Prevent duplicate values in LEFT JOIN

Question

I faced a situation where I got duplicate values from LEFT JOIN. I think this might be a desired behavior but unlike from what I want.

I have three tabl

Answer 1

Although the tables are obviously simplified for discussion, it appears they are structurally flawed. Tables should be structured to show relationships between entities, rather than be merely lists of entities and/or attributes. And I would consider a phone number to be an attribute (of a person or department entity) in this case.

The first step would be to create tables with relationships, each having a primary key and possibly a foreign key. In this example, it would be helpful to have the person table use person_id for the primary key, and the department table to use department_id for its primary key. Next look for one-to-many or many-to-many relationships, and set your foreign keys accordingly:

• If one person can only be in one department at a time, then you have a one(department)-to-many(persons). No foreign key in the department table, but department_id will be a foreign key in the persons table.
• If one person can be in more than one department, they you have a many-to-many, and you'll need an additional junction table with person_id and department_id as foreign keys.

To summarize, there should only be two tables in your scenario: one table for the person and the other table for the department. Even allowing for personal phone numbers (a column in the persons table) and department numbers in the department table, this would be a better approach.

The only caveat is when one department has many numbers (or more than one department shares a single phone number), but this would be beyond the scope of the original question.

Answer 2

Use this type of query: SQL Server
(You can change id of ORDER BY id to each column that you want it)

SELECT 
    p.id, 
    p.person_name, 
    d.department_name, 
    c.phone_number
FROM
    person p
    LEFT JOIN 
    (SELECT *, ROW_NUMBER() OVER (PARTITION BY person_id ORDER BY id) AS seq
     FROM department) d 
    ON d.person_id = p.id And d.seq = 1
    LEFT JOIN 
    ( SELECT *, ROW_NUMBER() OVER (PARTITION BY person_id ORDER BY id) AS seq
     FROM contact) c 
    ON c.person_id = p.id And c.seq = 1;

Answer 3

I like to call this problem "cross join by proxy". Since there is no information (WHERE or JOIN condition) how the tables department and contact are supposed to match up, they are cross-joined via the proxy table person - giving you the Cartesian product. Very similar to this one:

• Two SQL LEFT JOINS produce incorrect result

More explanation there.

Solution for your query:

SELECT p.id, p.person_name, d.department_name, c.phone_number
FROM   person p
LEFT   JOIN (
   SELECT person_id, min(department_name) AS department_name
   FROM   department
   GROUP  BY person_id
   ) d ON d.person_id = p.id
LEFT   JOIN (
   SELECT person_id, min(phone_number) AS phone_number
   FROM   contact
   GROUP  BY person_id
   ) c ON c.person_id = p.id;

You did not define which department or phone number to pick, so I arbitrarily chose the minimum. You can have it any other way ...

Answer 4

SELECT p.id, p.person_name, d.department_name, c.phone_number 
FROM person p
  LEFT JOIN department d 
    ON p.id = d.person_id
  LEFT JOIN contact c 
    ON p.id = c.person_id 
group by p.id, p.person_name, d.department_name, c.phone_number

Answer 5

I think you just need to get lists of departments and phones for particular person. So just use array_agg (or string_agg or json_agg):

SELECT
    p.id,
    p.person_name,
    array_agg(d.department_name) as "department_names",
    array_agg(c.phone_number) as "phone_numbers"
FROM person AS p
LEFT JOIN department AS d ON p.id = d.person_id
LEFT JOIN contact AS c on p.id = c.person_id
GROUP BY p.id, p.person_name