# TODO:
# --md5sum - display md5 of each file as read from disk
+import apiclient.errors
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
CAUGHT_SIGNALS = [signal.SIGINT, signal.SIGQUIT, signal.SIGTERM]
-def parse_arguments(arguments):
- parser = argparse.ArgumentParser(
- description='Copy data from the local filesystem to Keep.')
-
- parser.add_argument('paths', metavar='path', type=str, nargs='*',
- help="""
- Local file or directory. Default: read from standard input.
- """)
-
- parser.add_argument('--max-manifest-depth', type=int, metavar='N',
- default=-1, help="""
- Maximum depth of directory tree to represent in the manifest
- structure. A directory structure deeper than this will be represented
- as a single stream in the manifest. If N=0, the manifest will contain
- a single stream. Default: -1 (unlimited), i.e., exactly one manifest
- stream per filesystem directory that contains files.
- """)
-
- group = parser.add_mutually_exclusive_group()
-
- group.add_argument('--as-stream', action='store_true', dest='stream',
- help="""
- Synonym for --stream.
- """)
-
- group.add_argument('--stream', action='store_true',
- help="""
- Store the file content and display the resulting manifest on
- stdout. Do not write the manifest to Keep or save a Collection object
- in Arvados.
- """)
-
- group.add_argument('--as-manifest', action='store_true', dest='manifest',
- help="""
- Synonym for --manifest.
- """)
-
- group.add_argument('--in-manifest', action='store_true', dest='manifest',
- help="""
- Synonym for --manifest.
- """)
-
- group.add_argument('--manifest', action='store_true',
- help="""
- Store the file data and resulting manifest in Keep, save a Collection
- object in Arvados, and display the manifest locator (Collection uuid)
- on stdout. This is the default behavior.
- """)
-
- group.add_argument('--as-raw', action='store_true', dest='raw',
- help="""
- Synonym for --raw.
- """)
-
- group.add_argument('--raw', action='store_true',
- help="""
- Store the file content and display the data block locators on stdout,
- separated by commas, with a trailing newline. Do not store a
- manifest.
- """)
-
- parser.add_argument('--use-filename', type=str, default=None,
- dest='filename', help="""
- Synonym for --filename.
- """)
-
- parser.add_argument('--filename', type=str, default=None,
- help="""
- Use the given filename in the manifest, instead of the name of the
- local file. This is useful when "-" or "/dev/stdin" is given as an
- input file. It can be used only if there is exactly one path given and
- it is not a directory. Implies --manifest.
- """)
+upload_opts = argparse.ArgumentParser(add_help=False)
+
+upload_opts.add_argument('paths', metavar='path', type=str, nargs='*',
+ help="""
+Local file or directory. Default: read from standard input.
+""")
+
+upload_opts.add_argument('--max-manifest-depth', type=int, metavar='N',
+ default=-1, help="""
+Maximum depth of directory tree to represent in the manifest
+structure. A directory structure deeper than this will be represented
+as a single stream in the manifest. If N=0, the manifest will contain
+a single stream. Default: -1 (unlimited), i.e., exactly one manifest
+stream per filesystem directory that contains files.
+""")
+
+upload_opts.add_argument('--project-uuid', metavar='UUID', help="""
+Store the collection in the specified project, instead of your Home
+project.
+""")
+
+upload_opts.add_argument('--name', help="""
+Save the collection with the specified name, rather than the default
+generic name "Saved at {time} by {username}@{host}".
+""")
+
+_group = upload_opts.add_mutually_exclusive_group()
+
+_group.add_argument('--as-stream', action='store_true', dest='stream',
+ help="""
+Synonym for --stream.
+""")
+
+_group.add_argument('--stream', action='store_true',
+ help="""
+Store the file content and display the resulting manifest on
+stdout. Do not write the manifest to Keep or save a Collection object
+in Arvados.
+""")
+
+_group.add_argument('--as-manifest', action='store_true', dest='manifest',
+ help="""
+Synonym for --manifest.
+""")
+
+_group.add_argument('--in-manifest', action='store_true', dest='manifest',
+ help="""
+Synonym for --manifest.
+""")
+
+_group.add_argument('--manifest', action='store_true',
+ help="""
+Store the file data and resulting manifest in Keep, save a Collection
+object in Arvados, and display the manifest locator (Collection uuid)
+on stdout. This is the default behavior.
+""")
+
+_group.add_argument('--as-raw', action='store_true', dest='raw',
+ help="""
+Synonym for --raw.
+""")
+
+_group.add_argument('--raw', action='store_true',
+ help="""
+Store the file content and display the data block locators on stdout,
+separated by commas, with a trailing newline. Do not store a
+manifest.
+""")
+
+upload_opts.add_argument('--use-filename', type=str, default=None,
+ dest='filename', help="""
+Synonym for --filename.
+""")
+
+upload_opts.add_argument('--filename', type=str, default=None,
+ help="""
+Use the given filename in the manifest, instead of the name of the
+local file. This is useful when "-" or "/dev/stdin" is given as an
+input file. It can be used only if there is exactly one path given and
+it is not a directory. Implies --manifest.
