sdk/python/arvados/commands/put.py

   1 # Copyright (C) The Arvados Authors. All rights reserved.
   2 #
   3 # SPDX-License-Identifier: Apache-2.0
   4
   5 from __future__ import division
   6 from future.utils import listitems, listvalues
   7 from builtins import str
   8 from builtins import object
   9 import argparse
  10 import arvados
  11 import arvados.collection
  12 import base64
  13 import ciso8601
  14 import copy
  15 import datetime
  16 import errno
  17 import fcntl
  18 import fnmatch
  19 import hashlib
  20 import json
  21 import logging
  22 import os
  23 import pwd
  24 import re
  25 import signal
  26 import socket
  27 import sys
  28 import tempfile
  29 import threading
  30 import time
  31 import traceback
  32
  33 from apiclient import errors as apiclient_errors
  34 from arvados._version import __version__
  35 from arvados.util import keep_locator_pattern
  36
  37 import arvados.commands._util as arv_cmd
  38
  39 api_client = None
  40
  41 upload_opts = argparse.ArgumentParser(add_help=False)
  42
  43 upload_opts.add_argument('--version', action='version',
  44                          version="%s %s" % (sys.argv[0], __version__),
  45                          help='Print version and exit.')
  46 upload_opts.add_argument('paths', metavar='path', type=str, nargs='*',
  47                          help="""
  48 Local file or directory. If path is a directory reference with a trailing
  49 slash, then just upload the directory's contents; otherwise upload the
  50 directory itself. Default: read from standard input.
  51 """)
  52
  53 _group = upload_opts.add_mutually_exclusive_group()
  54
  55 _group.add_argument('--max-manifest-depth', type=int, metavar='N',
  56                     default=-1, help=argparse.SUPPRESS)
  57
  58 _group.add_argument('--normalize', action='store_true',
  59                     help="""
  60 Normalize the manifest by re-ordering files and streams after writing
  61 data.
  62 """)
  63
  64 _group.add_argument('--dry-run', action='store_true', default=False,
  65                     help="""
  66 Don't actually upload files, but only check if any file should be
  67 uploaded. Exit with code=2 when files are pending for upload.
  68 """)
  69
  70 _group = upload_opts.add_mutually_exclusive_group()
  71
  72 _group.add_argument('--as-stream', action='store_true', dest='stream',
  73                     help="""
  74 Synonym for --stream.
  75 """)
  76
  77 _group.add_argument('--stream', action='store_true',
  78                     help="""
  79 Store the file content and display the resulting manifest on
  80 stdout. Do not save a Collection object in Arvados.
  81 """)
  82
  83 _group.add_argument('--as-manifest', action='store_true', dest='manifest',
  84                     help="""
  85 Synonym for --manifest.
  86 """)
  87
  88 _group.add_argument('--in-manifest', action='store_true', dest='manifest',
  89                     help="""
  90 Synonym for --manifest.
  91 """)
  92
  93 _group.add_argument('--manifest', action='store_true',
  94                     help="""
  95 Store the file data and resulting manifest in Keep, save a Collection
  96 object in Arvados, and display the manifest locator (Collection uuid)
  97 on stdout. This is the default behavior.
  98 """)
  99
 100 _group.add_argument('--as-raw', action='store_true', dest='raw',
 101                     help="""
 102 Synonym for --raw.
 103 """)
 104
 105 _group.add_argument('--raw', action='store_true',
 106                     help="""
 107 Store the file content and display the data block locators on stdout,
 108 separated by commas, with a trailing newline. Do not store a
 109 manifest.
 110 """)
 111
 112 upload_opts.add_argument('--update-collection', type=str, default=None,
 113                          dest='update_collection', metavar="UUID", help="""
 114 Update an existing collection identified by the given Arvados collection
 115 UUID. All new local files will be uploaded.
 116 """)
 117
 118 upload_opts.add_argument('--use-filename', type=str, default=None,
 119                          dest='filename', help="""
 120 Synonym for --filename.
 121 """)
 122
 123 upload_opts.add_argument('--filename', type=str, default=None,
 124                          help="""
 125 Use the given filename in the manifest, instead of the name of the
 126 local file. This is useful when "-" or "/dev/stdin" is given as an
 127 input file. It can be used only if there is exactly one path given and
 128 it is not a directory. Implies --manifest.
 129 """)
 130
 131 upload_opts.add_argument('--portable-data-hash', action='store_true',
 132                          help="""
 133 Print the portable data hash instead of the Arvados UUID for the collection
 134 created by the upload.
 135 """)
 136
 137 upload_opts.add_argument('--replication', type=int, metavar='N', default=None,
 138                          help="""
 139 Set the replication level for the new collection: how many different
 140 physical storage devices (e.g., disks) should have a copy of each data
 141 block. Default is to use the server-provided default (if any) or 2.
 142 """)
 143
 144 upload_opts.add_argument('--storage-classes', help="""
 145 Specify comma separated list of storage classes to be used when saving data to Keep.
 146 """)
 147
 148 upload_opts.add_argument('--threads', type=int, metavar='N', default=None,
 149                          help="""
 150 Set the number of upload threads to be used. Take into account that
 151 using lots of threads will increase the RAM requirements. Default is
 152 to use 2 threads.
 153 On high latency installations, using a greater number will improve
 154 overall throughput.
 155 """)
 156
 157 upload_opts.add_argument('--exclude', metavar='PATTERN', default=[],
 158                       action='append', help="""
 159 Exclude files and directories whose names match the given glob pattern. When
 160 using a path-like pattern like 'subdir/*.txt', all text files inside 'subdir'
 161 directory, relative to the provided input dirs will be excluded.
 162 When using a filename pattern like '*.txt', any text file will be excluded
 163 no matter where it is placed.
 164 For the special case of needing to exclude only files or dirs directly below
 165 the given input directory, you can use a pattern like './exclude_this.gif'.
 166 You can specify multiple patterns by using this argument more than once.
 167 """)
 168
 169 _group = upload_opts.add_mutually_exclusive_group()
 170 _group.add_argument('--follow-links', action='store_true', default=True,
 171                     dest='follow_links', help="""
 172 Follow file and directory symlinks (default).
 173 """)
 174 _group.add_argument('--no-follow-links', action='store_false', dest='follow_links',
 175                     help="""
 176 Ignore file and directory symlinks. Even paths given explicitly on the
 177 command line will be skipped if they are symlinks.
 178 """)
 179
 180
 181 run_opts = argparse.ArgumentParser(add_help=False)
 182
 183 run_opts.add_argument('--project-uuid', metavar='UUID', help="""
 184 Store the collection in the specified project, instead of your Home
 185 project.
 186 """)
 187
 188 run_opts.add_argument('--name', help="""
 189 Save the collection with the specified name.
 190 """)
 191
 192 _group = run_opts.add_mutually_exclusive_group()
 193 _group.add_argument('--progress', action='store_true',
 194                     help="""
 195 Display human-readable progress on stderr (bytes and, if possible,
 196 percentage of total data size). This is the default behavior when
 197 stderr is a tty.
