我有一张表（RECORDS），其中包含从何时到项目链接到父项的记录：
CHILD STARTDATE ENDDATE PARENT
207 2013-12-18 12:45:59.017 2014-01-09 18:16:01.227 NULL
207 2014-01-09 18:16:01.227 2016-03-03 09:54:28.757 NULL
207 2016-03-03 09:54:28.757 2100-01-01 00:00:00.000 NULL
10558 2016-03-03 18:10:34.877 2016-04-05 10:25:22.860 10557
10558 2016-04-05 10:25:22.860 2016-04-05 11:33:10.493 10557
10558 2016-04-05 11:33:10.493 2100-01-01 00:00:00.000 10557
10596 2016-05-15 16:55:15.970 2016-05-16 10:00:00.000 10592
10596 2016-05-16 10:00:00.000 2016-05-17 10:00:00.000 10593
10596 2016-05-17 10:00:00.000 2100-01-01 00:00:00.000 10592
10600 2012-12-18 12:45:59.017 2100-01-01 00:00:00.000 10599
我在4个主要案例上面显示：
对于同一个孩子，具有空值的新记录（207）
对于同一个孩子，具有相同父值的新记录（10558）
对于同一个孩子，父母的新记录会更改（10596）
对于同一个孩子，只有一个记录（10600）。
由于子项目可以更改回较旧的父项，使用：
SELECT
CHILD
, MIN(STARTDATE) as STARTDATE
, MAX(ENDDATE) as ENDDATE
, PARENT
FROM RECORDS
GROUP BY CHILD, PARENT
会导致项目10596出现以下不良后果：
207 2013-12-18 12:45:59.017 2100-01-01 00:00:00.000 NULL
10558 2016-03-03 18:10:34.877 2100-01-01 00:00:00.000 10557
10596 2016-05-15 16:55:15.970 2100-01-01 00:00:00.000 10592
10596 2016-05-16 10:00:00.000 2016-05-17 10:00:00.000 10593
10600 2012-12-18 12:45:59.017 2100-01-01 00:00:00.000 10599
期望的结果是：
207 2013-12-18 12:45:59.017 2100-01-01 00:00:00.000 NULL
10558 2016-03-03 18:10:34.877 2100-01-01 00:00:00.000 10557
10596 2016-05-15 16:55:15.970 2016-05-16 10:00:00.000 10592
10596 2016-05-16 10:00:00.000 2016-05-17 10:00:00.000 10593
10596 2016-05-17 10:00:00.000 2100-01-01 00:00:00.000 10592
10600 2012-12-18 12:45:59.017 2100-01-01 00:00:00.000 10599
知道如何实现这个目标吗？
答案 0 :(得分：0)
以下假设您的记录中存在连续性（无重叠）：
WITH cte
AS
(
SELECT
*
, ROW_NUMBER() OVER (PARTITION BY CHILD ORDER BY STARTDATE) r
FROM RECORDS
)
,
cte2
AS
(
SELECT
t1.CHILD CHILD1
, t1.STARTDATE STARTDATE1
, t1.ENDDATE ENDDATE1
, t1.PARENT PARENT1
, t1.r r1
, t2.CHILD CHILD2
, t2.STARTDATE STARTDATE2
, t2.ENDDATE ENDDATE2
, t2.PARENT PARENT2
, t2.r r2
FROM
cte t1
LEFT JOIN cte t2 ON
t1.CHILD = t2.CHILD
AND t1.r = t2.r - 1
WHERE
t1.PARENT != t2.PARENT
OR (t1.PARENT IS NULL AND t2.PARENT IS NOT NULL)
OR (t2.PARENT IS NULL AND t1.PARENT IS NOT NULL)
OR t2.r IS NULL
)
SELECT
CHILD1 CHILD
,
ISNULL
(
(
SELECT MIN(STARTDATE)
FROM RECORDS
WHERE
CHILD = CHILD1
AND (PARENT = PARENT1 OR (PARENT IS NULL AND PARENT1 IS NULL))
AND STARTDATE >=
(
SELECT ENDDATE1
FROM cte2 cte2INNER
WHERE
cte2INNER.CHILD1 = cte2OUTER.CHILD1
AND cte2INNER.r2 =
