Same Join Constraints on Multiple Columns

Question

We have a situation that I have been able to recreate with the following simple example. I have the following two sample tables:

CREATE TABLE contact_info
(
  id INT UNSIGNED AUTO_INCREMENT,
  priContactId INT,
  secContactId INT,
  blahBlah VARCHAR(32),

  PRIMARY KEY(id)  
);  

and

CREATE TABLE name_lookup
(
  id INT UNSIGNED AUTO_INCREMENT,
  contactID INT,
  contactName VARCHAR(32),

  PRIMARY KEY(id)  
);  

I populate them as follows:

INSERT INTO contact_info(priContactId, secContactId, blahBlah) VALUES(1, 3, "Team A"), (4, 2, "Team B");
INSERT INTO name_lookup(contactID, contactName) VALUES(1, "John Doe"), (2, "Mary Smith"), (3, "Jose Garcia"), (4, "Larry Brown");

Obviously, the contents of the tables are as follows:

+----+--------------+--------------+----------+
| id | priContactId | secContactId | blahBlah |
+----+--------------+--------------+----------+
|  1 |            1 |            3 | Team A   |
|  2 |            4 |            2 | Team B   |
+----+--------------+--------------+----------+

+----+-----------+-------------+
| id | contactID | contactName |
+----+-----------+-------------+
|  1 |         1 | John Doe    |
|  2 |         2 | Mary Smith  |
|  3 |         3 | Jose Garcia |
|  4 |         4 | Larry Brown |
+----+-----------+-------------+

We would like to perform a JOIN operation so that we get output like this:

+-------------+-------------+--------+
| John Doe    | Jose Garcia | Team A |
+-------------+-------------+--------+
| Larry Brown | Mary Smith  | Team B |
+-------------+-------------+--------+

The join constraints for both the priContactId and secContactId columns are the same and I am having tough time figuring out what the JOIN query should look like.

FYI, we are using MySQL version 5.6.49.

Answer 1

This is an interesting example of where an (indirect) SELF-JOINs prove useful!

In order to answer your problem, I did the following (all the SQL below is available on the fiddle here):

I used your DDL and DML as supplied in the question - thanks for that (and a +1 - you've only asked two questions, so I'd consider you a New Contributor and it's great to see that some people take the trouble to supply DDL and DML - if only all OP's did the same!).

CREATE TABLE name_lookup
(
  id INT UNSIGNED AUTO_INCREMENT,
  contact_id INT,
  contact_name VARCHAR(32),

  PRIMARY KEY(id)  
);



CREATE TABLE contact_info
(
  id INT UNSIGNED AUTO_INCREMENT,
  pri_contact_id INT,
  sec_contact_id INT,
  blah_blah VARCHAR(32),

  PRIMARY KEY(id)  
);

Populate them:

INSERT INTO name_lookup(contact_id, contact_name) 
VALUES
(1, "John Doe"), (2, "Mary Smith"), 
(3, "Jose Garcia"), (4, "Larry Brown");

and

INSERT INTO contact_info(pri_contact_id, sec_contact_id, blah_blah) 
VALUES(1, 3, "Team A"), (4, 2, "Team B"), (1, NULL, "Team A");

Notice the final record which has sec_contact_id = NULL - see the discussion in the EDIT below. I'm implicitly assuming that the team identity is determined by the pri_contact_id - define as necessary.

You'll also notice that I use snake_case or lower_case_with_underscores - personal preference - pick a style and stick to it!.

I've left extra fields in the SQL in the fiddle so that you can see the thought processes involved and how I arrived at a solution!

SELECT 
  nl1.id, nl1.contact_id, nl1.contact_name, 
  ci1.pri_contact_id, ci1.sec_contact_id, ci1.blah_blah
FROM name_lookup nl1
JOIN contact_info ci1 
  ON nl1.contact_id = ci1.pri_contact_id;

Result:

id  contact_id  contact_name    pri_contact_id  sec_contact_id  blah_blah
 1           1      John Doe                 1               3     Team A
 4           4   Larry Brown                 4               2     Team B

So, now we join name_lookup back on itself using contact_info as the link between them.

