Mysql group_concat(id) as ids in a left join and using ids to select all columns in id group

Question

i have a table that contains some articles with it's own ID and shared SKU key. I've tried to make the query with a left join and using group result to take all ids returned from the query.

My data structure is like that:

id -    name   - sku - ...
1  - felix     - cat
2  - tom       - cat - ...
3  - sylvester - cat - ...
4  - red       - pen - ...
5  - blue      - pen - ...

I tried to use this query:

SELECT * FROM `test` 
[LEFT/RIGHT/INNER] JOIN 
(
    SELECT GROUP_CONCAT(DISTINCT id) AS idsgroup FROM `test` WHERE (attribute_name = 'sku') GROUP BY value_name LIMIT 0, 3 
) bind
ON id IN (bind.idsgroup);

this query is wrong, it return only 1 id per group instead all ids selected from concat or in LEFT JOIN case, obviously all rows.

Any suggestion workaround to achieve the right result?

EDIT:

here a fiddle with the structure: http://sqlfiddle.com/#!9/b6747a

And the query i tried into:

   SELECT * FROM `view_test` 
    INNER JOIN 
    (
        SELECT GROUP_CONCAT(DISTINCT entity_id) AS idsgroup FROM `view_test` WHERE (attribute_name = 'sku') GROUP BY value_name LIMIT 0, 3 
    ) bind
    ON entity_id IN (bind.idsgroup);

As this pic show, my result lost some ids, part of the group.

EDIT 2: after i used FIND_IN_SET() suggested by Kickstart the result is the expected:

SELECT * FROM `view_test` 
INNER JOIN 
(
    SELECT GROUP_CONCAT(DISTINCT entity_id) AS idsgroup FROM `view_test` WHERE (attribute_name = 'sku') GROUP BY value_name LIMIT 0, 3 
) bind
ON FIND_IN_SET(entity_id, bind.idsgroup);

Answer 1

The simple fix would appear to be to use FIND_IN_SET for the join. But this is a bit of a hack and will not be that quick.

SELECT * 
FROM `view_test` 
INNER JOIN 
(
    SELECT GROUP_CONCAT(DISTINCT entity_id) AS idsgroup 
    FROM `view_test` 
    WHERE (attribute_name = 'sku') 
    GROUP BY value_name 
    LIMIT 0, 3 
) bind
ON FIND_IN_SET(entity_id, bind.idsgroup);

Further not sure why you have a LIMIT on the sub query, especially without an order clause.

Possibly better to use a sub query to just get the DISTINCT entity_id with an attribute_name of sku and join against that.

SELECT * 
FROM `view_test` 
INNER JOIN 
(
    SELECT DISTINCT entity_id 
    FROM `view_test` 
    WHERE (attribute_name = 'sku') 
) bind
ON view_test.entity_id = bind.entity_id

Answer 2

Something like this?

SELECT t.*, group_concat(t2.id) FROM `test` t
LEFT JOIN test t2 ON t2.attribute_name = 'sku' and t.id != t2.id
group by t.id;

row, and the list of all ids that have same SKU