arvados.collection

   1# Copyright (C) The Arvados Authors. All rights reserved.
   2#
   3# SPDX-License-Identifier: Apache-2.0
   4"""Tools to work with Arvados collections
   5
   6This module provides high-level interfaces to create, read, and update
   7Arvados collections. Most users will want to instantiate `Collection`
   8objects, and use methods like `Collection.open` and `Collection.mkdirs` to
   9read and write data in the collection. Refer to the Arvados Python SDK
  10cookbook for [an introduction to using the Collection class][cookbook].
  11
  12[cookbook]: https://doc.arvados.org/sdk/python/cookbook.html#working-with-collections
  13"""
  14
  15from __future__ import absolute_import
  16from future.utils import listitems, listvalues, viewkeys
  17from builtins import str
  18from past.builtins import basestring
  19from builtins import object
  20import ciso8601
  21import datetime
  22import errno
  23import functools
  24import hashlib
  25import io
  26import logging
  27import os
  28import re
  29import sys
  30import threading
  31import time
  32
  33from collections import deque
  34from stat import *
  35
  36from .arvfile import split, _FileLikeObjectBase, ArvadosFile, ArvadosFileWriter, ArvadosFileReader, WrappableFile, _BlockManager, synchronized, must_be_writable, NoopLock
  37from .keep import KeepLocator, KeepClient
  38from .stream import StreamReader
  39from ._normalize_stream import normalize_stream, escape
  40from ._ranges import Range, LocatorAndRange
  41from .safeapi import ThreadSafeApiCache
  42import arvados.config as config
  43import arvados.errors as errors
  44import arvados.util
  45import arvados.events as events
  46from arvados.retry import retry_method
  47
  48from typing import (
  49    Any,
  50    Callable,
  51    Dict,
  52    IO,
  53    Iterator,
  54    List,
  55    Mapping,
  56    Optional,
  57    Tuple,
  58    Union,
  59)
  60
  61if sys.version_info < (3, 8):
  62    from typing_extensions import Literal
  63else:
  64    from typing import Literal
  65
  66_logger = logging.getLogger('arvados.collection')
  67
  68ADD = "add"
  69"""Argument value for `Collection` methods to represent an added item"""
  70DEL = "del"
  71"""Argument value for `Collection` methods to represent a removed item"""
  72MOD = "mod"
  73"""Argument value for `Collection` methods to represent a modified item"""
  74TOK = "tok"
  75"""Argument value for `Collection` methods to represent an item with token differences"""
  76FILE = "file"
  77"""`create_type` value for `Collection.find_or_create`"""
  78COLLECTION = "collection"
  79"""`create_type` value for `Collection.find_or_create`"""
  80
  81ChangeList = List[Union[
  82    Tuple[Literal[ADD, DEL], str, 'Collection'],
  83    Tuple[Literal[MOD, TOK], str, 'Collection', 'Collection'],
  84]]
  85ChangeType = Literal[ADD, DEL, MOD, TOK]
  86CollectionItem = Union[ArvadosFile, 'Collection']
  87ChangeCallback = Callable[[ChangeType, 'Collection', str, CollectionItem], object]
  88CreateType = Literal[COLLECTION, FILE]
  89Properties = Dict[str, Any]
  90StorageClasses = List[str]
  91
  92class CollectionBase(object):
  93    """Abstract base class for Collection classes
  94
  95    .. ATTENTION:: Internal
  96       This class is meant to be used by other parts of the SDK. User code
  97       should instantiate or subclass `Collection` or one of its subclasses
  98       directly.
  99    """
 100
 101    def __enter__(self):
 102        """Enter a context block with this collection instance"""
 103        return self
 104
 105    def __exit__(self, exc_type, exc_value, traceback):
 106        """Exit a context block with this collection instance"""
 107        pass
 108
 109    def _my_keep(self):
 110        if self._keep_client is None:
 111            self._keep_client = KeepClient(api_client=self._api_client,
 112                                           num_retries=self.num_retries)
 113        return self._keep_client
 114
 115    def stripped_manifest(self) -> str:
 116        """Create a copy of the collection manifest with only size hints
 117
 118        This method returns a string with the current collection's manifest
 119        text with all non-portable locator hints like permission hints and
 120        remote cluster hints removed. The only hints in the returned manifest
 121        will be size hints.
 122        """
 123        raw = self.manifest_text()
 124        clean = []
 125        for line in raw.split("\n"):
 126            fields = line.split()
 127            if fields:
 128                clean_fields = fields[:1] + [
 129                    (re.sub(r'\+[^\d][^\+]*', '', x)
 130                     if re.match(arvados.util.keep_locator_pattern, x)
 131                     else x)
 132                    for x in fields[1:]]
 133                clean += [' '.join(clean_fields), "\n"]
 134        return ''.join(clean)
 135
 136
 137class _WriterFile(_FileLikeObjectBase):
 138    def __init__(self, coll_writer, name):
 139        super(_WriterFile, self).__init__(name, 'wb')
 140        self.dest = coll_writer
 141
 142    def close(self):
 143        super(_WriterFile, self).close()
 144        self.dest.finish_current_file()
 145
 146    @_FileLikeObjectBase._before_close
 147    def write(self, data):
 148        self.dest.write(data)
 149
 150    @_FileLikeObjectBase._before_close
 151    def writelines(self, seq):
 152        for data in seq:
 153            self.write(data)
 154
 155    @_FileLikeObjectBase._before_close
 156    def flush(self):
 157        self.dest.flush_data()
 158
 159
 160class RichCollectionBase(CollectionBase):
 161    """Base class for Collection classes
 162
 163    .. ATTENTION:: Internal
 164       This class is meant to be used by other parts of the SDK. User code
 165       should instantiate or subclass `Collection` or one of its subclasses
 166       directly.
 167    """
 168
 169    def __init__(self, parent=None):
 170        self.parent = parent
 171        self._committed = False
 172        self._has_remote_blocks = False
 173        self._callback = None
 174        self._items = {}
 175
 176    def _my_api(self):
 177        raise NotImplementedError()
 178
 179    def _my_keep(self):
 180        raise NotImplementedError()
 181
 182    def _my_block_manager(self):
 183        raise NotImplementedError()
 184
 185    def writable(self) -> bool:
 186        """Indicate whether this collection object can be modified
 187
 188        This method returns `False` if this object is a `CollectionReader`,
 189        else `True`.
 190        """
 191        raise NotImplementedError()
 192
 193    def root_collection(self) -> 'Collection':
 194        """Get this collection's root collection object
 195
 196        If you open a subcollection with `Collection.find`, calling this method
 197        on that subcollection returns the source Collection object.
 198        """
 199        raise NotImplementedError()
 200
 201    def stream_name(self) -> str:
 202        """Get the name of the manifest stream represented by this collection
 203
 204        If you open a subcollection with `Collection.find`, calling this method
 205        on that subcollection returns the name of the stream you opened.
 206        """
 207        raise NotImplementedError()
 208
 209    @synchronized
 210    def has_remote_blocks(self) -> bool:
 211        """Indiciate whether the collection refers to remote data
 212
 213        Returns `True` if the collection manifest includes any Keep locators
 214        with a remote hint (`+R`), else `False`.
 215        """
 216        if self._has_remote_blocks:
 217            return True
 218        for item in self:
 219            if self[item].has_remote_blocks():
 220                return True
 221        return False
 222
 223    @synchronized
 224    def set_has_remote_blocks(self, val: bool) -> None:
 225        """Cache whether this collection refers to remote blocks
 226
 227        .. ATTENTION:: Internal
 228           This method is only meant to be used by other Collection methods.
 229
 230        Set this collection's cached "has remote blocks" flag to the given
 231        value.
 232        """
 233        self._has_remote_blocks = val
 234        if self.parent:
 235            self.parent.set_has_remote_blocks(val)
 236
 237    @must_be_writable
 238    @synchronized
 239    def find_or_create(
 240            self,
 241            path: str,
 242            create_type: CreateType,
 243    ) -> CollectionItem:
 244        """Get the item at the given path, creating it if necessary
 245
 246        If `path` refers to a stream in this collection, returns a
 247        corresponding `Subcollection` object. If `path` refers to a file in
 248        this collection, returns a corresponding
 249        `arvados.arvfile.ArvadosFile` object. If `path` does not exist in
 250        this collection, then this method creates a new object and returns
 251        it, creating parent streams as needed. The type of object created is
 252        determined by the value of `create_type`.
 253
 254        Arguments:
 255
 256        * path: str --- The path to find or create within this collection.
 257
 258        * create_type: Literal[COLLECTION, FILE] --- The type of object to
 259          create at `path` if one does not exist. Passing `COLLECTION`
 260          creates a stream and returns the corresponding
 261          `Subcollection`. Passing `FILE` creates a new file and returns the
 262          corresponding `arvados.arvfile.ArvadosFile`.
 263        """
 264        pathcomponents = path.split("/", 1)
 265        if pathcomponents[0]:
 266            item = self._items.get(pathcomponents[0])
 267            if len(pathcomponents) == 1:
 268                if item is None:
 269                    # create new file
 270                    if create_type == COLLECTION:
 271                        item = Subcollection(self, pathcomponents[0])
 272                    else:
 273                        item = ArvadosFile(self, pathcomponents[0])
 274                    self._items[pathcomponents[0]] = item
 275                    self.set_committed(False)
 276                    self.notify(ADD, self, pathcomponents[0], item)
 277                return item
 278            else:
 279                if item is None:
 280                    # create new collection
 281                    item = Subcollection(self, pathcomponents[0])
 282                    self._items[pathcomponents[0]] = item
 283                    self.set_committed(False)
 284                    self.notify(ADD, self, pathcomponents[0], item)
 285                if isinstance(item, RichCollectionBase):
 286                    return item.find_or_create(pathcomponents[1], create_type)
 287                else:
 288                    raise IOError(errno.ENOTDIR, "Not a directory", pathcomponents[0])
 289        else:
 290            return self
 291
 292    @synchronized
 293    def find(self, path: str) -> CollectionItem:
 294        """Get the item at the given path
 295
 296        If `path` refers to a stream in this collection, returns a
 297        corresponding `Subcollection` object. If `path` refers to a file in
 298        this collection, returns a corresponding
 299        `arvados.arvfile.ArvadosFile` object. If `path` does not exist in
 300        this collection, then this method raises `NotADirectoryError`.
 301
 302        Arguments:
 303
 304        * path: str --- The path to find or create within this collection.
 305        """
 306        if not path:
 307            raise errors.ArgumentError("Parameter 'path' is empty.")
