def __init__(self, adjacency_map): """ TopologicalSorter constructor. @param adjacency_map: dict form adjacency map @type adjacency_map: dict """ object.__init__(self) self.__adjacency_map = adjacency_map self.__stack = Lifo()
def __init__(self, uri): """ EntropyTransceiver constructor, just pass the friggin URI(tm). @param uri: URI to handle @type uri: string """ self._uri = uri self._speed_limit = 0 self._verbose = False self._timeout = None self._silent = None self._output_interface = None self.__with_stack = Lifo()
def __topological_sort(self, graph): """ Effectively executes topological sorting on given graph. """ # initialize count map count = dict((node, 0) for node in graph) for node in graph: for successor in graph[node]: count[successor] += 1 ready_stack = Lifo() for node in graph: if count[node] == 0: ready_stack.push(node) dep_level = 1 result = {} while ready_stack.is_filled(): node = ready_stack.pop() result[dep_level] = node dep_level += 1 for successor in graph[node]: count[successor] -= 1 if count[successor] == 0: ready_stack.push(successor) return result
def init_singleton(self): """ Singleton overloaded method. Equals to __init__. This is the place where all the properties initialization takes place. """ self.__copy = copy self.__alive = False self.__cache_writer = None self.__cache_buffer = Lifo() self.__stashing_cache = {} self.__inside_with_stmt = 0 self.__dump_data_lock = threading.Lock() self.__worker_sem = threading.Semaphore(0) # this lock ensures that all the writes are hold while it's acquired self.__enter_context_lock = threading.RLock()
def _graph_to_stdout(entropy_client, graph, start_item, item_translation_callback, show_already_pulled_in, quiet): if not quiet: entropy_client.output("="*40, level="generic") sorted_data = graph.solve_nodes() stack = Lifo() for dep_level in sorted(sorted_data.keys(), reverse = True): stack.push(sorted_data[dep_level]) # required to make sure that our first pkg is user required one stack.push((start_item,)) out_data = { 'cache': set(), 'lvl': 0, 'txc_cb': item_translation_callback, 'show_already_pulled_in': show_already_pulled_in, } first_tree_item = True while stack.is_filled(): stack_items = stack.pop() # cleanup already printed items items = [x for x in stack_items if x not in out_data['cache']] if not items: continue out_data['cache'].update(stack_items) # print items and its deps for item in items: old_level = out_data['lvl'] _print_graph_item_deps( entropy_client, item, out_data, colorize = blue) out_data['lvl'] = old_level if first_tree_item: out_data['lvl'] += 1 first_tree_item = False del stack
def _get_flat_deps(self, ids_to_check, get_deps_func): """ Return a set (frozenset) of package IDs that are dependencies of provided packages, recursively. """ stack = Lifo() result_deps = set() for pkg_id in ids_to_check: stack.push(pkg_id) while stack.is_filled(): pkg_id = stack.pop() if pkg_id in result_deps: continue result_deps.add(pkg_id) for dep_id in get_deps_func(pkg_id): if dep_id not in result_deps: stack.push(dep_id) return frozenset(result_deps)
class EntropyCacher(Singleton): # Max number of cache objects written at once _OBJS_WRITTEN_AT_ONCE = 250 # Number of seconds between cache writeback to disk WRITEBACK_TIMEOUT = 5 # If True, in-ram cache will be used to mitigate # concurrent push/pop executions with push() not # yet able to write data to disk. STASHING_CACHE = True """ Entropy asynchronous and synchronous cache writer and reader. This class is a Singleton and contains a thread doing the cache writes asynchronously, thus it must be stopped before your application is terminated calling the stop() method. Sample code: >>> # import module >>> from entropy.cache import EntropyCacher ... >>> # first EntropyCacher load, start it >>> cacher = EntropyCacher() >>> cacher.start() ... >>> # now store something into its cache >>> cacher.push('my_identifier1', [1, 2, 3]) >>> # now store something synchronously >>> cacher.push('my_identifier2', [1, 2, 3], async_mode = False) ... >>> # now flush all the caches to disk, and make sure all >>> # is written >>> cacher.sync() ... >>> # now fetch something from the cache >>> data = cacher.pop('my_identifier1') [1, 2, 3] ... >>> # now discard all the cached (async_mode) writes >>> cacher.discard() ... >>> # and stop EntropyCacher >>> cacher.stop() """ class SemaphoreTimeScheduled(TimeScheduled): def __init__(self, sem, *args, **kwargs): self._sem = sem TimeScheduled.