def readline(self, size=-1):
"""Read one line or up to `size` bytes from the file.
:Parameters:
- `size` (optional): the maximum number of bytes to read
"""
if size == 0:
return b''
remainder = int(self.length) - self.__position
if size < 0 or size > remainder:
size = remainder
received = 0
data = StringIO()
while received < size:
chunk_data = self.readchunk()
pos = chunk_data.find(NEWLN, 0, size)
if pos != -1:
size = received + pos + 1
received += len(chunk_data)
data.write(chunk_data)
if pos != -1:
break
self.__position -= received - size
# Return 'size' bytes and store the rest.
data.seek(size)
self.__buffer = data.read()
data.seek(0)
return data.read(size)
def write(self, data):
"""Write data to the file. There is no return value.
`data` can be either a string of bytes or a file-like object
(implementing :meth:`read`). If the file has an
:attr:`encoding` attribute, `data` can also be a
:class:`unicode` (:class:`str` in python 3) instance, which
will be encoded as :attr:`encoding` before being written.
Due to buffering, the data may not actually be written to the
database until the :meth:`close` method is called. Raises
:class:`ValueError` if this file is already closed. Raises
:class:`TypeError` if `data` is not an instance of
:class:`str` (:class:`bytes` in python 3), a file-like object,
or an instance of :class:`unicode` (:class:`str` in python 3).
Unicode data is only allowed if the file has an :attr:`encoding`
attribute.
:Parameters:
- `data`: string of bytes or file-like object to be written
to the file
"""
if self._closed:
raise ValueError("cannot write to a closed file")
try:
# file-like
read = data.read
except AttributeError:
# string
if not isinstance(data, (text_type, bytes)):
raise TypeError("can only write strings or file-like objects")
if isinstance(data, text_type):
try:
data = data.encode(self.encoding)
except AttributeError:
raise TypeError("must specify an encoding for file in "
"order to write %s" % (text_type.__name__,))
read = StringIO(data).read
if self._buffer.tell() > 0:
# Make sure to flush only when _buffer is complete
space = self.chunk_size - self._buffer.tell()
if space:
try:
to_write = read(space)
except:
self.abort()
raise
self._buffer.write(to_write)
if len(to_write) < space:
return # EOF or incomplete
self.__flush_buffer()
to_write = read(self.chunk_size)
while to_write and len(to_write) == self.chunk_size:
self.__flush_data(to_write)
to_write = read(self.chunk_size)
self._buffer.write(to_write)
def read(self, size=-1):
"""Read at most `size` bytes from the file (less if there
isn't enough data).
The bytes are returned as an instance of :class:`str` (:class:`bytes`
in python 3). If `size` is negative or omitted all data is read.
:Parameters:
- `size` (optional): the number of bytes to read
"""
self._ensure_file()
if size == 0:
return EMPTY
remainder = int(self.length) - self.__position
if size < 0 or size > remainder:
size = remainder
received = 0
data = StringIO()
while received < size:
chunk_data = self.readchunk()
received += len(chunk_data)
data.write(chunk_data)
# Detect extra chunks.
max_chunk_n = math.ceil(self.length / float(self.chunk_size))
chunk = self.__chunks.find_one({"files_id": self._id,
"n": {"$gte": max_chunk_n}})
# According to spec, ignore extra chunks if they are empty.
if chunk is not None and len(chunk['data']):
raise CorruptGridFile(
"Extra chunk found: expected %i chunks but found "
"chunk with n=%i" % (max_chunk_n, chunk['n']))
self.__position -= received - size
# Return 'size' bytes and store the rest.
data.seek(size)
self.__buffer = data.read()
data.seek(0)
return data.read(size)
class GridIn(object):
"""Class to write data to GridFS.
"""
def __init__(self, root_collection, **kwargs):
"""Write a file to GridFS
Application developers should generally not need to
instantiate this class directly - instead see the methods
provided by :class:`~gridfs.GridFS`.
Raises :class:`TypeError` if `root_collection` is not an
instance of :class:`~pymongo.collection.Collection`.
Any of the file level options specified in the `GridFS Spec
<http://dochub.mongodb.org/core/gridfsspec>`_ may be passed as
keyword arguments. Any additional keyword arguments will be
set as additional fields on the file document. Valid keyword
arguments include:
- ``"_id"``: unique ID for this file (default:
:class:`~bson.objectid.ObjectId`) - this ``"_id"`` must
not have already been used for another file
- ``"filename"``: human name for the file
- ``"contentType"`` or ``"content_type"``: valid mime-type
for the file
- ``"chunkSize"`` or ``"chunk_size"``: size of each of the
chunks, in bytes (default: 255 kb)
- ``"encoding"``: encoding used for this file. In Python 2,
any :class:`unicode` that is written to the file will be
converted to a :class:`str`. In Python 3, any :class:`str`
that is written to the file will be converted to
:class:`bytes`.
