+ def writable(self):
+ raise NotImplementedError()
+
+ def root_collection(self):
+ raise NotImplementedError()
+
+ def notify(self, event, collection, name, item):
+ raise NotImplementedError()
+
+ def stream_name(self):
+ raise NotImplementedError()
+
+ @must_be_writable
+ @synchronized
+ def find_or_create(self, path, create_type):
+ """Recursively search the specified file path.
+
+ May return either a `Collection` or `ArvadosFile`. If not found, will
+ create a new item at the specified path based on `create_type`. Will
+ create intermediate subcollections needed to contain the final item in
+ the path.
+
+ :create_type:
+ One of `arvados.collection.FILE` or
+ `arvados.collection.COLLECTION`. If the path is not found, and value
+ of create_type is FILE then create and return a new ArvadosFile for
+ the last path component. If COLLECTION, then create and return a new
+ Collection for the last path component.
+
+ """
+
+ pathcomponents = path.split("/", 1)
+ if pathcomponents[0]:
+ item = self._items.get(pathcomponents[0])
+ if len(pathcomponents) == 1:
+ if item is None:
+ # create new file
+ if create_type == COLLECTION:
+ item = Subcollection(self)
+ else:
+ item = ArvadosFile(self)
+ self._items[pathcomponents[0]] = item
+ self._modified = True
+ self.notify(ADD, self, pathcomponents[0], item)
+ return item
+ else:
+ if item is None:
+ # create new collection
+ item = Subcollection(self)
+ self._items[pathcomponents[0]] = item
+ self._modified = True
+ self.notify(ADD, self, pathcomponents[0], item)
+ if isinstance(item, RichCollectionBase):
+ return item.find_or_create(pathcomponents[1], create_type)
+ else:
+ raise IOError((errno.ENOTDIR, "Interior path components must be subcollection"))
+ else:
+ return self
+
+ @synchronized
+ def find(self, path):
+ """Recursively search the specified file path.
+
+ May return either a Collection or ArvadosFile. Return None if not
+ found.
+
+ """
+ if not path:
+ raise errors.ArgumentError("Parameter 'path' must not be empty.")
+
+ pathcomponents = path.split("/", 1)
+ item = self._items.get(pathcomponents[0])
+ if len(pathcomponents) == 1:
+ return item
+ else:
+ if isinstance(item, RichCollectionBase):
+ if pathcomponents[1]:
+ return item.find(pathcomponents[1])
+ else:
+ return item
+ else:
+ raise IOError((errno.ENOTDIR, "Interior path components must be subcollection"))
+
+ def mkdirs(path):
+ """Recursive subcollection create.
+
+ Like `os.mkdirs()`. Will create intermediate subcollections needed to
+ contain the leaf subcollection path.
+
+ """
+ return self.find_or_create(path, COLLECTION)
+
+ def open(self, path, mode="r"):
+ """Open a file-like object for access.
+
+ :path:
+ path to a file in the collection
+ :mode:
+ one of "r", "r+", "w", "w+", "a", "a+"
+ :"r":
+ opens for reading
+ :"r+":
+ opens for reading and writing. Reads/writes share a file pointer.
+ :"w", "w+":
+ truncates to 0 and opens for reading and writing. Reads/writes share a file pointer.
+ :"a", "a+":
+ opens for reading and writing. All writes are appended to
+ the end of the file. Writing does not affect the file pointer for
+ reading.
+ """
+ mode = mode.replace("b", "")
+ if len(mode) == 0 or mode[0] not in ("r", "w", "a"):
+ raise errors.ArgumentError("Bad mode '%s'" % mode)
+ create = (mode != "r")
+
+ if create and not self.writable():
+ raise IOError((errno.EROFS, "Collection is read only"))
+
+ if create:
+ arvfile = self.find_or_create(path, FILE)
+ else:
+ arvfile = self.find(path)
+
+ if arvfile is None:
+ raise IOError((errno.ENOENT, "File not found"))
+ if not isinstance(arvfile, ArvadosFile):
+ raise IOError((errno.EISDIR, "Path must refer to a file."))
+
+ if mode[0] == "w":
+ arvfile.truncate(0)
+
+ name = os.path.basename(path)
+
+ if mode == "r":
+ return ArvadosFileReader(arvfile, name, mode, num_retries=self.num_retries)
+ else:
+ return ArvadosFileWriter(arvfile, name, mode, num_retries=self.num_retries)
+
+ @synchronized
+ def modified(self):
+ """Test if the collection (or any subcollection or file) has been modified."""