__init__(self, *args, **kwargs) def kill(self): self._sem.release() return TimeScheduled.kill(self) def init_singleton(self): """ Singleton overloaded method. Equals to __init__. This is the place where all the properties initialization takes place. """ self.__copy = copy self.__alive = False self.__cache_writer = None self.__cache_buffer = Lifo() self.__stashing_cache = {} self.__inside_with_stmt = 0 self.__dump_data_lock = threading.Lock() self.__worker_sem = threading.Semaphore(0) # this lock ensures that all the writes are hold while it's acquired self.__enter_context_lock = threading.RLock() def __enter__(self): """ When used with the with statement, pause cacher on-disk writes. """ self.__enter_context_lock.acquire() self.__inside_with_stmt += 1 def __exit__(self, exc_type, exc_value, traceback): """ When used with the with statement, pause cacher on-disk writes. """ self.__inside_with_stmt -= 1 self.__enter_context_lock.release() def __copy_obj(self, obj): """ Return a copy of an object done by the standard library "copy" module. @param obj: object to copy @type obj: any Python object @rtype: copied object @return: copied object """ return self.__copy.deepcopy(obj) def __cacher(self, run_until_empty=False, sync=False, _loop=False): """ This is where the actual asynchronous copy takes place. __cacher runs on a different threads and all the operations done by this are atomic and thread-safe. It just loops over and over until __alive becomes False. """ try: if self.__inside_with_stmt != 0: return except AttributeError: # interpreter shutdown pass # make sure our set delay is respected try: self.__cache_writer.set_delay(EntropyCacher.WRITEBACK_TIMEOUT) except AttributeError: # can be None pass # sleep if there's nothing to do if _loop: try: # CANBLOCK self.__worker_sem.acquire() # we just consumed one acquire() # that was dedicated to actual data, # put it back self.__worker_sem.release() except AttributeError: pass def _commit_data(_massive_data): for (key, cache_dir), data in _massive_data: d_o = entropy.dump.dumpobj if d_o is not None: d_o(key, data, dump_dir=cache_dir) while self.__alive or run_until_empty: if const_debug_enabled(): const_debug_write( __name__, "EntropyCacher.__cacher: loop: %s, alive: %s, empty: %s" % ( _loop, self.__alive, run_until_empty, )) with self.__enter_context_lock: massive_data = [] try: massive_data_count = EntropyCacher._OBJS_WRITTEN_AT_ONCE except AttributeError: # interpreter shutdown break while massive_data_count > 0: if _loop: # extracted an item from worker_sem # call down() on the semaphore without caring # can't sleep here because we're in a critical region # holding __enter_context_lock self.__worker_sem.acquire(False) massive_data_count -= 1 try: data = self.__cache_buffer.pop() except ( ValueError, TypeError, ): # TypeError is when objects are being destroyed break # stack empty massive_data.append(data) if not massive_data: break task = ParallelTask(_commit_data, massive_data) task.name = "EntropyCacherCommitter" task.daemon = not sync task.start() if sync: task.join() if const_debug_enabled(): const_debug_write( __name__, "EntropyCacher.__cacher [%s], writing %s objs" % ( task, len(massive_data), )) if EntropyCacher.STASHING_CACHE: for (key, cache_dir), data in massive_data: try: del self.__stashing_cache[(key, cache_dir)] except ( AttributeError, KeyError, ): continue del massive_data[:] del massive_data @classmethod def current_directory(cls): """ Return the path to current EntropyCacher cache storage directory. """ return entropy.dump.D_DIR def start(self): """ This is the method used to start the asynchronous cache writer but also the whole cacher. If this method is not called, the instance will always trash and cache write request. @return: None """ self.__cache_buffer.clear() self.__cache_writer = EntropyCacher.SemaphoreTimeScheduled( self.__worker_sem, EntropyCacher.WRITEBACK_TIMEOUT, self.__cacher, _loop=True) self.__cache_writer.daemon = True self.__cache_writer.name = "EntropyCacheWriter" self.__cache_writer.set_delay_before(True) self.__cache_writer.start() while not self.__cache_writer.isAlive(): continue self.__alive = True def is_started(self): """ Return whether start is called or not. This equals to checking if the cacher is running, thus is writing cache to disk. @return: None """ return self.