 198 """)
 199
 200 _group.add_argument('--no-progress', action='store_true',
 201                     help="""
 202 Do not display human-readable progress on stderr, even if stderr is a
 203 tty.
 204 """)
 205
 206 _group.add_argument('--batch-progress', action='store_true',
 207                     help="""
 208 Display machine-readable progress on stderr (bytes and, if known,
 209 total data size).
 210 """)
 211
 212 run_opts.add_argument('--silent', action='store_true',
 213                       help="""
 214 Do not print any debug messages to console. (Any error messages will
 215 still be displayed.)
 216 """)
 217
 218 _group = run_opts.add_mutually_exclusive_group()
 219 _group.add_argument('--resume', action='store_true', default=True,
 220                     help="""
 221 Continue interrupted uploads from cached state (default).
 222 """)
 223 _group.add_argument('--no-resume', action='store_false', dest='resume',
 224                     help="""
 225 Do not continue interrupted uploads from cached state.
 226 """)
 227
 228 _group = run_opts.add_mutually_exclusive_group()
 229 _group.add_argument('--cache', action='store_true', dest='use_cache', default=True,
 230                     help="""
 231 Save upload state in a cache file for resuming (default).
 232 """)
 233 _group.add_argument('--no-cache', action='store_false', dest='use_cache',
 234                     help="""
 235 Do not save upload state in a cache file for resuming.
 236 """)
 237
 238 _group = upload_opts.add_mutually_exclusive_group()
 239 _group.add_argument('--trash-at', metavar='YYYY-MM-DDTHH:MM', default=None,
 240                     help="""
 241 Set the trash date of the resulting collection to an absolute date in the future.
 242 The accepted format is defined by the ISO 8601 standard. Examples: 20090103, 2009-01-03, 20090103T181505, 2009-01-03T18:15:05.\n
 243 Timezone information can be added. If not, the provided date/time is assumed as being in the local system's timezone.
 244 """)
 245 _group.add_argument('--trash-after', type=int, metavar='DAYS', default=None,
 246                     help="""
 247 Set the trash date of the resulting collection to an amount of days from the
 248 date/time that the upload process finishes.
 249 """)
 250
 251 arg_parser = argparse.ArgumentParser(
 252     description='Copy data from the local filesystem to Keep.',
 253     parents=[upload_opts, run_opts, arv_cmd.retry_opt])
 254
 255 def parse_arguments(arguments):
 256     args = arg_parser.parse_args(arguments)
 257
 258     if len(args.paths) == 0:
 259         args.paths = ['-']
 260
 261     args.paths = ["-" if x == "/dev/stdin" else x for x in args.paths]
 262
 263     if args.filename and (len(args.paths) != 1 or os.path.isdir(args.paths[0])):
 264         arg_parser.error("""
 265     --filename argument cannot be used when storing a directory or
 266     multiple files.
 267     """)
 268
 269     # Turn on --progress by default if stderr is a tty.
 270     if (not (args.batch_progress or args.no_progress or args.silent)
 271         and os.isatty(sys.stderr.fileno())):
 272         args.progress = True
 273
 274     # Turn off --resume (default) if --no-cache is used.
 275     if not args.use_cache:
 276         args.resume = False
 277
 278     if args.paths == ['-']:
 279         if args.update_collection:
 280             arg_parser.error("""
 281     --update-collection cannot be used when reading from stdin.
 282     """)
 283         args.resume = False
 284         args.use_cache = False
 285         if not args.filename:
 286             args.filename = 'stdin'
 287
 288     # Remove possible duplicated patterns
 289     if len(args.exclude) > 0:
 290         args.exclude = list(set(args.exclude))
 291
 292     return args
 293
 294
 295 class PathDoesNotExistError(Exception):
 296     pass
 297
 298
 299 class CollectionUpdateError(Exception):
 300     pass
 301
 302
 303 class ResumeCacheConflict(Exception):
 304     pass
 305
 306
 307 class ResumeCacheInvalidError(Exception):
 308     pass
 309
 310 class ArvPutArgumentConflict(Exception):
 311     pass
 312
 313
 314 class ArvPutUploadIsPending(Exception):
 315     pass
 316
 317
 318 class ArvPutUploadNotPending(Exception):
 319     pass
 320
 321
 322 class FileUploadList(list):
 323     def __init__(self, dry_run=False):
 324         list.__init__(self)
 325         self.dry_run = dry_run
 326
 327     def append(self, other):
 328         if self.dry_run:
 329             raise ArvPutUploadIsPending()
 330         super(FileUploadList, self).append(other)
 331
 332
 333 # Appends the X-Request-Id to the log message when log level is ERROR or DEBUG
 334 class ArvPutLogFormatter(logging.Formatter):
 335     std_fmtr = logging.Formatter(arvados.log_format, arvados.log_date_format)
 336     err_fmtr = None
 337     request_id_informed = False
 338
 339     def __init__(self, request_id):
 340         self.err_fmtr = logging.Formatter(
 341             arvados.log_format+' (X-Request-Id: {})'.format(request_id),
 342             arvados.log_date_format)
 343
 344     def format(self, record):
 345         if (not self.request_id_informed) and (record.levelno in (logging.DEBUG, logging.ERROR)):
 346             self.request_id_informed = True
 347             return self.err_fmtr.format(record)
 348         return self.std_fmtr.format(record)
 349
 350
 351 class ResumeCache(object):
 352     CACHE_DIR = '.cache/arvados/arv-put'
 353
 354     def __init__(self, file_spec):
 355         self.cache_file = open(file_spec, 'a+')
 356         self._lock_file(self.cache_file)
 357         self.filename = self.cache_file.name
 358
 359     @classmethod
 360     def make_path(cls, args):
 361         md5 = hashlib.md5()
 362         md5.update(arvados.config.get('ARVADOS_API_HOST', '!nohost').encode())
 363         realpaths = sorted(os.path.realpath(path) for path in args.paths)
 364         md5.update(b'\0'.join([p.encode() for p in realpaths]))
 365         if any(os.path.isdir(path) for path in realpaths):
 366             md5.update(b'-1')
 367         elif args.filename:
 368             md5.update(args.filename.encode())
 369         return os.path.join(
 370             arv_cmd.make_home_conf_dir(cls.CACHE_DIR, 0o700, 'raise'),
 371             md5.hexdigest())
 372
 373     def _lock_file(self, fileobj):
 374         try:
 375             fcntl.flock(fileobj, fcntl.LOCK_EX | fcntl.LOCK_NB)
 376         except IOError:
 377             raise ResumeCacheConflict(u"{} locked".format(fileobj.name))
 378
 379     def load(self):
 380         self.cache_file.seek(0)
 381         return json.load(self.cache_file)
 382
 383     def check_cache(self, api_client=None, num_retries=0):
 384         try:
 385             state = self.load()
 386             locator = None
 387             try:
 388                 if "_finished_streams" in state and len(state["_finished_streams"]) > 0:
 389                     locator = state["_finished_streams"][0][1][0]
 390                 elif "_current_stream_locators" in state and len(state["_current_stream_locators"]) > 0:
 391                     locator = state["_current_stream_locators"][0]
 392                 if locator is not None:
 393                     kc = arvados.keep.KeepClient(api_client=api_client)
 394                     kc.head(locator, num_retries=num_retries)
 395             except Exception as e:
 396                 self.restart()
 397         except (ValueError):
 398             pass
 399
 400     def save(self, data):
 401         try:
 402             new_cache_fd, new_cache_name = tempfile.mkstemp(
 403                 dir=os.path.dirname(self.filename))
 404             self._lock_file(new_cache_fd)
 405             new_cache = os.fdopen(new_cache_fd, 'r+')
 406             json.dump(data, new_cache)
 407             os.rename(new_cache_name, self.filename)
 408         except (IOError, OSError, ResumeCacheConflict):
 409             try:
 410                 os.unlink(new_cache_name)
 411             except NameError:  # mkstemp failed.
