Python in a Nutshell, 2nd Edition

Threads in Python

Python offers multithreading on platforms that support threads, such as Win32, Linux, and most other variants of Unix. The Classic Python interpreter does not freely switch threads. In the current implementation, Python uses a global interpreter lock (GIL) to ensure that switching between threads happens only between bytecode instructions or when C code deliberately releases the GIL (Python’s own C code releases the GIL around blocking I/O and sleep operations). An action is said to be atomic if it’s guaranteed that no thread switching within Python’s process occurs between the start and the end of the action. In practice, in the current implementation, operations that look atomic (such as simple assignments and accesses) actually are atomic when executed on objects of built-in types (augmented and multiple assignments, however, are not atomic). In general, though, it is not a good idea to rely on atomicity. For example, you never know when you might be dealing with a derived class rather than an object of a built-in type, meaning there might be callbacks to Python code, and any assumptions of atomicity would then prove unwarranted. Moreover, relying on implementation-depended atomicity would lock your code into a specific implementation and might preclude future upgrades. A much better strategy is to use the synchronization facilities covered in the rest of this chapter.

Python offers multithreading in two different flavors. An older and lower-level module, thread ...