mirror of
https://github.com/predis/predis.git
synced 2026-08-24 12:09:32 +00:00
Daniele Alessandri
dd5d665156
This is a complete overhaul of how aggregate connections are created and initialized, now everything is self-contained in our usual 3 supported client options: "aggregate", "cluster" and "replication". The usage of callables acting as connection initializerss is now more consistent through the various options. When the callable is invoked it receives 3 arguments (the original set of connection parameters passed by reference, the options container, the current option) and must return an instance of Predis\Connection\AggregateConnectionInterface otherwise an InvalidArgumentException is thrown. When using "cluster" and "replication" the returned aggregate connection is automatically populated by adding the list of nodes in $parameters, on the other hand "aggregate" skips this automatism so it is up to the user. In any case the user-supplied callable receives $parameters as a reference, setting $parameters to NULL inside the body of the callable makes the client skip automatic aggregation regardless of the option in use. In addition to this the actual procedure of adding nodes to an aggregate connection has been moved directly into the respective options instead of being spread between the client (which instead should just pass a set of parameters and get back a fully-configured aggregate connection) and the connection factory (and the scope of a connection factory is only to create new connetion instances to single Redis servers).