+# Copyright (C) The Arvados Authors. All rights reserved.
+#
+# SPDX-License-Identifier: Apache-2.0
+
+from __future__ import absolute_import
+from future.utils import listitems, listvalues, viewkeys
+from builtins import str
+from past.builtins import basestring
+from builtins import object
import functools
import logging
import os
from stat import *
from .arvfile import split, _FileLikeObjectBase, ArvadosFile, ArvadosFileWriter, ArvadosFileReader, _BlockManager, synchronized, must_be_writable, NoopLock
-from keep import KeepLocator, KeepClient
+from .keep import KeepLocator, KeepClient
from .stream import StreamReader
from ._normalize_stream import normalize_stream
from ._ranges import Range, LocatorAndRange
from .safeapi import ThreadSafeApiCache
-import config
-import errors
-import util
-import events
+import arvados.config as config
+import arvados.errors as errors
+import arvados.util
+import arvados.events as events
from arvados.retry import retry_method
_logger = logging.getLogger('arvados.collection')
class CollectionBase(object):
+ """Abstract base class for Collection classes."""
+
def __enter__(self):
return self
if fields:
clean_fields = fields[:1] + [
(re.sub(r'\+[^\d][^\+]*', '', x)
- if re.match(util.keep_locator_pattern, x)
+ if re.match(arvados.util.keep_locator_pattern, x)
else x)
for x in fields[1:]]
clean += [' '.join(clean_fields), "\n"]
class CollectionWriter(CollectionBase):
+ """Deprecated, use Collection instead."""
+
def __init__(self, api_client=None, num_retries=0, replication=None):
"""Instantiate a CollectionWriter.
def _work_trees(self):
path, stream_name, max_manifest_depth = self._queued_trees[0]
- d = util.listdir_recursive(
+ d = arvados.util.listdir_recursive(
path, max_depth = (None if max_manifest_depth == 0 else 0))
if d:
self._queue_dirents(stream_name, d)
self.do_queued_work()
def write(self, newdata):
- if hasattr(newdata, '__iter__'):
+ if isinstance(newdata, bytes):
+ pass
+ elif isinstance(newdata, str):
+ newdata = newdata.encode()
+ elif hasattr(newdata, '__iter__'):
for s in newdata:
self.write(s)
return
return self._last_open
def flush_data(self):
- data_buffer = ''.join(self._data_buffer)
+ data_buffer = b''.join(self._data_buffer)
if data_buffer:
self._current_stream_locators.append(
self._my_keep().put(
sending manifest_text() to the API server's "create
collection" endpoint.
"""
- return self._my_keep().put(self.manifest_text(), copies=self.replication)
+ return self._my_keep().put(self.manifest_text().encode(),
+ copies=self.replication)
def portable_data_hash(self):
- stripped = self.stripped_manifest()
- return hashlib.md5(stripped).hexdigest() + '+' + str(len(stripped))
+ stripped = self.stripped_manifest().encode()
+ return '{}+{}'.format(hashlib.md5(stripped).hexdigest(), len(stripped))
def manifest_text(self):
self.finish_current_stream()
ret += locators
return ret
+ def save_new(self, name=None):
+ return self._api_client.collections().create(
+ ensure_unique_name=True,
+ body={
+ 'name': name,
+ 'manifest_text': self.manifest_text(),
+ }).execute(num_retries=self.num_retries)
+
class ResumableCollectionWriter(CollectionWriter):
+ """Deprecated, use Collection instead."""
+
STATE_PROPS = ['_current_stream_files', '_current_stream_length',
'_current_stream_locators', '_current_stream_name',
'_current_file_name', '_current_file_pos', '_close_file',
return writer
def check_dependencies(self):
- for path, orig_stat in self._dependencies.items():
+ for path, orig_stat in listitems(self._dependencies):
if not S_ISREG(orig_stat[ST_MODE]):
raise errors.StaleWriterStateError("{} not file".format(path))
try:
else:
item = ArvadosFile(self, pathcomponents[0])
self._items[pathcomponents[0]] = item
- self._committed = False
+ self.set_committed(False)
self.notify(ADD, self, pathcomponents[0], item)
return item
else:
# create new collection
item = Subcollection(self, pathcomponents[0])
self._items[pathcomponents[0]] = item
- self._committed = False
+ self.set_committed(False)
self.notify(ADD, self, pathcomponents[0], item)
if isinstance(item, RichCollectionBase):
return item.find_or_create(pathcomponents[1], create_type)
:path:
path to a file in the collection
:mode:
- one of "r", "r+", "w", "w+", "a", "a+"
+ a string consisting of "r", "w", or "a", optionally followed
+ by "b" or "t", optionally followed by "+".