 308
 309        pathcomponents = path.split("/", 1)
 310        if pathcomponents[0] == '':
 311            raise IOError(errno.ENOTDIR, "Not a directory", pathcomponents[0])
 312
 313        item = self._items.get(pathcomponents[0])
 314        if item is None:
 315            return None
 316        elif len(pathcomponents) == 1:
 317            return item
 318        else:
 319            if isinstance(item, RichCollectionBase):
 320                if pathcomponents[1]:
 321                    return item.find(pathcomponents[1])
 322                else:
 323                    return item
 324            else:
 325                raise IOError(errno.ENOTDIR, "Not a directory", pathcomponents[0])
 326
 327    @synchronized
 328    def mkdirs(self, path: str) -> 'Subcollection':
 329        """Create and return a subcollection at `path`
 330
 331        If `path` exists within this collection, raises `FileExistsError`.
 332        Otherwise, creates a stream at that path and returns the
 333        corresponding `Subcollection`.
 334        """
 335        if self.find(path) != None:
 336            raise IOError(errno.EEXIST, "Directory or file exists", path)
 337
 338        return self.find_or_create(path, COLLECTION)
 339
 340    def open(
 341            self,
 342            path: str,
 343            mode: str="r",
 344            encoding: Optional[str]=None,
 345    ) -> IO:
 346        """Open a file-like object within the collection
 347
 348        This method returns a file-like object that can read and/or write the
 349        file located at `path` within the collection. If you attempt to write
 350        a `path` that does not exist, the file is created with `find_or_create`.
 351        If the file cannot be opened for any other reason, this method raises
 352        `OSError` with an appropriate errno.
 353
 354        Arguments:
 355
 356        * path: str --- The path of the file to open within this collection
 357
 358        * mode: str --- The mode to open this file. Supports all the same
 359          values as `builtins.open`.
 360
 361        * encoding: str | None --- The text encoding of the file. Only used
 362          when the file is opened in text mode. The default is
 363          platform-dependent.
 364        """
 365        if not re.search(r'^[rwa][bt]?\+?$', mode):
 366            raise errors.ArgumentError("Invalid mode {!r}".format(mode))
 367
 368        if mode[0] == 'r' and '+' not in mode:
 369            fclass = ArvadosFileReader
 370            arvfile = self.find(path)
 371        elif not self.writable():
 372            raise IOError(errno.EROFS, "Collection is read only")
 373        else:
 374            fclass = ArvadosFileWriter
 375            arvfile = self.find_or_create(path, FILE)
 376
 377        if arvfile is None:
 378            raise IOError(errno.ENOENT, "File not found", path)
 379        if not isinstance(arvfile, ArvadosFile):
 380            raise IOError(errno.EISDIR, "Is a directory", path)
 381
 382        if mode[0] == 'w':
 383            arvfile.truncate(0)
 384
 385        binmode = mode[0] + 'b' + re.sub('[bt]', '', mode[1:])
 386        f = fclass(arvfile, mode=binmode, num_retries=self.num_retries)
 387        if 'b' not in mode:
 388            bufferclass = io.BufferedRandom if f.writable() else io.BufferedReader
 389            f = io.TextIOWrapper(bufferclass(WrappableFile(f)), encoding=encoding)
 390        return f
 391
 392    def modified(self) -> bool:
 393        """Indicate whether this collection has an API server record
 394
 395        Returns `False` if this collection corresponds to a record loaded from
 396        the API server, `True` otherwise.
 397        """
 398        return not self.committed()
 399
 400    @synchronized
 401    def committed(self):
 402        """Indicate whether this collection has an API server record
 403
 404        Returns `True` if this collection corresponds to a record loaded from
 405        the API server, `False` otherwise.
 406        """
 407        return self._committed
 408
 409    @synchronized
 410    def set_committed(self, value: bool=True):
 411        """Cache whether this collection has an API server record
 412
 413        .. ATTENTION:: Internal
 414           This method is only meant to be used by other Collection methods.
 415
 416        Set this collection's cached "committed" flag to the given
 417        value and propagates it as needed.
 418        """
 419        if value == self._committed:
 420            return
 421        if value:
 422            for k,v in listitems(self._items):
 423                v.set_committed(True)
 424            self._committed = True
 425        else:
 426            self._committed = False
 427            if self.parent is not None:
 428                self.parent.set_committed(False)
 429
 430    @synchronized
 431    def __iter__(self) -> Iterator[str]:
 432        """Iterate names of streams and files in this collection
 433
 434        This method does not recurse. It only iterates the contents of this
 435        collection's corresponding stream.
 436        """
 437        return iter(viewkeys(self._items))
 438
 439    @synchronized
 440    def __getitem__(self, k: str) -> CollectionItem:
 441        """Get a `arvados.arvfile.ArvadosFile` or `Subcollection` in this collection
 442
 443        This method does not recurse. If you want to search a path, use
 444        `RichCollectionBase.find` instead.
 445        """
 446        return self._items[k]
 447
 448    @synchronized
 449    def __contains__(self, k: str) -> bool:
 450        """Indicate whether this collection has an item with this name
 451
 452        This method does not recurse. It you want to check a path, use
 453        `RichCollectionBase.exists` instead.
 454        """
 455        return k in self._items
 456
 457    @synchronized
 458    def __len__(self):
 459        """Get the number of items directly contained in this collection
 460
 461        This method does not recurse. It only counts the streams and files
 462        in this collection's corresponding stream.
 463        """
 464        return len(self._items)
 465
 466    @must_be_writable
 467    @synchronized
 468    def __delitem__(self, p: str) -> None:
 469        """Delete an item from this collection's stream
 470
 471        This method does not recurse. If you want to remove an item by a
 472        path, use `RichCollectionBase.remove` instead.
 473        """
 474        del self._items[p]
 475        self.set_committed(False)
 476        self.notify(DEL, self, p, None)
 477
 478    @synchronized
 479    def keys(self) -> Iterator[str]:
 480        """Iterate names of streams and files in this collection
 481
 482        This method does not recurse. It only iterates the contents of this
 483        collection's corresponding stream.
 484        """
 485        return self._items.keys()
 486
 487    @synchronized
 488    def values(self) -> List[CollectionItem]:
 489        """Get a list of objects in this collection's stream
 490
 491        The return value includes a `Subcollection` for every stream, and an
 492        `arvados.arvfile.ArvadosFile` for every file, directly within this
 493        collection's stream.  This method does not recurse.
 494        """
 495        return listvalues(self._items)
 496
 497    @synchronized
 498    def items(self) -> List[Tuple[str, CollectionItem]]:
 499        """Get a list of `(name, object)` tuples from this collection's stream
 500
 501        The return value includes a `Subcollection` for every stream, and an
 502        `arvados.arvfile.ArvadosFile` for every file, directly within this
 503        collection's stream.  This method does not recurse.
 504        """
 505        return listitems(self._items)
 506
 507    def exists(self, path: str) -> bool:
 508        """Indicate whether this collection includes an item at `path`
 509
 510        This method returns `True` if `path` refers to a stream or file within
 511        this collection, else `False`.
 512
 513        Arguments:
 514
 515        * path: str --- The path to check for existence within this collection
 516        """
 517        return self.find(path) is not None
 518
 519    @must_be_writable
 520    @synchronized
 521    def remove(self, path: str, recursive: bool=False) -> None:
 522        """Remove the file or stream at `path`
 523
 524        Arguments:
 525
 526        * path: str --- The path of the item to remove from the collection
 527
 528        * recursive: bool --- Controls the method's behavior if `path` refers
 529          to a nonempty stream. If `False` (the default), this method raises
 530          `OSError` with errno `ENOTEMPTY`. If `True`, this method removes all
 531          items under the stream.
 532        """
 533        if not path:
 534            raise errors.ArgumentError("Parameter 'path' is empty.")
 535
 536        pathcomponents = path.split("/", 1)
 537        item = self._items.get(pathcomponents[0])
 538        if item is None:
 539            raise IOError(errno.ENOENT, "File not found", path)
 540        if len(pathcomponents) == 1:
 541            if isinstance(self._items[pathcomponents[0]], RichCollectionBase) and len(self._items[pathcomponents[0]]) > 0 and not recursive:
 542                raise IOError(errno.ENOTEMPTY, "Directory not empty", path)
 543            deleteditem = self._items[pathcomponents[0]]
 544            del self._items[pathcomponents[0]]
 545            self.set_committed(False)
 546            self.notify(DEL, self, pathcomponents[0], deleteditem)
 547        else:
 548            item.remove(pathcomponents[1], recursive=recursive)
 549
 550    def _clonefrom(self, source):
 551        for k,v in listitems(source):
 552            self._items[k] = v.clone(self, k)
 553
 554    def clone(self):
 555        raise NotImplementedError()
 556
 557    @must_be_writable
 558    @synchronized
 559    def add(
 560            self,
 561            source_obj: CollectionItem,
 562            target_name: str,
 563            overwrite: bool=False,
 564            reparent: bool=False,
 565    ) -> None:
 566        """Copy or move a file or subcollection object to this collection
 567
 568        Arguments:
 569
 570        * source_obj: arvados.arvfile.ArvadosFile | Subcollection --- The file or subcollection
 571          to add to this collection
 572
 573        * target_name: str --- The path inside this collection where
 574          `source_obj` should be added.
 575
 576        * overwrite: bool --- Controls the behavior of this method when the
 577          collection already contains an object at `target_name`. If `False`
 578          (the default), this method will raise `FileExistsError`. If `True`,
 579          the object at `target_name` will be replaced with `source_obj`.
 580
 581        * reparent: bool --- Controls whether this method copies or moves
 582          `source_obj`. If `False` (the default), `source_obj` is copied into
 583          this collection. If `True`, `source_obj` is moved into this
 584          collection.
 585        """
 586        if target_name in self and not overwrite:
 587            raise IOError(errno.EEXIST, "File already exists", target_name)
 588
 589        modified_from = None
 590        if target_name in self:
 591            modified_from = self[target_name]
 592
 593        # Actually make the move or copy.
 594        if reparent:
 595            source_obj._reparent(self, target_name)
 596            item = source_obj
 597        else:
 598            item = source_obj.clone(self, target_name)
 599
 600        self._items[target_name] = item
 601        self.set_committed(False)
 602        if not self._has_remote_blocks and source_obj.has_remote_blocks():
 603            self.set_has_remote_blocks(True)
 604
 605        if modified_from:
 606            self.notify(MOD, self, target_name, (modified_from, item))
 607        else:
 608            self.notify(ADD, self, target_name, item)
 609
 610    def _get_src_target(self, source, target_path, source_collection, create_dest):
 611        if source_collection is None:
 612            source_collection = self
 613
 614        # Find the object
 615        if isinstance(source, basestring):
 616            source_obj = source_collection.find(source)
 617            if source_obj is None:
 618                raise IOError(errno.ENOENT, "File not found", source)
 619            sourcecomponents = source.split("/")
 620        else:
 621            source_obj = source
 622            sourcecomponents = None
 623
 624        # Find parent collection the target path
 625        targetcomponents = target_path.split("/")
 626
 627        # Determine the name to use.
