The threading module builds on the low-level features of thread to make working with threads even easier and more pythonic. Using threads allows a program to run multiple operations concurrently in the same process space.
The simplest way to use a Thread is to instantiate it with a target function and call start() to let it begin working.
In [1]:
# %load threading_simple.py
import threading
def worker():
"""thread worker function"""
print ('> Worker')
sleep(2000)
return
threads = []
for i in range(5):
t = threading.Thread(target=worker)
threads.append(t)
t.start()
It is useful to be able to spawn a thread and pass it arguments to tell it what work to do. This example passes a number, which the thread then prints.
In [2]:
# %load threading_simpleargs.py
import threading
def worker(num):
"""thread worker function"""
print ('Worker: %s' % num)
return
threads = []
for i in range(5):
t = threading.Thread(target=worker, args=(i,))
threads.append(t)
t.start()
In [3]:
# %load threading_simpleargs.py
import threading
def worker(num):
"""thread worker function"""
print ('Worker:', num)
return
threads = []
for i in range(5):
t = threading.Thread(target=worker, args=(i,))
threads.append(t)
t.start()
Using arguments to identify or name the thread is cumbersome, and unnecessary. Each Thead instance has a name with a default value that can be changed as the thread is created. Naming threads is useful in server processes with multiple service threads handling different operations.
In [ ]:
# %load threading_names.py
import threading
import time
def worker():
print (threading.currentThread().getName() + 'Starting')
time.sleep(2)
print (threading.currentThread().getName()+'Exiting')
def my_service():
print (threading.currentThread().getName()+ 'Starting')
time.sleep(3)
print (threading.currentThread().getName()+'Exiting')
t = threading.Thread(name='my_service', target=my_service)
w = threading.Thread(name='worker bee', target=worker)
w2 = threading.Thread(target=worker) # use default name
w3 = threading.Thread(target=worker) # use default name
w.start()
w2.start()
w3.start()
t.start()
The debug output includes the name of the current thread on each line. The lines with "Thread-XX" in the thread name column correspond to the unamed theads w2 and w3.
Note: The number in thread name may change because IPython notebook records every thread.
Most programs do not use print to debug. The logging moudule supports embedding the thread name in every log message using the formatter code (threadName)s. Including threading names in log messages makes it easier to trace those messages back to their source.
In [4]:
# %load threading_names_log.py
import logging
import threading
import time
logging.basicConfig(level=logging.DEBUG,
format='[%(levelname)s] (%(threadName)-10s) %(message)s',
)
def worker():
logging.debug('Starting')
time.sleep(2)
logging.debug('Exiting')
def my_service():
logging.debug('Starting')
time.sleep(3)
logging.debug('Exiting')
t = threading.Thread(name='my_service', target=my_service)
w = threading.Thread(name='worker', target=worker)
w2 = threading.Thread(target=worker) # use default name
w.start()
w2.start()
t.start()
NOTE:
logging is also thread-safe, so messages from different threads are kept distinct in the output.
The output is as follow:
$ python threading_names_log.py
[DEBUG] (worker ) Starting
[DEBUG] (Thread-1 ) Starting
[DEBUG] (my_service) Starting
[DEBUG] (Thread-1 ) Exiting
[DEBUG] (worker ) Exiting
[DEBUG] (my_service) Exiting
Up to this point, the example programs have implicitly waited to exit until all threads have completed their work. Sometimes programs spawn a thread as a daemon that runs without blocking the main program from exiting. Using daemon threads is useful for services where there may not be an easy way to interrupt the thread or where letting the thread die in the middle of its work does not lose or corrupt data (for example, a thread that generates "heart beats" for a service monitoring tool). To mark a thread as a daemon, call its setDeamon() method with a boolean argument. The default is for threads to not be daemons, so passing True turns the daemon mode on.
In [ ]:
# %load threading_daemon.py
import threading
import time
import logging
logging.basicConfig(level=logging.DEBUG,
format='(%(threadName)-10s) %(message)s',
)
def daemon():
logging.debug('Starting')
time.sleep(2)
logging.debug('Exiting')
d = threading.Thread(name='daemon', target=daemon)
d.setDaemon(True)
def non_daemon():
logging.debug('Starting')
logging.debug('Exiting')
t = threading.Thread(name='non-daemon', target=non_daemon)
d.start()
t.start()
Notice that the output does not include the "Exiting" message from the daemon thread, since all of the non-daemon threads (including the mian thread) exit before the deamon thread wakes up from its two seconds sleep.
