当同一类型的承诺之间存在差距时，事实应包含多个记录，显示每个连续承诺的准确开始和结束日期。 一个例子是 patid 1001，当相同状态连续存在而没有间隙时，它应该是单个记录。
CREATE TABLE #legal_data (
ClaimID VARCHAR(20)
,dim_legalstat_key int -- dimensionkey
,[order_start_date] DATE
,[order_end_date] DATE
,[days_committed] int -- days between order_start_date & order_end_date
)
INSERT INTO #legal_data
VALUES
('1001','11','2022-05-11','2022-10-29','171')
,('1001','131','2022-07-15','2023-03-19','247')
,('1001','116','2023-03-14','2023-03-20','6')
,('1001','11','2023-03-20','2023-03-23','3')
,('1207','11','2022-09-13','2023-03-12','180')
,('1207','11','2023-03-10','2023-03-23','13')
,('1924','2','2021-12-18','2022-06-19','183')
,('1924','2','2022-06-19','2023-12-20','184')
,('1842','77','2021-02-20','2022-06-17','482')
,('1842','77','2022-06-18','2023-12-20','550')
,('1661','22','2022-02-14','2023-03-20','399')
,('1661','22','2022-02-14','2023-03-23','402')
,('1553','4','2022-01-14','2022-02-12','29')
,('1553','4','2022-02-14','2023-03-23','402')
----- desired result
CREATE TABLE #legal_Result (
ClaimID VARCHAR(20)
,dim_legalstat_key int-- dimensionkey
,[order_start_date] DATE
,[order_end_date] DATE
,[days_committed] int --days between order_start_date & order_end_date
)
INSERT INTO #legal_Result
VALUES
('1001','11','2022-05-11','2022-10-29','171')
,('1001','131','2022-07-15','2023-03-19','247')
,('1001','116','2023-03-14','2023-03-20','6')
,('1001','11','2023-03-20','2023-03-23','3')
,('1207','11','2022-09-13','2023-03-23','191')
,('1924','2','2021-12-18','2023-12-20','732')
,('1842','77','2021-02-20','2023-12-20','1033') --not working
,('1661','22','2022-02-14','2023-03-23','402') ---
,('1553','4','2022-01-14','2022-02-12','29') --anything the
,('1553','4','2022-02-14','2023-03-23','402')
select * from #legal_data
select * from #legal_Result
这是一个差距和岛屿问题。 您已经定义了范围，现在需要对它们进行分组，识别岛屿并执行聚合。
您的预期输出似乎不正确。 ClaimID 1207 的最短日期为 2022-03-10，而不是 2023-09-13。 此外，2022年7月15日至2023年3月29日之间有257天。 假设这些是您的问题中的缺陷，那么这段代码应该适合您。
回复评论
如果您的范围尚未完全整合（ValNik 指出的一种可能性），您可以在组 CTE 之前使用此 CTE 来完成此操作：
ranges as (
select ClaimID
, dim_legalstat_key
, MIN(order_start_date) OVER (PARTITION BY ClaimID, dim_legalstat_key ORDER BY order_start_date, order_end_date ROWS BETWEEN 0 FOLLOWING AND UNBOUNDED FOLLOWING) as order_start_date
, MAX(order_end_date) OVER (PARTITION BY ClaimID, dim_legalstat_key ORDER BY order_start_date, order_end_date ROWS BETWEEN UNBOUNDED PRECEDING AND 0 PRECEDING) as order_end_date
from #legal_data f
),
那么，当然，组中的表引用必须更改。
编辑结束
;
WITH
Groups as (
SELECT ClaimID
, dim_legalstat_key
, order_start_date
, order_end_date
, LAG(order_end_date,1) OVER (partition by ClaimID, dim_legalstat_key ORDER BY order_start_date, order_end_date) AS PreviousEndDate
FROM #legal_data
),
IslandID as (
SELECT ClaimID
, dim_legalstat_key
, order_start_date
, order_end_date
, CASE WHEN PreviousEndDate >= order_start_date THEN 0 ELSE 1 END AS IslandStartInd
, SUM(CASE WHEN PreviousEndDate >= order_start_date THEN 0 ELSE 1 END) OVER ( ORDER BY claimid, dim_legalstat_key, order_start_date, order_end_date) AS IslandId
FROM Groups
),
Islands as (
SELECT ClaimID
, dim_legalstat_key
, MIN(order_start_date) AS order_start_date
, MAX(order_end_date) AS order_end_date
FROM IslandId
GROUP BY IslandId
, ClaimID
, dim_legalstat_key
)
select *
, datediff(day,order_start_date, order_end_date) as days_committed
from Islands
order by ClaimID
, order_start_date
, order_end_date
我查看了@dougp提出的解决方案。它适用于“正常”排序的数据。我认为，问题在于“混乱”的有序数据，如下所示：
period1 ---------------------------
period2 -----
period3 ---------------------------
以这个问题为例
,('2925','5','2022-12-10','2022-12-20','x')
,('2925','5','2022-12-15','2022-12-18','x')
,('2925','5','2022-12-19','2022-12-29','x')
该行的查询结果
| 索赔ID | dim_legalstat_key | 订单开始日期 | 订单结束日期 | 承诺天数 |
|---|---|---|---|---|
| 2925 | 5 | 2022-12-10 | 2022-12-20 | 10 |
| 2925 | 5 | 2022-12-19 | 2022-12-29 | 10 |
