# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. __all__ = ['buffered', 'compose', 'chain', 'shuffle', 'ComposeNotAligned'] from Queue import Queue from threading import Thread import itertools import random def shuffle(reader_creator, buf_size): """Creates a data reader creator whose data output is suffled. Output from the iterator that created by original reader creator will be buffered into shuffle buffer, and then shuffled. The size of shuffle buffer is determined by argument buf_size. Args: reader_creator: the original reader creator whose output will be shuffled. buf_size: shuffle buffer size. Returns: the new reader creator whose output is shuffled. """ def create_reader_creator(): buf = [] for e in reader_creator(): buf.append(e) if len(buf) >= buf_size: random.shuffle(buf) for b in buf: yield b buf = [] if len(buf) > 0: random.shuffle(buf) for b in buf: yield b return create_reader_creator def chain(*reader_creators): """Creates a data reader creator whose output is the outputs of input data reader creators chained together. If input reader creators output following data entries: [0, 0, 0] [1, 1, 1] [2, 2, 2] The chained reader creator will output: [0, 0, 0, 1, 1, 1, 2, 2, 2] Args: readers_creators: input reader creators Returns: the new data reader creator. """ def create_reader_creator(): rs = [] for r in reader_creators: rs.append(r()) for e in itertools.chain(*rs): yield e return create_reader_creator class ComposeNotAligned: pass def compose(*reader_creators, **kwargs): """Creates a data reader creator whose output is the combination of input readers creators. If input reader creators output following data entries: (1, 2) 3 (4, 5) The composed reader creator will output: (1, 2, 3, 4, 5) Args: *reader_creators: reader creators that will be composed together. check_alignment: If True, will check if input reader creators are aligned correctly. If False, will not check alignment and trailing outputs will be discarded. Defaults to True. Returns: the new data reader creator. Raises: ComposeNotAligned: outputs of reader creators are not aligned. Will not raise when check_alignment is set to False. """ check_alignment = kwargs.pop('check_alignment', True) def make_tuple(x): if isinstance(x, tuple): return x else: return (x, ) def create_reader_creator(): rs = [] for r in reader_creators: rs.append(r()) if not check_alignment: for outputs in itertools.izip(*rs): yield sum(map(make_tuple, outputs), ()) else: for outputs in itertools.izip_longest(*rs): for o in outputs: if o is None: # None will be not be present if compose is aligned raise ComposeNotAligned yield sum(map(make_tuple, outputs), ()) return create_reader_creator def buffered(reader_creator, size): """Creates a buffered data reader creator. The buffered data reader creator will read and save data entries into a buffer. Reading from the buffered data reader creator will proceed as long as the buffer is not empty. Args: reader_creator: the data reader creator to read from. size: max buffer size. Returns: The buffered data reader creator. """ class EndSignal(): pass end = EndSignal() def read_worker(r, q): for d in r: q.put(d) q.put(end) def create_reader_creator(): r = reader_creator() q = Queue(maxsize=size) t = Thread( target=read_worker, args=( r, q, )) t.daemon = True t.start() e = q.get() while e != end: yield e e = q.get() return create_reader_creator