Sorting over two columns in entity-attribute-value schema

Question

We are using a entity-attribute-value schema to store objects in our project. It's basically like this:

entity- dummy storage of all entries prototypes (car, house etc.)

entry (representation of single entity object)

• id
• entity_id
• ...

parameter (all entity attributes color, name, brand etc.)

• id
• name
• …

value (single value for one entry and parameter)

• id
• entry_id (foreign key to entry.id)
• parameter_id (foreign key to parameter.id)
• value (actual value)

These tables represent dynamic tables stored in MySQL database.

The only problem with this scheme is sorting over 2 (or more) parameters. For example sort all cars by manufacturing_year DESC and brand ASC.

The final result we need is the sorted list of entry ids.

For these objects:

entity

id | name
------------------------------
1  | Car
------------------------------

entry

id| entity_id
-------------
1 | 1
-------------
2 | 1
-------------
3 | 1
-------------

parameter

id| entity_id | name
-----------------------
1 | 1         | Brand
-----------------------
2 | 1         | Year 
-----------------------

value

id | entry_id | parameter_id | value 
----------------------------------------------
1  | 1        | 1            | Tatra
----------------------------------------------
2  | 1        | 2            | 2005
----------------------------------------------
3  | 2        | 1            | Aston Martin
----------------------------------------------
4  | 2        | 2            | 1999
----------------------------------------------
5  | 3        | 1            | Man
----------------------------------------------
6  | 3        | 2            | 2005
----------------------------------------------

The correct order of these entries by given criteria shoud be 3, 1, 2.

Answer 1

You can join parameter table as many times as you have parameters to sort on, with parameter id's as join conditions, so you get a simple rowset with values as columns for sorting.

SELECT entry.id as entry_id, 
       v_brand.value as brand,
       v_year.value as year 
FROM entity
  JOIN entry 
    ON entity.id = entry.entity_id
  JOIN value v_brand 
    ON v_brand.parameter_id = 1 
    AND v_brand.entry_id = entry.id
  JOIN value v_year 
    ON v_year.parameter_id = 2 
    AND v_year.entry_id = entry.id
WHERE entity.name = 'Car'
ORDER BY year DESC, brand ASC

Answer 2

So, to recap, you have table something like this...

DROP TABLE IF EXISTS eav_hell;

CREATE TABLE eav_hell
(entity INT NOT NULL
,attribute VARCHAR(12) NOT NULL
,value VARCHAR(12)
);

INSERT INTO eav_hell VALUES
(1,'Brand','Tatra'),
(1,'Year','2005'),
(2,'Brand','Aston Martin'),
(2,'Year','1999'),
(3,'Brand','Man'),
(3,'Year','2005');

...from which you can obtain a result something like this...

SELECT entity
     , MAX(CASE WHEN attribute = 'Brand' THEN value END) Brand
     , MAX(CASE WHEN attribute = 'Year' THEN value END) Year
  FROM eav_hell
 GROUP
    BY entity;   

+--------+--------------+------+
| entity | Brand        | Year |
+--------+--------------+------+
|      1 | Tatra        | 2005 |
|      2 | Aston Martin | 1999 |
|      3 | Man          | 2005 |
+--------+--------------+------+

...so what was the problem again?

(and pray pity the poor data types)