__alive def stop(self): """ This method stops the execution of the cacher, which won't accept cache writes anymore. The thread responsible of writing to disk is stopped here and the Cacher will be back to being inactive. A watchdog will avoid the thread to freeze the call if the write buffer is overloaded. @return: None """ self.__alive = False if self.__cache_writer is not None: self.__cache_writer.kill() # make sure it unblocks self.__worker_sem.release() self.__cache_writer.join() self.__cache_writer = None self.sync() def sync(self): """ This method can be called anytime and forces the instance to flush all the cache writes queued to disk. If wait == False a watchdog prevents this call to get stuck in case of write buffer overloads. """ self.__cacher(run_until_empty=True, sync=True) def discard(self): """ This method makes buffered cache to be discarded synchronously. @return: None """ self.__cache_buffer.clear() self.__stashing_cache.clear() def save(self, key, data, cache_dir=None): """ Save data object to cache asynchronously and in any case. This method guarantees that cached data is stored even if cacher is not started. If data cannot be stored, IOError will be raised. @param key: cache data identifier @type key: string @param data: picklable object @type data: any picklable object @keyword cache_dir: alternative cache directory @type cache_dir: string """ if cache_dir is None: cache_dir = self.current_directory() try: with self.__dump_data_lock: entropy.dump.dumpobj(key, data, dump_dir=cache_dir, ignore_exceptions=False) except (EOFError, IOError, OSError) as err: raise IOError("cannot store %s to %s. err: %s" % (key, cache_dir, repr(err))) def push(self, key, data, async_mode=True, cache_dir=None): """ This is the place where data is either added to the write queue or written to disk (if async_mode == False) only and only if start() method has been called. @param key: cache data identifier @type key: string @param data: picklable object @type data: any picklable object @keyword async_mode: store cache asynchronously or not @type async_mode: bool @keyword cache_dir: alternative cache directory @type cache_dir: string """ if not self.__alive: return if cache_dir is None: cache_dir = self.current_directory() if async_mode: try: obj_copy = self.__copy_obj(data) self.__cache_buffer.push(( ( key, cache_dir, ), obj_copy, )) self.__worker_sem.release() if EntropyCacher.STASHING_CACHE: self.__stashing_cache[(key, cache_dir)] = obj_copy except TypeError: # sometimes, very rarely, copy.deepcopy() is unable # to properly copy an object (blame Python bug) sys.stdout.write("!!! cannot cache object with key %s\n" % (key, )) sys.stdout.flush() #if const_debug_enabled(): # const_debug_write(__name__, # "EntropyCacher.push, async_mode push %s, into %s" % ( # key, cache_dir,)) else: #if const_debug_enabled(): # const_debug_write(__name__, # "EntropyCacher.push, sync push %s, into %s" % ( # key, cache_dir,)) with self.__dump_data_lock: entropy.dump.dumpobj(key, data, dump_dir=cache_dir) def pop(self, key, cache_dir=None, aging_days=None): """ This is the place where data is retrieved from cache. You must know the cache identifier used when push() was called. @param key: cache data identifier @type key: string @keyword cache_dir: alternative cache directory @type cache_dir: string @rtype: Python object @return: object stored into the stack or None (if stack is empty) """ if cache_dir is None: cache_dir = self.current_directory() if EntropyCacher.STASHING_CACHE: # object is being saved on disk, it's in RAM atm ram_obj = self.__stashing_cache.get((key, cache_dir)) if ram_obj is not None: return ram_obj l_o = entropy.dump.loadobj if not l_o: return return l_o(key, dump_dir=cache_dir, aging_days=aging_days) @classmethod def clear_cache_item(cls, cache_item, cache_dir=None): """ Clear Entropy Cache item from on-disk cache. @param cache_item: Entropy Cache item identifier @type cache_item: string @keyword cache_dir: alternative cache directory @type cache_dir: string """ if cache_dir is None: cache_dir = cls.current_directory() dump_path = os.path.join(cache_dir, cache_item) dump_dir = os.path.dirname(dump_path) for currentdir, subdirs, files in os.walk(dump_dir): path = os.path.join(dump_dir, currentdir) for item in files: if item.endswith(entropy.dump.D_EXT): item = os.path.join(path, item) try: os.remove(item) except ( OSError, IOError, ): pass try: if not os.listdir(path): os.rmdir(path) except ( OSError, IOError, ): pass