 628        target_name = targetcomponents[-1] if targetcomponents[-1] else sourcecomponents[-1]
 629
 630        if not target_name:
 631            raise errors.ArgumentError("Target path is empty and source is an object.  Cannot determine destination filename to use.")
 632
 633        if create_dest:
 634            target_dir = self.find_or_create("/".join(targetcomponents[0:-1]), COLLECTION)
 635        else:
 636            if len(targetcomponents) > 1:
 637                target_dir = self.find("/".join(targetcomponents[0:-1]))
 638            else:
 639                target_dir = self
 640
 641        if target_dir is None:
 642            raise IOError(errno.ENOENT, "Target directory not found", target_name)
 643
 644        if target_name in target_dir and isinstance(target_dir[target_name], RichCollectionBase) and sourcecomponents:
 645            target_dir = target_dir[target_name]
 646            target_name = sourcecomponents[-1]
 647
 648        return (source_obj, target_dir, target_name)
 649
 650    @must_be_writable
 651    @synchronized
 652    def copy(
 653            self,
 654            source: Union[str, CollectionItem],
 655            target_path: str,
 656            source_collection: Optional['RichCollectionBase']=None,
 657            overwrite: bool=False,
 658    ) -> None:
 659        """Copy a file or subcollection object to this collection
 660
 661        Arguments:
 662
 663        * source: str | arvados.arvfile.ArvadosFile |
 664          arvados.collection.Subcollection --- The file or subcollection to
 665          add to this collection. If `source` is a str, the object will be
 666          found by looking up this path from `source_collection` (see
 667          below).
 668
 669        * target_path: str --- The path inside this collection where the
 670          source object should be added.
 671
 672        * source_collection: arvados.collection.Collection | None --- The
 673          collection to find the source object from when `source` is a
 674          path. Defaults to the current collection (`self`).
 675
 676        * overwrite: bool --- Controls the behavior of this method when the
 677          collection already contains an object at `target_path`. If `False`
 678          (the default), this method will raise `FileExistsError`. If `True`,
 679          the object at `target_path` will be replaced with `source_obj`.
 680        """
 681        source_obj, target_dir, target_name = self._get_src_target(source, target_path, source_collection, True)
 682        target_dir.add(source_obj, target_name, overwrite, False)
 683
 684    @must_be_writable
 685    @synchronized
 686    def rename(
 687            self,
 688            source: Union[str, CollectionItem],
 689            target_path: str,
 690            source_collection: Optional['RichCollectionBase']=None,
 691            overwrite: bool=False,
 692    ) -> None:
 693        """Move a file or subcollection object to this collection
 694
 695        Arguments:
 696
 697        * source: str | arvados.arvfile.ArvadosFile |
 698          arvados.collection.Subcollection --- The file or subcollection to
 699          add to this collection. If `source` is a str, the object will be
 700          found by looking up this path from `source_collection` (see
 701          below).
 702
 703        * target_path: str --- The path inside this collection where the
 704          source object should be added.
 705
 706        * source_collection: arvados.collection.Collection | None --- The
 707          collection to find the source object from when `source` is a
 708          path. Defaults to the current collection (`self`).
 709
 710        * overwrite: bool --- Controls the behavior of this method when the
 711          collection already contains an object at `target_path`. If `False`
 712          (the default), this method will raise `FileExistsError`. If `True`,
 713          the object at `target_path` will be replaced with `source_obj`.
 714        """
 715        source_obj, target_dir, target_name = self._get_src_target(source, target_path, source_collection, False)
 716        if not source_obj.writable():
 717            raise IOError(errno.EROFS, "Source collection is read only", source)
 718        target_dir.add(source_obj, target_name, overwrite, True)
 719
 720    def portable_manifest_text(self, stream_name: str=".") -> str:
 721        """Get the portable manifest text for this collection
 722
 723        The portable manifest text is normalized, and does not include access
 724        tokens. This method does not flush outstanding blocks to Keep.
 725
 726        Arguments:
 727
 728        * stream_name: str --- The name to use for this collection's stream in
 729          the generated manifest. Default `'.'`.
 730        """
 731        return self._get_manifest_text(stream_name, True, True)
 732
 733    @synchronized
 734    def manifest_text(
 735            self,
 736            stream_name: str=".",
 737            strip: bool=False,
 738            normalize: bool=False,
 739            only_committed: bool=False,
 740    ) -> str:
 741        """Get the manifest text for this collection
 742
 743        Arguments:
 744
 745        * stream_name: str --- The name to use for this collection's stream in
 746          the generated manifest. Default `'.'`.
 747
 748        * strip: bool --- Controls whether or not the returned manifest text
 749          includes access tokens. If `False` (the default), the manifest text
 750          will include access tokens. If `True`, the manifest text will not
 751          include access tokens.
 752
 753        * normalize: bool --- Controls whether or not the returned manifest
 754          text is normalized. Default `False`.
 755
 756        * only_committed: bool --- Controls whether or not this method uploads
 757          pending data to Keep before building and returning the manifest text.
 758          If `False` (the default), this method will finish uploading all data
 759          to Keep, then return the final manifest. If `True`, this method will
 760          build and return a manifest that only refers to the data that has
 761          finished uploading at the time this method was called.
 762        """
 763        if not only_committed:
 764            self._my_block_manager().commit_all()
 765        return self._get_manifest_text(stream_name, strip, normalize,
 766                                       only_committed=only_committed)
 767
 768    @synchronized
 769    def _get_manifest_text(self, stream_name, strip, normalize, only_committed=False):
 770        """Get the manifest text for this collection, sub collections and files.
 771
 772        :stream_name:
 773          Name to use for this stream (directory)
 774
 775        :strip:
 776          If True, remove signing tokens from block locators if present.
 777          If False (default), block locators are left unchanged.
 778
 779        :normalize:
 780          If True, always export the manifest text in normalized form
 781          even if the Collection is not modified.  If False (default) and the collection
 782          is not modified, return the original manifest text even if it is not
 783          in normalized form.
 784
 785        :only_committed:
 786          If True, only include blocks that were already committed to Keep.
 787
 788        """
 789
 790        if not self.committed() or self._manifest_text is None or normalize:
 791            stream = {}
 792            buf = []
 793            sorted_keys = sorted(self.keys())
 794            for filename in [s for s in sorted_keys if isinstance(self[s], ArvadosFile)]:
 795                # Create a stream per file `k`
 796                arvfile = self[filename]
 797                filestream = []
 798                for segment in arvfile.segments():
 799                    loc = segment.locator
 800                    if arvfile.parent._my_block_manager().is_bufferblock(loc):
 801                        if only_committed:
 802                            continue
 803                        loc = arvfile.parent._my_block_manager().get_bufferblock(loc).locator()
 804                    if strip:
 805                        loc = KeepLocator(loc).stripped()
 806                    filestream.append(LocatorAndRange(loc, KeepLocator(loc).size,
 807                                         segment.segment_offset, segment.range_size))
 808                stream[filename] = filestream
 809            if stream:
 810                buf.append(" ".join(normalize_stream(stream_name, stream)) + "\n")
 811            for dirname in [s for s in sorted_keys if isinstance(self[s], RichCollectionBase)]:
 812                buf.append(self[dirname].manifest_text(
 813                    stream_name=os.path.join(stream_name, dirname),
 814                    strip=strip, normalize=True, only_committed=only_committed))
 815            return "".join(buf)
 816        else:
 817            if strip:
 818                return self.stripped_manifest()
 819            else:
 820                return self._manifest_text
 821
 822    @synchronized
 823    def _copy_remote_blocks(self, remote_blocks={}):
 824        """Scan through the entire collection and ask Keep to copy remote blocks.
 825
 826        When accessing a remote collection, blocks will have a remote signature
 827        (+R instead of +A). Collect these signatures and request Keep to copy the
 828        blocks to the local cluster, returning local (+A) signatures.
 829
 830        :remote_blocks:
 831          Shared cache of remote to local block mappings. This is used to avoid
 832          doing extra work when blocks are shared by more than one file in
 833          different subdirectories.
 834
 835        """
 836        for item in self:
 837            remote_blocks = self[item]._copy_remote_blocks(remote_blocks)
 838        return remote_blocks
 839
 840    @synchronized
 841    def diff(
 842            self,
 843            end_collection: 'RichCollectionBase',
 844            prefix: str=".",
 845            holding_collection: Optional['Collection']=None,
 846    ) -> ChangeList:
 847        """Build a list of differences between this collection and another
 848
 849        Arguments:
 850
 851        * end_collection: arvados.collection.RichCollectionBase --- A
 852          collection object with the desired end state. The returned diff
 853          list will describe how to go from the current collection object
 854          `self` to `end_collection`.
 855
 856        * prefix: str --- The name to use for this collection's stream in
 857          the diff list. Default `'.'`.
 858
 859        * holding_collection: arvados.collection.Collection | None --- A
 860          collection object used to hold objects for the returned diff
 861          list. By default, a new empty collection is created.
 862        """
 863        changes = []
 864        if holding_collection is None:
 865            holding_collection = Collection(api_client=self._my_api(), keep_client=self._my_keep())
 866        for k in self:
 867            if k not in end_collection:
 868               changes.append((DEL, os.path.join(prefix, k), self[k].clone(holding_collection, "")))
 869        for k in end_collection:
 870            if k in self:
 871                if isinstance(end_collection[k], Subcollection) and isinstance(self[k], Subcollection):
 872                    changes.extend(self[k].diff(end_collection[k], os.path.join(prefix, k), holding_collection))
 873                elif end_collection[k] != self[k]:
 874                    changes.append((MOD, os.path.join(prefix, k), self[k].clone(holding_collection, ""), end_collection[k].clone(holding_collection, "")))
 875                else:
 876                    changes.append((TOK, os.path.join(prefix, k), self[k].clone(holding_collection, ""), end_collection[k].clone(holding_collection, "")))
 877            else:
 878                changes.append((ADD, os.path.join(prefix, k), end_collection[k].clone(holding_collection, "")))
 879        return changes
 880
 881    @must_be_writable
 882    @synchronized
 883    def apply(self, changes: ChangeList) -> None:
 884        """Apply a list of changes from to this collection
 885
 886        This method takes a list of changes generated by
 887        `RichCollectionBase.diff` and applies it to this
 888        collection. Afterward, the state of this collection object will
 889        match the state of `end_collection` passed to `diff`. If a change
 890        conflicts with a local change, it will be saved to an alternate path
 891        indicating the conflict.
 892
 893        Arguments:
 894
 895        * changes: arvados.collection.ChangeList --- The list of differences
 896          generated by `RichCollectionBase.diff`.
