""" Utilities for fast persistence of big data, with optional compression. """ # Author: Gael Varoquaux # Copyright (c) 2009 Gael Varoquaux # License: BSD Style, 3 clauses. import pickle import traceback import os import zlib import warnings from io import BytesIO from ._compat import _basestring, PY3_OR_LATER if PY3_OR_LATER: Unpickler = pickle._Unpickler Pickler = pickle._Pickler def asbytes(s): if isinstance(s, bytes): return s return s.encode('latin1') else: Unpickler = pickle.Unpickler Pickler = pickle.Pickler asbytes = str def hex_str(an_int): """Converts an int to an hexadecimal string """ return '{0:#x}'.format(an_int) _MEGA = 2 ** 20 # Compressed pickle header format: _ZFILE_PREFIX followed by _MAX_LEN # bytes which contains the length of the zlib compressed data as an # hexadecimal string. For example: 'ZF0x139 ' _ZFILE_PREFIX = asbytes('ZF') _MAX_LEN = len(hex_str(2 ** 64)) ############################################################################### # Compressed file with Zlib def _read_magic(file_handle): """ Utility to check the magic signature of a file identifying it as a Zfile """ magic = file_handle.read(len(_ZFILE_PREFIX)) # Pickling needs file-handles at the beginning of the file file_handle.seek(0) return magic def read_zfile(file_handle): """Read the z-file and return the content as a string Z-files are raw data compressed with zlib used internally by joblib for persistence. Backward compatibility is not guaranteed. Do not use for external purposes. """ file_handle.seek(0) assert _read_magic(file_handle) == _ZFILE_PREFIX, \ "File does not have the right magic" header_length = len(_ZFILE_PREFIX) + _MAX_LEN length = file_handle.read(header_length) length = length[len(_ZFILE_PREFIX):] length = int(length, 16) # With python2 and joblib version <= 0.8.4 compressed pickle header is one # character wider so we need to ignore an additional space if present. # Note: the first byte of the zlib data is guaranteed not to be a # space according to # https://tools.ietf.org/html/rfc6713#section-2.1 next_byte = file_handle.read(1) if next_byte != b' ': # The zlib compressed data has started and we need to go back # one byte file_handle.seek(header_length) # We use the known length of the data to tell Zlib the size of the # buffer to allocate. data = zlib.decompress(file_handle.read(), 15, length) assert len(data) == length, ( "Incorrect data length while decompressing %s." "The file could be corrupted." % file_handle) return data def write_zfile(file_handle, data, compress=1): """Write the data in the given file as a Z-file. Z-files are raw data compressed with zlib used internally by joblib for persistence. Backward compatibility is not guarantied. Do not use for external purposes. """ file_handle.write(_ZFILE_PREFIX) length = hex_str(len(data)) # Store the length of the data file_handle.write(asbytes(length.ljust(_MAX_LEN))) file_handle.write(zlib.compress(asbytes(data), compress)) ############################################################################### # Utility objects for persistence. class NDArrayWrapper(object): """ An object to be persisted instead of numpy arrays. The only thing this object does, is to carry the filename in which the array has been persisted, and the array subclass. """ def __init__(self, filename, subclass, allow_mmap=True): "Store the useful information for later" self.filename = filename self.subclass = subclass self.allow_mmap = allow_mmap def read(self, unpickler): "Reconstruct the array" filename = os.path.join(unpickler._dirname, self.filename) # Load the array from the disk np_ver = [int(x) for x in unpickler.np.__version__.split('.', 2)[:2]] # use getattr instead of self.allow_mmap to ensure backward compat # with NDArrayWrapper instances pickled with joblib < 0.9.0 allow_mmap = getattr(self, 'allow_mmap', True) memmap_kwargs = ({} if not allow_mmap else {'mmap_mode': unpickler.mmap_mode}) array = unpickler.np.load(filename, **memmap_kwargs) # Reconstruct subclasses. This does not work with old # versions of numpy if (hasattr(array, '__array_prepare__') and not self.subclass in (unpickler.np.ndarray, unpickler.np.memmap)): # We need to reconstruct another subclass new_array = unpickler.np.core.multiarray._reconstruct( self.subclass, (0,), 'b') new_array.