 412                 pass
 413         else:
 414             self.cache_file.close()
 415             self.cache_file = new_cache
 416
 417     def close(self):
 418         self.cache_file.close()
 419
 420     def destroy(self):
 421         try:
 422             os.unlink(self.filename)
 423         except OSError as error:
 424             if error.errno != errno.ENOENT:  # That's what we wanted anyway.
 425                 raise
 426         self.close()
 427
 428     def restart(self):
 429         self.destroy()
 430         self.__init__(self.filename)
 431
 432
 433 class ArvPutUploadJob(object):
 434     CACHE_DIR = '.cache/arvados/arv-put'
 435     EMPTY_STATE = {
 436         'manifest' : None, # Last saved manifest checkpoint
 437         'files' : {} # Previous run file list: {path : {size, mtime}}
 438     }
 439
 440     def __init__(self, paths, resume=True, use_cache=True, reporter=None,
 441                  name=None, owner_uuid=None, api_client=None,
 442                  ensure_unique_name=False, num_retries=None,
 443                  put_threads=None, replication_desired=None, filename=None,
 444                  update_time=60.0, update_collection=None, storage_classes=None,
 445                  logger=logging.getLogger('arvados.arv_put'), dry_run=False,
 446                  follow_links=True, exclude_paths=[], exclude_names=None,
 447                  trash_at=None):
 448         self.paths = paths
 449         self.resume = resume
 450         self.use_cache = use_cache
 451         self.update = False
 452         self.reporter = reporter
 453         # This will set to 0 before start counting, if no special files are going
 454         # to be read.
 455         self.bytes_expected = None
 456         self.bytes_written = 0
 457         self.bytes_skipped = 0
 458         self.name = name
 459         self.owner_uuid = owner_uuid
 460         self.ensure_unique_name = ensure_unique_name
 461         self.num_retries = num_retries
 462         self.replication_desired = replication_desired
 463         self.put_threads = put_threads
 464         self.filename = filename
 465         self.storage_classes = storage_classes
 466         self._api_client = api_client
 467         self._state_lock = threading.Lock()
 468         self._state = None # Previous run state (file list & manifest)
 469         self._current_files = [] # Current run file list
 470         self._cache_file = None
 471         self._collection_lock = threading.Lock()
 472         self._remote_collection = None # Collection being updated (if asked)
 473         self._local_collection = None # Collection from previous run manifest
 474         self._file_paths = set() # Files to be updated in remote collection
 475         self._stop_checkpointer = threading.Event()
 476         self._checkpointer = threading.Thread(target=self._update_task)
 477         self._checkpointer.daemon = True
 478         self._update_task_time = update_time  # How many seconds wait between update runs
 479         self._files_to_upload = FileUploadList(dry_run=dry_run)
 480         self._upload_started = False
 481         self.logger = logger
 482         self.dry_run = dry_run
 483         self._checkpoint_before_quit = True
 484         self.follow_links = follow_links
 485         self.exclude_paths = exclude_paths
 486         self.exclude_names = exclude_names
 487         self._trash_at = trash_at
 488
 489         if self._trash_at is not None:
 490             if type(self._trash_at) not in [datetime.datetime, datetime.timedelta]:
 491                 raise TypeError('trash_at should be None, timezone-naive datetime or timedelta')
 492             if type(self._trash_at) == datetime.datetime and self._trash_at.tzinfo is not None:
 493                 raise TypeError('provided trash_at datetime should be timezone-naive')
 494
 495         if not self.use_cache and self.resume:
 496             raise ArvPutArgumentConflict('resume cannot be True when use_cache is False')
 497
 498         # Check for obvious dry-run responses
 499         if self.dry_run and (not self.use_cache or not self.resume):
 500             raise ArvPutUploadIsPending()
 501
 502         # Load cached data if any and if needed
 503         self._setup_state(update_collection)
 504
 505         # Build the upload file list, excluding requested files and counting the
 506         # bytes expected to be uploaded.
 507         self._build_upload_list()
 508
 509     def _build_upload_list(self):
 510         """
 511         Scan the requested paths to count file sizes, excluding requested files
 512         and dirs and building the upload file list.
 513         """
 514         # If there aren't special files to be read, reset total bytes count to zero
 515         # to start counting.
 516         if not any([p for p in self.paths
 517                     if not (os.path.isfile(p) or os.path.isdir(p))]):
 518             self.bytes_expected = 0
 519
 520         for path in self.paths:
 521             # Test for stdin first, in case some file named '-' exist
 522             if path == '-':
 523                 if self.dry_run:
 524                     raise ArvPutUploadIsPending()
 525                 self._write_stdin(self.filename or 'stdin')
 526             elif not os.path.exists(path):
 527                  raise PathDoesNotExistError(u"file or directory '{}' does not exist.".format(path))
 528             elif (not self.follow_links) and os.path.islink(path):
 529                 self.logger.warning("Skipping symlink '{}'".format(path))
 530                 continue
 531             elif os.path.isdir(path):
 532                 # Use absolute paths on cache index so CWD doesn't interfere
 533                 # with the caching logic.
 534                 orig_path = path
 535                 path = os.path.abspath(path)
 536                 if orig_path[-1:] == os.sep:
 537                     # When passing a directory reference with a trailing slash,
 538                     # its contents should be uploaded directly to the
 539                     # collection's root.
 540                     prefixdir = path
 541                 else:
 542                     # When passing a directory reference with no trailing slash,
 543                     # upload the directory to the collection's root.
