summaryrefslogtreecommitdiff
path: root/Doc/library/asyncio-policy.rst
blob: aa8f8f13eae02165d336ef0efa534f9005930137 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
.. currentmodule:: asyncio


.. _asyncio-policies:

========
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
:func:`get_event_loop`, :func:`set_event_loop`, and
:func:`new_event_loop` functions can be customized.

Policy objects should implement the APIs defined
in the :class:`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:

.. function:: get_event_loop_policy()

   Return the current process-wide policy.

.. function:: 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:: AbstractEventLoopPolicy

   An abstract base class for asyncio policies.

   .. method:: get_event_loop()

      Get the event loop for the current context.

      Return an event loop object implementing the
      :class:`AbstractEventLoop` interface.

      This method should never return ``None``.

      .. versionchanged:: 3.6

   .. method:: set_event_loop(loop)

      Set the event loop for the current context to *loop*.

   .. method:: new_event_loop()

      Create and return a new event loop object.

      This method should never return ``None``.

   .. method:: get_child_watcher()

      Get a child process watcher object.

      Return a watcher object implementing the
      :class:`AbstractChildWatcher` interface.

      This function is Unix specific.

   .. method:: 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:: DefaultEventLoopPolicy

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

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

   .. versionchanged:: 3.8

      On Windows, :class:`ProactorEventLoop` is now used by default.


.. class:: WindowsSelectorEventLoopPolicy

   An alternative event loop policy that uses the
   :class:`SelectorEventLoop` event loop implementation.

   .. availability:: Windows.


.. class:: WindowsProactorEventLoopPolicy

   An alternative event loop policy that uses the
   :class:`ProactorEventLoop` event loop implementation.

   .. availability:: Windows.

.. _asyncio-watchers:

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
:func:`create_subprocess_exec` and :meth:`loop.subprocess_exec`
functions.

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

See also the :ref:`Subprocess and Threads <asyncio-subprocess-threads>`
section.

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

.. function:: get_child_watcher()

   Return the current child watcher for the current policy.

.. function:: set_child_watcher(watcher)

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

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

.. class:: AbstractChildWatcher

   .. method:: 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.

   .. method:: 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.

   .. method:: 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``.

   .. method:: is_active()

      Return ``True`` if the watcher is ready to use.

      Spawning a subprocess with *inactive* current child watcher raises
      :exc:`RuntimeError`.

      .. versionadded:: 3.8

   .. method:: close()

      Close the watcher.

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

.. class:: 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 stating a thread per process requires extra memory.

   This watcher is used by default.

   .. versionadded:: 3.8

.. class:: MultiLoopChildWatcher

   This implementation registers a :py:data:`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 :py:data:`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
   :py:data:`SIGCHLD` is received).

   .. versionadded:: 3.8

.. class:: SafeChildWatcher

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

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

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

.. class:: 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
   :class:`SafeChildWatcher`.


Custom Policies
===============

To implement a new event loop policy, it is recommended to subclass
:class:`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())