What is the meaning of 'key' in psycopg connection pooling?

Question

I am going through Psycopg Connections Pooling documentation, and don't understand what is the purpose of 'key' argument in getconn and putconn?

Answer 1

Perusing the source code of psycopg2 (lib/pool.py) you can see that the psycopg2.pool.AbstractConnectionPool class has a dict attribute named _used where the connections in the pool are referenced. The key parameter in the getconn and putconn methods is the the key to the items in that dictionary. By default, if the value of the key parameter of these methods is None the id of the connection object is used as the key.

Basically the key parameter allows implementers of concrete connection pool classes to identify connections. For example, in the psycopg2.pool.PersistentConnectionPool class, a single connection is meant to be shared in a single thread, so the thread ID is used as the key.