__array_prepare__(array) array = new_array return array #def __reduce__(self): # return None class ZNDArrayWrapper(NDArrayWrapper): """An object to be persisted instead of numpy arrays. This object store the Zfile filename in which the data array has been persisted, and the meta information to retrieve it. The reason that we store the raw buffer data of the array and the meta information, rather than array representation routine (tostring) is that it enables us to use completely the strided model to avoid memory copies (a and a.T store as fast). In addition saving the heavy information separately can avoid creating large temporary buffers when unpickling data with large arrays. """ def __init__(self, filename, init_args, state): "Store the useful information for later" self.filename = filename self.state = state self.init_args = init_args def read(self, unpickler): "Reconstruct the array from the meta-information and the z-file" # Here we a simply reproducing the unpickling mechanism for numpy # arrays filename = os.path.join(unpickler._dirname, self.filename) array = unpickler.np.core.multiarray._reconstruct(*self.init_args) with open(filename, 'rb') as f: data = read_zfile(f) state = self.state + (data,) array.__setstate__(state) return array ############################################################################### # Pickler classes class NumpyPickler(Pickler): """A pickler to persist of big data efficiently. The main features of this object are: * persistence of numpy arrays in separate .npy files, for which I/O is fast. * optional compression using Zlib, with a special care on avoid temporaries. """ dispatch = Pickler.dispatch.copy() def __init__(self, filename, compress=0, cache_size=10, protocol=None): self._filename = filename self._filenames = [filename, ] self.cache_size = cache_size self.compress = compress if not self.compress: self.file = open(filename, 'wb') else: self.file = BytesIO() # Count the number of npy files that we have created: self._npy_counter = 1 # By default we want a pickle protocol that only changes with # the major python version and not the minor one if protocol is None: protocol = (pickle.DEFAULT_PROTOCOL if PY3_OR_LATER else pickle.HIGHEST_PROTOCOL) Pickler.__init__(self, self.file, protocol=protocol) # delayed import of numpy, to avoid tight coupling try: import numpy as np except ImportError: np = None self.np = np def _write_array(self, array, filename): if not self.compress: self.np.save(filename, array) allow_mmap = not array.dtype.hasobject container = NDArrayWrapper(os.path.basename(filename), type(array), allow_mmap=allow_mmap) else: filename += '.z' # Efficient compressed storage: # The meta data is stored in the container, and the core # numerics in a z-file _, init_args, state = array.__reduce__() # the last entry of 'state' is the data itself with open(filename, 'wb') as zfile: write_zfile(zfile, state[-1], compress=self.compress) state = state[:-1] container = ZNDArrayWrapper(os.path.basename(filename), init_args, state) return container, filename def save(self, obj): """ Subclass the save method, to save ndarray subclasses in npy files, rather than pickling them. Of course, this is a total abuse of the Pickler class. """ if (self.np is not None and type(obj) in (self.np.ndarray, self.np.matrix, self.np.memmap)): size = obj.size * obj.itemsize if self.compress and size < self.cache_size * _MEGA: # When compressing, as we are not writing directly to the # disk, it is more efficient to use standard pickling if type(obj) is self.np.memmap: # Pickling doesn't work with memmaped arrays obj = self.np.asarray(obj) return Pickler.save(self, obj) if not obj.dtype.hasobject: try: filename = '%s_%02i.npy' % (self._filename, self._npy_counter) # This converts the array in a container obj, filename = self._write_array(obj, filename) self._filenames.append(filename) self._npy_counter += 1 except Exception: # XXX: We should have a logging mechanism print('Failed to save %s to .npy file:\n%s' % ( type(obj), traceback.format_exc())) return Pickler.save(self, obj) def close(self): if self.compress: with open(self._filename, 'wb') as zfile: write_zfile(zfile, self.file.getvalue(), self.compress) class NumpyUnpickler(Unpickler): """A subclass of the Unpickler to unpickle our numpy pickles. """ dispatch = Unpickler.dispatch.copy() def __init__(self, filename, file_handle, mmap_mode=None): self._filename = os.path.basename(filename) self._dirname = os.path.dirname(filename) self.mmap_mode = mmap_mode self.file_handle = self._open_pickle(file_handle) Unpickler.__init__(self, self.file_handle) try: import numpy as np except ImportError: np = None self.np = np def _open_pickle(self, file_handle): return file_handle def load_build(self): """ This method is called to set the state of a newly created object. We capture it to replace our place-holder objects, NDArrayWrapper, by the array we are interested in. We replace them directly in the stack of pickler. """ Unpickler.load_build(self) if isinstance(self.stack[-1], NDArrayWrapper): if self.np is None: raise ImportError('Trying to unpickle an ndarray, ' "but numpy didn't import correctly") nd_array_wrapper = self.stack.pop() array = nd_array_wrapper.read(self) self.stack.append(array) # Be careful to register our new method. if PY3_OR_LATER: dispatch[pickle.BUILD[0]] = load_build else: dispatch[pickle.BUILD] = load_build class ZipNumpyUnpickler(NumpyUnpickler): """A subclass of our Unpickler to unpickle on the fly from compressed storage.""" def __init__(self, filename, file_handle): NumpyUnpickler.