1"""Create portable serialized representations of Python objects. 2 3See module copyreg for a mechanism for registering custom picklers. 4See module pickletools source for extensive comments. 5 6Classes: 7 8 Pickler 9 Unpickler 10 11Functions: 12 13 dump(object, file) 14 dumps(object) -> string 15 load(file) -> object 16 loads(bytes) -> object 17 18Misc variables: 19 20 __version__ 21 format_version 22 compatible_formats 23 24""" 25 26from types import FunctionType 27from copyreg import dispatch_table 28from copyreg import _extension_registry, _inverted_registry, _extension_cache 29from itertools import islice 30from functools import partial 31import sys 32from sys import maxsize 33from struct import pack, unpack 34import re 35import io 36import codecs 37import _compat_pickle 38 39__all__ = ["PickleError", "PicklingError", "UnpicklingError", "Pickler", 40 "Unpickler", "dump", "dumps", "load", "loads"] 41 42try: 43 from _pickle import PickleBuffer 44 __all__.append("PickleBuffer") 45 _HAVE_PICKLE_BUFFER = True 46except ImportError: 47 _HAVE_PICKLE_BUFFER = False 48 49 50# Shortcut for use in isinstance testing 51bytes_types = (bytes, bytearray) 52 53# These are purely informational; no code uses these. 54format_version = "4.0" # File format version we write 55compatible_formats = ["1.0", # Original protocol 0 56 "1.1", # Protocol 0 with INST added 57 "1.2", # Original protocol 1 58 "1.3", # Protocol 1 with BINFLOAT added 59 "2.0", # Protocol 2 60 "3.0", # Protocol 3 61 "4.0", # Protocol 4 62 "5.0", # Protocol 5 63 ] # Old format versions we can read 64 65# This is the highest protocol number we know how to read. 66HIGHEST_PROTOCOL = 5 67 68# The protocol we write by default. May be less than HIGHEST_PROTOCOL. 69# Only bump this if the oldest still supported version of Python already 70# includes it. 71DEFAULT_PROTOCOL = 4 72 73class PickleError(Exception): 74 """A common base class for the other pickling exceptions.""" 75 pass 76 77class PicklingError(PickleError): 78 """This exception is raised when an unpicklable object is passed to the 79 dump() method. 80 81 """ 82 pass 83 84class UnpicklingError(PickleError): 85 """This exception is raised when there is a problem unpickling an object, 86 such as a security violation. 87 88 Note that other exceptions may also be raised during unpickling, including 89 (but not necessarily limited to) AttributeError, EOFError, ImportError, 90 and IndexError. 91 92 """ 93 pass 94 95# An instance of _Stop is raised by Unpickler.load_stop() in response to 96# the STOP opcode, passing the object that is the result of unpickling. 97class _Stop(Exception): 98 def __init__(self, value): 99 self.value = value 100 101# Jython has PyStringMap; it's a dict subclass with string keys 102try: 103 from org.python.core import PyStringMap 104except ImportError: 105 PyStringMap = None 106 107# Pickle opcodes. See pickletools.py for extensive docs. The listing 108# here is in kind-of alphabetical order of 1-character pickle code. 109# pickletools groups them by purpose. 110 111MARK = b'(' # push special markobject on stack 112STOP = b'.' # every pickle ends with STOP 113POP = b'0' # discard topmost stack item 114POP_MARK = b'1' # discard stack top through topmost markobject 115DUP = b'2' # duplicate top stack item 116FLOAT = b'F' # push float object; decimal string argument 117INT = b'I' # push integer or bool; decimal string argument 118BININT = b'J' # push four-byte signed int 119BININT1 = b'K' # push 1-byte unsigned int 120LONG = b'L' # push long; decimal string argument 121BININT2 = b'M' # push 2-byte unsigned int 122NONE = b'N' # push None 123PERSID = b'P' # push persistent object; id is taken from string arg 124BINPERSID = b'Q' # " " " ; " " " " stack 125REDUCE = b'R' # apply callable to argtuple, both on stack 126STRING = b'S' # push string; NL-terminated string argument 127BINSTRING = b'T' # push string; counted binary string argument 128SHORT_BINSTRING= b'U' # " " ; " " " " < 256 bytes 129UNICODE = b'V' # push Unicode string; raw-unicode-escaped'd argument 130BINUNICODE = b'X' # " " " ; counted UTF-8 string argument 131APPEND = b'a' # append stack top to list below it 132BUILD = b'b' # call __setstate__ or __dict__.update() 133GLOBAL = b'c' # push self.find_class(modname, name); 2 string args 134DICT = b'd' # build a dict from stack items 135EMPTY_DICT = b'}' # push empty dict 136APPENDS = b'e' # extend list on stack by topmost stack slice 137GET = b'g' # push item from memo on stack; index is string arg 138BINGET = b'h' # " " " " " " ; " " 1-byte arg 139INST = b'i' # build & push class instance 140LONG_BINGET = b'j' # push item from memo on stack; index is 4-byte arg 141LIST = b'l' # build list from topmost stack items 142EMPTY_LIST = b']' # push empty list 143OBJ = b'o' # build & push class instance 144PUT = b'p' # store stack top in memo; index is string arg 145BINPUT = b'q' # " " " " " ; " " 1-byte arg 146LONG_BINPUT = b'r' # " " " " " ; " " 4-byte arg 147SETITEM = b's' # add key+value pair to dict 148TUPLE = b't' # build tuple from topmost stack items 149EMPTY_TUPLE = b')' # push empty tuple 150SETITEMS = b'u' # modify dict by adding topmost key+value pairs 151BINFLOAT = b'G' # push float; arg is 8-byte float encoding 152 153TRUE = b'I01\n' # not an opcode; see INT docs in pickletools.py 154FALSE = b'I00\n' # not an opcode; see INT docs in pickletools.py 155 156# Protocol 2 157 158PROTO = b'\x80' # identify pickle protocol 159NEWOBJ = b'\x81' # build object by applying cls.__new__ to argtuple 160EXT1 = b'\x82' # push object from extension registry; 1-byte index 161EXT2 = b'\x83' # ditto, but 2-byte index 162EXT4 = b'\x84' # ditto, but 4-byte index 163TUPLE1 = b'\x85' # build 1-tuple from stack top 164TUPLE2 = b'\x86' # build 2-tuple from two topmost stack items 165TUPLE3 = b'\x87' # build 3-tuple from three topmost stack items 166NEWTRUE = b'\x88' # push True 167NEWFALSE = b'\x89' # push False 168LONG1 = b'\x8a' # push long from < 256 bytes 169LONG4 = b'\x8b' # push really big long 170 171_tuplesize2code = [EMPTY_TUPLE, TUPLE1, TUPLE2, TUPLE3] 172 173# Protocol 3 (Python 3.x) 174 175BINBYTES = b'B' # push bytes; counted binary string argument 176SHORT_BINBYTES = b'C' # " " ; " " " " < 256 bytes 177 178# Protocol 4 179 180SHORT_BINUNICODE = b'\x8c' # push short string; UTF-8 length < 256 bytes 181BINUNICODE8 = b'\x8d' # push very long string 182BINBYTES8 = b'\x8e' # push very long bytes string 183EMPTY_SET = b'\x8f' # push empty set on the stack 184ADDITEMS = b'\x90' # modify set by adding topmost stack items 185FROZENSET = b'\x91' # build frozenset from topmost stack items 186NEWOBJ_EX = b'\x92' # like NEWOBJ but work with keyword only arguments 187STACK_GLOBAL = b'\x93' # same as GLOBAL but using names on the stacks 188MEMOIZE = b'\x94' # store top of the stack in memo 189FRAME = b'\x95' # indicate the beginning of a new frame 190 191# Protocol 5 192 193BYTEARRAY8 = b'\x96' # push bytearray 194NEXT_BUFFER = b'\x97' # push next out-of-band buffer 195READONLY_BUFFER = b'\x98' # make top of stack readonly 196 197__all__.extend([x for x in dir() if re.match("[A-Z][A-Z0-9_]+$", x)]) 198 199 200class _Framer: 201 202 _FRAME_SIZE_MIN = 4 203 _FRAME_SIZE_TARGET = 64 * 1024 204 205 def __init__(self, file_write): 206 self.file_write = file_write 207 self.current_frame = None 208 209 def start_framing(self): 210 self.current_frame = io.BytesIO() 211 212 def end_framing(self): 213 if self.current_frame and self.current_frame.tell() > 0: 214 self.commit_frame(force=True) 215 self.current_frame = None 216 217 def commit_frame(self, force=False): 218 if self.current_frame: 219 f = self.current_frame 220 if f.tell() >= self._FRAME_SIZE_TARGET or force: 221 data = f.getbuffer() 222 write = self.file_write 223 if len(data) >= self._FRAME_SIZE_MIN: 224 # Issue a single call to the write method of the underlying 225 # file object for the frame opcode with the size of the 226 # frame. The concatenation is expected to be less expensive 227 # than issuing an additional call to write. 