def abort_previous_upload(data_config):
if data_config.tarball_path and os.path.exists(data_config.tarball_path):
rmtree(os.path.dirname(data_config.tarball_path))
data_config.set_tarball_path("")
data_config.set_data_endpoint("")
DataConfigManager.set_config(data_config)
def upload():
"""
Upload data in the current dir to Floyd.
"""
data_config = DataConfigManager.get_config()
access_token = AuthConfigManager.get_access_token()
version = data_config.version
# Create data object
data_name = "{}/{}:{}".format(access_token.username, data_config.name,
version)
data = DataRequest(name=data_name, description=version, version=version)
data_id = DataClient().create(data)
floyd_logger.debug("Created data with id : {}".format(data_id))
floyd_logger.info("Upload finished")
# Update expt config including predecessor
data_config.increment_version()
data_config.set_data_predecessor(data_id)
DataConfigManager.set_config(data_config)
# Print output
table_output = [["DATA ID", "NAME", "VERSION"],
[data_id, data_name, version]]
floyd_logger.info(tabulate(table_output, headers="firstrow"))
def init(dataset_name):
"""
Initialize a new dataset at the current dir.
After init ensure that your data files are in this directory.
Then you can upload them to Floyd. Example:
floyd data upload
"""
dataset_obj = DatasetClient().get_by_name(dataset_name)
if not dataset_obj:
create_dataset_base_url = "{}/datasets/create".format(
floyd.floyd_web_host)
create_dataset_url = "{}?name={}".format(create_dataset_base_url,
dataset_name)
floyd_logger.error(
("Dataset name does not match your list of datasets. "
"Create your new dataset in the web dashboard:\n\t%s"),
create_dataset_base_url)
webbrowser.open(create_dataset_url)
return
data_config = DataConfig(name=dataset_name, family_id=dataset_obj.id)
DataConfigManager.set_config(data_config)
floyd_logger.info(
"Data source \"{}\" initialized in current directory".format(
dataset_name))
floyd_logger.info("""
You can now upload your data to Floyd by:
floyd data upload
""")
def init(dataset_name):
"""
Initialize a new dataset at the current dir.
Then run the upload command to copy all the files in this
directory to FloydHub.
floyd data upload
"""
dataset_obj = DatasetClient().get_by_name(dataset_name)
if not dataset_obj:
namespace, name = get_namespace_from_name(dataset_name)
create_dataset_base_url = "{}/datasets/create".format(
floyd.floyd_web_host)
create_dataset_url = "{}?name={}&namespace={}".format(
create_dataset_base_url, name, namespace)
floyd_logger.info(
("Dataset name does not match your list of datasets. "
"Create your new dataset in the web dashboard:\n\t%s"),
create_dataset_base_url)
webbrowser.open(create_dataset_url)
name = click.prompt(
'Press ENTER to use dataset name "%s" or enter a different name' %
dataset_name,
default=dataset_name,
show_default=False)
dataset_name = name.strip() or dataset_name
dataset_obj = DatasetClient().get_by_name(dataset_name)
if not dataset_obj:
raise FloydException(
'Dataset "%s" does not exist on floydhub.com. Ensure it exists before continuing.'
% dataset_name)
namespace, name = get_namespace_from_name(dataset_name)
data_config = DataConfig(name=name,
namespace=namespace,
family_id=dataset_obj.id)
DataConfigManager.set_config(data_config)
floyd_logger.info(
"Data source \"{}\" initialized in current directory".format(
dataset_name))
floyd_logger.info("""
You can now upload your data to Floyd by:
floyd data upload
""")
def init(name):
"""
Initialize a new data upload.
After init ensure that your data files are in this directory.
Then you can upload them to Floyd. Example:
floyd data upload
"""
data_config = DataConfig(name=name, family_id=generate_uuid())
DataConfigManager.set_config(data_config)
floyd_logger.info("Data source \"{}\" initialized in current directory".format(name))
floyd_logger.info("""
You can now upload your data to Floyd by:
floyd data upload
""")
def add_data(self, source):
data_config = DataConfigManager.get_config()
dataset_id = data_config.family_id
if not dataset_id:
sys.exit(
'Please initialize current directory with \'floyd data init DATASET_NAME\' first.'
)
re = self.request('POST',
'%s/%s' % (self.url, dataset_id),
json={'source': source})
return re.json()
def upload(resume, tar_file, message):
"""
Upload data in the current dir to Floyd.
"""
data_config = DataConfigManager.get_config()
if not upload_is_resumable(data_config) or not opt_to_resume(resume):
abort_previous_upload(data_config)
access_token = AuthConfigManager.get_access_token()
initialize_new_upload(data_config, access_token, tar_file, message)
complete_upload(data_config)
def initialize_new_upload(data_config, access_token, description=None, source_dir='.'):
# TODO: hit upload server to check for liveness before moving on
data_config.set_tarball_path(None)
data_config.set_data_endpoint(None)
data_config.set_resource_id(None)
namespace = data_config.namespace or access_token.username
data_name = "{}/{}".format(namespace, data_config.name)
# Create tarball of the data using the ID returned from the API
# TODO: allow to the users to change directory for the compression
temp_dir = tempfile.mkdtemp()
tarball_path = os.path.join(temp_dir, "floydhub_data.tar.gz")
floyd_logger.debug("Creating tarfile with contents of current directory: %s",
tarball_path)
floyd_logger.info("Compressing data...")
