12 import arvados.commands._util as arv_cmd
13 import arvados.util as util
15 from arvados._version import __version__
18 logger = logging.getLogger('arvados.arv-get')
20 parser = argparse.ArgumentParser(
21 description='Copy data from Keep to a local file or pipe.',
22 parents=[arv_cmd.retry_opt])
23 parser.add_argument('--version', action='version',
24 version="%s %s" % (sys.argv[0], __version__),
25 help='Print version and exit.')
26 parser.add_argument('locator', type=str,
28 Collection locator, optionally with a file path or prefix.
30 parser.add_argument('destination', type=str, nargs='?', default='-',
32 Local file or directory where the data is to be written. Default: stdout.
34 group = parser.add_mutually_exclusive_group()
35 group.add_argument('--progress', action='store_true',
37 Display human-readable progress on stderr (bytes and, if possible,
38 percentage of total data size). This is the default behavior when it
39 is not expected to interfere with the output: specifically, stderr is
40 a tty _and_ either stdout is not a tty, or output is being written to
41 named files rather than stdout.
43 group.add_argument('--no-progress', action='store_true',
45 Do not display human-readable progress on stderr.
47 group.add_argument('--batch-progress', action='store_true',
49 Display machine-readable progress on stderr (bytes and, if known,
52 group = parser.add_mutually_exclusive_group()
53 group.add_argument('--hash',
55 Display the hash of each file as it is read from Keep, using the given
56 hash algorithm. Supported algorithms include md5, sha1, sha224,
57 sha256, sha384, and sha512.
59 group.add_argument('--md5sum', action='store_const',
60 dest='hash', const='md5',
62 Display the MD5 hash of each file as it is read from Keep.
64 parser.add_argument('-n', action='store_true',
66 Do not write any data -- just read from Keep, and report md5sums if
69 parser.add_argument('-r', action='store_true',
71 Retrieve all files in the specified collection/prefix. This is the
72 default behavior if the "locator" argument ends with a forward slash.
74 group = parser.add_mutually_exclusive_group()
75 group.add_argument('-f', action='store_true',
77 Overwrite existing files while writing. The default behavior is to
78 refuse to write *anything* if any of the output files already
79 exist. As a special case, -f is not needed to write to stdout.
81 group.add_argument('--skip-existing', action='store_true',
83 Skip files that already exist. The default behavior is to refuse to
84 write *anything* if any files exist that would have to be
85 overwritten. This option causes even devices, sockets, and fifos to be
88 group.add_argument('--strip-manifest', action='store_true', default=False,
90 When getting a collection manifest, strip its access tokens before writing
94 def parse_arguments(arguments, stdout, stderr):
95 args = parser.parse_args(arguments)
97 if args.locator[-1] == os.sep:
101 not (args.destination and
102 os.path.isdir(args.destination))):
103 parser.error('Destination is not a directory.')
104 if not args.r and (os.path.isdir(args.destination) or
105 args.destination[-1] == os.path.sep):
106 args.destination = os.path.join(args.destination,
107 os.path.basename(args.locator))
108 logger.debug("Appended source file name to destination directory: %s",
111 if args.destination == '/dev/stdout':
112 args.destination = "-"
114 if args.destination == '-':
115 # Normally you have to use -f to write to a file (or device) that
116 # already exists, but "-" and "/dev/stdout" are common enough to
117 # merit a special exception.
