def reduce(self, func, output=None, options: dict = None): total_partitions = self.__store._store_locator._total_partitions job_id = generate_job_id(self.__session_id, tag=RollPair.REDUCE) serialized_func = ErFunctor(name=RollPair.REDUCE, serdes=SerdesTypes.CLOUD_PICKLE, body=cloudpickle.dumps(func)) job = ErJob(id=job_id, name=RollPair.REDUCE, inputs=[self.ctx.populate_processor(self.__store)], functors=[serialized_func]) args = list() for i in range(total_partitions): partition_input = job._inputs[0]._partitions[i] task = ErTask(id=generate_task_id(job_id, i), name=job._name, inputs=[partition_input], job=job) args.append(([task], partition_input._processor._command_endpoint)) futures = self.__command_client.async_call( args=args, output_types=[ErPair], command_uri=CommandURI( f'{RollPair.EGG_PAIR_URI_PREFIX}/{RollPair.RUN_TASK}')) done = wait(futures, return_when=FIRST_EXCEPTION).done result = None first = True for future in done: pair = future.result()[0] seq_op_result = self.functor_serdes.deserialize(pair._value) if seq_op_result is not None: if not first: result = func(result, seq_op_result) else: result = seq_op_result first = False return result
def glom(self, output=None, options: dict = None): if options is None: options = {} outputs = [] if output: RollPair.__check_partition(self.get_partitions(), output._store_locator._total_partitions) outputs.append(output) functor = ErFunctor(name=RollPair.GLOM, serdes=SerdesTypes.CLOUD_PICKLE) job = ErJob(id=generate_job_id(self.__session_id, RollPair.GLOM), name=RollPair.GLOM, inputs=[self.__store], outputs=outputs, functors=[functor]) task_results = self._run_job(job=job) er_store = self.__get_output_from_result(task_results) return RollPair(er_store, self.ctx)
def get_all(self, limit=None, options: dict = None): if options is None: options = {} if limit is not None and not isinstance(limit, int) and limit <= 0: raise ValueError(f"limit:{limit} must be positive int") job_id = generate_job_id(self.__session_id, RollPair.GET_ALL) er_pair = ErPair(key=create_serdes( self.__store._store_locator._serdes).serialize(limit) if limit is not None else None, value=None) def send_command(): job = ErJob(id=job_id, name=RollPair.GET_ALL, inputs=[self.__store], outputs=[self.__store], functors=[ ErFunctor(name=RollPair.GET_ALL, body=cloudpickle.dumps(er_pair)) ]) task_results = self._run_job(job=job) er_store = self.__get_output_from_result(task_results) return er_store send_command() populated_store = self.ctx.populate_processor(self.__store) transfer_pair = TransferPair(transfer_id=job_id) done_cnt = 0 for k, v in transfer_pair.gather(populated_store): done_cnt += 1 yield self.key_serdes.deserialize( k), self.value_serdes.deserialize(v) L.trace( f"get_all: namespace={self.get_namespace()} name={self.get_name()}, count={done_cnt}" )
def get(self, k, options: dict = None): if options is None: options = {} k = create_serdes(self.__store._store_locator._serdes).serialize(k) er_pair = ErPair(key=k, value=None) partition_id = self.partitioner(k) egg = self.ctx.route_to_egg(self.__store._partitions[partition_id]) inputs = [ ErPartition(id=partition_id, store_locator=self.__store._store_locator) ] outputs = [ ErPartition(id=partition_id, store_locator=self.__store._store_locator) ] job_id = generate_job_id(self.__session_id, RollPair.GET) job = ErJob(id=job_id, name=RollPair.GET, inputs=[self.__store], outputs=[self.__store], functors=[ ErFunctor(name=RollPair.GET, body=cloudpickle.dumps(er_pair)) ]) task = ErTask(id=generate_task_id(job_id, partition_id), name=RollPair.GET, inputs=inputs, outputs=outputs, job=job) job_resp = self.__command_client.simple_sync_send( input=task, output_type=ErPair, endpoint=egg._command_endpoint, command_uri=self.RUN_TASK_URI, serdes_type=self.__command_serdes) return self.value_serdes.deserialize( job_resp._value) if job_resp._value != b'' else None
def sample(self, fraction, seed=None, output=None, options: dict = None): if options is None: options = {} outputs = [] if output: RollPair.__check_partition(self.get_partitions(), output._store_locator._total_partitions) outputs.append(output) er_fraction = ErFunctor(name=RollPair.REDUCE, serdes=SerdesTypes.CLOUD_PICKLE, body=cloudpickle.dumps(fraction)) er_seed = ErFunctor(name=RollPair.REDUCE, serdes=SerdesTypes.CLOUD_PICKLE, body=cloudpickle.dumps(seed)) job = ErJob(id=generate_job_id(self.__session_id, RollPair.SAMPLE), name=RollPair.SAMPLE, inputs=[self.__store], outputs=outputs, functors=[er_fraction, er_seed]) task_results = self._run_job(job=job) er_store = self.__get_output_from_result(task_results) return RollPair(er_store, self.ctx)