228 write(FRAME + pack("<Q", len(data))) 229 230 # Issue a separate call to write to append the frame 231 # contents without concatenation to the above to avoid a 232 # memory copy. 233 write(data) 234 235 # Start the new frame with a new io.BytesIO instance so that 236 # the file object can have delayed access to the previous frame 237 # contents via an unreleased memoryview of the previous 238 # io.BytesIO instance. 239 self.current_frame = io.BytesIO() 240 241 def write(self, data): 242 if self.current_frame: 243 return self.current_frame.write(data) 244 else: 245 return self.file_write(data) 246 247 def write_large_bytes(self, header, payload): 248 write = self.file_write 249 if self.current_frame: 250 # Terminate the current frame and flush it to the file. 251 self.commit_frame(force=True) 252 253 # Perform direct write of the header and payload of the large binary 254 # object. Be careful not to concatenate the header and the payload 255 # prior to calling 'write' as we do not want to allocate a large 256 # temporary bytes object. 257 # We intentionally do not insert a protocol 4 frame opcode to make 258 # it possible to optimize file.read calls in the loader. 259 write(header) 260 write(payload) 261 262 263class _Unframer: 264 265 def __init__(self, file_read, file_readline, file_tell=None): 266 self.file_read = file_read 267 self.file_readline = file_readline 268 self.current_frame = None 269 270 def readinto(self, buf): 271 if self.current_frame: 272 n = self.current_frame.readinto(buf) 273 if n == 0 and len(buf) != 0: 274 self.current_frame = None 275 n = len(buf) 276 buf[:] = self.file_read(n) 277 return n 278 if n < len(buf): 279 raise UnpicklingError( 280 "pickle exhausted before end of frame") 281 return n 282 else: 283 n = len(buf) 284 buf[:] = self.file_read(n) 285 return n 286 287 def read(self, n): 288 if self.current_frame: 289 data = self.current_frame.read(n) 290 if not data and n != 0: 291 self.current_frame = None 292 return self.file_read(n) 293 if len(data) < n: 294 raise UnpicklingError( 295 "pickle exhausted before end of frame") 296 return data 297 else: 298 return self.file_read(n) 299 300 def readline(self): 301 if self.current_frame: 302 data = self.current_frame.readline() 303 if not data: 304 self.current_frame = None 305 return self.file_readline() 306 if data[-1] != b'\n'[0]: 307 raise UnpicklingError( 308 "pickle exhausted before end of frame") 309 return data 310 else: 311 return self.file_readline() 312 313 def load_frame(self, frame_size): 314 if self.current_frame and self.current_frame.read() != b'': 315 raise UnpicklingError( 316 "beginning of a new frame before end of current frame") 317 self.current_frame = io.BytesIO(self.file_read(frame_size)) 318 319 320# Tools used for pickling. 321 322def _getattribute(obj, name): 323 for subpath in name.split('.'): 324 if subpath == '<locals>': 325 raise AttributeError("Can't get local attribute {!r} on {!r}" 326 .format(name, obj)) 327 try: 328 parent = obj 329 obj = getattr(obj, subpath) 330 except AttributeError: 331 raise AttributeError("Can't get attribute {!r} on {!r}" 332 .format(name, obj)) from None 333 return obj, parent 334 335def whichmodule(obj, name): 336 """Find the module an object belong to.""" 337 module_name = getattr(obj, '__module__', None) 338 if module_name is not None: 339 return module_name 340 # Protect the iteration by using a list copy of sys.modules against dynamic 341 # modules that trigger imports of other modules upon calls to getattr. 342 for module_name, module in sys.modules.copy().items(): 343 if (module_name == '__main__' 344 or module_name == '__mp_main__' # bpo-42406 345 or module is None): 346 continue 347 try: 348 if _getattribute(module, name)[0] is obj: 349 return module_name 350 except AttributeError: 351 pass 352 return '__main__' 353 354def encode_long(x): 355 r"""Encode a long to a two's complement little-endian binary string. 356 Note that 0 is a special case, returning an empty string, to save a 357 byte in the LONG1 pickling context. 358 359 >>> encode_long(0) 360 b'' 361 >>> encode_long(255) 362 b'\xff\x00' 363 >>> encode_long(32767) 364 b'\xff\x7f' 365 >>> encode_long(-256) 366 b'\x00\xff' 367 >>> encode_long(-32768) 368 b'\x00\x80' 369 >>> encode_long(-128) 370 b'\x80' 371 >>> encode_long(127) 372 b'\x7f' 373 >>> 374 """ 375 if x == 0: 376 return b'' 377 nbytes = (x.bit_length() >> 3) + 1 378 result = x.to_bytes(nbytes, byteorder='little', signed=True) 379 if x < 0 and nbytes > 1: 380 if result[-1] == 0xff and (result[-2] & 0x80) != 0: 381 result = result[:-1] 382 return result 383 384def decode_long(data): 385 r"""Decode a long from a two's complement little-endian binary string. 386 387 >>> decode_long(b'') 388 0 389 >>> decode_long(b"\xff\x00") 390 255 391 >>> decode_long(b"\xff\x7f") 392 32767 393 >>> decode_long(b"\x00\xff") 394 -256 395 >>> decode_long(b"\x00\x80") 396 -32768 397 >>> decode_long(b"\x80") 398 -128 399 >>> decode_long(b"\x7f") 400 127 401 """ 402 return int.from_bytes(data, byteorder='little', signed=True) 403 404 405# Pickling machinery 406 407class _Pickler: 408 409 def __init__(self, file, protocol=None, *, fix_imports=True, 410 buffer_callback=None): 411 """This takes a binary file for writing a pickle data stream. 412 413 The optional *protocol* argument tells the pickler to use the 414 given protocol; supported protocols are 0, 1, 2, 3, 4 and 5. 415 The default protocol is 4. It was introduced in Python 3.4, and 416 is incompatible with previous versions. 417 418 Specifying a negative protocol version selects the highest 419 protocol version supported. The higher the protocol used, the 420 more recent the version of Python needed to read the pickle 421 produced. 422 423 The *file* argument must have a write() method that accepts a 424 single bytes argument. It can thus be a file object opened for 425 binary writing, an io.BytesIO instance, or any other custom 426 object that meets this interface. 427 428 If *fix_imports* is True and *protocol* is less than 3, pickle 429 will try to map the new Python 3 names to the old module names 430 used in Python 2, so that the pickle data stream is readable 431 with Python 2. 432 433 If *buffer_callback* is None (the default), buffer views are 434 serialized into *file* as part of the pickle stream. 