 897        """
 898        if changes:
 899            self.set_committed(False)
 900        for change in changes:
 901            event_type = change[0]
 902            path = change[1]
 903            initial = change[2]
 904            local = self.find(path)
 905            conflictpath = "%s~%s~conflict~" % (path, time.strftime("%Y%m%d-%H%M%S",
 906                                                                    time.gmtime()))
 907            if event_type == ADD:
 908                if local is None:
 909                    # No local file at path, safe to copy over new file
 910                    self.copy(initial, path)
 911                elif local is not None and local != initial:
 912                    # There is already local file and it is different:
 913                    # save change to conflict file.
 914                    self.copy(initial, conflictpath)
 915            elif event_type == MOD or event_type == TOK:
 916                final = change[3]
 917                if local == initial:
 918                    # Local matches the "initial" item so it has not
 919                    # changed locally and is safe to update.
 920                    if isinstance(local, ArvadosFile) and isinstance(final, ArvadosFile):
 921                        # Replace contents of local file with new contents
 922                        local.replace_contents(final)
 923                    else:
 924                        # Overwrite path with new item; this can happen if
 925                        # path was a file and is now a collection or vice versa
 926                        self.copy(final, path, overwrite=True)
 927                else:
 928                    # Local is missing (presumably deleted) or local doesn't
 929                    # match the "start" value, so save change to conflict file
 930                    self.copy(final, conflictpath)
 931            elif event_type == DEL:
 932                if local == initial:
 933                    # Local item matches "initial" value, so it is safe to remove.
 934                    self.remove(path, recursive=True)
 935                # else, the file is modified or already removed, in either
 936                # case we don't want to try to remove it.
 937
 938    def portable_data_hash(self) -> str:
 939        """Get the portable data hash for this collection's manifest"""
 940        if self._manifest_locator and self.committed():
 941            # If the collection is already saved on the API server, and it's committed
 942            # then return API server's PDH response.
 943            return self._portable_data_hash
 944        else:
 945            stripped = self.portable_manifest_text().encode()
 946            return '{}+{}'.format(hashlib.md5(stripped).hexdigest(), len(stripped))
 947
 948    @synchronized
 949    def subscribe(self, callback: ChangeCallback) -> None:
 950        """Set a notify callback for changes to this collection
 951
 952        Arguments:
 953
 954        * callback: arvados.collection.ChangeCallback --- The callable to
 955          call each time the collection is changed.
 956        """
 957        if self._callback is None:
 958            self._callback = callback
 959        else:
 960            raise errors.ArgumentError("A callback is already set on this collection.")
 961
 962    @synchronized
 963    def unsubscribe(self) -> None:
 964        """Remove any notify callback set for changes to this collection"""
 965        if self._callback is not None:
 966            self._callback = None
 967
 968    @synchronized
 969    def notify(
 970            self,
 971            event: ChangeType,
 972            collection: 'RichCollectionBase',
 973            name: str,
 974            item: CollectionItem,
 975    ) -> None:
 976        """Notify any subscribed callback about a change to this collection
 977
 978        .. ATTENTION:: Internal
 979           This method is only meant to be used by other Collection methods.
 980
 981        If a callback has been registered with `RichCollectionBase.subscribe`,
 982        it will be called with information about a change to this collection.
 983        Then this notification will be propagated to this collection's root.
 984
 985        Arguments:
 986
 987        * event: Literal[ADD, DEL, MOD, TOK] --- The type of modification to
 988          the collection.
 989
 990        * collection: arvados.collection.RichCollectionBase --- The
 991          collection that was modified.
 992
 993        * name: str --- The name of the file or stream within `collection` that
 994          was modified.
 995
 996        * item: arvados.arvfile.ArvadosFile |
 997          arvados.collection.Subcollection --- The new contents at `name`
 998          within `collection`.
 999        """
1000        if self._callback:
1001            self._callback(event, collection, name, item)
1002        self.root_collection().notify(event, collection, name, item)
1003
1004    @synchronized
1005    def __eq__(self, other: Any) -> bool:
1006        """Indicate whether this collection object is equal to another"""
1007        if other is self:
1008            return True
1009        if not isinstance(other, RichCollectionBase):
1010            return False
1011        if len(self._items) != len(other):
1012            return False
1013        for k in self._items:
1014            if k not in other:
1015                return False
1016            if self._items[k] != other[k]:
1017                return False
1018        return True
1019
1020    def __ne__(self, other: Any) -> bool:
1021        """Indicate whether this collection object is not equal to another"""
1022        return not self.__eq__(other)
1023
1024    @synchronized
1025    def flush(self) -> None:
1026        """Upload any pending data to Keep"""
1027        for e in listvalues(self):
1028            e.flush()
1029
1030
1031class Collection(RichCollectionBase):
1032    """Read and manipulate an Arvados collection
1033
1034    This class provides a high-level interface to create, read, and update
1035    Arvados collections and their contents. Refer to the Arvados Python SDK
1036    cookbook for [an introduction to using the Collection class][cookbook].
1037
1038    [cookbook]: https://doc.arvados.org/sdk/python/cookbook.html#working-with-collections
1039    """
1040
1041    def __init__(self, manifest_locator_or_text: Optional[str]=None,
1042                 api_client: Optional['arvados.api_resources.ArvadosAPIClient']=None,
1043                 keep_client: Optional['arvados.keep.KeepClient']=None,
1044                 num_retries: int=10,
1045                 parent: Optional['Collection']=None,
1046                 apiconfig: Optional[Mapping[str, str]]=None,
1047                 block_manager: Optional['arvados.arvfile._BlockManager']=None,
1048                 replication_desired: Optional[int]=None,
1049                 storage_classes_desired: Optional[List[str]]=None,
1050                 put_threads: Optional[int]=None):
1051        """Initialize a Collection object
1052
1053        Arguments:
1054
1055        * manifest_locator_or_text: str | None --- This string can contain a
1056          collection manifest text, portable data hash, or UUID. When given a
1057          portable data hash or UUID, this instance will load a collection
1058          record from the API server. Otherwise, this instance will represent a
1059          new collection without an API server record. The default value `None`
1060          instantiates a new collection with an empty manifest.
1061
1062        * api_client: arvados.api_resources.ArvadosAPIClient | None --- The
1063          Arvados API client object this instance uses to make requests. If
1064          none is given, this instance creates its own client using the
1065          settings from `apiconfig` (see below). If your client instantiates
1066          many Collection objects, you can help limit memory utilization by
1067          calling `arvados.api.api` to construct an
1068          `arvados.safeapi.ThreadSafeApiCache`, and use that as the `api_client`
1069          for every Collection.
1070
1071        * keep_client: arvados.keep.KeepClient | None --- The Keep client
1072          object this instance uses to make requests. If none is given, this
1073          instance creates its own client using its `api_client`.
1074
1075        * num_retries: int --- The number of times that client requests are
1076          retried. Default 10.
1077
1078        * parent: arvados.collection.Collection | None --- The parent Collection
1079          object of this instance, if any. This argument is primarily used by
1080          other Collection methods; user client code shouldn't need to use it.
1081
1082        * apiconfig: Mapping[str, str] | None --- A mapping with entries for
1083          `ARVADOS_API_HOST`, `ARVADOS_API_TOKEN`, and optionally
1084          `ARVADOS_API_HOST_INSECURE`. When no `api_client` is provided, the
1085          Collection object constructs one from these settings. If no
1086          mapping is provided, calls `arvados.config.settings` to get these
1087          parameters from user configuration.
1088
1089        * block_manager: arvados.arvfile._BlockManager | None --- The
1090          _BlockManager object used by this instance to coordinate reading
1091          and writing Keep data blocks. If none is given, this instance
1092          constructs its own. This argument is primarily used by other
1093          Collection methods; user client code shouldn't need to use it.
1094
1095        * replication_desired: int | None --- This controls both the value of
1096          the `replication_desired` field on API collection records saved by
1097          this class, as well as the number of Keep services that the object
1098          writes new data blocks to. If none is given, uses the default value
1099          configured for the cluster.
1100
1101        * storage_classes_desired: list[str] | None --- This controls both
1102          the value of the `storage_classes_desired` field on API collection
1103          records saved by this class, as well as selecting which specific
1104          Keep services the object writes new data blocks to. If none is
1105          given, defaults to an empty list.
1106
1107        * put_threads: int | None --- The number of threads to run
1108          simultaneously to upload data blocks to Keep. This value is used when
1109          building a new `block_manager`. It is unused when a `block_manager`
1110          is provided.
1111        """
1112
1113        if storage_classes_desired and type(storage_classes_desired) is not list:
1114            raise errors.ArgumentError("storage_classes_desired must be list type.")
1115
1116        super(Collection, self).__init__(parent)
1117        self._api_client = api_client
1118        self._keep_client = keep_client
1119
1120        # Use the keep client from ThreadSafeApiCache
1121        if self._keep_client is None and isinstance(self._api_client, ThreadSafeApiCache):
1122            self._keep_client = self._api_client.keep
1123
1124        self._block_manager = block_manager
1125        self.replication_desired = replication_desired
1126        self._storage_classes_desired = storage_classes_desired
1127        self.put_threads = put_threads
1128
1129        if apiconfig:
1130            self._config = apiconfig
1131        else:
1132            self._config = config.settings()
1133
1134        self.num_retries = num_retries
1135        self._manifest_locator = None
1136        self._manifest_text = None
1137        self._portable_data_hash = None
1138        self._api_response = None
1139        self._past_versions = set()
1140
1141        self.lock = threading.RLock()
1142        self.events = None
1143
1144        if manifest_locator_or_text:
1145            if re.match(arvados.util.keep_locator_pattern, manifest_locator_or_text):
1146                self._manifest_locator = manifest_locator_or_text
1147            elif re.match(arvados.util.collection_uuid_pattern, manifest_locator_or_text):
1148                self._manifest_locator = manifest_locator_or_text
1149                if not self._has_local_collection_uuid():
1150                    self._has_remote_blocks = True
1151            elif re.match(arvados.util.manifest_pattern, manifest_locator_or_text):
1152                self._manifest_text = manifest_locator_or_text
1153                if '+R' in self._manifest_text:
1154                    self._has_remote_blocks = True
1155            else:
1156                raise errors.ArgumentError(
1157                    "Argument to CollectionReader is not a manifest or a collection UUID")
1158
1159            try:
1160                self._populate()
1161            except errors.SyntaxError as e:
1162                raise errors.ArgumentError("Error processing manifest text: %s", str(e)) from None
1163
1164    def storage_classes_desired(self) -> List[str]:
1165        """Get this collection's `storage_classes_desired` value"""
1166        return self._storage_classes_desired or []
1167
1168    def root_collection(self) -> 'Collection':
1169        return self
1170
1171    def get_properties(self) -> Properties:
1172        """Get this collection's properties
1173
1174        This method always returns a dict. If this collection object does not
1175        have an associated API record, or that record does not have any
1176        properties set, this method returns an empty dict.