 544                     prefixdir = os.path.dirname(path)
 545                 prefixdir += os.sep
 546                 for root, dirs, files in os.walk(path,
 547                                                  followlinks=self.follow_links):
 548                     root_relpath = os.path.relpath(root, path)
 549                     if root_relpath == '.':
 550                         root_relpath = ''
 551                     # Exclude files/dirs by full path matching pattern
 552                     if self.exclude_paths:
 553                         dirs[:] = [d for d in dirs
 554                                    if not any(pathname_match(
 555                                            os.path.join(root_relpath, d), pat)
 556                                               for pat in self.exclude_paths)]
 557                         files = [f for f in files
 558                                  if not any(pathname_match(
 559                                          os.path.join(root_relpath, f), pat)
 560                                             for pat in self.exclude_paths)]
 561                     # Exclude files/dirs by name matching pattern
 562                     if self.exclude_names is not None:
 563                         dirs[:] = [d for d in dirs
 564                                    if not self.exclude_names.match(d)]
 565                         files = [f for f in files
 566                                  if not self.exclude_names.match(f)]
 567                     # Make os.walk()'s dir traversing order deterministic
 568                     dirs.sort()
 569                     files.sort()
 570                     for f in files:
 571                         filepath = os.path.join(root, f)
 572                         # Add its size to the total bytes count (if applicable)
 573                         if self.follow_links or (not os.path.islink(filepath)):
 574                             if self.bytes_expected is not None:
 575                                 self.bytes_expected += os.path.getsize(filepath)
 576                         self._check_file(filepath,
 577                                          os.path.join(root[len(prefixdir):], f))
 578             else:
 579                 filepath = os.path.abspath(path)
 580                 # Add its size to the total bytes count (if applicable)
 581                 if self.follow_links or (not os.path.islink(filepath)):
 582                     if self.bytes_expected is not None:
 583                         self.bytes_expected += os.path.getsize(filepath)
 584                 self._check_file(filepath,
 585                                  self.filename or os.path.basename(path))
 586         # If dry-mode is on, and got up to this point, then we should notify that
 587         # there aren't any file to upload.
 588         if self.dry_run:
 589             raise ArvPutUploadNotPending()
 590         # Remove local_collection's files that don't exist locally anymore, so the
 591         # bytes_written count is correct.
 592         for f in self.collection_file_paths(self._local_collection,
 593                                             path_prefix=""):
 594             if f != 'stdin' and f != self.filename and not f in self._file_paths:
 595                 self._local_collection.remove(f)
 596
 597     def start(self, save_collection):
 598         """
 599         Start supporting thread & file uploading
 600         """
 601         self._checkpointer.start()
 602         try:
 603             # Update bytes_written from current local collection and
 604             # report initial progress.
 605             self._update()
 606             # Actual file upload
 607             self._upload_started = True # Used by the update thread to start checkpointing
 608             self._upload_files()
 609         except (SystemExit, Exception) as e:
 610             self._checkpoint_before_quit = False
 611             # Log stack trace only when Ctrl-C isn't pressed (SIGINT)
 612             # Note: We're expecting SystemExit instead of
 613             # KeyboardInterrupt because we have a custom signal
 614             # handler in place that raises SystemExit with the catched
 615             # signal's code.
 616             if isinstance(e, PathDoesNotExistError):
 617                 # We aren't interested in the traceback for this case
 618                 pass
 619             elif not isinstance(e, SystemExit) or e.code != -2:
 620                 self.logger.warning("Abnormal termination:\n{}".format(
 621                     traceback.format_exc()))
 622             raise
 623         finally:
 624             if not self.dry_run:
 625                 # Stop the thread before doing anything else
 626                 self._stop_checkpointer.set()
 627                 self._checkpointer.join()
 628                 if self._checkpoint_before_quit:
 629                     # Commit all pending blocks & one last _update()
 630                     self._local_collection.manifest_text()
 631                     self._update(final=True)
 632                     if save_collection:
 633                         self.save_collection()
 634             if self.use_cache:
 635                 self._cache_file.close()
 636
 637     def _collection_trash_at(self):
 638         """
 639         Returns the trash date that the collection should use at save time.
 640         Takes into account absolute/relative trash_at values requested
 641         by the user.
 642         """
 643         if type(self._trash_at) == datetime.timedelta:
 644             # Get an absolute datetime for trash_at
 645             return datetime.datetime.utcnow() + self._trash_at
 646         return self._trash_at
 647
 648     def save_collection(self):
 649         if self.update:
 650             # Check if files should be updated on the remote collection.
 651             for fp in self._file_paths:
 652                 remote_file = self._remote_collection.find(fp)
 653                 if not remote_file:
 654                     # File don't exist on remote collection, copy it.
 655                     self._remote_collection.copy(fp, fp, self._local_collection)
 656                 elif remote_file != self._local_collection.find(fp):
 657                     # A different file exist on remote collection, overwrite it.
 658                     self._remote_collection.copy(fp, fp, self._local_collection, overwrite=True)
 659                 else:
 660                     # The file already exist on remote collection, skip it.
 661                     pass
 662             self._remote_collection.save(storage_classes=self.storage_classes,
 663                                          num_retries=self.num_retries,
 664                                          trash_at=self._collection_trash_at())
 665         else:
 666             if self.storage_classes is None:
 667                 self.storage_classes = ['default']
 668             self._local_collection.save_new(
 669                 name=self.name, owner_uuid=self.owner_uuid,
 670                 storage_classes=self.storage_classes,
 671                 ensure_unique_name=self.ensure_unique_name,
 672                 num_retries=self.num_retries,
 673                 trash_at=self._collection_trash_at())
 674
 675     def destroy_cache(self):
 676         if self.use_cache:
 677             try:
 678                 os.unlink(self._cache_filename)
 679             except OSError as error:
 680                 # That's what we wanted anyway.
 681                 if error.errno != errno.ENOENT:
 682                     raise
 683             self._cache_file.close()
 684
 685     def _collection_size(self, collection):
 686         """
 687         Recursively get the total size of the collection
 688         """
 689         size = 0
 690         for item in listvalues(collection):
 691             if isinstance(item, arvados.collection.Collection) or isinstance(item, arvados.collection.Subcollection):
 692                 size += self._collection_size(item)
 693             else:
 694                 size += item.size()
 695         return size
 696
 697     def _update_task(self):
 698         """
 699         Periodically called support task. File uploading is
 700         asynchronous so we poll status from the collection.