def _graph_package(match, package, entropy_intf, show_complete = False, quiet = False): include_sys_pkgs = False show_already_pulled_in = False if show_complete: include_sys_pkgs = True show_already_pulled_in = True graph = Graph() stack = Lifo() start_item = (match, package, None) stack.push(start_item) stack_cache = set() # ensure package availability in graph, initialize now graph.add(start_item, []) depsorter = lambda x: entropy.dep.dep_getcpv(x[0]) while stack.is_filled(): item = stack.pop() if item in stack_cache: continue stack_cache.add(item) ((pkg_id, repo_id,), _was_dep, dep_type) = item # deps repodb = entropy_intf.open_repository(repo_id) deps = repodb.retrieveDependencies(pkg_id, extended = True, resolve_conditional_deps = False) graph_deps = [] for dep, x_dep_type in sorted(deps, key = depsorter): if dep.startswith("!"): # conflict continue dep_item = entropy_intf.atom_match(dep) if dep_item[0] == -1: continue do_include = True if not include_sys_pkgs: dep_repodb = entropy_intf.open_repository(dep_item[1]) do_include = not dep_repodb.isSystemPackage(dep_item[0]) g_item = (dep_item, dep, x_dep_type) if do_include: stack.push(g_item) graph_deps.append(g_item) graph.add(item, graph_deps) def item_translation_func(match): value = "%s" % (match[1],) if match[2] is not None: value += " %s%s%s" % (teal("{"), brown(str(match[2])), teal("}"),) return value _graph_to_stdout(entropy_intf, graph, graph.get_node(start_item), item_translation_func, show_already_pulled_in, quiet) if not quiet: _show_graph_legend(entropy_intf) show_dependencies_legend(entropy_intf) del stack graph.destroy() return 0
def _revgraph_package(entropy_client, installed_pkg_id, package, dbconn, show_complete = False, quiet = False): include_sys_pkgs = False show_already_pulled_in = False include_build_deps = False if show_complete: include_sys_pkgs = True show_already_pulled_in = True include_build_deps = True excluded_dep_types = [etpConst['dependency_type_ids']['bdepend_id']] if not include_build_deps: excluded_dep_types = None graph = Graph() stack = Lifo() inst_item = (installed_pkg_id, package) stack.push(inst_item) stack_cache = set() # ensure package availability in graph, initialize now graph.add(inst_item, set()) def rev_pkgs_sorter(_pkg_id): return dbconn.retrieveAtom(_pkg_id) while stack.is_filled(): item = stack.pop() if item in stack_cache: continue stack_cache.add(item) pkg_id, _was_dep = item rev_deps = dbconn.retrieveReverseDependencies(pkg_id, exclude_deptypes = excluded_dep_types) graph_deps = [] for rev_pkg_id in sorted(rev_deps, key = rev_pkgs_sorter): dep = dbconn.retrieveAtom(rev_pkg_id) do_include = True if not include_sys_pkgs: do_include = not dbconn.isSystemPackage(rev_pkg_id) g_item = (rev_pkg_id, dep) if do_include: stack.push(g_item) graph_deps.append(g_item) graph.add(item, graph_deps) def item_translation_func(match): return match[1] _graph_to_stdout(entropy_client, graph, graph.get_node(inst_item), item_translation_func, show_already_pulled_in, quiet) if not quiet: _show_graph_legend(entropy_client) del stack graph.destroy() return 0
class EntropyTransceiver(TextInterface): """ Base class for Entropy transceivers. This provides a common API across all the available URI handlers. How to use this class: Let's consider that we have a valid EntropyUriHandler for ftp:// protocol already installed via "add_uri_handler". >> txc = EntropyTransceiver("ftp://*****:*****@myhost") >> txc.set_speed_limit(150) # set speed limit to 150kb/sec >> handler = txc.swallow() >> handler.download("ftp://*****:*****@myhost/myfile.txt", "/tmp") # download """ _URI_HANDLERS = [] @staticmethod def add_uri_handler(entropy_uri_handler_class): """ Add custom URI handler to EntropyTransceiver class. @param entropy_uri_handler_class: EntropyUriHandler based class @type entropy_uri_handler_class; EntropyUriHandler instance """ if not issubclass(entropy_uri_handler_class, EntropyUriHandler): raise AttributeError("EntropyUriHandler based class expected") EntropyTransceiver._URI_HANDLERS.append(entropy_uri_handler_class) @staticmethod def remove_uri_handler(entropy_uri_handler_class): """ Remove custom URI handler to EntropyTransceiver class. @param