package documentation
Twisted integration with operating system threads.
Module | _convenience |
Common functionality used within the implementation of various workers. |
Module | _ithreads |
Interfaces related to threads. |
Module | _memory |
Implementation of an in-memory worker that defers execution. |
Module | _pool |
Top level thread pool interface, used to implement twisted.python.threadpool . |
Module | _team |
Implementation of a Team of workers; a thread-pool that can allocate work to workers. |
Module | _threadworker |
Implementation of an IWorker based on native threads and queues. |
From __init__.py
:
Interface |
|
A worker that can perform some work concurrently. |
Class |
|
An IWorker implemented based on a mutual-exclusion lock. |
Class |
|
A composite IWorker implementation. |
Class |
|
An IExclusiveWorker implemented based on a single thread and a queue. |
Exception |
|
This worker worker is dead and cannot execute more instructions. |
Function | create |
Create an IWorker that does nothing but defer work, to be performed later. |
Function | pool |
Construct a Team that spawns threads as a thread pool, with the given limiting function. |
def pool(currentLimit:
Callable[ [], int]
, threadFactory: _ThreadFactory
= Thread) -> Team
:
(source)
¶
Construct a Team
that spawns threads as a thread pool, with the given limiting function.
Parameters | |
currentint | a callable that returns the current limit on the number of workers that the returned Team should create; if it already has more workers than that value, no new workers will be created. |
threadthreading.Thread | Factory that, when given a target keyword argument, returns a threading.Thread that will run that target. |
Returns | |
Team | a new Team . |
Note | |
Future maintainers: while the public API for the eventual move to twisted.threads should look something like this, and while this function is necessary to implement the API described by twisted.python.threadpool , I am starting to think the idea of a hard upper limit on threadpool size is just bad (turning memory performance issues into correctness issues well before we run into memory pressure), and instead we should build something with reactor integration for slowly releasing idle threads when they're not needed and rate limiting the creation of new threads rather than just hard-capping it. |