435 436 If *buffer_callback* is not None, then it can be called any number 437 of times with a buffer view. If the callback returns a false value 438 (such as None), the given buffer is out-of-band; otherwise the 439 buffer is serialized in-band, i.e. inside the pickle stream. 440 441 It is an error if *buffer_callback* is not None and *protocol* 442 is None or smaller than 5. 443 """ 444 if protocol is None: 445 protocol = DEFAULT_PROTOCOL 446 if protocol < 0: 447 protocol = HIGHEST_PROTOCOL 448 elif not 0 <= protocol <= HIGHEST_PROTOCOL: 449 raise ValueError("pickle protocol must be <= %d" % HIGHEST_PROTOCOL) 450 if buffer_callback is not None and protocol < 5: 451 raise ValueError("buffer_callback needs protocol >= 5") 452 self._buffer_callback = buffer_callback 453 try: 454 self._file_write = file.write 455 except AttributeError: 456 raise TypeError("file must have a 'write' attribute") 457 self.framer = _Framer(self._file_write) 458 self.write = self.framer.write 459 self._write_large_bytes = self.framer.write_large_bytes 460 self.memo = {} 461 self.proto = int(protocol) 462 self.bin = protocol >= 1 463 self.fast = 0 464 self.fix_imports = fix_imports and protocol < 3 465 466 def clear_memo(self): 467 """Clears the pickler's "memo". 468 469 The memo is the data structure that remembers which objects the 470 pickler has already seen, so that shared or recursive objects 471 are pickled by reference and not by value. This method is 472 useful when re-using picklers. 473 """ 474 self.memo.clear() 475 476 def dump(self, obj): 477 """Write a pickled representation of obj to the open file.""" 478 # Check whether Pickler was initialized correctly. This is 479 # only needed to mimic the behavior of _pickle.Pickler.dump(). 480 if not hasattr(self, "_file_write"): 481 raise PicklingError("Pickler.__init__() was not called by " 482 "%s.__init__()" % (self.__class__.__name__,)) 483 if self.proto >= 2: 484 self.write(PROTO + pack("<B", self.proto)) 485 if self.proto >= 4: 486 self.framer.start_framing() 487 self.save(obj) 488 self.write(STOP) 489 self.framer.end_framing() 490 491 def memoize(self, obj): 492 """Store an object in the memo.""" 493 494 # The Pickler memo is a dictionary mapping object ids to 2-tuples 495 # that contain the Unpickler memo key and the object being memoized. 496 # The memo key is written to the pickle and will become 497 # the key in the Unpickler's memo. The object is stored in the 498 # Pickler memo so that transient objects are kept alive during 499 # pickling. 500 501 # The use of the Unpickler memo length as the memo key is just a 502 # convention. The only requirement is that the memo values be unique. 503 # But there appears no advantage to any other scheme, and this 504 # scheme allows the Unpickler memo to be implemented as a plain (but 505 # growable) array, indexed by memo key. 506 if self.fast: 507 return 508 assert id(obj) not in self.memo 509 idx = len(self.memo) 510 self.write(self.put(idx)) 511 self.memo[id(obj)] = idx, obj 512 513 # Return a PUT (BINPUT, LONG_BINPUT) opcode string, with argument i. 514 def put(self, idx): 515 if self.proto >= 4: 516 return MEMOIZE 517 elif self.bin: 518 if idx < 256: 519 return BINPUT + pack("<B", idx) 520 else: 521 return LONG_BINPUT + pack("<I", idx) 522 else: 523 return PUT + repr(idx).encode("ascii") + b'\n' 524 525 # Return a GET (BINGET, LONG_BINGET) opcode string, with argument i. 526 def get(self, i): 527 if self.bin: 528 if i < 256: 529 return BINGET + pack("<B", i) 530 else: 531 return LONG_BINGET + pack("<I", i) 532 533 return GET + repr(i).encode("ascii") + b'\n' 534 535 def save(self, obj, save_persistent_id=True): 536 self.framer.commit_frame() 537 538 # Check for persistent id (defined by a subclass) 539 pid = self.persistent_id(obj) 540 if pid is not None and save_persistent_id: 541 self.save_pers(pid) 542 return 543 544 # Check the memo 545 x = self.memo.get(id(obj)) 546 if x is not None: 547 self.write(self.get(x[0])) 548 return 549 550 rv = NotImplemented 551 reduce = getattr(self, "reducer_override", None) 552 if reduce is not None: 553 rv = reduce(obj) 554 555 if rv is NotImplemented: 556 # Check the type dispatch table 557 t = type(obj) 558 f = self.dispatch.get(t) 559 if f is not None: 560 f(self, obj) # Call unbound method with explicit self 561 return 562 563 # Check private dispatch table if any, or else 564 # copyreg.dispatch_table 565 reduce = getattr(self, 'dispatch_table', dispatch_table).get(t) 566 if reduce is not None: 567 rv = reduce(obj) 568 else: 569 # Check for a class with a custom metaclass; treat as regular 570 # class 571 if issubclass(t, type): 572 self.save_global(obj) 573 return 574 575 # Check for a __reduce_ex__ method, fall back to __reduce__ 576 reduce = getattr(obj, "__reduce_ex__", None) 577 if reduce is not None: 578 rv = reduce(self.proto) 579 else: 580 reduce = getattr(obj, "__reduce__", None) 581 if reduce is not None: 582 rv = reduce() 583 else: 584 raise PicklingError("Can't pickle %r object: %r" % 585 (t.__name__, obj)) 586 587 # Check for string returned by reduce(), meaning "save as global" 588 if isinstance(rv, str): 589 self.save_global(obj, rv) 590 return 591 592 # Assert that reduce() returned a tuple 593 if not isinstance(rv, tuple): 594 raise PicklingError("%s must return string or tuple" % reduce) 595 596 # Assert that it returned an appropriately sized tuple 597 l = len(rv) 598 if not (2 <= l <= 6): 599 raise PicklingError("Tuple returned by %s must have " 600 "two to six elements" % reduce) 601 602 # Save the reduce() output and finally memoize the object 603 self.save_reduce(obj=obj, *rv) 604 605 def persistent_id(self, obj): 606 # This exists so a subclass can override it 607 return None 608 609 def save_pers(self, pid): 610 # Save a persistent id reference 611 if self.bin: 612 self.save(pid, save_persistent_id=False) 613 self.write(BINPERSID) 614 else: 615 try: 616 self.write(PERSID + str(pid).encode("ascii") + b'\n') 617 except UnicodeEncodeError: 618 raise PicklingError( 619 "persistent IDs in protocol 0 must be ASCII strings") 620 621 def save_reduce(self, func, args, state=None, listitems=None, 622 dictitems=None, state_setter=None, *, obj=None): 623 # This API is called by some subclasses 624 625 if not isinstance(args, tuple): 626 raise PicklingError("args from save_reduce() must be a tuple") 627 if not callable(func): 628 raise PicklingError("func from save_reduce() must be callable") 629 630 save = self.save 631 write = self.write 632 633 func_name = getattr(func, "__name__", "") 634 if self.proto >= 2 and func_name == "__newobj_ex__": 635 cls, args, kwargs = args 636 if not hasattr(cls, "__new__"): 637 raise PicklingError("args[0] from {} args has no __new__" 638 .format(func_name)) 639 if obj is not None and cls is not obj.__class__: 640 raise PicklingError("args[0] from {} args has the wrong class" 641 .format(func_name)) 642 if self.proto >= 4: 643 save(cls) 644 save(args) 645 save(kwargs) 646 write(NEWOBJ_EX) 647 else: 648 func = partial(cls.