def execute(self, write_concern):
"""Execute operations.
"""
if not self.ops:
raise InvalidOperation('No operations to execute')
if self.executed:
raise InvalidOperation('Bulk operations can '
'only be executed once.')
self.executed = True
write_concern = (WriteConcern(**write_concern)
if write_concern else self.collection.write_concern)
if self.ordered:
generator = self.gen_ordered()
else:
generator = self.gen_unordered()
client = self.collection.database.client
with client._socket_for_writes() as sock_info:
if sock_info.max_wire_version < 5 and self.uses_collation:
raise ConfigurationError(
'Must be connected to MongoDB 3.4+ to use a collation.')
if not write_concern.acknowledged:
if self.uses_collation:
raise ConfigurationError(
'Collation is unsupported for unacknowledged writes.')
self.execute_no_results(sock_info, generator)
elif sock_info.max_wire_version > 1:
return self.execute_command(sock_info, generator,
write_concern)
else:
return self.execute_legacy(sock_info, generator, write_concern)
def _raise_if_unacknowledged(self, property_name):
"""Raise an exception on property access if unacknowledged."""
if not self.__acknowledged:
raise InvalidOperation("A value for %s is not available when "
"the write is unacknowledged. Check the "
"acknowledged attribute to avoid this "
"error." % (property_name, ))
def add_option(self, mask):
"""Set arbitrary query flags using a bitmask.
To set the tailable flag:
cursor.add_option(2)
"""
if not isinstance(mask, int):
raise TypeError("mask must be an int")
self.__check_okay_to_chain()
if mask & _QUERY_OPTIONS["exhaust"]:
if self.__limit:
raise InvalidOperation("Can't use limit and exhaust together.")
if self.__collection.database.client.is_mongos:
raise InvalidOperation('Exhaust cursors are '
'not supported by mongos')
self.__exhaust = True
self.__query_flags |= mask
return self
def insert(collection_name, docs, check_keys,
safe, last_error_args, continue_on_error, opts):
"""Get an **insert** message."""
options = 0
if continue_on_error:
options += 1
data = struct.pack("<i", options)
data += bson._make_c_string(collection_name)
encoded = [bson.BSON.encode(doc, check_keys, opts) for doc in docs]
if not encoded:
raise InvalidOperation("cannot do an empty bulk insert")
max_bson_size = max(map(len, encoded))
data += _EMPTY.join(encoded)
if safe:
(_, insert_message) = __pack_message(2002, data)
(request_id, error_message, _) = __last_error(collection_name,
last_error_args)
return (request_id, insert_message + error_message, max_bson_size)
else:
(request_id, insert_message) = __pack_message(2002, data)
return (request_id, insert_message, max_bson_size)
def limit(self, limit):
"""Limits the number of results to be returned by this cursor.
Raises :exc:`TypeError` if `limit` is not an integer. Raises
:exc:`~pymongo.errors.InvalidOperation` if this :class:`Cursor`
has already been used. The last `limit` applied to this cursor
takes precedence. A limit of ``0`` is equivalent to no limit.
:Parameters:
- `limit`: the number of results to return
.. mongodoc:: limit
"""
if not isinstance(limit, integer_types):
raise TypeError("limit must be an integer")
if self.__exhaust:
raise InvalidOperation("Can't use limit and exhaust together.")
self.__check_okay_to_chain()
self.__empty = False
self.__limit = limit
return self
def __check_okay_to_chain(self):
"""Check if it is okay to chain more options onto this cursor.
"""
if self.__retrieved or self.__id is not None:
raise InvalidOperation("cannot set options after executing query")
def __init__(self, collection, filter=None, projection=None, skip=0,
limit=0, no_cursor_timeout=False,
cursor_type=CursorType.NON_TAILABLE,
sort=None, allow_partial_results=False, oplog_replay=False,
modifiers=None, batch_size=0, manipulate=True,
collation=None, hint=None, max_scan=None, max_time_ms=None,
max=None, min=None, return_key=False, show_record_id=False,
snapshot=False, comment=None):
"""Create a new cursor.
Should not be called directly by application developers - see
:meth:`~pymongo.collection.Collection.find` instead.