The output is as follow:
$ python threading_daemon.py
(daemon ) Starting
(non-daemon) Starting
(non-daemon) Exiting
To wait until a daemon thread has completed its work, use the join() method.
In [ ]:
# %load threading_daemon_join.py
import threading
import time
import logging
logging.basicConfig(level=logging.DEBUG,
format='(%(threadName)-10s) %(message)s',
)
def daemon():
logging.debug('Starting')
time.sleep(2)
logging.debug('Exiting')
d = threading.Thread(name='daemon', target=daemon)
d.setDaemon(True)
def non_daemon():
logging.debug('Starting')
logging.debug('Exiting')
t = threading.Thread(name='non-daemon', target=non_daemon)
d.start()
t.start()
d.join()
t.join()
Waiting for the daemon thread to exit using join() means it has a chance to produce its "Exiting" message.
$ python threading_daemon_join.py
(daemon ) Starting
(non-daemon) Starting
(non-daemon) Exiting
(daemon ) Exiting
By default, join() blocks indefinitely. It is also possible to pass a timeout argument (a float representing the number of seconds to wait for the thread to become inactive). If the thread does not complete within the timeout period, join() returns anyway.
In [ ]:
# %load threading_daemon_join_timeout.py
import threading
import time
import logging
logging.basicConfig(level=logging.DEBUG,
format='(%(threadName)-10s) %(message)s',
)
def daemon():
logging.debug('Starting')
time.sleep(2)
logging.debug('Exiting')
d = threading.Thread(name='daemon', target=daemon)
d.setDaemon(True)
def non_daemon():
logging.debug('Starting')
logging.debug('Exiting')
t = threading.Thread(name='non-daemon', target=non_daemon)
d.start()
t.start()
d.join(1)
print('d.isAlive()', d.isAlive())
t.join()
print('d.isAlive()', d.isAlive())
Since the timeout passed is less than the amount of time the daemon thread sleeps, the thread is still "alive" after join() returns.
$ python threading_daemon_join_timeout.py
(daemon ) Starting
(non-daemon) Starting
(non-daemon) Exiting
d.isAlive() True
It is not necessary to retain an explicit handle to all the daemon threads in order to ensure they have completed before exiting the main process.
enumerate() returns a list of active Thread instances. The list includes the current thread, and since joining the current thread is not allowed (it introduces a deadlock situation), it must be skipped.
In [ ]:
# %load threading_enumerate.py
import random
import threading
import time
import logging
logging.basicConfig(level=logging.DEBUG,
format='(%(threadName)-10s) %(message)s',
)
def worker():
"""thread worker function"""
t = threading.currentThread()
pause = random.randint(1, 5)
logging.debug('sleeping %s', pause)
time.sleep(pause)
logging.debug('ending')
return
for i in range(3):
t = threading.Thread(target=worker)
t.setDaemon(True)
t.start()
main_thread = threading.currentThread()
for t in threading.enumerate():
if t is main_thread:
continue
logging.debug('joining %s', t.getName())
t.join()
Since the worker is sleeping for a random amount of time, the output from this program may vary. It should look something like this:
$ python threading_enumerate.py
(Thread-1 ) sleeping 3
(Thread-2 ) sleeping 4
(Thread-3 ) sleeping 4
(MainThread) joining Thread-1
(Thread-1 ) ending
(MainThread) joining Thread-3
(Thread-2 ) ending
(Thread-3 ) ending
(MainThread) joining Thread-2
At start-up, a Thread does some basic initialization and then calls its run() method, which calls the target function passed to the constructor. To create a subclass of Thread, override run() to do whatever is necesssary.
In [ ]:
# %load threading_subclass.py
import threading
import logging
logging.basicConfig(level=logging.DEBUG,
format='(%(threadName)-10s) %(message)s',
)
class MyThread(threading.Thread):
def run(self):
logging.debug('running')
return
for i in range(5):
t = MyThread()
t.start()
The return value of run() is ignored.
