1 | /* pickle accelerator C extensor: _pickle module. |
2 | * |
3 | * It is built as a built-in module (Py_BUILD_CORE_BUILTIN define) on Windows |
4 | * and as an extension module (Py_BUILD_CORE_MODULE define) on other |
5 | * platforms. */ |
6 | |
7 | #if !defined(Py_BUILD_CORE_BUILTIN) && !defined(Py_BUILD_CORE_MODULE) |
8 | # error "Py_BUILD_CORE_BUILTIN or Py_BUILD_CORE_MODULE must be defined" |
9 | #endif |
10 | |
11 | #include "Python.h" |
12 | #include "pycore_moduleobject.h" // _PyModule_GetState() |
13 | #include "structmember.h" // PyMemberDef |
14 | |
15 | PyDoc_STRVAR(pickle_module_doc, |
16 | "Optimized C implementation for the Python pickle module." ); |
17 | |
18 | /*[clinic input] |
19 | module _pickle |
20 | class _pickle.Pickler "PicklerObject *" "&Pickler_Type" |
21 | class _pickle.PicklerMemoProxy "PicklerMemoProxyObject *" "&PicklerMemoProxyType" |
22 | class _pickle.Unpickler "UnpicklerObject *" "&Unpickler_Type" |
23 | class _pickle.UnpicklerMemoProxy "UnpicklerMemoProxyObject *" "&UnpicklerMemoProxyType" |
24 | [clinic start generated code]*/ |
25 | /*[clinic end generated code: output=da39a3ee5e6b4b0d input=4b3e113468a58e6c]*/ |
26 | |
27 | /* Bump HIGHEST_PROTOCOL when new opcodes are added to the pickle protocol. |
28 | Bump DEFAULT_PROTOCOL only when the oldest still supported version of Python |
29 | already includes it. */ |
30 | enum { |
31 | HIGHEST_PROTOCOL = 5, |
32 | DEFAULT_PROTOCOL = 4 |
33 | }; |
34 | |
35 | /* Pickle opcodes. These must be kept updated with pickle.py. |
36 | Extensive docs are in pickletools.py. */ |
37 | enum opcode { |
38 | MARK = '(', |
39 | STOP = '.', |
40 | POP = '0', |
41 | POP_MARK = '1', |
42 | DUP = '2', |
43 | FLOAT = 'F', |
44 | INT = 'I', |
45 | BININT = 'J', |
46 | BININT1 = 'K', |
47 | LONG = 'L', |
48 | BININT2 = 'M', |
49 | NONE = 'N', |
50 | PERSID = 'P', |
51 | BINPERSID = 'Q', |
52 | REDUCE = 'R', |
53 | STRING = 'S', |
54 | BINSTRING = 'T', |
55 | SHORT_BINSTRING = 'U', |
56 | UNICODE = 'V', |
57 | BINUNICODE = 'X', |
58 | APPEND = 'a', |
59 | BUILD = 'b', |
60 | GLOBAL = 'c', |
61 | DICT = 'd', |
62 | EMPTY_DICT = '}', |
63 | APPENDS = 'e', |
64 | GET = 'g', |
65 | BINGET = 'h', |
66 | INST = 'i', |
67 | LONG_BINGET = 'j', |
68 | LIST = 'l', |
69 | EMPTY_LIST = ']', |
70 | OBJ = 'o', |
71 | PUT = 'p', |
72 | BINPUT = 'q', |
73 | LONG_BINPUT = 'r', |
74 | SETITEM = 's', |
75 | TUPLE = 't', |
76 | EMPTY_TUPLE = ')', |
77 | SETITEMS = 'u', |
78 | BINFLOAT = 'G', |
79 | |
80 | /* Protocol 2. */ |
81 | PROTO = '\x80', |
82 | NEWOBJ = '\x81', |
83 | EXT1 = '\x82', |
84 | EXT2 = '\x83', |
85 | EXT4 = '\x84', |
86 | TUPLE1 = '\x85', |
87 | TUPLE2 = '\x86', |
88 | TUPLE3 = '\x87', |
89 | NEWTRUE = '\x88', |
90 | NEWFALSE = '\x89', |
91 | LONG1 = '\x8a', |
92 | LONG4 = '\x8b', |
93 | |
94 | /* Protocol 3 (Python 3.x) */ |
95 | BINBYTES = 'B', |
96 | SHORT_BINBYTES = 'C', |
97 | |
98 | /* Protocol 4 */ |
99 | SHORT_BINUNICODE = '\x8c', |
100 | BINUNICODE8 = '\x8d', |
101 | BINBYTES8 = '\x8e', |
102 | EMPTY_SET = '\x8f', |
103 | ADDITEMS = '\x90', |
104 | FROZENSET = '\x91', |
105 | NEWOBJ_EX = '\x92', |
106 | STACK_GLOBAL = '\x93', |
107 | MEMOIZE = '\x94', |
108 | FRAME = '\x95', |
109 | |
110 | /* Protocol 5 */ |
111 | BYTEARRAY8 = '\x96', |
112 | NEXT_BUFFER = '\x97', |
113 | READONLY_BUFFER = '\x98' |
114 | }; |
115 | |
116 | enum { |
117 | /* Keep in synch with pickle.Pickler._BATCHSIZE. This is how many elements |
118 | batch_list/dict() pumps out before doing APPENDS/SETITEMS. Nothing will |
119 | break if this gets out of synch with pickle.py, but it's unclear that would |
120 | help anything either. */ |
