"""Implementation of a thread-pool working with channels""" from thread import WorkerThread from threading import ( Lock, _Condition, _sleep, _time, ) from task import InputChannelTask from Queue import Queue, Empty from collections import deque from graph import ( Graph, ) from channel import ( Channel, WChannel, RChannel ) import weakref import sys #{ Utilities class SyncQueue(deque): """Adapter to allow using a deque like a queue, without locking""" def get(self, block=True, timeout=None): try: return self.pop() except IndexError: raise Empty # END raise empty def empty(self): return len(self) == 0 put = deque.append class HSCondition(_Condition): """An attempt to make conditions less blocking, which gains performance in return by sleeping less""" delay = 0.00002 # reduces wait times, but increases overhead def wait(self, timeout=None): waiter = Lock() waiter.acquire() self.__dict__['_Condition__waiters'].append(waiter) saved_state = self._release_save() try: # restore state no matter what (e.g., KeyboardInterrupt) if timeout is None: waiter.acquire() else: # Balancing act: We can't afford a pure busy loop, so we # have to sleep; but if we sleep the whole timeout time, # we'll be unresponsive. The scheme here sleeps very # little at first, longer as time goes on, but never longer # than 20 times per second (or the timeout time remaining). endtime = _time() + timeout delay = self.delay acquire = waiter.acquire while True: gotit = acquire(0) if gotit: break remaining = endtime - _time() if remaining <= 0: break delay = min(delay * 2, remaining, .05) _sleep(delay) # END endless loop if not gotit: try: self.__dict__['_Condition__waiters'].remove(waiter) except ValueError: pass # END didn't ever get it finally: self._acquire_restore(saved_state) def notify(self, n=1): __waiters = self.__dict__['_Condition__waiters'] if not __waiters: return if n == 1: __waiters[0].release() try: __waiters.pop(0) except IndexError: pass else: waiters = __waiters[:n] for waiter in waiters: waiter.release() try: __waiters.remove(waiter) except ValueError: pass # END handle n = 1 case faster class PerfQueue(Queue): """A queue using different condition objects to gain multithreading performance""" def __init__(self, maxsize=0): Queue.__init__(self, maxsize) self.not_empty = HSCondition(self.mutex) self.not_full = HSCondition(self.mutex) self.all_tasks_done = HSCondition(self.mutex) #} END utilities class RPoolChannel(RChannel): """ A read-only pool channel may not be wrapped or derived from, but it provides slots to call before and after an item is to be read. It acts like a handle to the underlying task in the pool.""" __slots__ = ('_task', '_pool', '_pre_cb', '_post_cb') def __init__(self, wchannel, task, pool): RChannel.__init__(self, wchannel) self._task = task self._pool = pool self._pre_cb = None self._post_cb = None def __del__(self): """Assures that our task will be deleted if we were the last reader""" del(self._wc) # decrement ref-count self._pool._del_task_if_orphaned(self._task) def set_pre_cb(self, fun = lambda count: None): """Install a callback to call with the item count to be read before any item is actually read from the channel. If it fails, the read will fail with an IOError If a function is not provided, the call is effectively uninstalled.""" self._pre_cb = fun def set_post_cb(self, fun = lambda item: item): """Install a callback to call after the items were read. The function returns a possibly changed item list. If it raises, the exception will be propagated. If a function is not provided, the call is effectively uninstalled.""" self._post_cb = fun def read(self, count=0, block=True, timeout=None): """Read an item that was processed by one of our threads :note: Triggers task dependency handling needed to provide the necessary input""" if self._pre_cb: self._pre_cb() # END pre callback # if we have count items, don't do any queue preparation - if someone # depletes the queue in the meanwhile, the channel will close and # we will unblock naturally have_enough = False if count > 0: # explicitly > count, as we want a certain safe range have_enough = self._wc._queue.qsize() > count # END risky game ########## prepare ############################## if not have_enough: self._pool._prepare_channel_read(self._task, count) ######### read data ###### # read actual items, tasks were setup to put their output into our channel ( as well ) items = RChannel.read(self, count, block, timeout) if self._post_cb: items = self._post_cb(items) ####### Finalize ######## self._pool._post_channel_read(self._task) return items #{ Internal def _read(self, count=0, block=False, timeout=None): """Calls the underlying channel's read directly, without triggering the pool""" return RChannel.read(self, count, block, timeout) #} END internal class Pool(object): """A thread pool maintains a set of one or more worker threads, but supports a fully serial mode in which case the amount of threads is zero. Work is distributed via Channels, which form a dependency graph. The evaluation is lazy, as work will only be done once an output is requested. The thread pools inherent issue is the global interpreter lock that it will hit, which gets worse considering a few c extensions specifically lock their part globally as well. The only way this will