A persistent process pool in Python for use with Twisted. Provides the ability to run Python callables asynchronously within a pool of persistent processes, as long as the callables, their arguments, and their return values are all picklable.
pip install txpool
Here are some simple examples to give you the idea:
import glob
from twisted.internet import reactor
from twisted.internet.defer import inlineCallbacks
import txpool
pool = txpool.Pool()
@inlineCallbacks
def main():
result = yield pool.apply_async(glob.glob, ('*.pdf',))
print result
reactor.stop()
reactor.callWhenRunning(main)
reactor.run()
The callable can instead be specified as a string, using dotted notation to specify the full path to the callable.
from twisted.internet import reactor
from twisted.internet.defer import inlineCallbacks
import txpool
pool = txpool.Pool()
@inlineCallbacks
def main():
# You can provide an optional timeout (in seconds) for the call
# (the default is None).
try:
result = yield pool.apply_async('glob.glob', ('*.pdf',), timeout=5)
except txpool.PoolTimeout as e:
result = e
print result
reactor.stop()
reactor.callWhenRunning(main)
reactor.run()
The txpool.Pool class can be explicitly sized, asked to log its actions, and/or given a custom name.
import logging
from twisted.internet import reactor
from twisted.internet.defer import inlineCallbacks, gatherResults
import txpool
logger = logging.getLogger('example')
logger.addHandler(logging.StreamHandler())
logger.setLevel(logging.DEBUG)
pool = txpool.Pool(size=5, log=logger, name='twisting-by-the-pool')
@inlineCallbacks
def main():
calls = ('math.factorial',) * 5
args = [(n,) for n in range(150780, 150785)]
# You can wait until the pool is at full-strength (providing an
# optional timeout if desired), but it's not required before
# calling the "apply_async" method. Jobs are queued until a
# worker process is available.
try:
yield pool.on_ready(timeout=10)
except txpool.PoolTimeout as e:
results = e
else:
results = yield gatherResults(map(pool.apply_async, calls, args))
print results
try:
# You can gracefully close the pool, which ensures all jobs
# already queued are completed before shutting down...
yield pool.close(timeout=10)
except txpool.PoolTimeout as e:
print e
# ...or you can use force and immediately send SIGKILL to each
# process in the pool.
yield pool.terminate(timeout=10)
reactor.stop()
reactor.callWhenRunning(main)
reactor.run()