Source code for tornado.platform.asyncio

"""Bridges between the `asyncio` module and Tornado IOLoop.

.. versionadded:: 3.2

This module integrates Tornado with the ``asyncio`` module introduced
in Python 3.4. This makes it possible to combine the two libraries on
the same event loop.

.. deprecated:: 5.0

   While the code in this module is still used, it is now enabled
   automatically when `asyncio` is available, so applications should
   no longer need to refer to this module directly.

.. note::

   Tornado is designed to use a selector-based event loop. On Windows,
   where a proactor-based event loop has been the default since Python 3.8,
   a selector event loop is emulated by running ``select`` on a separate thread.
   Configuring ``asyncio`` to use a selector event loop may improve performance
   of Tornado (but may reduce performance of other ``asyncio``-based libraries
   in the same process).

import asyncio
import atexit
import concurrent.futures
import errno
import functools
import select
import socket
import sys
import threading
import typing
from tornado.gen import convert_yielded
from tornado.ioloop import IOLoop, _Selectable

from typing import Any, TypeVar, Awaitable, Callable, Union, Optional, List, Tuple, Dict

if typing.TYPE_CHECKING:
    from typing import Set  # noqa: F401
    from typing_extensions import Protocol

    class _HasFileno(Protocol):
        def fileno(self) -> int:

    _FileDescriptorLike = Union[int, _HasFileno]

_T = TypeVar("_T")

# Collection of selector thread event loops to shut down on exit.
_selector_loops = set()  # type: Set[AddThreadSelectorEventLoop]

def _atexit_callback() -> None:
    for loop in _selector_loops:
        with loop._select_cond:
            loop._closing_selector = True
        except BlockingIOError:
        # If we don't join our (daemon) thread here, we may get a deadlock
        # during interpreter shutdown. I don't really understand why. This
        # deadlock happens every time in CI (both travis and appveyor) but
        # I've never been able to reproduce locally.


class BaseAsyncIOLoop(IOLoop):
    def initialize(  # type: ignore
        self, asyncio_loop: asyncio.AbstractEventLoop, **kwargs: Any
    ) -> None:
        # asyncio_loop is always the real underlying IOLoop. This is used in
        # to maintain the asyncio-to-ioloop mappings.
        self.asyncio_loop = asyncio_loop
        # selector_loop is an event loop that implements the add_reader family of
        # methods. Usually the same as asyncio_loop but differs on platforms such
        # as windows where the default event loop does not implement these methods.
        self.selector_loop = asyncio_loop
        if hasattr(asyncio, "ProactorEventLoop") and isinstance(
            asyncio_loop, asyncio.ProactorEventLoop  # type: ignore
            # Ignore this line for mypy because the abstract method checker
            # doesn't understand dynamic proxies.
            self.selector_loop = AddThreadSelectorEventLoop(asyncio_loop)  # type: ignore
        # Maps fd to (fileobj, handler function) pair (as in IOLoop.add_handler)
        self.handlers = {}  # type: Dict[int, Tuple[Union[int, _Selectable], Callable]]
        # Set of fds listening for reads/writes
        self.readers = set()  # type: Set[int]
        self.writers = set()  # type: Set[int]
        self.closing = False
        # If an asyncio loop was closed through an asyncio interface
        # instead of IOLoop.close(), we'd never hear about it and may
        # have left a dangling reference in our map. In case an
        # application (or, more likely, a test suite) creates and
        # destroys a lot of event loops in this way, check here to
        # ensure that we don't have a lot of dead loops building up in
        # the map.
        # TODO(bdarnell): consider making self.asyncio_loop a weakref
        # for AsyncIOMainLoop and make _ioloop_for_asyncio a
        # WeakKeyDictionary.
        for loop in list(IOLoop._ioloop_for_asyncio):
            if loop.is_closed():
                del IOLoop._ioloop_for_asyncio[loop]
        IOLoop._ioloop_for_asyncio[asyncio_loop] = self

        self._thread_identity = 0


        def assign_thread_identity() -> None:
            self._thread_identity = threading.get_ident()


    def close(self, all_fds: bool = False) -> None:
        self.closing = True
        for fd in list(self.handlers):
            fileobj, handler_func = self.handlers[fd]
            if all_fds:
        # Remove the mapping before closing the asyncio loop. If this
        # happened in the other order, we could race against another
        # initialize() call which would see the closed asyncio loop,
        # assume it was closed from the asyncio side, and do this
        # cleanup for us, leading to a KeyError.
        del IOLoop._ioloop_for_asyncio[self.asyncio_loop]
        if self.selector_loop is not self.asyncio_loop:

