gunthercox
commented
8 years ago It might be a good idea to weight the advantages of threading vs multiprocessing for this task. Most computers nowadays are built with multi-core processors.
A multiprocessor implementation of this example would allow multi_adapter.py to run the processes for several adapters in parallel on separate processors. Of course the adapter would still need to wait for each adapter to finish processing, but running them in parallel would still accomplish this faster even if a few adapters were lagging behind the others.
Also, (I have to verify this) but I believe that threading would be restricted to a single core. I do think threading would be quite useful for tasks in some adapters such as where a bit of background processing could provide an advantage. (For instance, an email IO adapter could use a thread to periodically check for new emails, if there were new ones it would add them to a queue for the next time it was called to retrieve data. This would remove the need to wait for a web request to return with the updated information every time the adapter is used.)
To increase the speed of processing, multithreading support is really needed.
In the future, when more logic adapters are added including more advanced ones (such as a developer assistant), https://github.com/gunthercox/ChatterBot/blob/master/chatterbot/adapters/logic/multi_adapter.py#L11 will consume a lot of computing resources. If this is all in one thread, it could cause a slowdown of the program this is very noticeable.
Here is my solution: We enable threading at that stage of the processing. In addition, if that is not enough, individual logic adapters can also implement threading, but that is for another issue