同一表上的2 LEFT OUTER JOIN会冻结服务器

Question

I am trying to run a query on table 'apst_mailings' storing the content of each newsletter we send to our subscribers. Each time we attempt to send the email to an individual we insert a row in apst_mailings_accuses reporting the time and state of the sending, and the ID of the newletter. I wanna list the newsletters and count the total sent and successfully sent for each.

    SELECT m.id AS code_mailing, 
    COUNT(a.adh_code) AS num, COUNT(b.adh_code) AS succes
    FROM apst_mailings AS m
        LEFT OUTER JOIN apst_mailings_accuses AS a
            ON a.id_mailing = m.id
        LEFT OUTER JOIN apst_mailings_accuses AS b
            ON b.id_mailing = m.id
            AND b.etat = 'succes'
    GROUP BY m.id

And it simply hangs the server forever. I have tried each join in separated queries and it worked without problem:

// Counts the email sent per mailing
    SELECT m.id AS code_mailing, 
    COUNT(a.adh_code) AS num
    FROM apst_mailings AS m
        LEFT OUTER JOIN apst_mailings_accuses AS a
            ON a.id_mailing = m.id
    GROUP BY m.id

    SELECT m.id AS code_mailing, 
    COUNT(b.adh_code) AS succes
    FROM apst_mailings AS m
        LEFT OUTER JOIN apst_mailings_accuses AS b
            ON b.id_mailing = m.id
            AND b.etat = 'succes'
    GROUP BY m.id

I could split my query but the reason why it doesn't work really bothers me. Can anyone explain please?

Thanks!

Answer 1

You can what you want in a much simpler way by using a single join and using SUM to perform a conditional count.

SELECT
   m.id AS code_mailing, 
   COUNT(a.adh_code) AS num,
   SUM(a.etat = 'succes') AS succes
FROM apst_mailings AS m
LEFT OUTER JOIN apst_mailings_accuses AS a
ON a.id_mailing = m.id
GROUP BY m.id

But the reason why your query doesn't work is because you're joining ALL the rows in subquery a with ALL the matching rows in subquery b in a huge cross join. This could generate a huge temporary result set, which is presumably why the query takes forever to terminate. And even if it did terminate, your counts would be completely off - they would be the product of the two counts.

To solve it do the GROUP BY first. Then JOIN the result of that to your main table.

SELECT
   m.id AS code_mailing, 
   IFNULL(a.num, 0) AS num,
   IFNULL(b.succes, 0) AS succes
FROM apst_mailings AS m
LEFT OUTER JOIN (
   SELECT id_mailing, COUNT(adh_code) AS num
   FROM apst_mailings_accuses 
   GROUP BY id_mailing
) a
ON a.id_mailing = m.id
LEFT OUTER JOIN (
    SELECT id_mailing, COUNT(adh_code) AS succes
    FROM apst_mailings_accuses
    WHERE etat = 'succes'
    GROUP BY id_mailing
) b
ON b.id_mailing = m.id