CODEWITHSUNDEEP
sql-serveraggregate
jalperin
jalperin

Reputation: 2674

Querying aggregate columns in a SQL Server SELECT statement

I have a SQL Server query which runs just fine -- until I add a computed column to the SELECT statement. Then I get an odd SQL Server error.

Here's the SQL:

SELECT
    outmail_.MessageID_, 
    CONVERT(VARCHAR(10),outmail_.Created_,120) AS 'Issue', 
    lyrReportSummaryData.mailed, 
    lyrReportSummaryData.successes, 
    COUNT(*) AS 'opens',
    COUNT(DISTINCT clicktracking_.MemberID_) AS 'unique_opens', 
    convert(decimal(3,1),((convert(float,[unique_opens]))/[successes]) * 100) AS 'Rate'
FROM 
    outmail_ 
RIGHT JOIN 
    clicktracking_ ON clicktracking_.MessageID_ = outmail_.MessageID_
RIGHT JOIN 
    lyrReportSummaryData ON lyrReportSummaryData.id = clicktracking_.MessageID_ 
GROUP BY  
    outmail_.MessageID_, CONVERT(VARCHAR(10), outmail_.Created_,120), 
    lyrReportSummaryData.mailed, lyrReportSummaryData.successes

The problem is the line beginning with the convert(decimal ... When it is included, I get the following error:

Error 8120: Column 'lyrReportSummaryData.unique_opens' is invalid in the select list because it is not contained in either an aggregate function or the GROUP BY clause.

I'm not sure how to resolve the error since I don't know how to use it in a GROUP BY clause (and it doesn't seem that I should need to do so).

Any suggestions for how to proceed? Thanks.

Upvotes: 0

Views: 484

Answers (1)

user111013
user111013

Reputation:

I'm sure someone with better DBA skills than me can point out a more efficient way of doing this, but...

If you perform the bulk of your query as an sub-query, you can then do the calculations on the result of your sub-query: 

 SELECT 
    MessageID_, 
    Issue, 
    mailed, 
    successes, 
    opens, 
    unique_opens, 
    convert(decimal(3,1),((convert(float,[unique_opens]))/[successes]) * 100) AS 'Rate'

 FROM 
 (SELECT
      outmail_.MessageID_, 
      CONVERT(VARCHAR(10),outmail_.Created_,120) AS 'Issue', 
      lyrReportSummaryData.mailed, 
      lyrReportSummaryData.successes, 
      COUNT(*) AS 'opens',
      COUNT(DISTINCT clicktracking_.MemberID_) AS 'unique_opens'


    FROM outmail_ 
    RIGHT JOIN clicktracking_ ON clicktracking_.MessageID_ = outmail_.MessageID_
    RIGHT JOIN lyrReportSummaryData ON lyrReportSummaryData.id = clicktracking_.MessageID_ 

    GROUP BY  outmail_.MessageID_, CONVERT(VARCHAR(10), outmail_.Created_,120), lyrReportSummaryData.mailed, lyrReportSummaryData.successes
   ) subquery /* was 'g' */

Effectively what this does is runs the grouping, and then based on that, does the calculation afterwards.

Subqueries must be given an alias (in this instance 'subquery') - even if you don't use that alias name.

Upvotes: 2

Related Questions