+ if self._modified:
+ return True
+ for k,v in self._items.items():
+ if v.modified():
+ return True
+ return False
+
+ @synchronized
+ def set_unmodified(self):
+ """Recursively clear modified flag."""
+ self._modified = False
+ for k,v in self._items.items():
+ v.set_unmodified()
+
+ @synchronized
+ def __iter__(self):
+ """Iterate over names of files and collections contained in this collection."""
+ return iter(self._items.keys())
+
+ @synchronized
+ def __getitem__(self, k):
+ """Get a file or collection that is directly contained by this collection.
+
+ If you want to search a path, use `find()` instead.
+
+ """
+ return self._items[k]
+
+ @synchronized
+ def __contains__(self, k):
+ """Test if there is a file or collection a directly contained by this collection."""
+ return k in self._items
+
+ @synchronized
+ def __len__(self):
+ """Get the number of items directly contained in this collection."""
+ return len(self._items)
+
+ @must_be_writable
+ @synchronized
+ def __delitem__(self, p):
+ """Delete an item by name which is directly contained by this collection."""
+ del self._items[p]
+ self._modified = True
+ self.notify(DEL, self, p, None)
+
+ @synchronized
+ def keys(self):
+ """Get a list of names of files and collections directly contained in this collection."""
+ return self._items.keys()
+
+ @synchronized
+ def values(self):
+ """Get a list of files and collection objects directly contained in this collection."""
+ return self._items.values()
+
+ @synchronized
+ def items(self):
+ """Get a list of (name, object) tuples directly contained in this collection."""
+ return self._items.items()
+
+ def exists(self, path):
+ """Test if there is a file or collection at `path`."""
+ return self.find(path) is not None
+
+ @must_be_writable
+ @synchronized
+ def remove(self, path, recursive=False):
+ """Remove the file or subcollection (directory) at `path`.
+
+ :recursive:
+ Specify whether to remove non-empty subcollections (True), or raise an error (False).
+ """
+
+ if not path:
+ raise errors.ArgumentError("Parameter 'path' must not be empty.")
+
+ pathcomponents = path.split("/", 1)
+ item = self._items.get(pathcomponents[0])
+ if item is None:
+ raise IOError((errno.ENOENT, "File not found"))
+ if len(pathcomponents) == 1:
+ if isinstance(self._items[pathcomponents[0]], RichCollectionBase) and len(self._items[pathcomponents[0]]) > 0 and not recursive:
+ raise IOError((errno.ENOTEMPTY, "Subcollection not empty"))
+ deleteditem = self._items[pathcomponents[0]]
+ del self._items[pathcomponents[0]]
+ self._modified = True
+ self.notify(DEL, self, pathcomponents[0], deleteditem)
+ else:
+ item.remove(pathcomponents[1])
+
+ def _clonefrom(self, source):
+ for k,v in source.items():
+ self._items[k] = v.clone(self)
+
+ def clone(self):
+ raise NotImplementedError()
+
+ @must_be_writable
+ @synchronized
+ def add(self, source_obj, target_name, overwrite=False):
+ """Copy a file or subcollection to this collection.
+
+ :source_obj:
+ An ArvadosFile, or Subcollection object
+
+ :target_name:
+ Destination item name. If the target name already exists and is a
+ file, this will raise an error unless you specify `overwrite=True`.
+
+ :overwrite:
+ Whether to overwrite target file if it already exists.
+
+ """
+
+ if target_name in self and not overwrite:
+ raise IOError((errno.EEXIST, "File already exists"))
+
+ modified_from = None
+ if target_name in self:
+ modified_from = self[target_name]
+
+ # Actually make the copy.
+ dup = source_obj.clone(self)
+ self._items[target_name] = dup
+ self._modified = True
+
+ if modified_from:
+ self.notify(MOD, self, target_name, (modified_from, dup))
+ else:
+ self.notify(ADD, self, target_name, dup)
+
+ @must_be_writable
+ @synchronized
+ def copy(self, source, target_path, source_collection=None, overwrite=False):
+ """Copy a file or subcollection to a new path in this collection.
+
+ :source:
+ A string with a path to source file or subcollection, or an actual ArvadosFile or Subcollection object.