$ python threading_subclass.py
(Thread-1 ) running
(Thread-2 ) running
(Thread-3 ) running
(Thread-4 ) running
(Thread-5 ) running
Because the args and kwargs values passed to the Thread constructor are saved in private variables, they are not easily accessed from a subclass. To pass arguments to a custom thread type, redefine the constructor to save the values in an instance attribute that can be seen in the subclass.
In [ ]:
# %load threading_subclass_args.py
import threading
import logging
logging.basicConfig(level=logging.DEBUG,
format='(%(threadName)-10s) %(message)s',
)
class MyThreadWithArgs(threading.Thread):
def __init__(self, group=None, target=None, name=None,
args=(), kwargs=None):
threading.Thread.__init__(self, group=group, target=target)
self.args = args
self.kwargs = kwargs
return
def run(self):
logging.debug('running with %s and %s', self.args, self.kwargs)
return
for i in range(5):
t = MyThreadWithArgs(args=(i,), kwargs={'a':'A', 'b':'B'})
t.start()
The ouput is as follow:
$ python threading_subclass_args.py
(Thread-1 ) running with (0,) and {'a': 'A', 'b': 'B'}
(Thread-2 ) running with (1,) and {'a': 'A', 'b': 'B'}
(Thread-3 ) running with (2,) and {'a': 'A', 'b': 'B'}
(Thread-4 ) running with (3,) and {'a': 'A', 'b': 'B'}
(Thread-5 ) running with (4,) and {'a': 'A', 'b': 'B'}
In [ ]:
# %load threading_timer.py
import threading
import time
import logging
logging.basicConfig(level=logging.DEBUG,
format='(%(threadName)-10s) %(message)s',
)
def delayed():
logging.debug('worker running')
return
t1 = threading.Timer(3, delayed)
t1.setName('t1')
t2 = threading.Timer(3, delayed)
t2.setName('t2')
logging.debug('starting timers')
t1.start()
t2.start()
logging.debug('waiting before canceling %s', t2.getName())
time.sleep(2)
logging.debug('cancaling %s', t2.getName())
t2.cancel()
logging.debug('done')
Notice that the second timer is never run, and the first timer appears to run after the rest of the main program is done. Since it is not a daemon thread, it is joined implicitly when the main thread is done.
$ python threading_timer.py
(MainThread) starting timers
(MainThread) waiting before canceling t2
(MainThread) cancaling t2
(MainThread) done
(t1 ) worker running
Although the point of using multiple threads is to spin separate operations off to run concurrently, there are times when it is important to be able to synchronize the operations in two or more threads. A simple way to communicate between threads is using Event objects. An Event manages an internal flag that callers can either set() or clear(). Other threads can wait() for the flag to be set(), effectively blocking progress until allowed to continue.
In [ ]:
# %load threading_event.py
import logging
import threading
import time
logging.basicConfig(level=logging.DEBUG,
format='(%(threadName)-10s) %(message)s',
)
def wait_for_event(e):
"""Wait for the event to be set before doing anything"""
logging.debug('wait_for_event starting')
event_is_set = e.wait()
logging.debug('event set: %s', event_is_set)
def wait_for_event_timeout(e, t):
"""Wait t seconds and then timeout"""
while not e.isSet():
logging.debug('wait_for_event_timeout starting')
event_is_set = e.wait(t)
logging.debug('event set: %s', event_is_set)
if event_is_set:
logging.debug('processing event')
else:
logging.debug('doing other work')
e = threading.Event()
t1 = threading.Thread(name='block',
target=wait_for_event,
args=(e,))
t1.start()
t2 = threading.Thread(name='non-block',
target=wait_for_event_timeout,
args=(e,2))
t2.start()
logging.debug('Waiting before calling Event.set()')
time.sleep(3)
e.set()
logging.debug('Event is set')
The wait() method takes an argument representing the number of seconds to wait for the event before timing out. It returns a boolean indicating whether or not the event is set, so the caller knows why wait() returned. The isSet() method can be used separately on the event without fear of blocking.
In this example, wait_for_event_timeout() checks the event status without blocking indefinitely. The wait_for_event() blocks on the call to wait(), which does not return until the event status changes.