121 | BATCHSIZE = 1000, |
122 | |
123 | /* Nesting limit until Pickler, when running in "fast mode", starts |
124 | checking for self-referential data-structures. */ |
125 | FAST_NESTING_LIMIT = 50, |
126 | |
127 | /* Initial size of the write buffer of Pickler. */ |
128 | WRITE_BUF_SIZE = 4096, |
129 | |
130 | /* Prefetch size when unpickling (disabled on unpeekable streams) */ |
131 | PREFETCH = 8192 * 16, |
132 | |
133 | FRAME_SIZE_MIN = 4, |
134 | FRAME_SIZE_TARGET = 64 * 1024, |
135 | = 9 |
136 | }; |
137 | |
138 | /*************************************************************************/ |
139 | |
140 | /* State of the pickle module, per PEP 3121. */ |
141 | typedef struct { |
142 | /* Exception classes for pickle. */ |
143 | PyObject *PickleError; |
144 | PyObject *PicklingError; |
145 | PyObject *UnpicklingError; |
146 | |
147 | /* copyreg.dispatch_table, {type_object: pickling_function} */ |
148 | PyObject *dispatch_table; |
149 | |
150 | /* For the extension opcodes EXT1, EXT2 and EXT4. */ |
151 | |
152 | /* copyreg._extension_registry, {(module_name, function_name): code} */ |
153 | PyObject *extension_registry; |
154 | /* copyreg._extension_cache, {code: object} */ |
155 | PyObject *extension_cache; |
156 | /* copyreg._inverted_registry, {code: (module_name, function_name)} */ |
157 | PyObject *inverted_registry; |
158 | |
159 | /* Import mappings for compatibility with Python 2.x */ |
160 | |
161 | /* _compat_pickle.NAME_MAPPING, |
162 | {(oldmodule, oldname): (newmodule, newname)} */ |
163 | PyObject *name_mapping_2to3; |
164 | /* _compat_pickle.IMPORT_MAPPING, {oldmodule: newmodule} */ |
165 | PyObject *import_mapping_2to3; |
166 | /* Same, but with REVERSE_NAME_MAPPING / REVERSE_IMPORT_MAPPING */ |
167 | PyObject *name_mapping_3to2; |
168 | PyObject *import_mapping_3to2; |
169 | |
170 | /* codecs.encode, used for saving bytes in older protocols */ |
171 | PyObject *codecs_encode; |
172 | /* builtins.getattr, used for saving nested names with protocol < 4 */ |
173 | PyObject *getattr; |
174 | /* functools.partial, used for implementing __newobj_ex__ with protocols |
175 | 2 and 3 */ |
176 | PyObject *partial; |
177 | } PickleState; |
178 | |
179 | /* Forward declaration of the _pickle module definition. */ |
180 | static struct PyModuleDef _picklemodule; |
181 | |
182 | /* Given a module object, get its per-module state. */ |
183 | static PickleState * |
184 | _Pickle_GetState(PyObject *module) |
185 | { |
186 | return (PickleState *)_PyModule_GetState(module); |
187 | } |
188 | |
189 | /* Find the module instance imported in the currently running sub-interpreter |
190 | and get its state. */ |
191 | static PickleState * |
192 | _Pickle_GetGlobalState(void) |
193 | { |
194 | return _Pickle_GetState(PyState_FindModule(&_picklemodule)); |
195 | } |
196 | |
197 | /* Clear the given pickle module state. */ |
198 | static void |
199 | _Pickle_ClearState(PickleState *st) |
200 | { |
201 | Py_CLEAR(st->PickleError); |
202 | Py_CLEAR(st->PicklingError); |
203 | Py_CLEAR(st->UnpicklingError); |
204 | Py_CLEAR(st->dispatch_table); |
205 | Py_CLEAR(st->extension_registry); |
206 | Py_CLEAR(st->extension_cache); |
207 | Py_CLEAR(st->inverted_registry); |
208 | Py_CLEAR(st->name_mapping_2to3); |
209 | Py_CLEAR(st->import_mapping_2to3); |
210 | Py_CLEAR(st->name_mapping_3to2); |
211 | Py_CLEAR(st->import_mapping_3to2); |
212 | Py_CLEAR(st->codecs_encode); |
213 | Py_CLEAR(st->getattr); |
214 | Py_CLEAR(st->partial); |
215 | } |
216 | |
217 | /* Initialize the given pickle module state. */ |
218 | static int |
219 | _Pickle_InitState(PickleState *st) |
