def storage_path_hierarchy_test(self):
""" StoragePath should be read first from global and then from cloud section """
c = ("[Global]", "LogLevel =", "DataDir = ", "CacheDir = ",
"StateDir = ", "Cloud = aws",
"StoragePath = globalsb/global_storage_path", "[AWS]",
"StoragePath = awssb/aws_storage_path", "CredentialPath =",
"[GCP]", "StoragePath = googlesb/google_storage_path")
s = StringIO('\n'.join(c))
conf = self._conf(s)
cloud = DataCloud(None, conf)
self.assertEqual(cloud.typ, 'AWS')
self.assertEqual(cloud._cloud.storage_bucket, 'globalsb')
self.assertEqual(cloud._cloud.storage_prefix, 'global_storage_path')
c = ("[Global]", "LogLevel =", "DataDir = ", "CacheDir = ",
"StateDir = ", "Cloud = Aws", "[AWS]",
"StoragePath = awssb/aws_storage_path", "CredentialPath =",
"[GCP]", "StoragePath = googlesb/google_storage_path")
s = StringIO('\n'.join(c))
conf = self._conf(s)
cloud = DataCloud(None, conf)
self.assertEqual(cloud.typ, 'AWS')
self.assertEqual(cloud._cloud.storage_bucket, 'awssb')
self.assertEqual(cloud._cloud.storage_prefix, 'aws_storage_path')
def __init__(self, root_dir):
self.root_dir = os.path.abspath(os.path.realpath(root_dir))
self.dvc_dir = os.path.join(self.root_dir, self.DVC_DIR)
self.scm = SCM(self.root_dir)
self.lock = Lock(self.dvc_dir)
self.cache = Cache(self.dvc_dir)
self.state = State(self.root_dir, self.dvc_dir)
self.config = Config(self.dvc_dir)
self.logger = Logger(self.config._config)
self.cloud = DataCloud(self.cache, self.config._config)
def test(self):
for k,v in DataCloud.CLOUD_MAP.items():
config = {Config.SECTION_CORE: {'Cloud': k},
k: {'StoragePath': 'a/b',
'ProjectName': 'name'}}
cloud = DataCloud(self.dvc.cache, config)
self.assertIsInstance(cloud._cloud, v)
with self.assertRaises(ConfigError) as cx:
config = {Config.SECTION_CORE: {'Cloud': 'not_supported_type'}}
DataCloud(self.dvc.cache, config)
def __init__(self, root_dir):
self.root_dir = os.path.abspath(os.path.realpath(root_dir))
self.dvc_dir = os.path.join(self.root_dir, self.DVC_DIR)
self.scm = SCM(self.root_dir)
self.lock = Lock(self.dvc_dir)
self.cache = Cache(self.dvc_dir)
self.state = State(self.root_dir, self.dvc_dir)
self.config = Config(self.dvc_dir)
self.logger = Logger(self.config._config[Config.SECTION_CORE].get(
Config.SECTION_CORE_LOGLEVEL, None))
self.cloud = DataCloud(cache=self.cache,
state=self.state,
config=self.config._config)
def aws_credentials_default_tes_t_DISABLED(self):
""" in absence of [AWS] -> CredentialPath, aws creds should be read from ~/.aws/credentials """
default_path = os.path.expanduser('~/.aws/credentials')
c = (
"[Global]",
"LogLevel =",
"DataDir = ",
"CacheDir = ",
"StateDir = ",
"Cloud = aws",
"[AWS]",
"StoragePath = awssb/aws_storage_path",
"CredentialPath =",
)
s = StringIO('\n'.join(c))
patcher = mock.patch(
builtin_module_name + '.open',
side_effect=self.mocked_open_aws_default_credentials)
# patcher.start()
conf = self._conf(s)
cloud = DataCloud(None, conf)
aws_creds = cloud._cloud._get_credentials()
patcher.stop()
self.assertEqual(aws_creds[0], 'default_access_id')
self.assertEqual(aws_creds[1], 'default_sekret')
def aws_credentials_specified_tes_t_DISABLED(self, isfile_function):
""" in presence of [AWS] -> CredentialPath, use those credentials """
