For tasks that are I/O bound, such as network calls, parallelizing across multiple threads can greatly increase throughput. Python 3 comes with a ThreadPoolExecutor
class that makes it easy to queue up units of work that will by run by a pool of threads. ThreadPoolExecutor
has an unbounded queue though, and sometimes I want to submit only a certain number of tasks before blocking on the submit()
method. For example, sending a large number of requests to a server as fast as possible without buffering all the requests in memory. The below BoundedExecutor
class below accomplishes this. Note that it delegates to a ThreadPoolExecutor
instance instead of subclassing ThreadPoolExecutor
, this decouples it from the underlying implementation of ThreadPoolExecutor
(see Effective Java [affiliate link] Item #16: Favor composition over inheritance).
Continue reading “TheadPoolExecutor with a bounded queue in Python”