Question

I have persons and a person can contact multiple other persons, so basically the “default” tables would be:

persons (id)
contacts (person1_id, person2_id)

With this schema, I’d have to issue queries like

SELECT   * 
FROM     contacts c
WHERE    ( person1_id = *id of person1* AND person2_id = *id of person2* )
         OR
         ( person1_id = *id of person2* AND person2_id = *id of person1* )

to get the relation between two persons when I insert such a relation only once.

What is the common practice to deal with this situation?

1. Insert data once and do such an OR query
2. Insert the relation twice so that person1_id = id of person1 AND person2_id = id of person2 is enough
3. An entirely different approach?

Assuming:

• The m:n table actually contains additional data, so if I create a relation for both ways, I’d have to duplicate the data
• This is a core part of the application and most non-trivial queries involve at least a sub query that determines whether or not such a relation exists

Answer 1

If you write your insert logic such that person1_id < person2_id is true for all rows, then you can just write

SELECT *
FROM contacts c
WHERE person1_id = min(*id_of_person_1*, *id_of_person_2*)
AND person2_id = max(*id_of_person_1*, *id_of_person_2*)