在一对多关系查询中使用LIMIT

Question

I have the two tables tabA and tabB , and there is a one-to-many relationship from tabA to tabB . I have the query:

SELECT * FROM `tabA` LEFT JOIN `tabB` ON `tabA`.`aID` = `tabB`.`aID`

and the rows that are returned is a large set with multiple duplicates from tabA for each tabB reference to tabA .

I am aware that I can use GROUP BY to limit the tabA rows to unique elements, unless I use custom field(s) using the GROUP_CONCAT function, combined with two REPLACE functions for escaping (which seriously impacts performance), I loose all bar one of the rows contained in tabB . An example query looks like:

   SELECT `tabA`.*,
          GROUP_CONCAT(REPLACE(REPLACE(`tabB`.`tabBCol1`, '/', '//'), ',', '/,')) AS `tabBCol`,
          GROUP_CONCAT(REPLACE(REPLACE(`tabB`.`tabBCol2`, '/', '//'), ',', '/,')) AS `tabBCo2` 
     FROM `tabA` 
LEFT JOIN `tabB` ON `tabA`.`aID` = `tabB`.`aID` 
 GROUP BY `tabA`.`aID`

That query will allow me to use the LIMIT syntax so I can (for example) only show 5 entries, starting after 5 (ie LIMIT 5,5 ). And when I apply that to the former query, then I won't get the next 5 queries, but a random set of data based on the numbers of associations.

So, apart from the second query, is there any way that I can fetch the rows, with there associations, but allow the use of the LIMIT syntax, and without the performance hit of excessive REPLACE functions?

ADDITIONAL

Although I can use multiple subqueries for each row, using the first query with GROUP BY syntax (which would allow me to apply any WHERE conditions for the associations), I am trying to find a way to avoid the N+1 Selects Problem (although in this example, my LIMIT syntax is LIMIT 5,5 , I will be applying this to much larger LIMIT s (upto 1000 rows at a time)).

Answer 1

Try two queries:

// get those 5 records
SELECT * FROM Cars   WHERE some_conditon = blabla LIMIT 5;

// get all associated records from related table
SELECT * FROM Wheels WHERE car_id IN (1, 3, 5, 123, 16);

In the result there will not be any N problem as you will always have two queries. Even if you will have 1000 records in 1st query it will always be better to use this simple method, than joins/groups by/concats/etc.