"""\ Pickling Algorithm ------------------ This module implements a basic but powerful algorithm for "pickling" (a.k.a. serializing, marshalling or flattening) nearly arbitrary Python objects. This is a more primitive notion than persistency -- although pickle reads and writes file objects, it does not handle the issue of naming persistent objects, nor the (even more complicated) area of concurrent access to persistent objects. The pickle module can transform a complex object into a byte stream and it can transform the byte stream into an object with the same internal structure. The most obvious thing to do with these byte streams is to write them onto a file, but it is also conceivable to send them across a network or store them in a database. Unlike the built-in marshal module, pickle handles the following correctly: - recursive objects - pointer sharing - classes and class instances Pickle is Python-specific. This has the advantage that there are no restrictions imposed by external standards such as CORBA (which probably can't represent pointer sharing or recursive objects); however it means that non-Python programs may not be able to reconstruct pickled Python objects. Pickle uses a printable ASCII representation. This is slightly more voluminous than a binary representation. However, small integers actually take *less* space when represented as minimal-size decimal strings than when represented as 32-bit binary numbers, and strings are only much longer if they contain control characters or 8-bit characters. The big advantage of using printable ASCII (and of some other characteristics of pickle's representation) is that for debugging or recovery purposes it is possible for a human to read the pickled file with a standard text editor. (I could have gone a step further and used a notation like S-expressions, but the parser would have been considerably more complicated and slower, and the files would probably have become much larger.) Pickle doesn't handle code objects, which marshal does. I suppose pickle could, and maybe it should, but there's probably no great need for it right now (as long as marshal continues to be used for reading and writing code objects), and at least this avoids the possibility of smuggling Trojan horses into a program. For the benefit of persistency modules written using pickle, it supports the notion of a reference to an object outside the pickled data stream. Such objects are referenced by a name, which is an arbitrary string of printable ASCII characters. The resolution of such names is not defined by the pickle module -- the persistent object module will have to implement a method "persistent_load". To write references to persistent objects, the persistent module must define a method "persistent_id" which returns either None or the persistent ID of the object. There are some restrictions on the pickling of class instances. First of all, the class must be defined at the top level in a module. Next, it must normally be possible to create class instances by calling the class without arguments. Usually, this is best accomplished by providing default values for all arguments to its __init__ method (if it has one). If this is undesirable, the class can define a method __getinitargs__, which should return a *tuple* containing the arguments to be passed to the class constructor. Classes can influence how their instances are pickled -- if the class defines the method __getstate__, it is called and the return state is pickled as the contents for the instance, and if the class defines the method __setstate__, it is called with the unpickled state. (Note that these methods can also be used to implement copying class instances.) If there is no __getstate__ method, the instance's __dict__ is pickled. If there is no __setstate__ method, the pickled object must be a dictionary and its items are assigned to the new instance's dictionary. (If a class defines both __getstate__ and __setstate__, the state object needn't be a dictionary -- these methods can do what