1177        """
1178        if self._api_response and self._api_response["properties"]:
1179            return self._api_response["properties"]
1180        else:
1181            return {}
1182
1183    def get_trash_at(self) -> Optional[datetime.datetime]:
1184        """Get this collection's `trash_at` field
1185
1186        This method parses the `trash_at` field of the collection's API
1187        record and returns a datetime from it. If that field is not set, or
1188        this collection object does not have an associated API record,
1189        returns None.
1190        """
1191        if self._api_response and self._api_response["trash_at"]:
1192            try:
1193                return ciso8601.parse_datetime(self._api_response["trash_at"])
1194            except ValueError:
1195                return None
1196        else:
1197            return None
1198
1199    def stream_name(self) -> str:
1200        return "."
1201
1202    def writable(self) -> bool:
1203        return True
1204
1205    @synchronized
1206    def known_past_version(
1207            self,
1208            modified_at_and_portable_data_hash: Tuple[Optional[str], Optional[str]]
1209    ) -> bool:
1210        """Indicate whether an API record for this collection has been seen before
1211
1212        As this collection object loads records from the API server, it records
1213        their `modified_at` and `portable_data_hash` fields. This method accepts
1214        a 2-tuple with values for those fields, and returns `True` if the
1215        combination was previously loaded.
1216        """
1217        return modified_at_and_portable_data_hash in self._past_versions
1218
1219    @synchronized
1220    @retry_method
1221    def update(
1222            self,
1223            other: Optional['Collection']=None,
1224            num_retries: Optional[int]=None,
1225    ) -> None:
1226        """Merge another collection's contents into this one
1227
1228        This method compares the manifest of this collection instance with
1229        another, then updates this instance's manifest with changes from the
1230        other, renaming files to flag conflicts where necessary.
1231
1232        When called without any arguments, this method reloads the collection's
1233        API record, and updates this instance with any changes that have
1234        appeared server-side. If this instance does not have a corresponding
1235        API record, this method raises `arvados.errors.ArgumentError`.
1236
1237        Arguments:
1238
1239        * other: arvados.collection.Collection | None --- The collection
1240          whose contents should be merged into this instance. When not
1241          provided, this method reloads this collection's API record and
1242          constructs a Collection object from it.  If this instance does not
1243          have a corresponding API record, this method raises
1244          `arvados.errors.ArgumentError`.
1245
1246        * num_retries: int | None --- The number of times to retry reloading
1247          the collection's API record from the API server. If not specified,
1248          uses the `num_retries` provided when this instance was constructed.
1249        """
1250        if other is None:
1251            if self._manifest_locator is None:
1252                raise errors.ArgumentError("`other` is None but collection does not have a manifest_locator uuid")
1253            response = self._my_api().collections().get(uuid=self._manifest_locator).execute(num_retries=num_retries)
1254            if (self.known_past_version((response.get("modified_at"), response.get("portable_data_hash"))) and
1255                response.get("portable_data_hash") != self.portable_data_hash()):
1256                # The record on the server is different from our current one, but we've seen it before,
1257                # so ignore it because it's already been merged.
1258                # However, if it's the same as our current record, proceed with the update, because we want to update
1259                # our tokens.
1260                return
1261            else:
1262                self._remember_api_response(response)
1263            other = CollectionReader(response["manifest_text"])
1264        baseline = CollectionReader(self._manifest_text)
1265        self.apply(baseline.diff(other))
1266        self._manifest_text = self.manifest_text()
1267
1268    @synchronized
1269    def _my_api(self):
1270        if self._api_client is None:
1271            self._api_client = ThreadSafeApiCache(self._config, version='v1')
1272            if self._keep_client is None:
1273                self._keep_client = self._api_client.keep
1274        return self._api_client
1275
1276    @synchronized
1277    def _my_keep(self):
1278        if self._keep_client is None:
1279            if self._api_client is None:
1280                self._my_api()
1281            else:
1282                self._keep_client = KeepClient(api_client=self._api_client)
1283        return self._keep_client
1284
1285    @synchronized
1286    def _my_block_manager(self):
1287        if self._block_manager is None:
1288            copies = (self.replication_desired or
1289                      self._my_api()._rootDesc.get('defaultCollectionReplication',
1290                                                   2))
1291            self._block_manager = _BlockManager(self._my_keep(),
1292                                                copies=copies,
1293                                                put_threads=self.put_threads,
1294                                                num_retries=self.num_retries,
1295                                                storage_classes_func=self.storage_classes_desired)
1296        return self._block_manager
1297
1298    def _remember_api_response(self, response):
1299        self._api_response = response
1300        self._past_versions.add((response.get("modified_at"), response.get("portable_data_hash")))
1301
1302    def _populate_from_api_server(self):
1303        # As in KeepClient itself, we must wait until the last
1304        # possible moment to instantiate an API client, in order to
1305        # avoid tripping up clients that don't have access to an API
1306        # server.  If we do build one, make sure our Keep client uses
1307        # it.  If instantiation fails, we'll fall back to the except
1308        # clause, just like any other Collection lookup
1309        # failure. Return an exception, or None if successful.
1310        self._remember_api_response(self._my_api().collections().get(
1311            uuid=self._manifest_locator).execute(
1312                num_retries=self.num_retries))
1313        self._manifest_text = self._api_response['manifest_text']
1314        self._portable_data_hash = self._api_response['portable_data_hash']
1315        # If not overriden via kwargs, we should try to load the
1316        # replication_desired and storage_classes_desired from the API server
1317        if self.replication_desired is None:
1318            self.replication_desired = self._api_response.get('replication_desired', None)
1319        if self._storage_classes_desired is None:
1320            self._storage_classes_desired = self._api_response.get('storage_classes_desired', None)
1321
1322    def _populate(self):
1323        if self._manifest_text is None:
1324            if self._manifest_locator is None:
1325                return
1326            else:
1327                self._populate_from_api_server()
1328        self._baseline_manifest = self._manifest_text
1329        self._import_manifest(self._manifest_text)
1330
1331    def _has_collection_uuid(self):
1332        return self._manifest_locator is not None and re.match(arvados.util.collection_uuid_pattern, self._manifest_locator)
1333
1334    def _has_local_collection_uuid(self):
1335        return self._has_collection_uuid and \
1336            self._my_api()._rootDesc['uuidPrefix'] == self._manifest_locator.split('-')[0]
1337
1338    def __enter__(self):
1339        return self
1340
1341    def __exit__(self, exc_type, exc_value, traceback):
1342        """Exit a context with this collection instance
1343
1344        If no exception was raised inside the context block, and this
1345        collection is writable and has a corresponding API record, that
1346        record will be updated to match the state of this instance at the end
1347        of the block.
1348        """
1349        if exc_type is None:
1350            if self.writable() and self._has_collection_uuid():
1351                self.save()
1352        self.stop_threads()
1353
1354    def stop_threads(self) -> None:
1355        """Stop background Keep upload/download threads"""
1356        if self._block_manager is not None:
1357            self._block_manager.stop_threads()
1358
1359    @synchronized
1360    def manifest_locator(self) -> Optional[str]:
1361        """Get this collection's manifest locator, if any
1362
1363        * If this collection instance is associated with an API record with a
1364          UUID, return that.
1365        * Otherwise, if this collection instance was loaded from an API record
1366          by portable data hash, return that.
1367        * Otherwise, return `None`.
1368        """
1369        return self._manifest_locator
1370
1371    @synchronized
1372    def clone(
1373            self,
1374            new_parent: Optional['Collection']=None,
1375            new_name: Optional[str]=None,
1376            readonly: bool=False,
1377            new_config: Optional[Mapping[str, str]]=None,
1378    ) -> 'Collection':
1379        """Create a Collection object with the same contents as this instance
1380
1381        This method creates a new Collection object with contents that match
1382        this instance's. The new collection will not be associated with any API
1383        record.
1384
1385        Arguments:
1386
1387        * new_parent: arvados.collection.Collection | None --- This value is
1388          passed to the new Collection's constructor as the `parent`
1389          argument.
1390
1391        * new_name: str | None --- This value is unused.
1392
1393        * readonly: bool --- If this value is true, this method constructs and
1394          returns a `CollectionReader`. Otherwise, it returns a mutable
1395          `Collection`. Default `False`.
1396
1397        * new_config: Mapping[str, str] | None --- This value is passed to the
1398          new Collection's constructor as `apiconfig`. If no value is provided,
1399          defaults to the configuration passed to this instance's constructor.
1400        """
1401        if new_config is None:
1402            new_config = self._config
1403        if readonly:
1404            newcollection = CollectionReader(parent=new_parent, apiconfig=new_config)
1405        else:
1406            newcollection = Collection(parent=new_parent, apiconfig=new_config)
1407
1408        newcollection._clonefrom(self)
1409        return newcollection
1410
1411    @synchronized
1412    def api_response(self) -> Optional[Dict[str, Any]]:
1413        """Get this instance's associated API record
1414
1415        If this Collection instance has an associated API record, return it.
1416        Otherwise, return `None`.
1417        """
1418        return self._api_response
1419
1420    def find_or_create(
1421            self,
1422            path: str,
1423            create_type: CreateType,
1424    ) -> CollectionItem:
1425        if path == ".":
1426            return self
1427        else:
1428            return super(Collection, self).find_or_create(path[2:] if path.startswith("./") else path, create_type)
1429
1430    def find(self, path: str) -> CollectionItem:
1431        if path == ".":
1432            return self
1433        else:
1434            return super(Collection, self).find(path[2:] if path.startswith("./") else path)
1435
1436    def remove(self, path: str, recursive: bool=False) -> None:
1437        if path == ".":
1438            raise errors.ArgumentError("Cannot remove '.'")
1439        else:
1440            return super(Collection, self).remove(path[2:] if path.startswith("./") else path, recursive)
1441
1442    @must_be_writable
1443    @synchronized
1444    @retry_method
1445    def save(
1446            self,
1447            properties: Optional[Properties]=None,
1448            storage_classes: Optional[StorageClasses]=None,
1449            trash_at: Optional[datetime.datetime]=None,
1450            merge: bool=True,
1451            num_retries: Optional[int]=None,
1452            preserve_version: bool=False,
1453    ) -> str:
1454        """Save collection to an existing API record
1455
1456        This method updates the instance's corresponding API record to match
1457        the instance's state. If this instance does not have a corresponding API
1458        record yet, raises `AssertionError`. (To create a new API record, use
1459        `Collection.save_new`.) This method returns the saved collection
1460        manifest.
1461
1462        Arguments:
1463
1464        * properties: dict[str, Any] | None --- If provided, the API record will
1465          be updated with these properties. Note this will completely replace
1466          any existing properties.
1467
1468        * storage_classes: list[str] | None --- If provided, the API record will
1469          be updated with this value in the `storage_classes_desired` field.
1470          This value will also be saved on the instance and used for any
1471          changes that follow.
1472
1473        * trash_at: datetime.datetime | None --- If provided, the API record
1474          will be updated with this value in the `trash_at` field.