(
SELECT MAX(cteINNER2.r2)
FROM cte2 cteINNER2
WHERE
cteINNER2.CHILD1 = cte2OUTER.CHILD1
AND cteINNER2.r2 <= cte2OUTER.r1
)
)
)
,
(
SELECT MIN(STARTDATE)
FROM RECORDS
WHERE CHILD = CHILD1
)
)
STARTDATE
, ENDDATE1 ENDDATE
, PARENT1 PARENT
FROM cte2 cte2OUTER
它使用临时表快速运行：
WITH cte
AS
(
SELECT
*
, ROW_NUMBER() OVER (PARTITION BY CHILD ORDER BY STARTDATE) r
FROM RECORDS
)
SELECT
t1.CHILD CHILD1
, t1.STARTDATE STARTDATE1
, t1.ENDDATE ENDDATE1
, t1.PARENT PARENT1
, t1.r r1
, t2.CHILD CHILD2
, t2.STARTDATE STARTDATE2
, t2.ENDDATE ENDDATE2
, t2.PARENT PARENT2
, t2.r r2
INTO #RECORDSTEMP
FROM
cte t1
LEFT JOIN cte t2 ON
t1.CHILD = t2.CHILD
AND t1.r = t2.r - 1
WHERE
t1.PARENT != t2.PARENT
OR (t1.PARENT IS NULL AND t2.PARENT IS NOT NULL)
OR (t2.PARENT IS NULL AND t1.PARENT IS NOT NULL)
OR t2.r IS NULL
SELECT
CHILD1 CHILD
,
ISNULL
(
(
SELECT MIN(STARTDATE)
FROM RECORDS
WHERE
CHILD = CHILD1
AND (PARENT = PARENT1 OR (PARENT IS NULL AND PARENT1 IS NULL))
AND STARTDATE >=
(
SELECT ENDDATE1
FROM #RECORDSTEMP rtmpINNER
WHERE
rtmpINNER.CHILD1 = rtmpOUTER.CHILD1
AND rtmpINNER.r2 =
(
SELECT MAX(rtmpINNER2.r2)
FROM #RECORDSTEMP rtmpINNER2
WHERE
rtmpINNER2.CHILD1 = rtmpOUTER.CHILD1
AND rtmpINNER2.r2 <= rtmpOUTER.r1
)
)
)
,
(
SELECT MIN(STARTDATE)
FROM RECORDS
WHERE CHILD = CHILD1
)
)
STARTDATE
, ENDDATE1 ENDDATE
, PARENT1 PARENT
FROM #RECORDSTEMP rtmpOUTER
DROP TABLE #RECORDSTEMP
答案 1 :(得分：0)
感谢您的回答，它确实有效。
我也收到了这个解决方案：
WITH DISTINCTRECORDS as (
SELECT
CHILD
,
CASE
WHEN
(
CHILD = lag(CHILD) over(order by CHILD,STARTDATE, ENDDATE, PARENT)
AND
STARTDATE = lag(ENDDATE) over(order by CHILD,STARTDATE, ENDDATE, PARENT)
AND
(PARENT = lag(PARENT) over(order by CHILD,STARTDATE, ENDDATE, PARENT) OR ( PARENT is null AND lag(PARENT) over(order by CHILD,STARTDATE, ENDDATE, PARENT) is null ))
)
OR
(
CHILD = lead(CHILD) over(order by CHILD,STARTDATE, ENDDATE, PARENT)
AND
ENDDATE = lead(STARTDATE) over(order by CHILD,STARTDATE, ENDDATE, PARENT)
AND
(PARENT = lead(PARENT) over(order by CHILD,STARTDATE, ENDDATE, PARENT) OR ( PARENT is null AND lead(PARENT) over(order by CHILD,STARTDATE, ENDDATE, PARENT) is null ))
)
THEN 1
ELSE 0
END ToBeMerged
, STARTDATE
, ENDDATE
, PARENT
FROM RECORDS
)
SELECT CHILD, min(STARTDATE) STARTDATE, max(ENDDATE) ENDDATE, PARENT FROM DISTINCTRECORDS WHERE ToBeMerged = 1 GROUP BY CHILD, PARENT
UNION ALL
SELECT CHILD, STARTDATE, ENDDATE, PARENT FROM DISTINCTRECORDS WHERE ToBeMerged = 0
哪一个会有最好的表现？