c = (
"[Global]",
"LogLevel =",
"DataDir = ",
"CacheDir = ",
"StateDir = ",
"Cloud = aws",
"[AWS]",
"StoragePath = awssb/aws_storage_path",
"CredentialPath = some_credential_path",
)
s = StringIO('\n'.join(c))
patcher = mock.patch(
builtin_module_name + '.open',
side_effect=self.mocked_open_aws_default_credentials)
patcher.start()
conf = self._conf(s)
cloud = DataCloud(None, conf)
aws_creds = cloud._cloud._get_credentials()
patcher.stop()
self.assertEqual(aws_creds[0], 'override_access_id')
self.assertEqual(aws_creds[1], 'override_sekret')
def test_unsupported(self):
with self.assertRaises(ConfigError) as cx:
config = TEST_CONFIG
config[TEST_SECTION][
Config.SECTION_REMOTE_URL] = 'notsupportedscheme://a/b'
DataCloud(cache=self.dvc.cache,
state=self.dvc.state,
config=config)
def __init__(self, root_dir):
self.root_dir = os.path.abspath(os.path.realpath(root_dir))
self.dvc_dir = os.path.join(self.root_dir, self.DVC_DIR)
self.config = Config(self.dvc_dir)
self.scm = SCM(self.root_dir)
self.lock = Lock(self.dvc_dir)
self.cache = Cache(self.root_dir, self.dvc_dir, cache_dir=self.config._config[Config.SECTION_CACHE].get(Config.SECTION_CACHE_DIR, None),
cache_type=self.config._config[Config.SECTION_CACHE].get(Config.SECTION_CACHE_TYPE, None))
self.state = State(self.dvc_dir)
self.logger = Logger(self.config._config[Config.SECTION_CORE].get(Config.SECTION_CORE_LOGLEVEL, None))
self.cloud = DataCloud(cache=self.cache, config=self.config._config)
self.updater = Updater(self.dvc_dir)
self._ignore()
self.updater.check()
def storage_path_hierarchy_test(self):
""" StoragePath should be read first from global and then from cloud section """
c = ("[core]", "loglevel =", "cloud = aws",
"storagepath = globalsb/global_storage_path", "[aws]",
"storagepath = awssb/aws_storage_path", "credentialpath =",
"[gcp]", "storagepath = googlesb/google_storage_path")
s = StringIO('\n'.join(c))
conf = self._conf(s)
cloud = DataCloud(None, conf)
self.assertIsInstance(cloud._cloud, DataCloudAWS)
self.assertEqual(cloud._cloud.storage_bucket, 'globalsb')
self.assertEqual(cloud._cloud.storage_prefix, 'global_storage_path')
c = ("[core]", "loglevel =", "cloud = Aws", "[aws]",
"storagepath = awssb/aws_storage_path", "credentialpath =",
"[gcp]", "storagepath = googlesb/google_storage_path")
s = StringIO('\n'.join(c))
conf = self._conf(s)
cloud = DataCloud(None, conf)
self.assertIsInstance(cloud._cloud, DataCloudAWS)
self.assertEqual(cloud._cloud.storage_bucket, 'awssb')
self.assertEqual(cloud._cloud.storage_prefix, 'aws_storage_path')
class Project(object):
DVC_DIR = '.dvc'
def __init__(self, root_dir):
self.root_dir = os.path.abspath(os.path.realpath(root_dir))
self.dvc_dir = os.path.join(self.root_dir, self.DVC_DIR)
self.scm = SCM(self.root_dir)
self.lock = Lock(self.dvc_dir)
self.cache = Cache(self.dvc_dir)
self.state = State(self.root_dir, self.dvc_dir)
self.config = Config(self.dvc_dir)
self.logger = Logger(self.config._config[Config.SECTION_CORE].get(
Config.SECTION_CORE_LOGLEVEL, None))
self.cloud = DataCloud(cache=self.cache,
state=self.state,
config=self.config._config)
@staticmethod
def init(root_dir=os.curdir):
"""
Initiate dvc project in directory.
