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 Do not follow file and directory symlinks.
 177 """)
 178
 179
 180 run_opts = argparse.ArgumentParser(add_help=False)
 181
 182 run_opts.add_argument('--project-uuid', metavar='UUID', help="""
 183 Store the collection in the specified project, instead of your Home
 184 project.
 185 """)
 186
 187 run_opts.add_argument('--name', help="""
 188 Save the collection with the specified name.
 189 """)
 190
 191 _group = run_opts.add_mutually_exclusive_group()
 192 _group.add_argument('--progress', action='store_true',
 193                     help="""
 194 Display human-readable progress on stderr (bytes and, if possible,
 195 percentage of total data size). This is the default behavior when
 196 stderr is a tty.
 197 """)
 198
 199 _group.add_argument('--no-progress', action='store_true',
 200                     help="""
 201 Do not display human-readable progress on stderr, even if stderr is a
 202 tty.
 203 """)
 204
 205 _group.add_argument('--batch-progress', action='store_true',
 206                     help="""
 207 Display machine-readable progress on stderr (bytes and, if known,
 208 total data size).
 209 """)
 210
 211 run_opts.add_argument('--silent', action='store_true',
 212                       help="""
 213 Do not print any debug messages to console. (Any error messages will
 214 still be displayed.)
 215 """)
 216
 217 _group = run_opts.add_mutually_exclusive_group()
 218 _group.add_argument('--resume', action='store_true', default=True,
 219                     help="""
 220 Continue interrupted uploads from cached state (default).
 221 """)
 222 _group.add_argument('--no-resume', action='store_false', dest='resume',
 223                     help="""
 224 Do not continue interrupted uploads from cached state.
 225 """)
 226
 227 _group = run_opts.add_mutually_exclusive_group()
 228 _group.add_argument('--cache', action='store_true', dest='use_cache', default=True,
 229                     help="""
 230 Save upload state in a cache file for resuming (default).
 231 """)
 232 _group.add_argument('--no-cache', action='store_false', dest='use_cache',
 233                     help="""
 234 Do not save upload state in a cache file for resuming.
 235 """)
 236
 237 _group = upload_opts.add_mutually_exclusive_group()
 238 _group.add_argument('--trash-at', metavar='YYYY-MM-DDTHH:MM', default=None,
 239                     help="""
 240 Set the trash date of the resulting collection to an absolute date in the future.
 241 The accepted format is defined by the ISO 8601 standard. Examples: 20090103, 2009-01-03, 20090103T181505, 2009-01-03T18:15:05.\n
 242 Timezone information can be added. If not, the provided date/time is assumed as being in the local system's timezone.
 243 """)
 244 _group.add_argument('--trash-after', type=int, metavar='DAYS', default=None,
 245                     help="""
 246 Set the trash date of the resulting collection to an amount of days from the
 247 date/time that the upload process finishes.
 248 """)
 249
 250 arg_parser = argparse.ArgumentParser(
 251     description='Copy data from the local filesystem to Keep.',
 252     parents=[upload_opts, run_opts, arv_cmd.retry_opt])
 253
 254 def parse_arguments(arguments):
 255     args = arg_parser.parse_args(arguments)
 256
 257     if len(args.paths) == 0:
 258         args.paths = ['-']
 259
 260     args.paths = ["-" if x == "/dev/stdin" else x for x in args.paths]
 261
 262     if len(args.paths) != 1 or os.path.isdir(args.paths[0]):
 263         if args.filename:
 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 os.path.isdir(path):
 529                 # Use absolute paths on cache index so CWD doesn't interfere
 530                 # with the caching logic.
 531                 orig_path = path
 532                 path = os.path.abspath(path)
 533                 if orig_path[-1:] == os.sep:
 534                     # When passing a directory reference with a trailing slash,
 535                     # its contents should be uploaded directly to the
 536                     # collection's root.
 537                     prefixdir = path
 538                 else:
 539                     # When passing a directory reference with no trailing slash,
 540                     # upload the directory to the collection's root.