220 | { |
221 | PyObject *copyreg = NULL; |
222 | PyObject *compat_pickle = NULL; |
223 | PyObject *codecs = NULL; |
224 | PyObject *functools = NULL; |
225 | _Py_IDENTIFIER(getattr); |
226 | |
227 | st->getattr = _PyEval_GetBuiltinId(&PyId_getattr); |
228 | if (st->getattr == NULL) |
229 | goto error; |
230 | |
231 | copyreg = PyImport_ImportModule("copyreg" ); |
232 | if (!copyreg) |
233 | goto error; |
234 | st->dispatch_table = PyObject_GetAttrString(copyreg, "dispatch_table" ); |
235 | if (!st->dispatch_table) |
236 | goto error; |
237 | if (!PyDict_CheckExact(st->dispatch_table)) { |
238 | PyErr_Format(PyExc_RuntimeError, |
239 | "copyreg.dispatch_table should be a dict, not %.200s" , |
240 | Py_TYPE(st->dispatch_table)->tp_name); |
241 | goto error; |
242 | } |
243 | st->extension_registry = \ |
244 | PyObject_GetAttrString(copyreg, "_extension_registry" ); |
245 | if (!st->extension_registry) |
246 | goto error; |
247 | if (!PyDict_CheckExact(st->extension_registry)) { |
248 | PyErr_Format(PyExc_RuntimeError, |
249 | "copyreg._extension_registry should be a dict, " |
250 | "not %.200s" , Py_TYPE(st->extension_registry)->tp_name); |
251 | goto error; |
252 | } |
253 | st->inverted_registry = \ |
254 | PyObject_GetAttrString(copyreg, "_inverted_registry" ); |
255 | if (!st->inverted_registry) |
256 | goto error; |
257 | if (!PyDict_CheckExact(st->inverted_registry)) { |
258 | PyErr_Format(PyExc_RuntimeError, |
259 | "copyreg._inverted_registry should be a dict, " |
260 | "not %.200s" , Py_TYPE(st->inverted_registry)->tp_name); |
261 | goto error; |
262 | } |
263 | st->extension_cache = PyObject_GetAttrString(copyreg, "_extension_cache" ); |
264 | if (!st->extension_cache) |
265 | goto error; |
266 | if (!PyDict_CheckExact(st->extension_cache)) { |
267 | PyErr_Format(PyExc_RuntimeError, |
268 | "copyreg._extension_cache should be a dict, " |
269 | "not %.200s" , Py_TYPE(st->extension_cache)->tp_name); |
270 | goto error; |
271 | } |
272 | Py_CLEAR(copyreg); |
273 | |
274 | /* Load the 2.x -> 3.x stdlib module mapping tables */ |
275 | compat_pickle = PyImport_ImportModule("_compat_pickle" ); |
276 | if (!compat_pickle) |
277 | goto error; |
278 | st->name_mapping_2to3 = \ |
279 | PyObject_GetAttrString(compat_pickle, "NAME_MAPPING" ); |
280 | if (!st->name_mapping_2to3) |
281 | goto error; |
282 | if (!PyDict_CheckExact(st->name_mapping_2to3)) { |
283 | PyErr_Format(PyExc_RuntimeError, |
284 | "_compat_pickle.NAME_MAPPING should be a dict, not %.200s" , |
285 | Py_TYPE(st->name_mapping_2to3)->tp_name); |
286 | goto error; |
287 | } |
288 | st->import_mapping_2to3 = \ |
289 | PyObject_GetAttrString(compat_pickle, "IMPORT_MAPPING" ); |
290 | if (!st->import_mapping_2to3) |
291 | goto error; |
292 | if (!PyDict_CheckExact(st->import_mapping_2to3)) { |
293 | PyErr_Format(PyExc_RuntimeError, |
294 | "_compat_pickle.IMPORT_MAPPING should be a dict, " |
295 | "not %.200s" , Py_TYPE(st->import_mapping_2to3)->tp_name); |
296 | goto error; |
297 | } |
298 | /* ... and the 3.x -> 2.x mapping tables */ |
299 | st->name_mapping_3to2 = \ |
300 | PyObject_GetAttrString(compat_pickle, "REVERSE_NAME_MAPPING" ); |
301 | if (!st->name_mapping_3to2) |
302 | goto error; |
303 | if (!PyDict_CheckExact(st->name_mapping_3to2)) { |
304 | PyErr_Format(PyExc_RuntimeError, |
305 | "_compat_pickle.REVERSE_NAME_MAPPING should be a dict, " |
306 | "not %.200s" , Py_TYPE(st->name_mapping_3to2)->tp_name); |
307 | goto error; |
308 | } |
309 | st->import_mapping_3to2 = \ |
310 | PyObject_GetAttrString(compat_pickle, "REVERSE_IMPORT_MAPPING" ); |
311 | if (!st->import_mapping_3to2) |