entropy_uri_handler_class: EntropyUriHandler based instance @type entropy_uri_handler_class; EntropyUriHandler instance @raise ValueError: if provided EntropyUriHandler is not in storage. """ if not issubclass(entropy_uri_handler_class, EntropyUriHandler): raise AttributeError("EntropyUriHandler based class expected") EntropyTransceiver._URI_HANDLERS.remove(entropy_uri_handler_class) @staticmethod def get_uri_handlers(): """ Return a copy of the internal list of URI handler instances. @return: URI handlers instances list @rtype: list """ return EntropyTransceiver._URI_HANDLERS[:] @staticmethod def get_uri_name(uri): """ Given an URI, extract and return the URI name (hostname). @param uri: URI to handle @type uri: string @return: URI name @rtype: string @raise UriHandlerNotFound: if no URI handlers can deal with given URI string """ handlers = EntropyTransceiver.get_uri_handlers() for handler in handlers: if handler.approve_uri(uri): return handler.get_uri_name(uri) raise UriHandlerNotFound( "no URI handler available for %s" % (uri,)) @staticmethod def hide_sensible_data(uri): """ Given an URI, hide sensible data from string and return it back. @param uri: URI to handle @type uri: string @return: URI cleaned @rtype: string @raise UriHandlerNotFound: if no URI handlers can deal with given URI string """ handlers = EntropyTransceiver.get_uri_handlers() for handler in handlers: if handler.approve_uri(uri): return handler.hide_sensible_data(uri) raise UriHandlerNotFound( "no URI handler available for %s" % (uri,)) def __init__(self, uri): """ EntropyTransceiver constructor, just pass the friggin URI(tm). @param uri: URI to handle @type uri: string """ self._uri = uri self._speed_limit = 0 self._verbose = False self._timeout = None self._silent = None self._output_interface = None self.__with_stack = Lifo() def __enter__(self): """ Support for "with" statement, this method will execute swallow() and return a valid EntropyUriHandler instance. """ handler = self.swallow() self.__with_stack.push(handler) return handler def __exit__(self, exc_type, exc_value, traceback): """ Support for "with" statement, this method will automagically close the previously created EntropyUriHandler instance connection. """ handler = self.__with_stack.pop() # if this fails, it's not a good sign handler.close() def set_output_interface(self, output_interface): """ Provide alternative Entropy output interface (must be based on entropy.output.TextInterface) @param output_interface: new entropy.output.TextInterface instance to use @type output_interface: entropy.output.TextInterface based instance @raise AttributeError: if argument passed is not correct """ if not isinstance(output_interface, TextInterface): raise AttributeError( "expected a valid TextInterface based instance") self._output_interface = output_interface def set_speed_limit(self, speed_limit): """ Set download/upload speed limit in kb/sec form. Zero value will be considered as "disable speed limiter". @param speed_limit: speed limit in kb/sec form. @type speed_limit: int @raise AttributeError: if speed_limit is not an integer """ if not const_isnumber(speed_limit): raise AttributeError("expected a valid number") self._speed_limit = speed_limit def set_timeout(self, timeout): """ Set transceiver tx/rx timeout value in seconds. @param timeout: timeout in seconds @type timeout: int """ if not const_isnumber(timeout): raise AttributeError("not a number") self._timeout = timeout def set_silent(self, silent): """ Disable transceivers verbosity. @param verbosity: verbosity value @type verbosity: bool """ self._silent = silent def set_verbosity(self, verbosity): """ Set transceiver verbosity. @param verbosity: verbosity value @type verbosity: bool """ if not isinstance(verbosity, bool): raise AttributeError("expected a valid bool") self._verbose = verbosity def swallow(self): """ Given the URI at the constructor, this method returns the first valid URI handler instance found that can be used to do required action. @raise entropy.exceptions.UriHandlerNotFound: when URI handler for given URI is not available. """ handlers = EntropyTransceiver.get_uri_handlers() for handler in handlers: if handler.approve_uri(self._uri): handler_instance = handler(self._uri) if self._output_interface is not None: handler_instance.set_output_interface( self._output_interface) if const_isnumber(self._speed_limit): handler_instance.set_speed_limit(self._speed_limit) handler_instance.set_verbosity(self._verbose) handler_instance.set_silent(self._silent) if const_isnumber(self._timeout): handler_instance.set_timeout(self._timeout) return handler_instance raise UriHandlerNotFound( "no URI handler available for %s" % (self._uri,))