+ :"b":
+ binary mode: write() accepts bytes, read() returns bytes.
+ :"t":
+ text mode (default): write() accepts strings, read() returns strings.
:"r":
opens for reading
:"r+":
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 not re.search(r'^[rwa][bt]?\+?$', mode):
+ raise errors.ArgumentError("Invalid mode {!r}".format(mode))
- if create:
- arvfile = self.find_or_create(path, FILE)
- else:
+ if mode[0] == 'r' and '+' not in mode:
+ fclass = ArvadosFileReader
arvfile = self.find(path)
+ elif not self.writable():
+ raise IOError(errno.EROFS, "Collection is read only")
+ else:
+ fclass = ArvadosFileWriter
+ arvfile = self.find_or_create(path, FILE)
if arvfile is None:
raise IOError(errno.ENOENT, "File not found", path)
if not isinstance(arvfile, ArvadosFile):
raise IOError(errno.EISDIR, "Is a directory", path)
- if mode[0] == "w":
+ if mode[0] == 'w':
arvfile.truncate(0)
- name = os.path.basename(path)
-
- if mode == "r":
- return ArvadosFileReader(arvfile, num_retries=self.num_retries)
- else:
- return ArvadosFileWriter(arvfile, mode, num_retries=self.num_retries)
+ return fclass(arvfile, mode=mode, num_retries=self.num_retries)
def modified(self):
"""Determine if the collection has been modified since last commited."""
@synchronized
def committed(self):
"""Determine if the collection has been committed to the API server."""
-
- if self._committed is False:
- return False
- for v in self._items.values():
- if v.committed() is False:
- return False
- return True
+ return self._committed
@synchronized
- def set_committed(self):
- """Recursively set committed flag to True."""
- self._committed = True
- for k,v in self._items.items():
- v.set_committed()
+ def set_committed(self, value=True):
+ """Recursively set committed flag.
+
+ If value is True, set committed to be True for this and all children.
+
+ If value is False, set committed to be False for this and all parents.
+ """
+ if value == self._committed:
+ return
+ if value:
+ for k,v in listitems(self._items):
+ v.set_committed(True)
+ self._committed = True
+ else:
+ self._committed = False
+ if self.parent is not None:
+ self.parent.set_committed(False)
@synchronized
def __iter__(self):
"""Iterate over names of files and collections contained in this collection."""
- return iter(self._items.keys())
+ return iter(viewkeys(self._items))
@synchronized
def __getitem__(self, k):
def __delitem__(self, p):
"""Delete an item by name which is directly contained by this collection."""
del self._items[p]
- self._committed = False
+ self.set_committed(False)
self.notify(DEL, self, p, None)
@synchronized
@synchronized
def values(self):
"""Get a list of files and collection objects directly contained in this collection."""
- return self._items.values()
+ return listvalues(self._items)
@synchronized
def items(self):
"""Get a list of (name, object) tuples directly contained in this collection."""
- return self._items.items()
+ return listitems(self._items)
def exists(self, path):
"""Test if there is a file or collection at `path`."""
raise IOError(errno.ENOTEMPTY, "Directory not empty", path)
deleteditem = self._items[pathcomponents[0]]
del self._items[pathcomponents[0]]
- self._committed = False
+ self.set_committed(False)
self.notify(DEL, self, pathcomponents[0], deleteditem)
else:
item.remove(pathcomponents[1])
def _clonefrom(self, source):
- for k,v in source.items():
+ for k,v in listitems(source):
self._items[k] = v.clone(self, k)
def clone(self):
item = source_obj.clone(self, target_name)
self._items[target_name] = item
- self._committed = False
+ self.set_committed(False)
if modified_from:
self.notify(MOD, self, target_name, (modified_from, item))
"""
if changes:
- self._committed = False
+ self.set_committed(False)
for change in changes:
event_type = change[0]
path = change[1]
# then return API server's PDH response.
return self._portable_data_hash
else:
- stripped = self.portable_manifest_text()
- return hashlib.md5(stripped).hexdigest() + '+' + str(len(stripped))
+ stripped = self.portable_manifest_text().encode()
+ return '{}+{}'.format(hashlib.md5(stripped).hexdigest(), len(stripped))
@synchronized
def subscribe(self, callback):
@synchronized
def flush(self):
"""Flush bufferblocks to Keep."""
