In commit b3b4e39, the 'disque' directory was removed but I think the 'pydisque' directory was not git add'ed.

According to the disque docs the working command should be sent to the Node which holds the Job

I am more then happy to tackle this one myself, but I noticed quite a few hiccups when I was trying to use this with a Python 3.2 project.

Leaving it as an "issue" for now, and I will try to resolve it "the right way," before this module gets too popular :)

Can you add a brief comment in README or add LICENSE file?

It would be easier to see that the license is MIT

Thanks.

It is required at times to keep the disque server password protected. Is there a way or shouldn't there be one to pass password as parameter while listening to the queue through disque client?

SUMMARY: Just a warning for other people who may be interested in using this library with Python 3.2!

I found a few more bugs in Python 3.2 execution while writing some new unit tests for new Disque commands.

Mostly, I was brutally sloppy in implementing the 'six' library, but as learning some more complete Py2->Py3 transitional code is something of interest to me, I'm more then happy to take the task upon myself to fix it.

Right now client.connect() does not take any redis client connection paramaters. We should be able to pass all the redis config parameters like tcp sock connect time out etc..

Heya ybrs,

What do you think about incrementing the version number and doing a release to PyPI?

I started breaking out some milestones a few weeks ago (https://github.com/ybrs/pydisque/milestones) and I think we've hit what I originally put into 0.1.2. I'm always up for more + feedback.

Is there a plan to make a more "humane" interface?

By this, for example, I mean returning a job object from get_job:

jobs = c.get_job(['q'])
job = jobs[0]
print job.payload
job.ack()

Other niceties would be returning a dictionary from qstat. It seems "info" returns a dictionary, but I think the redis library does that.

python3

from pydisque.client import Client
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python3.4/dist-packages/pydisque/client.py", line 6, in <module>
    from itertools import izip_longest
ImportError: cannot import name 'izip_longest'

If anyone else can get to this before me... Got a lot going on out here.

...

PAUSE option1 [option2 ... optionN]

Control the paused state of a queue, possibly broadcasting the command to other nodes in the cluster. Disque queues can be paused in both directions, input and output, or both. Pausing a queue makes it not available for input or output operations. Specifically:

A queue paused in input will have change behavior in the following way:

ADDJOB returns a -PAUSED error for queues paused in input.
The node where the queue is paused, no longer accept to replicate jobs for this queue when requested by other nodes. Since ADDJOB by default uses synchronous replication, it means that if the queue is paused in enough nodes, adding jobs with a specified level of replication may fail. In general the node where the queue is paused will not create new jobs in the local node about this queue.
The job no longer accepts ENQUEUE messages from other nodes. Those messages are usually used by nodes in out of memory conditions that replicate jobs externally (not holding a copy), in order to put the job in the queue of some random node, among the nodes having a copy of a job.
Active jobs that reach their retry time, are not put back into the queue. Instead their retry timer is updated and the node will try again later.

Basically a queue paused in input never creates new jobs for this queue, and never puts active jobs (jobs for which the node have a copy but are not currently queued) back in the queue, for all the time the queue is paused.

A queue paused in output instead will behave in the following way:

GETJOB will block even if there are jobs available in the specified queue, instead of serving the jobs. But GETJOB will unblock if the queue output pause will be cleared later.
The node will not provide jobs to other nodes in the context of nodes federation, for paused queues.

So a queue paused in output will stop to act as a source of messages for both local and non local clients.

The paused state can be set for each queue using the PAUSE command followed by options to specify how to change the paused state. Possible options are:

in: pause the queue in input.
out: pause the queue in output.
all: pause the queue in input and output (same as specifying both the in and out options).
none: clear the paused state in input and output.
state: just report the current queue state.
bcast: send a PAUSE command to all the reachable nodes of the cluster to set the same queue in the other nodes to the same state.

The command always returns the state of the queue after the execution of the specified options, so the return value is one of in, out, all, none.

Queues paused in input or output are never evicted to reclaim memory, even if they are empty and inactive for a long time, since otherwise the paused state would be forgot.

For example in order to block output for the queue myqueue in all the nodes currently reachable, the following command should be send to a single node:

PAUSE myqueue out bcast

To specify all is the same as to specify both in and out, so the two following forms are equivalent:

PAUSE myqueue in out
PAUSE myqueue all

To just get the current state use:

PAUSE myqueue state
"none"

async is a reserved word in python 3.7.

This breaks:

def add_job(self, queue_name, job, timeout=200, replicate=None, delay=None,
retry=None, ttl=None, maxlen=None, async=None):