$ python threading_event.py
(block ) wait_for_event starting
(non-block ) wait_for_event_timeout starting
(MainThread) Waiting before calling Event.set()
(non-block ) event set: False
(non-block ) doing other work
(non-block ) wait_for_event_timeout starting
(MainThread) Event is set
(block ) event set: True
(non-block ) event set: True
(non-block ) processing event
In addition to synchronizing the operations of threads, it is also important to be able to control access to shared resources to prevent corruption or missed data. Python's built-in data structures (lists, dictionaries, etc.) are thread-safe as a side-effect of having atomic byte-codes for manuplating them (the GIL is not released in the middle of an update). Other data structures implemented in Python, or simpler types like integers and floats, don't have that protection. To guard against simultaneous access to an object, use a Lock object.
In [ ]:
# %load threading_lock.py
import logging
import random
import threading
import time
logging.basicConfig(level=logging.DEBUG,
format='(%(threadName)-10s) %(message)s',
)
class Counter(object):
def __init__(self, start=0):
self.lock = threading.Lock()
self.value = start
def increment(self):
logging.debug('Waiting for lock')
self.lock.acquire()
try:
logging.debug('Acquired lock')
self.value = self.value + 1
finally:
self.lock.release()
def worker(c):
for i in range(2):
pause = random.random()
logging.debug('Sleeping %0.02f', pause)
time.sleep(pause)
c.increment()
logging.debug('Done')
counter = Counter()
for i in range(2):
t = threading.Thread(target=worker, args=(counter,))
t.start()
logging.debug('Waiting for worker threads')
main_thread = threading.currentThread()
for t in threading.enumerate():
if t is not main_thread:
t.join()
logging.debug('Counter: %d', counter.value)
In this example, the worker() function increments a Counter instance, which manages a Lock to prevent two threads from changing its internal state at the same time. If the Lock was not used, there is a possibility of missing a change to the value attribute.
$ python threading_lock.py
(Thread-1 ) Sleeping 0.63
(Thread-2 ) Sleeping 0.37
(MainThread) Waiting for worker threads
(Thread-2 ) Waiting for lock
(Thread-2 ) Acquired lock
(Thread-2 ) Sleeping 0.44
(Thread-1 ) Waiting for lock
(Thread-1 ) Acquired lock
(Thread-1 ) Sleeping 0.24
(Thread-2 ) Waiting for lock
(Thread-2 ) Acquired lock
(Thread-2 ) Done
(Thread-1 ) Waiting for lock
(Thread-1 ) Acquired lock
(Thread-1 ) Done
(MainThread) Counter: 4
To find out whether another thread has acquired the lock without holding up the current thread, pass False for the blocking argument to acquire(). In the next example, worker() tries to make to do so. In the mean time, lock_holder() cycles between holding and releasing the lock, with short pauses in each state used to simulate load.
In [ ]:
# %load threading_lock_noblock.py
import logging
import threading
import time
logging.basicConfig(level=logging.DEBUG,
format='(%(threadName)-10s) %(message)s',
)
def lock_holder(lock):
logging.debug('Starting')
while True:
lock.acquire()
try:
logging.debug('Holding')
time.sleep(0.5)
finally:
logging.debug('Not holding')
lock.release()
time.sleep(0.5)
return
def worker(lock):
logging.debug('Starting')
num_tries = 0
num_acquires = 0
while num_acquires < 3:
time.sleep(1)
logging.debug('Trying to acquire')
have_it = lock.acquire(0)
try:
num_tries += 1
if have_it:
logging.debug('Iteration %d: Acquired', num_tries)
num_acquires += 1
else:
logging.debug('Iteration %d: Not acquired', num_tries)
finally:
if have_it:
lock.release()
logging.debug('Done after %d iterations', num_tries)
lock =threading.Lock()
holder = threading.Thread(target=lock_holder, args=(lock,), name='LockHolder')
holder.setDaemon(True)
holder.start()
worker = threading.Thread(target=worker, args=(lock,), name='Worker')
worker.start()
It takes worker() more than three iterations to acquire the lock three separate times.