.. mongodoc:: cursors
"""
self.__id = None
self.__exhaust = False
self.__exhaust_mgr = None
spec = filter
if spec is None:
spec = {}
validate_is_mapping("filter", spec)
if not isinstance(skip, int):
raise TypeError("skip must be an instance of int")
if not isinstance(limit, int):
raise TypeError("limit must be an instance of int")
validate_boolean("no_cursor_timeout", no_cursor_timeout)
if cursor_type not in (CursorType.NON_TAILABLE, CursorType.TAILABLE,
CursorType.TAILABLE_AWAIT, CursorType.EXHAUST):
raise ValueError("not a valid value for cursor_type")
validate_boolean("allow_partial_results", allow_partial_results)
validate_boolean("oplog_replay", oplog_replay)
if modifiers is not None:
warnings.warn("the 'modifiers' parameter is deprecated",
DeprecationWarning, stacklevel=2)
validate_is_mapping("modifiers", modifiers)
if not isinstance(batch_size, integer_types):
raise TypeError("batch_size must be an integer")
if batch_size < 0:
raise ValueError("batch_size must be >= 0")
if projection is not None:
if not projection:
projection = {"_id": 1}
projection = helpers._fields_list_to_dict(projection, "projection")
self.__collection = collection
self.__spec = spec
self.__projection = projection
self.__skip = skip
self.__limit = limit
self.__batch_size = batch_size
self.__modifiers = modifiers and modifiers.copy() or {}
self.__ordering = sort and helpers._index_document(sort) or None
self.__max_scan = max_scan
self.__explain = False
self.__comment = comment
self.__max_time_ms = max_time_ms
self.__max_await_time_ms = None
self.__max = max
self.__min = min
self.__manipulate = manipulate
self.__collation = validate_collation_or_none(collation)
self.__return_key = return_key
self.__show_record_id = show_record_id
self.__snapshot = snapshot
self.__set_hint(hint)
# Exhaust cursor support
if cursor_type == CursorType.EXHAUST:
if self.__collection.database.client.is_mongos:
raise InvalidOperation('Exhaust cursors are '
'not supported by mongos')
if limit:
raise InvalidOperation("Can't use limit and exhaust together.")
self.__exhaust = True
# This is ugly. People want to be able to do cursor[5:5] and
# get an empty result set (old behavior was an
# exception). It's hard to do that right, though, because the
# server uses limit(0) to mean 'no limit'. So we set __empty
# in that case and check for it when iterating. We also unset
# it anytime we change __limit.
self.__empty = False
self.__data = deque()
self.__address = None
self.__retrieved = 0
self.__killed = False
self.__codec_options = collection.codec_options
self.__read_preference = collection.read_preference
self.__read_concern = collection.read_concern
self.__query_flags = cursor_type
if self.__read_preference != ReadPreference.PRIMARY:
self.__query_flags |= _QUERY_OPTIONS["slave_okay"]
if no_cursor_timeout:
self.__query_flags |= _QUERY_OPTIONS["no_timeout"]
if allow_partial_results:
self.__query_flags |= _QUERY_OPTIONS["partial"]
if oplog_replay:
self.__query_flags |= _QUERY_OPTIONS["oplog_replay"]
def _do_batched_write_command(namespace, operation, command,
docs, check_keys, opts, ctx):
"""Execute a batch of insert, update, or delete commands.
"""
max_bson_size = ctx.max_bson_size
max_write_batch_size = ctx.max_write_batch_size
# Max BSON object size + 16k - 2 bytes for ending NUL bytes.
# Server guarantees there is enough room: SERVER-10643.
max_cmd_size = max_bson_size + _COMMAND_OVERHEAD
ordered = command.get('ordered', True)
buf = StringIO()
# Save space for message length and request id
buf.write(_ZERO_64)
# responseTo, opCode
buf.write(b"\x00\x00\x00\x00\xd4\x07\x00\x00")
# No options
buf.write(_ZERO_32)
# Namespace as C string
buf.write(b(namespace))
buf.write(_ZERO_8)
# Skip: 0, Limit: -1
buf.write(_SKIPLIM)
# Where to write command document length
command_start = buf.tell()
buf.write(bson.BSON.encode(command))
# Start of payload
buf.seek(-1, 2)
# Work around some Jython weirdness.
buf.truncate()
try:
buf.write(_OP_MAP[operation])
except KeyError:
raise InvalidOperation('Unknown command')
if operation in (_UPDATE, _DELETE):
check_keys = False
# Where to write list document length
list_start = buf.tell() - 4
to_send = []
def send_message():
"""Finalize and send the current OP_QUERY message.
