Policies

An event loop policy is a global per-process object that controls the management of the event loop. Each event loop has a default policy, which can be changed and customized using the policy API.

A policy defines the notion of context and manages a separate event loop per context. The default policy defines context to be the current thread.

By using a custom event loop policy, the behavior of get_event_loop(), set_event_loop(), and new_event_loop() functions can be customized.

Policy objects should implement the APIs defined in the AbstractEventLoopPolicy abstract base class.

Getting and Setting the Policy

The following functions can be used to get and set the policy for the current process:

asyncio.get_event_loop_policy()

Return the current process-wide policy.

asyncio.set_event_loop_policy(policy)

Set the current process-wide policy to policy.

If policy is set to None, the default policy is restored.

Policy Objects

The abstract event loop policy base class is defined as follows:

class asyncio.AbstractEventLoopPolicy

An abstract base class for asyncio policies.

get_event_loop()

Get the event loop for the current context.

Return an event loop object implementing the AbstractEventLoop interface.

This method should never return None.

Zmienione w wersji 3.6.

set_event_loop(loop)

Set the event loop for the current context to loop.

new_event_loop()

Create and return a new event loop object.

This method should never return None.

get_child_watcher()

Get a child process watcher object.

Return a watcher object implementing the AbstractChildWatcher interface.

This function is Unix specific.

set_child_watcher(watcher)

Set the current child process watcher to watcher.

This function is Unix specific.

asyncio ships with the following built-in policies:

class asyncio.DefaultEventLoopPolicy

The default asyncio policy. Uses SelectorEventLoop on Unix and ProactorEventLoop on Windows.

There is no need to install the default policy manually. asyncio is configured to use the default policy automatically.

Zmienione w wersji 3.8: On Windows, ProactorEventLoop is now used by default.

class asyncio.WindowsSelectorEventLoopPolicy

An alternative event loop policy that uses the SelectorEventLoop event loop implementation.

Availability: Windows.

class asyncio.WindowsProactorEventLoopPolicy

An alternative event loop policy that uses the ProactorEventLoop event loop implementation.

Availability: Windows.

Process Watchers

A process watcher allows customization of how an event loop monitors child processes on Unix. Specifically, the event loop needs to know when a child process has exited.

In asyncio, child processes are created with create_subprocess_exec() and loop.subprocess_exec() functions.

asyncio defines the AbstractChildWatcher abstract base class, which child watchers should implement, and has four different implementations: ThreadedChildWatcher (configured to be used by default), MultiLoopChildWatcher, SafeChildWatcher, and FastChildWatcher.

See also the Subprocess and Threads section.

The following two functions can be used to customize the child process watcher implementation used by the asyncio event loop:

asyncio.get_child_watcher()

Return the current child watcher for the current policy.

asyncio.set_child_watcher(watcher)

Set the current child watcher to watcher for the current policy. watcher must implement methods defined in the AbstractChildWatcher base class.

Informacja

Third-party event loops implementations might not support custom child watchers. For such event loops, using set_child_watcher() might be prohibited or have no effect.

class asyncio.AbstractChildWatcher
add_child_handler(pid, callback, *args)

Register a new child handler.

Arrange for callback(pid, returncode, *args) to be called when a process with PID equal to pid terminates. Specifying another callback for the same process replaces the previous handler.

The callback callable must be thread-safe.

remove_child_handler(pid)

Removes the handler for process with PID equal to pid.

The function returns True if the handler was successfully removed, False if there was nothing to remove.

attach_loop(loop)

Attach the watcher to an event loop.

If the watcher was previously attached to an event loop, then it is first detached before attaching to the new loop.

Note: loop may be None.

is_active()

Return True if the watcher is ready to use.

Spawning a subprocess with inactive current child watcher raises RuntimeError.

Nowe w wersji 3.8.

close()

Close the watcher.

This method has to be called to ensure that underlying resources are cleaned-up.

class asyncio.ThreadedChildWatcher

This implementation starts a new waiting thread for every subprocess spawn.

It works reliably even when the asyncio event loop is run in a non-main OS thread.

There is no noticeable overhead when handling a big number of children (O(1) each time a child terminates), but starting a thread per process requires extra memory.

This watcher is used by default.

Nowe w wersji 3.8.

class asyncio.MultiLoopChildWatcher

This implementation registers a SIGCHLD signal handler on instantiation. That can break third-party code that installs a custom handler for SIGCHLD signal.

The watcher avoids disrupting other code spawning processes by polling every process explicitly on a SIGCHLD signal.

There is no limitation for running subprocesses from different threads once the watcher is installed.

The solution is safe but it has a significant overhead when handling a big number of processes (O(n) each time a SIGCHLD is received).

Nowe w wersji 3.8.

class asyncio.SafeChildWatcher

This implementation uses active event loop from the main thread to handle SIGCHLD signal. If the main thread has no running event loop another thread cannot spawn a subprocess (RuntimeError is raised).

The watcher avoids disrupting other code spawning processes by polling every process explicitly on a SIGCHLD signal.

This solution is as safe as MultiLoopChildWatcher and has the same O(N) complexity but requires a running event loop in the main thread to work.

class asyncio.FastChildWatcher

This implementation reaps every terminated processes by calling os.waitpid(-1) directly, possibly breaking other code spawning processes and waiting for their termination.

There is no noticeable overhead when handling a big number of children (O(1) each time a child terminates).

This solution requires a running event loop in the main thread to work, as SafeChildWatcher.

Custom Policies

To implement a new event loop policy, it is recommended to subclass DefaultEventLoopPolicy and override the methods for which custom behavior is wanted, e.g.:

class MyEventLoopPolicy(asyncio.DefaultEventLoopPolicy):

    def get_event_loop(self):
        """Get the event loop.

        This may be None or an instance of EventLoop.
        """
        loop = super().get_event_loop()
        # Do something with loop ...
        return loop

asyncio.set_event_loop_policy(MyEventLoopPolicy())