+
+ :target_path:
+ Destination file or path. If the target path already exists and is a
+ subcollection, the item will be placed inside the subcollection. If
+ the target path already exists and is a file, this will raise an error
+ unless you specify `overwrite=True`.
+
+ :source_collection:
+ Collection to copy `source_path` from (default `self`)
+
+ :overwrite:
+ Whether to overwrite target file if it already exists.
+ """
+ if source_collection is None:
+ source_collection = self
+
+ # Find the object to copy
+ if isinstance(source, basestring):
+ source_obj = source_collection.find(source)
+ if source_obj is None:
+ raise IOError((errno.ENOENT, "File not found"))
+ sourcecomponents = source.split("/")
+ else:
+ source_obj = source
+ sourcecomponents = None
+
+ # Find parent collection the target path
+ targetcomponents = target_path.split("/")
+
+ # Determine the name to use.
+ target_name = targetcomponents[-1] if targetcomponents[-1] else (sourcecomponents[-1] if sourcecomponents else None)
+
+ if not target_name:
+ raise errors.ArgumentError("Target path is empty and source is an object. Cannot determine destination filename to use.")
+
+ target_dir = self.find_or_create("/".join(targetcomponents[0:-1]), COLLECTION)
+
+ if target_name in target_dir and isinstance(self[target_name], RichCollectionBase) and sourcecomponents:
+ target_dir = target_dir[target_name]
+ target_name = sourcecomponents[-1]
+
+ target_dir.add(source_obj, target_name, overwrite)
+
+ @synchronized
+ def manifest_text(self, stream_name=".", strip=False, normalize=False):
+ """Get the manifest text for this collection, sub collections and files.
+
+ :stream_name:
+ Name of the stream (directory)
+
+ :strip:
+ If True, remove signing tokens from block locators if present.
+ If False (default), block locators are left unchanged.
+
+ :normalize:
+ If True, always export the manifest text in normalized form
+ even if the Collection is not modified. If False (default) and the collection
+ is not modified, return the original manifest text even if it is not
+ in normalized form.
+
+ """
+
+ if self.modified() or self._manifest_text is None or normalize:
+ stream = {}
+ buf = []
+ sorted_keys = sorted(self.keys())
+ for filename in [s for s in sorted_keys if isinstance(self[s], ArvadosFile)]:
+ # Create a stream per file `k`
+ arvfile = self[filename]
+ filestream = []
+ for segment in arvfile.segments():
+ loc = segment.locator
+ if arvfile.parent._my_block_manager().is_bufferblock(loc):
+ loc = arvfile.parent._my_block_manager().get_bufferblock(loc).locator()
+ if strip:
+ loc = KeepLocator(loc).stripped()
+ filestream.append(LocatorAndRange(loc, KeepLocator(loc).size,
+ segment.segment_offset, segment.range_size))
+ stream[filename] = filestream
+ if stream:
+ buf.append(" ".join(normalize_stream(stream_name, stream)) + "\n")
+ for dirname in [s for s in sorted_keys if isinstance(self[s], RichCollectionBase)]:
+ buf.append(self[dirname].manifest_text(stream_name=os.path.join(stream_name, dirname), strip=strip))
+ return "".join(buf)
+ else:
+ if strip:
+ return self.stripped_manifest()
+ else:
+ return self._manifest_text
+
+ @synchronized
+ def diff(self, end_collection, prefix=".", holding_collection=None):
+ """Generate list of add/modify/delete actions.
+
+ When given to `apply`, will change `self` to match `end_collection`
+
+ """
+ changes = []
+ if holding_collection is None:
+ holding_collection = Collection(api_client=self._my_api(), keep_client=self._my_keep())
+ for k in self:
+ if k not in end_collection:
+ changes.append((DEL, os.path.join(prefix, k), self[k].clone(holding_collection)))
+ for k in end_collection:
+ if k in self:
+ if isinstance(end_collection[k], Subcollection) and isinstance(self[k], Subcollection):
+ changes.extend(self[k].diff(end_collection[k], os.path.join(prefix, k), holding_collection))
+ elif end_collection[k] != self[k]:
+ changes.append((MOD, os.path.join(prefix, k), self[k].clone(holding_collection), end_collection[k].clone(holding_collection)))
+ else:
+ changes.append((ADD, os.path.join(prefix, k), end_collection[k].clone(holding_collection)))
+ return changes
+
+ @must_be_writable
+ @synchronized
+ def apply(self, changes):
+ """Apply changes from `diff`.