- for e in self.values():
+ for e in listvalues(self):
e.flush()
"""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).
+ An Arvados collection UUID, portable data hash, raw manifest
+ text, or (if creating an empty collection) None.
+
:parent:
the parent Collection, may be None.
self.events = None
if manifest_locator_or_text:
- if re.match(util.keep_locator_pattern, manifest_locator_or_text):
+ if re.match(arvados.util.keep_locator_pattern, manifest_locator_or_text):
self._manifest_locator = manifest_locator_or_text
- elif re.match(util.collection_uuid_pattern, manifest_locator_or_text):
+ elif re.match(arvados.util.collection_uuid_pattern, manifest_locator_or_text):
self._manifest_locator = manifest_locator_or_text
- elif re.match(util.manifest_pattern, manifest_locator_or_text):
+ elif re.match(arvados.util.manifest_pattern, manifest_locator_or_text):
self._manifest_text = manifest_locator_or_text
else:
raise errors.ArgumentError(
# it. If instantiation fails, we'll fall back to the except
# clause, just like any other Collection lookup
# failure. Return an exception, or None if successful.
- try:
- self._remember_api_response(self._my_api().collections().get(
- uuid=self._manifest_locator).execute(
- num_retries=self.num_retries))
- self._manifest_text = self._api_response['manifest_text']
- # If not overriden via kwargs, we should try to load the
- # replication_desired from the API server
- if self.replication_desired is None:
- self.replication_desired = self._api_response.get('replication_desired', None)
- return None
- except Exception as e:
- return e
-
- def _populate_from_keep(self):
- # Retrieve a manifest directly from Keep. This has a chance of
- # working if [a] the locator includes a permission signature
- # or [b] the Keep services are operating in world-readable
- # mode. Return an exception, or None if successful.
- try:
- self._manifest_text = self._my_keep().get(
- self._manifest_locator, num_retries=self.num_retries)
- except Exception as e:
- return e
+ self._remember_api_response(self._my_api().collections().get(
+ uuid=self._manifest_locator).execute(
+ num_retries=self.num_retries))
+ self._manifest_text = self._api_response['manifest_text']
+ self._portable_data_hash = self._api_response['portable_data_hash']
+ # If not overriden via kwargs, we should try to load the
+ # replication_desired from the API server
+ if self.replication_desired is None:
+ self.replication_desired = self._api_response.get('replication_desired', None)
def _populate(self):
- if self._manifest_locator is None and self._manifest_text is None:
- return
- error_via_api = None
- error_via_keep = None
- should_try_keep = ((self._manifest_text is None) and
- util.keep_locator_pattern.match(
- self._manifest_locator))
- if ((self._manifest_text is None) and
- util.signed_locator_pattern.match(self._manifest_locator)):
- error_via_keep = self._populate_from_keep()
- if self._manifest_text is None:
- error_via_api = self._populate_from_api_server()
- if error_via_api is not None and not should_try_keep:
- raise error_via_api
- if ((self._manifest_text is None) and
- not error_via_keep and
- should_try_keep):
- # Looks like a keep locator, and we didn't already try keep above
- error_via_keep = self._populate_from_keep()
if self._manifest_text is None:
- # Nothing worked!
- raise errors.NotFoundError(
- ("Failed to retrieve collection '{}' " +
- "from either API server ({}) or Keep ({})."
- ).format(
- self._manifest_locator,
- error_via_api,
- error_via_keep))
- # populate
+ if self._manifest_locator is None:
+ return
+ else:
+ self._populate_from_api_server()
self._baseline_manifest = self._manifest_text
self._import_manifest(self._manifest_text)
-
def _has_collection_uuid(self):
- return self._manifest_locator is not None and re.match(util.collection_uuid_pattern, self._manifest_locator)
+ return self._manifest_locator is not None and re.match(arvados.util.collection_uuid_pattern, self._manifest_locator)
def __enter__(self):
return self
@must_be_writable
@synchronized
@retry_method
- def save(self, merge=True, num_retries=None):
+ def save(self, storage_classes=None, merge=True, num_retries=None):
"""Save collection to an existing collection record.