"""
# Close list and command documents
buf.write(_ZERO_16)
# Write document lengths and request id
length = buf.tell()
buf.seek(list_start)
buf.write(struct.pack('<i', length - list_start - 1))
buf.seek(command_start)
buf.write(struct.pack('<i', length - command_start))
buf.seek(4)
request_id = _randint()
buf.write(struct.pack('<i', request_id))
buf.seek(0)
buf.write(struct.pack('<i', length))
return ctx.write_command(request_id, buf.getvalue(), to_send)
# If there are multiple batches we'll
# merge results in the caller.
results = []
idx = 0
idx_offset = 0
has_docs = False
for doc in docs:
has_docs = True
# Encode the current operation
key = b(str(idx))
value = bson.BSON.encode(doc, check_keys, opts)
# Send a batch?
enough_data = (buf.tell() + len(key) + len(value) + 2) >= max_cmd_size
enough_documents = (idx >= max_write_batch_size)
if enough_data or enough_documents:
if not idx:
write_op = "insert" if operation == _INSERT else None
_raise_document_too_large(
write_op, len(value), max_bson_size)
result = send_message()
results.append((idx_offset, result))
if ordered and "writeErrors" in result:
return results
# Truncate back to the start of list elements
buf.seek(list_start + 4)
buf.truncate()
idx_offset += idx
idx = 0
key = b'0'
to_send = []
buf.write(_BSONOBJ)
buf.write(key)
buf.write(_ZERO_8)
buf.write(value)
to_send.append(doc)
idx += 1
if not has_docs:
raise InvalidOperation("cannot do an empty bulk write")
results.append((idx_offset, send_message()))
return results
def _do_batched_insert(collection_name, docs, check_keys,
safe, last_error_args, continue_on_error, opts,
ctx):
"""Insert `docs` using multiple batches.
"""
def _insert_message(insert_message, send_safe):
"""Build the insert message with header and GLE.
"""
request_id, final_message = __pack_message(2002, insert_message)
if send_safe:
request_id, error_message, _ = __last_error(collection_name,
last_error_args)
final_message += error_message
return request_id, final_message
send_safe = safe or not continue_on_error
last_error = None
data = StringIO()
data.write(struct.pack("<i", int(continue_on_error)))
data.write(bson._make_c_string(collection_name))
message_length = begin_loc = data.tell()
has_docs = False
to_send = []
for doc in docs:
encoded = bson.BSON.encode(doc, check_keys, opts)
encoded_length = len(encoded)
too_large = (encoded_length > ctx.max_bson_size)
message_length += encoded_length
if message_length < ctx.max_message_size and not too_large:
data.write(encoded)
to_send.append(doc)
has_docs = True
continue
if has_docs:
# We have enough data, send this message.
try:
request_id, msg = _insert_message(data.getvalue(), send_safe)
ctx.legacy_write(request_id, msg, 0, send_safe, to_send)
# Exception type could be OperationFailure or a subtype
# (e.g. DuplicateKeyError)
except OperationFailure as exc:
# Like it says, continue on error...
if continue_on_error:
# Store exception details to re-raise after the final batch.
last_error = exc
# With unacknowledged writes just return at the first error.
elif not safe:
return
# With acknowledged writes raise immediately.
else:
raise
if too_large:
_raise_document_too_large(
"insert", encoded_length, ctx.max_bson_size)
message_length = begin_loc + encoded_length
data.seek(begin_loc)
data.truncate()
data.write(encoded)
to_send = [doc]
if not has_docs:
raise InvalidOperation("cannot do an empty bulk insert")
request_id, msg = _insert_message(data.getvalue(), safe)
ctx.legacy_write(request_id, msg, 0, safe, to_send)
# Re-raise any exception stored due to continue_on_error
if last_error is not None:
raise last_error