 541                     prefixdir = os.path.dirname(path)
 542                 prefixdir += os.sep
 543                 for root, dirs, files in os.walk(path,
 544                                                  followlinks=self.follow_links):
 545                     root_relpath = os.path.relpath(root, path)
 546                     if root_relpath == '.':
 547                         root_relpath = ''
 548                     # Exclude files/dirs by full path matching pattern
 549                     if self.exclude_paths:
 550                         dirs[:] = [d for d in dirs
 551                                    if not any(pathname_match(
 552                                            os.path.join(root_relpath, d), pat)
 553                                               for pat in self.exclude_paths)]
 554                         files = [f for f in files
 555                                  if not any(pathname_match(
 556                                          os.path.join(root_relpath, f), pat)
 557                                             for pat in self.exclude_paths)]
 558                     # Exclude files/dirs by name matching pattern
 559                     if self.exclude_names is not None:
 560                         dirs[:] = [d for d in dirs
 561                                    if not self.exclude_names.match(d)]
 562                         files = [f for f in files
 563                                  if not self.exclude_names.match(f)]
 564                     # Make os.walk()'s dir traversing order deterministic
 565                     dirs.sort()
 566                     files.sort()
 567                     for f in files:
 568                         filepath = os.path.join(root, f)
 569                         # Add its size to the total bytes count (if applicable)
 570                         if self.follow_links or (not os.path.islink(filepath)):
 571                             if self.bytes_expected is not None:
 572                                 self.bytes_expected += os.path.getsize(filepath)
 573                         self._check_file(filepath,
 574                                          os.path.join(root[len(prefixdir):], f))
 575             else:
 576                 filepath = os.path.abspath(path)
 577                 # Add its size to the total bytes count (if applicable)
 578                 if self.follow_links or (not os.path.islink(filepath)):
 579                     if self.bytes_expected is not None:
 580                         self.bytes_expected += os.path.getsize(filepath)
 581                 self._check_file(filepath,
 582                                  self.filename or os.path.basename(path))
 583         # If dry-mode is on, and got up to this point, then we should notify that
 584         # there aren't any file to upload.
 585         if self.dry_run:
 586             raise ArvPutUploadNotPending()
 587         # Remove local_collection's files that don't exist locally anymore, so the
 588         # bytes_written count is correct.
 589         for f in self.collection_file_paths(self._local_collection,
 590                                             path_prefix=""):
 591             if f != 'stdin' and f != self.filename and not f in self._file_paths:
 592                 self._local_collection.remove(f)
 593
 594     def start(self, save_collection):
 595         """
 596         Start supporting thread & file uploading
 597         """
 598         self._checkpointer.start()
 599         try:
 600             # Update bytes_written from current local collection and
 601             # report initial progress.
 602             self._update()
 603             # Actual file upload
 604             self._upload_started = True # Used by the update thread to start checkpointing
 605             self._upload_files()
 606         except (SystemExit, Exception) as e:
 607             self._checkpoint_before_quit = False
 608             # Log stack trace only when Ctrl-C isn't pressed (SIGINT)
 609             # Note: We're expecting SystemExit instead of
 610             # KeyboardInterrupt because we have a custom signal
 611             # handler in place that raises SystemExit with the catched
 612             # signal's code.
 613             if isinstance(e, PathDoesNotExistError):
 614                 # We aren't interested in the traceback for this case
 615                 pass
 616             elif not isinstance(e, SystemExit) or e.code != -2:
 617                 self.logger.warning("Abnormal termination:\n{}".format(
 618                     traceback.format_exc()))
 619             raise
 620         finally:
 621             if not self.dry_run:
 622                 # Stop the thread before doing anything else
 623                 self._stop_checkpointer.set()
 624                 self._checkpointer.join()
 625                 if self._checkpoint_before_quit:
 626                     # Commit all pending blocks & one last _update()
 627                     self._local_collection.manifest_text()
 628                     self._update(final=True)
 629                     if save_collection:
 630                         self.save_collection()
 631             if self.use_cache:
 632                 self._cache_file.close()
 633
 634     def _collection_trash_at(self):
 635         """
 636         Returns the trash date that the collection should use at save time.
 637         Takes into account absolute/relative trash_at values requested
 638         by the user.
 639         """
 640         if type(self._trash_at) == datetime.timedelta:
 641             # Get an absolute datetime for trash_at
 642             return datetime.datetime.utcnow() + self._trash_at
 643         return self._trash_at
 644
 645     def save_collection(self):
 646         if self.update:
 647             # Check if files should be updated on the remote collection.
 648             for fp in self._file_paths:
 649                 remote_file = self._remote_collection.find(fp)
 650                 if not remote_file:
 651                     # File don't exist on remote collection, copy it.
 652                     self._remote_collection.copy(fp, fp, self._local_collection)
 653                 elif remote_file != self._local_collection.find(fp):
 654                     # A different file exist on remote collection, overwrite it.
 655                     self._remote_collection.copy(fp, fp, self._local_collection, overwrite=True)
 656                 else:
 657                     # The file already exist on remote collection, skip it.
 658                     pass
 659             self._remote_collection.save(storage_classes=self.storage_classes,
 660                                          num_retries=self.num_retries,
 661                                          trash_at=self._collection_trash_at())
 662         else:
 663             if self.storage_classes is None:
 664                 self.storage_classes = ['default']
 665             self._local_collection.save_new(
 666                 name=self.name, owner_uuid=self.owner_uuid,
 667                 storage_classes=self.storage_classes,
 668                 ensure_unique_name=self.ensure_unique_name,
 669                 num_retries=self.num_retries,
 670                 trash_at=self._collection_trash_at())
 671
 672     def destroy_cache(self):
 673         if self.use_cache:
 674             try:
 675                 os.unlink(self._cache_filename)
 676             except OSError as error:
 677                 # That's what we wanted anyway.
