HAVING MAX(Value) is not working in SQL Server query

sycxhyv7  于 12个月前  发布在  SQL Server
关注(0)|答案(4)|浏览(135)

I have a table ExchangeUserMailbox with columns

  • OrganisationID (GUID)
  • MailboxLocationName (STRING)
  • MailboxCode(STRING)
  • DayofMon (INT)
  • UPN(STRING)

The possible values for DayofMon are "15" and "30" (I have data coming to this table on every two days of month.. eg:- 14 th and 30th , or 15th and 31st )

I am running the following SQL script

SELECT 
    [OrganisationID] 
    ,[MailboxLocationName] AS Location 
    ,MailboxCode as LocationCode 
    ,DayofMon 
    ,COUNT(DISTINCT [UPN]) AS [Count]
FROM 
    [ExchangeUserMailbox]
GROUP BY 
    [OrganisationID], [MailboxLocationName], DayofMon, MailboxCode
HAVING  
    DayofMon = MAX(DayofMon)

My intention is to get all records with the maximum value of DayofMon and aggregate the count of UPNs.

But in the output window I am getting

OrganisationID                              Location        LocationCode  DayofMon Count

2EC8650F-8902-4CFB-BE0E-A218982EDEEC        Diffraction         DIF          1       3
2EC8650F-8902-4CFB-BE0E-A218982EDEEC        Diffraction         DIF          30      10

As you can see it aggregated dayofmon= 1 and dayofmon=30 . But I want to aggregate only dayofmon=30

What am I doing wrong here?

ia2d9nvy

ia2d9nvy1#

Try a hard_code ... having DayofMon > 15 ... I'm more of an Oracle guy, so from my experience there you can't put a Windowed function in the result of Having.

mwngjboj

mwngjboj2#

Try this: use max function on select.

SELECT 
    [OrganisationID] 
    ,[MailboxLocationName] AS Location 
    ,MailboxCode as LocationCode 
    ,Max(DayofMon) as [DayofMon]
    ,COUNT(DISTINCT [UPN]) AS [Count]
FROM 
    [ExchangeUserMailbox]
GROUP BY 
    [OrganisationID], [MailboxLocationName], MailboxCode
wtlkbnrh

wtlkbnrh3#

The HAVING clause is designed for for filtering by aggregated calculations, hence it is performed AFTER the GROUP BY is complete so it can access the aggregated values. What you need requires filtering BEFORE grouping is performed, so the having clause isn't helpful.

You could just filter the data where DayOfMon is in (28, 29, 30 or 31)

SELECT
        OrganisationID
      , MailboxLocationName AS Location
      , MailboxCode         AS LocationCode
      , 30                  AS DayofMon 
      , COUNT(DISTINCT UPN) AS [Count]
FROM [ExchangeUserMailbox]
WHERE DayofMon IN (28,29,30,31)
GROUP BY
        OrganisationID
      , MailboxLocationName
      , MailboxCode

or you could make use of ROW_NUMBER() like this:

SELECT
        OrganisationID
      , Location
      , LocationCode
      , 30                  AS DayofMon 
      , COUNT(DISTINCT UPN) AS [Count]
FROM (
        SELECT
                OrganisationID
              , MailboxLocationName   AS Location
              , MailboxCode           AS LocationCode
              , UPN
              , ROW_NUMBER() OVER (PARTITION BY OrganisationID
                                                , MailboxLocationName
                                                , MailboxCode
                                   ORDER BY DayofMon DESC) AS rn
        FROM [ExchangeUserMailbox]
        ) AS d
WHERE d.rn = 1
GROUP BY
        OrganisationID
      , Location
      , LocationCode
;

Which will give the number 1 to each row having the largest DayOfMon for each "partition"

jslywgbw

jslywgbw4#

The most straightforward way is using a subquery:

SELECT 
    eum.[OrganisationID] 
    ,eum.[MailboxLocationName] AS Location 
    ,eum.MailboxCode as LocationCode 
    ,eum.DayofMon 
    ,COUNT(DISTINCT [eum.UPN]) AS [Count]
FROM 
    [ExchangeUserMailbox] AS eum
WHERE eum.DayofMon =
      (SELECT MAX(eum2.DayofMon)
       FROM   [ExchangeUserMailbox] AS eum2
       WHERE  eum2.[OrganisationID] = eum.[OrganisationID]
       AND    eum2.[MailboxLocationName] = eum.[MailboxLocationName]
       AND    eum2.MailboxCode = eum.MailboxCode
      )
GROUP BY 
    eum.[OrganisationID], eum.[MailboxLocationName], eum.DayofMon, eum.MailboxCode

The "hard coded" solutions work, but are more brittle. This way, you are sure to always get the records with the highest DayofMon value (even if it is an unexpected value) and you get the actual value of DayofMon instead of a hard coded value.

相关问题