__new__, cls, *args, **kwargs) 649 save(func) 650 save(()) 651 write(REDUCE) 652 elif self.proto >= 2 and func_name == "__newobj__": 653 # A __reduce__ implementation can direct protocol 2 or newer to 654 # use the more efficient NEWOBJ opcode, while still 655 # allowing protocol 0 and 1 to work normally. For this to 656 # work, the function returned by __reduce__ should be 657 # called __newobj__, and its first argument should be a 658 # class. The implementation for __newobj__ 659 # should be as follows, although pickle has no way to 660 # verify this: 661 # 662 # def __newobj__(cls, *args): 663 # return cls.__new__(cls, *args) 664 # 665 # Protocols 0 and 1 will pickle a reference to __newobj__, 666 # while protocol 2 (and above) will pickle a reference to 667 # cls, the remaining args tuple, and the NEWOBJ code, 668 # which calls cls.__new__(cls, *args) at unpickling time 669 # (see load_newobj below). If __reduce__ returns a 670 # three-tuple, the state from the third tuple item will be 671 # pickled regardless of the protocol, calling __setstate__ 672 # at unpickling time (see load_build below). 673 # 674 # Note that no standard __newobj__ implementation exists; 675 # you have to provide your own. This is to enforce 676 # compatibility with Python 2.2 (pickles written using 677 # protocol 0 or 1 in Python 2.3 should be unpicklable by 678 # Python 2.2). 679 cls = args[0] 680 if not hasattr(cls, "__new__"): 681 raise PicklingError( 682 "args[0] from __newobj__ args has no __new__") 683 if obj is not None and cls is not obj.__class__: 684 raise PicklingError( 685 "args[0] from __newobj__ args has the wrong class") 686 args = args[1:] 687 save(cls) 688 save(args) 689 write(NEWOBJ) 690 else: 691 save(func) 692 save(args) 693 write(REDUCE) 694 695 if obj is not None: 696 # If the object is already in the memo, this means it is 697 # recursive. In this case, throw away everything we put on the 698 # stack, and fetch the object back from the memo. 699 if id(obj) in self.memo: 700 write(POP + self.get(self.memo[id(obj)][0])) 701 else: 702 self.memoize(obj) 703 704 # More new special cases (that work with older protocols as 705 # well): when __reduce__ returns a tuple with 4 or 5 items, 706 # the 4th and 5th item should be iterators that provide list 707 # items and dict items (as (key, value) tuples), or None. 708 709 if listitems is not None: 710 self._batch_appends(listitems) 711 712 if dictitems is not None: 713 self._batch_setitems(dictitems) 714 715 if state is not None: 716 if state_setter is None: 717 save(state) 718 write(BUILD) 719 else: 720 # If a state_setter is specified, call it instead of load_build 721 # to update obj's with its previous state. 722 # First, push state_setter and its tuple of expected arguments 723 # (obj, state) onto the stack. 724 save(state_setter) 725 save(obj) # simple BINGET opcode as obj is already memoized. 726 save(state) 727 write(TUPLE2) 728 # Trigger a state_setter(obj, state) function call. 729 write(REDUCE) 730 # The purpose of state_setter is to carry-out an 731 # inplace modification of obj. We do not care about what the 732 # method might return, so its output is eventually removed from 733 # the stack. 734 write(POP) 735 736 # Methods below this point are dispatched through the dispatch table 737 738 dispatch = {} 739 740 def save_none(self, obj): 741 self.write(NONE) 742 dispatch[type(None)] = save_none 743 744 def save_bool(self, obj): 745 if self.proto >= 2: 746 self.write(NEWTRUE if obj else NEWFALSE) 747 else: 748 self.write(TRUE if obj else FALSE) 749 dispatch[bool] = save_bool 750 751 def save_long(self, obj): 752 if self.bin: 753 # If the int is small enough to fit in a signed 4-byte 2's-comp 754 # format, we can store it more efficiently than the general 755 # case. 756 # First one- and two-byte unsigned ints: 757 if obj >= 0: 758 if obj <= 0xff: 759 self.write(BININT1 + pack("<B", obj)) 760 return 761 if obj <= 0xffff: 762 self.write(BININT2 + pack("<H", obj)) 763 return 764 # Next check for 4-byte signed ints: 765 if -0x80000000 <= obj <= 0x7fffffff: 766 self.write(BININT + pack("<i", obj)) 767 return 768 if self.proto >= 2: 769 encoded = encode_long(obj) 770 n = len(encoded) 771 if n < 256: 772 self.write(LONG1 + pack("<B", n) + encoded) 773 else: 774 self.write(LONG4 + pack("<i", n) + encoded) 775 return 776 if -0x80000000 <= obj <= 0x7fffffff: 777 self.write(INT + repr(obj).encode("ascii") + b'\n') 778 else: 779 self.write(LONG + repr(obj).encode("ascii") + b'L\n') 780 dispatch[int] = save_long 781 782 def save_float(self, obj): 783 if self.bin: 784 self.write(BINFLOAT + pack('>d', obj)) 785 else: 786 self.write(FLOAT + repr(obj).encode("ascii") + b'\n') 787 dispatch[float] = save_float 788 789 def save_bytes(self, obj): 790 if self.proto < 3: 791 if not obj: # bytes object is empty 792 self.save_reduce(bytes, (), obj=obj) 793 else: 794 self.save_reduce(codecs.encode, 795 (str(obj, 'latin1'), 'latin1'), obj=obj) 796 return 797 n = len(obj) 798 if n <= 0xff: 799 self.write(SHORT_BINBYTES + pack("<B", n) + obj) 800 elif n > 0xffffffff and self.proto >= 4: 801 self._write_large_bytes(BINBYTES8 + pack("<Q", n), obj) 802 elif n >= self.framer._FRAME_SIZE_TARGET: 803 self._write_large_bytes(BINBYTES + pack("<I", n), obj) 804 else: 805 self.write(BINBYTES + pack("<I", n) + obj) 806 self.memoize(obj) 807 dispatch[bytes] = save_bytes 808 809 def save_bytearray(self, obj): 810 if self.proto < 5: 811 if not obj: # bytearray is empty 812 self.save_reduce(bytearray, (), obj=obj) 813 else: 814 self.save_reduce(bytearray, (bytes(obj),), obj=obj) 815 return 816 n = len(obj) 817 if n >= self.framer._FRAME_SIZE_TARGET: 818 self._write_large_bytes(BYTEARRAY8 + pack("<Q", n), obj) 819 else: 820 self.write(BYTEARRAY8 + pack("<Q", n) + obj) 821 self.memoize(obj) 822 dispatch[bytearray] = save_bytearray 823 824 if _HAVE_PICKLE_BUFFER: 825 def save_picklebuffer(self, obj): 826 if self.proto < 5: 827 raise PicklingError("PickleBuffer can only pickled with " 828 "protocol >= 5") 829 with obj.raw() as m: 830 if not m.contiguous: 831 raise PicklingError("PickleBuffer can not be pickled when " 832 "pointing to a non-contiguous buffer") 833 in_band = True 834 if self._buffer_callback is not None: 835 in_band = bool(self._buffer_callback(obj)) 836 if in_band: 837 # Write data in-band 838 # XXX The C implementation avoids a copy here 839 if m.readonly: 840 self.save_bytes(m.tobytes()) 841 else: 842 self.save_bytearray(m.tobytes()) 843 else: 844 # Write data out-of-band 845 self.write(NEXT_BUFFER) 846 if m.readonly: 847 self.write(READONLY_BUFFER) 848 849 dispatch[PickleBuffer] = save_picklebuffer 850 851 def save_str(self, obj): 852 if self.bin: 853 encoded = obj.encode('utf-8', 'surrogatepass') 854 n = len(encoded) 855 if n <= 0xff and self.proto >= 4: 856 self.write(SHORT_BINUNICODE + pack("<B", n) + encoded) 857 elif n > 0xffffffff and self.proto >= 4: 858 self._write_large_bytes(BINUNICODE8 + pack("<Q", n), encoded) 859 elif n >= self.framer._FRAME_SIZE_TARGET: 860 self._write_large_bytes(BINUNICODE + pack("<I", n), encoded) 861 else: 862 self.write(BINUNICODE + pack("<I", n) + encoded) 863 else: 864 obj = obj.replace("\\", "\\u005c") 865 obj = obj.replace("\0", "\\u0000") 866 obj = obj.replace("\n", "\\u000a") 867 obj = obj.replace("\r", "\\u000d") 868 obj = obj.replace("\x1a", "\\u001a") # EOF on DOS 869 self.write(UNICODE + obj.encode('raw-unicode-escape') + 870 b'\n') 871 self.memoize(obj) 872 dispatch[str] = save_str 873 874 def save_tuple(self, obj): 875 if not obj: # tuple is empty 876 if self.bin: 877 self.write(EMPTY_TUPLE) 878 else: 879 self.write(MARK + TUPLE) 880 return 881 882 n = len(obj) 883 save = self.save 884 memo = self.memo 885 if n <= 3 and self.proto >= 2: 886 for element in obj: 887 save(element) 888 # Subtle. Same as in the big comment below. 