 701         """
 702         while not self._stop_checkpointer.wait(1 if not self._upload_started else self._update_task_time):
 703             self._update()
 704
 705     def _update(self, final=False):
 706         """
 707         Update cached manifest text and report progress.
 708         """
 709         if self._upload_started:
 710             with self._collection_lock:
 711                 self.bytes_written = self._collection_size(self._local_collection)
 712                 if self.use_cache:
 713                     if final:
 714                         manifest = self._local_collection.manifest_text()
 715                     else:
 716                         # Get the manifest text without comitting pending blocks
 717                         manifest = self._local_collection.manifest_text(strip=False,
 718                                                                         normalize=False,
 719                                                                         only_committed=True)
 720                     # Update cache
 721                     with self._state_lock:
 722                         self._state['manifest'] = manifest
 723             if self.use_cache:
 724                 try:
 725                     self._save_state()
 726                 except Exception as e:
 727                     self.logger.error("Unexpected error trying to save cache file: {}".format(e))
 728             # Keep remote collection's trash_at attribute synced when using relative expire dates
 729             if self._remote_collection is not None and type(self._trash_at) == datetime.timedelta:
 730                 try:
 731                     self._api_client.collections().update(
 732                         uuid=self._remote_collection.manifest_locator(),
 733                         body={'trash_at': self._collection_trash_at().strftime("%Y-%m-%dT%H:%M:%S.%fZ")}
 734                     ).execute(num_retries=self.num_retries)
 735                 except Exception as e:
 736                     self.logger.error("Unexpected error trying to update remote collection's expire date: {}".format(e))
 737         else:
 738             self.bytes_written = self.bytes_skipped
 739         # Call the reporter, if any
 740         self.report_progress()
 741
 742     def report_progress(self):
 743         if self.reporter is not None:
 744             self.reporter(self.bytes_written, self.bytes_expected)
 745
 746     def _write_stdin(self, filename):
 747         output = self._local_collection.open(filename, 'wb')
 748         self._write(sys.stdin.buffer, output)
 749         output.close()
 750
 751     def _check_file(self, source, filename):
 752         """
 753         Check if this file needs to be uploaded
 754         """
 755         # Ignore symlinks when requested
 756         if (not self.follow_links) and os.path.islink(source):
 757             return
 758         resume_offset = 0
 759         should_upload = False
 760         new_file_in_cache = False
 761         # Record file path for updating the remote collection before exiting
 762         self._file_paths.add(filename)
 763
 764         with self._state_lock:
 765             # If no previous cached data on this file, store it for an eventual
 766             # repeated run.
 767             if source not in self._state['files']:
 768                 self._state['files'][source] = {
 769                     'mtime': os.path.getmtime(source),
 770                     'size' : os.path.getsize(source)
 771                 }
 772                 new_file_in_cache = True
 773             cached_file_data = self._state['files'][source]
 774
 775         # Check if file was already uploaded (at least partially)
 776         file_in_local_collection = self._local_collection.find(filename)
 777
 778         # If not resuming, upload the full file.
 779         if not self.resume:
 780             should_upload = True
 781         # New file detected from last run, upload it.
 782         elif new_file_in_cache:
 783             should_upload = True
 784         # Local file didn't change from last run.
 785         elif cached_file_data['mtime'] == os.path.getmtime(source) and cached_file_data['size'] == os.path.getsize(source):
 786             if not file_in_local_collection:
 787                 # File not uploaded yet, upload it completely
 788                 should_upload = True
 789             elif file_in_local_collection.permission_expired():
 790                 # Permission token expired, re-upload file. This will change whenever
 791                 # we have a API for refreshing tokens.
 792                 self.logger.warning(u"Uploaded file '{}' access token expired, will re-upload it from scratch".format(filename))
 793                 should_upload = True
 794                 self._local_collection.remove(filename)
 795             elif cached_file_data['size'] == file_in_local_collection.size():
 796                 # File already there, skip it.
 797                 self.bytes_skipped += cached_file_data['size']
 798             elif cached_file_data['size'] > file_in_local_collection.size():
 799                 # File partially uploaded, resume!
 800                 resume_offset = file_in_local_collection.size()
 801                 self.bytes_skipped += resume_offset
 802                 should_upload = True
 803             else:
 804                 # Inconsistent cache, re-upload the file
 805                 should_upload = True
 806                 self._local_collection.remove(filename)
 807                 self.logger.warning(u"Uploaded version of file '{}' is bigger than local version, will re-upload it from scratch.".format(source))
 808         # Local file differs from cached data, re-upload it.
 809         else:
 810             if file_in_local_collection:
 811                 self._local_collection.remove(filename)
 812             should_upload = True
 813
 814         if should_upload:
 815             try:
 816                 self._files_to_upload.append((source, resume_offset, filename))
 817             except ArvPutUploadIsPending:
 818                 # This could happen when running on dry-mode, close cache file to
 819                 # avoid locking issues.
 820                 self._cache_file.close()
 821                 raise
 822
 823     def _upload_files(self):
 824         for source, resume_offset, filename in self._files_to_upload:
 825             with open(source, 'rb') as source_fd:
 826                 with self._state_lock:
 827                     self._state['files'][source]['mtime'] = os.path.getmtime(source)
 828                     self._state['files'][source]['size'] = os.path.getsize(source)
 829                 if resume_offset > 0:
 830                     # Start upload where we left off
 831                     output = self._local_collection.open(filename, 'ab')
 832                     source_fd.seek(resume_offset)
 833                 else:
 834                     # Start from scratch
 835                     output = self._local_collection.open(filename, 'wb')
 836                 self._write(source_fd, output)
 837                 output.close(flush=False)
 838
 839     def _write(self, source_fd, output):
 840         while True:
 841             data = source_fd.read(arvados.config.KEEP_BLOCK_SIZE)
 842             if not data:
 843                 break
 844             output.write(data)
 845
 846     def _my_collection(self):
 847         return self._remote_collection if self.update else self._local_collection
 848
 849     def _get_cache_filepath(self):
 850         # Set up cache file name from input paths.
 851         md5 = hashlib.md5()
 852         md5.update(arvados.config.get('ARVADOS_API_HOST', '!nohost').encode())
 853         realpaths = sorted(os.path.realpath(path) for path in self.paths)
 854         md5.update(b'\0'.join([p.encode() for p in realpaths]))
 855         if self.filename:
 856             md5.update(self.filename.encode())
 857         cache_filename = md5.hexdigest()
 858         cache_filepath = os.path.join(
 859             arv_cmd.make_home_conf_dir(self.CACHE_DIR, 0o700, 'raise'),
 860             cache_filename)
 861         return cache_filepath
 862
 863     def _setup_state(self, update_collection):
 864         """
 865         Create a new cache file or load a previously existing one.
 866         """
 867         # Load an already existing collection for update
 868         if update_collection and re.match(arvados.util.collection_uuid_pattern,
 869                                           update_collection):
 870             try:
 871                 self._remote_collection = arvados.collection.Collection(
 872                     update_collection,
 873                     api_client=self._api_client,
 874                     num_retries=self.num_retries)
 875             except arvados.errors.ApiError as error:
 876                 raise CollectionUpdateError("Cannot read collection {} ({})".format(update_collection, error))
 877             else:
 878                 self.update = True
 879         elif update_collection:
 880             # Collection locator provided, but unknown format
 881             raise CollectionUpdateError("Collection locator unknown: '{}'".format(update_collection))
 882
 883         if self.use_cache:
 884             cache_filepath = self._get_cache_filepath()
 885             if self.resume and os.path.exists(cache_filepath):
 886                 self.logger.info(u"Resuming upload from cache file {}".format(cache_filepath))
 887                 self._cache_file = open(cache_filepath, 'a+')
 888             else:
 889                 # --no-resume means start with a empty cache file.
