SQL：根据记录创建日期和其他条件过滤记录

Question

I am struggling to find a better solution to pick unique records from my user call data table. My table structure is as follows:

SELECT [MarketName],
       [WebsiteName] ,
       [ID] ,
       [UserID],
       [CreationDate],
       [CallDuration],
       [FromPhone] ,
       [ToPhone],
       [IsAnswered],
       [Source]
FROM [dbo].[UserCallData]

There are multiple entries in this table with different and same ID's. I wanted to check if [FromPhone] and [ToPhone] exists multiple times within last 3 months, if yes, I wanted to pick the first record with all columns based on [CreationDate], count the number of occurrences as TotalCallCount and sum the totalCallDuration as a single record. If [FromPhone] and [ToPhone] does not occur multiple times, I wanted to pick all columns as such. I have been able to put up partial query like below. But this doesn't return all columns without including in group by clause and also it doesn't satisfy my entire criteria. Any help on this would be highly appreciated.

select  [FromPhone], 
        MIN([CreationDate]),
        [ToPhone], 
        marketname, 
        count(*) as TotalCallCount , 
        sum(CallDuration) as TotalCallDuration 
from [dbo].[UserCallData]
where  [CreationDate] >= DATEADD(MONTH, -3, GETDATE())
group by  [FromPhone],[ToPhone], marketname 
having count([FromPhone]) > 1 and count([ToPhone]) >1

Answer 1

Try to use ROW_NUMBER()

;with cte as
(
    select *, ROW_NUMBER() OVER(PARTITION BY FromPhone, ToPhone ORDER BY CreationDate) as RN
    from UserCallData
    where CreationDate >= DATEADD(MONTH, -3, GETDATE())
),
cte_totals as
(
    select C1.FromPhone, C1.ToPhone, COUNT(*) as TotalCallCount, SUM(CallDuration) as TotalCallDuration
    from cte C1
    where exists(select * from cte C2 where C1.FromPhone = C2.FromPhone and C1.ToPhone = C2.ToPhone and C2.RN > 1)
    group by C1.FromPhone, C1.ToPhone
)
select C1.*, TotalCallCount, TotalCallDuration
from cte C1
    inner join cte_totals C2 on C1.FromPhone = C2.FromPhone and C1.ToPhone = C2.ToPhone
where C1.RN = 1 

I wrote query right in here so it can have some mistakes or mistypes, but the main idea might be clear.

Answer 2

I'm not entirely sure I've understood the question, but if I have the following may be what you want (or be a useful starting point):

SELECT
       ucd.FromPhone,
       min(ucd.CreationDate) as MinCreationDate,
       ucd.ToPhone,
       ucd.MarketName,
       count(*) as TotalCallCount,
       sum(ucd.CallDuration) as TotalCallDuration,
       case
           when min(ucd.WebsiteName) = max(ucd.WebsiteName) then min(ucd.WebsiteName)
           else '* Various'
       end as WebsiteName,
       case
           when min(ucd.ID) = max(ucd.ID) then min(ucd.ID)
           else '* Various'
       end as ID,
       case
           when min(ucd.UserID) = max(ucd.UserID) then min(ucd.UserID)
           else '* Various'
       end as UserID,
       case
           when min(ucd.IsAnswered) = max(ucd.IsAnswered) then min(ucd.IsAnswered)
           else '* Some'
       end as IsAnswered,
       case
           when min(ucd.Source) = max(ucd.Source) then min(ucd.Source)
           else '* Various'
       end as Source
FROM
    dbo.UserCallData ucd
WHERE
    ucd.CreationDate >= DATEADD(MONTH, -3, GETDATE())
GROUP BY
    ucd.FromPhone,
    ucd.ToPhone,
    ucd.MarketName

Where we are collapsing rows together, if all the rows agree on a given column (so min(Field) = max(Field) ), I return the min(Field) value (which is the same all the others, but avoid problems with needing additional "group by" clauses which would interfere with the other cases). Where they don't all agree, I've returned "* something" .

The code assumes that all the columns are text type columns (you haven't said), you may get conversion errors. It also assumes that none of these fields are null . You / we can adapt the code if those assumptions aren't correct. If you aren't able to do that for yourself, let me know about issues, I'll be happy to do what I can.