889 if id(obj) in memo: 890 get = self.get(memo[id(obj)][0]) 891 self.write(POP * n + get) 892 else: 893 self.write(_tuplesize2code[n]) 894 self.memoize(obj) 895 return 896 897 # proto 0 or proto 1 and tuple isn't empty, or proto > 1 and tuple 898 # has more than 3 elements. 899 write = self.write 900 write(MARK) 901 for element in obj: 902 save(element) 903 904 if id(obj) in memo: 905 # Subtle. d was not in memo when we entered save_tuple(), so 906 # the process of saving the tuple's elements must have saved 907 # the tuple itself: the tuple is recursive. The proper action 908 # now is to throw away everything we put on the stack, and 909 # simply GET the tuple (it's already constructed). This check 910 # could have been done in the "for element" loop instead, but 911 # recursive tuples are a rare thing. 912 get = self.get(memo[id(obj)][0]) 913 if self.bin: 914 write(POP_MARK + get) 915 else: # proto 0 -- POP_MARK not available 916 write(POP * (n+1) + get) 917 return 918 919 # No recursion. 920 write(TUPLE) 921 self.memoize(obj) 922 923 dispatch[tuple] = save_tuple 924 925 def save_list(self, obj): 926 if self.bin: 927 self.write(EMPTY_LIST) 928 else: # proto 0 -- can't use EMPTY_LIST 929 self.write(MARK + LIST) 930 931 self.memoize(obj) 932 self._batch_appends(obj) 933 934 dispatch[list] = save_list 935 936 _BATCHSIZE = 1000 937 938 def _batch_appends(self, items): 939 # Helper to batch up APPENDS sequences 940 save = self.save 941 write = self.write 942 943 if not self.bin: 944 for x in items: 945 save(x) 946 write(APPEND) 947 return 948 949 it = iter(items) 950 while True: 951 tmp = list(islice(it, self._BATCHSIZE)) 952 n = len(tmp) 953 if n > 1: 954 write(MARK) 955 for x in tmp: 956 save(x) 957 write(APPENDS) 958 elif n: 959 save(tmp[0]) 960 write(APPEND) 961 # else tmp is empty, and we're done 962 if n < self._BATCHSIZE: 963 return 964 965 def save_dict(self, obj): 966 if self.bin: 967 self.write(EMPTY_DICT) 968 else: # proto 0 -- can't use EMPTY_DICT 969 self.write(MARK + DICT) 970 971 self.memoize(obj) 972 self._batch_setitems(obj.items()) 973 974 dispatch[dict] = save_dict 975 if PyStringMap is not None: 976 dispatch[PyStringMap] = save_dict 977 978 def _batch_setitems(self, items): 979 # Helper to batch up SETITEMS sequences; proto >= 1 only 980 save = self.save 981 write = self.write 982 983 if not self.bin: 984 for k, v in items: 985 save(k) 986 save(v) 987 write(SETITEM) 988 return 989 990 it = iter(items) 991 while True: 992 tmp = list(islice(it, self._BATCHSIZE)) 993 n = len(tmp) 994 if n > 1: 995 write(MARK) 996 for k, v in tmp: 997 save(k) 998 save(v) 999 write(SETITEMS) 1000 elif n: 1001 k, v = tmp[0] 1002 save(k) 1003 save(v) 1004 write(SETITEM) 1005 # else tmp is empty, and we're done 1006 if n < self._BATCHSIZE: 1007 return 1008 1009 def save_set(self, obj): 1010 save = self.save 1011 write = self.write 1012 1013 if self.proto < 4: 1014 self.save_reduce(set, (list(obj),), obj=obj) 1015 return 1016 1017 write(EMPTY_SET) 1018 self.memoize(obj) 1019 1020 it = iter(obj) 1021 while True: 1022 batch = list(islice(it, self._BATCHSIZE)) 1023 n = len(batch) 1024 if n > 0: 1025 write(MARK) 1026 for item in batch: 1027 save(item) 1028 write(ADDITEMS) 1029 if n < self._BATCHSIZE: 1030 return 1031 dispatch[set] = save_set 1032 1033 def save_frozenset(self, obj): 1034 save = self.save 1035 write = self.write 1036 1037 if self.proto < 4: 1038 self.save_reduce(frozenset, (list(obj),), obj=obj) 1039 return 1040 1041 write(MARK) 1042 for item in obj: 1043 save(item) 1044 1045 if id(obj) in self.memo: 1046 # If the object is already in the memo, this means it is 1047 # recursive. In this case, throw away everything we put on the 1048 # stack, and fetch the object back from the memo. 1049 write(POP_MARK + self.get(self.memo[id(obj)][0])) 1050 return 1051 1052 write(FROZENSET) 1053 self.memoize(obj) 1054 dispatch[frozenset] = save_frozenset 1055 1056 def save_global(self, obj, name=None): 1057 write = self.write 1058 memo = self.memo 1059 1060 if name is None: 1061 name = getattr(obj, '__qualname__', None) 1062 if name is None: 1063 name = obj.__name__ 1064 1065 module_name = whichmodule(obj, name) 1066 try: 1067 __import__(module_name, level=0) 1068 module = sys.modules[module_name] 1069 obj2, parent = _getattribute(module, name) 1070 except (ImportError, KeyError, AttributeError): 1071 raise PicklingError( 1072 "Can't pickle %r: it's not found as %s.%s" % 1073 (obj, module_name, name)) from None 1074 else: 1075 if obj2 is not obj: 1076 raise PicklingError( 1077 "Can't pickle %r: it's not the same object as %s.%s" % 1078 (obj, module_name, name)) 1079 1080 if self.proto >= 2: 1081 code = _extension_registry.get((module_name, name)) 1082 if code: 1083 assert code > 0 1084 if code <= 0xff: 1085 write(EXT1 + pack("<B", code)) 1086 elif code <= 0xffff: 1087 write(EXT2 + pack("<H", code)) 1088 else: 1089 write(EXT4 + pack("<i", code)) 1090 return 1091 lastname = name.rpartition('.')[2] 1092 if parent is module: 1093 name = lastname 1094 # Non-ASCII identifiers are supported only with protocols >= 3. 1095 if self.proto >= 4: 1096 self.save(module_name) 1097 self.save(name) 1098 write(STACK_GLOBAL) 1099 elif parent is not module: 1100 self.save_reduce(getattr, (parent, lastname)) 1101 elif self.proto >= 3: 1102 write(GLOBAL + bytes(module_name, "utf-8") + b'\n' + 1103 bytes(name, "utf-8") + b'\n') 1104 else: 1105 if self.fix_imports: 1106 r_name_mapping = _compat_pickle.REVERSE_NAME_MAPPING 1107 r_import_mapping = _compat_pickle.REVERSE_IMPORT_MAPPING 1108 if (module_name, name) in r_name_mapping: 1109 module_name, name = r_name_mapping[(module_name, name)] 1110 elif module_name in r_import_mapping: 1111 module_name = r_import_mapping[module_name] 1112 try: 1113 write(GLOBAL + bytes(module_name, "ascii") + b'\n' + 1114 bytes(name, "ascii") + b'\n') 1115 except UnicodeEncodeError: 1116 raise PicklingError( 1117 "can't pickle global identifier '%s.%s' using " 1118 "pickle protocol %i" % (module, name, self.proto)) from None 1119 1120 self.memoize(obj) 1121 1122 def save_type(self, obj): 1123 if obj is type(None): 1124 return self.save_reduce(type, (None,), obj=obj) 1125 elif obj is type(NotImplemented): 1126 return self.save_reduce(type, (NotImplemented,), obj=obj) 1127 elif obj is type(...): 1128 return self.save_reduce(type, (...,), obj=obj) 1129 return self.save_global(obj) 1130 1131 dispatch[FunctionType] = save_global 1132 dispatch[type] = save_type 1133 1134 1135# Unpickling machinery 1136 1137class _Unpickler: 1138 1139 def __init__(self, file, *, fix_imports=True, 1140 encoding="ASCII", errors="strict", buffers=None): 1141 """This takes a binary file for reading a pickle data stream. 