MYSQL - 选择两个表之间的唯一公共列 - 最有效的查询

Question

I have two tables:

db_contacts

Phone | Name | Last_Name
--------------------
111   | Foo  | Foo
222   | Bar  | Bar
333   | John | Smith
444   | Tomy | Smith

users_contacts

User_ID | Phone
--------------------
1       | 123
1       | 111
2       | 222
2       | 333
3       | 111
3       | 333
4       | 444

Notice from above that:

• User with ID 2 is the only one that have the phone number 222
• User with ID 4 is the only one that have the phone number 444

I need to obtain these results with a MySQL query.

In other words: How can I select all the users that have a unique phone number in condition that this number exists in the db_contacts.

I need my end result to be something like that:

User_ID | Phone | Name | Last_Name
------------------------------------
2       | 222   | Bar  | Bar
4       | 444   | Tomy | Smith

PS: There is no Foreign key between the Phone columns, as a User can have a phone that is not in the db_contacts.

In real life, db_contacts contains about 1 million records and users_contacts about 5 million records.

What I tried and failed and taking a lot of time to execute:

SELECT * 
FROM users_contacts 
WHERE users_contacts.phone IN (
    SELECT users_contacts.phone 
    FROM `users_contacts`
    JOIN db_contacts ON db_contacts.phone = users_contacts.phone
    GROUP BY users_contacts.phone
    HAVING COUNT(users_contacts.phone) = 1
)

Update:

Thank you for your replies, I have provided my solution that fits my case perfectly.

Answer 1

I think you want:

select uc.*
from user_contacts uc
where not exists (select 1
                  from user_contacts uc2
                  where uc2.phone = uc.phone and uc2.user_id <> uc.user_id
                 );

For performance, you want an index on user_contacts(phone, user_id) .

Another method is:

select max(user_id) as user_id, phone
from user_contacts
group by phone
having count(*) = 1;

The not exists version is probably going to be faster.

Answer 2

I would use a simple JOIN with a NOT EXISTS condition. This is usually the most efficient way to check that something has no duplicates ; compared to your solution, this has the advantage of avoiding aggregation.

SELECT uc.User_ID, dc.*
FROM users_contacts uc
INNER JOIN db_contacts dc ON uc.Phone = dc.Phone
WHERE NOT EXISTS (
    SELECT 1 
    FROM users_contacts uc1 
    WHERE uc1.Phone = dc.Phone AND uc1.User_ID != uc2.User_ID
)

Hint: consider setting the following indexes:

• users_contacts(Phone, User_ID)
• db_contacts(Phone)

Answer 3

I first would like to thank everyone that posted solutions, they all worked.

But I was a bit crucial on response times, and solutions provided by the fellows took a lot of time to execute, couple of seconds.

In case anyone was having a similar problem, I ended up by creating a new table calling it users_unique_contacts, and created a trigger AFTER INSERT on users_contacts that checks if the newly created contact existed in the users_unique_contacts, if it didn't exist, add it, else remove it as it means the number is not unique anymore.

My Trigger went like this:

BEGIN
    IF EXISTS (SELECT 1 = 1 FROM users_unique_contacts WHERE phone = new.phone LIMIT 1) THEN
        BEGIN
                DELETE FROM users_unique_contacts WHERE phone = new.phone LIMIT 1;
        END;
    ELSE
        BEGIN
                INSERT INTO users_unique_contacts (user_id,phone) VALUES (new.user_id, new.phone);
        END;
    END IF;
END

Now everytime I want the unique numbers of a user, I query the users_unique_contacts and execution time is milliseconds.