SQL Statement - SQL Matrix - mysql

Question

For the following Table, is it possible, to create an SQL Statement to create a data Matrix or view?
Table:
TeamA|TeamB|Won|Lost
--------------------
A | B | 5 | 3
A | C | 2 | 4
A | D | 9 | 1
B | E | 5 | 5
C | A | 2 | 4
Result-Matrix:
| A | B | C | D | E
----------------------------
A | 0 | 2 | -2 | 8 | 0
B | 0 | 0 | 0 | 0 | 0
C | -2 | 0 | 0 | 0 | 0

Answer 1

There are two ways that you can pivot data in MySQL. If you know the values ahead of time (teams) then you will hard-code the values or you can use a prepared statement to generate dynamic sql.
A static version would be:
select TeamA,
max(case when TeamB = 'A' then won - lost else 0 end) as A,
max(case when TeamB = 'B' then won - lost else 0 end) as B,
max(case when TeamB = 'C' then won - lost else 0 end) as C,
max(case when TeamB = 'D' then won - lost else 0 end) as D,
max(case when TeamB = 'E' then won - lost else 0 end) as E
from yourtable
group by TeamA;
See SQL Fiddle with Demo
If you want to use a dynamic version with a prepared statement, the code would be:
SET #sql = NULL;
SELECT
GROUP_CONCAT(DISTINCT
CONCAT(
'MAX(CASE WHEN TeamB = ''',
TeamB,
''' THEN won - lost else 0 END) AS `',
TeamB, '`'
)
) INTO #sql
from
(
select *
from yourtable
order by teamb
) x;
SET #sql
= CONCAT('SELECT TeamA, ', #sql, '
from yourtable
group by TeamA');
PREPARE stmt FROM #sql;
EXECUTE stmt;
DEALLOCATE PREPARE stmt;
See SQL Fiddle with Demo.
Edit #1, after thinking about this I would actually do this a slight bit different. I would generate a true matrix os the data where the teams appeared in both the row and the column. To do this you would first use a UNION ALL query to get all teams in two columns:
select teama Team1, teamb Team2,
won-lost Total
from yourtable
union all
select teamb, teama,
won-lost
from yourtable
See SQL Fiddle with Demo. Once that is done, then you would pivot the data:
select Team1,
coalesce(max(case when Team2 = 'A' then Total end), 0) as A,
coalesce(max(case when Team2 = 'B' then Total end), 0) as B,
coalesce(max(case when Team2 = 'C' then Total end), 0) as C,
coalesce(max(case when Team2 = 'D' then Total end), 0) as D,
coalesce(max(case when Team2 = 'E' then Total end), 0) as E
from
(
select teama Team1, teamb Team2,
won-lost Total
from yourtable
union all
select teamb, teama,
won-lost
from yourtable
) src
group by Team1;
See SQL Fiddle with Demo. Which gives a more detailed result of:
| TEAM1 | A | B | C | D | E |
-------------------------------
| A | 0 | 2 | -2 | 8 | 0 |
| B | 2 | 0 | 0 | 0 | 0 |
| C | -2 | 0 | 0 | 0 | 0 |
| D | 8 | 0 | 0 | 0 | 0 |
| E | 0 | 0 | 0 | 0 | 0 |

Answer 2

You cannot create a SQL statement or view that has a variable number of columns. In standard SQL, you can pivot the data by doing something like:
select TeamA,
max(case when TeamB = 'A' then won - lost end) as A,
max(case when TeamB = 'B' then won - lost end) as B,
max(case when TeamB = 'C' then won - lost end) as C,
max(case when TeamB = 'D' then won - lost end) as D,
max(case when TeamB = 'E' then won - lost end) as E
from t
group by TeamA
order by 1
Some databases support a pivot statement.
To do this generically, you would have to create a SQL statement as a string, and then execute it (often called dynamic SQL). Such a statement could be produced by SQL, by a stored procedure, in Excel, or some other programming tool. It then would need to be executed.
Let me repeat: any given SQL statement has a pre-defined set of columns. You cannot vary the number of columns.