+""")
+
+run_opts = argparse.ArgumentParser(add_help=False)
+_group = run_opts.add_mutually_exclusive_group()
+_group.add_argument('--progress', action='store_true',
+ help="""
+Display human-readable progress on stderr (bytes and, if possible,
+percentage of total data size). This is the default behavior when
+stderr is a tty.
+""")
+
+_group.add_argument('--no-progress', action='store_true',
+ help="""
+Do not display human-readable progress on stderr, even if stderr is a
+tty.
+""")
+
+_group.add_argument('--batch-progress', action='store_true',
+ help="""
+Display machine-readable progress on stderr (bytes and, if known,
+total data size).
+""")
+
+_group = run_opts.add_mutually_exclusive_group()
+_group.add_argument('--resume', action='store_true', default=True,
+ help="""
+Continue interrupted uploads from cached state (default).
+""")
+_group.add_argument('--no-resume', action='store_false', dest='resume',
+ help="""
+Do not continue interrupted uploads from cached state.
+""")
+
+arg_parser = argparse.ArgumentParser(
+ description='Copy data from the local filesystem to Keep.',
+ parents=[upload_opts, run_opts])
- group = parser.add_mutually_exclusive_group()
- group.add_argument('--progress', action='store_true',
- help="""
- Display human-readable progress on stderr (bytes and, if possible,
- percentage of total data size). This is the default behavior when
- stderr is a tty.
- """)
-
- group.add_argument('--no-progress', action='store_true',
- help="""
- Do not display human-readable progress on stderr, even if stderr is a
- tty.
- """)
-
- group.add_argument('--batch-progress', action='store_true',
- help="""
- Display machine-readable progress on stderr (bytes and, if known,
- total data size).
- """)
-
- group = parser.add_mutually_exclusive_group()
- group.add_argument('--resume', action='store_true', default=True,
- help="""
- Continue interrupted uploads from cached state (default).
- """)
- group.add_argument('--no-resume', action='store_false', dest='resume',
- help="""
- Do not continue interrupted uploads from cached state.
- """)
-
- args = parser.parse_args(arguments)
+def parse_arguments(arguments):
+ args = arg_parser.parse_args(arguments)
if len(args.paths) == 0:
args.paths += ['/dev/stdin']
if len(args.paths) != 1 or os.path.isdir(args.paths[0]):
if args.filename:
- parser.error("""
+ arg_parser.error("""
--filename argument cannot be used when storing a directory or
multiple files.
""")
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:
def exit_signal_handler(sigcode, frame):
sys.exit(-sigcode)
-def main(arguments=None):
+def check_project_exists(project_uuid):
+ try:
+ arvados.api('v1').groups().get(uuid=project_uuid).execute()
+ except (apiclient.errors.Error, arvados.errors.NotFoundError) as error:
+ raise ValueError("Project {} not found ({})".format(project_uuid,
+ error))
+ else:
+ return True
+
+def prep_project_link(args, stderr, project_exists=check_project_exists):
+ # Given the user's command line arguments, return a dictionary with data
+ # to create the desired project link for this Collection, or None.
+ # Raises ValueError if the arguments request something impossible.
+ making_collection = not (args.raw or args.stream)
+ if not making_collection:
+ if args.name or args.project_uuid:
+ raise ValueError("Requested a Link without creating a Collection")
+ return None
+ link = {'tail_uuid': args.project_uuid,
+ 'link_class': 'name',
+ 'name': args.name}
+ if not link['tail_uuid']:
+ link['tail_uuid'] = arvados.api('v1').users().current().execute()['uuid']
+ elif not project_exists(link['tail_uuid']):
+ raise ValueError("Project {} not found".format(args.project_uuid))
+ if not link['name']:
+ link['name'] = "Saved at {} by {}@{}".format(
+ datetime.datetime.utcnow().strftime("%Y-%m-%d %H:%M:%S UTC"),
+ pwd.getpwuid(os.getuid()).pw_name,
+ socket.gethostname())
+ stderr.write(
+ "arv-put: No --name specified. Saving as \"%s\"\n" % link['name'])
+ link['owner_uuid'] = link['tail_uuid']
+ return link
+
+def create_project_link(locator, link):
+ link['head_uuid'] = locator
+ return arvados.api('v1').links().create(body=link).execute()
+
+def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr):
+ status = 0
+
args = parse_arguments(arguments)
+ try:
+ project_link = prep_project_link(args, stderr)
+ except ValueError as error:
+ print >>stderr, "arv-put: {}.".format(error)
+ sys.exit(2)
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)
else:
- if not args.resume:
- resume_cache.restart()
writer = ArvPutCollectionWriter.from_cache(
resume_cache, reporter, bytes_expected)
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']
+ output = collection['uuid']
+ if project_link is not None:
+ try:
+ if 'name' in collection:
+ arvados.api().collections().update(uuid=output,
+ body={"owner_uuid": project_link["tail_uuid"],
+ "name": project_link["name"]}).execute()
+ else:
+ create_project_link(output, project_link)
+ except apiclient.errors.Error as error:
+ print >>stderr, (
+ "arv-put: Error adding Collection to project: {}.".format(
+ error))
+ status = 1
+
+ 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()
projects / arvados.git / blobdiff