they want.) Note that when class instances are pickled, their class's code and data is not pickled along with them. Only the instance data is pickled. This is done on purpose, so you can fix bugs in a class or add methods and still load objects that were created with an earlier version of the class. If you plan to have long-lived objects that will see many versions of a class, it may be worth to put a version number in the objects so that suitable conversions can be made by the class's __setstate__ method. The interface is as follows: To pickle an object x onto a file f, open for writing: p = pickle.Pickler(f) p.dump(x) To unpickle an object x from a file f, open for reading: u = pickle.Unpickler(f) x = u.load() The Pickler class only calls the method f.write with a string argument (XXX possibly the interface should pass f.write instead of f). The Unpickler calls the methods f.read(with an integer argument) and f.readline(without argument), both returning a string. It is explicitly allowed to pass non-file objects here, as long as they have the right methods. The following types can be pickled: - None - integers, long integers, floating point numbers - strings - tuples, lists and dictionaries containing only picklable objects - class instances whose __dict__ or __setstate__() is picklable - classes Attempts to pickle unpicklable objects will raise an exception after having written an unspecified number of bytes to the file argument. It is possible to make multiple calls to Pickler.dump() or to Unpickler.load(), as long as there is a one-to-one correspondence between pickler and Unpickler objects and between dump and load calls for any pair of corresponding Pickler and Unpicklers. WARNING: this is intended for pickleing multiple objects without intervening modifications to the objects or their parts. If you modify an object and then pickle it again using the same Pickler instance, the object is not pickled again -- a reference to it is pickled and the Unpickler will return the old value, not the modified one. (XXX There are two problems here: (a) detecting changes, and (b) marshalling a minimal set of changes. I have no answers. Garbage Collection may also become a problem here.) """ __version__ = "1.7" # Code version from types import * from copy_reg import * import string, marshal format_version = "1.2" # File format version we write compatible_formats = ["1.0", "1.1"] # Old format versions we can read mdumps = marshal.dumps mloads = marshal.loads PicklingError = "pickle.PicklingError" UnpicklingError = "pickle.UnpicklingError" MARK = '(' STOP = '.' POP = '0' POP_MARK = '1' DUP = '2' FLOAT = 'F' INT = 'I' BININT = 'J' BININT1 = 'K' LONG = 'L' BININT2 = 'M' NONE = 'N' PERSID = 'P' BINPERSID = 'Q' REDUCE = 'R' STRING = 'S' BINSTRING = 'T' SHORT_BINSTRING = 'U' APPEND = 'a' BUILD = 'b' GLOBAL = 'c' DICT = 'd' EMPTY_DICT = '}' APPENDS = 'e' GET = 'g' BINGET = 'h' INST = 'i' LONG_BINGET = 'j' LIST = 'l' EMPTY_LIST = ']' OBJ = 'o' PUT = 'p' BINPUT = 'q' LONG_BINPUT = 'r' SETITEM = 's' TUPLE = 't' EMPTY_TUPLE = ')' SETITEMS = 'u' class Pickler: def __init__(self, file, bin = 0): self.write = file.write self.memo = {} self.bin = bin def dump(self, object): self.save(object) self.write(STOP) def dump_special(self, callable, args, state = None): if (type(args) is not TupleType): raise PicklingError, "Second argument to dump_special " \ "must be a tuple" self.save_reduce(callable, args, state) self.write(STOP) def put(self, i): if (self.bin): s = mdumps(i)[1:] if (i < 256): return BINPUT + s[0] return LONG_BINPUT + s return PUT + `i` + '\n' def get(self, i): if (self.bin): s = mdumps(i)[1:] if (i < 256): return BINGET + s[0] return LONG_BINGET + s return GET + `i` + '\n' def save(self, object, pers_save = 0): memo = self.memo if (not pers_save): pid = self.persistent_id(object) if (pid is not None): self.save_pers(pid) return d = id(object) t = type(object) if ((t is TupleType) and (len(object) == 0)): if (self.bin): self.save_empty_tuple(object) else: self.save_tuple(object) return if memo.has_key(d): self.write(self.get(memo[d][0])) return try: f = self.dispatch[t] except KeyError: pid = self.inst_persistent_id(object) if pid is not None: self.save_pers(pid) return try: reduce = dispatch_table[t] except KeyError: try: reduce = object.