312 | goto error; |
313 | if (!PyDict_CheckExact(st->import_mapping_3to2)) { |
314 | PyErr_Format(PyExc_RuntimeError, |
315 | "_compat_pickle.REVERSE_IMPORT_MAPPING should be a dict, " |
316 | "not %.200s" , Py_TYPE(st->import_mapping_3to2)->tp_name); |
317 | goto error; |
318 | } |
319 | Py_CLEAR(compat_pickle); |
320 | |
321 | codecs = PyImport_ImportModule("codecs" ); |
322 | if (codecs == NULL) |
323 | goto error; |
324 | st->codecs_encode = PyObject_GetAttrString(codecs, "encode" ); |
325 | if (st->codecs_encode == NULL) { |
326 | goto error; |
327 | } |
328 | if (!PyCallable_Check(st->codecs_encode)) { |
329 | PyErr_Format(PyExc_RuntimeError, |
330 | "codecs.encode should be a callable, not %.200s" , |
331 | Py_TYPE(st->codecs_encode)->tp_name); |
332 | goto error; |
333 | } |
334 | Py_CLEAR(codecs); |
335 | |
336 | functools = PyImport_ImportModule("functools" ); |
337 | if (!functools) |
338 | goto error; |
339 | st->partial = PyObject_GetAttrString(functools, "partial" ); |
340 | if (!st->partial) |
341 | goto error; |
342 | Py_CLEAR(functools); |
343 | |
344 | return 0; |
345 | |
346 | error: |
347 | Py_CLEAR(copyreg); |
348 | Py_CLEAR(compat_pickle); |
349 | Py_CLEAR(codecs); |
350 | Py_CLEAR(functools); |
351 | _Pickle_ClearState(st); |
352 | return -1; |
353 | } |
354 | |
355 | /* Helper for calling a function with a single argument quickly. |
356 | |
357 | This function steals the reference of the given argument. */ |
358 | static PyObject * |
359 | _Pickle_FastCall(PyObject *func, PyObject *obj) |
360 | { |
361 | PyObject *result; |
362 | |
363 | result = PyObject_CallOneArg(func, obj); |
364 | Py_DECREF(obj); |
365 | return result; |
366 | } |
367 | |
368 | /*************************************************************************/ |
369 | |
370 | /* Retrieve and deconstruct a method for avoiding a reference cycle |
371 | (pickler -> bound method of pickler -> pickler) */ |
372 | static int |
373 | init_method_ref(PyObject *self, _Py_Identifier *name, |
374 | PyObject **method_func, PyObject **method_self) |
375 | { |
376 | PyObject *func, *func2; |
377 | int ret; |
378 | |
379 | /* *method_func and *method_self should be consistent. All refcount decrements |
380 | should be occurred after setting *method_self and *method_func. */ |
381 | ret = _PyObject_LookupAttrId(self, name, &func); |
382 | if (func == NULL) { |
383 | *method_self = NULL; |
384 | Py_CLEAR(*method_func); |
385 | return ret; |
386 | } |
387 | |
388 | if (PyMethod_Check(func) && PyMethod_GET_SELF(func) == self) { |
389 | /* Deconstruct a bound Python method */ |
390 | func2 = PyMethod_GET_FUNCTION(func); |
391 | Py_INCREF(func2); |
392 | *method_self = self; /* borrowed */ |
393 | Py_XSETREF(*method_func, func2); |
394 | Py_DECREF(func); |
395 | return 0; |
396 | } |
397 | else { |
398 | *method_self = NULL; |
399 | Py_XSETREF(*method_func, func); |
400 | return 0; |
401 | } |
402 | } |
403 | |
404 | /* Bind a method if it was deconstructed */ |
405 | static PyObject * |
406 | reconstruct_method(PyObject *func, PyObject *self) |
407 | { |
408 | if (self) { |
409 | return PyMethod_New(func, self); |
410 | } |
411 | else { |
412 | Py_INCREF(func); |
413 | return func; |
414 | } |
415 | } |
416 | |
417 | static PyObject * |
418 | call_method(PyObject *func, PyObject *self, PyObject *obj) |
419 | { |
420 | if (self) { |
421 | return PyObject_CallFunctionObjArgs(func, self, obj, NULL); |
422 | } |
423 | else { |
424 | return PyObject_CallOneArg(func, obj); |
425 | } |
426 | } |
427 | |
428 | /*************************************************************************/ |
429 | |