    def add_handler(
        self, fd: Union[int, _Selectable], handler: Callable[..., None], events: int
    ) -> None:
        fd, fileobj = self.split_fd(fd)
        if fd in self.handlers:
            raise ValueError("fd %s added twice" % fd)
        self.handlers[fd] = (fileobj, handler)
        if events & IOLoop.READ:
            self.selector_loop.add_reader(fd, self._handle_events, fd, IOLoop.READ)
        if events & IOLoop.WRITE:
            self.selector_loop.add_writer(fd, self._handle_events, fd, IOLoop.WRITE)

    def update_handler(self, fd: Union[int, _Selectable], events: int) -> None:
        fd, fileobj = self.split_fd(fd)
        if events & IOLoop.READ:
            if fd not in self.readers:
                self.selector_loop.add_reader(fd, self._handle_events, fd, IOLoop.READ)
            if fd in self.readers:
        if events & IOLoop.WRITE:
            if fd not in self.writers:
                self.selector_loop.add_writer(fd, self._handle_events, fd, IOLoop.WRITE)
            if fd in self.writers:

    def remove_handler(self, fd: Union[int, _Selectable]) -> None:
        fd, fileobj = self.split_fd(fd)
        if fd not in self.handlers:
        if fd in self.readers:
        if fd in self.writers:
        del self.handlers[fd]

    def _handle_events(self, fd: int, events: int) -> None:
        fileobj, handler_func = self.handlers[fd]
        handler_func(fileobj, events)

    def start(self) -> None:
            old_loop = asyncio.get_event_loop()
        except (RuntimeError, AssertionError):
            old_loop = None  # type: ignore

    def stop(self) -> None:

    def call_at(
        self, when: float, callback: Callable[..., None], *args: Any, **kwargs: Any
    ) -> object:
        # asyncio.call_at supports *args but not **kwargs, so bind them here.
        # We do not synchronize self.time and asyncio_loop.time, so
        # convert from absolute to relative.
        return self.asyncio_loop.call_later(
            max(0, when - self.time()),
            functools.partial(callback, *args, **kwargs),

    def remove_timeout(self, timeout: object) -> None:
        timeout.cancel()  # type: ignore

    def add_callback(self, callback: Callable, *args: Any, **kwargs: Any) -> None:
        if threading.get_ident() == self._thread_identity:
            call_soon = self.asyncio_loop.call_soon
            call_soon = self.asyncio_loop.call_soon_threadsafe
            call_soon(self._run_callback, functools.partial(callback, *args, **kwargs))
        except RuntimeError:
            # "Event loop is closed". Swallow the exception for
            # consistency with PollIOLoop (and logical consistency
            # with the fact that we can't guarantee that an
            # add_callback that completes without error will
            # eventually execute).
        except AttributeError:
            # ProactorEventLoop may raise this instead of RuntimeError
            # if call_soon_threadsafe races with a call to close().
            # Swallow it too for consistency.

    def add_callback_from_signal(
        self, callback: Callable, *args: Any, **kwargs: Any
    ) -> None:
                self._run_callback, functools.partial(callback, *args, **kwargs)
        except RuntimeError:

    def run_in_executor(
        executor: Optional[concurrent.futures.Executor],
        func: Callable[..., _T],
        *args: Any
    ) -> Awaitable[_T]:
        return self.asyncio_loop.run_in_executor(executor, func, *args)

    def set_default_executor(self, executor: concurrent.futures.Executor) -> None:
        return self.asyncio_loop.set_default_executor(executor)