def union(self, other, func=lambda v1, v2: v1, output=None, options: dict = None): if options is None: options = {} inputs = self.__repartition_with(other) outputs = [] if output: RollPair.__check_partition(inputs[0]._store_locator._total_partitions, output._store_locator._total_partitions) outputs.append(output) functor = ErFunctor(name=RollPair.UNION, serdes=SerdesTypes.CLOUD_PICKLE, body=cloudpickle.dumps(func)) job = ErJob(id=generate_job_id(self.__session_id, RollPair.UNION), name=RollPair.UNION, inputs=inputs, outputs=outputs, functors=[functor]) task_future = self._run_job(job=job) er_store = self.__get_output_from_result(task_future) return RollPair(er_store, self.ctx)
def subtract_by_key(self, other, output=None, options: dict = None): if options is None: options = {} inputs = self.__repartition_with(other) outputs = [] if output: RollPair.__check_partition(inputs[0]._store_locator._total_partitions, output._store_locator._total_partitions) outputs.append(output) functor = ErFunctor(name=RollPair.SUBTRACT_BY_KEY, serdes=SerdesTypes.CLOUD_PICKLE) job = ErJob(id=generate_job_id(self.__session_id, RollPair.SUBTRACT_BY_KEY), name=RollPair.SUBTRACT_BY_KEY, inputs=inputs, outputs=outputs, functors=[functor]) task_future = self._run_job(job=job) er_store = self.__get_output_from_result(task_future) return RollPair(er_store, self.ctx)
def delete(self, k, options: dict = None): if options is None: options = {} key = create_serdes(self.__store._store_locator._serdes).serialize(k) er_pair = ErPair(key=key, value=None) outputs = [] value = None partition_id = self.partitioner(key) egg = self.ctx.route_to_egg(self.__store._partitions[partition_id]) L.info(egg._command_endpoint) L.info(f"count: {self.__store._store_locator._total_partitions}") inputs = [ ErPartition(id=partition_id, store_locator=self.__store._store_locator) ] output = [ ErPartition(id=partition_id, store_locator=self.__store._store_locator) ] job_id = generate_job_id(self.__session_id, RollPair.DELETE) job = ErJob(id=job_id, name=RollPair.DELETE, inputs=[self.__store], outputs=outputs, functors=[ErFunctor(body=cloudpickle.dumps(er_pair))]) task = ErTask(id=generate_task_id(job_id, partition_id), name=RollPair.DELETE, inputs=inputs, outputs=output, job=job) L.info("start send req") job_resp = self.__command_client.simple_sync_send( input=task, output_type=ErPair, endpoint=egg._command_endpoint, command_uri=CommandURI( f'{RollPair.EGG_PAIR_URI_PREFIX}/{RollPair.RUN_TASK}'), serdes_type=self.__command_serdes)
def destroy(self, options: dict = None): if len(self.ctx.get_session()._cluster_manager_client.get_store( self.get_store())._partitions) == 0: L.exception(f"store:{self.get_store()} has been destroyed before") raise ValueError( f"store:{self.get_store()} has been destroyed before") if options is None: options = {} job = ErJob(id=generate_job_id(self.__session_id, RollPair.DESTROY), name=RollPair.DESTROY, inputs=[self.__store], outputs=[self.__store], functors=[], options=options) task_results = self._run_job(job=job, create_output_if_missing=False) self.ctx.get_session()._cluster_manager_client.delete_store( self.__store) L.debug(f'{RollPair.DESTROY}={self.__store}') self.destroyed = True
def put(self, k, v, options: dict = None): if options is None: options = {} k, v = create_serdes(self.__store._store_locator._serdes).serialize(k), \ create_serdes(self.__store._store_locator._serdes).serialize(v) er_pair = ErPair(key=k, value=v) outputs = [] partition_id = self.partitioner(k) egg = self.ctx.route_to_egg(self.__store._partitions[partition_id]) inputs = [ ErPartition(id=partition_id, store_locator=self.__store._store_locator) ] output = [ErPartition(id=0, store_locator=self.__store._store_locator)] job_id = generate_job_id(self.__session_id, RollPair.PUT) job = ErJob(id=job_id, name=RollPair.PUT, inputs=[self.__store], outputs=outputs, functors=[ ErFunctor(name=RollPair.PUT, body=cloudpickle.dumps(er_pair)) ]) task = ErTask(id=generate_task_id(job_id, partition_id), name=RollPair.PUT, inputs=inputs, outputs=output, job=job) job_resp = self.__command_client.simple_sync_send( input=task, output_type=ErPair, endpoint=egg._command_endpoint, command_uri=CommandURI( f'{RollPair.EGG_PAIR_URI_PREFIX}/{RollPair.RUN_TASK}'), serdes_type=self.__command_serdes) value = job_resp._value return value
