Prioritize JOIN on multiple columns with an OR

Question

I'm trying to track the source of comments on a blog site and my question relates to priority JOINing on multiple columns with an OR statement.

The comments table:

      Column        |          Type
--------------------+-------------------------
 id                 | integer
 text               | character varying(1024)
 session_source_id  | integer
 user_source_id     | integer

If I have a query like so:

SELECT s.type, COUNT(c.id) FROM comments c
   LEFT JOIN sources s ON (c.session_source_id = s.id OR c.user_source_id = s.id) 
GROUP BY 1;

There are situations where we know the user source but NOT the session source (which is why I used OR), but I want to prioritize the session_source_id if we also have a user_source_id and those two ids differ.

(there are also situations where we don't know either sources and both columns are null, hence the LEFT JOIN)

Does that query prioritize the JOIN on session_source_id since it's listed first in the joining OR statement? How does Postgres handle OR conditionals in JOIN statements?

Answer 1

Posting GarethD's comment as answer and Wiki, so that it won't be lost:

Try COALESCE instead of OR...

LEFT JOIN sources s 
ON s.id = COALESCE(c.session_source_id, c.user_source_id)

Answer 2

Using OR won't prioritize the conditions, if both conditions are true, both records will be returned.
If you want any priority, you can use CASE EXPRESSION which will break the moment a condition is met :

SELECT s.type, COUNT(c.id) FROM comments c
LEFT JOIN sources s 
 ON (CASE WHEN c.session_source_id = s.id THEN 1
          WHEN c.user_source_id = s.id THEN 1
          ELSE 0
     END = 1) 
GROUP BY 1;