+
+ If a change conflicts with a local change, it will be saved to an
+ alternate path indicating the conflict.
+
+ """
+ for change in changes:
+ event_type = change[0]
+ path = change[1]
+ initial = change[2]
+ local = self.find(path)
+ conflictpath = "%s~conflict-%s~" % (path, time.strftime("%Y-%m-%d-%H:%M:%S",
+ time.gmtime()))
+ if event_type == ADD:
+ if local is None:
+ # No local file at path, safe to copy over new file
+ self.copy(initial, path)
+ elif local is not None and local != initial:
+ # There is already local file and it is different:
+ # save change to conflict file.
+ self.copy(initial, conflictpath)
+ elif event_type == MOD:
+ final = change[3]
+ if local == initial:
+ # Local matches the "initial" item so it has not
+ # changed locally and is safe to update.
+ if isinstance(local, ArvadosFile) and isinstance(final, ArvadosFile):
+ # Replace contents of local file with new contents
+ local.replace_contents(final)
+ else:
+ # Overwrite path with new item; this can happen if
+ # path was a file and is now a collection or vice versa
+ self.copy(final, path, overwrite=True)
+ else:
+ # Local is missing (presumably deleted) or local doesn't
+ # match the "start" value, so save change to conflict file
+ self.copy(final, conflictpath)
+ elif event_type == DEL:
+ if local == initial:
+ # Local item matches "initial" value, so it is safe to remove.
+ self.remove(path, recursive=True)
+ # else, the file is modified or already removed, in either
+ # case we don't want to try to remove it.
+
+ def portable_data_hash(self):
+ """Get the portable data hash for this collection's manifest."""
+ stripped = self.manifest_text(strip=True)
+ return hashlib.md5(stripped).hexdigest() + '+' + str(len(stripped))
+
+ @synchronized
+ def __eq__(self, other):
+ if other is self:
+ return True
+ if not isinstance(other, RichCollectionBase):
+ return False
+ if len(self._items) != len(other):
+ return False
+ for k in self._items:
+ if k not in other:
+ return False
+ if self._items[k] != other[k]:
+ return False
+ return True
+
+ def __ne__(self, other):
+ return not self.__eq__(other)
+
+
+class Collection(RichCollectionBase):
+ """Represents the root of an Arvados Collection.
+
+ This class is threadsafe. The root collection object, all subcollections
+ and files are protected by a single lock (i.e. each access locks the entire
+ collection).
+
+ Brief summary of
+ useful methods:
+
+ :To read an existing file:
+ `c.open("myfile", "r")`
+
+ :To write a new file:
+ `c.open("myfile", "w")`
+
+ :To determine if a file exists:
+ `c.find("myfile") is not None`
+
+ :To copy a file:
+ `c.copy("source", "dest")`
+
+ :To delete a file:
+ `c.remove("myfile")`
+
+ :To save to an existing collection record:
+ `c.save()`
+
+ :To save a new collection record:
+ `c.save_new()`
+
+ :To merge remote changes into this object:
+ `c.update()`
+
+ Must be associated with an API server Collection record (during
+ initialization, or using `save_new`) to use `save` or `update`
+
+ """
+
+ def __init__(self, manifest_locator_or_text=None,
+ api_client=None,
+ keep_client=None,
+ num_retries=None,
+ parent=None,
+ apiconfig=None,
+ block_manager=None):
+ """Collection constructor.
+
+ :manifest_locator_or_text:
+ One of Arvados collection UUID, block locator of
+ a manifest, raw manifest text, or None (to create an empty collection).
+ :parent:
+ the parent Collection, may be None.
+ :apiconfig:
+ A dict containing keys for ARVADOS_API_HOST and ARVADOS_API_TOKEN.
+ Prefer this over supplying your own api_client and keep_client (except in testing).
+ Will use default config settings if not specified.
+ :api_client:
+ The API client object to use for requests. If not specified, create one using `apiconfig`.
+ :keep_client:
+ the Keep client to use for requests. If not specified, create one using `apiconfig`.
+ :num_retries:
+ the number of retries for API and Keep requests.
+ :block_manager:
+ the block manager to use. If not specified, create one.
+
+ """
+ super(Collection, self).__init__(parent)
projects / arvados.git / blobdiff