预期结果
| 索赔ID | dim_legalstat_key | 订单开始日期 | 订单结束日期 | 承诺天数 |
|---|---|---|---|---|
| 2925 | 5 | 2022-12-10 | 2022-12-29 | 19 |
也许数据的排序是“正确”的，不会出现这种情况。
这个任务很有趣。我明白了，我对解决方案的看法与@dougp 的解决方案类似。看到测试数据可能有错误， 我将提出一个递归解决方案供考虑。
with ndata as(
select ClaimID,dim_legalstat_key,order_start_date
,max(order_end_date)order_end_date
,datediff(d,order_start_date,max(order_end_date))days_committed
,row_number()over(partition by ClaimId order by order_start_date) rn
from #legal_data
group by ClaimID,dim_legalstat_key,order_start_date
)
,r as( --Islands head rows
select rn headrow,rn,1 lvl,ClaimID,dim_legalstat_key
,order_start_date,order_end_date,days_committed
,cast(rn as varchar(1000)) rowlist
from ndata t1
where not exists
(
select * from ndata t2
where t2.ClaimId=t1.ClaimId and t2.rn<>t1.rn
and t2.dim_legalstat_key=t1.dim_legalstat_key
and t2.order_start_date<=t1.order_start_date
and t2.order_end_date>=t1.order_start_date
)
union all --iterate through all possible rows
select r.headrow,t2.rn,r.lvl+1 lvl,r.ClaimID,t2.dim_legalstat_key
,case when r.order_start_date<t2.order_start_date then r.order_start_date
else t2.order_start_date end order_start_date
,case when r.order_end_date>t2.order_end_date then r.order_end_date
else t2.order_end_date end order_end_date
,r.days_committed
,cast(concat(r.rowlist,',',cast(t2.rn as varchar)) as varchar(1000))rowlist
from r inner join ndata t2
on t2.ClaimId=r.ClaimId and t2.dim_legalstat_key=r.dim_legalstat_key
and r.rn<>t2.rn
and charindex(','+cast(t2.rn as varchar)+',',','+r.rowlist+',')=0
and t2.order_start_date>=r.order_start_date
and t2.order_start_date<=r.order_end_date
and lvl<100
)
select ClaimID,dim_legalstat_key
,min(order_start_date) order_start_date
,max(order_end_date)order_end_date
,datediff(d,min(order_start_date),max(order_end_date)) days_committed
from r
group by ClaimID,dim_legalstat_key,headrow
order by ClaimID,dim_legalstat_key
更新1。
对于串联范围（start_date = end_date = end_date + 1day 的下一天） 我会添加一些检查。 （例如 ClaimId 为 1842 的案例）
第一个 CTE
ndata递归查询的基础部分
rconsecutively combines更正查询
with ndata as(
select ClaimID,dim_legalstat_key,order_start_date
,max(order_end_date)order_end_date
,datediff(d,order_start_date,max(order_end_date))days_committed
,row_number()over(partition by ClaimId order by order_start_date) rn
from #legal_data
group by ClaimID,dim_legalstat_key,order_start_date
)
,r as(
select rn headrow,rn,1 lvl,ClaimID,dim_legalstat_key
,order_start_date,order_end_date,days_committed
,cast(rn as varchar(1000)) rowlist
from ndata t1
where not exists(
select * from ndata t2
where t2.ClaimId=t1.ClaimId and t2.rn<>t1.rn
and t2.dim_legalstat_key=t1.dim_legalstat_key
and
(
( t2.order_start_date<=t1.order_start_date
and t2.order_end_date>=t1.order_start_date
)
or
( -- concatenated date ranges
dateadd(day,1,t2.order_end_date)=t1.order_start_date
)
)
)
union all
select r.headrow,t2.rn,r.lvl+1 lvl,r.ClaimID,t2.dim_legalstat_key
,case when r.order_start_date<t2.order_start_date then r.order_start_date
else t2.order_start_date end order_start_date
,case when r.order_end_date>t2.order_end_date then r.order_end_date
else t2.order_end_date end order_end_date
,r.days_committed
,cast(concat(r.rowlist,',',cast(t2.rn as varchar)) as varchar(1000))rowlist
from r inner join ndata t2
on t2.ClaimId=r.ClaimId and t2.dim_legalstat_key=r.dim_legalstat_key
and r.rn<>t2.rn
and charindex(','+cast(t2.rn as varchar)+',',','+r.rowlist+',')=0
and t2.order_start_date>=r.order_start_date
and t2.order_start_date<=dateadd(day,1,r.order_end_date) -- concatenated date ranges
and lvl<100
)
-- select * from r;
select ClaimID,dim_legalstat_key
,min(order_start_date) order_start_date
,max(order_end_date)order_end_date
,datediff(d,min(order_start_date),max(order_end_date)) days_committed
from r
group by ClaimID,dim_legalstat_key,headrow
order by ClaimID,dim_legalstat_key