 678                 if error.errno != errno.ENOENT:
 679                     raise
 680             self._cache_file.close()
 681
 682     def _collection_size(self, collection):
 683         """
 684         Recursively get the total size of the collection
 685         """
 686         size = 0
 687         for item in listvalues(collection):
 688             if isinstance(item, arvados.collection.Collection) or isinstance(item, arvados.collection.Subcollection):
 689                 size += self._collection_size(item)
 690             else:
 691                 size += item.size()
 692         return size
 693
 694     def _update_task(self):
 695         """
 696         Periodically called support task. File uploading is
 697         asynchronous so we poll status from the collection.
 698         """
 699         while not self._stop_checkpointer.wait(1 if not self._upload_started else self._update_task_time):
 700             self._update()
 701
 702     def _update(self, final=False):
 703         """
 704         Update cached manifest text and report progress.
 705         """
 706         if self._upload_started:
 707             with self._collection_lock:
 708                 self.bytes_written = self._collection_size(self._local_collection)
 709                 if self.use_cache:
 710                     if final:
 711                         manifest = self._local_collection.manifest_text()
 712                     else:
 713                         # Get the manifest text without comitting pending blocks
 714                         manifest = self._local_collection.manifest_text(strip=False,
 715                                                                         normalize=False,
 716                                                                         only_committed=True)
 717                     # Update cache
 718                     with self._state_lock:
 719                         self._state['manifest'] = manifest
 720             if self.use_cache:
 721                 try:
 722                     self._save_state()
 723                 except Exception as e:
 724                     self.logger.error("Unexpected error trying to save cache file: {}".format(e))
 725             # Keep remote collection's trash_at attribute synced when using relative expire dates
 726             if self._remote_collection is not None and type(self._trash_at) == datetime.timedelta:
 727                 try:
 728                     self._api_client.collections().update(
 729                         uuid=self._remote_collection.manifest_locator(),
 730                         body={'trash_at': self._collection_trash_at().strftime("%Y-%m-%dT%H:%M:%S.%fZ")}
 731                     ).execute(num_retries=self.num_retries)
 732                 except Exception as e:
 733                     self.logger.error("Unexpected error trying to update remote collection's expire date: {}".format(e))
 734         else:
 735             self.bytes_written = self.bytes_skipped
 736         # Call the reporter, if any
 737         self.report_progress()
 738
 739     def report_progress(self):
 740         if self.reporter is not None:
 741             self.reporter(self.bytes_written, self.bytes_expected)
 742
 743     def _write_stdin(self, filename):
 744         output = self._local_collection.open(filename, 'wb')
 745         self._write(sys.stdin.buffer, output)
 746         output.close()
 747
 748     def _check_file(self, source, filename):
 749         """
 750         Check if this file needs to be uploaded
 751         """
 752         # Ignore symlinks when requested
 753         if (not self.follow_links) and os.path.islink(source):
 754             return
 755         resume_offset = 0
 756         should_upload = False
 757         new_file_in_cache = False
 758         # Record file path for updating the remote collection before exiting
 759         self._file_paths.add(filename)
 760
 761         with self._state_lock:
 762             # If no previous cached data on this file, store it for an eventual
 763             # repeated run.
 764             if source not in self._state['files']:
 765                 self._state['files'][source] = {
 766                     'mtime': os.path.getmtime(source),
 767                     'size' : os.path.getsize(source)
 768                 }
 769                 new_file_in_cache = True
 770             cached_file_data = self._state['files'][source]
 771
 772         # Check if file was already uploaded (at least partially)
 773         file_in_local_collection = self._local_collection.find(filename)
 774
 775         # If not resuming, upload the full file.
 776         if not self.resume:
 777             should_upload = True
 778         # New file detected from last run, upload it.
 779         elif new_file_in_cache:
 780             should_upload = True
 781         # Local file didn't change from last run.
 782         elif cached_file_data['mtime'] == os.path.getmtime(source) and cached_file_data['size'] == os.path.getsize(source):
 783             if not file_in_local_collection:
 784                 # File not uploaded yet, upload it completely
 785                 should_upload = True
 786             elif file_in_local_collection.permission_expired():
 787                 # Permission token expired, re-upload file. This will change whenever
 788                 # we have a API for refreshing tokens.
 789                 self.logger.warning(u"Uploaded file '{}' access token expired, will re-upload it from scratch".format(filename))
 790                 should_upload = True
 791                 self._local_collection.remove(filename)
 792             elif cached_file_data['size'] == file_in_local_collection.size():
 793                 # File already there, skip it.
