def _prevalidate_file(config, file_descriptor, file_path, file_type, run_id,
schema, files_to_delete, logger):
"""Pre-validate the input file using the CSV validator."""
split_file_basename = '{0}_import_{1}_split'.format(file_type, run_id)
split_files = list(
split_file(file_descriptor, config.import_config.batch_size,
dirname(file_path), logger, split_file_basename))
files_to_delete.extend(split_files)
num_batches = len(split_files)
num_validated_batches = 0
with futures.ThreadPoolExecutor(max_workers=config.multiprocessing_config.
max_local_cpus) as prevalidator:
logger.info(
'Pre-validating file {0} ({1} pre-validation workers)'.format(
basename(file_path),
config.multiprocessing_config.max_local_cpus))
tasks = [
prevalidator.submit(prevalidate_file, f, schema)
for f in split_files
]
for f in futures.as_completed(tasks):
f.result() # will throw exception if this one was thrown in thread
num_validated_batches += 1
lvl = logging.DEBUG
if num_validated_batches % 50 == 0 or num_validated_batches == num_batches:
lvl = logging.INFO # Only print every 50 batches at INFO so as not to spam console
logger.log(
lvl,
'Pre-validated {validated_batches} of {num_batches} batches'.
format(validated_batches=num_validated_batches,
num_batches=num_batches))
logger.info('Successfully pre-validated file: {0}'.format(
basename(file_path)))
return True
def _split_file(self, file_to_split):
"""Method which unzips the input file into split files in the work directory."""
split_file_basename = '{0}_import_{1}_split'.format(
self._import_type, self.import_id)
num_batches = 0
for batch_filename in split_file(file_to_split, self._batch_size,
self._extract_dir, self._logger,
split_file_basename):
self._files_to_delete.append(batch_filename)
num_batches += 1
yield batch_filename
self._logger.info(
'Finished splitting input file into {num_batches} batches'.format(
num_batches=num_batches))
return
