日期范围交叉口

Question

我有一个BreakdownLog表具有以下的列：

EquipmentID, ProblemID, BreakdownDate, IssueFixedDate 

每个设备可以有多个故障，更重要的是，有可能是重叠的日期范围相同的设备！

例如，下面给出的数据：

EquipmentID|ProblemID|BreakdownDate|IssueFixedDate 
1   |1  |01-Jun-2011 |01-Sep-2011 
1   |2  |01-Jun-2011 |01-Oct-2011 
2   |1  |01-Jun-2011 |01-Oct-2011 
2   |2  |01-Jun-2011 |01-Oct-2011 
3   |1  |15-Jun-2011 |01-Sep-2011 
3   |2  |10-Jun-2011 |25-Aug-2011 
4   |1  |01-Jun-2011 |01-Aug-2011 
4   |2  |10-Sep-2011 |22-Oct-2011 
5   |1  |01-Jun-2011 |15-Jun-2011 
5   |2  |02-Jun-2011 |NULL 

现在我想查询它可以计算每个天是设备倒闭的数量。如果IssueFixedDate为NULL，那么我们假定设备仍然不存在，因此计算直到当前日期的Defunct天数。

预期的结果集应该是：

EquipmentID|DefunctDays 
1   |122 
2   |122 
3   |83 
4   |103 
5   |143 

我使用SQL Server 2008中因此，即使热膨胀系数，光标等都是可以接受的。

感谢
Raghu

Answer 1

这使用master..spt_values作为adhoc辅助数字表。您可以使用one of the techniques here（开始编号为零）创建您自己的专用号码表。

;WITH Numbers 
    AS (SELECT number 
     FROM master..spt_values 
     WHERE type = 'P') 
SELECT EquipmentID, 
     COUNT(DISTINCT number + DATEDIFF(DAY,0, BreakdownDate)) - 1 AS DefunctDays 
FROM BreakdownLog 
     JOIN Numbers N 
     ON number <= DATEDIFF(DAY, BreakdownDate, 
         ISNULL(IssueFixedDate, GETDATE())) 
GROUP BY EquipmentID 

Answer 2

它应该是这样的：

测试表初始化

DROP TABLE BreakdownLog 

CREATE TABLE BreakdownLog 
(
    EquipmentID INT, 
    ProblemID INT, 
    BreakdownDate DATETIME, 
    IssueFixedDate DATETIME NULL 
); 

INSERT INTO BreakDownLog VALUES (1, 1, '01-Jun-2011', '01-Sep-2011') 
INSERT INTO BreakDownLog VALUES (1, 2, '01-Jun-2011', '01-Oct-2011') 
INSERT INTO BreakDownLog VALUES (2, 1, '01-Jun-2011', '01-Oct-2011') 
INSERT INTO BreakDownLog VALUES (2, 2, '01-Jun-2011', '01-Oct-2011') 
INSERT INTO BreakDownLog VALUES (3, 1, '15-Jun-2011', '01-Sep-2011') 
INSERT INTO BreakDownLog VALUES (3, 2, '10-Jun-2011', '25-Aug-2011') 
INSERT INTO BreakDownLog VALUES (4, 1, '01-Jun-2011', '01-Aug-2011') 
INSERT INTO BreakDownLog VALUES (4, 2, '10-Sep-2011', '22-Oct-2011') 
INSERT INTO BreakDownLog VALUES (5, 1, '01-Jun-2011', '15-Jun-2011') 
INSERT INTO BreakDownLog VALUES (5, 2, '02-Jun-2011', NULL) 

真正的代码

-- We exchange the NULLs in IssueFixedDate with the current date 
; WITH Base AS (
    SELECT EquipmentID, ProblemID, BreakdownDate 
     , ISNULL(IssueFixedDate 
      , CONVERT(VARCHAR(10), GETDATE(), 101)) IssueFixedDate 
    -- The previous line generates the current date without time 
    FROM BreakDownLog 
) 

-- We generate a table with all the days the equipment was broken. 
-- This is done through a recursive CTE 
, BaseDays AS (
    SELECT EquipmentID, BreakdownDate AS DefunctDay, IssueFixedDate FROM Base 
    UNION ALL 
    SELECT EquipmentID, DefunctDay + 1 AS DefunctDay, IssueFixedDate 
     FROM BaseDays 
     WHERE DefunctDay + 1 <= IssueFixedDate 
    -- In T-SQL if you add 1 to a DateTime it's equivalent to adding a day 
) 

-- We make a distinct on the days where the equipment was broken, 
-- to delete days where the equipment was broken for two reasons 
, BaseDaysDistinct AS (
    SELECT DISTINCT EquipmentID, DefunctDay 
     FROM BaseDays 
) 

-- We group the equipment's DefunctDays by EquipmentID 
SELECT EquipmentID, COUNT(*) DefunctDays 
    FROM BaseDaysDistinct 
    GROUP BY EquipmentID 

我们可以在已经改变了过去两个选择：

SELECT EquipmentID, COUNT(DISTINCT DefunctDay) DefunctDays 
    FROM BaseDays 
    GROUP BY EquipmentID 

简化我生成使用递归CTE BreakdownDate和IssueFixedDate之间的天列表，删除多次出现并数着日子更天。