$ python threading_lock_noblock.py
(LockHolder) Starting
(Worker ) Starting
(LockHolder) Holding
(LockHolder) Not holding
(Worker ) Trying to acquire
(Worker ) Iteration 1: Acquired
(LockHolder) Holding
(LockHolder) Not holding
(Worker ) Trying to acquire
(LockHolder) Holding
(Worker ) Iteration 2: Not acquired
(LockHolder) Not holding
(Worker ) Trying to acquire
(Worker ) Iteration 3: Acquired
(LockHolder) Holding
(LockHolder) Not holding
(Worker ) Trying to acquire
(LockHolder) Holding
(Worker ) Iteration 4: Not acquired
(LockHolder) Not holding
(Worker ) Trying to acquire
(Worker ) Iteration 5: Acquired
(Worker ) Done after 5 iterations
(LockHolder) Holding
In [ ]:
# %load threading_lock_reacquire.py
import threading
lock = threading.Lock()
print ('First try :', lock.acquire())
print ('Second try:', lock.acquire(0))
In this case, since both functions are using the same global lock, and one calls the other, the second acquisition fails and would have blocked using the default arguments to acquire().
$ python threading_lock_reacquire.py
First try : True
Second try: False
In a situation where separate code from the same thread needs to "re-acquire" the lock, use an RLock instead.
In [1]:
# %load threading_rlock.py
import threading
lock = threading.RLock()
print ('First try :', lock.acquire())
print ('Second try:', lock.acquire(0))
The only change to the code from the previous example was substituting RLock for Lock.
$ python threading_rlock.py
First try : True
Second try: 1
Locks implement the context mananger API and are compatible with with statement. Using with removes the need to explicitly acquire and release the lock.
In [2]:
# %load threading_lock_with.py
import threading
import logging
import time
logging.basicConfig(level=logging.DEBUG,
format='(%(threadName)-10s) %(message)s',
)
def worker_with(lock):
with lock:
logging.debug('Lock acquired via with')
time.sleep(3)
def worker_no_with(lock):
lock.acquire()
try:
logging.debug('Lock acquired directly')
time.sleep(2)
finally:
lock.release()
lock = threading.Lock()
w = threading.Thread(target=worker_with, args=(lock,))
nw = threading.Thread(target=worker_no_with, args=(lock,))
w.start()
nw.start()
The two functions worker_with() and worker_no_with() manager the lock in equivalent ways.
$ python threading_lock_with.py
(Thread-1 ) Lock acquired via with
(Thread-2 ) Lock acquired directly
In addition to using Events, another way of synchronizing threads is through using a Condition object. Because the Condition uses a Lock, it can be tied to a shared resource. This allows threads to wait for the resource to be updated. In this example, the consume() threads wait() for the Condition to be set before continuing. The producer() thread is responsible for setting the condition and notifying the other threads that they can continue.
In [3]:
# %load threading_condition.py
import logging
import threading
import time
logging.basicConfig(level=logging.DEBUG,
format='%(asctime)s (%(threadName)-2s) %(message)s',
)
def consumer(cond):
"""wait for the condition and use the resource"""
logging.debug('Starting consumer thread')
t = threading.currentThread()
with cond:
cond.wait()
logging.debug('Resource is available to consumer')
def producer(cond):
"""set up the resource to be used by the consumer"""
logging.debug('Starting producer thread')
with cond:
logging.debug('Making resource available')
cond.notifyAll()
condition = threading.Condition()
c1 = threading.Thread(name='c1', target=consumer, args=(condition,))
c2 = threading.Thread(name='c2', target=consumer, args=(condition,))
p = threading.Thread(name='p', target=producer, args=(condition,))
c1.start()
time.sleep(2)
c2.start()
time.sleep(2)
p.start()
The threads use with to acquire the lock associated with the Condition. Using the acquire() and release() methods explicitly also works.
$ python threading_condition.py
2015-06-27 17:07:55,552 (c1) Starting consumer thread
2015-06-27 17:07:57,557 (c2) Starting consumer thread
2015-06-27 17:07:59,558 (p ) Starting producer thread
2015-06-27 17:07:59,558 (p ) Making resource available
2015-06-27 17:07:59,559 (c2) Resource is available to consumer
2015-06-27 17:07:59,559 (c1) Resource is available to consumer
Sometimes it is useful to allow more than one worker access to a resource at a time, while still limiting the overall number. For example, a connection pool might support a fixed number of simultaneous connnections, or a network application might support a fixed number of concurrent downloads. A Semaphore is one way to manage those connections.
