Select rows that match arbitrary criterias - mysql

Question

I have a MySql table like this:
ID Category Value
A LOCATION COPENHAGEN
A LOCATION MADRID
B LOCATION MADRID
C LOCATION MADRID
C LOCATION AARHUS
C POSITION WEBDESIGN
D POSITION SYSADMIN
D POSITION WEBDESIGN
A GENDER MALE
B GENDER MALE
C GENDER FEMALE
Id, Category and Values together forms the key. Note that they are dynamic (user defined) and the table will be large.
Now I want to get all ID that match some criteria. To be returned, it is enough for the ID to match one value within each Category.
For example:
GIVE ME ALL IDS WHO HAVE
((LOCATION = COPENHAGEN OR LOCATION = MADRID) AND (GENDER = MALE))
should return A, B
Another example:
GIVE ME ALL IDS WHO HAVE
((POSITION = SYSADMIN OR POSITION = WEBDESIGN) AND (GENDER = FEMALE) AND (LOCATION = COPENHAGEN OR LOCATION = MADRID))
should return C
The returned ids are to be used as a subquery for another query - so performance matters.
UPDATE
I've created this sql fiddle with sample data and a proposed not working solution.

Answer 1

[Removed initial response as it was non-functional. In making it functional I came up with the query in the second edit which is much preferable.]
Edit
It seems like you're trying to create a schema to keep track of people working in various locations, and from your data it seems like they might hold multiple titles among various offices. The below schema would allow you to define each person, job, and location only once and then connect them using a linking table.
TABLE People
p_id INT AI PK
p_name VARCHAR
p_sex BOOL
...
TABLE Offices
o_ID INT AI PK
o_name VARCHAR
o_location VARCHAR
...
TABLE Jobs
j_ID INT AI PK
j_name VARCHAR
...
TABLE People_Jobs_Offices --linking table
p_ID INT PK
o_ID INT PK
j_ID INT PK
Now get all of the Sysadmins in Madrid or Copenhagen:
SELECT *
FROM People_Jobs_Offices pjo
INNER JOIN People p
ON pjo.p_id = p.p_id
INNER JOIN Jobs j
on pjo.j_id = j.j_id
INNER JOIN Offices o
ON pjo.o_id = o.o_id
WHERE
j.j_name = 'SYSADMIN'
AND ( o.o_location = 'MADRID'
OR o.o_location = 'COPENHAGEN' )
This approach is called Database Normalization, and generally makes the best use of indexes in any RDBMS, as well as keeping the number of rows in your tables to a minimum by avoiding the duplication of data.
Edit²
I've re-fiddled your SQLfiddle. Much of this would have to be done porgrammatically, ie. determining which categories/how many joins/table aliases/column names, but you'd want to start out with whatever table the ID column references so you have a solid starting point for these joins, I've just created the users table for illustration.
CREATE TABLE users (`ID` varchar(16),`name` varchar(16));
INSERT INTO users (`ID`, `name`)
VALUES ('A', 'andrew'), ('B', 'bob'), ('C', 'charla');
And the query:
SELECT u.*,
gen.value 'gender',
pos.value 'position',
loc.value 'location'
FROM users u
LEFT JOIN yourtable gen
ON u.ID = gen.ID
AND gen.category = 'GENDER'
LEFT JOIN yourtable pos
ON u.id = pos.id
AND pos.category = 'POSITION'
LEFT JOIN yourtable loc
ON u.id = loc.id
AND loc.category = 'LOCATION'
And the output:
ID NAME GENDER POSITION LOCATION
A andrew MALE (null) COPENHAGEN
A andrew MALE (null) MADRID
B bob MALE WEBDESIGN MADRID
B bob MALE SYSADMIN MADRID
C charla FEMALE (null) MADRID
C charla FEMALE (null) AARHUS
I would also make certain that, with the potential for many joins, you have indexes on category as well as ID, if not all 3 columns in the table.

Answer 2

The solution - for the moment - is to make a subquery for each category. Like this:
SELECT DISTINCT ID
FROM yourtable
WHERE ID IN
(SELECT ID from yourtable WHERE (Category = 'POSITION' AND Value = 'SYSADMIN'))
AND ID IN
(SELECT ID from yourtable WHERE (Category = 'LOCATION' AND Value = 'MADRID') OR (Category = 'LOCATION' AND Value = 'COPENHAGEN'))
Working example can be seen here at SQL Fiddle.
It is working, but I am worried about performance when I have many rows or am checking in many categories.