Question

Sorry for the mega title… I was trying to be descriptive enough. I’ve got a table that contains event attendance data that has some erroneous data in it. The table definition is kind of like this:

id (row id)
date
company_name
attendees

It ended up with some cases where for a given date, there are two entries matching a company_name and date but one has attendees=0 and the other has attendees>0. In those cases, I want to discard the ones where attendees=0.

I know you can’t join on the same table while deleting, so please consider this query to be pseudocode that shows what I want to accomplish.

DELETE FROM attendance a WHERE a.attendees=0 AND a.date IN (SELECT b.date FROM attendance b WHERE b.attendees > 0 AND b.company_name = a.company_name);

I also tried to populate a temporary table with the ids of the rows I want to delete, but that query hangs because of the IN (SELECT …) clause. My table has thousands of rows so that just maxes out the CPU and then times out.

Answer 1

This ugly thing should work (using alias permit to avoid the You can't specify target table for update in FROM clause error)

DELETE FROM attendance
WHERE (attendees, date, company_name)
IN (SELECT c.a, c.d, c.c 
    FROM
     (SELECT MIN(attendees) a, date d, company_name c
      FROM attendance
      GROUP BY date, company_name
      HAVING COUNT(*) > 1) as c);

SqlFiddle