Commit pending buffer blocks to Keep, merge with remote record (if
self.update()
text = self.manifest_text(strip=False)
+ body={'manifest_text': text}
+ if storage_classes:
+ body["storage_classes_desired"] = storage_classes
+
self._remember_api_response(self._my_api().collections().update(
uuid=self._manifest_locator,
- body={'manifest_text': text}
+ body=body
).execute(
num_retries=num_retries))
self._manifest_text = self._api_response["manifest_text"]
self._portable_data_hash = self._api_response["portable_data_hash"]
- self.set_committed()
+ self.set_committed(True)
return self._manifest_text
def save_new(self, name=None,
create_collection_record=True,
owner_uuid=None,
+ storage_classes=None,
ensure_unique_name=False,
num_retries=None):
"""Save collection to a new collection record.
"replication_desired": self.replication_desired}
if owner_uuid:
body["owner_uuid"] = owner_uuid
+ if storage_classes:
+ body["storage_classes_desired"] = storage_classes
self._remember_api_response(self._my_api().collections().create(ensure_unique_name=ensure_unique_name, body=body).execute(num_retries=num_retries))
text = self._api_response["manifest_text"]
self._portable_data_hash = self._api_response["portable_data_hash"]
self._manifest_text = text
- self.set_committed()
+ self.set_committed(True)
return text
+ _token_re = re.compile(r'(\S+)(\s+|$)')
+ _block_re = re.compile(r'[0-9a-f]{32}\+(\d+)(\+\S+)*')
+ _segment_re = re.compile(r'(\d+):(\d+):(\S+)')
+
@synchronized
def _import_manifest(self, manifest_text):
"""Import a manifest into a `Collection`.
stream_name = None
state = STREAM_NAME
- for token_and_separator in re.finditer(r'(\S+)(\s+|$)', manifest_text):
+ for token_and_separator in self._token_re.finditer(manifest_text):
tok = token_and_separator.group(1)
sep = token_and_separator.group(2)
stream_name = tok.replace('\\040', ' ')
blocks = []
segments = []
- streamoffset = 0L
+ streamoffset = 0
state = BLOCKS
self.find_or_create(stream_name, COLLECTION)
continue
if state == BLOCKS:
- block_locator = re.match(r'[0-9a-f]{32}\+(\d+)(\+\S+)*', tok)
+ block_locator = self._block_re.match(tok)
if block_locator:
- blocksize = long(block_locator.group(1))
+ blocksize = int(block_locator.group(1))
blocks.append(Range(tok, streamoffset, blocksize, 0))
streamoffset += blocksize
else:
state = SEGMENTS
if state == SEGMENTS:
- file_segment = re.search(r'^(\d+):(\d+):(\S+)', tok)
+ file_segment = self._segment_re.match(tok)
if file_segment:
- pos = long(file_segment.group(1))
- size = long(file_segment.group(2))
+ pos = int(file_segment.group(1))
+ size = int(file_segment.group(2))
name = file_segment.group(3).replace('\\040', ' ')
filepath = os.path.join(stream_name, name)
afile = self.find_or_create(filepath, FILE)
stream_name = None
state = STREAM_NAME
- self.set_committed()
+ self.set_committed(True)
@synchronized
def notify(self, event, collection, name, item):
@must_be_writable
@synchronized
def _reparent(self, newparent, newname):
- self._committed = False
+ self.set_committed(False)
self.flush()
self.parent.remove(self.name, recursive=True)
self.parent = newparent
class CollectionReader(Collection):
"""A read-only collection object.
- Initialize from an api collection record locator, a portable data hash of a
- manifest, or raw manifest text. See `Collection` constructor for detailed
- options.
+ Initialize from a collection UUID or portable data hash, or raw
+ manifest text. See `Collection` constructor for detailed options.
"""
def __init__(self, manifest_locator_or_text, *args, **kwargs):
projects / arvados.git / blobdiff