120 args.destination = args.destination.rstrip(os.sep)
122 # Turn on --progress by default if stderr is a tty and output is
123 # either going to a named file, or going (via stdout) to something
125 if (not (args.batch_progress or args.no_progress)
127 and (args.destination != '-'
128 or not stdout.isatty())):
132 def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr):
135 args = parse_arguments(arguments, stdout, stderr)
136 if api_client is None:
137 api_client = arvados.api('v1')
139 r = re.search(r'^(.*?)(/.*)?$', args.locator)
141 get_prefix = r.group(2)
142 if args.r and not get_prefix:
145 reader = arvados.CollectionReader(col_loc, num_retries=args.retries)
146 except Exception as error:
147 logger.error("failed to read collection: {}".format(error))
150 # User asked to download the collection's manifest
153 open_flags = os.O_CREAT | os.O_WRONLY
155 open_flags |= os.O_EXCL
157 if args.destination == "-":
158 stdout.write(reader.manifest_text(strip=args.strip_manifest))
160 out_fd = os.open(args.destination, open_flags)
161 with os.fdopen(out_fd, 'wb') as out_file:
162 out_file.write(reader.manifest_text(strip=args.strip_manifest))
163 except (IOError, OSError) as error:
164 logger.error("can't write to '{}': {}".format(args.destination, error))
166 except (arvados.errors.ApiError, arvados.errors.KeepReadError) as error:
167 logger.error("failed to download '{}': {}".format(col_loc, error))
171 # Scan the collection. Make an array of (stream, file, local
172 # destination filename) tuples, and add up total size to extract.
176 if get_prefix == os.sep:
179 item = reader.find('.' + get_prefix)
181 if isinstance(item, arvados.collection.Subcollection) or isinstance(item, arvados.collection.CollectionReader):
182 # If the user asked for a file and we got a subcollection, error out.
183 if get_prefix[-1] != os.sep:
184 logger.error("requested file '{}' is in fact a subcollection. Append a trailing '/' to download it.".format('.' + get_prefix))
186 # If the user asked stdout as a destination, error out.
187 elif args.destination == '-':
188 logger.error("cannot use 'stdout' as destination when downloading multiple files.")
190 # User asked for a subcollection, and that's what was found. Add up total size
192 for s, f in files_in_collection(item):
193 dest_path = os.path.join(
195 os.path.join(s.stream_name(), f.name)[len(get_prefix)+1:])
196 if (not (args.n or args.f or args.skip_existing) and
197 os.path.exists(dest_path)):
198 logger.error('Local file %s already exists.' % (dest_path,))
200 todo += [(s, f, dest_path)]
201 todo_bytes += f.size()
202 elif isinstance(item, arvados.arvfile.ArvadosFile):
203 todo += [(item.parent, item, args.destination)]
204 todo_bytes += item.size()
206 logger.error("'{}' not found.".format('.' + get_prefix))
208 except (IOError, arvados.errors.NotFoundError) as e:
213 for s, f, outfilename in todo:
217 if outfilename == "-":
220 if args.skip_existing and os.path.exists(outfilename):
221 logger.debug('Local file %s exists. Skipping.', outfilename)
223 elif not args.f and (os.path.isfile(outfilename) or
224 os.path.isdir(outfilename)):
225 # Good thing we looked again: apparently this file wasn't
226 # here yet when we checked earlier.
227 logger.error('Local file %s already exists.' % (outfilename,))
230 arvados.util.mkdir_dash_p(os.path.dirname(outfilename))
232 outfile = open(outfilename, 'wb')
233 except Exception as error:
234 logger.error('Open(%s) failed: %s' % (outfilename, error))
237 digestor = hashlib.new(args.hash)
239 with s.open(f.name, 'r') as file_reader:
240 for data in file_reader.readall():
244 digestor.update(data)
245 out_bytes += len(data)
247 stderr.write('\r%d MiB / %d MiB %.1f%%' %
252 else 100.0*out_bytes/todo_bytes)))
253 elif args.batch_progress:
254 stderr.write('%s %d read %d total\n' %
255 (sys.argv[0], os.getpid(),
256 out_bytes, todo_bytes))
258 stderr.write("%s %s/%s\n"
259 % (digestor.hexdigest(), s.stream_name(), f.name))
260 except KeyboardInterrupt:
261 if outfile and (outfile.fileno() > 2) and not outfile.closed:
262 os.unlink(outfile.name)
265 if outfile != None and outfile != stdout:
272 def files_in_collection(c):
273 # Sort first by file type, then alphabetically by file path.
274 for i in sorted(c.keys(),
276 isinstance(c[k], arvados.collection.Subcollection),
278 if isinstance(c[i], arvados.arvfile.ArvadosFile):
280 elif isinstance(c[i], arvados.collection.Subcollection):
281 for s, f in files_in_collection(c[i]):