def main(self):
# Specify options.
options = [("n=", "The number of protocol buffers to save.", True)]
# Start main method here.
command_line = "%s --n=number_to_print input_shard output_shard"
options_hash, remainder = parseCommandLine(options, command_line=command_line)
if len(remainder) != 2:
print usage(sys.argv, command_line, options)
sys.exit()
input_shard = remainder[0]
output_shard = remainder[1]
num_in_file = countProtosInFile(input_shard)
reader = ProtocolBufferFileReader(None, filename=input_shard, return_byte_string_only=True)
writer = ProtocolBufferFileWriter(filename=output_shard, messages_are_byte_strings=True)
num_messages_to_print = int(options_hash["n"])
first_message_to_print = num_in_file - num_messages_to_print
if first_message_to_print < 0:
first_message_to_print = 0
num_messages = 0
for message in reader:
if num_messages >= first_message_to_print:
writer.write(message)
num_messages += 1
writer.close()
reader.close()
def run(self):
# Specify options.
options = [
('num_records=', 'Number of records to select.', True),
]
# Start main method here.
command_line = '%s --num_records=n input_shard output_shard'
options_hash, remainder = parseCommandLine(options, command_line=command_line)
if (len(remainder) != 2):
print usage(sys.argv, command_line, options)
sys.exit()
num_records = int(options_hash['num_records'])
input_file = remainder[0]
output_file = remainder[1]
total_records = countProtosInFile(input_file)
print 'Selecting %d records from %d total records.' % (num_records, total_records)
random.seed()
# Randomly select some records to use.
records_to_use = set(random.sample(range(total_records), num_records))
reader = ProtocolBufferFileReader(None, filename=input_file, return_byte_string_only=True)
writer = ProtocolBufferFileWriter(filename=output_file, messages_are_byte_strings=True)
for ii, message in enumerate(reader):
if ii in records_to_use:
writer.write(message)
reader.close()
writer.close()
def countFiles(self, files, verbose=False):
total_records = 0
for file in files:
num_records = countProtosInFile(file)
if verbose:
print '%s\t%d' % (file, num_records)
total_records += num_records
if len(files) > 1 and verbose:
print 'Total: %d' % (total_records)
return total_records
def countFiles(self, files, verbose=False): total_records = 0 for file in files: num_records = countProtosInFile(file) if verbose: print '%s\t%d' % (file, num_records) total_records += num_records if len(files) > 1 and verbose: print 'Total: %d' % (total_records) return total_records
def run(self):
# Specify options.
options = [
('num_records=', 'Number of records to select.', True),
]
# Start main method here.
command_line = '%s --num_records=n input_shard output_shard'
options_hash, remainder = parseCommandLine(options,
command_line=command_line)
if (len(remainder) != 2):
print usage(sys.argv, command_line, options)
sys.exit()
num_records = int(options_hash['num_records'])
input_file = remainder[0]
output_file = remainder[1]
total_records = countProtosInFile(input_file)
print 'Selecting %d records from %d total records.' % (num_records,
total_records)
random.seed()
# Randomly select some records to use.
records_to_use = set(random.sample(range(total_records), num_records))
reader = ProtocolBufferFileReader(None,
filename=input_file,
return_byte_string_only=True)
writer = ProtocolBufferFileWriter(filename=output_file,
messages_are_byte_strings=True)
for ii, message in enumerate(reader):
if ii in records_to_use:
writer.write(message)
reader.close()
writer.close()
def main(self):
# Specify options.
options = [
('n=', 'The number of protocol buffers to save.', True),
]
# Start main method here.
command_line = '%s --n=number_to_print input_shard output_shard'
options_hash, remainder = parseCommandLine(options,
command_line=command_line)
if (len(remainder) != 2):
print usage(sys.argv, command_line, options)
sys.exit()
input_shard = remainder[0]
output_shard = remainder[1]
num_in_file = countProtosInFile(input_shard)
reader = ProtocolBufferFileReader(None,
filename=input_shard,
return_byte_string_only=True)
writer = ProtocolBufferFileWriter(filename=output_shard,
messages_are_byte_strings=True)
num_messages_to_print = int(options_hash['n'])
first_message_to_print = num_in_file - num_messages_to_print
if first_message_to_print < 0:
first_message_to_print = 0
num_messages = 0
for message in reader:
if num_messages >= first_message_to_print:
writer.write(message)
num_messages += 1
writer.close()
reader.close()