430 | /* Internal data type used as the unpickling stack. */ |
431 | typedef struct { |
432 | PyObject_VAR_HEAD |
433 | PyObject **data; |
434 | int mark_set; /* is MARK set? */ |
435 | Py_ssize_t fence; /* position of top MARK or 0 */ |
436 | Py_ssize_t allocated; /* number of slots in data allocated */ |
437 | } Pdata; |
438 | |
439 | static void |
440 | Pdata_dealloc(Pdata *self) |
441 | { |
442 | Py_ssize_t i = Py_SIZE(self); |
443 | while (--i >= 0) { |
444 | Py_DECREF(self->data[i]); |
445 | } |
446 | PyMem_Free(self->data); |
447 | PyObject_Free(self); |
448 | } |
449 | |
450 | static PyTypeObject Pdata_Type = { |
451 | PyVarObject_HEAD_INIT(NULL, 0) |
452 | "_pickle.Pdata" , /*tp_name*/ |
453 | sizeof(Pdata), /*tp_basicsize*/ |
454 | sizeof(PyObject *), /*tp_itemsize*/ |
455 | (destructor)Pdata_dealloc, /*tp_dealloc*/ |
456 | }; |
457 | |
458 | static PyObject * |
459 | Pdata_New(void) |
460 | { |
461 | Pdata *self; |
462 | |
463 | if (!(self = PyObject_New(Pdata, &Pdata_Type))) |
464 | return NULL; |
465 | Py_SET_SIZE(self, 0); |
466 | self->mark_set = 0; |
467 | self->fence = 0; |
468 | self->allocated = 8; |
469 | self->data = PyMem_Malloc(self->allocated * sizeof(PyObject *)); |
470 | if (self->data) |
471 | return (PyObject *)self; |
472 | Py_DECREF(self); |
473 | return PyErr_NoMemory(); |
474 | } |
475 | |
476 | |
477 | /* Retain only the initial clearto items. If clearto >= the current |
478 | * number of items, this is a (non-erroneous) NOP. |
479 | */ |
480 | static int |
481 | Pdata_clear(Pdata *self, Py_ssize_t clearto) |
482 | { |
483 | Py_ssize_t i = Py_SIZE(self); |
484 | |
485 | assert(clearto >= self->fence); |
486 | if (clearto >= i) |
487 | return 0; |
488 | |
489 | while (--i >= clearto) { |
490 | Py_CLEAR(self->data[i]); |
491 | } |
492 | Py_SET_SIZE(self, clearto); |
493 | return 0; |
494 | } |
495 | |
496 | static int |
497 | Pdata_grow(Pdata *self) |
498 | { |
499 | PyObject **data = self->data; |
500 | size_t allocated = (size_t)self->allocated; |
501 | size_t new_allocated; |
502 | |
503 | new_allocated = (allocated >> 3) + 6; |
504 | /* check for integer overflow */ |
505 | if (new_allocated > (size_t)PY_SSIZE_T_MAX - allocated) |
506 | goto nomemory; |
507 | new_allocated += allocated; |
508 | PyMem_RESIZE(data, PyObject *, new_allocated); |
509 | if (data == NULL) |
510 | goto nomemory; |
511 | |
512 | self->data = data; |
513 | self->allocated = (Py_ssize_t)new_allocated; |
514 | return 0; |
515 | |
516 | nomemory: |
517 | PyErr_NoMemory(); |
518 | return -1; |
519 | } |
520 | |
521 | static int |
522 | Pdata_stack_underflow(Pdata *self) |
523 | { |
524 | PickleState *st = _Pickle_GetGlobalState(); |
525 | PyErr_SetString(st->UnpicklingError, |
526 | self->mark_set ? |
527 | "unexpected MARK found" : |
528 | "unpickling stack underflow" ); |
529 | return -1; |
530 | } |
531 | |
532 | /* D is a Pdata*. Pop the topmost element and store it into V, which |
533 | * must be an lvalue holding PyObject*. On stack underflow, UnpicklingError |
534 | * is raised and V is set to NULL. |
535 | */ |
536 | static PyObject * |
537 | Pdata_pop(Pdata *self) |
538 | { |
539 | if (Py_SIZE(self) <= self->fence) { |
540 | Pdata_stack_underflow(self); |
541 | return NULL; |
542 | } |
543 | Py_SET_SIZE(self, Py_SIZE(self) - 1); |
544 | return self->data[Py_SIZE(self)]; |
545 | } |
546 | #define PDATA_POP(D, V) do { (V) = Pdata_pop((D)); } while (0) |
547 | |
548 | static int |
549 | Pdata_push(Pdata *self, PyObject *obj) |
550 | { |
551 | if (Py_SIZE(self) == self->allocated && Pdata_grow(self) < 0) { |
552 | return -1; |
553 | } |
554 | self->data[Py_SIZE(self)] = obj; |
