twisted.internet.asyncioreactor.AsyncioSelectorReactor(PosixReactorBase)
class documentationtwisted.internet.asyncioreactor
View Source
(View In Hierarchy)
Implements interfaces: twisted.internet.interfaces.IReactorFDSet
Reactor running on top of asyncio.SelectorEventLoop
.
Method | __init__ | Undocumented |
Method | addReader | I add reader to the set of file descriptors to get read events for. |
Method | addWriter | I add writer to the set of file descriptors to get write events for. |
Method | removeReader | Removes an object previously added with addReader . |
Method | removeWriter | Removes an object previously added with addWriter . |
Method | removeAll | Remove all readers and writers. |
Method | getReaders | Return the list of file descriptors currently monitored for input events by the reactor. |
Method | getWriters | Return the list file descriptors currently monitored for output events by the reactor. |
Method | getDelayedCalls | No summary |
Method | iterate | See twisted.internet.interfaces.IReactorCore.iterate. |
Method | run | Fire 'startup' System Events, move the reactor to the 'running' state,
then run the main loop until it is stopped with stop() or
crash() . |
Method | stop | See twisted.internet.interfaces.IReactorCore.stop. |
Method | crash | See twisted.internet.interfaces.IReactorCore.crash. |
Method | seconds | Get the current time in seconds. |
Method | callLater | See twisted.internet.interfaces.IReactorTime.callLater. |
Method | callFromThread | Cause a function to be executed by the reactor thread. |
Method | _unregisterFDInAsyncio | Compensate for a bug in asyncio where it will not unregister a FD that it cannot handle in the epoll loop. It touches internal asyncio code. |
Method | _readOrWrite | Undocumented |
Inherited from PosixReactorBase:
Method | installWaker | Install a `waker' to allow threads and signals to wake up the IO thread. |
Method | spawnProcess | Spawn a process, with a process protocol. |
Method | listenUDP | Connects a given DatagramProtocol
to the given numeric UDP port. |
Method | listenMulticast | Connects a given DatagramProtocol to the given numeric UDP port. |
Method | connectUNIX | Connect a client protocol to a UNIX socket. |
Method | listenUNIX | Listen on a UNIX socket. |
Method | listenUNIXDatagram | Connects a given DatagramProtocol
to the given path. |
Method | connectUNIXDatagram | Connects a ConnectedDatagramProtocol
instance to a path. |
Method | adoptStreamPort | Create a new IListeningPort
from an already-initialized socket. |
Method | adoptStreamConnection | |
Method | adoptDatagramPort | Add an existing listening SOCK_DGRAM socket to the reactor to monitor for read and write readiness. |
Method | listenTCP | Connects a given protocol factory to the given numeric TCP/IP port. |
Method | connectTCP | Connect a TCP client. |
Method | connectSSL | Connect a client Protocol to a remote SSL socket. |
Method | listenSSL | Connects a given protocol factory to the given numeric TCP/IP port. The connection is a SSL one, using contexts created by the context factory. |
Instance Variable | _childWaker | None
or a reference to the _SIGCHLDWaker
which is used to properly notice child process termination. |
Method | _handleSignals | Extend the basic signal handling logic to also support handling SIGCHLD to know when to try to reap child processes. |
Method | _uninstallHandler | If a child waker was created and installed, uninstall it now. |
Method | _removeAll | Remove all readers and writers, and list of removed IReadDescriptor s
and IWriteDescriptor s. |
Inherited from _SignalReactorMixin (via PosixReactorBase):
Method | startRunning | Extend the base implementation in order to remember whether signal handlers should be installed later. |
Method | mainLoop | Undocumented |
Instance Variable | _installSignalHandlers | A flag which indicates whether any signal handlers will be installed during
startup. This includes handlers for SIGCHLD to monitor child processes,
and SIGINT, SIGTERM, and SIGBREAK to stop the reactor. (type: bool ) |
Method | _reallyStartRunning | Extend the base implementation by also installing signal handlers, if
self._installSignalHandlers is true. |
Inherited from _DisconnectSelectableMixin (via PosixReactorBase):
Method | _disconnectSelectable | Utility function for disconnecting a selectable. |
Inherited from ReactorBase (via PosixReactorBase):
Instance Variable | running | See IReactorCore.running |
Method | installResolver | See IReactorPluggableResolver . |
Method | installNameResolver | See IReactorPluggableNameResolver . |
Method | nameResolver | Implementation of read-only IReactorPluggableNameResolver.nameResolver . |
Method | wakeUp | Wake up the event loop. |
Method | doIteration | Do one iteration over the readers and writers which have been added. |
Method | resolve | Return a Deferred that will resolve a hostname. |
Method | sigInt | Handle a SIGINT interrupt. |
Method | sigBreak | Handle a SIGBREAK interrupt. |
Method | sigTerm | Handle a SIGTERM interrupt. |
Method | disconnectAll | Disconnect every reader, and writer in the system. |
Method | fireSystemEvent | See twisted.internet.interfaces.IReactorCore.fireSystemEvent. |
Method | addSystemEventTrigger | See twisted.internet.interfaces.IReactorCore.addSystemEventTrigger. |
Method | removeSystemEventTrigger | See twisted.internet.interfaces.IReactorCore.removeSystemEventTrigger. |
Method | callWhenRunning | See twisted.internet.interfaces.IReactorCore.callWhenRunning. |
Method | startRunning | Method called when reactor starts: do some initialization and fire startup events. |
Method | timeout | Determine the longest time the reactor may sleep (waiting on I/O notification, perhaps) before it must wake up to service a time-related event. |
Method | runUntilCurrent | Run all pending timed calls. |
Method | callFromThread 0 | See twisted.internet.interfaces.IReactorFromThreads.callFromThread . |
Method | getThreadPool | See twisted.internet.interfaces.IReactorThreads.getThreadPool . |
Method | callInThread | See twisted.internet.interfaces.IReactorInThreads.callInThread . |