:Parameters:
- `root_collection`: root collection to write to
- `**kwargs` (optional): file level options (see above)
.. versionchanged:: 3.0
`root_collection` must use an acknowledged
:attr:`~pymongo.collection.Collection.write_concern`
"""
if not isinstance(root_collection, Collection):
raise TypeError("root_collection must be an "
"instance of Collection")
# With w=0, 'filemd5' might run before the final chunks are written.
if not root_collection.write_concern.acknowledged:
raise ConfigurationError('root_collection must use '
'acknowledged write_concern')
# Handle alternative naming
if "content_type" in kwargs:
kwargs["contentType"] = kwargs.pop("content_type")
if "chunk_size" in kwargs:
kwargs["chunkSize"] = kwargs.pop("chunk_size")
coll = root_collection.with_options(
read_preference=ReadPreference.PRIMARY)
kwargs['md5'] = md5()
# Defaults
kwargs["_id"] = kwargs.get("_id", ObjectId())
kwargs["chunkSize"] = kwargs.get("chunkSize", DEFAULT_CHUNK_SIZE)
object.__setattr__(self, "_coll", coll)
object.__setattr__(self, "_chunks", coll.chunks)
object.__setattr__(self, "_file", kwargs)
object.__setattr__(self, "_buffer", StringIO())
object.__setattr__(self, "_position", 0)
object.__setattr__(self, "_chunk_number", 0)
object.__setattr__(self, "_closed", False)
object.__setattr__(self, "_ensured_index", False)
def __create_index(self, collection, index_key, unique):
doc = collection.find_one(projection={"_id": 1})
if doc is None:
try:
index_keys =[index_spec['key'] for index_spec in collection.list_indexes()]
except OperationFailure:
index_keys = []
if index_key not in index_keys:
collection.create_index(index_key.items(), unique=unique)
def __ensure_indexes(self):
if not object.__getattribute__(self, "_ensured_index"):
self.__create_index(self._coll.files, _F_INDEX, False)
self.__create_index(self._coll.chunks, _C_INDEX, True)
object.__setattr__(self, "_ensured_index", True)
def abort(self):
"""Remove all chunks/files that may have been uploaded and close.
"""
self._coll.chunks.delete_many({"files_id": self._file['_id']})
self._coll.files.delete_one({"_id": self._file['_id']})
object.__setattr__(self, "_closed", True)
@property
def closed(self):
"""Is this file closed?
"""
return self._closed
_id = _grid_in_property("_id", "The ``'_id'`` value for this file.",
read_only=True)
filename = _grid_in_property("filename", "Name of this file.")
name = _grid_in_property("filename", "Alias for `filename`.")
content_type = _grid_in_property("contentType", "Mime-type for this file.")
length = _grid_in_property("length", "Length (in bytes) of this file.",
closed_only=True)
chunk_size = _grid_in_property("chunkSize", "Chunk size for this file.",
read_only=True)
upload_date = _grid_in_property("uploadDate",
"Date that this file was uploaded.",
closed_only=True)
md5 = _grid_in_property("md5", "MD5 of the contents of this file "
"(generated on the server).",
closed_only=True)
def __getattr__(self, name):
if name in self._file:
return self._file[name]
raise AttributeError("GridIn object has no attribute '%s'" % name)
def __setattr__(self, name, value):
# For properties of this instance like _buffer, or descriptors set on
# the class like filename, use regular __setattr__
if name in self.__dict__ or name in self.__class__.__dict__:
object.__setattr__(self, name, value)
else:
# All other attributes are part of the document in db.fs.files.
# Store them to be sent to server on close() or if closed, send
# them now.
self._file[name] = value
if self._closed:
self._coll.files.update_one({"_id": self._file["_id"]},
{"$set": {name: value}})
def __flush_data(self, data):
"""Flush `data` to a chunk.
"""
# Ensure the index, even if there's nothing to write, so
# the filemd5 command always succeeds.
self.__ensure_indexes()
self._file['md5'].update(data)
if not data:
return
assert(len(data) <= self.chunk_size)
chunk = {"files_id": self._file["_id"],
"n": self._chunk_number,
"data": Binary(data)}
try:
self._chunks.insert_one(chunk)
except DuplicateKeyError:
self._raise_file_exists(self._file['_id'])
self._chunk_number += 1
self._position += len(data)
def __flush_buffer(self):
"""Flush the buffer contents out to a chunk.
"""
self.__flush_data(self._buffer.getvalue())
self._buffer.close()
self._buffer = StringIO()
def __flush(self):
"""Flush the file to the database.
"""
try:
self.__flush_buffer()
self._file['md5'] = self._file["md5"].hexdigest()
self._file["length"] = self._position
self._file["uploadDate"] = datetime.datetime.utcnow()
return self._coll.files.insert_one(self._file)
except DuplicateKeyError:
self._raise_file_exists(self._id)
def _raise_file_exists(self, file_id):
"""Raise a FileExists exception for the given file_id."""
raise FileExists("file with _id %r already exists" % file_id)
def close(self):
"""Flush the file and close it.
A closed file cannot be written any more. Calling
:meth:`close` more than once is allowed.