__reduce__ except AttributeError: raise PicklingError, \ "can't pickle %s objects" % `t.__name__` else: tup = reduce() else: tup = reduce(object) if (type(tup) is not TupleType): raise PicklingError, "Value returned by %s must be a " \ "tuple" % reduce l = len(tup) if ((l != 2) and (l != 3)): raise PicklingError, "tuple returned by %s must contain " \ "only two or three elements" % reduce callable = tup[0] arg_tup = tup[1] if (l > 2): state = tup[2] else: state = None if (type(arg_tup) is not TupleType): raise PicklingError, "Second element of tuple returned " \ "by %s must be a tuple" % reduce self.save_reduce(callable, arg_tup, state) return f(self, object) def persistent_id(self, object): return None def inst_persistent_id(self, object): return None def save_pers(self, pid): if (not self.bin): self.write(PERSID + str(pid) + '\n') else: self.save(pid, 1) self.write(BINPERSID) def save_reduce(self, callable, arg_tup, state = None): write = self.write save = self.save save(callable) save(arg_tup) write(REDUCE) if (state is not None): save(state) write(BUILD) dispatch = {} def save_none(self, object): self.write(NONE) dispatch[NoneType] = save_none def save_int(self, object): if (self.bin): i = mdumps(object)[1:] if (i[-2:] == '\000\000'): if (i[-3] == '\000'): self.write(BININT1 + i[:-3]) return self.write(BININT2 + i[:-2]) return self.write(BININT + i) else: self.write(INT + `object` + '\n') dispatch[IntType] = save_int def save_long(self, object): self.write(LONG + `object` + '\n') dispatch[LongType] = save_long def save_float(self, object): self.write(FLOAT + `object` + '\n') dispatch[FloatType] = save_float def save_string(self, object): d = id(object) memo = self.memo if (self.bin): l = len(object) s = mdumps(l)[1:] if (l < 256): self.write(SHORT_BINSTRING + s[0] + object) else: self.write(BINSTRING + s + object) else: self.write(STRING + `object` + '\n') memo_len = len(memo) self.write(self.put(memo_len)) memo[d] = (memo_len, object) dispatch[StringType] = save_string def save_tuple(self, object): write = self.write save = self.save memo = self.memo d = id(object) write(MARK) for element in object: save(element) if (len(object) and memo.has_key(d)): if (self.bin): write(POP_MARK + self.get(memo[d][0])) return write(POP * (len(object) + 1) + self.get(mem[d][0])) return memo_len = len(memo) self.write(TUPLE + self.put(memo_len)) memo[d] = (memo_len, object) dispatch[TupleType] = save_tuple def save_empty_tuple(self, object): self.write(EMPTY_TUPLE) def save_list(self, object): d = id(object) write = self.write save = self.save memo = self.memo if (self.bin): write(EMPTY_LIST) else: write(MARK + LIST) memo_len = len(memo) write(self.put(memo_len)) memo[d] = (memo_len, object) using_appends = (self.bin and (len(object) > 1)) if (using_appends): write(MARK) for element in object: save(element) if (not using_appends): write(APPEND) if (using_appends): write(APPENDS) dispatch[ListType] = save_list def save_dict(self, object): d = id(object) write = self.write save = self.save memo = self.memo if (self.bin): write(EMPTY_DICT) else: write(MARK + DICT) memo_len = len(memo) self.write(self.put(memo_len)) memo[d] = (memo_len, object) using_setitems = (self.bin and (len(object) > 1)) if (using_setitems): write(MARK) items = object.items() for key, value in items: save(key) save(value) if (not using_setitems): write(SETITEM) if (using_setitems): write(SETITEMS) dispatch[DictionaryType] = save_dict def save_inst(self, object): d = id(object) cls = object.__class__ memo = self.memo write = self.write save = self.save if hasattr(object, '__getinitargs__'): args = object.__getinitargs__() len(args) # XXX Assert it's a sequence else: args = () write(MARK) if (self.bin): save(cls) for arg in args: save(arg) memo_len = len(memo) if (self.bin): write(OBJ + self.put(memo_len)) else: module = whichmodule(cls, cls.