 794                 self.bytes_skipped += cached_file_data['size']
 795             elif cached_file_data['size'] > file_in_local_collection.size():
 796                 # File partially uploaded, resume!
 797                 resume_offset = file_in_local_collection.size()
 798                 self.bytes_skipped += resume_offset
 799                 should_upload = True
 800             else:
 801                 # Inconsistent cache, re-upload the file
 802                 should_upload = True
 803                 self._local_collection.remove(filename)
 804                 self.logger.warning(u"Uploaded version of file '{}' is bigger than local version, will re-upload it from scratch.".format(source))
 805         # Local file differs from cached data, re-upload it.
 806         else:
 807             if file_in_local_collection:
 808                 self._local_collection.remove(filename)
 809             should_upload = True
 810
 811         if should_upload:
 812             try:
 813                 self._files_to_upload.append((source, resume_offset, filename))
 814             except ArvPutUploadIsPending:
 815                 # This could happen when running on dry-mode, close cache file to
 816                 # avoid locking issues.
 817                 self._cache_file.close()
 818                 raise
 819
 820     def _upload_files(self):
 821         for source, resume_offset, filename in self._files_to_upload:
 822             with open(source, 'rb') as source_fd:
 823                 with self._state_lock:
 824                     self._state['files'][source]['mtime'] = os.path.getmtime(source)
 825                     self._state['files'][source]['size'] = os.path.getsize(source)
 826                 if resume_offset > 0:
 827                     # Start upload where we left off
 828                     output = self._local_collection.open(filename, 'ab')
 829                     source_fd.seek(resume_offset)
 830                 else:
 831                     # Start from scratch
 832                     output = self._local_collection.open(filename, 'wb')
 833                 self._write(source_fd, output)
 834                 output.close(flush=False)
 835
 836     def _write(self, source_fd, output):
 837         while True:
 838             data = source_fd.read(arvados.config.KEEP_BLOCK_SIZE)
 839             if not data:
 840                 break
 841             output.write(data)
 842
 843     def _my_collection(self):
 844         return self._remote_collection if self.update else self._local_collection
 845
 846     def _get_cache_filepath(self):
 847         # Set up cache file name from input paths.
 848         md5 = hashlib.md5()
 849         md5.update(arvados.config.get('ARVADOS_API_HOST', '!nohost').encode())
 850         realpaths = sorted(os.path.realpath(path) for path in self.paths)
 851         md5.update(b'\0'.join([p.encode() for p in realpaths]))
 852         if self.filename:
 853             md5.update(self.filename.encode())
 854         cache_filename = md5.hexdigest()
 855         cache_filepath = os.path.join(
 856             arv_cmd.make_home_conf_dir(self.CACHE_DIR, 0o700, 'raise'),
 857             cache_filename)
 858         return cache_filepath
 859
 860     def _setup_state(self, update_collection):
 861         """
 862         Create a new cache file or load a previously existing one.
 863         """
 864         # Load an already existing collection for update
 865         if update_collection and re.match(arvados.util.collection_uuid_pattern,
 866                                           update_collection):
 867             try:
 868                 self._remote_collection = arvados.collection.Collection(
 869                     update_collection,
 870                     api_client=self._api_client,
 871                     num_retries=self.num_retries)
 872             except arvados.errors.ApiError as error:
 873                 raise CollectionUpdateError("Cannot read collection {} ({})".format(update_collection, error))
 874             else:
 875                 self.update = True
 876         elif update_collection:
 877             # Collection locator provided, but unknown format
 878             raise CollectionUpdateError("Collection locator unknown: '{}'".format(update_collection))
 879
 880         if self.use_cache:
 881             cache_filepath = self._get_cache_filepath()
 882             if self.resume and os.path.exists(cache_filepath):
 883                 self.logger.info(u"Resuming upload from cache file {}".format(cache_filepath))
 884                 self._cache_file = open(cache_filepath, 'a+')
 885             else:
 886                 # --no-resume means start with a empty cache file.
 887                 self.logger.info(u"Creating new cache file at {}".format(cache_filepath))
 888                 self._cache_file = open(cache_filepath, 'w+')
 889             self._cache_filename = self._cache_file.name
 890             self._lock_file(self._cache_file)
 891             self._cache_file.seek(0)
 892
 893         with self._state_lock:
 894             if self.use_cache:
 895                 try:
 896                     self._state = json.load(self._cache_file)
 897                     if not set(['manifest', 'files']).issubset(set(self._state.keys())):
 898                         # Cache at least partially incomplete, set up new cache
 899                         self._state = copy.deepcopy(self.EMPTY_STATE)
 900                 except ValueError:
 901                     # Cache file empty, set up new cache
 902                     self._state = copy.deepcopy(self.EMPTY_STATE)
 903             else:
 904                 self.logger.info("No cache usage requested for this run.")
