MYSQL Counting matching results across multiple tables - mysql

Question

I have the following tables
Business
+-------------+--------------+------+-----+---------+----------------+
| Field | Type | Null | Key | Default | Extra |
+-------------+--------------+------+-----+---------+----------------+
| b_id | bigint(20) | NO | PRI | NULL | auto_increment |
| b_name | varchar(255) | NO | | NULL | |
+-------------+--------------+------+-----+---------+----------------+
Locations
+-------------+--------------+------+-----+---------+----------------+
| Field | Type | Null | Key | Default | Extra |
+-------------+--------------+------+-----+---------+----------------+
| l_id | bigint(20) | NO | PRI | NULL | auto_increment |
| l_name | varchar(255) | NO | | NULL | |
| b_id | big(20) | NO | | NULL | |
+-------------+--------------+------+-----+---------+----------------+
Jobs
+-------------+--------------+------+-----+---------+----------------+
| Field | Type | Null | Key | Default | Extra |
+-------------+--------------+------+-----+---------+----------------+
| j_id | bigint(20) | NO | PRI | NULL | auto_increment |
| j_name | varchar(255) | NO | | NULL | |
| b_id | bigint(20) | NO | | NULL | |
| l_id | bigint(20) | NO | | NULL | |
+-------------+--------------+------+-----+---------+----------------+
People
+-------------+---------------+------+-----+---------+----------------+
| Field | Type | Null | Key | Default | Extra |
+-------------+---------------+------+-----+---------+----------------+
| u_id | bigint(20) | NO | PRI | NULL | auto_increment |
| salutation | varchar(10) | NO | | NULL | |
| first_name | varchar(25) | NO | | NULL | |
| last_name | varchar(25) | NO | | NULL | |
+-------------+---------------+------+-----+---------+----------------+
People's Jobs
+-------------+------------+------+-----+---------+----------------+
| Field | Type | Null | Key | Default | Extra |
+-------------+------------+------+-----+---------+----------------+
| pj_id | bigint(20) | NO | PRI | NULL | auto_increment |
| u_id | bigint(20) | NO | | NULL | |
| j_id | bigint(20) | NO | | NULL | |
| l_id | bigint(20) | NO | MUL | NULL | |
+-------------+------------+------+-----+---------+----------------+
I need to produce a table that shows
+----------+-------------------------+------------+------------+------------+
| b_id | b_name | Locations | Jobs | People |
+----------+-------------------------+------------+------------+------------+
| 21 | Widgets Inc | 0 | x | 0 |
| 24 | Prince Privates | 0 | 0 | 0 |
| 23 | Halon plc | x | 0 | 0 |
| 18 | Stinky Hotels | x | x | x |
| 20 | Pylon Catering Corps | x | x | x |
| 22 | Skytrain Biscuits | 0 | 0 | 0 |
+----------+-------------------------+------------+------------+------------+
I can achieve a correct count of matching locations for each business with:
SELECT b.b_id,
b.b_name,
count(l.l_id) AS locations
FROM business AS b
LEFT JOIN locations AS l ON b.b_id=l.b_id
GROUP BY b.b_id
ORDER BY b_name
If I extend it to include a count of the jobs at each business and then the count of people at each business it all goes pear shaped.
I know that the following is inherently wrong with regards to getting the count of people (as people can hold more than 1 job). I don't know if I need to use sub selects or COALESCE?
SELECT b.b_id,
b.b_name,
count(l.l_id) AS locations,
count(j.j_id) AS jobs,
count(p.u_id) AS people
FROM business AS b
LEFT JOIN locations AS l ON b.b_id=l.b_id
LEFT JOIN job AS j ON b.b_id=j.b_id
LEFT JOIN people_jobs AS p ON l.l_id=p.l_id
GROUP BY b.b_id
ORDER BY b_name

Answer 1

I think you can do a quick-and-dirty fix of your query by using count(distinct):
SELECT b.b_id, b.b_name,
count(distinct l.l_id) AS locations,
count(distinct j.j_id) AS jobs,
count(distinct p.u_id) AS people
FROM business b LEFT JOIN
locations l
ON b.b_id = l.b_id LEFT JOIN
job j
ON b.b_id = j.b_id LEFT JOIN
people_jobs p
ON l.l_id = p.l_id
GROUP BY b.b_id
ORDER BY b_name ;
It is also possible that the problem is simply that the join to people_jobs needs more conditions:
people_jobs p
ON l.l_id = p.l_id and j.j_id = p.j_id
And maybe a condition on u.
Your problem is that you are trying to do aggregation across multiple dimensions and getting a cartesian product for each business. An alternative that is sometimes necessary is to do the counts in subqueries.

Answer 2

This query should do what you need:
SELECT
b.b_id,
b.b_name,
(SELECT COALESCE(COUNT(l_id ),0) FROM locations WHERE b_id=b.b_id) AS locations,
(SELECT COALESCE(COUNT(j_id ),0) FROM jobs WHERE b_id=b.b_id) AS jobs,
(SELECT COALESCE(COUNT(DISTINCT u_id),0)
FROM jobs j
JOIN people_jobs pj ON pj.j_id=j.j_id
WHERE j.b_id=b.b_id
) AS people
FROM business as b
ORDER BY b_name
You don't need the GROUP BY if you use subSELECTs, as the outer query will return 1 row per b_id, no more.
If instead you do JOIN the 4 tables at the main query level, like you were doing, you have two difficulties:
number of rows increases (avoidable with GROUP BY)
a simple COUNT does not work properly (avoidable with COUNT(DISTINCT
...))
(as shown in Gordon's answer)

Answer 3

You can try This Query:-
SELECT b.b_id,b.b_name,count(l.l_id) AS locations,count(j.j_id) AS jobs,count(p.u_id) AS people
FROM business as b LEFT JOIN locations as l ON b.b_id=l.b_id
LEFT JOIN job as j ON b.b_id=j.b_id
LEFT JOIN people_jobs as p ON l.l_id=p.l_id
GROUP BY b.b_id, b.b_name
ORDER BY b_name
I hope this will work for you.