555 | Py_SET_SIZE(self, Py_SIZE(self) + 1); |
556 | return 0; |
557 | } |
558 | |
559 | /* Push an object on stack, transferring its ownership to the stack. */ |
560 | #define PDATA_PUSH(D, O, ER) do { \ |
561 | if (Pdata_push((D), (O)) < 0) return (ER); } while(0) |
562 | |
563 | /* Push an object on stack, adding a new reference to the object. */ |
564 | #define PDATA_APPEND(D, O, ER) do { \ |
565 | Py_INCREF((O)); \ |
566 | if (Pdata_push((D), (O)) < 0) return (ER); } while(0) |
567 | |
568 | static PyObject * |
569 | Pdata_poptuple(Pdata *self, Py_ssize_t start) |
570 | { |
571 | PyObject *tuple; |
572 | Py_ssize_t len, i, j; |
573 | |
574 | if (start < self->fence) { |
575 | Pdata_stack_underflow(self); |
576 | return NULL; |
577 | } |
578 | len = Py_SIZE(self) - start; |
579 | tuple = PyTuple_New(len); |
580 | if (tuple == NULL) |
581 | return NULL; |
582 | for (i = start, j = 0; j < len; i++, j++) |
583 | PyTuple_SET_ITEM(tuple, j, self->data[i]); |
584 | |
585 | Py_SET_SIZE(self, start); |
586 | return tuple; |
587 | } |
588 | |
589 | static PyObject * |
590 | Pdata_poplist(Pdata *self, Py_ssize_t start) |
591 | { |
592 | PyObject *list; |
593 | Py_ssize_t len, i, j; |
594 | |
595 | len = Py_SIZE(self) - start; |
596 | list = PyList_New(len); |
597 | if (list == NULL) |
598 | return NULL; |
599 | for (i = start, j = 0; j < len; i++, j++) |
600 | PyList_SET_ITEM(list, j, self->data[i]); |
601 | |
602 | Py_SET_SIZE(self, start); |
603 | return list; |
604 | } |
605 | |
606 | typedef struct { |
607 | PyObject *me_key; |
608 | Py_ssize_t me_value; |
609 | } PyMemoEntry; |
610 | |
611 | typedef struct { |
612 | size_t mt_mask; |
613 | size_t mt_used; |
614 | size_t mt_allocated; |
615 | PyMemoEntry *mt_table; |
616 | } PyMemoTable; |
617 | |
618 | typedef struct PicklerObject { |
619 | PyObject_HEAD |
620 | PyMemoTable *memo; /* Memo table, keep track of the seen |
621 | objects to support self-referential objects |
622 | pickling. */ |
623 | PyObject *pers_func; /* persistent_id() method, can be NULL */ |
624 | PyObject *pers_func_self; /* borrowed reference to self if pers_func |
625 | is an unbound method, NULL otherwise */ |
626 | PyObject *dispatch_table; /* private dispatch_table, can be NULL */ |
627 | PyObject *reducer_override; /* hook for invoking user-defined callbacks |
628 | instead of save_global when pickling |
629 | functions and classes*/ |
630 | |
631 | PyObject *write; /* write() method of the output stream. */ |
632 | PyObject *output_buffer; /* Write into a local bytearray buffer before |
633 | flushing to the stream. */ |
634 | Py_ssize_t output_len; /* Length of output_buffer. */ |
635 | Py_ssize_t max_output_len; /* Allocation size of output_buffer. */ |
636 | int proto; /* Pickle protocol number, >= 0 */ |
637 | int bin; /* Boolean, true if proto > 0 */ |
638 | int framing; /* True when framing is enabled, proto >= 4 */ |
639 | Py_ssize_t frame_start; /* Position in output_buffer where the |
640 | current frame begins. -1 if there |
641 | is no frame currently open. */ |
642 | |
643 | Py_ssize_t buf_size; /* Size of the current buffered pickle data */ |
644 | int fast; /* Enable fast mode if set to a true value. |
645 | The fast mode disable the usage of memo, |
646 | therefore speeding the pickling process by |
647 | not generating superfluous PUT opcodes. It |
648 | should not be used if with self-referential |
649 | objects. */ |
650 | int fast_nesting; |
651 | int fix_imports; /* Indicate whether Pickler should fix |
652 | the name of globals for Python 2.x. */ |
653 | PyObject *fast_memo; |