 905                 # No cache file, set empty state
 906                 self._state = copy.deepcopy(self.EMPTY_STATE)
 907             if not self._cached_manifest_valid():
 908                 raise ResumeCacheInvalidError()
 909             # Load the previous manifest so we can check if files were modified remotely.
 910             self._local_collection = arvados.collection.Collection(
 911                 self._state['manifest'],
 912                 replication_desired=self.replication_desired,
 913                 put_threads=self.put_threads,
 914                 api_client=self._api_client,
 915                 num_retries=self.num_retries)
 916
 917     def _cached_manifest_valid(self):
 918         """
 919         Validate the oldest non-expired block signature to check if cached manifest
 920         is usable: checking if the cached manifest was not created with a different
 921         arvados account.
 922         """
 923         if self._state.get('manifest', None) is None:
 924             # No cached manifest yet, all good.
 925             return True
 926         now = datetime.datetime.utcnow()
 927         oldest_exp = None
 928         oldest_loc = None
 929         block_found = False
 930         for m in keep_locator_pattern.finditer(self._state['manifest']):
 931             loc = m.group(0)
 932             try:
 933                 exp = datetime.datetime.utcfromtimestamp(int(loc.split('@')[1], 16))
 934             except IndexError:
 935                 # Locator without signature
 936                 continue
 937             block_found = True
 938             if exp > now and (oldest_exp is None or exp < oldest_exp):
 939                 oldest_exp = exp
 940                 oldest_loc = loc
 941         if not block_found:
 942             # No block signatures found => no invalid block signatures.
 943             return True
 944         if oldest_loc is None:
 945             # Locator signatures found, but all have expired.
 946             # Reset the cache and move on.
 947             self.logger.info('Cache expired, starting from scratch.')
 948             self._state['manifest'] = ''
 949             return True
 950         kc = arvados.KeepClient(api_client=self._api_client,
 951                                 num_retries=self.num_retries)
 952         try:
 953             kc.head(oldest_loc)
 954         except arvados.errors.KeepRequestError:
 955             # Something is wrong, cached manifest is not valid.
 956             return False
 957         return True
 958
 959     def collection_file_paths(self, col, path_prefix='.'):
 960         """Return a list of file paths by recursively go through the entire collection `col`"""
 961         file_paths = []
 962         for name, item in listitems(col):
 963             if isinstance(item, arvados.arvfile.ArvadosFile):
 964                 file_paths.append(os.path.join(path_prefix, name))
 965             elif isinstance(item, arvados.collection.Subcollection):
 966                 new_prefix = os.path.join(path_prefix, name)
 967                 file_paths += self.collection_file_paths(item, path_prefix=new_prefix)
 968         return file_paths
 969
 970     def _lock_file(self, fileobj):
 971         try:
 972             fcntl.flock(fileobj, fcntl.LOCK_EX | fcntl.LOCK_NB)
 973         except IOError:
 974             raise ResumeCacheConflict(u"{} locked".format(fileobj.name))
 975
 976     def _save_state(self):
 977         """
 978         Atomically save current state into cache.
 979         """
 980         with self._state_lock:
 981             # We're not using copy.deepcopy() here because it's a lot slower
 982             # than json.dumps(), and we're already needing JSON format to be
 983             # saved on disk.
 984             state = json.dumps(self._state)
 985         try:
 986             new_cache = tempfile.NamedTemporaryFile(
 987                 mode='w+',
 988                 dir=os.path.dirname(self._cache_filename), delete=False)
 989             self._lock_file(new_cache)
 990             new_cache.write(state)
 991             new_cache.flush()
 992             os.fsync(new_cache)
 993             os.rename(new_cache.name, self._cache_filename)
 994         except (IOError, OSError, ResumeCacheConflict) as error:
 995             self.logger.error("There was a problem while saving the cache file: {}".format(error))
 996             try:
 997                 os.unlink(new_cache_name)
 998             except NameError:  # mkstemp failed.
 999                 pass
1000         else:
1001             self._cache_file.close()
1002             self._cache_file = new_cache
1003
1004     def collection_name(self):
1005         return self._my_collection().api_response()['name'] if self._my_collection().api_response() else None
1006
1007     def collection_trash_at(self):
1008         return self._my_collection().get_trash_at()
1009
1010     def manifest_locator(self):
1011         return self._my_collection().manifest_locator()
1012
1013     def portable_data_hash(self):
1014         pdh = self._my_collection().portable_data_hash()
1015         m = self._my_collection().stripped_manifest().encode()
1016         local_pdh = '{}+{}'.format(hashlib.md5(m).hexdigest(), len(m))
1017         if pdh != local_pdh:
1018             self.logger.warning("\n".join([
1019                 "arv-put: API server provided PDH differs from local manifest.",
1020                 "         This should not happen; showing API server version."]))