Args:
root_dir: Path to project's root directory.
Returns:
Project instance.
Raises:
KeyError: Raises an exception.
"""
root_dir = os.path.abspath(root_dir)
dvc_dir = os.path.join(root_dir, Project.DVC_DIR)
os.mkdir(dvc_dir)
config = Config.init(dvc_dir)
cache = Cache.init(dvc_dir)
state = State.init(root_dir, dvc_dir)
lock = Lock(dvc_dir)
scm = SCM(root_dir)
scm.ignore_list([cache.cache_dir, state.state_file, lock.lock_file])
ignore_file = os.path.join(dvc_dir, scm.ignore_file())
scm.add([config.config_file, ignore_file])
return Project(root_dir)
def to_dvc_path(self, path):
return os.path.relpath(path, self.root_dir)
def add(self, fname):
out = os.path.basename(fname)
stage_fname = out + Stage.STAGE_FILE_SUFFIX
cwd = os.path.dirname(os.path.abspath(fname))
stage = Stage.loads(project=self,
cmd=None,
deps=[],
outs=[out],
fname=stage_fname,
cwd=cwd)
stage.save()
stage.dump()
return stage
def remove(self, target):
if not Stage.is_stage_file(target):
raise StageNotFoundError(target)
stage = Stage.load(self, target)
for out in stage.outs:
out.remove()
return stage
def run(self,
cmd=None,
deps=[],
outs=[],
outs_no_cache=[],
fname=Stage.STAGE_FILE,
cwd=os.curdir,
no_exec=False):
stage = Stage.loads(project=self,
fname=fname,
cmd=cmd,
cwd=cwd,
outs=outs,
outs_no_cache=outs_no_cache,
deps=deps)
if not no_exec:
stage.run()
stage.dump()
return stage
def _reproduce_stage(self, stages, node, force):
stage = stages[node].reproduce(force=force)
if not stage:
return []
stage.dump()
return [stage]
def reproduce(self, target, recursive=True, force=False):
stages = nx.get_node_attributes(self.graph(), 'stage')
node = os.path.relpath(os.path.abspath(target), self.root_dir)
if node not in stages:
raise StageNotFoundError(target)
if recursive:
return self._reproduce_stages(stages, node, force)
return self._reproduce_stage(stages, node, force)
def _reproduce_stages(self, stages, node, force):
result = []
for n in nx.dfs_postorder_nodes(self.graph(), node):
try:
result += self._reproduce_stage(stages, n, force)
except Exception as ex:
raise ReproductionError(stages[n].relpath, ex)
return result
def _remove_untracked_hardlinks(self):
untracked = self.scm.untracked_files()
cache = dict((System.inode(c), c) for c in self.cache.all())
for file in untracked:
inode = System.inode(file)
if inode not in cache.keys():
continue
Logger.info(u'Remove \'{}\''.format(file))
os.remove(file)
dir = os.path.dirname(file)
if len(dir) != 0 and not os.listdir(dir):
Logger.info(u'Remove empty directory \'{}\''.format(dir))
os.removedirs(dir)
def checkout(self):
self._remove_untracked_hardlinks()
for stage in self.stages():
stage.checkout()
def _used_cache(self, target=None):
cache_set = set()
if target:
stages = [Stage.load(self, target)]
else:
stages = self.stages()
for stage in stages:
for out in stage.outs:
if not out.use_cache or not out.cache:
continue
cache_set |= set([out.cache])
if out.is_dir_cache(out.cache) and os.path.isfile(out.cache):
dir_cache = out.dir_cache()
cache_set |= set(dir_cache.values())
return list(cache_set)
def gc(self):