Method | suggestThreadPoolSize | See twisted.internet.interfaces.IReactorThreads.suggestThreadPoolSize . |
Instance Variable | _stopped | A flag which is true between paired calls to reactor.run and
reactor.stop . This should be replaced with an explicit state
machine. (type: bool ) |
Instance Variable | _justStopped | A flag which is true between the time reactor.stop is called
and the time the shutdown system event is fired. This is used to determine
whether that event should be fired after each iteration through the
mainloop. This should be replaced with an explicit state machine. (type: bool ) |
Instance Variable | _started | A flag which is true from the time reactor.run is called until
the time reactor.run returns. This is used to prevent calls
to reactor.run on a running reactor. This should be replaced
with an explicit state machine. (type: bool ) |
Instance Variable | _registerAsIOThread | A flag controlling whether the reactor will register the thread it is
running in as the I/O thread when it starts. If True ,
registration will be done, otherwise it will not be. |
Instance Variable | _exitSignal | See _ISupportsExitSignalCapturing._exitSignal |
Method | _reallyStartRunning | Method called to transition to the running state. This should happen in the during startup event trigger phase. |
Method | _moveCallLaterSooner | Undocumented |
Method | _cancelCallLater | Undocumented |
Method | _insertNewDelayedCalls | Undocumented |
Method | _checkProcessArgs | Check for valid arguments and environment to spawnProcess. |
Method | _initThreads | Undocumented |
Method | _initThreadPool | Create the threadpool accessible with callFromThread. |
Method | _stopThreadPool | No summary |
Compensate for a bug in asyncio where it will not unregister a FD that it cannot handle in the epoll loop. It touches internal asyncio code.
A description of the bug by markrwilliams:
The add_writer
method of asyncio event loops isn't atomic
because all the Selector classes in the selector module internally record a
file object before passing it to the platform's selector implementation. If
the platform's selector decides the file object isn't acceptable, the
resulting exception doesn't cause the Selector to un-track the file
object.
The failing/hanging stdio test goes through the following sequence of events (roughly):
* The first connection.write(intToByte(value))
call hits
the asyncio reactor's addWriter
method.
* addWriter
calls the asyncio loop's
add_writer
method, which happens to live on
_BaseSelectorEventLoop
.
* The asyncio loop's add_writer
method checks if the file
object has been registered before via the selector's get_key
method.
* It hasn't, so the KeyError block runs and calls the selector's register method
* Code examples that follow use EpollSelector, but the code flow holds true for any other selector implementation. The selector's register method first calls through to the next register method in the MRO
* That next method is always _BaseSelectorImpl.register
which creates a SelectorKey
instance for the file object,
stores it under the file object's file descriptor, and then returns it.
* Control returns to the concrete selector implementation, which asks the operating system to track the file descriptor using the right API.
* The operating system refuses! An exception is raised that, in this
case, the asyncio reactor handles by creating a
_ContinuousPolling
object to watch the file descriptor.
* The second connection.write(intToByte(value))
call hits
the asyncio reactor's addWriter
method, which hits the
add_writer
method. But the loop's selector's get_key method
now returns a SelectorKey
! Now the asyncio reactor's
addWriter
method thinks the asyncio loop will watch the file
descriptor, even though it won't.
I add reader to the set of file descriptors to get read events for.
Parameters | reader | An IReadDescriptor
provider that will be checked for read events until it is removed from the
reactor with removeReader . |
Returns | None . |
I add writer to the set of file descriptors to get write events for.
Parameters | writer | An IWriteDescriptor
provider that will be checked for write events until it is removed from the
reactor with removeWriter . |
Returns | None . |
Remove all readers and writers.
Should not remove reactor internal reactor connections (like a waker).
Returns | A list of IReadDescriptor
and IWriteDescriptor
providers which were removed. |
Return the list of file descriptors currently monitored for input events by the reactor.
Returns | the list of file descriptors monitored for input events. (type: list of IReadDescriptor ) |
Return the list file descriptors currently monitored for output events by the reactor.
Returns | the list of file descriptors monitored for output events. (type: list of IWriteDescriptor ) |
Return all the outstanding delayed calls in the system. They are returned in no particular order. This method is not efficient -- it is really only meant for test cases.
Returns | A list of outstanding delayed calls. |
Fire 'startup' System Events, move the reactor to the 'running' state,
then run the main loop until it is stopped with stop()
or
crash()
.
See twisted.internet.interfaces.IReactorCore.stop.
See twisted.internet.interfaces.IReactorCore.crash.
Reset reactor state tracking attributes and re-initialize certain
state-transition helpers which were set up in __init__
but
later destroyed (through use).
Get the current time in seconds.
Returns | A number-like object of some sort. |
See twisted.internet.interfaces.IReactorTime.callLater.
Cause a function to be executed by the reactor thread.
Use this method when you want to run a function in the reactor's thread
from another thread. Calling callFromThread
should wake up the main thread (where reactor.run()
is executing) and run the given callable in that thread.
If you're writing a multi-threaded application the callable
may need to be thread safe, but this method doesn't require it as such. If
you want to call a function in the next mainloop iteration, but you're in
the same thread, use callLater
with a delay of 0.