__name__) name = cls.__name__ write(INST + module + '\n' + name + '\n' + self.put(memo_len)) memo[d] = (memo_len, object) try: getstate = object.__getstate__ except AttributeError: stuff = object.__dict__ else: stuff = getstate() save(stuff) write(BUILD) dispatch[InstanceType] = save_inst def save_global(self, object, name = None): write = self.write memo = self.memo if (name is None): name = object.__name__ module = whichmodule(object, name) memo_len = len(memo) write(GLOBAL + module + '\n' + name + '\n' + self.put(memo_len)) memo[id(object)] = (memo_len, object) dispatch[ClassType] = save_global dispatch[FunctionType] = save_global dispatch[BuiltinFunctionType] = save_global classmap = {} def whichmodule(cls, clsname): """Figure out the module in which a class occurs. Search sys.modules for the module. Cache in classmap. Return a module name. If the class cannot be found, return __main__. """ if classmap.has_key(cls): return classmap[cls] import sys for name, module in sys.modules.items(): if hasattr(module, clsname) and \ getattr(module, clsname) is cls: break else: name = '__main__' classmap[cls] = name return name class Unpickler: def __init__(self, file): self.readline = file.readline self.read = file.read self.memo = {} def load(self): self.mark = ['spam'] # Any new unique object self.stack = [] self.append = self.stack.append read = self.read dispatch = self.dispatch try: while 1: key = read(1) dispatch[key](self) except STOP, value: return value def marker(self): stack = self.stack mark = self.mark k = len(stack)-1 while stack[k] is not mark: k = k-1 return k dispatch = {} def load_eof(self): raise EOFError dispatch[''] = load_eof def load_persid(self): pid = self.readline()[:-1] self.append(self.persistent_load(pid)) dispatch[PERSID] = load_persid def load_binpersid(self): stack = self.stack pid = stack[-1] del stack[-1] self.append(self.persistent_load(pid)) dispatch[BINPERSID] = load_binpersid def load_none(self): self.append(None) dispatch[NONE] = load_none def load_int(self): self.append(string.atoi(self.readline()[:-1], 0)) dispatch[INT] = load_int def load_binint(self): self.append(mloads('i' + self.read(4))) dispatch[BININT] = load_binint def load_binint1(self): self.append(mloads('i' + self.read(1) + '\000\000\000')) dispatch[BININT1] = load_binint1 def load_binint2(self): self.append(mloads('i' + self.read(2) + '\000\000')) dispatch[BININT2] = load_binint2 def load_long(self): self.append(string.atol(self.readline()[:-1], 0)) dispatch[LONG] = load_long def load_float(self): self.append(string.atof(self.readline()[:-1])) dispatch[FLOAT] = load_float def load_string(self): self.append(eval(self.readline()[:-1], {'__builtins__': {}})) # Let's be careful dispatch[STRING] = load_string def load_binstring(self): len = mloads('i' + self.read(4)) self.append(self.read(len)) dispatch[BINSTRING] = load_binstring def load_short_binstring(self): len = mloads('i' + self.read(1) + '\000\000\000') self.append(self.read(len)) dispatch[SHORT_BINSTRING] = load_short_binstring def load_tuple(self): k = self.marker() self.stack[k:] = [tuple(self.stack[k+1:])] dispatch[TUPLE] = load_tuple def load_empty_tuple(self): self.stack.append(()) dispatch[EMPTY_TUPLE] = load_empty_tuple def load_empty_list(self): self.stack.append([]) dispatch[EMPTY_LIST] = load_empty_list def load_empty_dictionary(self): self.stack.append({}) dispatch[EMPTY_DICT] = load_empty_dictionary def load_list(self): k = self.marker() self.stack[k:] = [self.stack[k+1:]] dispatch[LIST] = load_list def load_dict(self): k = self.marker() d = {} items = self.stack[k+1:] for i in range(0, len(items), 2): key = items[i] value = items[i+1] d[key] = value self.stack[k:] = [d] dispatch[DICT] = load_dict def load_inst(self): k = self.marker() args = tuple(self.stack[k+1:]) del self.stack[k:] module = self.readline()[:-1] name = self.readline()[:-1] klass = self.find_class(module, name) if (type(klass) is not ClassType): raise SystemError, "Imported