普通文本  |  197行  |  7.97 KB

# Copyright 2014 Google Inc. 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.
"""Helper class for streaming resumable uploads."""

import collections
import os

from gslib.exception import CommandException
from gslib.util import GetJsonResumableChunkSize


class ResumableStreamingJsonUploadWrapper(object):
  """Wraps an input stream in a buffer for resumable uploads.

  This class takes a non-seekable input stream, buffers it, and exposes it
  as a stream with limited seek capabilities such that it can be used in a
  resumable JSON API upload.

  max_buffer_size bytes of buffering is supported.
  """

  def __init__(self, stream, max_buffer_size, test_small_buffer=False):
    """Initializes the wrapper.

    Args:
      stream: Input stream.
      max_buffer_size: Maximum size of internal buffer; should be >= the chunk
          size of the resumable upload API to ensure that at least one full
          chunk write can be replayed in the event of a server error.
      test_small_buffer: Skip check for buffer size vs. chunk size, for testing.
    """
    self._orig_fp = stream

    if not test_small_buffer and max_buffer_size < GetJsonResumableChunkSize():
      raise CommandException('Resumable streaming upload created with buffer '
                             'size %s, JSON resumable upload chunk size %s. '
                             'Buffer size must be >= JSON resumable upload '
                             'chunk size to ensure that uploads can be '
                             'resumed.' % (max_buffer_size,
                                           GetJsonResumableChunkSize()))

    self._max_buffer_size = max_buffer_size
    self._buffer = collections.deque()
    self._buffer_start = 0
    self._buffer_end = 0
    self._position = 0

  def read(self, size=-1):  # pylint: disable=invalid-name
    """"Reads from the wrapped stream.

    Args:
      size: The amount of bytes to read. If omitted or negative, the entire
          contents of the stream will be read and returned.

    Returns:
      Bytes from the wrapped stream.
    """
    read_all_bytes = size is None or size < 0
    if read_all_bytes:
      bytes_remaining = self._max_buffer_size
    else:
      bytes_remaining = size
    data = b''
    buffered_data = []
    if self._position < self._buffer_end:
      # There was a backwards seek, so read from the buffer first.

      # TODO: Performance test to validate if it is worth re-aligning
      # the buffers in this case.  Also, seeking through the buffer for
      # each read on a long catch-up is probably not performant, but we'd
      # need a more complex data structure than a deque to get around this.
      pos_in_buffer = self._buffer_start
      buffer_index = 0
      # First, find the start position in the buffer.
      while pos_in_buffer + len(self._buffer[buffer_index]) < self._position:
        # When this loop exits, buffer_index will refer to a buffer that
        # has at least some overlap with self._position, and
        # pos_in_buffer will be >= self._position
        pos_in_buffer += len(self._buffer[buffer_index])
        buffer_index += 1

      # Read until we've read enough or we're out of buffer.
      while pos_in_buffer < self._buffer_end and bytes_remaining > 0:
        buffer_len = len(self._buffer[buffer_index])
        # This describes how far into the current buffer self._position is.
        offset_from_position = self._position - pos_in_buffer
        bytes_available_this_buffer = buffer_len - offset_from_position
        read_size = min(bytes_available_this_buffer, bytes_remaining)
        buffered_data.append(
            self._buffer[buffer_index]
            [offset_from_position:offset_from_position + read_size])
        bytes_remaining -= read_size
        pos_in_buffer += buffer_len
        buffer_index += 1
        self._position += read_size

    # At this point we're guaranteed that if there are any bytes left to read,
    # then self._position == self._buffer_end, and we can read from the
    # wrapped stream if needed.
    if read_all_bytes:
      # TODO: The user is requesting reading until the end of an
      # arbitrary length stream, which is bad we'll need to return data
      # with no size limits; if the stream is sufficiently long, we could run
      # out of memory. We could break this down into smaller reads and
      # buffer it as we go, but we're still left returning the data all at
      # once to the caller.  We could raise, but for now trust the caller to
      # be sane and have enough memory to hold the remaining stream contents.
      new_data = self._orig_fp.read(size)
      data_len = len(new_data)
      if not buffered_data:
        data = new_data
      else:
        buffered_data.append(new_data)
        data = b''.join(buffered_data)
      self._position += data_len
    elif bytes_remaining:
      new_data = self._orig_fp.read(bytes_remaining)
      if not buffered_data:
        data = new_data
      else:
        buffered_data.append(new_data)
        data = b''.join(buffered_data)
      data_len = len(new_data)
      if data_len:
        self._position += data_len
        self._buffer.append(new_data)
        self._buffer_end += data_len
        oldest_data = None
        while self._buffer_end - self._buffer_start > self._max_buffer_size:
          oldest_data = self._buffer.popleft()
          self._buffer_start += len(oldest_data)
        if oldest_data:
          refill_amount = self._max_buffer_size - (self._buffer_end -
                                                   self._buffer_start)
          if refill_amount:
            self._buffer.appendleft(oldest_data[-refill_amount:])
            self._buffer_start -= refill_amount
    else:
      data = b''.join(buffered_data) if buffered_data else b''

    return data

  def tell(self):  # pylint: disable=invalid-name
    """Returns the current stream position."""
    return self._position

  def seekable(self):  # pylint: disable=invalid-name
    """Returns true since limited seek support exists."""
    return True

  def seek(self, offset, whence=os.SEEK_SET):  # pylint: disable=invalid-name
    """Seeks on the buffered stream.

    Args:
      offset: The offset to seek to; must be within the buffer bounds.
      whence: Must be os.SEEK_SET.

    Raises:
      CommandException if an unsupported seek mode or position is used.
    """
    if whence == os.SEEK_SET:
      if offset < self._buffer_start or offset > self._buffer_end:
        raise CommandException('Unable to resume upload because of limited '
                               'buffering available for streaming uploads. '
                               'Offset %s was requested, but only data from '
                               '%s to %s is buffered.' %
                               (offset, self._buffer_start, self._buffer_end))
      # Move to a position within the buffer.
      self._position = offset
    elif whence == os.SEEK_END:
      if offset > self._max_buffer_size:
        raise CommandException('Invalid SEEK_END offset %s on streaming '
                               'upload. Only %s can be buffered.' %
                               (offset, self._max_buffer_size))
      # Read to the end and rely on buffering to handle the offset.
      while self.read(self._max_buffer_size):
        pass
      # Now we're at the end.
      self._position -= offset
    else:
      raise CommandException('Invalid seek mode on streaming upload. '
                             '(mode %s, offset %s)' % (whence, offset))

  def close(self):  # pylint: disable=invalid-name
    return self._orig_fp.close()