469,306 Members | 1,987 Online
Bytes | Developer Community
New Post

Home Posts Topics Members FAQ

Post your question to a community of 469,306 developers. It's quick & easy.

use of Queue

how is Queue intended to be used? I found the following code in python
manual, but I don't understand how to stop consumers after all items
have been produced. I tried different approaches but all of them
seemed incorrect (race, deadlock or duplicating queue functionality)
def worker():
while True:
item = q.get()
do_work(item)
q.task_done()

q = Queue()
for i in range(num_worker_threads):
t = Thread(target=worker)
t.setDaemon(True)
t.start()

for item in source():
q.put(item)

q.join() # block until all tasks are done
Aug 27 '08 #1
13 1634
Alexandru Mosoi wrote:
how is Queue intended to be used? I found the following code in python
manual, but I don't understand how to stop consumers after all items
have been produced. I tried different approaches but all of them
seemed incorrect (race, deadlock or duplicating queue functionality)
def worker():
while True:
item = q.get()
do_work(item)
q.task_done()

q = Queue()
for i in range(num_worker_threads):
t = Thread(target=worker)
t.setDaemon(True)
t.start()

for item in source():
q.put(item)

q.join() # block until all tasks are done
Put a sentinel into the queue that gets interpreted as "terminate" for the
workers. You need of course to put it in there once for each worker.

Diez
Aug 27 '08 #2
Alexandru Mosoi wrote:
how is Queue intended to be used? I found the following code in python
manual, but I don't understand how to stop consumers after all items
have been produced. I tried different approaches but all of them
seemed incorrect (race, deadlock or duplicating queue functionality)
def worker():
while True:
item = q.get()
if item is None:
break
do_work(item)
q.task_done()

q = Queue()
for i in range(num_worker_threads):
t = Thread(target=worker)
t.setDaemon(True)
t.start()

for item in source():
q.put(item)
# stop all consumers
for i in range(num_worker_threads):
q.put(None)
>
q.join() # block until all tasks are done
This is how I do it.

-- Gerhard

Aug 27 '08 #3

DiezPut a sentinel into the queue that gets interpreted as "terminate"
Diezfor the workers. You need of course to put it in there once for
Diezeach worker.

Or make the consumers daemon threads so that when the producers are finished
an all non-daemon threads exit, the consumers do as well.

Skip
Aug 27 '08 #4

skipOr make the consumers daemon threads so that when the producers
skipare finished an all non-daemon threads exit, the consumers do as
skipwell.

Forget that I wrote this. If they happen to be working on the token they've
consumed at the time the other threads exit, they will as well. Use the
sentinel token idea instead.

Skip
Aug 27 '08 #5
On Aug 27, 1:06*pm, Gerhard Häring <g...@ghaering.dewrote:
Alexandru Mosoi wrote:
how is Queue intended to be used? I found the following code in python
manual, but I don't understand how to stop consumers after all items
have been produced. I tried different approaches but all of them
seemed incorrect (race, deadlock or duplicating queue functionality)
* * def worker():
* * * * while True:
* * * * * * item = q.get()

* * * * * * * *if item is None:
* * * * * * * * * *break
* * * * * * do_work(item)
* * * * * * q.task_done()
* * q = Queue()
* * for i in range(num_worker_threads):
* * * * *t = Thread(target=worker)
* * * * *t.setDaemon(True)
* * * * *t.start()
* * for item in source():
* * * * q.put(item)

# stop all consumers
for i in range(num_worker_threads):
* * *q.put(None)
* * q.join() * * * # block until all tasks are done

This is how I do it.

-- Gerhard

Your solution works assuming that you know how many consumer threads
you have :). I don't :). More than that, it's not correct if you have
more than one producer :). Having a sentinel was my very first idea,
but as you see... it's a race condition (there are cases in which not
all items are processed).
Aug 27 '08 #6
Your solution works assuming that you know how many consumer threads
you have :). I don't :). More than that, it's not correct if you have
more than one producer :). Having a sentinel was my very first idea,
but as you see... it's a race condition (there are cases in which not
all items are processed).
Queue raises an Empty exception when there are no items left in the
queue. Put the q.get() call in a try block and exit in the except
block.

You can also use a condition variable to signal threads to terminate.
Aug 27 '08 #7
On Aug 27, 2:54*pm, Jeff <jeffo...@gmail.comwrote:
Queue raises an Empty exception when there are no items left in the
queue. *Put the q.get() call in a try block and exit in the except
block.
Wrong. What if producer takes a long time to produce an item?
Consumers
will find the queue empty and exit instead of waiting.
You can also use a condition variable to signal threads to terminate.
This is the solution I want to avoid because it duplicates Queue's
functionality.
I prefer having a clean solution with nice design to hacking Queue
class.
Aug 27 '08 #8
On Aug 27, 12:45*pm, Alexandru Mosoi <brtz...@gmail.comwrote:
how is Queue intended to be used? I found the following code in python
manual, but I don't understand how to stop consumers after all items
have been produced. I tried different approaches but all of them
seemed incorrect (race, deadlock or duplicating queue functionality)

* * def worker():
* * * * while True:
* * * * * * item = q.get()
* * * * * * do_work(item)
* * * * * * q.task_done()

* * q = Queue()
* * for i in range(num_worker_threads):
* * * * *t = Thread(target=worker)
* * * * *t.setDaemon(True)
* * * * *t.start()