1475
1476        * merge: bool --- If `True` (the default), this method will first
1477          reload this collection's API record, and merge any new contents into
1478          this instance before saving changes. See `Collection.update` for
1479          details.
1480
1481        * num_retries: int | None --- The number of times to retry reloading
1482          the collection's API record from the API server. If not specified,
1483          uses the `num_retries` provided when this instance was constructed.
1484
1485        * preserve_version: bool --- This value will be passed to directly
1486          to the underlying API call. If `True`, the Arvados API will
1487          preserve the versions of this collection both immediately before
1488          and after the update. If `True` when the API server is not
1489          configured with collection versioning, this method raises
1490          `arvados.errors.ArgumentError`.
1491        """
1492        if properties and type(properties) is not dict:
1493            raise errors.ArgumentError("properties must be dictionary type.")
1494
1495        if storage_classes and type(storage_classes) is not list:
1496            raise errors.ArgumentError("storage_classes must be list type.")
1497        if storage_classes:
1498            self._storage_classes_desired = storage_classes
1499
1500        if trash_at and type(trash_at) is not datetime.datetime:
1501            raise errors.ArgumentError("trash_at must be datetime type.")
1502
1503        if preserve_version and not self._my_api().config()['Collections'].get('CollectionVersioning', False):
1504            raise errors.ArgumentError("preserve_version is not supported when CollectionVersioning is not enabled.")
1505
1506        body={}
1507        if properties:
1508            body["properties"] = properties
1509        if self.storage_classes_desired():
1510            body["storage_classes_desired"] = self.storage_classes_desired()
1511        if trash_at:
1512            t = trash_at.strftime("%Y-%m-%dT%H:%M:%S.%fZ")
1513            body["trash_at"] = t
1514        if preserve_version:
1515            body["preserve_version"] = preserve_version
1516
1517        if not self.committed():
1518            if self._has_remote_blocks:
1519                # Copy any remote blocks to the local cluster.
1520                self._copy_remote_blocks(remote_blocks={})
1521                self._has_remote_blocks = False
1522            if not self._has_collection_uuid():
1523                raise AssertionError("Collection manifest_locator is not a collection uuid.  Use save_new() for new collections.")
1524            elif not self._has_local_collection_uuid():
1525                raise AssertionError("Collection manifest_locator is from a remote cluster. Use save_new() to save it on the local cluster.")
1526
1527            self._my_block_manager().commit_all()
1528
1529            if merge:
1530                self.update()
1531
1532            text = self.manifest_text(strip=False)
1533            body['manifest_text'] = text
1534
1535            self._remember_api_response(self._my_api().collections().update(
1536                uuid=self._manifest_locator,
1537                body=body
1538                ).execute(num_retries=num_retries))
1539            self._manifest_text = self._api_response["manifest_text"]
1540            self._portable_data_hash = self._api_response["portable_data_hash"]
1541            self.set_committed(True)
1542        elif body:
1543            self._remember_api_response(self._my_api().collections().update(
1544                uuid=self._manifest_locator,
1545                body=body
1546                ).execute(num_retries=num_retries))
1547
1548        return self._manifest_text
1549
1550
1551    @must_be_writable
1552    @synchronized
1553    @retry_method
1554    def save_new(
1555            self,
1556            name: Optional[str]=None,
1557            create_collection_record: bool=True,
1558            owner_uuid: Optional[str]=None,
1559            properties: Optional[Properties]=None,
1560            storage_classes: Optional[StorageClasses]=None,
1561            trash_at: Optional[datetime.datetime]=None,
1562            ensure_unique_name: bool=False,
1563            num_retries: Optional[int]=None,
1564            preserve_version: bool=False,
1565    ):
1566        """Save collection to a new API record
1567
1568        This method finishes uploading new data blocks and (optionally)
1569        creates a new API collection record with the provided data. If a new
1570        record is created, this instance becomes associated with that record
1571        for future updates like `save()`. This method returns the saved
1572        collection manifest.
1573
1574        Arguments:
1575
1576        * name: str | None --- The `name` field to use on the new collection
1577          record. If not specified, a generic default name is generated.
1578
1579        * create_collection_record: bool --- If `True` (the default), creates a
1580          collection record on the API server. If `False`, the method finishes
1581          all data uploads and only returns the resulting collection manifest
1582          without sending it to the API server.
1583
1584        * owner_uuid: str | None --- The `owner_uuid` field to use on the
1585          new collection record.
1586
1587        * properties: dict[str, Any] | None --- The `properties` field to use on
1588          the new collection record.
1589
1590        * storage_classes: list[str] | None --- The
1591          `storage_classes_desired` field to use on the new collection record.
1592
1593        * trash_at: datetime.datetime | None --- The `trash_at` field to use
1594          on the new collection record.
1595
1596        * ensure_unique_name: bool --- This value is passed directly to the
1597          Arvados API when creating the collection record. If `True`, the API
1598          server may modify the submitted `name` to ensure the collection's
1599          `name`+`owner_uuid` combination is unique. If `False` (the default),
1600          if a collection already exists with this same `name`+`owner_uuid`
1601          combination, creating a collection record will raise a validation
1602          error.
1603
1604        * num_retries: int | None --- The number of times to retry reloading
1605          the collection's API record from the API server. If not specified,
1606          uses the `num_retries` provided when this instance was constructed.
1607
1608        * preserve_version: bool --- This value will be passed to directly
1609          to the underlying API call. If `True`, the Arvados API will
1610          preserve the versions of this collection both immediately before
1611          and after the update. If `True` when the API server is not
1612          configured with collection versioning, this method raises
1613          `arvados.errors.ArgumentError`.
1614        """
1615        if properties and type(properties) is not dict:
1616            raise errors.ArgumentError("properties must be dictionary type.")
1617
1618        if storage_classes and type(storage_classes) is not list:
1619            raise errors.ArgumentError("storage_classes must be list type.")
1620
1621        if trash_at and type(trash_at) is not datetime.datetime:
1622            raise errors.ArgumentError("trash_at must be datetime type.")
1623
1624        if preserve_version and not self._my_api().config()['Collections'].get('CollectionVersioning', False):
1625            raise errors.ArgumentError("preserve_version is not supported when CollectionVersioning is not enabled.")
1626
1627        if self._has_remote_blocks:
1628            # Copy any remote blocks to the local cluster.
1629            self._copy_remote_blocks(remote_blocks={})
1630            self._has_remote_blocks = False
1631
1632        if storage_classes:
1633            self._storage_classes_desired = storage_classes
1634
1635        self._my_block_manager().commit_all()
1636        text = self.manifest_text(strip=False)
1637
1638        if create_collection_record:
1639            if name is None:
1640                name = "New collection"
1641                ensure_unique_name = True
1642
1643            body = {"manifest_text": text,
1644                    "name": name,
1645                    "replication_desired": self.replication_desired}
1646            if owner_uuid:
1647                body["owner_uuid"] = owner_uuid
1648            if properties:
1649                body["properties"] = properties
1650            if self.storage_classes_desired():
1651                body["storage_classes_desired"] = self.storage_classes_desired()
1652            if trash_at:
1653                t = trash_at.strftime("%Y-%m-%dT%H:%M:%S.%fZ")
1654                body["trash_at"] = t
1655            if preserve_version:
1656                body["preserve_version"] = preserve_version
1657
1658            self._remember_api_response(self._my_api().collections().create(ensure_unique_name=ensure_unique_name, body=body).execute(num_retries=num_retries))
1659            text = self._api_response["manifest_text"]
1660
1661            self._manifest_locator = self._api_response["uuid"]
1662            self._portable_data_hash = self._api_response["portable_data_hash"]
1663
1664            self._manifest_text = text
1665            self.set_committed(True)
1666
1667        return text
1668
1669    _token_re = re.compile(r'(\S+)(\s+|$)')
1670    _block_re = re.compile(r'[0-9a-f]{32}\+(\d+)(\+\S+)*')
1671    _segment_re = re.compile(r'(\d+):(\d+):(\S+)')
1672
1673    def _unescape_manifest_path(self, path):
1674        return re.sub(r'\\([0-3][0-7][0-7])', lambda m: chr(int(m.group(1), 8)), path)
1675
1676    @synchronized
1677    def _import_manifest(self, manifest_text):
1678        """Import a manifest into a `Collection`.
1679
1680        :manifest_text:
1681          The manifest text to import from.
1682
1683        """
1684        if len(self) > 0:
1685            raise ArgumentError("Can only import manifest into an empty collection")
1686
1687        STREAM_NAME = 0
1688        BLOCKS = 1
1689        SEGMENTS = 2
1690
1691        stream_name = None
1692        state = STREAM_NAME
1693
1694        for token_and_separator in self._token_re.finditer(manifest_text):
1695            tok = token_and_separator.group(1)
1696            sep = token_and_separator.group(2)
1697
1698            if state == STREAM_NAME:
1699                # starting a new stream
1700                stream_name = self._unescape_manifest_path(tok)
1701                blocks = []
1702                segments = []
1703                streamoffset = 0
1704                state = BLOCKS
1705                self.find_or_create(stream_name, COLLECTION)
1706                continue
1707
1708            if state == BLOCKS:
1709                block_locator = self._block_re.match(tok)
1710                if block_locator:
1711                    blocksize = int(block_locator.group(1))
1712                    blocks.append(Range(tok, streamoffset, blocksize, 0))
1713                    streamoffset += blocksize
1714                else:
1715                    state = SEGMENTS
1716
1717            if state == SEGMENTS:
1718                file_segment = self._segment_re.match(tok)
1719                if file_segment:
1720                    pos = int(file_segment.group(1))
1721                    size = int(file_segment.group(2))
1722                    name = self._unescape_manifest_path(file_segment.group(3))
1723                    if name.split('/')[-1] == '.':
1724                        # placeholder for persisting an empty directory, not a real file
1725                        if len(name) > 2:
1726                            self.find_or_create(os.path.join(stream_name, name[:-2]), COLLECTION)
1727                    else:
1728                        filepath = os.path.join(stream_name, name)
1729                        try:
1730                            afile = self.find_or_create(filepath, FILE)
1731                        except IOError as e:
1732                            if e.errno == errno.ENOTDIR:
1733                                raise errors.SyntaxError("Dir part of %s conflicts with file of the same name.", filepath) from None
1734                            else:
1735                                raise e from None
1736                        if isinstance(afile, ArvadosFile):
1737                            afile.add_segment(blocks, pos, size)
1738                        else:
1739                            raise errors.SyntaxError("File %s conflicts with stream of the same name.", filepath)
1740                else:
1741                    # error!