object %s from module %s is " \ "not a class" % (name, module) value = apply(klass, args) self.append(value) dispatch[INST] = load_inst def load_obj(self): stack = self.stack k = self.marker() klass = stack[k + 1] del stack[k + 1] args = tuple(stack[k + 1:]) del stack[k:] value = apply(klass, args) self.append(value) dispatch[OBJ] = load_obj def load_global(self): module = self.readline()[:-1] name = self.readline()[:-1] klass = self.find_class(module, name) self.append(klass) dispatch[GLOBAL] = load_global def find_class(self, module, name): env = {} try: exec 'from %s import %s' % (module, name) in env except ImportError: raise SystemError, \ "Failed to import class %s from module %s" % \ (name, module) klass = env[name] return klass def load_reduce(self): stack = self.stack callable = stack[-2] arg_tup = stack[-1] del stack[-2:] if (type(callable) is not ClassType): if (not safe_constructors.has_key(callable)): try: safe = callable.__safe_for_unpickling__ except AttributeError: safe = None if (not safe): raise UnpicklingError, "%s is not safe for " \ "unpickling" % callable value = apply(callable, arg_tup) self.append(value) dispatch[REDUCE] = load_reduce def load_pop(self): del self.stack[-1] dispatch[POP] = load_pop def load_pop_mark(self): k = self.marker() del self.stack[k:] dispatch[POP_MARK] = load_pop_mark def load_dup(self): self.append(stack[-1]) dispatch[DUP] = load_dup def load_get(self): self.append(self.memo[self.readline()[:-1]]) dispatch[GET] = load_get def load_binget(self): i = mloads('i' + self.read(1) + '\000\000\000') self.append(self.memo[`i`]) dispatch[BINGET] = load_binget def load_long_binget(self): i = mloads('i' + self.read(4)) self.append(self.memo[`i`]) dispatch[LONG_BINGET] = load_long_binget def load_put(self): self.memo[self.readline()[:-1]] = self.stack[-1] dispatch[PUT] = load_put def load_binput(self): i = mloads('i' + self.read(1) + '\000\000\000') self.memo[`i`] = self.stack[-1] dispatch[BINPUT] = load_binput def load_long_binput(self): i = mloads('i' + self.read(4)) self.memo[`i`] = self.stack[-1] dispatch[LONG_BINPUT] = load_long_binput def load_append(self): stack = self.stack value = stack[-1] del stack[-1] list = stack[-1] list.append(value) dispatch[APPEND] = load_append def load_appends(self): stack = self.stack mark = self.marker() list = stack[mark - 1] for i in range(mark + 1, len(stack)): list.append(stack[i]) del stack[mark:] dispatch[APPENDS] = load_appends def load_setitem(self): stack = self.stack value = stack[-1] key = stack[-2] del stack[-2:] dict = stack[-1] dict[key] = value dispatch[SETITEM] = load_setitem def load_setitems(self): stack = self.stack mark = self.marker() dict = stack[mark - 1] for i in range(mark + 1, len(stack), 2): dict[stack[i]] = stack[i + 1] del stack[mark:] dispatch[SETITEMS] = load_setitems def load_build(self): stack = self.stack value = stack[-1] del stack[-1] inst = stack[-1] try: setstate = inst.__setstate__ except AttributeError: for key in value.keys(): setattr(inst, key, value[key]) else: setstate(value) dispatch[BUILD] = load_build def load_mark(self): self.append(self.mark) dispatch[MARK] = load_mark def load_stop(self): value = self.stack[-1] del self.stack[-1] raise STOP, value dispatch[STOP] = load_stop # Shorthands from StringIO import StringIO def dump(object, file, bin = 0): Pickler(file, bin).dump(object) def dumps(object, bin = 0): file = StringIO() Pickler(file, bin).dump(object) return file.getvalue() def load(file): return Unpickler(file).load() def loads(str): file = StringIO(str) return Unpickler(file).load() # The rest is used for testing only class C: def __cmp__(self, other): return cmp(self.__dict__, other.__dict__) def test(): fn = 'out' c = C() c.foo = 1 c.bar = 2 x = [0, 1, 2, 3] y = ('abc', 'abc', c, c) x.append(y) x.append(y) x.append(5) f = open(fn, 'w') F = Pickler(f) F.dump(x) f.close() f = open(fn, 'r') U = Unpickler(f) x2 = U.load() print x print x2 print x == x2 print map(id, x) print map(id, x2) print F.memo print U.memo if __name__ == '__main__': test()