12 import arvados.commands._util as arv_cmd
14 from arvados._version import __version__
17 logger = logging.getLogger('arvados.arv-get')
19 parser = argparse.ArgumentParser(
20 description='Copy data from Keep to a local file or pipe.',
21 parents=[arv_cmd.retry_opt])
22 parser.add_argument('--version', action='version',
23 version="%s %s" % (sys.argv[0], __version__),
24 help='Print version and exit.')
25 parser.add_argument('locator', type=str,
27 Collection locator, optionally with a file path or prefix.
29 parser.add_argument('destination', type=str, nargs='?', default='-',
31 Local file or directory where the data is to be written. Default: stdout.
33 group = parser.add_mutually_exclusive_group()
34 group.add_argument('--progress', action='store_true',
36 Display human-readable progress on stderr (bytes and, if possible,
37 percentage of total data size). This is the default behavior when it
38 is not expected to interfere with the output: specifically, stderr is
39 a tty _and_ either stdout is not a tty, or output is being written to
40 named files rather than stdout.
42 group.add_argument('--no-progress', action='store_true',
44 Do not display human-readable progress on stderr.
46 group.add_argument('--batch-progress', action='store_true',
48 Display machine-readable progress on stderr (bytes and, if known,
51 group = parser.add_mutually_exclusive_group()
52 group.add_argument('--hash',
54 Display the hash of each file as it is read from Keep, using the given
55 hash algorithm. Supported algorithms include md5, sha1, sha224,
56 sha256, sha384, and sha512.
58 group.add_argument('--md5sum', action='store_const',
59 dest='hash', const='md5',
61 Display the MD5 hash of each file as it is read from Keep.
63 parser.add_argument('-n', action='store_true',
65 Do not write any data -- just read from Keep, and report md5sums if
68 parser.add_argument('-r', action='store_true',
70 Retrieve all files in the specified collection/prefix. This is the
71 default behavior if the "locator" argument ends with a forward slash.
73 group = parser.add_mutually_exclusive_group()
74 group.add_argument('-f', action='store_true',
76 Overwrite existing files while writing. The default behavior is to
77 refuse to write *anything* if any of the output files already
78 exist. As a special case, -f is not needed to write to stdout.
80 group.add_argument('--skip-existing', action='store_true',
82 Skip files that already exist. The default behavior is to refuse to
83 write *anything* if any files exist that would have to be
84 overwritten. This option causes even devices, sockets, and fifos to be
88 def parse_arguments(arguments, stdout, stderr):
89 args = parser.parse_args(arguments)
91 if args.locator[-1] == os.sep:
95 not (args.destination and
96 os.path.isdir(args.destination))):
97 parser.error('Destination is not a directory.')
98 if not args.r and (os.path.isdir(args.destination) or
99 args.destination[-1] == os.path.sep):
100 args.destination = os.path.join(args.destination,
101 os.path.basename(args.locator))
102 logger.debug("Appended source file name to destination directory: %s",
105 if args.destination == '/dev/stdout':
106 args.destination = "-"
108 if args.destination == '-':
109 # Normally you have to use -f to write to a file (or device) that
110 # already exists, but "-" and "/dev/stdout" are common enough to
111 # merit a special exception.
114 args.destination = args.destination.rstrip(os.sep)
116 # Turn on --progress by default if stderr is a tty and output is
117 # either going to a named file, or going (via stdout) to something
119 if (not (args.batch_progress or args.no_progress)
121 and (args.destination != '-'
122 or not stdout.isatty())):
126 def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr):
129 if stdout is sys.stdout and hasattr(stdout, 'buffer'):
130 # in Python 3, write to stdout as binary
131 stdout = stdout.buffer
133 args = parse_arguments(arguments, stdout, stderr)
134 if api_client is None:
135 api_client = arvados.api('v1')
137 r = re.search(r'^(.*?)(/.*)?$', args.locator)
138 collection = r.group(1)
139 get_prefix = r.group(2)
140 if args.r and not get_prefix:
143 reader = arvados.CollectionReader(collection, num_retries=args.retries)
144 except Exception as error:
145 logger.error("failed to read collection: {}".format(error))
148 # User asked to download the collection's manifest
151 open_flags = os.O_CREAT | os.O_WRONLY
153 open_flags |= os.O_EXCL
155 if args.destination == "-":
156 stdout.write(reader.manifest_text().encode())
158 out_fd = os.open(args.destination, open_flags)
159 with os.fdopen(out_fd, 'wb') as out_file:
160 out_file.write(reader.manifest_text().encode())
161 except (IOError, OSError) as error:
162 logger.error("can't write to '{}': {}".format(args.destination, error))
164 except (arvados.errors.ApiError, arvados.errors.KeepReadError) as error:
165 logger.error("failed to download '{}': {}".format(collection, error))
169 # Scan the collection. Make an array of (stream, file, local
170 # destination filename) tuples, and add up total size to extract.