clist = self._used_cache()
for cache in self.cache.all():
if cache in clist:
continue
os.unlink(cache)
self.logger.info(u'\'{}\' was removed'.format(
self.to_dvc_path(cache)))
def push(self, target=None, jobs=1, remote=None):
return self.cloud.push(self._used_cache(target), jobs, remote=remote)
def fetch(self, target=None, jobs=1, remote=None):
return self.cloud.pull(self._used_cache(target), jobs, remote=remote)
def pull(self, target=None, jobs=1, remote=None):
ret = self.fetch(target, jobs, remote=remote)
self.checkout()
return ret
def _local_status(self, target=None):
status = {}
if target:
stages = [Stage.load(self, target)]
else:
stages = self.stages()
for stage in self.stages():
status.update(stage.status())
return status
def _cloud_status(self, target=None, jobs=1, remote=None):
status = {}
for target, ret in self.cloud.status(self._used_cache(target),
jobs,
remote=remote):
if ret == cloud.STATUS_UNKNOWN or ret == cloud.STATUS_OK:
continue
prefix_map = {
cloud.STATUS_DELETED: 'deleted',
cloud.STATUS_MODIFIED: 'modified',
cloud.STATUS_NEW: 'new',
}
path = os.path.relpath(target, self.cache.cache_dir)
status[path] = prefix_map[ret]
return status
def status(self, target=None, jobs=1, cloud=False, remote=None):
if cloud:
return self._cloud_status(target, jobs, remote=remote)
return self._local_status(target)
def graph(self):
G = nx.DiGraph()
for stage in self.stages():
node = os.path.relpath(stage.path, self.root_dir)
G.add_node(node, stage=stage)
for dep in stage.deps:
dep_stage = dep.stage()
if not dep_stage:
continue
dep_node = os.path.relpath(dep_stage.path, self.root_dir)
G.add_node(dep_node, stage=dep_stage)
G.add_edge(node, dep_node)
return G
def stages(self):
stages = []
for root, dirs, files in os.walk(self.root_dir):
for fname in files:
path = os.path.join(root, fname)
if not Stage.is_stage_file(path):
continue
stages.append(Stage.load(self, path))
return stages
def outs(self):
outs = []
for stage in self.stages():
outs += stage.outs
return outs
def supported_url(url):
from dvc.cloud.data_cloud import DataCloud
return DataCloud.supported(url) != None
def _test_cloud(self, config, cl):
cloud = DataCloud(cache=self.dvc.cache,
state=self.dvc.state,
config=config)
self.assertIsInstance(cloud._cloud, cl)
class Project(object):
DVC_DIR = '.dvc'
def __init__(self, root_dir):
self.root_dir = os.path.abspath(os.path.realpath(root_dir))
self.dvc_dir = os.path.join(self.root_dir, self.DVC_DIR)
self.config = Config(self.dvc_dir)
self.scm = SCM(self.root_dir)
self.lock = Lock(self.dvc_dir)
self.link_state = LinkState(self.root_dir, self.dvc_dir)
self.logger = Logger(self.config._config[Config.SECTION_CORE].get(
Config.SECTION_CORE_LOGLEVEL, None))
self.cache = Cache(self)
self.cloud = DataCloud(cache=self.cache, config=self.config._config)
self.updater = Updater(self.dvc_dir)
self._ignore()
self.updater.check()
@staticmethod
def init(root_dir=os.curdir, no_scm=False):
"""
Initiate dvc project in directory.
Args:
root_dir: Path to project's root directory.
Returns:
Project instance.
Raises:
KeyError: Raises an exception.
