def _split_file(self, path: GcsfsFilePath,
file_contents_handle: GcsfsFileContentsHandle) -> None:
output_dir = GcsfsDirectoryPath.from_file_path(path)
upload_paths_and_df = []
for i, df in enumerate(
pd.read_csv(file_contents_handle.local_file_path,
dtype=str,
chunksize=self.file_split_line_limit,
keep_default_na=False)):
upload_path = self._create_split_file_path(path,
output_dir,
split_num=i)
upload_paths_and_df.append((upload_path, df))
for output_path, df in upload_paths_and_df:
logging.info("Writing file split [%s] to Cloud Storage.",
output_path.abs_path())
self.fs.upload_from_string(output_path, df.to_csv(index=False),
'text/csv')
logging.info("Done splitting file [%s] into [%s] paths, returning.",
path.abs_path(), len(upload_paths_and_df))
self.fs.mv_path_to_storage(path, self.storage_directory_path)
def test_processing_continues_if_there_are_subfolders_in_ingest_dir(self):
controller = build_gcsfs_controller_for_tests(
StateTestGcsfsDirectIngestController,
self.FIXTURE_PATH_PREFIX,
run_async=False)
if not isinstance(controller.fs, FakeDirectIngestGCSFileSystem):
raise ValueError(f"Controller fs must have type "
f"FakeDirectIngestGCSFileSystem. Found instead "
f"type [{type(controller.fs)}]")
subdir_path = \
path_for_fixture_file(controller, f'subdir/',
should_normalize=False)
paths = [
subdir_path,
path_for_fixture_file(controller,
f'subdir/Unexpected_Tag.csv',
should_normalize=False),
path_for_fixture_file(controller,
f'tagA.csv',
should_normalize=False),
path_for_fixture_file(controller,
f'tagB.csv',
should_normalize=False),
path_for_fixture_file(controller,
f'tagC.csv',
should_normalize=False),
path_for_fixture_file(controller,
f'subdir/tagC_2.csv',
should_normalize=False),
]
for path in paths:
controller.fs.test_add_path(path)
run_task_queues_to_empty(controller)
dir_paths_found = []
storage_file_paths = []
ingest_file_paths = []
for path in controller.fs.all_paths:
if isinstance(path, GcsfsDirectoryPath):
dir_paths_found.append(path)
continue
if path.abs_path().startswith(
controller.storage_directory_path.abs_path()):
storage_file_paths.append(path)
else:
self.assertTrue(path.abs_path().startswith(
controller.ingest_directory_path.abs_path()))
ingest_file_paths.append(path)
self.assertEqual(1, len(dir_paths_found))
self.assertEqual(subdir_path, dir_paths_found[0])
self.assertEqual(3, len(storage_file_paths))
storage_tags = {
filename_parts_from_path(path).file_tag
for path in storage_file_paths
}
self.assertEqual({'tagA', 'tagB', 'tagC'}, storage_tags)
for path in storage_file_paths:
self.assertTrue(controller.fs.is_normalized_file_path(path))
self.assertTrue(controller.fs.is_processed_file(path))
self.assertEqual(2, len(ingest_file_paths))
ingest_tags = {
filename_parts_from_path(path).file_tag
for path in ingest_file_paths
}
self.assertEqual({'tagC', 'Unexpected_Tag'}, ingest_tags)
for path in ingest_file_paths:
self.assertTrue(controller.fs.is_normalized_file_path(path))
self.assertTrue(controller.fs.is_seen_unprocessed_file(path))
self.assertEqual(subdir_path,
GcsfsDirectoryPath.from_file_path(path))
def _split_file_if_necessary(self, path: GcsfsFilePath) -> bool:
"""Checks if the given file needs to be split according to this controller's |file_split_line_limit|.
Returns True if the file was split, False if splitting was not necessary.
"""
should_split = self._should_split_file(path)
if not should_split:
logging.info("No need to split file path [%s].", path.abs_path())
return False
logging.info("Proceeding to file splitting for path [%s].",
path.abs_path())
original_metadata = None
if self.region.are_ingest_view_exports_enabled_in_env():
original_metadata = self.file_metadata_manager.get_file_metadata(
path)
output_dir = GcsfsDirectoryPath.from_file_path(path)
split_contents_paths = self._split_file(path)
upload_paths = []
for i, split_contents_path in enumerate(split_contents_paths):
upload_path = self._create_split_file_path(path,
output_dir,
split_num=i)
logging.info(
"Copying split [%s] to direct ingest directory at path [%s].",
i, upload_path.abs_path())
upload_paths.append(upload_path)
try:
self.fs.mv(split_contents_path, upload_path)
except Exception as e:
logging.error(
'Threw error while copying split files from temp bucket - attempting to clean up before rethrowing.'