__init__(self, filename, file_handle, mmap_mode=None) def _open_pickle(self, file_handle): return BytesIO(read_zfile(file_handle)) ############################################################################### # Utility functions def dump(value, filename, compress=0, cache_size=100, protocol=None): """Fast persistence of an arbitrary Python object into one or multiple files, with dedicated storage for numpy arrays. Parameters ----------- value: any Python object The object to store to disk filename: string The name of the file in which it is to be stored compress: integer for 0 to 9, optional Optional compression level for the data. 0 is no compression. Higher means more compression, but also slower read and write times. Using a value of 3 is often a good compromise. See the notes for more details. cache_size: positive number, optional Fixes the order of magnitude (in megabytes) of the cache used for in-memory compression. Note that this is just an order of magnitude estimate and that for big arrays, the code will go over this value at dump and at load time. protocol: positive int Pickle protocol, see pickle.dump documentation for more details. Returns ------- filenames: list of strings The list of file names in which the data is stored. If compress is false, each array is stored in a different file. See Also -------- joblib.load : corresponding loader Notes ----- Memmapping on load cannot be used for compressed files. Thus using compression can significantly slow down loading. In addition, compressed files take extra extra memory during dump and load. """ if compress is True: # By default, if compress is enabled, we want to be using 3 by # default compress = 3 if not isinstance(filename, _basestring): # People keep inverting arguments, and the resulting error is # incomprehensible raise ValueError( 'Second argument should be a filename, %s (type %s) was given' % (filename, type(filename)) ) try: pickler = NumpyPickler(filename, compress=compress, cache_size=cache_size, protocol=protocol) pickler.dump(value) pickler.close() finally: if 'pickler' in locals() and hasattr(pickler, 'file'): pickler.file.flush() pickler.file.close() return pickler._filenames def load(filename, mmap_mode=None): """Reconstruct a Python object from a file persisted with joblib.dump. Parameters ----------- filename: string The name of the file from which to load the object mmap_mode: {None, 'r+', 'r', 'w+', 'c'}, optional If not None, the arrays are memory-mapped from the disk. This mode has no effect for compressed files. Note that in this case the reconstructed object might not longer match exactly the originally pickled object. Returns ------- result: any Python object The object stored in the file. See Also -------- joblib.dump : function to save an object Notes ----- This function can load numpy array files saved separately during the dump. If the mmap_mode argument is given, it is passed to np.load and arrays are loaded as memmaps. As a consequence, the reconstructed object might not match the original pickled object. Note that if the file was saved with compression, the arrays cannot be memmaped. """ with open(filename, 'rb') as file_handle: # We are careful to open the file handle early and keep it open to # avoid race-conditions on renames. That said, if data are stored in # companion files, moving the directory will create a race when # joblib tries to access the companion files. if _read_magic(file_handle) == _ZFILE_PREFIX: if mmap_mode is not None: warnings.warn('file "%(filename)s" appears to be a zip, ' 'ignoring mmap_mode "%(mmap_mode)s" flag passed' % locals(), Warning, stacklevel=2) unpickler = ZipNumpyUnpickler(filename, file_handle=file_handle) else: unpickler = NumpyUnpickler(filename, file_handle=file_handle, mmap_mode=mmap_mode) try: obj = unpickler.load() except UnicodeDecodeError as exc: # More user-friendly error message if PY3_OR_LATER: new_exc = ValueError( 'You may be trying to read with ' 'python 3 a joblib pickle generated with python 2. ' 'This feature is not supported by joblib.') new_exc.__cause__ = exc raise new_exc finally: if hasattr(unpickler, 'file_handle'): unpickler.file_handle.close() return obj