MySQL with JOINS is very slow - mysql

Question

I have this query:
select
b.user as user1, b.timestamp as ts1,
c.user as user2, c.timestamp as ts2,
d.user as user3, d.timestamp as ts3,
e.user as user4, e.timestamp as ts4,
f.user as user5, f.timestamp as ts5,
g.user as user6, g.timestamp as ts6,
h.user as user7, h.timestamp as ts7,
i.user as user8, i.timestamp as ts8,
j.user as user9, j.timestamp as ts9,
k.user as user10, k.timestamp as ts10,
a.beschreibung, a.auftragsnummer, a.faellig, a.subkunde,
(SELECT firma from kunden where id=a.kunde limit 0,1) as kunde,
(SELECT kommision from kommision where id=a.kommision limit 0,1) as kommision
from auftrag a
left join details b on (b.beschreibung='Step1' AND b.auftrags_id=a.id)
left join details c on (c.beschreibung='Step2' AND c.auftrags_id=a.id)
left join details d on (d.beschreibung='Step3' AND d.auftrags_id=a.id)
left join details e on (e.beschreibung='Step4' AND e.auftrags_id=a.id)
left join details f on (f.beschreibung='Step5' AND f.auftrags_id=a.id)
left join details g on (g.beschreibung='Step6' AND g.auftrags_id=a.id)
left join details h on (h.beschreibung='Step7' AND h.auftrags_id=a.id)
left join details i on (i.beschreibung='Step8' AND i.auftrags_id=a.id)
left join details j on (j.beschreibung='Step9' AND j.auftrags_id=a.id)
left join details k on (k.beschreibung='Step10' AND k.auftrags_id=a.id)
where a.erledigt='1'
It runs very, very slow and take about 1 Minute to get a resultset.
Table "auftrag" has 820 rows and table "details" about 7000 rows.
What am I doing wrong?
Thanks
Patrick
CREATE TABLE auftrag
(
id INT(10) NOT NULL AUTO_INCREMENT,
beschreibung VARCHAR(255) NULL DEFAULT '0',
auftragsnummer VARCHAR(50) NULL DEFAULT '0',
faellig VARCHAR(50) NULL DEFAULT NULL,
kunde INT(11) NULL DEFAULT NULL,
subkunde VARCHAR(50) NULL DEFAULT NULL,
kommision VARCHAR(50) NULL DEFAULT NULL,
notiz TEXT NULL,
werbeanbringung TEXT NULL,
erledigt INT(1) NULL DEFAULT NULL,
INDEX Schlüssel 1 (id),
INDEX Schlüssel 2 (id, auftragsnummer, kunde, subkunde, beschreibung)
) COLLATE='latin1_swedish_ci' ENGINE=MyISAM AUTO_INCREMENT=850 ;
CREATE TABLE details
(
id INT(11) NOT NULL AUTO_INCREMENT,
auftrags_id VARCHAR(50) NULL DEFAULT '0',
beschreibung VARCHAR(50) NULL DEFAULT '0',
user VARCHAR(50) NULL DEFAULT '0',
timestamp VARCHAR(50) NULL DEFAULT '0',
notiz VARCHAR(255) NULL DEFAULT NULL,
INDEX Schlüssel 2 (user, timestamp, beschreibung, auftrags_id),
INDEX Schlüssel 1 (id, user, timestamp, beschreibung)
) COLLATE='latin1_swedish_ci' ENGINE=MyISAM AUTO_INCREMENT=7260 ;

Answer 1

You have no indexes on your tables that are useful for this query, and you are joining one table against another repeatedly.
On the auftrag table add an index on the erledigt column.
On the details table add an index covering both the beschreibungand the auftrags_id columns
It may also be possible to avoid most of the joins if you are prepared to process the returned data afterwards (ie, split up a returned field into an array). But a bit messy:-
SELECT
GROUP_CONCAT(CONCAT_WS('##', b.beschreibung, b.user, b.timestamp)),
a.beschreibung, a.auftragsnummer, a.faellig, a.subkunde,
(SELECT firma from kunden where id=a.kunde limit 0,1) as kunde,
(SELECT kommision from kommision where id=a.kommision limit 0,1) as kommision
FROM auftrag a
LEFT OUTER JOIN details b on b.auftrags_id=a.id AND b.beschreibung IN ('Step1', 'Step2', 'Step3', 'Step4', 'Step5', 'Step6', 'Step7', 'Step8', 'Step9', 'Step10')
WHERE a.erledigt='1'
GROUP BY a.id
EDIT - I have just noticed that the id on the auftrag table is an INT field (as I would expect) but the auftrags_id column on the details table is a VARCHAR(50) field. This will dramatically affect performance as it needs to convert the values for every single comparison.
Note that Gordons comment above on the order of the fields on the index does apply (how much will depend on the actual data - for example how many other values of beschreibung there are), but without fixing this problem with the mismatched data types changing the index will almost certainly not help.

Answer 2

For your query, you want the following indexes:
auftrag(erledigt, id)
details(auftrags_id, beschreibung)
kunden(kunde, firma)
You can also rewrite the query as a conditional aggregation, but the right indexes will be a big help.
Note: the order of the columns in the composite indexes is important.

Answer 3

This is probably a case where InnoDB will be a clear winner over MyISAM. In details, have PRIMARY KEY(auftrags_id, beschreibung).
Meanwhile, in auftrag, PRIMARY KEY(id). And the indexes Gordon suggests for auftrag and kunden.