08 08:23:00',  'test13'),
('2017-10-08 09:35:00',  'test9'),
('2017-10-08 12:38:00',  'test9'),
('2017-10-08 15:07:00',  'test2'),
('2017-10-10 05:09:00',  'test4')

declare @min_dt date, @max_dt date;
select @min_dt = min(Created), @max_dt = max(Created)
from @t;

with calendar as
(
select dateadd(day, n, @min_dt) as dt
from #nums
where dateadd(day, n, @min_dt) <= @max_dt
)

select c.dt, isnull(count(t.keyword), 0) as cnt
from calendar c left join @t t
        on c.dt = cast(t.Created as date)
group by c.dt; 

在我的情况下，我没有表格日历，但我有带有数字 (Nums) 的固定表格，但如果您没有数字表格，您可以像我在 #nums 中所做的那样生成它（您应该限制生成的数字到一个合理的数字）

Answer 3

试试这个：

declare @startDate date = '2017-10-01'
declare @endDate date = '2017-10-31'

;with cte as (
select cast(@startDate as date) [dayOfYear]
union all
select DATEADD(day, 1, [dayOfYear]) from cte
where [dayOfYear] < @endDate
)

select dayOfYear, SUM(case when Created is null then 0 else 1 end) from cte
left join MY_TABLE [T] on cte.dayOfYear = CAST(T.Created as date)
group by dayOfYear

逻辑如下：

获取@startDate 和@endDate 之间所有日期的表格（CTE - 我指定了十月的第一天和最后一天）。然后我们离开加入您的表格，当天数不匹配时，我们将相应的值定义为 0，否则为 1。然后每天对这些值求和就足够了。