1021         return pdh
1022
1023     def manifest_text(self, stream_name=".", strip=False, normalize=False):
1024         return self._my_collection().manifest_text(stream_name, strip, normalize)
1025
1026     def _datablocks_on_item(self, item):
1027         """
1028         Return a list of datablock locators, recursively navigating
1029         through subcollections
1030         """
1031         if isinstance(item, arvados.arvfile.ArvadosFile):
1032             if item.size() == 0:
1033                 # Empty file locator
1034                 return ["d41d8cd98f00b204e9800998ecf8427e+0"]
1035             else:
1036                 locators = []
1037                 for segment in item.segments():
1038                     loc = segment.locator
1039                     locators.append(loc)
1040                 return locators
1041         elif isinstance(item, arvados.collection.Collection):
1042             l = [self._datablocks_on_item(x) for x in listvalues(item)]
1043             # Fast list flattener method taken from:
1044             # http://stackoverflow.com/questions/952914/making-a-flat-list-out-of-list-of-lists-in-python
1045             return [loc for sublist in l for loc in sublist]
1046         else:
1047             return None
1048
1049     def data_locators(self):
1050         with self._collection_lock:
1051             # Make sure all datablocks are flushed before getting the locators
1052             self._my_collection().manifest_text()
1053             datablocks = self._datablocks_on_item(self._my_collection())
1054         return datablocks
1055
1056 _machine_format = "{} {}: {{}} written {{}} total\n".format(sys.argv[0],
1057                                                             os.getpid())
1058
1059 # Simulate glob.glob() matching behavior without the need to scan the filesystem
1060 # Note: fnmatch() doesn't work correctly when used with pathnames. For example the
1061 # pattern 'tests/*.py' will match 'tests/run_test.py' and also 'tests/subdir/run_test.py',
1062 # so instead we're using it on every path component.
1063 def pathname_match(pathname, pattern):
1064     name = pathname.split(os.sep)
1065     # Fix patterns like 'some/subdir/' or 'some//subdir'
1066     pat = [x for x in pattern.split(os.sep) if x != '' and x != '.']
1067     if len(name) != len(pat):
1068         return False
1069     for i in range(len(name)):
1070         if not fnmatch.fnmatch(name[i], pat[i]):
1071             return False
1072     return True
1073
1074 def machine_progress(bytes_written, bytes_expected):
1075     return _machine_format.format(
1076         bytes_written, -1 if (bytes_expected is None) else bytes_expected)
1077
1078 def human_progress(bytes_written, bytes_expected):
1079     if bytes_expected:
1080         return "\r{}M / {}M {:.1%} ".format(
1081             bytes_written >> 20, bytes_expected >> 20,
1082             float(bytes_written) / bytes_expected)
1083     else:
1084         return "\r{} ".format(bytes_written)
1085
1086 def progress_writer(progress_func, outfile=sys.stderr):
1087     def write_progress(bytes_written, bytes_expected):
1088         outfile.write(progress_func(bytes_written, bytes_expected))
1089     return write_progress
1090
1091 def desired_project_uuid(api_client, project_uuid, num_retries):
1092     if not project_uuid:
1093         query = api_client.users().current()
1094     elif arvados.util.user_uuid_pattern.match(project_uuid):
1095         query = api_client.users().get(uuid=project_uuid)
1096     elif arvados.util.group_uuid_pattern.match(project_uuid):
1097         query = api_client.groups().get(uuid=project_uuid)
1098     else:
1099         raise ValueError("Not a valid project UUID: {}".format(project_uuid))
1100     return query.execute(num_retries=num_retries)['uuid']
1101
1102 def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr,
1103          install_sig_handlers=True):
1104     global api_client
1105
1106     args = parse_arguments(arguments)
1107     logger = logging.getLogger('arvados.arv_put')
1108     if args.silent:
1109         logger.setLevel(logging.WARNING)
1110     else:
1111         logger.setLevel(logging.INFO)
1112     status = 0
1113
1114     request_id = arvados.util.new_request_id()
1115
1116     formatter = ArvPutLogFormatter(request_id)
1117     logging.getLogger('arvados').handlers[0].setFormatter(formatter)
1118
1119     if api_client is None:
1120         api_client = arvados.api('v1', request_id=request_id)
1121
1122     if install_sig_handlers:
1123         arv_cmd.install_signal_handlers()
1124
1125     # Trash arguments validation
1126     trash_at = None
1127     if args.trash_at is not None:
1128         # ciso8601 considers YYYYMM as invalid but YYYY-MM as valid, so here we
1129         # make sure the user provides a complete YYYY-MM-DD date.
1130         if not re.match(r'^\d{4}(?P<dash>-?)\d{2}?(?P=dash)\d{2}', args.trash_at):
1131             logger.error("--trash-at argument format invalid, use --help to see examples.")
1132             sys.exit(1)
1133         # Check if no time information was provided. In that case, assume end-of-day.
