import argparse
import arvados
import base64
+import datetime
import errno
import fcntl
import hashlib
import json
import os
+import pwd
import signal
+import socket
import sys
import tempfile
+from apiclient import errors as apiclient_errors
import arvados.commands._util as arv_cmd
CAUGHT_SIGNALS = [signal.SIGINT, signal.SIGQUIT, signal.SIGTERM]
+api_client = None
upload_opts = argparse.ArgumentParser(add_help=False)
it is not a directory. Implies --manifest.
""")
+upload_opts.add_argument('--portable-data-hash', action='store_true',
+ help="""
+Print the portable data hash instead of the Arvados UUID for the collection
+created by the upload.
+""")
+
run_opts = argparse.ArgumentParser(add_help=False)
+
+run_opts.add_argument('--project-uuid', metavar='UUID', help="""
+Store the collection in the specified project, instead of your Home
+project.
+""")
+
+run_opts.add_argument('--name', help="""
+Save the collection with the specified name.
+""")
+
_group = run_opts.add_mutually_exclusive_group()
_group.add_argument('--progress', action='store_true',
help="""
arg_parser = argparse.ArgumentParser(
description='Copy data from the local filesystem to Keep.',
- parents=[upload_opts, run_opts])
+ parents=[upload_opts, run_opts, arv_cmd.retry_opt])
def parse_arguments(arguments):
args = arg_parser.parse_args(arguments)
class ResumeCache(object):
CACHE_DIR = '.cache/arvados/arv-put'
- @classmethod
- def setup_user_cache(cls):
- return arv_cmd.make_home_conf_dir(cls.CACHE_DIR, 0o700)
-
def __init__(self, file_spec):
self.cache_file = open(file_spec, 'a+')
self._lock_file(self.cache_file)
md5.update(str(max(args.max_manifest_depth, -1)))
elif args.filename:
md5.update(args.filename)
- return os.path.join(cls.CACHE_DIR, md5.hexdigest())
+ return os.path.join(
+ arv_cmd.make_home_conf_dir(cls.CACHE_DIR, 0o700, 'raise'),
+ md5.hexdigest())
def _lock_file(self, fileobj):
try:
STATE_PROPS = (arvados.ResumableCollectionWriter.STATE_PROPS +
['bytes_written', '_seen_inputs'])
- def __init__(self, cache=None, reporter=None, bytes_expected=None):
+ def __init__(self, cache=None, reporter=None, bytes_expected=None,
+ api_client=None, num_retries=0):
self.bytes_written = 0
self._seen_inputs = []
self.cache = cache
self.reporter = reporter
self.bytes_expected = bytes_expected
- super(ArvPutCollectionWriter, self).__init__()
+ super(ArvPutCollectionWriter, self).__init__(
+ api_client, num_retries=num_retries)
@classmethod
- def from_cache(cls, cache, reporter=None, bytes_expected=None):
+ def from_cache(cls, cache, reporter=None, bytes_expected=None,
+ num_retries=0):
try:
state = cache.load()
state['_data_buffer'] = [base64.decodestring(state['_data_buffer'])]
- writer = cls.from_state(state, cache, reporter, bytes_expected)
+ writer = cls.from_state(state, cache, reporter, bytes_expected,
+ num_retries=num_retries)
except (TypeError, ValueError,
arvados.errors.StaleWriterStateError) as error:
- return cls(cache, reporter, bytes_expected)
+ return cls(cache, reporter, bytes_expected, num_retries=num_retries)
else:
return writer
def exit_signal_handler(sigcode, frame):
sys.exit(-sigcode)
-def main(arguments=None):
+def desired_project_uuid(api_client, project_uuid, num_retries):
+ if not project_uuid:
+ query = api_client.users().current()
+ elif arvados.util.user_uuid_pattern.match(project_uuid):
+ query = api_client.users().get(uuid=project_uuid)
+ elif arvados.util.group_uuid_pattern.match(project_uuid):
+ query = api_client.groups().get(uuid=project_uuid)
+ else:
+ raise ValueError("Not a valid project UUID: {}".format(project_uuid))
+ return query.execute(num_retries=num_retries)['uuid']
+
+def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr):
+ global api_client
+
args = parse_arguments(arguments)
+ status = 0
+ if api_client is None:
+ api_client = arvados.api('v1')
+
+ # Determine the name to use
+ if args.name:
+ if args.stream or args.raw:
+ print >>stderr, "Cannot use --name with --stream or --raw"
+ sys.exit(1)
+ collection_name = args.name
+ else:
+ collection_name = "Saved at {} by {}@{}".format(
+ datetime.datetime.utcnow().strftime("%Y-%m-%d %H:%M:%S UTC"),
+ pwd.getpwuid(os.getuid()).pw_name,
+ socket.gethostname())
+
+ if args.project_uuid and (args.stream or args.raw):
+ print >>stderr, "Cannot use --project-uuid with --stream or --raw"
+ sys.exit(1)
+
+ # Determine the parent project
+ try:
+ project_uuid = desired_project_uuid(api_client, args.project_uuid,
+ args.retries)
+ except (apiclient_errors.Error, ValueError) as error:
+ print >>stderr, error
+ sys.exit(1)
if args.progress:
reporter = progress_writer(human_progress)
bytes_expected = expected_bytes_for(args.paths)
resume_cache = None
- try:
- if ResumeCache.setup_user_cache() is not None:
+ if args.resume:
+ try:
resume_cache = ResumeCache(ResumeCache.make_path(args))
- except (IOError, OSError):
- pass # Couldn't open cache directory/file. Continue without it.