class TopologicalSorter(object): """ This class implements the topological sorting algorithm presented by R. E. Tarjan in 1972. """ def __init__(self, adjacency_map): """ TopologicalSorter constructor. @param adjacency_map: dict form adjacency map @type adjacency_map: dict """ object.__init__(self) self.__adjacency_map = adjacency_map self.__stack = Lifo() def __topological_sort_visit_node(self, node, low, result): """ Internal method, visits a node ad push to stack. """ if node in low: return num = len(low) low[node] = num stack_pos = len(self.__stack) self.__stack.push(node) for successor in self.__adjacency_map[node]: self.__topological_sort_visit_node(successor, low, result) low[node] = min(low[node], low[successor]) if num == low[node]: component = tuple() while len(self.__stack) > stack_pos: component += (self.__stack.pop(), ) result.append(component) for item in component: low[item] = len(self.__adjacency_map) def __strongly_connected_nodes(self): """ Find the strongly connected nodes in a adjacency_map using Tarjan's algorithm. adjacency_map should be a dictionary mapping node names to lists of successor nodes. """ result = [] low = {} for node in self.__adjacency_map: self.__topological_sort_visit_node(node, low, result) return result def __topological_sort(self, graph): """ Effectively executes topological sorting on given graph. """ # initialize count map count = dict((node, 0) for node in graph) for node in graph: for successor in graph[node]: count[successor] += 1 ready_stack = Lifo() for node in graph: if count[node] == 0: ready_stack.push(node) dep_level = 1 result = {} while ready_stack.is_filled(): node = ready_stack.pop() result[dep_level] = node dep_level += 1 for successor in graph[node]: count[successor] -= 1 if count[successor] == 0: ready_stack.push(successor) return result def get_stored_adjacency_map(self): """ Return stored adjacency map used for sorting. @return: stored adjacency map @rtype: dict """ return self.__adjacency_map def sort(self): """ Given an adjacency map, identify strongly connected nodes, then perform a topological sort on them. @return: sorted graph representation @rtype: dict """ # clear stack self.__stack.clear() components = self.__strongly_connected_nodes() node_component = {} for component in components: for node in component: node_component[node] = component component_graph = {} for node in self.__adjacency_map: node_c = node_component[node] obj = component_graph.setdefault(node_c, []) for successor in self.__adjacency_map[node]: successor_c = node_component[successor] if node_c != successor_c: obj.append(successor_c) return self.__topological_sort(component_graph)
class EntropyCacher(Singleton): # Max number of cache objects written at once _OBJS_WRITTEN_AT_ONCE = 250 # Number of seconds between cache writeback to disk WRITEBACK_TIMEOUT = 5 # If True, in-ram cache will be used to mitigate # concurrent push/pop executions with push() not # yet able to write data to disk. STASHING_CACHE = True """ Entropy asynchronous and synchronous cache writer and reader. This class is a Singleton and contains a thread doing the cache writes asynchronously, thus it must be stopped before your application is terminated calling the stop() method. Sample code: >>> # import module >>> from entropy.cache import EntropyCacher ... >>> # first EntropyCacher load, start it >>> cacher = EntropyCacher() >>> cacher.start() ... >>> # now store something into its cache >>> cacher.push('my_identifier1', [1, 2, 3]) >>> # now store something synchronously >>> cacher.push('my_identifier2', [1, 2, 3], async = False) ... >>> # now flush all the caches to disk, and make sure all >>> # is written >>> cacher.sync() ... >>> # now fetch something from the cache >>> data = cacher.pop('my_identifier1') [1, 2, 3] ... >>> # now discard all the cached (async) writes >>> cacher.discard() ... >>> # and stop EntropyCacher >>> cacher.stop() """ class SemaphoreTimeScheduled(TimeScheduled): def __init__(self, sem, *args, **kwargs): self._sem = sem TimeScheduled.