"""
root_dir = os.path.abspath(root_dir)
dvc_dir = os.path.join(root_dir, Project.DVC_DIR)
scm = SCM(root_dir)
if type(scm) == Base and not no_scm:
msg = "{} is not tracked by any supported scm tool(e.g. git).".format(
root_dir)
raise InitError(msg)
os.mkdir(dvc_dir)
config = Config.init(dvc_dir)
proj = Project(root_dir)
scm.add([config.config_file])
if scm.ignore_file():
scm.add([os.path.join(dvc_dir, scm.ignore_file())])
return proj
def _ignore(self):
l = [
self.link_state.state_file, self.link_state._lock_file.lock_file,
self.lock.lock_file, self.config.config_local_file,
self.updater.updater_file
]
if self.cache.local.cache_dir.startswith(self.root_dir):
l += [self.cache.local.cache_dir]
self.scm.ignore_list(l)
def install(self):
self.scm.install()
def to_dvc_path(self, path):
return os.path.relpath(path, self.root_dir)
def add(self, fname):
out = os.path.basename(os.path.normpath(fname))
stage_fname = out + Stage.STAGE_FILE_SUFFIX
cwd = os.path.dirname(os.path.abspath(fname))
stage = Stage.loads(project=self,
cmd=None,
deps=[],
outs=[out],
fname=stage_fname,
cwd=cwd)
stage.save()
stage.dump()
return stage
def remove(self, target):
if not Stage.is_stage_file(target):
raise StageNotFoundError(target)
stage = Stage.load(self, target)
for out in stage.outs:
out.remove()
return stage
def run(self,
cmd=None,
deps=[],
outs=[],
outs_no_cache=[],
fname=Stage.STAGE_FILE,
cwd=os.curdir,
no_exec=False):
stage = Stage.loads(project=self,
fname=fname,
cmd=cmd,
cwd=cwd,
outs=outs,
outs_no_cache=outs_no_cache,
deps=deps)
if not no_exec:
stage.run()
stage.dump()
return stage
def _reproduce_stage(self, stages, node, force):
stage = stages[node].reproduce(force=force)
if not stage:
return []
stage.dump()
return [stage]
def reproduce(self, target, recursive=True, force=False):
stages = nx.get_node_attributes(self.graph(), 'stage')
node = os.path.relpath(os.path.abspath(target), self.root_dir)
if node not in stages:
raise StageNotFoundError(target)
if recursive:
return self._reproduce_stages(stages, node, force)
return self._reproduce_stage(stages, node, force)
def _reproduce_stages(self, stages, node, force):
result = []
for n in nx.dfs_postorder_nodes(self.graph(), node):
try:
result += self._reproduce_stage(stages, n, force)
except Exception as ex:
raise ReproductionError(stages[n].relpath, ex)
return result
def checkout(self, target=None):
if target:
if not Stage.is_stage_file(target):
raise StageNotFoundError(target)
stages = [Stage.load(self, target)]
else:
self.link_state.remove_all()
stages = self.stages()
for stage in stages:
stage.checkout()
def _used_cache(self, target=None):
cache_set = set()
if target:
stages = [Stage.load(self, target)]
else:
stages = self.stages()
for stage in stages:
for out in stage.outs:
if out.path_info['scheme'] != 'local':
continue
if not out.use_cache or not out.cache:
continue
cache_set |= set([out.cache])
if self.cache.local.is_dir_cache(out.cache) and os.path.isfile(
out.cache):
dir_cache = self.cache.local.dir_cache(out.cache)
cache_set |= set(dir_cache.values())
return list(cache_set)
def gc(self):
clist = self._used_cache()
for cache in self.cache.local.all():
if cache in clist:
continue
os.unlink(cache)
self.logger.info(u'\'{}\' was removed'.format(
self.to_dvc_path(cache)))
def push(self, target=None, jobs=1, remote=None):
return self.cloud.push(self._used_cache(target), jobs, remote=remote)
def fetch(self, target=None, jobs=1, remote=None):
return self.cloud.pull(self._used_cache(target), jobs, remote=remote)
def pull(self, target=None, jobs=1, remote=None):
ret = self.fetch(target, jobs, remote=remote)
self.checkout()
return ret
def _local_status(self, target=None):
status = {}