 890                 self.logger.info(u"Creating new cache file at {}".format(cache_filepath))
 891                 self._cache_file = open(cache_filepath, 'w+')
 892             self._cache_filename = self._cache_file.name
 893             self._lock_file(self._cache_file)
 894             self._cache_file.seek(0)
 895
 896         with self._state_lock:
 897             if self.use_cache:
 898                 try:
 899                     self._state = json.load(self._cache_file)
 900                     if not set(['manifest', 'files']).issubset(set(self._state.keys())):
 901                         # Cache at least partially incomplete, set up new cache
 902                         self._state = copy.deepcopy(self.EMPTY_STATE)
 903                 except ValueError:
 904                     # Cache file empty, set up new cache
 905                     self._state = copy.deepcopy(self.EMPTY_STATE)
 906             else:
 907                 self.logger.info("No cache usage requested for this run.")
 908                 # No cache file, set empty state
 909                 self._state = copy.deepcopy(self.EMPTY_STATE)
 910             if not self._cached_manifest_valid():
 911                 raise ResumeCacheInvalidError()
 912             # Load the previous manifest so we can check if files were modified remotely.
 913             self._local_collection = arvados.collection.Collection(
 914                 self._state['manifest'],
 915                 replication_desired=self.replication_desired,
 916                 put_threads=self.put_threads,
 917                 api_client=self._api_client,
 918                 num_retries=self.num_retries)
 919
 920     def _cached_manifest_valid(self):
 921         """
 922         Validate the oldest non-expired block signature to check if cached manifest
 923         is usable: checking if the cached manifest was not created with a different
 924         arvados account.
 925         """
 926         if self._state.get('manifest', None) is None:
 927             # No cached manifest yet, all good.
 928             return True
 929         now = datetime.datetime.utcnow()
 930         oldest_exp = None
 931         oldest_loc = None
 932         block_found = False
 933         for m in keep_locator_pattern.finditer(self._state['manifest']):
 934             loc = m.group(0)
 935             try:
 936                 exp = datetime.datetime.utcfromtimestamp(int(loc.split('@')[1], 16))
 937             except IndexError:
 938                 # Locator without signature
 939                 continue
 940             block_found = True
 941             if exp > now and (oldest_exp is None or exp < oldest_exp):
 942                 oldest_exp = exp
 943                 oldest_loc = loc
 944         if not block_found:
 945             # No block signatures found => no invalid block signatures.
 946             return True
 947         if oldest_loc is None:
 948             # Locator signatures found, but all have expired.
 949             # Reset the cache and move on.
 950             self.logger.info('Cache expired, starting from scratch.')
 951             self._state['manifest'] = ''
 952             return True
 953         kc = arvados.KeepClient(api_client=self._api_client,
 954                                 num_retries=self.num_retries)
 955         try:
 956             kc.head(oldest_loc)
 957         except arvados.errors.KeepRequestError:
 958             # Something is wrong, cached manifest is not valid.
 959             return False
 960         return True
 961
 962     def collection_file_paths(self, col, path_prefix='.'):
 963         """Return a list of file paths by recursively go through the entire collection `col`"""
 964         file_paths = []
 965         for name, item in listitems(col):
 966             if isinstance(item, arvados.arvfile.ArvadosFile):
 967                 file_paths.append(os.path.join(path_prefix, name))
 968             elif isinstance(item, arvados.collection.Subcollection):
 969                 new_prefix = os.path.join(path_prefix, name)
 970                 file_paths += self.collection_file_paths(item, path_prefix=new_prefix)
 971         return file_paths
 972
 973     def _lock_file(self, fileobj):
 974         try:
 975             fcntl.flock(fileobj, fcntl.LOCK_EX | fcntl.LOCK_NB)
 976         except IOError:
 977             raise ResumeCacheConflict(u"{} locked".format(fileobj.name))
 978
 979     def _save_state(self):
 980         """
 981         Atomically save current state into cache.
 982         """
 983         with self._state_lock:
 984             # We're not using copy.deepcopy() here because it's a lot slower
 985             # than json.dumps(), and we're already needing JSON format to be
 986             # saved on disk.
 987             state = json.dumps(self._state)
 988         try:
 989             new_cache = tempfile.NamedTemporaryFile(
 990                 mode='w+',
 991                 dir=os.path.dirname(self._cache_filename), delete=False)
 992             self._lock_file(new_cache)
 993             new_cache.write(state)
 994             new_cache.flush()
 995             os.fsync(new_cache)
 996             os.rename(new_cache.name, self._cache_filename)
 997         except (IOError, OSError, ResumeCacheConflict) as error:
 998             self.logger.error("There was a problem while saving the cache file: {}".format(error))
 999             try:
1000                 os.unlink(new_cache_name)
1001             except NameError:  # mkstemp failed.
1002                 pass
1003         else:
1004             self._cache_file.close()
1005             self._cache_file = new_cache
1006
1007     def collection_name(self):
1008         return self._my_collection().api_response()['name'] if self._my_collection().api_response() else None
1009
1010     def collection_trash_at(self):
1011         return self._my_collection().get_trash_at()
1012
1013     def manifest_locator(self):
1014         return self._my_collection().manifest_locator()
1015
1016     def portable_data_hash(self):
1017         pdh = self._my_collection().portable_data_hash()
1018         m = self._my_collection().stripped_manifest().encode()
1019         local_pdh = '{}+{}'.format(hashlib.md5(m).hexdigest(), len(m))
1020         if pdh != local_pdh:
1021             self.logger.warning("\n".join([
1022                 "arv-put: API server provided PDH differs from local manifest.",
1023                 "         This should not happen; showing API server version."]))