improve is if custom c extensions are written which do some bulk work, but release the GIL once they have acquired their resources. Due to the nature of having multiple objects in git, its easy to distribute that work cleanly among threads. :note: the current implementation returns channels which are meant to be used only from the main thread, hence you cannot consume their results from multiple threads unless you use a task for it.""" __slots__ = ( '_tasks', # a graph of tasks '_consumed_tasks', # a queue with tasks that are done or had an error '_workers', # list of worker threads '_queue', # master queue for tasks '_taskorder_cache', # map task id -> ordered dependent tasks '_taskgraph_lock', # lock for accessing the task graph ) # CONFIGURATION # The type of worker to create - its expected to provide the Thread interface, # taking the taskqueue as only init argument # as well as a method called stop_and_join() to terminate it WorkerCls = None # The type of lock to use to protect critical sections, providing the # threading.Lock interface LockCls = None # the type of the task queue to use - it must provide the Queue interface TaskQueueCls = None def __init__(self, size=0): self._tasks = Graph() self._consumed_tasks = None self._workers = list() self._queue = SyncQueue() # start with a sync queue self._taskgraph_lock = self.LockCls() self._taskorder_cache = dict() self.set_size(size) def __del__(self): self.set_size(0) #{ Internal def _prepare_channel_read(self, task, count): """Process the tasks which depend on the given one to be sure the input channels are filled with data once we process the actual task Tasks have two important states: either they are done, or they are done and have an error, so they are likely not to have finished all their work. Either way, we will put them onto a list of tasks to delete them, providng information about the failed ones. Tasks which are not done will be put onto the queue for processing, which is fine as we walked them depth-first.""" # for the walk, we must make sure the ordering does not change. Even # when accessing the cache, as it is related to graph changes self._taskgraph_lock.acquire() try: try: dfirst_tasks = self._taskorder_cache[id(task)] except KeyError: # have to retrieve the list from the graph dfirst_tasks = list() self._tasks.visit_input_inclusive_depth_first(task, lambda n: dfirst_tasks.append(n)) self._taskorder_cache[id(task)] = dfirst_tasks # END handle cached order retrieval finally: self._taskgraph_lock.release() # END handle locking # check the min count on all involved tasks, and be sure that we don't # have any task which produces less than the maximum min-count of all tasks # The actual_count is used when chunking tasks up for the queue, whereas # the count is usued to determine whether we still have enough output # on the queue, checking qsize ( ->revise ) # ABTRACT: If T depends on T-1, and the client wants 1 item, T produces # at least 10, T-1 goes with 1, then T will block after 1 item, which # is read by the client. On the next read of 1 item, we would find T's # queue empty and put in another 10, which could put another thread into # blocking state. T-1 produces one more item, which is consumed right away # by the two threads running T. Although this works in the end, it leaves # many threads blocking and waiting for input, which is not desired. # Setting the min-count to the max of the mincount of all tasks assures # we have enough items for all. # Addition: in serial mode, we would enter a deadlock if one task would # ever wait for items ! actual_count = count min_counts = (((t.min_count is not None and t.min_count) or count) for t in dfirst_tasks) min_count = reduce(lambda m1, m2: max(m1, m2), min_counts) if 0 < count < min_count: actual_count = min_count # END set actual count # the list includes our tasks - the first one to evaluate first, the # requested one last for task in dfirst_tasks: if task.error() or task.is_done(): self._consumed_tasks.put(task) continue # END skip processing # if the task does not have the required output on its queue, schedule # it for processing. If we should process all, we don't care about the # amount as it should process until its all done. #if count > 1 and task._out_wc.size() >= count: # continue # END skip if we have enough # but use the actual count to produce the output, we may produce # more than requested numchunks = 1 chunksize = actual_count remainder = 0 # we need the count set for this - can't chunk up unlimited items # In serial mode we could do this by checking for empty input channels, # but in dispatch mode its impossible ( == not easily possible ) # Only try it if we have enough demand if task.max_chunksize and actual_count > task.max_chunksize: numchunks = actual_count / task.max_chunksize chunksize = task.max_chunksize remainder = actual_count - (numchunks * chunksize) # END handle chunking # the following loops are kind of unrolled - code duplication # should make things execute faster. Putting the if statements # into the loop would be less code, but ... slower # DEBUG # print actual_count, numchunks, chunksize, remainder, task._out_wc.size() if self._workers: # respect the chunk size, and split the task up if we want # to process too much. This can be defined per task queue = self._queue if numchunks > 1: for i in xrange(numchunks): queue.put((task.process, chunksize)) # END for each chunk to put else: queue.put((task.process, chunksize)) # END try efficient looping if remainder: queue.put((task.process, remainder)) # END handle chunksize else: # no workers, so we have to do the work ourselves if numchunks > 1: for i in xrange(numchunks): task.process(chunksize) # END for each chunk to put else: task.process(chunksize) # END try efficient looping if remainder: task.process(remainder) # END handle chunksize # END handle serial mode # END for each task to process def _post_channel_read(self, task): """Called after we processed a read to cleanup""" # check whether we consumed the task, and schedule it for deletion # This could have happend after the read returned ( even though the pre-read # checks it as well ) if task.error() or task.is_done(): self._consumed_tasks.put(task) # END handle consumption # delete consumed tasks to cleanup try: while True: ct = self._consumed_tasks.get(False) self.del_task(ct) # END for each task to delete except Empty: pass # END pop queue empty def _del_task_if_orphaned(self, task): """Check the task, and delete it if it is orphaned""" if sys.getrefcount(task._out_wc) < 3: self.del_task(task) #} END internal #{ Interface def size(self): """:return: amount of workers in the pool""" return len(self._workers) def set_size(self, size=0): """Set the amount of workers to use in this pool. When reducing the size, the call may block as it waits for threads to finish. When reducing the size to zero, this thread will process all remaining items on the queue. :return: self :param size: if 0, the pool will do all work itself in the calling thread, otherwise the work will be distributed among the given amount of threads :note: currently NOT threadsafe !""" assert size > -1, "Size cannot be negative" # either start new threads, or kill existing ones. # If we end up with no threads, we process the remaining chunks on the queue # ourselves cur_count = len(self._workers) if cur_count < size: # make sure we have a real queue, and can store our consumed tasks properly if not isinstance(self._queue, self.TaskQueueCls): if self._queue is not None and not self._queue.empty(): raise AssertionError("Expected empty queue when switching the queue type") # END safety check self._queue = self.TaskQueueCls() self._consumed_tasks = Queue() # END init queue for i in range(size - cur_count): worker = self.WorkerCls(self._queue) worker.start() self._workers.append(worker) # END for each new worker to create elif cur_count > size: del_count = cur_count - size for i in range(del_count): self._workers[i].stop_and_join() # END for each thread to stop del(self._workers[:del_count]) # END handle count if size == 0: while not self._queue.empty(): try: taskmethod, count = self._queue.get(False) taskmethod(count) except Queue.Empty: continue # END while there are tasks on the queue # use a serial queue, its faster if not isinstance(self._queue, SyncQueue): self._queue = SyncQueue() # END handle queue type if self._consumed_tasks and not self._consumed_tasks.empty(): self._post_channel_read(self._consumed_tasks.pop()) # END assure consumed tasks are empty self._consumed_tasks = SyncQueue() # END process queue return self def num_tasks(self): """:return: amount of tasks""" return len(self._tasks.nodes) def del_task(self, task): """Delete the task Additionally we will remove orphaned tasks, which can be identified if their output channel is only held by themselves, so no one will ever consume its items. :return: self""" # now delete our actual node - must set it done os it closes its channels. # Otherwise further reads of output tasks will block. # Actually they may still block if anyone wants to read all ... without # a timeout # keep its input nodes as we check whether they were orphaned in_tasks = task.in_nodes task.set_done() self._taskgraph_lock.acquire() try: self._taskorder_cache.clear() self._tasks.del_node(task) finally: self._taskgraph_lock.release() # END locked deletion for t in in_tasks: self._del_task_if_orphaned(t) # END handle orphans recursively return self def add_task(self, task): """Add a new task to be processed. :return: a read channel to retrieve processed items. If that handle is lost, the task will be considered orphaned and will be deleted on the next occasion.""" # create a write channel for it wc, rc = Channel() rc = RPoolChannel(wc, task, self) task._out_wc = wc has_input_channel = isinstance(task, InputChannelTask) if has_input_channel: task._pool_ref = weakref.ref(self) # END init input channel task self._taskgraph_lock.acquire() try: self._taskorder_cache.clear() self._tasks.add_node(task) finally: self._taskgraph_lock.release() # END sync task addition # If the input channel is one of our read channels, we add the relation if has_input_channel: ic = task.in_rc if isinstance(ic, RPoolChannel) and ic._pool is self: self._taskgraph_lock.acquire() try: self._tasks.add_edge(ic._task, task) finally: self._taskgraph_lock.release() # END handle edge-adding # END add task relation # END handle input channels for connections return rc #} END interface class ThreadPool(Pool): """A pool using threads as worker""" WorkerCls = WorkerThread LockCls = Lock TaskQueueCls = PerfQueue