1742                    raise errors.SyntaxError("Invalid manifest format, expected file segment but did not match format: '%s'" % tok)
1743
1744            if sep == "\n":
1745                stream_name = None
1746                state = STREAM_NAME
1747
1748        self.set_committed(True)
1749
1750    @synchronized
1751    def notify(
1752            self,
1753            event: ChangeType,
1754            collection: 'RichCollectionBase',
1755            name: str,
1756            item: CollectionItem,
1757    ) -> None:
1758        if self._callback:
1759            self._callback(event, collection, name, item)
1760
1761
1762class Subcollection(RichCollectionBase):
1763    """Read and manipulate a stream/directory within an Arvados collection
1764
1765    This class represents a single stream (like a directory) within an Arvados
1766    `Collection`. It is returned by `Collection.find` and provides the same API.
1767    Operations that work on the API collection record propagate to the parent
1768    `Collection` object.
1769    """
1770
1771    def __init__(self, parent, name):
1772        super(Subcollection, self).__init__(parent)
1773        self.lock = self.root_collection().lock
1774        self._manifest_text = None
1775        self.name = name
1776        self.num_retries = parent.num_retries
1777
1778    def root_collection(self) -> 'Collection':
1779        return self.parent.root_collection()
1780
1781    def writable(self) -> bool:
1782        return self.root_collection().writable()
1783
1784    def _my_api(self):
1785        return self.root_collection()._my_api()
1786
1787    def _my_keep(self):
1788        return self.root_collection()._my_keep()
1789
1790    def _my_block_manager(self):
1791        return self.root_collection()._my_block_manager()
1792
1793    def stream_name(self) -> str:
1794        return os.path.join(self.parent.stream_name(), self.name)
1795
1796    @synchronized
1797    def clone(
1798            self,
1799            new_parent: Optional['Collection']=None,
1800            new_name: Optional[str]=None,
1801    ) -> 'Subcollection':
1802        c = Subcollection(new_parent, new_name)
1803        c._clonefrom(self)
1804        return c
1805
1806    @must_be_writable
1807    @synchronized
1808    def _reparent(self, newparent, newname):
1809        self.set_committed(False)
1810        self.flush()
1811        self.parent.remove(self.name, recursive=True)
1812        self.parent = newparent
1813        self.name = newname
1814        self.lock = self.parent.root_collection().lock
1815
1816    @synchronized
1817    def _get_manifest_text(self, stream_name, strip, normalize, only_committed=False):
1818        """Encode empty directories by using an \056-named (".") empty file"""
1819        if len(self._items) == 0:
1820            return "%s %s 0:0:\\056\n" % (
1821                escape(stream_name), config.EMPTY_BLOCK_LOCATOR)
1822        return super(Subcollection, self)._get_manifest_text(stream_name,
1823                                                             strip, normalize,
1824                                                             only_committed)
1825
1826
1827class CollectionReader(Collection):
1828    """Read-only `Collection` subclass
1829
1830    This class will never create or update any API collection records. You can
1831    use this class for additional code safety when you only need to read
1832    existing collections.
1833    """
1834    def __init__(self, manifest_locator_or_text, *args, **kwargs):
1835        self._in_init = True
1836        super(CollectionReader, self).__init__(manifest_locator_or_text, *args, **kwargs)
1837        self._in_init = False
1838
1839        # Forego any locking since it should never change once initialized.
1840        self.lock = NoopLock()
1841
1842        # Backwards compatability with old CollectionReader
1843        # all_streams() and all_files()
1844        self._streams = None
1845
1846    def writable(self) -> bool:
1847        return self._in_init
1848
1849    def _populate_streams(orig_func):
1850        @functools.wraps(orig_func)
1851        def populate_streams_wrapper(self, *args, **kwargs):
1852            # Defer populating self._streams until needed since it creates a copy of the manifest.
1853            if self._streams is None:
1854                if self._manifest_text:
1855                    self._streams = [sline.split()
1856                                     for sline in self._manifest_text.split("\n")
1857                                     if sline]
1858                else:
1859                    self._streams = []
1860            return orig_func(self, *args, **kwargs)
1861        return populate_streams_wrapper
1862
1863    @arvados.util._deprecated('3.0', 'Collection iteration')
1864    @_populate_streams
1865    def normalize(self):
1866        """Normalize the streams returned by `all_streams`"""
1867        streams = {}
1868        for s in self.all_streams():
1869            for f in s.all_files():
1870                streamname, filename = split(s.name() + "/" + f.name())
1871                if streamname not in streams:
1872                    streams[streamname] = {}
1873                if filename not in streams[streamname]:
1874                    streams[streamname][filename] = []
1875                for r in f.segments:
1876                    streams[streamname][filename].extend(s.locators_and_ranges(r.locator, r.range_size))
1877
1878        self._streams = [normalize_stream(s, streams[s])
1879                         for s in sorted(streams)]
1880
1881    @arvados.util._deprecated('3.0', 'Collection iteration')
1882    @_populate_streams
1883    def all_streams(self):
1884        return [StreamReader(s, self._my_keep(), num_retries=self.num_retries)
1885                for s in self._streams]
1886
1887    @arvados.util._deprecated('3.0', 'Collection iteration')
1888    @_populate_streams
1889    def all_files(self):
1890        for s in self.all_streams():
1891            for f in s.all_files():
1892                yield f
1893
1894
1895class CollectionWriter(CollectionBase):
1896    """Create a new collection from scratch
1897
1898    .. WARNING:: Deprecated
1899       This class is deprecated. Prefer `arvados.collection.Collection`
1900       instead.
1901    """
1902
1903    @arvados.util._deprecated('3.0', 'arvados.collection.Collection')
1904    def __init__(self, api_client=None, num_retries=0, replication=None):
1905        """Instantiate a CollectionWriter.
1906
1907        CollectionWriter lets you build a new Arvados Collection from scratch.
1908        Write files to it.  The CollectionWriter will upload data to Keep as
1909        appropriate, and provide you with the Collection manifest text when
1910        you're finished.
1911
1912        Arguments:
1913        * api_client: The API client to use to look up Collections.  If not
1914          provided, CollectionReader will build one from available Arvados
1915          configuration.
1916        * num_retries: The default number of times to retry failed
1917          service requests.  Default 0.  You may change this value
1918          after instantiation, but note those changes may not
1919          propagate to related objects like the Keep client.
1920        * replication: The number of copies of each block to store.
1921          If this argument is None or not supplied, replication is
1922          the server-provided default if available, otherwise 2.
1923        """
1924        self._api_client = api_client
1925        self.num_retries = num_retries
1926        self.replication = (2 if replication is None else replication)
1927        self._keep_client = None
1928        self._data_buffer = []
1929        self._data_buffer_len = 0
1930        self._current_stream_files = []
1931        self._current_stream_length = 0
1932        self._current_stream_locators = []
1933        self._current_stream_name = '.'
1934        self._current_file_name = None
1935        self._current_file_pos = 0
1936        self._finished_streams = []
1937        self._close_file = None
1938        self._queued_file = None
1939        self._queued_dirents = deque()
1940        self._queued_trees = deque()
1941        self._last_open = None
1942
1943    def __exit__(self, exc_type, exc_value, traceback):
1944        if exc_type is None:
1945            self.finish()
1946
1947    def do_queued_work(self):
1948        # The work queue consists of three pieces:
1949        # * _queued_file: The file object we're currently writing to the
1950        #   Collection.
1951        # * _queued_dirents: Entries under the current directory
1952        #   (_queued_trees[0]) that we want to write or recurse through.
1953        #   This may contain files from subdirectories if
1954        #   max_manifest_depth == 0 for this directory.
1955        # * _queued_trees: Directories that should be written as separate
1956        #   streams to the Collection.
1957        # This function handles the smallest piece of work currently queued
1958        # (current file, then current directory, then next directory) until
1959        # no work remains.  The _work_THING methods each do a unit of work on
1960        # THING.  _queue_THING methods add a THING to the work queue.
1961        while True:
1962            if self._queued_file:
1963                self._work_file()
1964            elif self._queued_dirents:
1965                self._work_dirents()
1966            elif self._queued_trees:
1967                self._work_trees()
1968            else:
1969                break
1970
1971    def _work_file(self):
1972        while True:
1973            buf = self._queued_file.read(config.KEEP_BLOCK_SIZE)
1974            if not buf:
1975                break
1976            self.write(buf)
1977        self.finish_current_file()
1978        if self._close_file:
1979            self._queued_file.close()
1980        self._close_file = None
1981        self._queued_file = None
1982
1983    def _work_dirents(self):
1984        path, stream_name, max_manifest_depth = self._queued_trees[0]
1985        if stream_name != self.current_stream_name():
1986            self.start_new_stream(stream_name)
1987        while self._queued_dirents:
1988            dirent = self._queued_dirents.popleft()
1989            target = os.path.join(path, dirent)
1990            if os.path.isdir(target):
1991                self._queue_tree(target,
1992                                 os.path.join(stream_name, dirent),
1993                                 max_manifest_depth - 1)
1994            else:
1995                self._queue_file(target, dirent)
1996                break
1997        if not self._queued_dirents:
1998            self._queued_trees.popleft()
1999
2000    def _work_trees(self):
2001        path, stream_name, max_manifest_depth = self._queued_trees[0]
2002        d = arvados.util.listdir_recursive(
2003            path, max_depth = (None if max_manifest_depth == 0 else 0))
2004        if d:
2005            self._queue_dirents(stream_name, d)
2006        else:
2007            self._queued_trees.popleft()
2008
2009    def _queue_file(self, source, filename=None):
2010        assert (self._queued_file is None), "tried to queue more than one file"
2011        if not hasattr(source, 'read'):
2012            source = open(source, 'rb')
2013            self._close_file = True
2014        else:
2015            self._close_file = False
2016        if filename is None:
2017            filename = os.path.basename(source.name)
2018        self.start_new_file(filename)
2019        self._queued_file = source
2020
2021    def _queue_dirents(self, stream_name, dirents):
2022        assert (not self._queued_dirents), "tried to queue more than one tree"
2023        self._queued_dirents = deque(sorted(dirents))
2024
2025    def _queue_tree(self, path, stream_name, max_manifest_depth):
2026        self._queued_trees.append((path, stream_name, max_manifest_depth))
2027
2028    def write_file(self, source, filename=None):
2029        self._queue_file(source, filename)
2030        self.do_queued_work()
2031
2032    def write_directory_tree(self,
2033                             path, stream_name='.', max_manifest_depth=-1):
2034        self._queue_tree(path, stream_name, max_manifest_depth)
2035        self.do_queued_work()
2036
2037    def write(self, newdata):
2038        if isinstance(newdata, bytes):
2039            pass
2040        elif isinstance(newdata, str):
2041            newdata = newdata.encode()
2042        elif hasattr(newdata, '__iter__'):
2043            for s in newdata:
2044                self.write(s)
2045            return
2046        self._data_buffer.append(newdata)
2047        self._data_buffer_len += len(newdata)
2048        self._current_stream_length += len(newdata)
2049        while self._data_buffer_len >= config.KEEP_BLOCK_SIZE:
2050            self.flush_data()
2051
2052    def open(self, streampath, filename=None):
2053        """open(streampath[, filename]) -> file-like object
2054
2055        Pass in the path of a file to write to the Collection, either as a
2056        single string or as two separate stream name and file name arguments.
