def test_paths(self):
config = config_load()
self.assertTrue(len(get_index_path(config)) > 0)
self.assertTrue(len(get_objects_path(config)) > 0)
self.assertTrue(len(get_cache_path(config)) > 0)
self.assertTrue(len(get_metadata_path(config)) > 0)
self.assertTrue('.ml-git' in get_refs_path(config))
def garbage_collector(self):
any_metadata = False
removed_files = 0
reclaimed_space = 0
for entity in EntityType:
repo_type = entity.value
if self.metadata_exists(repo_type):
log.info(output_messages['INFO_STARTING_GC'] % repo_type,
class_name=REPOSITORY_CLASS_NAME)
any_metadata = True
index_path = get_index_path(self.__config, repo_type)
objects_path = get_objects_path(self.__config, repo_type)
blobs_hashes = self._get_blobs_hashes(index_path, objects_path,
repo_type)
cache = Cache(get_cache_path(self.__config, repo_type))
count_removed_cache, reclaimed_cache_space = cache.garbage_collector(
blobs_hashes)
objects = Objects('', objects_path)
count_removed_objects, reclaimed_objects_space = objects.garbage_collector(
blobs_hashes)
reclaimed_space += reclaimed_objects_space + reclaimed_cache_space
removed_files += count_removed_objects + count_removed_cache
if not any_metadata:
log.error(output_messages['ERROR_UNINITIALIZED_METADATA'],
class_name=REPOSITORY_CLASS_NAME)
return
log.info(output_messages['INFO_REMOVED_FILES'] %
(humanize.intword(removed_files),
os.path.join(get_root_path(), '.ml-git')),
class_name=REPOSITORY_CLASS_NAME)
log.info(output_messages['INFO_RECLAIMED_SPACE'] %
humanize.naturalsize(reclaimed_space),
class_name=REPOSITORY_CLASS_NAME)
def log(self, spec, stat=False, fullstat=False):
try:
repo_type = self.__repo_type
metadata_path = get_metadata_path(self.__config, repo_type)
metadata = Metadata(spec, metadata_path, self.__config, repo_type)
index_path = get_index_path(self.__config, repo_type)
log_info = metadata.get_log_info(spec, fullstat)
except Exception as e:
log.error(e, class_name=REPOSITORY_CLASS_NAME)
return
fidx = FullIndex(spec, index_path)
if stat or fullstat:
workspace_size = fidx.get_total_size()
amount_message = 'Total of files: %s' % fidx.get_total_count()
size_message = 'Workspace size: %s' % humanize.naturalsize(
workspace_size)
workspace_info = '------------------------------------------------- \n{}\t{}' \
.format(amount_message, size_message)
log_info = '{}\n{}'.format(log_info, workspace_info)
log.info(log_info, class_name=REPOSITORY_CLASS_NAME)
def unlock_file(self, spec, file_path):
repo_type = self.__repo_type
if not validate_config_spec_hash(self.__config):
log.error(
'.ml-git/config.yaml invalid. It should look something like this:\n%s'
% get_yaml_str(
get_sample_config_spec('somebucket', 'someprofile',
'someregion')),
class_name=REPOSITORY_CLASS_NAME)
return None
path, file = None, None
try:
refs_path = get_refs_path(self.__config, repo_type)
objects_path = get_objects_path(self.__config, repo_type)
index_path = get_index_path(self.__config, repo_type)
cache_path = get_cache_path(self.__config, repo_type)
ref = Refs(refs_path, spec, repo_type)
tag, sha = ref.branch()
categories_path = get_path_with_categories(tag)
path, file = search_spec_file(self.__repo_type, spec,
categories_path)
except Exception as e:
log.error(e, class_name=REPOSITORY_CLASS_NAME)
return
if path is None:
return
spec_path = os.path.join(path, file)
spec_file = yaml_load(spec_path)
try:
mutability = spec_file[repo_type]['mutability']
if mutability not in Mutability.list():
log.error('Invalid mutability type.',
class_name=REPOSITORY_CLASS_NAME)
return
except Exception:
log.info(
'The spec does not have the \'mutability\' property set. Default: strict.',
class_name=REPOSITORY_CLASS_NAME)
return
if mutability != Mutability.STRICT.value:
try:
local = LocalRepository(self.__config, objects_path, repo_type)
local.unlock_file(path, file_path, index_path, objects_path,
spec, cache_path)
except Exception as e:
log.error(e, class_name=REPOSITORY_CLASS_NAME)
return
else:
log.error(
'You cannot use this command for this entity because mutability cannot be strict.',
class_name=REPOSITORY_CLASS_NAME)
def fsck(self):
repo_type = self.__repo_type
try:
objects_path = get_objects_path(self.__config, repo_type)
index_path = get_index_path(self.__config, repo_type)
