Select count field of rows that are related to another table - sql

Question

I'm struggling with this issue from a long time and don't know how to solve it. It's hard for me to describe, so please be patient. There are two tables:
Table "Users"
UserId PK
Gender
Table "Forms"
FormId PK
UserId1 FK
UserId2 FK
Type
Forms are always related to two users, but not all users have related forms. Now I want to count specified gender only of those users, who have related forms.
So as a result, I want to have sth. like this:
# | Gender | GenderCount
1 | male | 43
2 | female | 12
3 | trans | 2
I tried the following SQL-Script but the result isn't distinct (the sum of all GenderCount is greater then the actual number of users)
SELECT u.Gender AS 'Gender', COUNT(u.Gender) AS 'GenderCount'
FROM Users u, Forms f
WHERE ((f.UserId1 = u.UserId)
OR (f.UserId2 = u.UserId))
AND (Type = 'Foo')
GROUP BY Gender
ORDER BY GenderCount
DESC
Any tips to solve this?

Answer 1

Let's take a look at what you want:
How many of each gender answered any form?
Note: each user should only be counted once, no matter how many forms they've filled out.
Phrased like this, the answer becomes fairly obvious, at least in pseudo-code:
SELECT
u.Gender,
COUNT(u.Gender)
FROM
Users u
WHERE
[User has answered a form]
GROUP BY
u.Gender
The easiest way to determine if a user has answered a form depends on the specific flavour of SQL being used. You'll need to use a subquery. There are a couple of options for how to access it.
IN is the most common method:
SELECT
u.Gender Gender,
COUNT(u.Gender) GenderCount
FROM
Users u
WHERE
u.id IN (
SELECT f.UserId1 user_id FROM Forms f WHERE Type = 'Foo'
UNION
SELECT f.UserId2 user_id FROM Forms f WHERE Type = 'Foo'
)
GROUP BY
Gender
ORDER BY
GenderCount DESC
Where available, EXISTS is more natural to read, and is sometimes faster:
SELECT
u.Gender Gender,
COUNT(u.Gender) GenderCount
FROM
Users u
WHERE
EXISTS(
SELECT '1'
FROM Forms f
WHERE
(f.UserId1 = u.id OR f.UserId2 = u.id)
AND Type = 'Foo'
)
GROUP BY
Gender
ORDER BY
GenderCount DESC
Regarding speed: The query optimiser will often convert IN to EXISTS where possible, to avoid selecting extra rows unnecessarily. However, the use of multiple columns necessitates either an OR or a UNION, so it may be pretty even in this case. ie: neither OR nor UNION play nicely with indexes.

Answer 2

SELECT u1.Gender AS 'Gender', COUNT(*) AS 'GenderCount'
FROM
Users u1
INNER JOIN
(SELECT DISTINCT u.UserId
FROM
Users u
INNER JOIN Forms f ON ((f.UserId1 = u.UserId)
OR (f.UserId2 = u.UserId))
AND (f.Type = 'Foo')) T ON T.UserId = u1.UserId
GROUP BY Gender
ORDER BY GenderCount DESC

Answer 3

Skip the join which is generating multiple rows per user:
SELECT Gender, COUNT(Gender) AS 'GenderCount'
FROM Users
WHERE UserId IN (SELECT UserId1 FROM Forms WHERE Type = 'Foo'
UNION
SELECT UserId2 FROM Forms WHERE Type = 'Foo')
GROUP BY Gender
ORDER BY GenderCount DESC
Or if you prefer to avoid a UNION (which is perfectly valid in this scenario BTW) you can use OR like this:
SELECT Gender, COUNT(Gender) AS 'GenderCount'
FROM Users
WHERE UserId IN (SELECT UserId1 FROM Forms WHERE Type = 'Foo')
OR UserId IN (SELECT UserId2 FROM Forms WHERE Type = 'Foo')
GROUP BY Gender
ORDER BY GenderCount DESC
As others have pointed out, there are ways to do this using a JOIN as well. However, a JOIN adds needless complexity for the DBMS engine as it will first need to match up the rows, and then reduce to DISTINCT values.

Answer 4

You should use
count(distinct u.UserId)
that way users only get counted once: count(distinct field_name) counts the number of unique values contained in field_name, so counting distinct on the primary key gives you the number of unique users, which is what you're looking for.
Also, instead of joining, you probably would be better off using an in clause like this
select Gender, count(distinct UserId) as GenderCount
from Users
where u.UserId in (select UserId1 from Forms) or u.UserId in (select UserId2 from Forms)
It's probably also going to be slightly faster.