def flat_map(self, func, output=None, options: dict = None): if options is None: options = {} outputs = [] if output: RollPair.__check_partition(self.get_partitions(), output._store_locator._total_partitions) outputs.append(output) shuffle = options.get('shuffle', True) functor = ErFunctor(name=RollPair.FLAT_MAP, serdes=SerdesTypes.CLOUD_PICKLE, body=cloudpickle.dumps(func)) need_shuffle = ErFunctor(name=RollPair.FLAT_MAP, serdes=SerdesTypes.CLOUD_PICKLE, body=cloudpickle.dumps(shuffle)) job = ErJob(id=generate_job_id(self.__session_id, RollPair.FLAT_MAP), name=RollPair.FLAT_MAP, inputs=[self.__store], outputs=outputs, functors=[functor, need_shuffle]) task_results = self._run_job(job=job) er_store = self.__get_output_from_result(task_results) return RollPair(er_store, self.ctx)
def with_stores(self, func, others=None, options: dict = None): if options is None: options = {} tag = "withStores" if others is None: others = [] total_partitions = self.get_partitions() for other in others: if other.get_partitions() != total_partitions: raise ValueError(f"diff partitions: expected:{total_partitions}, actual:{other.get_partitions()}") job_id = generate_job_id(self.__session_id, tag=tag) job = ErJob(id=job_id, name=tag, inputs=[self.ctx.populate_processor(rp.get_store()) for rp in [self] + others], functors=[ErFunctor(name=tag, serdes=SerdesTypes.CLOUD_PICKLE, body=cloudpickle.dumps(func))], options=options) args = list() for i in range(total_partitions): partition_self = job._inputs[0]._partitions[i] task = ErTask(id=generate_task_id(job_id, i), name=job._name, inputs=[store._partitions[i] for store in job._inputs], job=job) args.append(([task], partition_self._processor._command_endpoint)) futures = self.__command_client.async_call( args=args, output_types=[ErPair], command_uri=CommandURI(f'{RollPair.EGG_PAIR_URI_PREFIX}/{RollPair.RUN_TASK}')) result = list() for future in futures: ret_pair = future.result()[0] result.append((self.functor_serdes.deserialize(ret_pair._key), self.functor_serdes.deserialize(ret_pair._value))) return result
def map_values(self, func, output=None, options: dict = None): if options is None: options = {} outputs = self._maybe_set_output(output) functor = ErFunctor(name=RollPair.MAP_VALUES, serdes=SerdesTypes.CLOUD_PICKLE, body=cloudpickle.dumps(func)) # todo:1: options issues. refer to line 77 final_options = {} final_options.update(self.__store._options) final_options.update(options) job = ErJob(id=generate_job_id(self.__session_id, RollPair.MAP_VALUES), name=RollPair.MAP_VALUES, inputs=[self.__store], outputs=outputs, functors=[functor], options=final_options) task_results = self._run_job(job=job) er_store = self.__get_output_from_result(task_results) return RollPair(er_store, self.ctx)
def cleanup(self, name, namespace, options: dict = None): if not namespace: raise ValueError('namespace cannot be blank') L.debug(f'cleaning up namespace={namespace}, name={name}') if options is None: options = {} total_partitions = options.get('total_partitions', 1) partitioner = options.get('partitioner', PartitionerTypes.BYTESTRING_HASH) store_serdes = options.get('serdes', self.default_store_serdes) if name == '*': store_type = options.get('store_type', '*') L.debug(f'cleaning up whole store_type={store_type}, namespace={namespace}, name={name}') er_store = ErStore(store_locator=ErStoreLocator(namespace=namespace, name=name, store_type=store_type)) job_id = generate_job_id(namespace, tag=RollPair.CLEANUP) job = ErJob(id=job_id, name=RollPair.DESTROY, inputs=[er_store], options=options) args = list() cleanup_partitions = [ErPartition(id=-1, store_locator=er_store._store_locator)] for server_node, eggs in self.__session._eggs.items(): egg = eggs[0] task = ErTask(id=generate_task_id(job_id, egg._command_endpoint._host), name=job._name, inputs=cleanup_partitions, job=job) args.append(([task], egg._command_endpoint)) futures = self.__command_client.async_call( args=args, output_types=[ErTask], command_uri=CommandURI(f'{RollPair.EGG_PAIR_URI_PREFIX}/{RollPair.RUN_TASK}')) for future in futures: result = future.result() self.get_session()._cluster_manager_client.delete_store(er_store) else: # todo:1: add combine options to pass it through store_options = self.__session.get_all_options() store_options.update(options) final_options = store_options.copy() store = ErStore( store_locator=ErStoreLocator( store_type=StoreTypes.ROLLPAIR_LMDB, namespace=namespace, name=name, total_partitions=total_partitions, partitioner=partitioner, serdes=store_serdes), options=final_options) task_results = self.__session._cluster_manager_client.get_store_from_namespace(store) L.trace('res={}'.format(task_results._stores)) if task_results._stores is not None: L.trace("item count={}".format(len(task_results._stores))) for item in task_results._stores: L.trace("item namespace={} name={}".format(item._store_locator._namespace, item._store_locator._name)) rp = RollPair(er_store=item, rp_ctx=self) rp.destroy()