Question

I have this database structure:

TBL_A  |  TBL_B  |  TBL_C  |  TBL_D  | TBL_E -------+---------+---------+---------+---------- id     | id_tbla | id_tbla | id_tbla | id name   | id_user | id_user | id_user | name_tbla ...    | is_bool |         | weight  | id_user 

Here is what I’m trying to achieve :

SELECT      a.id,      a.name,      b.is_bool,     count(c.id_user) AS nb_views,      sum(d.weight) AS total_weight,     count(distinct e.id_user) AS distinct_users, FROM TBL_A AS a  LEFT JOIN (TBL_B AS b) on (b.id_tbla = a.id) LEFT JOIN (TBL_C AS c) on (c.id_tbla = a.id) LEFT JOIN (TBL_D AS d) on (d.id_tbla = a.id) LEFT JOIN (TBL_E AS e) on (e.name_tbla = a.name) where a.id = 1 and e.id_user = 1 

The query is performed but the results (nb_views, total_weight, distinct_users) are wrong. Any idea why?

Answer 1

You’re trying to compute too many aggregates in one query.

Enita non sunt multiplicanda praeter necessitatem

(Latin, ‘entities are not to be multiplied beyond necessity’)

Your tables B, C, D, and E are produced Cartesian Products against each other. Suppose the given row in A matches:

• 3 rows in B
• 6 rows in C
• 4 rows in D
• 1 row in E

The total number of rows in the result is 3 * 6 * 4 * 1 = 72 rows. So your count(c.id_user) is 12 times what it should be, your sum(d.weight) is 18 times what it should be, etc.

The simplest remedy is to compute each of these aggregates in a separate query:

SELECT a.id, a.name, COALESCE(b.is_bool, FALSE) AS is_bool FROM TBL_A AS a LEFT JOIN TBL_B AS b ON (b.id_tbla = a.id) WHERE a.id = 1;  SELECT a.id, COUNT(c.id_user) AS nb_views FROM TBL_A AS a LEFT JOIN TBL_C AS c ON (c.id_tbla = a.id) WHERE a.id = 1;  SELECT a.id, SUM(d.weight) AS total_weight, FROM TBL_A AS a LEFT JOIN TBL_D AS d ON (d.id_tbla = a.id) WHERE a.id = 1;  SELECT a.id, COUNT(DISTINCT e.id_user) AS distinct_users, FROM TBL_A AS a LEFT JOIN TBL_E AS e    ON (e.name_tbla = a.name AND e.id_user = 1) WHERE a.id = 1;