1142 1143 The protocol version of the pickle is detected automatically, so 1144 no proto argument is needed. 1145 1146 The argument *file* must have two methods, a read() method that 1147 takes an integer argument, and a readline() method that requires 1148 no arguments. Both methods should return bytes. Thus *file* 1149 can be a binary file object opened for reading, an io.BytesIO 1150 object, or any other custom object that meets this interface. 1151 1152 The file-like object must have two methods, a read() method 1153 that takes an integer argument, and a readline() method that 1154 requires no arguments. Both methods should return bytes. 1155 Thus file-like object can be a binary file object opened for 1156 reading, a BytesIO object, or any other custom object that 1157 meets this interface. 1158 1159 If *buffers* is not None, it should be an iterable of buffer-enabled 1160 objects that is consumed each time the pickle stream references 1161 an out-of-band buffer view. Such buffers have been given in order 1162 to the *buffer_callback* of a Pickler object. 1163 1164 If *buffers* is None (the default), then the buffers are taken 1165 from the pickle stream, assuming they are serialized there. 1166 It is an error for *buffers* to be None if the pickle stream 1167 was produced with a non-None *buffer_callback*. 1168 1169 Other optional arguments are *fix_imports*, *encoding* and 1170 *errors*, which are used to control compatibility support for 1171 pickle stream generated by Python 2. If *fix_imports* is True, 1172 pickle will try to map the old Python 2 names to the new names 1173 used in Python 3. The *encoding* and *errors* tell pickle how 1174 to decode 8-bit string instances pickled by Python 2; these 1175 default to 'ASCII' and 'strict', respectively. *encoding* can be 1176 'bytes' to read these 8-bit string instances as bytes objects. 1177 """ 1178 self._buffers = iter(buffers) if buffers is not None else None 1179 self._file_readline = file.readline 1180 self._file_read = file.read 1181 self.memo = {} 1182 self.encoding = encoding 1183 self.errors = errors 1184 self.proto = 0 1185 self.fix_imports = fix_imports 1186 1187 def load(self): 1188 """Read a pickled object representation from the open file. 1189 1190 Return the reconstituted object hierarchy specified in the file. 1191 """ 1192 # Check whether Unpickler was initialized correctly. This is 1193 # only needed to mimic the behavior of _pickle.Unpickler.dump(). 1194 if not hasattr(self, "_file_read"): 1195 raise UnpicklingError("Unpickler.__init__() was not called by " 1196 "%s.__init__()" % (self.__class__.__name__,)) 1197 self._unframer = _Unframer(self._file_read, self._file_readline) 1198 self.read = self._unframer.read 1199 self.readinto = self._unframer.readinto 1200 self.readline = self._unframer.readline 1201 self.metastack = [] 1202 self.stack = [] 1203 self.append = self.stack.append 1204 self.proto = 0 1205 read = self.read 1206 dispatch = self.dispatch 1207 try: 1208 while True: 1209 key = read(1) 1210 if not key: 1211 raise EOFError 1212 assert isinstance(key, bytes_types) 1213 dispatch[key[0]](self) 1214 except _Stop as stopinst: 1215 return stopinst.value 1216 1217 # Return a list of items pushed in the stack after last MARK instruction. 1218 def pop_mark(self): 1219 items = self.stack 1220 self.stack = self.metastack.pop() 1221 self.append = self.stack.append 1222 return items 1223 1224 def persistent_load(self, pid): 1225 raise UnpicklingError("unsupported persistent id encountered") 1226 1227 dispatch = {} 1228 1229 def load_proto(self): 1230 proto = self.read(1)[0] 1231 if not 0 <= proto <= HIGHEST_PROTOCOL: 1232 raise ValueError("unsupported pickle protocol: %d" % proto) 1233 self.proto = proto 1234 dispatch[PROTO[0]] = load_proto 1235 1236 def load_frame(self): 1237 frame_size, = unpack('<Q', self.read(8)) 1238 if frame_size > sys.maxsize: 1239 raise ValueError("frame size > sys.maxsize: %d" % frame_size) 1240 self._unframer.load_frame(frame_size) 1241 dispatch[FRAME[0]] = load_frame 1242 1243 def load_persid(self): 1244 try: 1245 pid = self.readline()[:-1].decode("ascii") 1246 except UnicodeDecodeError: 1247 raise UnpicklingError( 1248 "persistent IDs in protocol 0 must be ASCII strings") 1249 self.append(self.persistent_load(pid)) 1250 dispatch[PERSID[0]] = load_persid 1251 1252 def load_binpersid(self): 1253 pid = self.stack.pop() 1254 self.append(self.persistent_load(pid)) 1255 dispatch[BINPERSID[0]] = load_binpersid 1256 1257 def load_none(self): 1258 self.append(None) 1259 dispatch[NONE[0]] = load_none 1260 1261 def load_false(self): 1262 self.append(False) 1263 dispatch[NEWFALSE[0]] = load_false 1264 1265 def load_true(self): 1266 self.append(True) 1267 dispatch[NEWTRUE[0]] = load_true 1268 1269 def load_int(self): 1270 data = self.readline() 1271 if data == FALSE[1:]: 1272 val = False 1273 elif data == TRUE[1:]: 1274 val = True 1275 else: 1276 val = int(data, 0) 1277 self.append(val) 1278 dispatch[INT[0]] = load_int 1279 1280 def load_binint(self): 1281 self.append(unpack('<i', self.read(4))[0]) 1282 dispatch[BININT[0]] = load_binint 1283 1284 def load_binint1(self): 1285 self.append(self.read(1)[0]) 1286 dispatch[BININT1[0]] = load_binint1 1287 1288 def load_binint2(self): 1289 self.append(unpack('<H', self.read(2))[0]) 1290 dispatch[BININT2[0]] = load_binint2 1291 1292 def load_long(self): 1293 val = self.readline()[:-1] 1294 if val and val[-1] == b'L'[0]: 1295 val = val[:-1] 1296 self.append(int(val, 0)) 1297 dispatch[LONG[0]] = load_long 1298 1299 def load_long1(self): 1300 n = self.read(1)[0] 1301 data = self.read(n) 1302 self.append(decode_long(data)) 1303 dispatch[LONG1[0]] = load_long1 1304 1305 def load_long4(self): 1306 n, = unpack('<i', self.read(4)) 1307 if n < 0: 1308 # Corrupt or hostile pickle -- we never write one like this 1309 raise UnpicklingError("LONG pickle has negative byte count") 1310 data = self.read(n) 1311 self.append(decode_long(data)) 1312 dispatch[LONG4[0]] = load_long4 1313 1314 def load_float(self): 1315 self.append(float(self.readline()[:-1])) 1316 dispatch[FLOAT[0]] = load_float 1317 1318 def load_binfloat(self): 1319 self.append(unpack('>d', self.read(8))[0]) 1320 dispatch[BINFLOAT[0]] = load_binfloat 1321 1322 def _decode_string(self, value): 1323 # Used to allow strings from Python 2 to be decoded either as 1324 # bytes or Unicode strings. This should be used only with the 1325 # STRING, BINSTRING and SHORT_BINSTRING opcodes. 