Understanding SQL Joins - Crossed Wires (
Page 4 of 10 )
The examples on the previous page dealt with a single table, so the question of a join never arose. But often, using a single table returns an incomplete picture; you need to combine data from two or more tables in order to see a more accurate result.
That's where joins come in - they allow you to combine two or more tables, and to massage the data within those tables, in a variety of different ways. For example,
SELECT * FROM a,b;
Here's the output:
+----+------+----+------+
| a1 | a2 | b1 | b2 |
+----+------+----+------+
| 10 | u | 10 | p |
| 20 | v | 10 | p |
| 30 | w | 10 | p |
| 40 | x | 10 | p |
| 50 | y | 10 | p |
| 60 | z | 10 | p |
| 10 | u | 20 | q |
| 20 | v | 20 | q |
| 30 | w | 20 | q |
| 40 | x | 20 | q |
| 50 | y | 20 | q |
| 60 | z | 20 | q |
+----+------+----+------+
12 rows in set (0.00 sec)
And that's your very first join!
In this case, columns from both tables are combined to produce a resultset containing all possible combinations. This kind of join is referred to as a "cross join", and the number of rows in the joined table will be equal to the product of the number of rows in each of the tables used in the join. You can see this from the example above - table "a" has 6 rows, table "b" has 2 rows, and so the joined table has 6x2 = 12 rows.
There are two basic types of SQL joins - the "inner join" and the "outer join". Inner joins are the most common - the one you just saw was an example of an inner join - and also the most symmetrical, since they require a match in each table which forms a part of the join. Rows which do not match are excluded from the final resultset.
As you might imagine, a cross join like the one above can have huge implications for the performance of your database server. In order to illustrate, look what happens when I add a third table to the join above:
SELECT * FROM a,b,c;
Here's the output:
+----+------+----+------+-----+------+
| a1 | a2 | b1 | b2 | c1 | c2 |
+----+------+----+------+-----+------+
| 10 | u | 10 | p | 90 | m |
| 20 | v | 10 | p | 90 | m |
| 30 | w | 10 | p | 90 | m |
| 40 | x | 10 | p | 90 | m |
| 50 | y | 10 | p | 90 | m |
| 60 | z | 10 | p | 90 | m |
| 10 | u | 20 | q | 90 | m |
| 20 | v | 20 | q | 90 | m |
| 30 | w | 20 | q | 90 | m |
| 40 | x | 20 | q | 90 | m |
| 50 | y | 20 | q | 90 | m |
| 60 | z | 20 | q | 90 | m |
| 10 | u | 10 | p | 100 | n |
| 20 | v | 10 | p | 100 | n |
| 30 | w | 10 | p | 100 | n |
| 40 | x | 10 | p | 100 | n |
| 50 | y | 10 | p | 100 | n |
| 60 | z | 10 | p | 100 | n |
| 10 | u | 20 | q | 100 | n |
| 20 | v | 20 | q | 100 | n |
| 30 | w | 20 | q | 100 | n |
| 40 | x | 20 | q | 100 | n |
| 50 | y | 20 | q | 100 | n |
| 60 | z | 20 | q | 100 | n |
| 10 | u | 10 | p | 110 | o |
| 20 | v | 10 | p | 110 | o |
| 30 | w | 10 | p | 110 | o |
| 40 | x | 10 | p | 110 | o |
| 50 | y | 10 | p | 110 | o |
| 60 | z | 10 | p | 110 | o |
| 10 | u | 20 | q | 110 | o |
| 20 | v | 20 | q | 110 | o |
| 30 | w | 20 | q | 110 | o |
| 40 | x | 20 | q | 110 | o |
| 50 | y | 20 | q | 110 | o |
| 60 | z | 20 | q | 110 | o |
+----+------+----+------+-----+------+
36 rows in set (0.05 sec)
Though each of the tables used in the join contains less than ten records each, the final joined result contains 6x2x3=36 records. This might not seem like a big deal when all you're dealing with are three tables containing a total of 11 records, but imagine what would happen if you had three tables, each containing 100 records, and you decided to cross join them...
/ 229 