* * for item in source():
* * * * q.put(item)

* * q.join() * * * # block until all tasks are done

ok. I think I figured it out :). let me know what you think

global num_tasks, num_done, queue
num_tasks = 0
num_done = 0
queue = Queue()

# producer
num_tasks += 1
for i in items:
num_tasks += 1
queue.put(i)

num_tasks -= 1
if num_tasks == num_done:
queue.put(None)

# consumer
while True:
i = queue.get()
if i is None:
queue.put(None)
break

# do stuff

num_done += 1
if num_done == num_tasks:
queue.put(None)
break

Aug 27 '08 #9
>
Your solution works assuming that you know how many consumer threads
you have :). I don't :). More than that, it's not correct if you have
more than one producer :). Having a sentinel was my very first idea,
but as you see... it's a race condition (there are cases in which not
all items are processed).
If you have several producers, how do you coordinate when to shut down?

Apart from that, you can easily solve the problem of not knowing how many
consumers you have by making a consumer stuff back the sentinel into the
queue. Then it will ripple down until no consumer is left.

Diez
Aug 27 '08 #10
On Aug 27, 1:17 pm, Alexandru Mosoi <brtz...@gmail.comwrote:
On Aug 27, 12:45 pm, Alexandru Mosoi <brtz...@gmail.comwrote:
how is Queue intended to be used? I found the following code in python
manual, but I don't understand how to stop consumers after all items
have been produced. I tried different approaches but all of them
seemed incorrect (race, deadlock or duplicating queue functionality)
def worker():
while True:
item = q.get()
do_work(item)
q.task_done()
q = Queue()
for i in range(num_worker_threads):
t = Thread(target=worker)
t.setDaemon(True)
t.start()
for item in source():
q.put(item)
q.join() # block until all tasks are done

ok. I think I figured it out :). let me know what you think

global num_tasks, num_done, queue
num_tasks = 0
num_done = 0
queue = Queue()

# producer
num_tasks += 1
for i in items:
num_tasks += 1
queue.put(i)

num_tasks -= 1
if num_tasks == num_done:
queue.put(None)

# consumer
while True:
i = queue.get()
if i is None:
queue.put(None)
break

# do stuff

num_done += 1
if num_done == num_tasks:
queue.put(None)
break
Are you sure you want to put the final exit code in the consumer?
Shouldn't the producer place a None on the queue when it knows it's
finished? The way you have it, the producer could make 1 item, it
could get consumed, and the consumer exit before the producer makes
item 2.

Iain
Aug 27 '08 #11
Alexandru Mosoi wrote:
>how is Queue intended to be used? I found the following code in python
manual, but I don't understand how to stop consumers after all items
have been produced. I tried different approaches but all of them
seemed incorrect (race, deadlock or duplicating queue functionality)

def worker():
while True:
item = q.get()
do_work(item)
q.task_done()

q = Queue()
for i in range(num_worker_threads):
t = Thread(target=worker)
t.setDaemon(True)
t.start()

for item in source():
q.put(item)

q.join() # block until all tasks are done


ok. I think I figured it out :). let me know what you think

global num_tasks, num_done, queue
num_tasks = 0
num_done = 0
queue = Queue()

# producer
num_tasks += 1
for i in items:
num_tasks += 1
queue.put(i)
what's the point of using a thread-safe queue if you're going to use a
non-thread-safe counter? if you want to write broken code, you can do
that in a lot fewer lines ;-)

as others have mentioned, you can use sentinels:

http://effbot.org/librarybook/queue.htm

or, in Python 2.5 and later, the task_done/join pattern shown here:

http://docs.python.org/lib/QueueObjects.html

</F>

Aug 27 '08 #12
sk**@pobox.com writes:
Or make the consumers daemon threads so that when the producers are finished
an all non-daemon threads exit, the consumers do as well.
How are the consumers supposed to know when the producers are
finished? Yes, there are different approaches like sentinels, but the
absence of a unified approach built into the library really does seem
like a deficiency in the library.
Aug 27 '08 #13
On Aug 27, 4:55*pm, Paul Rubin <http://phr...@NOSPAM.invalidwrote:
s...@pobox.com writes:
Or make the consumers daemon threads so that when the producers are finished
an all non-daemon threads exit, the consumers do as well.

How are the consumers supposed to know when the producers are
finished? *Yes, there are different approaches like sentinels, but the
absence of a unified approach built into the library really does seem
like a deficiency in the library.
See effbot's reply. The task_done and join methods were put there for
exactly this use case.
Raymond
Aug 28 '08 #14

This discussion thread is closed

Replies have been disabled for this discussion.

Similar topics

9 posts views Thread by phil | last post: by
9 posts views Thread by Brian Henry | last post: by
4 posts views Thread by alisaee | last post: by
Rooro
5 posts views Thread by Rooro | last post: by
2 posts views Thread by lavender | last post: by
3 posts views Thread by jrpfinch | last post: by
4 posts views Thread by j_depp_99 | last post: by
reply views Thread by ecestd | last post: by
reply views Thread by zhoujie | last post: by
reply views Thread by suresh191 | last post: by
1 post views Thread by Geralt96 | last post: by
reply views Thread by harlem98 | last post: by
By using this site, you agree to our Privacy Policy and Terms of Use.