' [%s]', e)
for p in upload_paths:
self.fs.delete(p)
raise e
# We wait to register files with metadata manager until all files have been successfully copied to avoid leaving
# the metadata manager in an inconsistent state.
if self.region.are_ingest_view_exports_enabled_in_env():
if not isinstance(original_metadata,
DirectIngestIngestFileMetadata):
raise ValueError(
'Attempting to split a non-ingest view type file')
logging.info(
'Registering [%s] split files with the metadata manager.',
len(upload_paths))
for upload_path in upload_paths:
ingest_file_metadata = self.file_metadata_manager.register_ingest_file_split(
original_metadata, upload_path)
self.file_metadata_manager.mark_ingest_view_exported(
ingest_file_metadata)
self.file_metadata_manager.mark_file_as_processed(path)
logging.info(
"Done splitting file [%s] into [%s] paths, moving it to storage.",
path.abs_path(), len(split_contents_paths))
self.fs.mv_path_to_storage(path, self.storage_directory_path)
return True
def _split_file_if_necessary(self, path: GcsfsFilePath):
"""Checks if the given file needs to be split according to this
controller's |file_split_line_limit|.
"""
parts = filename_parts_from_path(path)
if self.region.is_raw_vs_ingest_file_name_detection_enabled() and \
parts.file_type != GcsfsDirectIngestFileType.INGEST_VIEW:
raise ValueError(f'Should not be attempting to split files other than ingest view files, found path with '
f'file type: {parts.file_type}')
if parts.file_tag not in self.get_file_tag_rank_list():
logging.info("File tag [%s] for path [%s] not in rank list - "
"not splitting.",
parts.file_tag,
path.abs_path())
return False
if parts.is_file_split and \
parts.file_split_size and \
parts.file_split_size <= self.file_split_line_limit:
logging.info("File [%s] already split with size [%s].",
path.abs_path(), parts.file_split_size)
return False
file_contents_handle = self._get_contents_handle_from_path(path)
if not file_contents_handle:
logging.info("File [%s] has no rows - not splitting.",
path.abs_path())
return False
if self._can_proceed_with_ingest_for_contents(file_contents_handle):
logging.info("No need to split file path [%s].", path.abs_path())
return False
logging.info("Proceeding to file splitting for path [%s].",
path.abs_path())
split_contents_handles = self._split_file(path, file_contents_handle)
original_metadata = None
if self.region.are_ingest_view_exports_enabled_in_env():
original_metadata = self.file_metadata_manager.get_file_metadata(path)
output_dir = GcsfsDirectoryPath.from_file_path(path)
for i, split_contents_handle in enumerate(split_contents_handles):
upload_path = self._create_split_file_path(path, output_dir, split_num=i)
ingest_file_metadata = None
if self.region.are_ingest_view_exports_enabled_in_env():
if not isinstance(original_metadata, DirectIngestIngestFileMetadata):
raise ValueError('Attempting to split a non-ingest view type file')
ingest_file_metadata = self.file_metadata_manager.register_ingest_file_split(original_metadata,
upload_path)
logging.info("Writing file split [%s] to Cloud Storage.", upload_path.abs_path())
self.fs.upload_from_contents_handle(upload_path, split_contents_handle, self._contents_type())
if self.region.are_ingest_view_exports_enabled_in_env():
if not ingest_file_metadata:
raise ValueError(f'Split file metadata for path unexpectedly none [{upload_path.abs_path()}]')
self.file_metadata_manager.mark_ingest_view_exported(ingest_file_metadata)
if self.region.are_ingest_view_exports_enabled_in_env():
self.file_metadata_manager.mark_file_as_processed(path)
logging.info("Done splitting file [%s] into [%s] paths, moving it to storage.",
path.abs_path(), len(split_contents_handles))
self.fs.mv_path_to_storage(path, self.storage_directory_path)
return True
def _split_file_if_necessary(self, path: GcsfsFilePath) -> bool:
"""Checks if the given file needs to be split according to this controller's |file_split_line_limit|.
Returns True if the file was split, False if splitting was not necessary.
"""
should_split = self._should_split_file(path)
if not should_split:
logging.info("No need to split file path [%s].", path.abs_path())
return False
logging.info("Proceeding to file splitting for path [%s].",
path.abs_path())
original_metadata = None
if self.region.are_ingest_view_exports_enabled_in_env():
original_metadata = self.file_metadata_manager.get_file_metadata(
path)
output_dir = GcsfsDirectoryPath.from_file_path(path)
split_contents_paths = self._split_file(path)
for i, split_contents_path in enumerate(split_contents_paths):
upload_path = self._create_split_file_path(path,
output_dir,
split_num=i)
ingest_file_metadata = None
if self.region.are_ingest_view_exports_enabled_in_env():
if not isinstance(original_metadata,
DirectIngestIngestFileMetadata):
raise ValueError(
'Attempting to split a non-ingest view type file')
ingest_file_metadata = self.file_metadata_manager.register_ingest_file_split(
original_metadata, upload_path)
logging.info(
"Copying split [%s] to direct ingest directory at path [%s].",
i, upload_path.abs_path())
self.fs.mv(split_contents_path, upload_path)
if self.region.are_ingest_view_exports_enabled_in_env():
if not ingest_file_metadata:
raise ValueError(
f'Split file metadata for path unexpectedly none [{upload_path.abs_path()}]'
)
self.file_metadata_manager.mark_ingest_view_exported(
ingest_file_metadata)
if self.region.are_ingest_view_exports_enabled_in_env():
self.file_metadata_manager.mark_file_as_processed(path)
logging.info(
"Done splitting file [%s] into [%s] paths, moving it to storage.",
path.abs_path(), len(split_contents_paths))
self.fs.mv_path_to_storage(path, self.storage_directory_path)
return True
