SQL Server Sum between two values

rjjhvcjd  于 2023-10-15  发布在  其他
关注(0)|答案(1)|浏览(107)

For my source table I have all the “Contact history for customers” in the correct order (ascending starting from earliest date at the top). What I need to do is group rows where the contacts are related to the same issue. The way to determine whether a issue is related to each other is based on the time. If the time difference <2 weeks then the two cases are related.

A 1 in the 'Pass or Fail' column highlights whether the case was fully resolved. A 0 indicates that the call from the below row is related to the same issue.

I need to produce a summary of this table to show where a customer has contacted multiple times against the same query

I am using SQL Server 2012

Here is an example of my source data. In my full table I will have multiple customers. You can see I have colour coded to show that there are a total of 9 contacts but just 4 different issues. There will be no fixed number of contacts or issues for each customer

Customer_IDstart_dateNext_DateTime_Difference_(Mins)total_contactsPass_or_failNEXT_Contact_same_day
1234503/03/202205/05/202290794110
1234505/05/202205/05/20223101
1234505/05/202206/05/20221409100
1234506/05/202206/05/20222101
1234506/05/202206/05/202211101
1234506/05/202220/11/2022284973110
1234520/11/202205/01/202366242110
1234505/01/202305/01/2023178101
1234505/01/2023NULLNULL110

I have tried using DENSE_RANK, ROW_NUMBER, LAG&LEAD on the Pass or Fail column, as well as MAX & MIN dates for the start and next dates, but am not having any luck.

E.g.

select Customer ID, sum(total_contacts), min(Start date), max(Next Date)
from (select *,
             sum(case when Pass or fail = 0 then 1 else 0 end) over (order by Customer ID) as grp
      from #Mastery
     ) #Mastery
where total_contacts = '1'
group by Customer ID, grp;

-- this returns all contacts

select Customer ID, sum(total_contacts) AS total_Contacts, SUM (Pass or fail) AS Pass#, min(Start date) AS Mindate, max(Next Date) AS Maxdate, 
       ROW_NUMBER () over(partition by Customer ID order by Pass or fail) as Contactranking
from  #Mastery2
group by Customer ID,  Pass or fail

-- this just returns the sum of passes and fails

This is the output I am looking for. You can see that against the same customer ID I have 4 rows, one row represents each issue. For each Row it should show the number of calls against the issue. To quantify the number of calls related to the same issue you have to use the 'Pass or Fail' column

fjaof16o

fjaof16o1#

This is a gaps and island style problem where you have the right idea about the grouping, but there are some details that makes this a bit tricker. Here's a potential solution:

SELECT  CustomerID, MIN(startDate) startDate, NULLIF(MAX(nextDate), '99991231') nextDate, COUNT(*) cnt
FROM    (
    select  *
    ,   SUM(passprev) OVER(partition by customerid order by startDate, nextdate) AS grp
    from (
        
        select  LAG(passorfail, 1, 0) over(partition by customerid order by d.startDate,  d.nextDate) AS passprev
        ,   d.*
        ,   customerID
        from (
            VALUES  (12345, N'03/03/2022', N'05/05/2022', N'90794', 1, 1, 0)
            ,   (12345, N'05/05/2022', N'05/05/2022', N'3', 1, 0, 1)
            ,   (12345, N'05/05/2022', N'06/05/2022', N'1409', 1, 0, 0)
            ,   (12345, N'06/05/2022', N'06/05/2022', N'2', 1, 0, 1)
            ,   (12345, N'06/05/2022', N'06/05/2022', N'11', 1, 0, 1)
            ,   (12345, N'06/05/2022', N'20/11/2022', N'284973', 1, 1, 0)
            ,   (12345, N'20/11/2022', N'05/01/2023', N'66242', 1, 1, 0)
            ,   (12345, N'05/01/2023', N'05/01/2023', N'178', 1, 0, 1)
            ,   (12345, N'05/01/2023', NULL, NULL, 1, 1, 0)
        ) t (CustomerID,Startdate,NextDate,[TimeDifference(Mins)],total_contacts,Passorfail,[NEXT_Contact_same_day])
        CROSS APPLY (
                SELECT  CONVERT(DATETIME, startdate, 103) AS startDate
                ,   ISNULL(CONVERT(DATETIME, NextDate, 103), '99991231') AS nextDate
            ) d
        ) x
    ) x
GROUP BY customerID, grp

First and foremost, the problem is that you have to get the previous passorfail column, because usually when you create a gaps and islands sum, you want to keep the value at 0, until a new group arrives, while here, 1 creates a new group. Therefore I use LAG(passorfail, 1, 0) .

The sorting here is also a bit tricky, because you have multiple duplicates dates. For this, I create the tiebreaker by ISNULL(CONVERT(DATETIME, NextDate, 103), '99991231') so the open dates goes to the end. It also simplifies the MAX handling later.

The rest is nothing special, except we flip back the date to null with NULLIF(MAX(nextDate), '99991231')

Just a few notes: you should really provide the test data in table format. Also, DD/MM/YYYY is a very unhandy date style to work with, YYYYMMDD is much nicer.

相关问题