Decoupling With Named Signals

Named signals are created with signal():

>>> from blinker import signal
>>> initialized = signal('initialized')
>>> initialized is signal('initialized')

Every call to signal('name') returns the same signal object, allowing unconnected parts of code (different modules, plugins, anything) to all use the same signal without requiring any code sharing or special imports.

Subscribing to Signals

Signal.connect() registers a function to be invoked each time the signal is emitted. Connected functions are always passed the object that caused the signal to be emitted.

>>> def subscriber(sender):
...     print("Got a signal sent by %r" % sender)
>>> ready = signal('ready')
>>> ready.connect(subscriber)
<function subscriber at 0x...>

Emitting Signals

Code producing events of interest can Signal.send() notifications to all connected receivers.

Below, a simple Processor class emits a ready signal when it’s about to process something, and complete when it is done. It passes self to the send() method, signifying that that particular instance was responsible for emitting the signal.

>>> class Processor:
...    def __init__(self, name):
... = name
...    def go(self):
...        ready = signal('ready')
...        ready.send(self)
...        print("Processing.")
...        complete = signal('complete')
...        complete.send(self)
...    def __repr__(self):
...        return '<Processor %s>' %
>>> processor_a = Processor('a')
>>> processor_a.go()
Got a signal sent by <Processor a>

Notice the complete signal in go()? No receivers have connected to complete yet, and that’s a-ok. Calling send() on a signal with no receivers will result in no notifications being sent, and these no-op sends are optimized to be as inexpensive as possible.

Subscribing to Specific Senders

The default connection to a signal invokes the receiver function when any sender emits it. The Signal.connect() function accepts an optional argument to restrict the subscription to one specific sending object:

>>> def b_subscriber(sender):
...     print("Caught signal from processor_b.")
...     assert == 'b'
>>> processor_b = Processor('b')
>>> ready.connect(b_subscriber, sender=processor_b)
<function b_subscriber at 0x...>

This function has been subscribed to ready but only when sent by processor_b:

>>> processor_a.go()
Got a signal sent by <Processor a>
>>> processor_b.go()
Got a signal sent by <Processor b>
Caught signal from processor_b.

Sending and Receiving Data Through Signals

Additional keyword arguments can be passed to send(). These will in turn be passed to the connected functions:

>>> send_data = signal('send-data')
>>> @send_data.connect
... def receive_data(sender, **kw):
...     print("Caught signal from %r, data %r" % (sender, kw))
...     return 'received!'
>>> result = send_data.send('anonymous', abc=123)
Caught signal from 'anonymous', data {'abc': 123}

The return value of send() collects the return values of each connected function as a list of (receiver function, return value) pairs:

>>> result
[(<function receive_data at 0x...>, 'received!')]

Anonymous Signals

Signals need not be named. The Signal constructor creates a unique signal each time it is invoked. For example, an alternative implementation of the Processor from above might provide the processing signals as class attributes:

>>> from blinker import Signal
>>> class AltProcessor:
...    on_ready = Signal()
...    on_complete = Signal()
...    def __init__(self, name):
... = name
...    def go(self):
...        self.on_ready.send(self)
...        print("Alternate processing.")
...        self.on_complete.send(self)
...    def __repr__(self):
...        return '<AltProcessor %s>' %

connect as a Decorator

You may have noticed the return value of connect() in the console output in the sections above. This allows connect to be used as a decorator on functions:

>>> apc = AltProcessor('c')
>>> @apc.on_complete.connect
... def completed(sender):
...     print "AltProcessor %s completed!" %
>>> apc.go()
Alternate processing.
AltProcessor c completed!

While convenient, this form unfortunately does not allow the sender or weak arguments to be customized for the connected function. For this, connect_via() can be used:

>>> dice_roll = signal('dice_roll')
>>> @dice_roll.connect_via(1)
... @dice_roll.connect_via(3)
... @dice_roll.connect_via(5)
... def odd_subscriber(sender):
...     print("Observed dice roll %r." % sender)
>>> result = dice_roll.send(3)
Observed dice roll 3.

Optimizing Signal Sending

Signals are optimized to send very quickly, whether receivers are connected or not. If the data to be sent down a signal is very expensive, it can be more efficient to check to see if any receivers are connected first by testing the receivers property:

>>> bool(signal('ready').receivers)
>>> bool(signal('complete').receivers)
>>> bool(AltProcessor.on_complete.receivers)

Checking for a receiver listening for a particular sender is also possible:

>>> signal('ready').has_receivers_for(processor_a)

Documenting Signals

Both named and anonymous signals can be passed a doc argument at construction to set the pydoc help text for the signal. This documentation will be picked up by most documentation generators (such as sphinx) and is nice for documenting any additional data parameters that will be sent down with the signal.

See the documentation of the receiver_connected built-in signal for an example.


Disconnecting receivers from signals, introspection of connected receivers, private namespaces for named signals and more are discussed in the API Documentation.