- except ResumeCacheConflict:
- print "arv-put: Another process is already uploading this data."
- sys.exit(1)
+ except (IOError, OSError, ValueError):
+ pass # Couldn't open cache directory/file. Continue without it.
+ except ResumeCacheConflict:
+ print >>stderr, "\n".join([
+ "arv-put: Another process is already uploading this data.",
+ " Use --no-resume if this is really what you want."])
+ sys.exit(1)
if resume_cache is None:
- writer = ArvPutCollectionWriter(resume_cache, reporter, bytes_expected)
+ writer = ArvPutCollectionWriter(resume_cache, reporter, bytes_expected,
+ num_retries=args.retries)
else:
- if not args.resume:
- resume_cache.restart()
writer = ArvPutCollectionWriter.from_cache(
- resume_cache, reporter, bytes_expected)
+ resume_cache, reporter, bytes_expected, num_retries=args.retries)
# Install our signal handler for each code in CAUGHT_SIGNALS, and save
# the originals.
for sigcode in CAUGHT_SIGNALS}
if writer.bytes_written > 0: # We're resuming a previous upload.
- print >>sys.stderr, "\n".join([
+ print >>stderr, "\n".join([
"arv-put: Resuming previous upload from last checkpoint.",
" Use the --no-resume option to start over."])
- writer.report_progress()
+ writer.report_progress()
writer.do_queued_work() # Do work resumed from cache.
for path in args.paths: # Copy file data to Keep.
if os.path.isdir(path):
writer.finish_current_stream()
if args.progress: # Print newline to split stderr from stdout for humans.
- print >>sys.stderr
+ print >>stderr
if args.stream:
- print writer.manifest_text(),
+ output = writer.manifest_text()
elif args.raw:
- print ','.join(writer.data_locators())
+ output = ','.join(writer.data_locators())
else:
- # Register the resulting collection in Arvados.
- collection = arvados.api().collections().create(
- body={
- 'uuid': writer.finish(),
- 'manifest_text': writer.manifest_text(),
- },
- ).execute()
-
- # Print the locator (uuid) of the new collection.
- print collection['uuid']
+ try:
+ # Register the resulting collection in Arvados.
+ collection = api_client.collections().create(
+ body={
+ 'owner_uuid': project_uuid,
+ 'name': collection_name,
+ 'manifest_text': writer.manifest_text()
+ },
+ ensure_unique_name=True
+ ).execute(num_retries=args.retries)
+
+ print >>stderr, "Collection saved as '%s'" % collection['name']
+
+ if args.portable_data_hash and 'portable_data_hash' in collection and collection['portable_data_hash']:
+ output = collection['portable_data_hash']
+ else:
+ output = collection['uuid']
+
+ except apiclient_errors.Error as error:
+ print >>stderr, (
+ "arv-put: Error creating Collection on project: {}.".format(
+ error))
+ status = 1
+
+ # Print the locator (uuid) of the new collection.
+ stdout.write(output)
+ if not output.endswith('\n'):
+ stdout.write('\n')
for sigcode, orig_handler in orig_signal_handlers.items():
signal.signal(sigcode, orig_handler)
+ if status != 0:
+ sys.exit(status)
+
if resume_cache is not None:
resume_cache.destroy()
+ return output
+
if __name__ == '__main__':
main()