Purpose of pool.join, pool.close in multiprocessing?

Question

In the link below there is an explanation of the map method on the Pool class.

It seems that it blocks until the result is ready. This implies that there is no need to do pool.close(); pool.join() after running pool.map, however it is demo'd in this way in this blog.

Am I missing something or is there a need to do pool.close after running pool.map (as opposed to pool.map_async? Note I am using [multiprocessing.dummy][2], which is provides a similar api to multiprocessing, but uses threading under the covers.

https://docs.python.org/2/library/multiprocessing.html#multiprocessing.pool.multiprocessing.Pool.map

Answer 1

pool.close tells the pool not to accept any new job.

pool.join tells the pool to wait until all jobs finished then exit, effectively cleaning up the pool.

So blocking the parent process is just a side effect of what pool.join is doing.

It's true that when you call pool.map(), the parent process is blocked until map returns result. But it's a good practice to close and join the pool after using it, to better manage resource and control exceptions.