1326 if self.encoding == "bytes": 1327 return value 1328 else: 1329 return value.decode(self.encoding, self.errors) 1330 1331 def load_string(self): 1332 data = self.readline()[:-1] 1333 # Strip outermost quotes 1334 if len(data) >= 2 and data[0] == data[-1] and data[0] in b'"\'': 1335 data = data[1:-1] 1336 else: 1337 raise UnpicklingError("the STRING opcode argument must be quoted") 1338 self.append(self._decode_string(codecs.escape_decode(data)[0])) 1339 dispatch[STRING[0]] = load_string 1340 1341 def load_binstring(self): 1342 # Deprecated BINSTRING uses signed 32-bit length 1343 len, = unpack('<i', self.read(4)) 1344 if len < 0: 1345 raise UnpicklingError("BINSTRING pickle has negative byte count") 1346 data = self.read(len) 1347 self.append(self._decode_string(data)) 1348 dispatch[BINSTRING[0]] = load_binstring 1349 1350 def load_binbytes(self): 1351 len, = unpack('<I', self.read(4)) 1352 if len > maxsize: 1353 raise UnpicklingError("BINBYTES exceeds system's maximum size " 1354 "of %d bytes" % maxsize) 1355 self.append(self.read(len)) 1356 dispatch[BINBYTES[0]] = load_binbytes 1357 1358 def load_unicode(self): 1359 self.append(str(self.readline()[:-1], 'raw-unicode-escape')) 1360 dispatch[UNICODE[0]] = load_unicode 1361 1362 def load_binunicode(self): 1363 len, = unpack('<I', self.read(4)) 1364 if len > maxsize: 1365 raise UnpicklingError("BINUNICODE exceeds system's maximum size " 1366 "of %d bytes" % maxsize) 1367 self.append(str(self.read(len), 'utf-8', 'surrogatepass')) 1368 dispatch[BINUNICODE[0]] = load_binunicode 1369 1370 def load_binunicode8(self): 1371 len, = unpack('<Q', self.read(8)) 1372 if len > maxsize: 1373 raise UnpicklingError("BINUNICODE8 exceeds system's maximum size " 1374 "of %d bytes" % maxsize) 1375 self.append(str(self.read(len), 'utf-8', 'surrogatepass')) 1376 dispatch[BINUNICODE8[0]] = load_binunicode8 1377 1378 def load_binbytes8(self): 1379 len, = unpack('<Q', self.read(8)) 1380 if len > maxsize: 1381 raise UnpicklingError("BINBYTES8 exceeds system's maximum size " 1382 "of %d bytes" % maxsize) 1383 self.append(self.read(len)) 1384 dispatch[BINBYTES8[0]] = load_binbytes8 1385 1386 def load_bytearray8(self): 1387 len, = unpack('<Q', self.read(8)) 1388 if len > maxsize: 1389 raise UnpicklingError("BYTEARRAY8 exceeds system's maximum size " 1390 "of %d bytes" % maxsize) 1391 b = bytearray(len) 1392 self.readinto(b) 1393 self.append(b) 1394 dispatch[BYTEARRAY8[0]] = load_bytearray8 1395 1396 def load_next_buffer(self): 1397 if self._buffers is None: 1398 raise UnpicklingError("pickle stream refers to out-of-band data " 1399 "but no *buffers* argument was given") 1400 try: 1401 buf = next(self._buffers) 1402 except StopIteration: 1403 raise UnpicklingError("not enough out-of-band buffers") 1404 self.append(buf) 1405 dispatch[NEXT_BUFFER[0]] = load_next_buffer 1406 1407 def load_readonly_buffer(self): 1408 buf = self.stack[-1] 1409 with memoryview(buf) as m: 1410 if not m.readonly: 1411 self.stack[-1] = m.toreadonly() 1412 dispatch[READONLY_BUFFER[0]] = load_readonly_buffer 1413 1414 def load_short_binstring(self): 1415 len = self.read(1)[0] 1416 data = self.read(len) 1417 self.append(self._decode_string(data)) 1418 dispatch[SHORT_BINSTRING[0]] = load_short_binstring 1419 1420 def load_short_binbytes(self): 1421 len = self.read(1)[0] 1422 self.append(self.read(len)) 1423 dispatch[SHORT_BINBYTES[0]] = load_short_binbytes 1424 1425 def load_short_binunicode(self): 1426 len = self.read(1)[0] 1427 self.append(str(self.read(len), 'utf-8', 'surrogatepass')) 1428 dispatch[SHORT_BINUNICODE[0]] = load_short_binunicode 1429 1430 def load_tuple(self): 1431 items = self.pop_mark() 1432 self.append(tuple(items)) 1433 dispatch[TUPLE[0]] = load_tuple 1434 1435 def load_empty_tuple(self): 1436 self.append(()) 1437 dispatch[EMPTY_TUPLE[0]] = load_empty_tuple 1438 1439 def load_tuple1(self): 1440 self.stack[-1] = (self.stack[-1],) 1441 dispatch[TUPLE1[0]] = load_tuple1 1442 1443 def load_tuple2(self): 1444 self.stack[-2:] = [(self.stack[-2], self.stack[-1])] 1445 dispatch[TUPLE2[0]] = load_tuple2 1446 1447 def load_tuple3(self): 1448 self.stack[-3:] = [(self.stack[-3], self.stack[-2], self.stack[-1])] 1449 dispatch[TUPLE3[0]] = load_tuple3 1450 1451 def load_empty_list(self): 1452 self.append([]) 1453 dispatch[EMPTY_LIST[0]] = load_empty_list 1454 1455 def load_empty_dictionary(self): 1456 self.append({}) 1457 dispatch[EMPTY_DICT[0]] = load_empty_dictionary 1458 1459 def load_empty_set(self): 1460 self.append(set()) 1461 dispatch[EMPTY_SET[0]] = load_empty_set 1462 1463 def load_frozenset(self): 1464 items = self.pop_mark() 1465 self.append(frozenset(items)) 1466 dispatch[FROZENSET[0]] = load_frozenset 1467 1468 def load_list(self): 1469 items = self.pop_mark() 1470 self.append(items) 1471 dispatch[LIST[0]] = load_list 1472 1473 def load_dict(self): 1474 items = self.pop_mark() 1475 d = {items[i]: items[i+1] 1476 for i in range(0, len(items), 2)} 1477 self.append(d) 1478 dispatch[DICT[0]] = load_dict 1479 1480 # INST and OBJ differ only in how they get a class object. It's not 1481 # only sensible to do the rest in a common routine, the two routines 1482 # previously diverged and grew different bugs. 1483 # klass is the class to instantiate, and k points to the topmost mark 1484 # object, following which are the arguments for klass.__init__. 1485 def _instantiate(self, klass, args): 1486 if (args or not isinstance(klass, type) or 1487 hasattr(klass, "__getinitargs__")): 1488 try: 1489 value = klass(*args) 1490 except TypeError as err: 1491 raise TypeError("in constructor for %s: %s" % 1492 (klass.__name__, str(err)), sys.exc_info()[2]) 1493 else: 1494 value = klass.__new__(klass) 1495 self.append(value) 1496 1497 def load_inst(self): 1498 module = self.readline()[:-1].decode("ascii") 1499 name = self.readline()[:-1].decode("ascii") 1500 klass = self.find_class(module, name) 1501 self._instantiate(klass, self.pop_mark()) 1502 dispatch[INST[0]] = load_inst 1503 1504 def load_obj(self): 1505 # Stack is ... markobject classobject arg1 arg2 ... 1506 args = self.pop_mark() 1507 cls = args.pop(0) 1508 self._instantiate(cls, args) 1509 dispatch[OBJ[0]] = load_obj 1510 1511 def load_newobj(self): 1512 args = self.stack.pop() 1513 cls = self.stack.pop() 1514 obj = cls.__new__(cls, *args) 1515 self.append(obj) 1516 dispatch[NEWOBJ[0]] = load_newobj 1517 1518 def load_newobj_ex(self): 1519 kwargs = self.stack.pop() 1520 args = self.stack.pop() 1521 cls = self.stack.pop() 1522 obj = cls.__new__(cls, *args, **kwargs) 1523 self.append(obj) 1524 dispatch[NEWOBJ_EX[0]] = load_newobj_ex 1525 1526 def load_global(self): 1527 module = self.readline()[:-1].decode("utf-8") 1528 name = self.readline()[:-1].decode("utf-8") 1529 klass = self.find_class(module, name) 1530 self.append(klass) 1531 dispatch[GLOBAL[0]] = load_global 1532 1533 def load_stack_global(self): 1534 name = self.stack.pop() 1535 module = self.stack.pop() 1536 if type(name) is not str or type(module) is not str: 1537 raise UnpicklingError("STACK_GLOBAL requires str") 1538 self.append(self.find_class(module, name)) 1539 dispatch[STACK_GLOBAL[0]] = load_stack_global 1540 1541 def load_ext1(self): 1542 code = self.read(1)[0] 1543 self.get_extension(code) 1544 dispatch[EXT1[0]] = load_ext1 1545 1546 def load_ext2(self): 1547 code, = unpack('<H', self.read(2)) 1548 self.get_extension(code) 1549 dispatch[EXT2[0]] = load_ext2 1550 1551 def load_ext4(self): 1552 code, = unpack('<i', self.read(4)) 1553 self.get_extension(code) 1554 dispatch[EXT4[0]] = load_ext4 1555 1556 def get_extension(self, code): 1557 nil = [] 1558 obj = _extension_cache.get(code, nil) 1559 if obj is not nil: 1560 self.append(obj) 1561 return 1562 key = _inverted_registry.get(code) 1563 if not key: 1564 if code <= 0: # note that 0 is forbidden 1565 # Corrupt or hostile pickle. 