Select rows grouped by a column having max aggregate

Question

Given the following data set, how would I find the email addresses that were references for the most ApplicationID s that have an "Accepted" decision?

CREATE TABLE IF NOT EXISTS `EmailReferences` (
  `ApplicationID` INT NOT NULL,
  `Email` VARCHAR(45) NOT NULL,
  PRIMARY KEY (`ApplicationID`, `Email`)
);
INSERT INTO EmailReferences (ApplicationID, Email)
VALUES
(1, 'ref10@test.org'), (1, 'ref11@test.org'), (1, 'ref12@test.org'),
(2, 'ref20@test.org'), (2, 'ref21@test.org'), (2, 'ref22@test.org'),
(3, 'ref11@test.org'), (3, 'ref31@test.org'), (3, 'ref32@test.org'),
(4, 'ref40@test.org'), (4, 'ref41@test.org'), (4, 'ref42@test.org'),
(5, 'ref50@test.org'), (5, 'ref51@test.org'), (5, 'ref52@test.org'),
(6, 'ref60@test.org'), (6, 'ref11@test.org'), (6, 'ref62@test.org'),
(7, 'ref70@test.org'), (7, 'ref71@test.org'), (7, 'ref72@test.org'),
(8, 'ref10@test.org'), (8, 'ref81@test.org'), (8, 'ref82@test.org')
;

CREATE TABLE IF NOT EXISTS `FinalDecision` (
  `ApplicationID` INT NOT NULL,
  `Decision` ENUM('Accepted', 'Denied') NOT NULL,
  PRIMARY KEY (`ApplicationID`)
);
INSERT INTO FinalDecision (ApplicationID, Decision)
VALUES
(1, 'Accepted'), (2, 'Denied'),
(3, 'Accepted'), (4, 'Denied'),
(5, 'Denied'),   (6, 'Denied'),
(7, 'Denied'),   (8, 'Accepted')
;

Fiddle of same: http://sqlfiddle.com/#!9/03bcf2/1

Initially, I was using LIMIT 1 and ORDER BY CountDecision DESC , like so:

SELECT  er.email, COUNT(fd.Decision) AS CountDecision
FROM    EmailReferences AS er
JOIN    FinalDecision AS fd ON er.ApplicationID = fd.ApplicationID
WHERE   fd.Decision = 'Accepted'
GROUP   BY er.email
ORDER   BY CountDecision DESC
LIMIT   1
;

However, it occurred to me that I could have multiple email addresses that referred different "most accepted" decisions (ie, a tie, so to speak), and those would be filtered out (is that the right phrasing?) with the LIMIT keyword.

I then tried a variation on the above query, replacing the ORDER BY and LIMIT lines with:

HAVING MAX(CountDecision)

But I realized that that's only half a statement: MAX(CountDecision) needs to be compared to something. I just don't know what.

Any pointers would be much appreciated. Thanks!

Note: this is for a homework assignment.

Update: To be clear, I'm trying to find value and count of Email s from EmailReferences . However, I only want rows that have FinalDecision.Decision = 'Accepted' (on matching ApplicantID s). Based on my data, the result should be:

Email          | CountDecision
---------------+--------------
ref10@test.org | 2
ref11@test.org | 2

Answer 1

Basically you need to do 2 things... first, you need to find what is the maxCount and then find the records with max count.

Now you can combine these two steps in a single nested query, or store the result in a variable and use it in a second query. Personally I try to avoid inner queries as they cause performance issue and are more complex to read, therefore I am using the variable option here:

-- Find out what max count is and store it in a variable
SELECT  @maxcount := COUNT(fd.Decision) AS CountDecision
FROM    EmailReferences AS er
JOIN    FinalDecision AS fd ON er.ApplicationID = fd.ApplicationID
WHERE   fd.Decision = 'Accepted'
GROUP   BY er.email
ORDER BY CountDecision desc
Limit 1;

-- get emails with @maxcount
SELECT  er.Email, COUNT(fd.Decision) AS CountDecision
FROM    EmailReferences AS er
JOIN    FinalDecision AS fd ON er.ApplicationID = fd.ApplicationID
WHERE   fd.Decision = 'Accepted'
GROUP   BY er.email
HAVING  COUNT(fd.Decision) = @maxcount;

Answer 2

MySQL still lack window functions, but when version 8 is production ready, this becomes easier. So for fuure reference, or for those databases like Mariadb that already have window functions:

 CREATE TABLE IF NOT EXISTS `EmailReferences` ( `ApplicationID` INT NOT NULL, `Email` VARCHAR(45) NOT NULL, PRIMARY KEY (`ApplicationID`, `Email`) ); 

 INSERT INTO EmailReferences (ApplicationID, Email) VALUES (1, 'ref10@test.org'), (1, 'ref11@test.org'), (1, 'ref12@test.org'), (2, 'ref20@test.org'), (2, 'ref21@test.org'), (2, 'ref22@test.org'), (3, 'ref30@test.org'), (3, 'ref31@test.org'), (3, 'ref32@test.org'), (4, 'ref40@test.org'), (4, 'ref41@test.org'), (4, 'ref42@test.org'), (5, 'ref50@test.org'), (5, 'ref51@test.org'), (5, 'ref52@test.org'), (6, 'ref60@test.org'), (6, 'ref11@test.org'), (6, 'ref62@test.org'), (7, 'ref70@test.org'), (7, 'ref71@test.org'), (7, 'ref72@test.org'), (8, 'ref10@test.org'), (8, 'ref81@test.org'), (8, 'ref82@test.org') ; 

 CREATE TABLE IF NOT EXISTS `FinalDecision` ( `ApplicationID` INT NOT NULL, `Decision` ENUM('Accepted', 'Denied') NOT NULL, PRIMARY KEY (`ApplicationID`) ); 

 INSERT INTO FinalDecision (ApplicationID, Decision) VALUES (1, 'Accepted'), (2, 'Denied'), (3, 'Accepted'), (4, 'Denied'), (5, 'Denied'), (6, 'Denied'), (7, 'Denied'), (8, 'Accepted') ; 

 select email, CountDecision from ( SELECT er.email, COUNT(fd.Decision) AS CountDecision , max(COUNT(fd.Decision)) over() maxCountDecision FROM EmailReferences AS er JOIN FinalDecision AS fd ON er.ApplicationID = fd.ApplicationID WHERE fd.Decision = 'Accepted' GROUP BY er.email ) d where CountDecision = maxCountDecision 

\nemail |  CountDecision \n:------------- |  ------------: \nref10@test.org |  2 \n

dbfiddle here

Answer 3

For example...

SELECT a.*
  FROM 
     ( SELECT x.email
            , COUNT(*) total
         FROM emailreferences x
         JOIN finaldecision y
           ON y.applicationid = x.applicationid
        WHERE y.decision = 'accepted'
        GROUP
           BY x.email
     ) a
  JOIN
     ( SELECT COUNT(*) total
         FROM emailreferences x
         JOIN finaldecision y
           ON y.applicationid = x.applicationid
        WHERE y.decision = 'accepted'
        GROUP
           BY x.email
        ORDER 
           BY total DESC 
        LIMIT 1
     ) b
    ON b.total = a.total;