Writing more better SQL - mysql

Question

I've got a query here that's painfully slow. Part of the problem may be that tableA in the sub-query has a quite substantial size in comparison to the other tables.
TABLES STRUCTURE
*-------------------*------------------*-------------------*
| ID_TABLE | DATA_TABLE | DATA_TABLE_EXT |
*-------------------*------------------*-------------------*
| id n<|>1 id 1<|>n owner_id |
| foreign_id | owner_id | information |
| foreign_id_source | date_field | ... |
| ... | ... | |
*-------------------*------------------*-------------------*
QUERY
SELECT ID_TABLE.foreign_id_source, count(ID_TABLE.id) as count
FROM DATA_TABLE
LEFT JOIN ID_TABLE ON DATA_TABLE.id = ID_TABLE.id
WHERE DATA_TABLE.owner_id = 'some_id'
AND DATA_TABLE.date_field > 'some_date'
AND DATA_TABLE.id IN (
SELECT DATA_TABLE_EXT.owner_id FROM DATA_TABLE_EXT
JOIN DATA_TABLE ON DATA_TABLE_EXT.owner_id = DATA_TABLE.id
WHERE DATA_TABLE.owner_id = 'some_id'
GROUP BY DATA_TABLE.id
HAVING SUM(ABS(DATA_TABLE_EXT.information)) <> 0
)
GROUP BY ID_TABLE.foreign_id_source
ORDER BY count ASC
REQUIRED RESULT
*-------------------*-------------*
| foreign_id_source | count |
*-------------------*-------------*
| source1 | 45 |
| source2 | 10 |
| ... | |
*-------------------*-------------*
Each id in DATA_TABLE may have multiple records in ID_TABLE.
many records in DATA_TABLE may have the same owner_id.
I'm looking for the number of records in data_table with a foreign_id_source, grouped by that foreign_id_source, where the record is after 'some_date' and it's DATA_TABLE_EXT records do not all have a value of 0 in the information field.
Short of creating indexes or other database manipulation is there a way to improve this query in terms of performance?
Any other suggestions are also welcome.

Answer 1

The point is: SUM(ABS(DATA_TABLE_EXT.information)) <> 0 can only be true if at least one DATA_TABLE_EXT.information is non-zero. So we don't have to sum() them, we only only need to check if a non-zero one exists.
[ I don't know if mysql is smart enough to handle the exists(), but in theory it is cheaper, and can be faster]
SELECT it.foreign_id_source, count(it.id) as count
FROM DATA_TABLE dt
LEFT JOIN ID_TABLE it ON dt.id = it.id
WHERE dt.owner_id = 'some_id'
AND dt.date_field > 'some_date'
AND EXISTS (
SELECT *
FROM DATA_TABLE_EXT x
JOIN DATA_TABLE dt2 ON x.owner_id = dt2.id
WHERE x.id =dt.id
AND dt2.owner_id = 'some_id'
AND x.information <> 0
)
GROUP BY it.foreign_id_source
ORDER BY count ASC
;

Answer 2

Often moving the subquery to the FROM will help:
SELECT ID_TABLE.foreign_id_source, count(DATA_TABLE.id) as count
FROM ID_TABLE LEFT JOIN
DATA_TABLE
ON DATA_TABLE.id = ID_TABLE.id JOIN
(SELECT DATA_TABLE.id
FROM DATA_TABLE_EXT JOIN
DATA_TABLE
ON DATA_TABLE_EXT.owner_id = DATA_TABLE.id
WHERE DATA_TABLE.owner_id = 'some_value'
GROUP BY DATA_TABLE.id
HAVING SUM(ABS(DATA_TABLE_EXT.information)) <> 0
) xx
ON DATA_TABLE.id = xx.id
WHERE DATA_TABLE.owner_id = 'some_value' AND
DATA_TABLE.date_field > 'some_date'
GROUP BY x.field1
ORDER BY count ASC;
Then, you can think about indexes. These would be tableX(field2, fieldZ, field1, fieldX), tableI(field1), tableX(field2, field1, fieldB), andtableA(field1)`.