views:

64

answers:

3

I have a master table called "parent" and a related table called "childs"

Now I run a query against the master table to update some values with the sum from the child table like this.

UPDATE master m SET
    quantity1 = (SELECT SUM(quantity1) FROM childs c WHERE c.master_id = m.id),
    quantity2 = (SELECT SUM(quantity2) FROM childs c WHERE c.master_id = m.id),
    count =  (SELECT COUNT(*) FROM childs c WHERE c.master_id = m.id)
WHERE master_id = 666;

Which works as expected but is not a good style because I basically make multiple SELECT querys on the same result. Is there a way to optimize that? (Making a query first and storing the values is not an option.

I tried this:

UPDATE master m SET (quantity1, quantity2, count) = (
    SELECT SUM(quantity1), SUM(quantity2), COUNT(*)
       FROM childs c WHERE c.master_id = m.id
) WHERE master_id = 666;

but that doesn't work.

Update: Here is the solution, thanks to everbody:

You can do something like this:

UPDATE master m
  INNER JOIN childs c ON m.master_id = c.master_id
SET master.quantity1 = c.quantity1,
    master.count = 1

If you have only one child record at a time. However if you want to use a group function like SUM() in the joined table that doesn't work. Either you get a "Invalid use of group function" if you leave the "group by" part or a "You have an error in your sql syntax if you use "GROUP BY c.master_id"

-- This doesnt work :(
UPDATE master m
  INNER JOIN childs c ON m.master_id = c.master_id
SET master.quantity1 = SUM(c.quantity1),
    master.count = COUNT(c.*)
GROUP by c.master_id

The solution is to use JOIN with a subquery:

UPDATE master m
  INNER JOIN 
  (
    SELECT   master_id,
             SUM(quantity1) as quantity1,
             COUNT(*) as count
    FROM     childs c 
    GROUP BY master_id
  ) c
  ON c.master_id = m.master_id
SET m.quantity1 = c.quantity1,
    m.count = c.count
WHERE   m.master_id = 666;

But since this pulls every row from the childtable the overhead would likely be bigger than using more subqueries like in the original sql. So you should add a WHERE clause to the joined table to get only the rows you need.

Another interesting approach is this syntax, which does the same as the JOIN with the WHERE clause but you should only use if if you want to update all rows with the same values and your subquery only returns one row, since the result from the subquery gets appended to the result and can be used like any column.

UPDATE master m,
    (
      SELECT SUM(c.quantity1) as sum_of_quantity,
      COUNT(*) as rowcount FROM child c WHERE c.master_id = 666
    ) as c
SET m.quantity1 = c.sum_of_quantity,
    m.count = c.rowcount
WHERE m.master_id = 666;
+1  A: 

I don't know if it is allowed in MySQL, but SQL Server allows you to use the result of a select in an update.

UPDATE master m SET
  quantity1 = c.quantity1
  , quantity2 = c.quantity2
  , count = c.count
FROM  master m
      INNER JOIN (
        SELECT  master_id
                , quantity1 = SUM(quantity1)
                , quantity2 = SUM(quantity2)
                , count = COUNT(*) 
        FROM    childs c 
        WHERE   master_id = 666
        GROUP BY
                master_id
      ) c ON c.master_id = m.master_id
Lieven
+1, that is exactly how I'd do it
KM
If you copy / paste the correct syntax for MySQL from my answer I'll delete that one.
Wrikken
@Wrikken, if you are certain this doesn't work with MySQL, by all means, OP should accept your answer.
Lieven
Wrikken
+1 Because this was the first one. However Wrikken is right. For MySQL the SET ... part has be be after the JOIN.
SchlaWiener
A: 

You could select your data into a temporary table, and then update using that data.

If you also want to insert "new" data in the same roundtrip, look into INSERT INTO ... SELECT FROM ... ON DUPLICATE KEY UPDATE ...

If you already are doing inserts if row doesn't exist, then that would be redundant with this example.

example:

INSERT INTO master m (id, quantity1, quantity2, count)
SELECT master_id, SUM(quantity1) q1, SUM(quantity2) q1, COUNT(*) c
   FROM childs
   GROUP BY master_id
ON DUPLICATE KEY UPDATE 
   m.quantity1 = q1,
   m.quantity2 = q2,
   m.count = c

NOTE! This is untested code, but I think it should be possible to backreference the select result in the UPDATE.

Syntax reference: http://dev.mysql.com/doc/refman/5.0/en/insert.html

jishi
+2  A: 

Rewriting Lieven's solution to MySQL:

UPDATE master m 
JOIN (
    SELECT  master_id
            , SUM(quantity1) as quantity1
            , SUM(quantity2) as quantity2 
            , COUNT(*) as count
    FROM    childs c 
    GROUP BY
            master_id
) c
ON c.master_id = m.master_id
SET
  m.quantity1 = c.quantity1
  ,m.quantity2 = c.quantity2
  ,m.count = c.count
WHERE   m.master_id = 666;
Wrikken
Works like a charm but (look at the updates in my question) don't forget to add a where clause to the subquery or you will pull millions of rows from the DB just to join one record ;)
SchlaWiener