Added - PartitionedModel separates CPU resource into .runtime and no longer runs multiple processes when the maximum number of parallel workers is 1. This will make unittests easier to manage.
0.0.47
Changed - The multiprocessing.Queue connecting sub-task processes in PartitionedModel to the parent to be a multiplexed queue with message types. This makes it simple to connect the logs from sub-tasks to the parent's logger. This is an internal change that opens up where log messages surface.
Added - A common patten module to show how multiple independent models can be run in parallel using a PartitionedModel
0.0.46
Changed - MultiConnector.add_engine_url to de-duplicate engine_urls and return a previously built connector when a duplicate engine_url is passed.
0.0.45
Added - overlay methods are now possible with MultiConnector and 'child_method_overlay' can be used to pass a callable to child connectors (of the parent multi connector).
0.0.44
Fixed - SqlAlchemyDatabaseConnector's .add() couldn't add ORM instance when in single model mode
0.0.43
Added - .last_modified to FileBasedConnector connectors. Thanks burnleyrob!
Changed - CsvConnector, ParquetConnector, NdjsonConnector, closes the file handle when iteration has finished with the file
Fixed - iterating within an iterator for a cloned data connector resulted in premature EOF