[docs]class AsyncIOMainLoop(BaseAsyncIOLoop): """``AsyncIOMainLoop`` creates an `.IOLoop` that corresponds to the current ``asyncio`` event loop (i.e. the one returned by ``asyncio.get_event_loop()``). .. deprecated:: 5.0 Now used automatically when appropriate; it is no longer necessary to refer to this class directly. .. versionchanged:: 5.0 Closing an `AsyncIOMainLoop` now closes the underlying asyncio loop. """ def initialize(self, **kwargs: Any) -> None: # type: ignore super().initialize(asyncio.get_event_loop(), **kwargs) def make_current(self) -> None: # AsyncIOMainLoop already refers to the current asyncio loop so # nothing to do here. pass
[docs]class AsyncIOLoop(BaseAsyncIOLoop): """``AsyncIOLoop`` is an `.IOLoop` that runs on an ``asyncio`` event loop. This class follows the usual Tornado semantics for creating new ``IOLoops``; these loops are not necessarily related to the ``asyncio`` default event loop. Each ``AsyncIOLoop`` creates a new ``asyncio.EventLoop``; this object can be accessed with the ``asyncio_loop`` attribute. .. versionchanged:: 5.0 When an ``AsyncIOLoop`` becomes the current `.IOLoop`, it also sets the current `asyncio` event loop. .. deprecated:: 5.0 Now used automatically when appropriate; it is no longer necessary to refer to this class directly. """ def initialize(self, **kwargs: Any) -> None: # type: ignore self.is_current = False loop = asyncio.new_event_loop() try: super().initialize(loop, **kwargs) except Exception: # If initialize() does not succeed (taking ownership of the loop), # we have to close it. loop.close() raise def close(self, all_fds: bool = False) -> None: if self.is_current: self.clear_current() super().close(all_fds=all_fds) def make_current(self) -> None: if not self.is_current: try: self.old_asyncio = asyncio.get_event_loop() except (RuntimeError, AssertionError): self.old_asyncio = None # type: ignore self.is_current = True asyncio.set_event_loop(self.asyncio_loop) def _clear_current_hook(self) -> None: if self.is_current: asyncio.set_event_loop(self.old_asyncio) self.is_current = False
[docs]def to_tornado_future(asyncio_future: asyncio.Future) -> asyncio.Future: """Convert an `asyncio.Future` to a `tornado.concurrent.Future`. .. versionadded:: 4.1 .. deprecated:: 5.0 Tornado ``Futures`` have been merged with `asyncio.Future`, so this method is now a no-op. """ return asyncio_future
[docs]def to_asyncio_future(tornado_future: asyncio.Future) -> asyncio.Future: """Convert a Tornado yieldable object to an `asyncio.Future`. .. versionadded:: 4.1 .. versionchanged:: 4.3 Now accepts any yieldable object, not just `tornado.concurrent.Future`. .. deprecated:: 5.0 Tornado ``Futures`` have been merged with `asyncio.Future`, so this method is now equivalent to `tornado.gen.convert_yielded`. """ return convert_yielded(tornado_future)
if sys.platform == "win32" and hasattr(asyncio, "WindowsSelectorEventLoopPolicy"): # "Any thread" and "selector" should be orthogonal, but there's not a clean # interface for composing policies so pick the right base. _BasePolicy = asyncio.WindowsSelectorEventLoopPolicy # type: ignore else: _BasePolicy = asyncio.DefaultEventLoopPolicy
[docs]class AnyThreadEventLoopPolicy(_BasePolicy): # type: ignore """Event loop policy that allows loop creation on any thread. The default `asyncio` event loop policy only automatically creates event loops in the main threads. Other threads must create event loops explicitly or `asyncio.get_event_loop` (and therefore `.IOLoop.current`) will fail. Installing this policy allows event loops to be created automatically on any thread, matching the behavior of Tornado versions prior to 5.0 (or 5.0 on Python 2). Usage:: asyncio.set_event_loop_policy(AnyThreadEventLoopPolicy()) .. versionadded:: 5.0 """ def get_event_loop(self) -> asyncio.AbstractEventLoop: try: return super().get_event_loop() except (RuntimeError, AssertionError): # This was an AssertionError in Python 3.4.2 (which ships with Debian Jessie) # and changed to a RuntimeError in 3.4.3. # "There is no current event loop in thread %r" loop = self.new_event_loop() self.set_event_loop(loop) return loop
[docs]class AddThreadSelectorEventLoop(asyncio.AbstractEventLoop): """Wrap an event loop to add implementations of the ``add_reader`` method family. Instances of this class start a second thread to run a selector. This thread is completely hidden from the user; all callbacks are run on the wrapped event loop's thread. This class is used automatically by Tornado; applications should not need to refer to it directly. It is safe to wrap any event loop with this class, although it only makes sense for event loops that do not implement the ``add_reader`` family of methods themselves (i.e. ``WindowsProactorEventLoop``) Closing the ``AddThreadSelectorEventLoop`` also closes the wrapped event loop. """ # This class is a __getattribute__-based proxy. All attributes other than those # in this set are proxied through to the underlying loop. MY_ATTRIBUTES = { "_consume_waker", "_select_cond", "_select_args", "_closing_selector", "_thread", "_handle_event", "_readers", "_real_loop", "_start_select", "_run_select", "_handle_select", "_wake_selector", "_waker_r", "_waker_w", "_writers", "add_reader", "add_writer", "close", "remove_reader", "remove_writer", } def __getattribute__(self, name: str) -> Any: if name in AddThreadSelectorEventLoop.MY_ATTRIBUTES: return super().