654 | PyObject *buffer_callback; /* Callback for out-of-band buffers, or NULL */ |
655 | } PicklerObject; |
656 | |
657 | typedef struct UnpicklerObject { |
658 | PyObject_HEAD |
659 | Pdata *stack; /* Pickle data stack, store unpickled objects. */ |
660 | |
661 | /* The unpickler memo is just an array of PyObject *s. Using a dict |
662 | is unnecessary, since the keys are contiguous ints. */ |
663 | PyObject **memo; |
664 | size_t memo_size; /* Capacity of the memo array */ |
665 | size_t memo_len; /* Number of objects in the memo */ |
666 | |
667 | PyObject *pers_func; /* persistent_load() method, can be NULL. */ |
668 | PyObject *pers_func_self; /* borrowed reference to self if pers_func |
669 | is an unbound method, NULL otherwise */ |
670 | |
671 | Py_buffer buffer; |
672 | char *input_buffer; |
673 | char *input_line; |
674 | Py_ssize_t input_len; |
675 | Py_ssize_t next_read_idx; |
676 | Py_ssize_t prefetched_idx; /* index of first prefetched byte */ |
677 | |
678 | PyObject *read; /* read() method of the input stream. */ |
679 | PyObject *readinto; /* readinto() method of the input stream. */ |
680 | PyObject *readline; /* readline() method of the input stream. */ |
681 | PyObject *peek; /* peek() method of the input stream, or NULL */ |
682 | PyObject *buffers; /* iterable of out-of-band buffers, or NULL */ |
683 | |
684 | char *encoding; /* Name of the encoding to be used for |
685 | decoding strings pickled using Python |
686 | 2.x. The default value is "ASCII" */ |
687 | char *errors; /* Name of errors handling scheme to used when |
688 | decoding strings. The default value is |
689 | "strict". */ |
690 | Py_ssize_t *marks; /* Mark stack, used for unpickling container |
691 | objects. */ |
692 | Py_ssize_t num_marks; /* Number of marks in the mark stack. */ |
693 | Py_ssize_t marks_size; /* Current allocated size of the mark stack. */ |
694 | int proto; /* Protocol of the pickle loaded. */ |
695 | int fix_imports; /* Indicate whether Unpickler should fix |
696 | the name of globals pickled by Python 2.x. */ |
697 | } UnpicklerObject; |
698 | |
699 | typedef struct { |
700 | PyObject_HEAD |
701 | PicklerObject *pickler; /* Pickler whose memo table we're proxying. */ |
702 | } PicklerMemoProxyObject; |
703 | |
704 | typedef struct { |
705 | PyObject_HEAD |
706 | UnpicklerObject *unpickler; |
707 | } UnpicklerMemoProxyObject; |
708 | |
709 | /* Forward declarations */ |
710 | static int save(PicklerObject *, PyObject *, int); |
711 | static int save_reduce(PicklerObject *, PyObject *, PyObject *); |
712 | static PyTypeObject Pickler_Type; |
713 | static PyTypeObject Unpickler_Type; |
714 | |
715 | #include "clinic/_pickle.c.h" |
716 | |
717 | /************************************************************************* |
718 | A custom hashtable mapping void* to Python ints. This is used by the pickler |
719 | for memoization. Using a custom hashtable rather than PyDict allows us to skip |
720 | a bunch of unnecessary object creation. This makes a huge performance |
721 | difference. */ |
722 | |
723 | #define MT_MINSIZE 8 |
724 | #define PERTURB_SHIFT 5 |
725 | |
726 | |
727 | static PyMemoTable * |
728 | PyMemoTable_New(void) |
729 | { |
730 | PyMemoTable *memo = PyMem_Malloc(sizeof(PyMemoTable)); |
731 | if (memo == NULL) { |
732 | PyErr_NoMemory(); |
733 | return NULL; |
734 | } |
735 | |
736 | memo->mt_used = 0; |
737 | memo->mt_allocated = MT_MINSIZE; |
738 | memo->mt_mask = MT_MINSIZE - 1; |
739 | memo->mt_table = PyMem_Malloc(MT_MINSIZE * sizeof(PyMemoEntry)); |
740 | if (memo->mt_table == NULL) { |
741 | PyMem_Free(memo); |
742 | PyErr_NoMemory(); |
743 | return NULL; |
744 | } |