1024         return pdh
1025
1026     def manifest_text(self, stream_name=".", strip=False, normalize=False):
1027         return self._my_collection().manifest_text(stream_name, strip, normalize)
1028
1029     def _datablocks_on_item(self, item):
1030         """
1031         Return a list of datablock locators, recursively navigating
1032         through subcollections
1033         """
1034         if isinstance(item, arvados.arvfile.ArvadosFile):
1035             if item.size() == 0:
1036                 # Empty file locator
1037                 return ["d41d8cd98f00b204e9800998ecf8427e+0"]
1038             else:
1039                 locators = []
1040                 for segment in item.segments():
1041                     loc = segment.locator
1042                     locators.append(loc)
1043                 return locators
1044         elif isinstance(item, arvados.collection.Collection):
1045             l = [self._datablocks_on_item(x) for x in listvalues(item)]
1046             # Fast list flattener method taken from:
1047             # http://stackoverflow.com/questions/952914/making-a-flat-list-out-of-list-of-lists-in-python
1048             return [loc for sublist in l for loc in sublist]
1049         else:
1050             return None
1051
1052     def data_locators(self):
1053         with self._collection_lock:
1054             # Make sure all datablocks are flushed before getting the locators
1055             self._my_collection().manifest_text()
1056             datablocks = self._datablocks_on_item(self._my_collection())
1057         return datablocks
1058
1059 _machine_format = "{} {}: {{}} written {{}} total\n".format(sys.argv[0],
1060                                                             os.getpid())
1061
1062 # Simulate glob.glob() matching behavior without the need to scan the filesystem
1063 # Note: fnmatch() doesn't work correctly when used with pathnames. For example the
1064 # pattern 'tests/*.py' will match 'tests/run_test.py' and also 'tests/subdir/run_test.py',
1065 # so instead we're using it on every path component.
1066 def pathname_match(pathname, pattern):
1067     name = pathname.split(os.sep)
1068     # Fix patterns like 'some/subdir/' or 'some//subdir'
1069     pat = [x for x in pattern.split(os.sep) if x != '' and x != '.']
1070     if len(name) != len(pat):
1071         return False
1072     for i in range(len(name)):
1073         if not fnmatch.fnmatch(name[i], pat[i]):
1074             return False
1075     return True
1076
1077 def machine_progress(bytes_written, bytes_expected):
1078     return _machine_format.format(
1079         bytes_written, -1 if (bytes_expected is None) else bytes_expected)
1080
1081 def human_progress(bytes_written, bytes_expected):
1082     if bytes_expected:
1083         return "\r{}M / {}M {:.1%} ".format(
1084             bytes_written >> 20, bytes_expected >> 20,
1085             float(bytes_written) / bytes_expected)
1086     else:
1087         return "\r{} ".format(bytes_written)
1088
1089 def progress_writer(progress_func, outfile=sys.stderr):
1090     def write_progress(bytes_written, bytes_expected):
1091         outfile.write(progress_func(bytes_written, bytes_expected))
1092     return write_progress
1093
1094 def desired_project_uuid(api_client, project_uuid, num_retries):
1095     if not project_uuid:
1096         query = api_client.users().current()
1097     elif arvados.util.user_uuid_pattern.match(project_uuid):
1098         query = api_client.users().get(uuid=project_uuid)
1099     elif arvados.util.group_uuid_pattern.match(project_uuid):
1100         query = api_client.groups().get(uuid=project_uuid)
1101     else:
1102         raise ValueError("Not a valid project UUID: {}".format(project_uuid))
1103     return query.execute(num_retries=num_retries)['uuid']
1104
1105 def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr,
1106          install_sig_handlers=True):
1107     global api_client
1108
1109     args = parse_arguments(arguments)
1110     logger = logging.getLogger('arvados.arv_put')
1111     if args.silent:
1112         logger.setLevel(logging.WARNING)
1113     else:
1114         logger.setLevel(logging.INFO)
1115     status = 0
1116
1117     request_id = arvados.util.new_request_id()
1118
1119     formatter = ArvPutLogFormatter(request_id)
1120     logging.getLogger('arvados').handlers[0].setFormatter(formatter)
1121
1122     if api_client is None:
1123         api_client = arvados.api('v1', request_id=request_id)
1124
1125     if install_sig_handlers:
1126         arv_cmd.install_signal_handlers()
1127
1128     # Trash arguments validation
1129     trash_at = None
1130     if args.trash_at is not None:
1131         # ciso8601 considers YYYYMM as invalid but YYYY-MM as valid, so here we
1132         # make sure the user provides a complete YYYY-MM-DD date.
1133         if not re.match(r'^\d{4}(?P<dash>-?)\d{2}?(?P=dash)\d{2}', args.trash_at):
1134             logger.error("--trash-at argument format invalid, use --help to see examples.")
1135             sys.exit(1)
1136         # Check if no time information was provided. In that case, assume end-of-day.
1137         if re.match(r'^\d{4}(?P<dash>-?)\d{2}?(?P=dash)\d{2}$', args.trash_at):
1138             args.trash_at += 'T23:59:59'
1139         try:
1140             trash_at = ciso8601.parse_datetime(args.trash_at)
1141         except:
1142             logger.error("--trash-at argument format invalid, use --help to see examples.")
1143             sys.exit(1)
1144         else:
1145             if trash_at.tzinfo is not None:
1146                 # Timezone aware datetime provided.
1147                 utcoffset = -trash_at.utcoffset()
1148             else:
1149                 # Timezone naive datetime provided. Assume is local.
1150                 if time.daylight:
1151                     utcoffset = datetime.timedelta(seconds=time.altzone)
1152                 else:
1153                     utcoffset = datetime.timedelta(seconds=time.timezone)
1154             # Convert to UTC timezone naive datetime.
1155             trash_at = trash_at.replace(tzinfo=None) + utcoffset
1156
1157         if trash_at <= datetime.datetime.utcnow():
1158             logger.error("--trash-at argument must be set in the future")
1159             sys.exit(1)
1160     if args.trash_after is not None:
1161         if args.trash_after < 1:
1162             logger.error("--trash-after argument must be >= 1")
1163             sys.exit(1)
1164         trash_at = datetime.timedelta(seconds=(args.trash_after * 24 * 60 * 60))
1165
1166     # Determine the name to use
1167     if args.name:
1168         if args.stream or args.raw:
1169             logger.error("Cannot use --name with --stream or --raw")
1170             sys.exit(1)
1171         elif args.update_collection:
1172             logger.error("Cannot use --name with --update-collection")
1173             sys.exit(1)
1174         collection_name = args.name
1175     else:
1176         collection_name = "Saved at {} by {}@{}".format(
1177             datetime.datetime.utcnow().strftime("%Y-%m-%d %H:%M:%S UTC"),
1178             pwd.getpwuid(os.getuid()).pw_name,
1179             socket.gethostname())
1180
1181     if args.project_uuid and (args.stream or args.raw):
1182         logger.error("Cannot use --project-uuid with --stream or --raw")
1183         sys.exit(1)
1184
1185     # Determine the parent project
1186     try:
1187         project_uuid = desired_project_uuid(api_client, args.project_uuid,
1188                                             args.retries)
1189     except (apiclient_errors.Error, ValueError) as error:
1190         logger.error(error)
1191         sys.exit(1)
1192
1193     if args.progress:
1194         reporter = progress_writer(human_progress)
1195     elif args.batch_progress:
1196         reporter = progress_writer(machine_progress)
1197     else:
1198         reporter = None
1199
1200     #  Split storage-classes argument
1201     storage_classes = None
1202     if args.storage_classes:
1203         storage_classes = args.storage_classes.strip().split(',')
1204         if len(storage_classes) > 1:
1205             logger.error("Multiple storage classes are not supported currently.")