__getattribute__(name) return getattr(self._real_loop, name) def __init__(self, real_loop: asyncio.AbstractEventLoop) -> None: self._real_loop = real_loop # Create a thread to run the select system call. We manage this thread # manually so we can trigger a clean shutdown from an atexit hook. Note # that due to the order of operations at shutdown, only daemon threads # can be shut down in this way (non-daemon threads would require the # introduction of a new hook: self._select_cond = threading.Condition() self._select_args = ( None ) # type: Optional[Tuple[List[_FileDescriptorLike], List[_FileDescriptorLike]]] self._closing_selector = False self._thread = threading.Thread( name="Tornado selector", daemon=True, target=self._run_select, ) self._thread.start() # Start the select loop once the loop is started. self._real_loop.call_soon(self._start_select) self._readers = {} # type: Dict[_FileDescriptorLike, Callable] self._writers = {} # type: Dict[_FileDescriptorLike, Callable] # Writing to _waker_w will wake up the selector thread, which # watches for _waker_r to be readable. self._waker_r, self._waker_w = socket.socketpair() self._waker_r.setblocking(False) self._waker_w.setblocking(False) _selector_loops.add(self) self.add_reader(self._waker_r, self._consume_waker) def __del__(self) -> None: # If the top-level application code uses asyncio interfaces to # start and stop the event loop, no objects created in Tornado # can get a clean shutdown notification. If we're just left to # be GC'd, we must explicitly close our sockets to avoid # logging warnings. _selector_loops.discard(self) self._waker_r.close() self._waker_w.close() def close(self) -> None: with self._select_cond: self._closing_selector = True self._select_cond.notify() self._wake_selector() self._thread.join() _selector_loops.discard(self) self._waker_r.close() self._waker_w.close() self._real_loop.close() def _wake_selector(self) -> None: try: self._waker_w.send(b"a") except BlockingIOError: pass def _consume_waker(self) -> None: try: self._waker_r.recv(1024) except BlockingIOError: pass def _start_select(self) -> None: # Capture reader and writer sets here in the event loop # thread to avoid any problems with concurrent # modification while the select loop uses them. with self._select_cond: assert self._select_args is None self._select_args = (list(self._readers.keys()), list(self._writers.keys())) self._select_cond.notify() def _run_select(self) -> None: while True: with self._select_cond: while self._select_args is None and not self._closing_selector: self._select_cond.wait() if self._closing_selector: return assert self._select_args is not None to_read, to_write = self._select_args self._select_args = None # We use the simpler interface of the select module instead of # the more stateful interface in the selectors module because # this class is only intended for use on windows, where # is the only option. The selector interface # does not have well-documented thread-safety semantics that # we can rely on so ensuring proper synchronization would be # tricky. try: # On windows, selecting on a socket for write will not # return the socket when there is an error (but selecting # for reads works). Also select for errors when selecting # for writes, and merge the results. # # This pattern is also used in # rs, ws, xs =, to_write, to_write) ws = ws + xs except OSError as e: # After remove_reader or remove_writer is called, the file # descriptor may subsequently be closed on the event loop # thread. It's possible that this select thread hasn't # gotten into the select system call by the time that # happens in which case (at least on macOS), select may # raise a "bad file descriptor" error. If we get that # error, check and see if we're also being woken up by # polling the waker alone. If we are, just return to the # event loop and we'll get the updated set of file # descriptors on the next iteration. Otherwise, raise the # original error. if e.errno == getattr(errno, "WSAENOTSOCK", errno.EBADF): rs, _, _ =[self._waker_r.fileno()], [], [], 0) if rs: ws = [] else: raise else: raise self._real_loop.call_soon_threadsafe(self._handle_select, rs, ws) def _handle_select( self, rs: List["_FileDescriptorLike"], ws: List["_FileDescriptorLike"] ) -> None: for r in rs: self._handle_event(r, self._readers) for w in ws: self._handle_event(w, self._writers) self._start_select() def _handle_event( self, fd: "_FileDescriptorLike", cb_map: Dict["_FileDescriptorLike", Callable], ) -> None: try: callback = cb_map[fd] except KeyError: return callback() def add_reader( self, fd: "_FileDescriptorLike", callback: Callable[..., None], *args: Any ) -> None: self._readers[fd] = functools.partial(callback, *args) self._wake_selector() def add_writer( self, fd: "_FileDescriptorLike", callback: Callable[..., None], *args: Any ) -> None: self._writers[fd] = functools.partial(callback, *args) self._wake_selector() def remove_reader(self, fd: "_FileDescriptorLike") -> None: del self._readers[fd] self._wake_selector() def remove_writer(self, fd: "_FileDescriptorLike") -> None: del self._writers[fd] self._wake_selector()