if target:
stages = [Stage.load(self, target)]
else:
stages = self.stages()
for stage in stages:
status.update(stage.status())
return status
def _cloud_status(self, target=None, jobs=1, remote=None):
status = {}
for target, ret in self.cloud.status(self._used_cache(target),
jobs,
remote=remote):
if ret == cloud.STATUS_UNKNOWN or ret == cloud.STATUS_OK:
continue
prefix_map = {
cloud.STATUS_DELETED: 'deleted',
cloud.STATUS_MODIFIED: 'modified',
cloud.STATUS_NEW: 'new',
}
path = os.path.relpath(target, self.cache.local.cache_dir)
status[path] = prefix_map[ret]
return status
def status(self, target=None, jobs=1, cloud=False, remote=None):
if cloud:
return self._cloud_status(target, jobs, remote=remote)
return self._local_status(target)
def _read_metric_json(self, fd, json_path):
parser = parse(json_path)
return [x.value for x in parser.find(json.load(fd))]
def _do_read_metric_tsv(self, reader, row, col):
if col != None and row != None:
return [reader[row][col]]
elif col != None:
return [r[col] for r in reader]
elif row != None:
return reader[row]
return None
def _read_metric_htsv(self, fd, htsv_path):
col, row = htsv_path.split(',')
row = int(row)
reader = list(csv.DictReader(fd, delimiter='\t'))
return self._do_read_metric_tsv(reader, row, col)
def _read_metric_tsv(self, fd, tsv_path):
col, row = tsv_path.split(',')
row = int(row)
col = int(col)
reader = list(csv.reader(fd, delimiter='\t'))
return self._do_read_metric_tsv(reader, row, col)
def _read_metric(self,
path,
json_path=None,
tsv_path=None,
htsv_path=None):
ret = None
try:
with open(path, 'r') as fd:
if json_path:
ret = self._read_metric_json(fd, json_path)
elif tsv_path:
ret = self._read_metric_tsv(fd, tsv_path)
elif htsv_path:
ret = self._read_metric_htsv(fd, htsv_path)
else:
ret = fd.read()
except Exception as exc:
self.logger.error('Unable to read metric in \'{}\''.format(path),
exc)
return ret
def metrics(self, path, json_path=None, tsv_path=None, htsv_path=None):
res = {}
saved = self.scm.active_branch()
for branch in self.scm.list_branches():
self.scm.checkout(branch)
self.checkout()
res[branch] = self._read_metric(path,
json_path=json_path,
tsv_path=tsv_path,
htsv_path=htsv_path)
self.scm.checkout(saved)
self.checkout()
return res
def graph(self):
G = nx.DiGraph()
stages = self.stages()
outs_map = {}
for stage in stages:
for o in stage.outs:
outs_map[o.path] = stage
for stage in stages:
node = os.path.relpath(stage.path, self.root_dir)
G.add_node(node, stage=stage)
for dep in stage.deps:
dep_stage = outs_map.get(dep.path, None)
if not dep_stage:
continue
dep_node = os.path.relpath(dep_stage.path, self.root_dir)
G.add_node(dep_node, stage=dep_stage)
G.add_edge(node, dep_node)
return G
def stages(self):
stages = []
for root, dirs, files in os.walk(self.root_dir):
for fname in files:
path = os.path.join(root, fname)
if not Stage.is_stage_file(path):
continue
stages.append(Stage.load(self, path))
return stages
def outs(self):
outs = []
for stage in self.stages():
outs += stage.outs
return outs
class Project(object):
DVC_DIR = '.dvc'
def __init__(self, root_dir):
self.root_dir = os.path.abspath(os.path.realpath(root_dir))
self.dvc_dir = os.path.join(self.root_dir, self.DVC_DIR)
self.scm = SCM(self.root_dir)
self.lock = Lock(self.dvc_dir)
self.cache = Cache(self.dvc_dir)
self.state = State(self.root_dir, self.dvc_dir)
self.config = Config(self.dvc_dir)
self.logger = Logger(self.config._config)
self.cloud = DataCloud(self.cache.cache_dir, self.config._config)
@staticmethod
def init(root_dir=os.curdir):
"""
Initiate dvc project in directory.
Args:
root_dir: Path to project's root directory.
Returns:
Project instance.
Raises:
KeyError: Raises an exception.