1206             sys.exit(1)
1207
1208
1209     # Setup exclude regex from all the --exclude arguments provided
1210     name_patterns = []
1211     exclude_paths = []
1212     exclude_names = None
1213     if len(args.exclude) > 0:
1214         # We're supporting 2 kinds of exclusion patterns:
1215         # 1)   --exclude '*.jpg'    (file/dir name patterns, will only match
1216         #                            the name, wherever the file is on the tree)
1217         # 2.1) --exclude 'foo/bar'  (file/dir path patterns, will match the
1218         #                            entire path, and should be relative to
1219         #                            any input dir argument)
1220         # 2.2) --exclude './*.jpg'  (Special case for excluding files/dirs
1221         #                            placed directly underneath the input dir)
1222         for p in args.exclude:
1223             # Only relative paths patterns allowed
1224             if p.startswith(os.sep):
1225                 logger.error("Cannot use absolute paths with --exclude")
1226                 sys.exit(1)
1227             if os.path.dirname(p):
1228                 # We don't support of path patterns with '..'
1229                 p_parts = p.split(os.sep)
1230                 if '..' in p_parts:
1231                     logger.error(
1232                         "Cannot use path patterns that include or '..'")
1233                     sys.exit(1)
1234                 # Path search pattern
1235                 exclude_paths.append(p)
1236             else:
1237                 # Name-only search pattern
1238                 name_patterns.append(p)
1239         # For name only matching, we can combine all patterns into a single
1240         # regexp, for better performance.
1241         exclude_names = re.compile('|'.join(
1242             [fnmatch.translate(p) for p in name_patterns]
1243         )) if len(name_patterns) > 0 else None
1244         # Show the user the patterns to be used, just in case they weren't
1245         # specified inside quotes and got changed by the shell expansion.
1246         logger.info("Exclude patterns: {}".format(args.exclude))
1247
1248     # If this is used by a human, and there's at least one directory to be
1249     # uploaded, the expected bytes calculation can take a moment.
1250     if args.progress and any([os.path.isdir(f) for f in args.paths]):
1251         logger.info("Calculating upload size, this could take some time...")
1252     try:
1253         writer = ArvPutUploadJob(paths = args.paths,
1254                                  resume = args.resume,
1255                                  use_cache = args.use_cache,
1256                                  filename = args.filename,
1257                                  reporter = reporter,
1258                                  api_client = api_client,
1259                                  num_retries = args.retries,
1260                                  replication_desired = args.replication,
1261                                  put_threads = args.threads,
1262                                  name = collection_name,
1263                                  owner_uuid = project_uuid,
1264                                  ensure_unique_name = True,
1265                                  update_collection = args.update_collection,
1266                                  storage_classes=storage_classes,
1267                                  logger=logger,
1268                                  dry_run=args.dry_run,
1269                                  follow_links=args.follow_links,
1270                                  exclude_paths=exclude_paths,
1271                                  exclude_names=exclude_names,
1272                                  trash_at=trash_at)
1273     except ResumeCacheConflict:
1274         logger.error("\n".join([
1275             "arv-put: Another process is already uploading this data.",
1276             "         Use --no-cache if this is really what you want."]))
1277         sys.exit(1)
1278     except ResumeCacheInvalidError:
1279         logger.error("\n".join([
1280             "arv-put: Resume cache contains invalid signature: it may have expired",
1281             "         or been created with another Arvados user's credentials.",
1282             "         Switch user or use one of the following options to restart upload:",
1283             "         --no-resume to start a new resume cache.",
1284             "         --no-cache to disable resume cache."]))
1285         sys.exit(1)
1286     except (CollectionUpdateError, PathDoesNotExistError) as error:
1287         logger.error("\n".join([
1288             "arv-put: %s" % str(error)]))
1289         sys.exit(1)
1290     except ArvPutUploadIsPending:
1291         # Dry run check successful, return proper exit code.
1292         sys.exit(2)
1293     except ArvPutUploadNotPending:
1294         # No files pending for upload
1295         sys.exit(0)
1296
1297     if not args.dry_run and not args.update_collection and args.resume and writer.bytes_written > 0:
1298         logger.warning("\n".join([
1299             "arv-put: Resuming previous upload from last checkpoint.",
1300             "         Use the --no-resume option to start over."]))
1301
1302     if not args.dry_run:
1303         writer.report_progress()
1304     output = None
1305     try:
1306         writer.start(save_collection=not(args.stream or args.raw))
1307     except arvados.errors.ApiError as error:
1308         logger.error("\n".join([
1309             "arv-put: %s" % str(error)]))
1310         sys.exit(1)
1311
1312     if args.progress:  # Print newline to split stderr from stdout for humans.
1313         logger.info("\n")
1314
1315     if args.stream:
1316         if args.normalize:
1317             output = writer.manifest_text(normalize=True)
1318         else:
1319             output = writer.manifest_text()
1320     elif args.raw:
1321         output = ','.join(writer.data_locators())
1322     else:
1323         try:
1324             expiration_notice = ""
1325             if writer.collection_trash_at() is not None:
1326                 # Get the local timezone-naive version, and log it with timezone information.
1327                 if time.daylight:
1328                     local_trash_at = writer.collection_trash_at().replace(tzinfo=None) - datetime.timedelta(seconds=time.altzone)
1329                 else:
1330                     local_trash_at = writer.collection_trash_at().replace(tzinfo=None) - datetime.timedelta(seconds=time.timezone)
1331                 expiration_notice = ". It will expire on {} {}.".format(
1332                     local_trash_at.strftime("%Y-%m-%d %H:%M:%S"), time.strftime("%z"))
1333             if args.update_collection:
1334                 logger.info(u"Collection updated: '{}'{}".format(
1335                     writer.collection_name(), expiration_notice))
1336             else:
1337                 logger.info(u"Collection saved as '{}'{}".format(
1338                     writer.collection_name(), expiration_notice))
1339             if args.portable_data_hash:
1340                 output = writer.portable_data_hash()
1341             else:
1342                 output = writer.manifest_locator()
1343         except apiclient_errors.Error as error:
1344             logger.error(
1345                 "arv-put: Error creating Collection on project: {}.".format(
1346                     error))
1347             status = 1
1348
1349     # Print the locator (uuid) of the new collection.
1350     if output is None:
1351         status = status or 1
1352     elif not args.silent:
1353         stdout.write(output)
1354         if not output.endswith('\n'):
1355             stdout.write('\n')
1356
1357     if install_sig_handlers:
1358         arv_cmd.restore_signal_handlers()
1359
1360     if status != 0:
1361         sys.exit(status)
1362
1363     # Success!
1364     return output
1365
1366
1367 if __name__ == '__main__':
1368     main()