How SQL's DISTINCT clause works?

问题

I'm looking for the answer on how DISTINCT clause works in SQL (SQL Server 2008 if that makes a difference) on a query with multiple tables joined?

I mean how the SQL engine handles the query with DISTINCT clause?

The reason I'm asking is that I was told by my far more experienced colleague that SQL applies DISTINCT to every field of every table. It seems unlikely for me, but I want to make sure....

For example having two tables:

CREATE TABLE users
(
u_id INT PRIMARY KEY,
u_name VARCHAR(30),
u_password VARCHAR(30)
)

CREATE TABLE roles
(
r_id INT PRIMARY KEY,
r_name VARCHAR(30)
)

CREATE TABLE users_l_roles
(
u_id INT FOREIGN KEY REFERENCES users(u_id) ,
r_id INT FOREIGN KEY REFERENCES roles(r_id) 
)

And then having this query:

SELECT          u_name
FROM            users 
INNER JOIN      users_l_roles ON users.u_id = users_l_roles.u_id
INNER JOIN      roles ON users_l_roles.r_id = roles.r_id

Assuming there was user with two roles then the above query will return two records with the same user name.

But this query with distinct:

SELECT DISTINCT u_name
FROM            users 
INNER JOIN      users_l_roles ON users.u_id = users_l_roles.u_id
INNER JOIN      roles ON users_l_roles.r_id = roles.r_id

will return only one user name.

The question is whether SQL will compare all the fields from all the joined tables (u_id, u_name, u_password, r_id, r_name) or it will compare only named fields in the query (u_name) and distinct the results?

回答1:

DISTINCT filters out duplicate values of your returned fields.

A really simplified way to look at it is:

It builds your overall result set (including duplicates) based on your FROM and WHERE clauses
It sorts that result set based on the fields you want to return
It removes any duplicate values in those fields

It's semantically equivalent to a GROUP BY where all returned fields are in the GROUP BY clause.