# TODO: purge tarball on Ctrl-C
create_tarfile(source_dir=source_dir, filename=tarball_path)
# If starting a new upload fails for some reason down the line, we don't
# want to re-tar, so save off the tarball path now
data_config.set_tarball_path(tarball_path)
DataConfigManager.set_config(data_config)
# Create data object using API
data = DataRequest(name=data_name,
description=description,
family_id=data_config.family_id,
data_type='gzip')
data_info = DataClient().create(data)
if not data_info:
rmtree(temp_dir)
sys.exit(1)
data_config.set_data_id(data_info['id'])
data_config.set_data_name(data_info['name'])
DataConfigManager.set_config(data_config)
# fetch auth token for upload server
creds = DataClient().new_tus_credentials(data_info['id'])
if not creds:
# TODO: delete module from server?
rmtree(temp_dir)
sys.exit(1)
data_resource_id = creds[0]
data_endpoint = TusDataClient().initialize_upload(
tarball_path,
metadata={"filename": data_resource_id},
auth=creds)
if not data_endpoint:
# TODO: delete module from server?
floyd_logger.error("Failed to get upload URL from Floydhub!")
rmtree(temp_dir)
sys.exit(1)
data_config.set_data_endpoint(data_endpoint)
DataConfigManager.set_config(data_config)
def get_all(self):
try:
data_config = DataConfigManager.get_config()
response = self.request("GET",
self.url,
params={
"module_type": "data",
"family_id": data_config.family_id
})
data_dict = response.json()
return [Data.from_dict(data) for data in data_dict]
except FloydException as e:
floyd_logger.error("Error while retrieving data: %s", e.message)
return []
def complete_upload(data_config):
data_endpoint = data_config.data_endpoint
data_id = data_config.data_id
tarball_path = data_config.tarball_path
if not data_id:
floyd_logger.error("Corrupted upload state, please start a new one.")
sys.exit(1)
# check for tarball upload, upload to server if not done
if not data_config.resource_id and (tarball_path and data_endpoint):
floyd_logger.debug("Getting fresh upload credentials")
creds = DataClient().new_tus_credentials(data_id)
if not creds:
sys.exit(1)
file_size = os.path.getsize(tarball_path)
# check for upload limit dimension
if file_size > MAX_UPLOAD_SIZE:
try:
floyd_logger.info("Removing compressed data...")
rmtree(os.path.dirname(tarball_path))
except (OSError, TypeError):
pass
sys.exit(("Data size too large to upload, please keep it under %s.\n") %
(sizeof_fmt(MAX_UPLOAD_SIZE)))
floyd_logger.info("Uploading compressed data. Total upload size: %s",
sizeof_fmt(file_size))
tus_client = TusDataClient()
if not tus_client.resume_upload(tarball_path, data_endpoint, auth=creds):
floyd_logger.error("Failed to finish upload!")
return
try:
floyd_logger.info("Removing compressed data...")
rmtree(os.path.dirname(tarball_path))
except (OSError, TypeError):
pass
floyd_logger.debug("Created data with id : %s", data_id)
floyd_logger.info("Upload finished.")
# Update data config
data_config.set_tarball_path(None)
data_config.set_data_endpoint(None)
data_source = DataClient().get(data_id)
data_config.set_resource_id(data_source.resource_id)
DataConfigManager.set_config(data_config)
# data tarball uploaded, check for server untar
if data_config.resource_id:
floyd_logger.info(
"Waiting for server to unpack data.\n"
"You can exit at any time and come back to check the status with:\n"
"\tfloyd data upload -r")
try:
for i in dots(ResourceWaitIter(data_config.resource_id),
label='Waiting for unpack...'):
pass
except WaitTimeoutException:
clint_STREAM.write('\n')
clint_STREAM.flush()
floyd_logger.info(
"Looks like it is going to take longer for Floydhub to unpack "
"your data. Please check back later.")
sys.exit(1)
else:
data_config.set_resource_id(None)
data_config.set_tarball_path(None)
data_config.set_data_endpoint(None)
data_config.set_resource_id(None)
data_config.set_data_id(None)
DataConfigManager.set_config(data_config)
# Print output
table_output = [["NAME"],
[normalize_data_name(data_config.data_name)]]
floyd_logger.info('')
floyd_logger.info(tabulate(table_output, headers="firstrow"))
def current_dataset_namespace():
return DataConfigManager.get_config().namespace or current_username()
def current_dataset_name():
return DataConfigManager.get_config().name
