Question

I have a PostgreSQL table containing users, a PostgreSQL table containing documents, and a table mapping the users to the documents they’ve read. Like so:

Table: users

------------
| oid      |
| username |
| ...      |
------------

Table: documents

------------
| oid      |
| title    |
| ...      |
------------

Table: users_documents

-----------------
| oid           |
| user_oid      |
| document_oid  |
-----------------

Whenever a user reads a document, a record is added to users_documents with their user id and the document id. This is all working fine.

What I want to do though is select a random unread document for a given user. I feel like I should be able to do this with a fairly simple JOIN, but I can’t get my head around exactly what the query should look like.

Can someone help please? Thanks.

Answer 1

For most data, you can go to the users_documents table. For instance,
a query of the documents read by user_oid = ? would be

SELECT document_oid
FROM users_documents
WHERE user_oid = ?

However, you want the records which are missing.

You can do an outer join, and find the NULLs in the results.

SELECT users_documents.document_oid
FROM users_documents
  RIGHT OUTER JOIN documents
    ON users_documents.document_oid = documents.oid AND users_documents.user_oid = ?

It is important that the “users_documents.user_oid = ?” be in the join in the FROM clause rather than in a WHERE clause, because it would not work in the WHERE clause.