Join Predicates And Comparison Operators - mysql

Question

In an attempt to better understand how SQL joins work, I've run into difficulty understanding the result set returned from a query -- The tables I'm using are:
Employees
Id Name Salary Gender City
1 Sam 2500 Male London
5 Todd 3100 Male Toronto
3 John 4500 Male New York
6 Jack 7000 Male Shangri La
4 Sara 5500 Female Tokyo
2 Pam 6500 Female Sydney
and Gender:
ID Gender
1 Male
2 Female
3 Unknown
The first query returns all the columns from an inner join on the Gender column -
SELECT *
FROM Employees
INNER JOIN Gender
ON Employees.Gender = Gender.Gender
The returned result -
Id Name Salary Gender City ID Gender
1 Sam 2500 Male London 1 Male
5 Todd 3100 Male Toronto 1 Male
3 John 4500 Male New York 1 Male
6 Jack 7000 Male Shangri La 1 Male
4 Sara 5500 Female Tokyo 2 Female
2 Pam 6500 Female Sydney 2 Female
Which is pretty much what I expected. However when I changed the comparison operator -
SELECT *
FROM Employees
INNER JOIN Gender
ON Employees.Gender != Gender.Gender
What I originally thought would return an empty set, returned this -
Id Name Salary Gender City ID Gender
4 Sara 5500 Female Tokyo 1 Male
2 Pam 6500 Female Sydney 1 Male
1 Sam 2500 Male London 2 Female
5 Todd 3100 Male Toronto 2 Female
3 John 4500 Male New York 2 Female
6 Jack 7000 Male Shangri La 2 Female
1 Sam 2500 Male London 3 Unknown
5 Todd 3100 Male Toronto 3 Unknown
3 John 4500 Male New York 3 Unknown
6 Jack 7000 Male Shangri La 3 Unknown
4 Sara 5500 Female Tokyo 3 Unknown
2 Pam 6500 Female Sydney 3 Unknown
While I can kinda see how the not-equals(!=) operator would return this result it begs the question of what type of comparisons are useful in join predicates and which aren't - does the type of join [inner, right, left...] impact the returned result adversely or can the join type and comparison be resolved to actionable behavior(in other words does it always have to be ==)? Also if there are any sources out there that could help me, that would be great. Thanks.

Answer 1

An inner join is a subset of the Cartesian product of both tables. That is, every row in one table is paired with every row of the other table.
So, if one table has n rows and the other m rows, then the Cartesian product has n * m rows.
The on clause filters these rows. Equality is the most common filter and such joins are often called equi-joins. They are offer the most optimization opportunities and are typically the most efficient.
(Outer joins are similar but have a mechanism to include unmatched rows.)
Normally, join predicates contain at least one equality comparison (although this is not necessary). Other comparisons -- including subqueries using exists/in -- are allowed and often useful.
Any decent documentation or tutorial or book should be able to explain this.
In your case, often not exists is the intention. To find employees whose gender is not in the reference table:
SELECT e.*
FROM Employees e
WHERE NOT EXISTS (SELECT 1 FROM Gender g WHERE e.Gender = g.Gender)
Of course, such a query would be unnecessary if you used the primary key to reference the table and included proper foreign key declarations.