PostgreSQL triggers in plpython

I wanted to run a query that inserted rows into table B whenever a new row was added to table A.

The first step took quite some time. You have to start off by adding the procedural language you wish to use to the database you're working with. I'm happy with Python, so I chose plpython:

$ psql

# CREATE LANGUAGE plpythonu;

but it didn't like that, I got an error:

ERROR:  could not access file "$libdir/plpython": No such file or directory

Googling failed me! I couldn't find the answer anywhere. This time, a bit of "aptitude foo" came in useful (I'm running Ubuntu):

$ aptitude search postgresql

there were a lot of results, but neatly nestled towards the bottom was:

p   postgresql-plpython-8.4         - PL/Python procedural language for PostgreS

so I installed that:

$ sudo aptitude install postgresql-plpython-8.4

back to psql and try creating the language again:

# CREATE LANGUAGE plpythonu;


Next to create the function that I wanted to use in a trigger, I discovered a useful command in psql:

# \e

this pulls up the environment default editor to make writing queries a little less horrible!

CREATE FUNCTION unread_trigger()
AS $$
qry = plpy.prepare("""INSERT INTO unread_messages
                      SELECT m.message_id, s.user_id
                      FROM messages m
                      JOIN conversation_subscriptions s
                        ON s.conversation_id = m.conversation_id
                      WHERE m.message_id = $1;""", ("INTEGER",))
res = plpy.execute(qry, (TD['new']['message_id'],))
$$ LANGUAGE plpythonu;

When you exit the editor and save the temporary file, psql executes the query. I'll explain what's going on here in a minute. Another useful command in psql lists the functions created:

# \df

The next step was to add the function as a trigger to my table:

# CREATE TRIGGER set_unread 
# AFTER INSERT ON messages 

so, every time a row is inserted into the messages table, the unread_trigger() function is called and inserts a new row in the unread_messages table for each user that is subscribed to the conversation that message belongs to. In that function, there is a dictionary called TD, this dictionary holds the various values relevant to the event that triggered the function, it's documented in the PL/Python PostgreSQL docs.

The PL/Python language includes a module that allows access to the database: plpy.

Finally, I didn't get the function right first time. I couldn't work out an easy way to modify a function, so I ended up DROPping and re-CREATEing it:

DROP TRIGGER set_unread ON messages;
DROP FUNCTION unread_trigger();

EDIT: I just realised you can use CREATE OR REPLACE FUNCTION ... and it will change the function definition, instead of dropping and re-creating it every time!