except RootPathException:
return
o = Objects('', objects_path)
corrupted_files_obj = o.fsck()
corrupted_files_obj_len = len(corrupted_files_obj)
idx = MultihashIndex('', index_path, objects_path)
corrupted_files_idx = idx.fsck()
corrupted_files_idx_len = len(corrupted_files_idx)
print('[%d] corrupted file(s) in Local Repository: %s' %
(corrupted_files_obj_len, corrupted_files_obj))
print('[%d] corrupted file(s) in Index: %s' %
(corrupted_files_idx_len, corrupted_files_idx))
print('Total of corrupted files: %d' %
(corrupted_files_obj_len + corrupted_files_idx_len))
def reset(self, spec, reset_type, head):
log.info(output_messages['INFO_INITIALIZING_RESET'] %
(reset_type, head),
class_name=REPOSITORY_CLASS_NAME)
if (reset_type == '--soft'
or reset_type == '--mixed') and head == HEAD:
return
try:
repo_type = self.__repo_type
metadata_path = get_metadata_path(self.__config, repo_type)
index_path = get_index_path(self.__config, repo_type)
refs_path = get_refs_path(self.__config, repo_type)
object_path = get_objects_path(self.__config, repo_type)
met = Metadata(spec, metadata_path, self.__config, repo_type)
ref = Refs(refs_path, spec, repo_type)
idx = MultihashIndex(spec, index_path, object_path)
fidx = FullIndex(spec, index_path)
except Exception as e:
log.error(e, class_name=REPOSITORY_CLASS_NAME)
return
# get tag before reset
tag = met.get_current_tag()
categories_path = get_path_with_categories(str(tag))
# current manifest file before reset
manifest_path = os.path.join(metadata_path, categories_path, spec,
MANIFEST_FILE)
_manifest = Manifest(manifest_path).load()
if head == HEAD_1: # HEAD~1
try:
# reset the repo
met.reset()
except Exception:
return
# get tag after reset
tag_after_reset = met.get_current_tag()
sha = met.sha_from_tag(tag_after_reset)
# update ml-git ref HEAD
ref.update_head(str(tag_after_reset), sha)
# # get path to reset workspace in case of --hard
path, file = None, None
try:
path, file = search_spec_file(self.__repo_type, spec,
categories_path)
except Exception as e:
log.error(e, class_name=REPOSITORY_CLASS_NAME)
if reset_type == '--hard' and path is None:
return
# get manifest from metadata after reset
_manifest_changed = Manifest(manifest_path)
hash_files, file_names = _manifest_changed.get_diff(_manifest)
idx_mf = idx.get_index().load()
if reset_type == '--soft':
# add in index/metadata/<entity-name>/MANIFEST
idx.update_index_manifest(idx_mf)
idx.update_index_manifest(hash_files)
fidx.update_index_status(file_names, Status.a.name)
else: # --hard or --mixed
# remove hash from index/hashsh/store.log
file_names.update(*idx_mf.values())
objs = MultihashFS(index_path)
for key_hash in hash_files:
objs.remove_hash(key_hash)
idx.remove_manifest()
fidx.remove_from_index_yaml(file_names)
fidx.remove_uncommitted()
if reset_type == '--hard': # reset workspace
remove_from_workspace(file_names, path, spec)
def add(self, spec, file_path, bump_version=False, run_fsck=False):
repo_type = self.__repo_type
is_shared_objects = 'objects_path' in self.__config[repo_type]
is_shared_cache = 'cache_path' in self.__config[repo_type]
if not validate_config_spec_hash(self.__config):
log.error(
'.ml-git/config.yaml invalid. It should look something like this:\n%s'
% get_yaml_str(
get_sample_config_spec('somebucket', 'someprofile',
'someregion')),
class_name=REPOSITORY_CLASS_NAME)
return None
path, file = None, None
try:
refs_path = get_refs_path(self.__config, repo_type)
index_path = get_index_path(self.__config, repo_type)
metadata_path = get_metadata_path(self.__config, repo_type)
cache_path = get_cache_path(self.__config, repo_type)
objects_path = get_objects_path(self.__config, repo_type)
repo = LocalRepository(self.__config, objects_path, repo_type)
mutability, check_mutability = repo.get_mutability_from_spec(
spec, repo_type)
sampling_flag = os.path.exists(
os.path.join(index_path, 'metadata', spec, 'sampling'))
if sampling_flag:
log.error(
'You cannot add new data to an entity that is based on a checkout with the --sampling option.',
class_name=REPOSITORY_CLASS_NAME)
return
if not mutability:
return
if not check_mutability:
log.error('Spec mutability cannot be changed.',
class_name=REPOSITORY_CLASS_NAME)
return
if not self._has_new_data(repo, spec):
return None