__init__(self, *args, **kwargs) def kill(self): self._sem.release() return TimeScheduled.kill(self) def init_singleton(self): """ Singleton overloaded method. Equals to __init__. This is the place where all the properties initialization takes place. """ self.__copy = copy self.__alive = False self.__cache_writer = None self.__cache_buffer = Lifo() self.__stashing_cache = {} self.__inside_with_stmt = 0 self.__dump_data_lock = threading.Lock() self.__worker_sem = threading.Semaphore(0) # this lock ensures that all the writes are hold while it's acquired self.__enter_context_lock = threading.RLock() def __enter__(self): """ When used with the with statement, pause cacher on-disk writes. """ self.__enter_context_lock.acquire() self.__inside_with_stmt += 1 def __exit__(self, exc_type, exc_value, traceback): """ When used with the with statement, pause cacher on-disk writes. """ self.__inside_with_stmt -= 1 self.__enter_context_lock.release() def __copy_obj(self, obj): """ Return a copy of an object done by the standard library "copy" module. @param obj: object to copy @type obj: any Python object @rtype: copied object @return: copied object """ return self.__copy.deepcopy(obj) def __cacher(self, run_until_empty=False, sync=False, _loop=False): """ This is where the actual asynchronous copy takes place. __cacher runs on a different threads and all the operations done by this are atomic and thread-safe. It just loops over and over until __alive becomes False. """ try: if self.__inside_with_stmt != 0: return except AttributeError: # interpreter shutdown pass # make sure our set delay is respected try: self.__cache_writer.set_delay(EntropyCacher.WRITEBACK_TIMEOUT) except AttributeError: # can be None pass # sleep if there's nothing to do if _loop: try: # CANBLOCK self.__worker_sem.acquire() # we just consumed one acquire() # that was dedicated to actual data, # put it back self.__worker_sem.release() except AttributeError: pass def _commit_data(_massive_data): for (key, cache_dir), data in _massive_data: d_o = entropy.dump.dumpobj if d_o is not None: d_o(key, data, dump_dir=cache_dir) while self.__alive or run_until_empty: if const_debug_enabled(): const_debug_write( __name__, "EntropyCacher.__cacher: loop: %s, alive: %s, empty: %s" % (_loop, self.__alive, run_until_empty), ) with self.__enter_context_lock: massive_data = [] try: massive_data_count = EntropyCacher._OBJS_WRITTEN_AT_ONCE except AttributeError: # interpreter shutdown break while massive_data_count > 0: if _loop: # extracted an item from worker_sem # call down() on the semaphore without caring # can't sleep here because we're in a critical region # holding __enter_context_lock self.__worker_sem.acquire(False) massive_data_count -= 1 try: data = self.__cache_buffer.pop() except (ValueError, TypeError): # TypeError is when objects are being destroyed break # stack empty massive_data.append(data) if not massive_data: break task = ParallelTask(_commit_data, massive_data) task.name = "EntropyCacherCommitter" task.daemon = not sync task.start() if sync: task.join() if const_debug_enabled(): const_debug_write( __name__, "EntropyCacher.__cacher [%s], writing %s objs" % (task, len(massive_data)) ) if EntropyCacher.STASHING_CACHE: for (key, cache_dir), data in massive_data: try: del self.__stashing_cache[(key, cache_dir)] except (AttributeError, KeyError): continue del massive_data[:] del massive_data @classmethod def current_directory(cls): """ Return the path to current EntropyCacher cache storage directory. """ return entropy.dump.D_DIR def start(self): """ This is the method used to start the asynchronous cache writer but also the whole cacher. If this method is not called, the instance will always trash and cache write request. @return: None """ self.__cache_buffer.clear() self.__cache_writer = EntropyCacher.SemaphoreTimeScheduled( self.__worker_sem, EntropyCacher.WRITEBACK_TIMEOUT, self.__cacher, _loop=True ) self.__cache_writer.daemon = True self.__cache_writer.name = "EntropyCacheWriter" self.__cache_writer.set_delay_before(True) self.__cache_writer.start() while not self.__cache_writer.isAlive(): continue self.__alive = True def is_started(self): """ Return whether start is called or not. This equals to checking if the cacher is running, thus is writing cache to disk. @return: None """ return self.