1566 raise UnpicklingError("EXT specifies code <= 0") 1567 raise ValueError("unregistered extension code %d" % code) 1568 obj = self.find_class(*key) 1569 _extension_cache[code] = obj 1570 self.append(obj) 1571 1572 def find_class(self, module, name): 1573 # Subclasses may override this. 1574 sys.audit('pickle.find_class', module, name) 1575 if self.proto < 3 and self.fix_imports: 1576 if (module, name) in _compat_pickle.NAME_MAPPING: 1577 module, name = _compat_pickle.NAME_MAPPING[(module, name)] 1578 elif module in _compat_pickle.IMPORT_MAPPING: 1579 module = _compat_pickle.IMPORT_MAPPING[module] 1580 __import__(module, level=0) 1581 if self.proto >= 4: 1582 return _getattribute(sys.modules[module], name)[0] 1583 else: 1584 return getattr(sys.modules[module], name) 1585 1586 def load_reduce(self): 1587 stack = self.stack 1588 args = stack.pop() 1589 func = stack[-1] 1590 stack[-1] = func(*args) 1591 dispatch[REDUCE[0]] = load_reduce 1592 1593 def load_pop(self): 1594 if self.stack: 1595 del self.stack[-1] 1596 else: 1597 self.pop_mark() 1598 dispatch[POP[0]] = load_pop 1599 1600 def load_pop_mark(self): 1601 self.pop_mark() 1602 dispatch[POP_MARK[0]] = load_pop_mark 1603 1604 def load_dup(self): 1605 self.append(self.stack[-1]) 1606 dispatch[DUP[0]] = load_dup 1607 1608 def load_get(self): 1609 i = int(self.readline()[:-1]) 1610 try: 1611 self.append(self.memo[i]) 1612 except KeyError: 1613 msg = f'Memo value not found at index {i}' 1614 raise UnpicklingError(msg) from None 1615 dispatch[GET[0]] = load_get 1616 1617 def load_binget(self): 1618 i = self.read(1)[0] 1619 try: 1620 self.append(self.memo[i]) 1621 except KeyError as exc: 1622 msg = f'Memo value not found at index {i}' 1623 raise UnpicklingError(msg) from None 1624 dispatch[BINGET[0]] = load_binget 1625 1626 def load_long_binget(self): 1627 i, = unpack('<I', self.read(4)) 1628 try: 1629 self.append(self.memo[i]) 1630 except KeyError as exc: 1631 msg = f'Memo value not found at index {i}' 1632 raise UnpicklingError(msg) from None 1633 dispatch[LONG_BINGET[0]] = load_long_binget 1634 1635 def load_put(self): 1636 i = int(self.readline()[:-1]) 1637 if i < 0: 1638 raise ValueError("negative PUT argument") 1639 self.memo[i] = self.stack[-1] 1640 dispatch[PUT[0]] = load_put 1641 1642 def load_binput(self): 1643 i = self.read(1)[0] 1644 if i < 0: 1645 raise ValueError("negative BINPUT argument") 1646 self.memo[i] = self.stack[-1] 1647 dispatch[BINPUT[0]] = load_binput 1648 1649 def load_long_binput(self): 1650 i, = unpack('<I', self.read(4)) 1651 if i > maxsize: 1652 raise ValueError("negative LONG_BINPUT argument") 1653 self.memo[i] = self.stack[-1] 1654 dispatch[LONG_BINPUT[0]] = load_long_binput 1655 1656 def load_memoize(self): 1657 memo = self.memo 1658 memo[len(memo)] = self.stack[-1] 1659 dispatch[MEMOIZE[0]] = load_memoize 1660 1661 def load_append(self): 1662 stack = self.stack 1663 value = stack.pop() 1664 list = stack[-1] 1665 list.append(value) 1666 dispatch[APPEND[0]] = load_append 1667 1668 def load_appends(self): 1669 items = self.pop_mark() 1670 list_obj = self.stack[-1] 1671 try: 1672 extend = list_obj.extend 1673 except AttributeError: 1674 pass 1675 else: 1676 extend(items) 1677 return 1678 # Even if the PEP 307 requires extend() and append() methods, 1679 # fall back on append() if the object has no extend() method 1680 # for backward compatibility. 1681 append = list_obj.append 1682 for item in items: 1683 append(item) 1684 dispatch[APPENDS[0]] = load_appends 1685 1686 def load_setitem(self): 1687 stack = self.stack 1688 value = stack.pop() 1689 key = stack.pop() 1690 dict = stack[-1] 1691 dict[key] = value 1692 dispatch[SETITEM[0]] = load_setitem 1693 1694 def load_setitems(self): 1695 items = self.pop_mark() 1696 dict = self.stack[-1] 1697 for i in range(0, len(items), 2): 1698 dict[items[i]] = items[i + 1] 1699 dispatch[SETITEMS[0]] = load_setitems 1700 1701 def load_additems(self): 1702 items = self.pop_mark() 1703 set_obj = self.stack[-1] 1704 if isinstance(set_obj, set): 1705 set_obj.update(items) 1706 else: 1707 add = set_obj.add 1708 for item in items: 1709 add(item) 1710 dispatch[ADDITEMS[0]] = load_additems 1711 1712 def load_build(self): 1713 stack = self.stack 1714 state = stack.pop() 1715 inst = stack[-1] 1716 setstate = getattr(inst, "__setstate__", None) 1717 if setstate is not None: 1718 setstate(state) 1719 return 1720 slotstate = None 1721 if isinstance(state, tuple) and len(state) == 2: 1722 state, slotstate = state 1723 if state: 1724 inst_dict = inst.__dict__ 1725 intern = sys.intern 1726 for k, v in state.items(): 1727 if type(k) is str: 1728 inst_dict[intern(k)] = v 1729 else: 1730 inst_dict[k] = v 1731 if slotstate: 1732 for k, v in slotstate.items(): 1733 setattr(inst, k, v) 1734 dispatch[BUILD[0]] = load_build 1735 1736 def load_mark(self): 1737 self.metastack.append(self.stack) 1738 self.stack = [] 1739 self.append = self.stack.append 1740 dispatch[MARK[0]] = load_mark 1741 1742 def load_stop(self): 1743 value = self.stack.pop() 1744 raise _Stop(value) 1745 dispatch[STOP[0]] = load_stop 1746 1747 1748# Shorthands 1749 1750def _dump(obj, file, protocol=None, *, fix_imports=True, buffer_callback=None): 1751 _Pickler(file, protocol, fix_imports=fix_imports, 1752 buffer_callback=buffer_callback).dump(obj) 1753 1754def _dumps(obj, protocol=None, *, fix_imports=True, buffer_callback=None): 1755 f = io.BytesIO() 1756 _Pickler(f, protocol, fix_imports=fix_imports, 1757 buffer_callback=buffer_callback).dump(obj) 1758 res = f.getvalue() 1759 assert isinstance(res, bytes_types) 1760 return res 1761 1762def _load(file, *, fix_imports=True, encoding="ASCII", errors="strict", 1763 buffers=None): 1764 return _Unpickler(file, fix_imports=fix_imports, buffers=buffers, 1765 encoding=encoding, errors=errors).load() 1766 1767def _loads(s, /, *, fix_imports=True, encoding="ASCII", errors="strict", 1768 buffers=None): 1769 if isinstance(s, str): 1770 raise TypeError("Can't load pickle from unicode string") 1771 file = io.BytesIO(s) 1772 return _Unpickler(file, fix_imports=fix_imports, buffers=buffers, 1773 encoding=encoding, errors=errors).load() 1774 1775# Use the faster _pickle if possible 1776try: 1777 from _pickle import ( 1778 PickleError, 1779 PicklingError, 1780 UnpicklingError, 1781 Pickler, 1782 Unpickler, 1783 dump, 1784 dumps, 1785 load, 1786 loads 1787 ) 1788except ImportError: 1789 Pickler, Unpickler = _Pickler, _Unpickler 1790 dump, dumps, load, loads = _dump, _dumps, _load, _loads 1791 1792# Doctest 1793def _test(): 1794 import doctest 1795 return doctest.testmod() 1796 1797if __name__ == "__main__": 1798 import argparse 1799 parser = argparse.ArgumentParser( 1800 description='display contents of the pickle files') 1801 parser.add_argument( 1802 'pickle_file', type=argparse.FileType('br'), 1803 nargs='*', help='the pickle file') 1804 parser.add_argument( 1805 '-t', '--test', action='store_true', 1806 help='run self-test suite') 1807 parser.add_argument( 1808 '-v', action='store_true', 1809 help='run verbosely; only affects self-test run') 1810 args = parser.parse_args() 1811 if args.test: 1812 _test() 1813 else: 1814 if not args.pickle_file: 1815 parser.print_help() 1816 else: 1817 import pprint 1818 for f in args.pickle_file: 1819 obj = load(f) 1820 pprint.pprint(obj) 1821