0

I have 3 tables, users (user info), mghours (volunteer time), mgtraining (training time). I want a result that lists some user columns, sum of volunteer hours and sum of two types of training:

uid lastname firstname mgstatus trainingyear vol at ce

I use the following and as long as I have the two fields from the mgtraining table and the LEFT OUTER JOIN for those in the query, I get skewed results. If I take that out of the query, my user info and vol columns are fine.

SELECT users.uid AS 'uid',
users.lastname AS 'lastname',
users.firstname AS 'firstname',
users.mgstatus AS 'mgstatus',
users.trainingyear AS 'trainingyear',
sum(mghours.hourstime) AS 'vol',
SUM(IF(mgtraining.trainingtype = 'AT', mgtraining.trainingtime, 0)) AS 'at',
SUM(IF(mgtraining.trainingtype = 'CE', mgtraining.trainingtime, 0)) AS 'ce'
FROM users
LEFT OUTER JOIN mghours ON users.uid = mghours.uid AND year(mghours.hoursdate) = YEAR(CURDATE())
LEFT OUTER JOIN mgtraining ON users.uid = mgtraining.uid AND year(mgtraining.trainingdate) = YEAR(CURDATE())
WHERE users.coid = 33
GROUP BY users.uid
ORDER BY users.lastname ASC

The users.coid = 33 references a county ID.

I've been pouring over this for hours. What am I doing wrong?

3
  • 1
    The 2 joins from the "master" (users) table to the 2 "details" tables (mghours and mgtraining) produces a mini cartesian product and you'll get wrong results for any user that has more than 1 related rows in each details table. Commented Jun 6, 2015 at 23:32
  • 1
    To solve the issue, you can use derived tables (where you do the group by, separately for each "details" table) and then left join users to these derived tables. See this answer of mine (you can skip the first paragraphs and go to Option 3): Help with this query Commented Jun 6, 2015 at 23:37
  • @ypercube I have to generate this kind of report a lot in my work and this particular DB (structure) was giving me headaches. I was getting a "mini-cartesian" product and was going about it all wrong. I had tried to do it with sub-queries but your reference made it much clearer. And, it worked! And fast! Thanks again.
    – RJ Jacob
    Commented Jun 7, 2015 at 15:23

1 Answer 1

1

JOINs together with aggregates (eg, SUM()) often give you bigger values than you expect. To see this, remove the SUMs and GROUP BY and stare at the output. You may see, for example, the same mghours.hourstime showing up multiple times. Hence, SUM(mghours.hourstime) is inflated.

This is happening because JOIN mgtraining is coming up with multiple rows in mgtraining for each row in mghours.

The solution is to do the query in steps. Notice how the inner query avoids inflating "vol":

SELECT  x.*,
        SUM(...) AS 'at
    FROM (
        SELECT  ...,
                SUM(mghours.hourstime) AS vol
            FROM ...
            GROUP BY ...
         ) x
    LEFT JOIN mgtraining ON ...
    GROUP BY ...
    ORDER BY ...;

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.