174 if get_prefix == os.sep:
177 item = reader.find('.' + get_prefix)
179 if isinstance(item, arvados.collection.Subcollection) or isinstance(item, arvados.collection.CollectionReader):
180 # If the user asked for a file and we got a subcollection, error out.
181 if get_prefix[-1] != os.sep:
182 logger.error("requested file '{}' is in fact a subcollection. Append a trailing '/' to download it.".format('.' + get_prefix))
184 # If the user asked stdout as a destination, error out.
185 elif args.destination == '-':
186 logger.error("cannot use 'stdout' as destination when downloading multiple files.")
188 # User asked for a subcollection, and that's what was found. Add up total size
190 for s, f in files_in_collection(item):
191 dest_path = os.path.join(
193 os.path.join(s.stream_name(), f.name)[len(get_prefix)+1:])
194 if (not (args.n or args.f or args.skip_existing) and
195 os.path.exists(dest_path)):
196 logger.error('Local file %s already exists.' % (dest_path,))
198 todo += [(s, f, dest_path)]
199 todo_bytes += f.size()
200 elif isinstance(item, arvados.arvfile.ArvadosFile):
201 todo += [(item.parent, item, args.destination)]
202 todo_bytes += item.size()
204 logger.error("'{}' not found.".format('.' + get_prefix))
206 except (IOError, arvados.errors.NotFoundError) as e:
211 for s, f, outfilename in todo:
215 if outfilename == "-":
218 if args.skip_existing and os.path.exists(outfilename):
219 logger.debug('Local file %s exists. Skipping.', outfilename)
221 elif not args.f and (os.path.isfile(outfilename) or
222 os.path.isdir(outfilename)):
223 # Good thing we looked again: apparently this file wasn't
224 # here yet when we checked earlier.
225 logger.error('Local file %s already exists.' % (outfilename,))
228 arvados.util.mkdir_dash_p(os.path.dirname(outfilename))
230 outfile = open(outfilename, 'wb')
231 except Exception as error:
232 logger.error('Open(%s) failed: %s' % (outfilename, error))
235 digestor = hashlib.new(args.hash)
237 with s.open(f.name, 'rb') as file_reader:
238 for data in file_reader.readall():
242 digestor.update(data)
243 out_bytes += len(data)
245 stderr.write('\r%d MiB / %d MiB %.1f%%' %
250 else 100.0*out_bytes/todo_bytes)))
251 elif args.batch_progress:
252 stderr.write('%s %d read %d total\n' %
253 (sys.argv[0], os.getpid(),
254 out_bytes, todo_bytes))
256 stderr.write("%s %s/%s\n"
257 % (digestor.hexdigest(), s.stream_name(), f.name))
258 except KeyboardInterrupt:
259 if outfile and (outfile.fileno() > 2) and not outfile.closed:
260 os.unlink(outfile.name)
263 if outfile != None and outfile != stdout:
270 def files_in_collection(c):
271 # Sort first by file type, then alphabetically by file path.
272 for i in sorted(list(c.keys()),
274 isinstance(c[k], arvados.collection.Subcollection),
276 if isinstance(c[i], arvados.arvfile.ArvadosFile):
278 elif isinstance(c[i], arvados.collection.Subcollection):
279 for s, f in files_in_collection(c[i]):