def next_file(self):
if self.__input_index == len(self.__inputs):
raise StopIteration()
file_input = self.__inputs[self.__input_index]
if file_input.range:
first_format = self.__file_format_root._formats[0]
if not first_format.can_split():
raise ValueError("Input range specified for a non splitable format %s" % first_format.NAME)
first_format._range = file_input.range
self.__previous_input_index = self.__input_index
self.__input_index += 1
return files.open(file_input.filename, "r", buffering=-1)
def next_file(self):
if self.__input_index == len(self.__inputs):
raise StopIteration()
file_input = self.__inputs[self.__input_index]
if file_input.range:
first_format = self.__file_format_root._formats[0]
if not first_format.can_split():
raise ValueError('Input range specified for a non splitable format %s'
% first_format.NAME)
first_format._range = file_input.range
self.__previous_input_index = self.__input_index
self.__input_index += 1
return files.open(file_input.filename, 'r', buffering=-1)
def _deep_split(filenames, size_per_shard, parsed_formats):
"""Split files into roots using the first FileFormat.
Deep split can split within a file. It tells the first format how big
a split it wants and the first format will do the actually splitting
because only the first format knows how to operate on this particular
format.
Args:
filenames: a list of input filenames.
size_per_shard: size per shard.
parsed_format: the parsed FileFormats.
Returns:
A list of FileFormatRoot.
"""
roots = []
inputs = []
size_left = size_per_shard
for filename in filenames:
index = 0
with files.open(filename) as f:
cache_for_split = {}
# Split a single file.
while True:
if size_left <= 0:
# Shard has been filled.
roots.append(FileFormatRoot(copy.deepcopy(parsed_formats), inputs))
size_left = size_per_shard
inputs = []
start_index = index
size_left, index = parsed_formats[0].split(size_left,
start_index,
f,
cache_for_split)
# File has been entirely covered.
if start_index == index:
break
inputs.append(_FileRange(filename, (start_index, index)))
if inputs:
roots.append(FileFormatRoot(copy.deepcopy(parsed_formats), inputs))
return roots
def _deep_split(filenames, size_per_shard, parsed_formats):
"""Split files into roots using the first FileFormat.
Deep split can split within a file. It tells the first format how big
a split it wants and the first format will do the actually splitting
because only the first format knows how to operate on this particular
format.
Args:
filenames: a list of input filenames.
size_per_shard: size per shard.
parsed_format: the parsed FileFormats.
Returns:
A list of FileFormatRoot.
"""
roots = []
inputs = []
size_left = size_per_shard
for filename in filenames:
index = 0
with files.open(filename) as f:
cache_for_split = {}
while True:
if size_left <= 0:
roots.append(FileFormatRoot(copy.deepcopy(parsed_formats), inputs))
size_left = size_per_shard
inputs = []
start_index = index
size_left, index = parsed_formats[0].split(size_left,
start_index,
f,
cache_for_split)
if start_index == index:
break
inputs.append(_FileRange(filename, (start_index, index)))
if inputs:
roots.append(FileFormatRoot(copy.deepcopy(parsed_formats), inputs))
return roots