"""
if not self._closed:
self.__flush()
object.__setattr__(self, "_closed", True)
def write(self, data):
"""Write data to the file. There is no return value.
`data` can be either a string of bytes or a file-like object
(implementing :meth:`read`). If the file has an
:attr:`encoding` attribute, `data` can also be a
:class:`unicode` (:class:`str` in python 3) instance, which
will be encoded as :attr:`encoding` before being written.
Due to buffering, the data may not actually be written to the
database until the :meth:`close` method is called. Raises
:class:`ValueError` if this file is already closed. Raises
:class:`TypeError` if `data` is not an instance of
:class:`str` (:class:`bytes` in python 3), a file-like object,
or an instance of :class:`unicode` (:class:`str` in python 3).
Unicode data is only allowed if the file has an :attr:`encoding`
attribute.
:Parameters:
- `data`: string of bytes or file-like object to be written
to the file
"""
if self._closed:
raise ValueError("cannot write to a closed file")
try:
# file-like
read = data.read
except AttributeError:
# string
if not isinstance(data, (text_type, bytes)):
raise TypeError("can only write strings or file-like objects")
if isinstance(data, text_type):
try:
data = data.encode(self.encoding)
except AttributeError:
raise TypeError("must specify an encoding for file in "
"order to write %s" % (text_type.__name__,))
read = StringIO(data).read
if self._buffer.tell() > 0:
# Make sure to flush only when _buffer is complete
space = self.chunk_size - self._buffer.tell()
if space:
try:
to_write = read(space)
except:
self.abort()
raise
self._buffer.write(to_write)
if len(to_write) < space:
return # EOF or incomplete
self.__flush_buffer()
to_write = read(self.chunk_size)
while to_write and len(to_write) == self.chunk_size:
self.__flush_data(to_write)
to_write = read(self.chunk_size)
self._buffer.write(to_write)
def writelines(self, sequence):
"""Write a sequence of strings to the file.
Does not add seperators.
"""
for line in sequence:
self.write(line)
def __enter__(self):
"""Support for the context manager protocol.
"""
return self
def __exit__(self, exc_type, exc_val, exc_tb):
"""Support for the context manager protocol.
Close the file and allow exceptions to propagate.
"""
self.close()
# propagate exceptions
return False
def __flush_buffer(self):
"""Flush the buffer contents out to a chunk.
"""
self.__flush_data(self._buffer.getvalue())
self._buffer.close()
self._buffer = StringIO()
def __init__(self, root_collection, **kwargs):
"""Write a file to GridFS
Application developers should generally not need to
instantiate this class directly - instead see the methods
provided by :class:`~gridfs.GridFS`.
Raises :class:`TypeError` if `root_collection` is not an
instance of :class:`~pymongo.collection.Collection`.
Any of the file level options specified in the `GridFS Spec
<http://dochub.mongodb.org/core/gridfsspec>`_ may be passed as
keyword arguments. Any additional keyword arguments will be
set as additional fields on the file document. Valid keyword
arguments include:
- ``"_id"``: unique ID for this file (default:
:class:`~bson.objectid.ObjectId`) - this ``"_id"`` must
not have already been used for another file
- ``"filename"``: human name for the file
- ``"contentType"`` or ``"content_type"``: valid mime-type
for the file
- ``"chunkSize"`` or ``"chunk_size"``: size of each of the
chunks, in bytes (default: 255 kb)
- ``"encoding"``: encoding used for this file. In Python 2,
any :class:`unicode` that is written to the file will be
converted to a :class:`str`. In Python 3, any :class:`str`
that is written to the file will be converted to
:class:`bytes`.
:Parameters:
- `root_collection`: root collection to write to
- `**kwargs` (optional): file level options (see above)
.. versionchanged:: 3.0
`root_collection` must use an acknowledged
:attr:`~pymongo.collection.Collection.write_concern`
"""
if not isinstance(root_collection, Collection):
raise TypeError("root_collection must be an "
"instance of Collection")
# With w=0, 'filemd5' might run before the final chunks are written.
if not root_collection.write_concern.acknowledged:
raise ConfigurationError('root_collection must use '
'acknowledged write_concern')
# Handle alternative naming
if "content_type" in kwargs:
kwargs["contentType"] = kwargs.pop("content_type")
if "chunk_size" in kwargs:
kwargs["chunkSize"] = kwargs.pop("chunk_size")
coll = root_collection.with_options(
read_preference=ReadPreference.PRIMARY)
kwargs['md5'] = md5()
# Defaults
kwargs["_id"] = kwargs.get("_id", ObjectId())
kwargs["chunkSize"] = kwargs.get("chunkSize", DEFAULT_CHUNK_SIZE)
object.__setattr__(self, "_coll", coll)
object.__setattr__(self, "_chunks", coll.chunks)
object.__setattr__(self, "_file", kwargs)
object.__setattr__(self, "_buffer", StringIO())
object.__setattr__(self, "_position", 0)
object.__setattr__(self, "_chunk_number", 0)
object.__setattr__(self, "_closed", False)
object.__setattr__(self, "_ensured_index", False)
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