"""
root_dir = os.path.abspath(root_dir)
dvc_dir = os.path.join(root_dir, Project.DVC_DIR)
os.mkdir(dvc_dir)
config = Config.init(dvc_dir)
cache = Cache.init(dvc_dir)
state = State.init(root_dir, dvc_dir)
lock = Lock(dvc_dir)
scm = SCM(root_dir)
scm.ignore_list([cache.cache_dir, state.state_file, lock.lock_file])
ignore_file = os.path.join(dvc_dir, scm.ignore_file())
scm.add([config.config_file, ignore_file])
return Project(root_dir)
def to_dvc_path(self, path):
return os.path.relpath(path, self.root_dir)
def add(self, fname):
out = os.path.basename(fname)
stage_fname = out + Stage.STAGE_FILE_SUFFIX
cwd = os.path.dirname(os.path.abspath(fname))
stage = Stage.loads(project=self,
cmd=None,
deps=[],
outs=[out],
fname=stage_fname,
cwd=cwd)
stage.save()
stage.dump()
return stage
def remove(self, fname):
stages = []
output = Output.loads(self, fname)
for out in self.outs():
if out.path == output.path:
stage = out.stage()
stages.append(stage)
if len(stages) == 0:
raise StageNotFoundError(fname)
for stage in stages:
stage.remove()
return stages
def run(self,
cmd=None,
deps=[],
outs=[],
outs_no_cache=[],
fname=Stage.STAGE_FILE,
cwd=os.curdir,
no_exec=False):
stage = Stage.loads(project=self,
fname=fname,
cmd=cmd,
cwd=cwd,
outs=outs,
outs_no_cache=outs_no_cache,
deps=deps)
if not no_exec:
stage.run()
stage.dump()
return stage
def _reproduce_stage(self, stages, node, force):
if not stages[node].changed():
return []
stages[node].reproduce(force=force)
stages[node].dump()
return [stages[node]]
def reproduce(self, target, recursive=True, force=False):
stages = nx.get_node_attributes(self.graph(), 'stage')
node = os.path.relpath(os.path.abspath(target), self.root_dir)
if node not in stages:
raise StageNotFoundError(target)
if recursive:
return self._reproduce_stages(stages, node, force)
return self._reproduce_stage(stages, node, force)
def _reproduce_stages(self, stages, node, force):
result = []
for n in nx.dfs_postorder_nodes(self.graph(), node):
try:
result += self._reproduce_stage(stages, n, force)
except Exception as ex:
raise ReproductionError(stages[n].relpath, ex)
return result
def checkout(self):
for stage in self.stages():
stage.checkout()
def _used_cache(self):
clist = []
for stage in self.stages():
for out in stage.outs:
if not out.use_cache:
continue
if out.cache not in clist:
clist.append(out.cache)
return clist
def _remove_cache_file(self, cache):
os.chmod(cache, stat.S_IWRITE)
os.unlink(cache)
def _remove_cache(self, cache):
if os.path.isfile(cache):
self._remove_cache_file(cache)
return
for root, dirs, files in os.walk(cache, topdown=False):
for dname in dirs:
path = os.path.join(root, dname)
os.rmdir(path)
for fname in files:
path = os.path.join(root, fname)
self._remove_cache_file(path)
os.rmdir(cache)
def gc(self):
clist = self._used_cache()
for cache in self.cache.all():
if cache in clist:
continue
self._remove_cache(cache)
self.logger.info(u'\'{}\' was removed'.format(
self.to_dvc_path(cache)))
def push(self, jobs=1):
self.cloud.push(self._used_cache(), jobs)
def pull(self, jobs=1):
self.cloud.pull(self._used_cache(), jobs)
self.checkout()
def status(self, jobs=1):
return self.cloud.status(self._used_cache(), jobs)
def graph(self):
G = nx.DiGraph()
for stage in self.stages():
node = os.path.relpath(stage.path, self.root_dir)
G.add_node(node, stage=stage)
for dep in stage.deps:
dep_stage = dep.stage()
if not dep_stage:
continue
dep_node = os.path.relpath(dep_stage.path, self.root_dir)
G.add_node(dep_node, stage=dep_stage)
G.add_edge(node, dep_node)
return G
def stages(self):
stages = []
for root, dirs, files in os.walk(self.root_dir):
for fname in files:
path = os.path.join(root, fname)
if not Stage.is_stage_file(path):
continue
stages.append(Stage.load(self, path))
return stages
def outs(self):
outs = []
for stage in self.stages():
outs += stage.outs
return outs