Question

I have a table called “users” which has info(first name,etc…) with the following rows:

username
----------
mikha
guy
maricela

Another table called “questions” with the following rows:

asker
----------
mikha
mikha
maricela
guy
maricela

Another table called connections with the following rows:

username1
----------
guy
mikha
mikha

I want to select info about the user ‘mikha’ including count when he is questions.asker and when he is connections.username1. I use the following query:

SELECT COUNT(questions.asker) AS asking,COUNT(connections.username1) AS following 
FROM users LEFT JOIN questions ON users.username = questions.asker 
LEFT JOIN connections ON users.username = connections.username1 
WHERE users.username = 'mikha'

Expected results:

asking: 2 (as mikha is found 2 times asking)
following:  2 (as mikha is found 2 times following)

Actual results:

asking: 4
following: 4

If I use COUNT(DISTINCT questions.asker) and COUNT(DISTINCT connections.username1) I get the result always as 1 as it counts the name only once. I tried GROUP BY also with no avail.

So, how can I use distinct to avoid duplicates and the same time count the same name all the times available not just one?

I created this fiddle to test the problem yourself.

Thanks in advance
Regards
Michael

Answer 1

Do you have an ID in questions table, and one ID in connections table? If you have, you could use this:

select
  users.username,
  count(distinct questions.id) as asking,
  count(distinct connections.id) as following
from
  users left join questions on users.username = questions.asker
  left join connections on users.username = connections.username1
where
  username='mikha'

EDIT: based on your comments, i think that what you need is this:

select
  COUNT(distinct questions.id) AS asking,
  COUNT(DISTINCT CONCAT(c1.username1,c1.username2)) AS following,
  COUNT(DISTINCT CONCAT(c2.username1,c2.username2)) AS followers
FROM
  users LEFT JOIN questions ON questions.asker=users.username
  LEFT JOIN connections c1 ON c1.username1=users.username
  LEFT JOIN connections c2 ON c2.username2=users.username
WHERE
  users.username = 'mikha'

the only problem here is that counting by CONCAT(...) does not take advantage of indexes and can be slow. And also it could happen that CONCAT(userA, userB) = CONCAT(userC, userD) even if userA<>userC and userB<>userD. You could use CONCAT(username1, ':', username2) to avoid this, but only if you make sure that no username contains : character. But if it is possible, i would suggest you to add an id also to your connections table.