2057        This method returns a file-like object you can write to add it to the
2058        Collection.
2059
2060        You may only have one file object from the Collection open at a time,
2061        so be sure to close the object when you're done.  Using the object in
2062        a with statement makes that easy:
2063
2064            with cwriter.open('./doc/page1.txt') as outfile:
2065                outfile.write(page1_data)
2066            with cwriter.open('./doc/page2.txt') as outfile:
2067                outfile.write(page2_data)
2068        """
2069        if filename is None:
2070            streampath, filename = split(streampath)
2071        if self._last_open and not self._last_open.closed:
2072            raise errors.AssertionError(
2073                u"can't open '{}' when '{}' is still open".format(
2074                    filename, self._last_open.name))
2075        if streampath != self.current_stream_name():
2076            self.start_new_stream(streampath)
2077        self.set_current_file_name(filename)
2078        self._last_open = _WriterFile(self, filename)
2079        return self._last_open
2080
2081    def flush_data(self):
2082        data_buffer = b''.join(self._data_buffer)
2083        if data_buffer:
2084            self._current_stream_locators.append(
2085                self._my_keep().put(
2086                    data_buffer[0:config.KEEP_BLOCK_SIZE],
2087                    copies=self.replication))
2088            self._data_buffer = [data_buffer[config.KEEP_BLOCK_SIZE:]]
2089            self._data_buffer_len = len(self._data_buffer[0])
2090
2091    def start_new_file(self, newfilename=None):
2092        self.finish_current_file()
2093        self.set_current_file_name(newfilename)
2094
2095    def set_current_file_name(self, newfilename):
2096        if re.search(r'[\t\n]', newfilename):
2097            raise errors.AssertionError(
2098                "Manifest filenames cannot contain whitespace: %s" %
2099                newfilename)
2100        elif re.search(r'\x00', newfilename):
2101            raise errors.AssertionError(
2102                "Manifest filenames cannot contain NUL characters: %s" %
2103                newfilename)
2104        self._current_file_name = newfilename
2105
2106    def current_file_name(self):
2107        return self._current_file_name
2108
2109    def finish_current_file(self):
2110        if self._current_file_name is None:
2111            if self._current_file_pos == self._current_stream_length:
2112                return
2113            raise errors.AssertionError(
2114                "Cannot finish an unnamed file " +
2115                "(%d bytes at offset %d in '%s' stream)" %
2116                (self._current_stream_length - self._current_file_pos,
2117                 self._current_file_pos,
2118                 self._current_stream_name))
2119        self._current_stream_files.append([
2120                self._current_file_pos,
2121                self._current_stream_length - self._current_file_pos,
2122                self._current_file_name])
2123        self._current_file_pos = self._current_stream_length
2124        self._current_file_name = None
2125
2126    def start_new_stream(self, newstreamname='.'):
2127        self.finish_current_stream()
2128        self.set_current_stream_name(newstreamname)
2129
2130    def set_current_stream_name(self, newstreamname):
2131        if re.search(r'[\t\n]', newstreamname):
2132            raise errors.AssertionError(
2133                "Manifest stream names cannot contain whitespace: '%s'" %
2134                (newstreamname))
2135        self._current_stream_name = '.' if newstreamname=='' else newstreamname
2136
2137    def current_stream_name(self):
2138        return self._current_stream_name
2139
2140    def finish_current_stream(self):
2141        self.finish_current_file()
2142        self.flush_data()
2143        if not self._current_stream_files:
2144            pass
2145        elif self._current_stream_name is None:
2146            raise errors.AssertionError(
2147                "Cannot finish an unnamed stream (%d bytes in %d files)" %
2148                (self._current_stream_length, len(self._current_stream_files)))
2149        else:
2150            if not self._current_stream_locators:
2151                self._current_stream_locators.append(config.EMPTY_BLOCK_LOCATOR)
2152            self._finished_streams.append([self._current_stream_name,
2153                                           self._current_stream_locators,
2154                                           self._current_stream_files])
2155        self._current_stream_files = []
2156        self._current_stream_length = 0
2157        self._current_stream_locators = []
2158        self._current_stream_name = None
2159        self._current_file_pos = 0
2160        self._current_file_name = None
2161
2162    def finish(self):
2163        """Store the manifest in Keep and return its locator.
2164
2165        This is useful for storing manifest fragments (task outputs)
2166        temporarily in Keep during a Crunch job.
2167
2168        In other cases you should make a collection instead, by
2169        sending manifest_text() to the API server's "create
2170        collection" endpoint.
2171        """
2172        return self._my_keep().put(self.manifest_text().encode(),
2173                                   copies=self.replication)
2174
2175    def portable_data_hash(self):
2176        stripped = self.stripped_manifest().encode()
2177        return '{}+{}'.format(hashlib.md5(stripped).hexdigest(), len(stripped))
2178
2179    def manifest_text(self):
2180        self.finish_current_stream()
2181        manifest = ''
2182
2183        for stream in self._finished_streams:
2184            if not re.search(r'^\.(/.*)?$', stream[0]):
2185                manifest += './'
2186            manifest += stream[0].replace(' ', '\\040')
2187            manifest += ' ' + ' '.join(stream[1])
2188            manifest += ' ' + ' '.join("%d:%d:%s" % (sfile[0], sfile[1], sfile[2].replace(' ', '\\040')) for sfile in stream[2])
2189            manifest += "\n"
2190
2191        return manifest
2192
2193    def data_locators(self):
2194        ret = []
2195        for name, locators, files in self._finished_streams:
2196            ret += locators
2197        return ret
2198
2199    def save_new(self, name=None):
2200        return self._api_client.collections().create(
2201            ensure_unique_name=True,
2202            body={
2203                'name': name,
2204                'manifest_text': self.manifest_text(),
2205            }).execute(num_retries=self.num_retries)
2206
2207
2208class ResumableCollectionWriter(CollectionWriter):
2209    """CollectionWriter that can serialize internal state to disk
2210
2211    .. WARNING:: Deprecated
2212       This class is deprecated. Prefer `arvados.collection.Collection`
2213       instead.
2214    """
2215
2216    STATE_PROPS = ['_current_stream_files', '_current_stream_length',
2217                   '_current_stream_locators', '_current_stream_name',
2218                   '_current_file_name', '_current_file_pos', '_close_file',
2219                   '_data_buffer', '_dependencies', '_finished_streams',
2220                   '_queued_dirents', '_queued_trees']
2221
2222    @arvados.util._deprecated('3.0', 'arvados.collection.Collection')
2223    def __init__(self, api_client=None, **kwargs):
2224        self._dependencies = {}
2225        super(ResumableCollectionWriter, self).__init__(api_client, **kwargs)
2226
2227    @classmethod
2228    def from_state(cls, state, *init_args, **init_kwargs):
2229        # Try to build a new writer from scratch with the given state.
2230        # If the state is not suitable to resume (because files have changed,
2231        # been deleted, aren't predictable, etc.), raise a
2232        # StaleWriterStateError.  Otherwise, return the initialized writer.
2233        # The caller is responsible for calling writer.do_queued_work()
2234        # appropriately after it's returned.
2235        writer = cls(*init_args, **init_kwargs)
2236        for attr_name in cls.STATE_PROPS:
2237            attr_value = state[attr_name]
2238            attr_class = getattr(writer, attr_name).__class__
2239            # Coerce the value into the same type as the initial value, if
2240            # needed.
2241            if attr_class not in (type(None), attr_value.__class__):
2242                attr_value = attr_class(attr_value)
2243            setattr(writer, attr_name, attr_value)
2244        # Check dependencies before we try to resume anything.
2245        if any(KeepLocator(ls).permission_expired()
2246               for ls in writer._current_stream_locators):
2247            raise errors.StaleWriterStateError(
2248                "locators include expired permission hint")
2249        writer.check_dependencies()
2250        if state['_current_file'] is not None:
2251            path, pos = state['_current_file']
2252            try:
2253                writer._queued_file = open(path, 'rb')
2254                writer._queued_file.seek(pos)
2255            except IOError as error:
2256                raise errors.StaleWriterStateError(
2257                    u"failed to reopen active file {}: {}".format(path, error))
2258        return writer
2259
2260    def check_dependencies(self):
2261        for path, orig_stat in listitems(self._dependencies):
2262            if not S_ISREG(orig_stat[ST_MODE]):
2263                raise errors.StaleWriterStateError(u"{} not file".format(path))
2264            try:
2265                now_stat = tuple(os.stat(path))
2266            except OSError as error:
2267                raise errors.StaleWriterStateError(
2268                    u"failed to stat {}: {}".format(path, error))
2269            if ((not S_ISREG(now_stat[ST_MODE])) or
2270                (orig_stat[ST_MTIME] != now_stat[ST_MTIME]) or
2271                (orig_stat[ST_SIZE] != now_stat[ST_SIZE])):
2272                raise errors.StaleWriterStateError(u"{} changed".format(path))
2273
2274    def dump_state(self, copy_func=lambda x: x):
2275        state = {attr: copy_func(getattr(self, attr))
2276                 for attr in self.STATE_PROPS}
2277        if self._queued_file is None:
2278            state['_current_file'] = None
2279        else:
2280            state['_current_file'] = (os.path.realpath(self._queued_file.name),
2281                                      self._queued_file.tell())
2282        return state
2283
2284    def _queue_file(self, source, filename=None):
2285        try:
2286            src_path = os.path.realpath(source)
2287        except Exception:
2288            raise errors.AssertionError(u"{} not a file path".format(source))
2289        try:
2290            path_stat = os.stat(src_path)
2291        except OSError as stat_error:
2292            path_stat = None
2293        super(ResumableCollectionWriter, self)._queue_file(source, filename)
2294        fd_stat = os.fstat(self._queued_file.fileno())
2295        if not S_ISREG(fd_stat.st_mode):
2296            # We won't be able to resume from this cache anyway, so don't
2297            # worry about further checks.
2298            self._dependencies[source] = tuple(fd_stat)
2299        elif path_stat is None:
2300            raise errors.AssertionError(
2301                u"could not stat {}: {}".format(source, stat_error))
2302        elif path_stat.st_ino != fd_stat.st_ino:
2303            raise errors.AssertionError(
2304                u"{} changed between open and stat calls".format(source))
2305        else:
2306            self._dependencies[src_path] = tuple(fd_stat)
2307
2308    def write(self, data):
2309        if self._queued_file is None:
2310            raise errors.AssertionError(
2311                "resumable writer can't accept unsourced data")
2312        return super(ResumableCollectionWriter, self).write(data)