ref = Refs(refs_path, spec, repo_type)
tag, sha = ref.branch()
categories_path = get_path_with_categories(tag)
path, file = search_spec_file(self.__repo_type, spec,
categories_path)
except Exception as e:
log.error(e, class_name=REPOSITORY_CLASS_NAME)
return
if path is None:
return
spec_path = os.path.join(path, file)
if not self._is_spec_valid(spec_path):
return None
# Check tag before anything to avoid creating unstable state
log.debug('Repository: check if tag already exists',
class_name=REPOSITORY_CLASS_NAME)
m = Metadata(spec, metadata_path, self.__config, repo_type)
if not m.check_exists():
log.error('The %s has not been initialized' % self.__repo_type,
class_name=REPOSITORY_CLASS_NAME)
return
try:
m.update()
except Exception:
pass
# get version of current manifest file
manifest = self._get_current_manifest_file(m, tag)
try:
# adds chunks to ml-git Index
log.info('%s adding path [%s] to ml-git index' % (repo_type, path),
class_name=REPOSITORY_CLASS_NAME)
with change_mask_for_routine(is_shared_objects):
idx = MultihashIndex(spec, index_path, objects_path,
mutability, cache_path)
idx.add(path, manifest, file_path)
# create hard links in ml-git Cache
self.create_hard_links_in_cache(cache_path, index_path,
is_shared_cache, mutability, path,
spec)
except Exception as e:
log.error(e, class_name=REPOSITORY_CLASS_NAME)
return None
if bump_version and not increment_version_in_spec(
spec_path, self.__repo_type):
return None
idx.add_metadata(path, file)
self._check_corrupted_files(spec, repo)
# Run file check
if run_fsck:
self.fsck()
def commit(self, spec, specs, version=None, run_fsck=False, msg=None):
# Move chunks from index to .ml-git/objects
repo_type = self.__repo_type
try:
index_path = get_index_path(self.__config, repo_type)
objects_path = get_objects_path(self.__config, repo_type)
metadata_path = get_metadata_path(self.__config, repo_type)
refs_path = get_refs_path(self.__config, repo_type)
repo = LocalRepository(self.__config, objects_path, repo_type)
mutability, check_mutability = repo.get_mutability_from_spec(
spec, repo_type)
if not mutability:
return
if not check_mutability:
log.error('Spec mutability cannot be changed.',
class_name=REPOSITORY_CLASS_NAME)
return
except Exception as e:
log.error(e, class_name=REPOSITORY_CLASS_NAME)
return
ref = Refs(refs_path, spec, repo_type)
tag, sha = ref.branch()
categories_path = get_path_with_categories(tag)
manifest_path = os.path.join(metadata_path, categories_path, spec,
MANIFEST_FILE)
path, file = None, None
try:
path, file = search_spec_file(self.__repo_type, spec,
categories_path)
except Exception as e:
log.error(e, class_name=REPOSITORY_CLASS_NAME)
if path is None:
return None, None, None
spec_path = os.path.join(path, file)
idx = MultihashIndex(spec, index_path, objects_path)
if version:
set_version_in_spec(version, spec_path, self.__repo_type)
idx.add_metadata(path, file)
# Check tag before anything to avoid creating unstable state
log.debug('Check if tag already exists',
class_name=REPOSITORY_CLASS_NAME)
m = Metadata(spec, metadata_path, self.__config, repo_type)
if not m.check_exists():
log.error('The %s has not been initialized' % self.__repo_type,
class_name=REPOSITORY_CLASS_NAME)
return
full_metadata_path, categories_sub_path, metadata = m.tag_exists(
index_path)
if metadata is None:
return None
log.debug('%s -> %s' % (index_path, objects_path),
class_name=REPOSITORY_CLASS_NAME)
# commit objects in index to ml-git objects
o = Objects(spec, objects_path)
changed_files, deleted_files = o.commit_index(index_path, path)
bare_mode = os.path.exists(
os.path.join(index_path, 'metadata', spec, 'bare'))
if not bare_mode:
manifest = m.get_metadata_manifest(manifest_path)
self._remove_deleted_files(idx, index_path, m, manifest, spec,
deleted_files)
m.remove_files_added_after_base_tag(manifest, path)
else:
tag, _ = ref.branch()
self._checkout_ref(tag)
# update metadata spec & README.md
# option --dataset-spec --labels-spec
tag, sha = m.commit_metadata(index_path, specs, msg, changed_files,
mutability, path)
# update ml-git ref spec HEAD == to new SHA-1 / tag
if tag is None:
return None
ref = Refs(refs_path, spec, repo_type)
ref.update_head(tag, sha)
# Run file check
if run_fsck:
self.fsck()
return tag