1134         if re.match(r'^\d{4}(?P<dash>-?)\d{2}?(?P=dash)\d{2}$', args.trash_at):
1135             args.trash_at += 'T23:59:59'
1136         try:
1137             trash_at = ciso8601.parse_datetime(args.trash_at)
1138         except:
1139             logger.error("--trash-at argument format invalid, use --help to see examples.")
1140             sys.exit(1)
1141         else:
1142             if trash_at.tzinfo is not None:
1143                 # Timezone aware datetime provided.
1144                 utcoffset = -trash_at.utcoffset()
1145             else:
1146                 # Timezone naive datetime provided. Assume is local.
1147                 if time.daylight:
1148                     utcoffset = datetime.timedelta(seconds=time.altzone)
1149                 else:
1150                     utcoffset = datetime.timedelta(seconds=time.timezone)
1151             # Convert to UTC timezone naive datetime.
1152             trash_at = trash_at.replace(tzinfo=None) + utcoffset
1153
1154         if trash_at <= datetime.datetime.utcnow():
1155             logger.error("--trash-at argument must be set in the future")
1156             sys.exit(1)
1157     if args.trash_after is not None:
1158         if args.trash_after < 1:
1159             logger.error("--trash-after argument must be >= 1")
1160             sys.exit(1)
1161         trash_at = datetime.timedelta(seconds=(args.trash_after * 24 * 60 * 60))
1162
1163     # Determine the name to use
1164     if args.name:
1165         if args.stream or args.raw:
1166             logger.error("Cannot use --name with --stream or --raw")
1167             sys.exit(1)
1168         elif args.update_collection:
1169             logger.error("Cannot use --name with --update-collection")
1170             sys.exit(1)
1171         collection_name = args.name
1172     else:
1173         collection_name = "Saved at {} by {}@{}".format(
1174             datetime.datetime.utcnow().strftime("%Y-%m-%d %H:%M:%S UTC"),
1175             pwd.getpwuid(os.getuid()).pw_name,
1176             socket.gethostname())
1177
1178     if args.project_uuid and (args.stream or args.raw):
1179         logger.error("Cannot use --project-uuid with --stream or --raw")
1180         sys.exit(1)
1181
1182     # Determine the parent project
1183     try:
1184         project_uuid = desired_project_uuid(api_client, args.project_uuid,
1185                                             args.retries)
1186     except (apiclient_errors.Error, ValueError) as error:
1187         logger.error(error)
1188         sys.exit(1)
1189
1190     if args.progress:
1191         reporter = progress_writer(human_progress)
1192     elif args.batch_progress:
1193         reporter = progress_writer(machine_progress)
1194     else:
1195         reporter = None
1196
1197     #  Split storage-classes argument
1198     storage_classes = None
1199     if args.storage_classes:
1200         storage_classes = args.storage_classes.strip().split(',')
1201         if len(storage_classes) > 1:
1202             logger.error("Multiple storage classes are not supported currently.")
1203             sys.exit(1)
1204
1205
1206     # Setup exclude regex from all the --exclude arguments provided
1207     name_patterns = []
1208     exclude_paths = []
1209     exclude_names = None
1210     if len(args.exclude) > 0:
1211         # We're supporting 2 kinds of exclusion patterns:
1212         # 1)   --exclude '*.jpg'    (file/dir name patterns, will only match
1213         #                            the name, wherever the file is on the tree)
1214         # 2.1) --exclude 'foo/bar'  (file/dir path patterns, will match the
1215         #                            entire path, and should be relative to
1216         #                            any input dir argument)
1217         # 2.2) --exclude './*.jpg'  (Special case for excluding files/dirs
1218         #                            placed directly underneath the input dir)
1219         for p in args.exclude:
1220             # Only relative paths patterns allowed
1221             if p.startswith(os.sep):
1222                 logger.error("Cannot use absolute paths with --exclude")
1223                 sys.exit(1)
1224             if os.path.dirname(p):
1225                 # We don't support of path patterns with '..'
1226                 p_parts = p.split(os.sep)
1227                 if '..' in p_parts:
1228                     logger.error(
1229                         "Cannot use path patterns that include or '..'")
1230                     sys.exit(1)
1231                 # Path search pattern
1232                 exclude_paths.append(p)
1233             else:
1234                 # Name-only search pattern
1235                 name_patterns.append(p)
1236         # For name only matching, we can combine all patterns into a single
1237         # regexp, for better performance.
1238         exclude_names = re.compile('|'.join(
1239             [fnmatch.translate(p) for p in name_patterns]
1240         )) if len(name_patterns) > 0 else None
1241         # Show the user the patterns to be used, just in case they weren't
1242         # specified inside quotes and got changed by the shell expansion.
1243         logger.info("Exclude patterns: {}".format(args.exclude))
1244
1245     # If this is used by a human, and there's at least one directory to be
1246     # uploaded, the expected bytes calculation can take a moment.
