Select distinct rows with max value of another column

问题

I am querying a document management database where documents can have many versions. So my query below returns the same document multiple times because of the versioning.

 SELECT distinct d.DOCNUM, d.DOCNAME,P.FolderName
    FROM DOCUMENT as d with (nolock)
    inner join Project_T as pt on d.docnum=pt.item_id
    inner join Projects p on pt.prj_id=p.prj_id
   where d.type = 'Personal' and d.owner like '%67360'

Results:

DOCNUM     | DOCNAME     | FOLDERNAME |
-----------+----------+---------------+
123         Article        Jonathan
123         Article part1  Jonathan
256         Meeting Notes  Jonathan
5697        Memo           Jonathan

Ideally docnum 123 should only return once.

I have joined the table that holds the versions for each document and am selecting the latest version (MAX(h.version) so my query should return unique document numbers per row.

    SELECT distinct d.DOCNUM, h.version, d.DOCNAME,P.FolderName
    FROM DOCUMENT as d with (nolock)
    inner join DOCHISTORY as h on h.DOCNUM = d.docnum
    inner join Project_T as pt on d.docnum=pt.item_id
    inner join Projects p on pt.prj_id=p.prj_id
   where d.type = 'Personal' and d.owner like '%67360'
   and d.SECURITY = 'P'
   AND h.VERSION = (SELECT MAX(x.version) FROM DOCHISTORY as X 
                    where x.docnum = d.DOCNUM)

but this time I get the latest versions however, where docname is slightly different, it returns the same docnum multiple times. see docnum 123 below.

DOCNUM     | Version      | DOCNAME     | FOLDERNAME |
-----------+--------------+-------------+------------+
123          9              Article        Jonathan
123          9              Article part1  Jonathan
256          1              Meeting Notes  Jonathan
5697         21             Memo           Jonathan

I need to display the docname column in my report. Is there a way aorund this please?

回答1:

Yes, you can select the top 1 docname. Replace your field selection list with this:

SELECT distinct d.DOCNUM
 ,h.version 
 ,DOCNAME = (SELECT TOP 1 d2.DOCNAME FROM DOCUMENT d2, DOCHISTORY h2 
 WHERE d2.DOCNUM = d.DOCNUM and h2.DOCNUM = d.docnum)
 ,P.FolderName

....(rest of code)

In the above output, doc 123 version 9 should appear as Article.

回答2:

What you are trying to do is much easier with window functions. Here is one way:

select DOCNUM, version, DOCNAME, FolderName
from (SELECT d.DOCNUM, h.version, d.DOCNAME, P.FolderName,
             max(h.version) over (partition by d.docnum) as maxversion
      FROM DOCUMENT  d with (nolock)
           inner join DOCHISTORY as h on h.DOCNUM = d.docnum
           inner join Project_T as pt on d.docnum=pt.item_id
           inner join Projects p on pt.prj_id=p.prj_id
     where d.type = 'Personal' and d.owner like '%67360' and d.SECURITY = 'P'
    ) d
where version = maxversion;

The expression max(h.version) over (partition by d.docnum) is a construct called a window function (or analytic function in some databases). What is does is return the maximum of h.version for each value of d.docnum (based on the partition by clause). This is ANSI standard SQL and most databases support this functionality.

Your version is probably not working because the subquery that calculates the maximum value is not using the same filtering criteria as the outer query.

来源：https://stackoverflow.com/questions/18213088/select-distinct-rows-with-max-value-of-another-column

标签

sql

sql-server

tsql

max

distinct