SELECT 
  nl1.id, nl1.contact_id, nl1.contact_name,
  nl2.id, nl2.contact_id, nl2.contact_name,
  ci1.pri_contact_id, ci1.sec_contact_id, ci1.blah_blah
FROM name_lookup nl1
JOIN contact_info ci1 
  ON nl1.contact_id = ci1.pri_contact_id
JOIN name_lookup nl2
  ON ci1.sec_contact_id = nl2.contact_id
ORDER BY nl1.id;

Result:

id  contact_id  contact_name    id  contact_id  contact_name    pri_contact_id  sec_contact_id  blah_blah
 1           1      John Doe     3           3   Jose Garcia                 1   3    Team A
 4           4   Larry Brown     2           2    Mary Smith                 4 2      Team B

So, having obtained our result, we can now clean up the SQL (only SELECT required fields - reduce any network traffic as well as I/O on the server) as follows:

SELECT 
  nl1.contact_name AS "Con_1 name",
  nl2.contact_name AS "Con_2 name",
  ci1.blah_blah AS "Team"
FROM name_lookup nl1
JOIN contact_info ci1 
  ON nl1.contact_id = ci1.pri_contact_id
JOIN name_lookup nl2
  ON ci1.sec_contact_id = nl2.contact_id
ORDER BY nl1.id;

Result:

Con_1 name     Con_2 name     Team
  John Doe     Jose Garcia  Team A
Larry Brown    Mary Smith   Team B

Et voilà - result as desired!

EDIT (NULLs in sec_contact_id):

It was pointed out to me that my answer wasn't as comprehensive as I might have liked it to be. What if sec_contact_id is NULL which, after all, is possible - you might have made the first one but a follow-up hasn't been completed yet?

So, I changed the table slightly and it now contains (as you'll see above - a considerably changed fiddle is available here - I wanted to run it on PostgreSQL also):

 Con_1 name   Con_2 name      Team
   John Doe  Jose Garcia    Team A
   John Doe         NULL    Team B
Larry Brown   Mary Smith    Team B

so, now, you have to use INNER JOINs thus:

  nl1.contact_name AS "Con_1 name",
  nl2.contact_name AS "Con_2 name",
  ci1.blah_blah AS "Team"
FROM contact_info ci1
    LEFT JOIN name_lookup nl1
        ON nl1.contact_id = ci1.pri_contact_id
    LEFT JOIN name_lookup nl2
        ON nl2.contact_id = ci1.sec_contact_id
ORDER BY nl1.id, ci1.blah_blah;

Result:

  Con_1 name      Con_2 name      Team
    John Doe     Jose Garcia    Team A
    John Doe            NULL    Team B
 Larry Brown      Mary Smith    Team B

So, now the record with the sec_contact_id = NULL datum appears in your resultset.

A couple of words of advice:

• You really should consider (strongly) consider upgrading from 5.6 to the current version of MySQL 8 - you'll get window functions, generated columns, CHECK constraints - it's at version 22 now, and I haven't heard many complaints, so it'd be a good option!

• Many people consider NULLs to be undesirable and avoid them like the plague - I tend to fall into this category. Therefore, you might wish to consider having two contact tables - pri_contact_info and sec_contact_info. You can decide if you wish to do that and ask a new question - if you do so, let me know here also!

• You might consider answering this question (a comment below your original post): Can there ever be more than two people on a team? I e. There would be 3 columns with team members names for the same row?!

• the PRIMARY KEY of the name_lookup table should be contact_id - the surrogate key id adds nothing to the mix! I assume that contact_id is assigned more or less randomly to employees as they are hired? So, it's essentially a surrogate key itself. Surrogate keys have their place, but sometimes they're not the way to go!

• MySQL is by far and away the worst of the major RDBMS's when it comes to standards support, plus it has a plethora of non-standard "extensions" - in future, maybe you could delimit your SQL strings with apostrophes ' and not double quotes "? My personal suggestion is that you use double quotes for aliases of field names for the final result (presentation, not content) as I have done - it will make your SQL more readable and portable!