1247     if args.progress and any([os.path.isdir(f) for f in args.paths]):
1248         logger.info("Calculating upload size, this could take some time...")
1249     try:
1250         writer = ArvPutUploadJob(paths = args.paths,
1251                                  resume = args.resume,
1252                                  use_cache = args.use_cache,
1253                                  filename = args.filename,
1254                                  reporter = reporter,
1255                                  api_client = api_client,
1256                                  num_retries = args.retries,
1257                                  replication_desired = args.replication,
1258                                  put_threads = args.threads,
1259                                  name = collection_name,
1260                                  owner_uuid = project_uuid,
1261                                  ensure_unique_name = True,
1262                                  update_collection = args.update_collection,
1263                                  storage_classes=storage_classes,
1264                                  logger=logger,
1265                                  dry_run=args.dry_run,
1266                                  follow_links=args.follow_links,
1267                                  exclude_paths=exclude_paths,
1268                                  exclude_names=exclude_names,
1269                                  trash_at=trash_at)
1270     except ResumeCacheConflict:
1271         logger.error("\n".join([
1272             "arv-put: Another process is already uploading this data.",
1273             "         Use --no-cache if this is really what you want."]))
1274         sys.exit(1)
1275     except ResumeCacheInvalidError:
1276         logger.error("\n".join([
1277             "arv-put: Resume cache contains invalid signature: it may have expired",
1278             "         or been created with another Arvados user's credentials.",
1279             "         Switch user or use one of the following options to restart upload:",
1280             "         --no-resume to start a new resume cache.",
1281             "         --no-cache to disable resume cache."]))
1282         sys.exit(1)
1283     except (CollectionUpdateError, PathDoesNotExistError) as error:
1284         logger.error("\n".join([
1285             "arv-put: %s" % str(error)]))
1286         sys.exit(1)
1287     except ArvPutUploadIsPending:
1288         # Dry run check successful, return proper exit code.
1289         sys.exit(2)
1290     except ArvPutUploadNotPending:
1291         # No files pending for upload
1292         sys.exit(0)
1293
1294     if not args.dry_run and not args.update_collection and args.resume and writer.bytes_written > 0:
1295         logger.warning("\n".join([
1296             "arv-put: Resuming previous upload from last checkpoint.",
1297             "         Use the --no-resume option to start over."]))
1298
1299     if not args.dry_run:
1300         writer.report_progress()
1301     output = None
1302     try:
1303         writer.start(save_collection=not(args.stream or args.raw))
1304     except arvados.errors.ApiError as error:
1305         logger.error("\n".join([
1306             "arv-put: %s" % str(error)]))
1307         sys.exit(1)
1308
1309     if args.progress:  # Print newline to split stderr from stdout for humans.
1310         logger.info("\n")
1311
1312     if args.stream:
1313         if args.normalize:
1314             output = writer.manifest_text(normalize=True)
1315         else:
1316             output = writer.manifest_text()
1317     elif args.raw:
1318         output = ','.join(writer.data_locators())
1319     else:
1320         try:
1321             expiration_notice = ""
1322             if writer.collection_trash_at() is not None:
1323                 # Get the local timezone-naive version, and log it with timezone information.
1324                 if time.daylight:
1325                     local_trash_at = writer.collection_trash_at().replace(tzinfo=None) - datetime.timedelta(seconds=time.altzone)
1326                 else:
1327                     local_trash_at = writer.collection_trash_at().replace(tzinfo=None) - datetime.timedelta(seconds=time.timezone)
1328                 expiration_notice = ". It will expire on {} {}.".format(
1329                     local_trash_at.strftime("%Y-%m-%d %H:%M:%S"), time.strftime("%z"))
1330             if args.update_collection:
1331                 logger.info(u"Collection updated: '{}'{}".format(
1332                     writer.collection_name(), expiration_notice))
1333             else:
1334                 logger.info(u"Collection saved as '{}'{}".format(
1335                     writer.collection_name(), expiration_notice))
1336             if args.portable_data_hash:
1337                 output = writer.portable_data_hash()
1338             else:
1339                 output = writer.manifest_locator()
1340         except apiclient_errors.Error as error:
1341             logger.error(
1342                 "arv-put: Error creating Collection on project: {}.".format(
1343                     error))
1344             status = 1
1345
1346     # Print the locator (uuid) of the new collection.
1347     if output is None:
1348         status = status or 1
1349     elif not args.silent:
1350         stdout.write(output)
1351         if not output.endswith('\n'):
1352             stdout.write('\n')
1353
1354     if install_sig_handlers:
1355         arv_cmd.restore_signal_handlers()
1356
1357     if status != 0:
1358         sys.exit(status)
1359
1360     # Success!
1361     return output
1362
1363
1364 if __name__ == '__main__':
1365     main()