SQL计算日历年内的日期段

Question

我需要的是在 SQL 中给定这样的表来计算日历年内缺失的时间段：

DatesTable
|ID|DateStart   |DateEnd   |
 1  NULL         NULL
 2  2015-1-1     2015-12-31
 3  2015-3-1     2015-12-31
 4  2015-1-1     2015-9-30
 5  2015-1-1     2015-3-31
 5  2015-6-1     2015-12-31
 6  2015-3-1     2015-6-30
 6  2015-7-1     2015-10-31

预期回报为：

 1  2015-1-1     2015-12-31
 3  2015-1-1     2015-2-28
 4  2015-10-1    2015-12-31
 5  2015-4-1     2015-5-31
 6  2015-1-1     2015-2-28
 6  2015-11-1    2015-12-31

本质上是工作块。我需要展示的是日历年中没有工作的部分。所以对于 ID = 3，他从 3 月 1 日一直工作到今年剩下的时间。但他从 1 月 1 日到 2 月 28 日都没有工作。这就是我要找的。​​p>

Answer 1

您可以使用 SQL Server 2012+ 中提供的 LEAD、LAG 窗口函数来实现：

;WITH CTE AS (
   SELECT ID, 
          LAG(DateEnd) OVER (PARTITION BY ID ORDER BY DateEnd) AS PrevEnd,
          DateStart,
          DateEnd,
          LEAD(DateStart) OVER (PARTITION BY ID ORDER BY DateEnd) AS NextStart
   FROM DatesTable
)
SELECT ID, DateStart, DateEnd
FROM (
-- Get interval right before current [DateStart, DateEnd] interval
SELECT ID, 
       CASE 
          WHEN DateStart IS NULL THEN '20150101'
          WHEN DateStart > start THEN start
          ELSE NULL
       END AS DateStart,
       CASE 
          WHEN DateStart IS NULL THEN '20151231'
          WHEN DateStart > start THEN DATEADD(d, -1, DateStart)
          ELSE NULL
       END AS DateEnd
FROM CTE
CROSS APPLY (SELECT COALESCE(DATEADD(d, 1, PrevEnd), '20150101')) x(start)

-- If there is no next interval then get interval right after current 
-- [DateStart, DateEnd] interval (up-to end of year)
UNION ALL

SELECT ID, DATEADD(d, 1, DateEnd) AS DateStart, '20151231' AS DateEnd       
FROM CTE
WHERE DateStart IS NOT NULl    -- Do not re-examine [Null, Null] interval
      AND NextStart IS NULL    -- There is no next [DateStart, DateEnd] interval 
      AND DateEnd < '20151231' -- Current [DateStart, DateEnd] interval  
                               -- does not terminate on 31/12/2015
) AS t
WHERE t.DateStart IS NOT NULL
ORDER BY ID, DateStart

上述查询背后的想法很简单：对于每个[DateStart, DateEnd] 间隔，在它之前获取'not working' 间隔。如果当前间隔之后没有间隔，则还获得连续的'not working'间隔（如果有）。

另外请注意，如果DateStart 是NULL，那么DateStart 也是NULL 对于相同的ID。

Demo here

Answer 2

如果您的数据不是太大，这种方法会奏效。它扩展了所有的天数和 ID，然后将它们重新分组：

with d as (
      select cast('2015-01-01' as date)
      union all
      select dateadd(day, 1, d)
      from d
      where d < cast('2015-12-31' as date)
     ),
     td as (
      select *
      from d cross join
           (select distinct id from t) t
      where not exists (select 1
                        from t t2
                        where d.d between t2.startdate and t2.enddate
                       )
     )
select id, min(d) as startdate, max(d) as enddate
from (select td.*,
             dateadd(day, - row_number() over (partition by id order by d), d) as grp
      from td
     ) td
group by id, grp
order by id, grp;

另一种方法依赖于累积和以及在 SQL Server 2012+ 中更容易表达的类似功能。

Answer 3

我认为有点简单的方法。

基本上为所有工作块范围 (A) 创建一个日期列表。然后为每个 ID (B) 创建一个全年日期列表。然后从 B 中删除 A。将剩余的日期列表编译为每个 ID 的日期范围。

DECLARE @startdate DATETIME, @enddate DATETIME
SET @startdate = '2015-01-01'
SET @enddate = '2015-12-31'

--Build date ranges from remaining date list
;WITH dateRange(ID, dates, Grouping)
AS
(
    SELECT dt1.id, dt1.Dates, dt1.Dates + row_number() over (order by dt1.id asc, dt1.Dates desc) AS Grouping
    FROM 
    (
        --Remove (A) from (B)
        SELECT distinct dt.ID, tmp.Dates FROM DatesTable dt
        CROSS APPLY
        (
            --GET (B) here
            SELECT DATEADD(DAY, number, @startdate) [Dates]
            FROM master..spt_values
            WHERE type = 'P' AND DATEADD(DAY, number, @startdate) <= @enddate
        ) tmp
        left join
        (
            --GET (A) here
            SELECT DISTINCT T.Id,
               D.Dates
            FROM DatesTable AS T
            INNER JOIN master..spt_values as N on N.number between 0 and datediff(day, T.DateStart, T.DateEnd)
            CROSS APPLY (select dateadd(day, N.number, T.DateStart)) as D(Dates)
            WHERE N.type ='P'
        ) dr
        ON dr.Id = dt.Id and dr.Dates = tmp.Dates
        WHERE dr.id is null
    ) dt1
)
SELECT ID, CAST(MIN(Dates) AS DATE) DateStart, CAST(MAX(Dates) AS DATE) DateEnd
FROM dateRange
GROUP BY ID, Grouping
ORDER BY ID

代码如下： http://sqlfiddle.com/#!3/f3615/1

我希望这会有所帮助！