In [ ]:
# %load threading_semaphore.py
import logging
import random
import threading
import time
logging.basicConfig(level=logging.DEBUG,
format='%(asctime)s (%(threadName)-2s) %(message)s',
)
class ActivePool(object):
def __init__(self):
super(ActivePool, self).__init__()
self.active = []
self.lock = threading.Lock()
def makeActive(self, name):
with self.lock:
self.active.append(name)
logging.debug('Running: %s', self.active)
def makeInactive(self, name):
with self.lock:
self.active.remove(name)
logging.debug('Running: %s', self.active)
def worker(s, pool):
logging.debug('Waiting to join the pool')
with s:
name = threading.currentThread().getName()
pool.makeActive(name)
time.sleep(0.1)
pool.makeInactive(name)
pool = ActivePool()
s = threading.Semaphore(2)
for i in range(4):
t = threading.Thread(target=worker, name=str(i), args=(s, pool))
t.start()
In this example, the ActivePool class simply serves as a convenient way to track which threads are able to run at a given moment. A real resource pool would allocate a connection or some other value to the newly active thread, and reclaim the value when the thread is done. Here it is just used to hold the names of the active threads to show the that only five are running concurrently.
$ python threading_semaphore.py
2015-06-27 17:15:56,441 (0 ) Waiting to join the pool
2015-06-27 17:15:56,442 (0 ) Running: ['0']
2015-06-27 17:15:56,443 (1 ) Waiting to join the pool
2015-06-27 17:15:56,443 (1 ) Running: ['0', '1']
2015-06-27 17:15:56,443 (2 ) Waiting to join the pool
2015-06-27 17:15:56,443 (3 ) Waiting to join the pool
2015-06-27 17:15:56,548 (1 ) Running: ['0']
2015-06-27 17:15:56,548 (0 ) Running: []
2015-06-27 17:15:56,549 (2 ) Running: ['2']
2015-06-27 17:15:56,549 (3 ) Running: ['2', '3']
2015-06-27 17:15:56,654 (2 ) Running: ['3']
2015-06-27 17:15:56,655 (3 ) Running: []
In [ ]:
# %load threading_local.py
import random
import threading
import logging
logging.basicConfig(level=logging.DEBUG,
format='(%(threadName)-10s) %(message)s',
)
def show_value(data):
try:
val = data.value
except AttributeError:
logging.debug('No value yet')
else:
logging.debug('value=%s', val)
def worker(data):
show_value(data)
data.value = random.randint(1, 100)
show_value(data)
local_data = threading.local()
show_value(local_data)
local_data.value = 1000
show_value(local_data)
for i in range(2):
t = threading.Thread(target=worker, args=(local_data,))
t.start()
Notice that local_data.value is not present for any thread until it is set in that thread.
$ python threading_local.py
(MainThread) No value yet
(MainThread) value=1000
(Thread-1 ) No value yet
(Thread-1 ) value=60
(Thread-2 ) No value yet
(Thread-2 ) value=58
To initialize the settings so all threads start with the same value, use a subclass and set the attributes in __init__().
In [ ]:
# %load threading_local_defaults.py
import random
import threading
import logging
logging.basicConfig(level=logging.DEBUG,
format='(%(threadName)-10s) %(message)s',
)
def show_value(data):
try:
val = data.value
except AttributeError:
logging.debug('No value yet')
else:
logging.debug('value=%s', val)
def worker(data):
show_value(data)
data.value = random.randint(1,100)
show_value(data)
class MyLocal(threading.local):
def __init__(self, value):
logging.debug('Initializing %r', self)
self.value = value
local_data = MyLocal(1000)
show_value(local_data)
for i in range(2):
t = threading.Thread(target=worker, args=(local_data,))
t.start()
show_value(local_data)
__init__() is invoked on the same object (note the id() value), once in each thread.
$ python threading_local_defaults.py
(MainThread) Initializing <__main__.MyLocal object at 0x10ae0de20>
(MainThread) value=1000
(Thread-1 ) Initializing <__main__.MyLocal object at 0x10ae0de20>
(Thread-1 ) value=1000
(Thread-1 ) value=58
(Thread-2 ) Initializing <__main__.MyLocal object at 0x10ae0de20>
(MainThread) value=1000
(Thread-2 ) value=1000
(Thread-2 ) value=77
Links :
1. https://pymotw.com/2/threading/
2. https://www.ibm.com/developerworks/aix/library/au-threadingpython/