__alive def stop(self): """ This method stops the execution of the cacher, which won't accept cache writes anymore. The thread responsible of writing to disk is stopped here and the Cacher will be back to being inactive. A watchdog will avoid the thread to freeze the call if the write buffer is overloaded. @return: None """ self.__alive = False if self.__cache_writer is not None: self.__cache_writer.kill() # make sure it unblocks self.__worker_sem.release() self.__cache_writer.join() self.__cache_writer = None self.sync() def sync(self): """ This method can be called anytime and forces the instance to flush all the cache writes queued to disk. If wait == False a watchdog prevents this call to get stuck in case of write buffer overloads. """ self.__cacher(run_until_empty=True, sync=True) def discard(self): """ This method makes buffered cache to be discarded synchronously. @return: None """ self.__cache_buffer.clear() self.__stashing_cache.clear() def save(self, key, data, cache_dir=None): """ Save data object to cache asynchronously and in any case. This method guarantees that cached data is stored even if cacher is not started. If data cannot be stored, IOError will be raised. @param key: cache data identifier @type key: string @param data: picklable object @type data: any picklable object @keyword cache_dir: alternative cache directory @type cache_dir: string """ if cache_dir is None: cache_dir = self.current_directory() try: with self.__dump_data_lock: entropy.dump.dumpobj(key, data, dump_dir=cache_dir, ignore_exceptions=False) except (EOFError, IOError, OSError) as err: raise IOError("cannot store %s to %s. err: %s" % (key, cache_dir, repr(err))) def push(self, key, data, async=True, cache_dir=None):
class TopologicalSorter(object): """ This class implements the topological sorting algorithm presented by R. E. Tarjan in 1972. """ def __init__(self, adjacency_map): """ TopologicalSorter constructor. @param adjacency_map: dict form adjacency map @type adjacency_map: dict """ object.__init__(self) self.__adjacency_map = adjacency_map self.__stack = Lifo() def __topological_sort_visit_node(self, node, low, result): """ Internal method, visits a node ad push to stack. """ if node in low: return num = len(low) low[node] = num stack_pos = len(self.__stack) self.__stack.push(node) for successor in self.__adjacency_map[node]: self.__topological_sort_visit_node(successor, low, result) low[node] = min(low[node], low[successor]) if num == low[node]: component = tuple() while len(self.__stack) > stack_pos: component += (self.__stack.pop(),) result.append(component) for item in component: low[item] = len(self.__adjacency_map) def __strongly_connected_nodes(self): """ Find the strongly connected nodes in a adjacency_map using Tarjan's algorithm. adjacency_map should be a dictionary mapping node names to lists of successor nodes. """ result = [] low = {} for node in self.__adjacency_map: self.__topological_sort_visit_node(node, low, result) return result def __topological_sort(self, graph): """ Effectively executes topological sorting on given graph. """ # initialize count map count = dict((node, 0) for node in graph) for node in graph: for successor in graph[node]: count[successor] += 1 ready_stack = Lifo() for node in graph: if count[node] == 0: ready_stack.push(node) dep_level = 1 result = {} while ready_stack.is_filled(): node = ready_stack.pop() result[dep_level] = node dep_level += 1 for successor in graph[node]: count[successor] -= 1 if count[successor] == 0: ready_stack.push(successor) return result def get_stored_adjacency_map(self): """ Return stored adjacency map used for sorting. @return: stored adjacency map @rtype: dict """ return self.__adjacency_map def sort(self): """ Given an adjacency map, identify strongly connected nodes, then perform a topological sort on them. @return: sorted graph representation @rtype: dict """ # clear stack self.__stack.clear() components = self.__strongly_connected_nodes() node_component = {} for component in components: for node in component: node_component[node] = component component_graph = {} for node in self.__adjacency_map: node_c = node_component[node] obj = component_graph.setdefault(node_c, []) for successor in self.__adjacency_map[node]: successor_c = node_component[successor] if node_c != successor_c: obj.append(successor_c) return self.__topological_sort(component_graph)