745 | memset(memo->mt_table, 0, MT_MINSIZE * sizeof(PyMemoEntry)); |
746 | |
747 | return memo; |
748 | } |
749 | |
750 | static PyMemoTable * |
751 | PyMemoTable_Copy(PyMemoTable *self) |
752 | { |
753 | PyMemoTable *new = PyMemoTable_New(); |
754 | if (new == NULL) |
755 | return NULL; |
756 | |
757 | new->mt_used = self->mt_used; |
758 | new->mt_allocated = self->mt_allocated; |
759 | new->mt_mask = self->mt_mask; |
760 | /* The table we get from _New() is probably smaller than we wanted. |
761 | Free it and allocate one that's the right size. */ |
762 | PyMem_Free(new->mt_table); |
763 | new->mt_table = PyMem_NEW(PyMemoEntry, self->mt_allocated); |
764 | if (new->mt_table == NULL) { |
765 | PyMem_Free(new); |
766 | PyErr_NoMemory(); |
767 | return NULL; |
768 | } |
769 | for (size_t i = 0; i < self->mt_allocated; i++) { |
770 | Py_XINCREF(self->mt_table[i].me_key); |
771 | } |
772 | memcpy(new->mt_table, self->mt_table, |
773 | sizeof(PyMemoEntry) * self->mt_allocated); |
774 | |
775 | return new; |
776 | } |
777 | |
778 | static Py_ssize_t |
779 | PyMemoTable_Size(PyMemoTable *self) |
780 | { |
781 | return self->mt_used; |
782 | } |
783 | |
784 | static int |
785 | PyMemoTable_Clear(PyMemoTable *self) |
786 | { |
787 | Py_ssize_t i = self->mt_allocated; |
788 | |
789 | while (--i >= 0) { |
790 | Py_XDECREF(self->mt_table[i].me_key); |
791 | } |
792 | self->mt_used = 0; |
793 | memset(self->mt_table, 0, self->mt_allocated * sizeof(PyMemoEntry)); |
794 | return 0; |
795 | } |
796 | |
797 | static void |
798 | PyMemoTable_Del(PyMemoTable *self) |
799 | { |
800 | if (self == NULL) |
801 | return; |
802 | PyMemoTable_Clear(self); |
803 | |
804 | PyMem_Free(self->mt_table); |
805 | PyMem_Free(self); |
806 | } |
807 | |
808 | /* Since entries cannot be deleted from this hashtable, _PyMemoTable_Lookup() |
809 | can be considerably simpler than dictobject.c's lookdict(). */ |
810 | static PyMemoEntry * |
811 | _PyMemoTable_Lookup(PyMemoTable *self, PyObject *key) |
812 | { |
813 | size_t i; |
814 | size_t perturb; |
815 | size_t mask = self->mt_mask; |
816 | PyMemoEntry *table = self->mt_table; |
817 | PyMemoEntry *entry; |
818 | Py_hash_t hash = (Py_hash_t)key >> 3; |
819 | |
820 | i = hash & mask; |
821 | entry = &table[i]; |
822 | if (entry->me_key == NULL || entry->me_key == key) |
823 | return entry; |
824 | |
825 | for (perturb = hash; ; perturb >>= PERTURB_SHIFT) { |
826 | i = (i << 2) + i + perturb + 1; |
827 | entry = &table[i & mask]; |
828 | if (entry->me_key == NULL || entry->me_key == key) |
829 | return entry; |
830 | } |
831 | Py_UNREACHABLE(); |
832 | } |
833 | |
834 | /* Returns -1 on failure, 0 on success. */ |
835 | static int |
836 | _PyMemoTable_ResizeTable(PyMemoTable *self, size_t min_size) |
837 | { |
838 | PyMemoEntry *oldtable = NULL; |
839 | PyMemoEntry *oldentry, *newentry; |
840 | size_t new_size = MT_MINSIZE; |
841 | size_t to_process; |
842 | |
843 | assert(min_size > 0); |
844 | |
845 | if (min_size > PY_SSIZE_T_MAX) { |
846 | PyErr_NoMemory(); |
847 | return -1; |
848 | } |
849 | |
850 | /* Find the smallest valid table size >= min_size. */ |
851 | while (new_size < min_size) { |
852 | new_size <<= 1; |
853 | } |
854 | /* new_size needs to be a power of two. */ |
855 | assert((new_size & (new_size - 1)) == 0); |
856 | |
857 | /* Allocate new table. */ |
858 | oldtable = self->mt_table; |
859 | self->mt_table = PyMem_NEW(PyMemoEntry, new_size); |
860 | if (self->mt_table == NULL) { |
861 | self->mt_table = oldtable; |
862 | PyErr_NoMemory(); |
863 | return -1; |
864 | } |
865 | self->mt_allocated = new_size; |
866 | self->mt_mask = new_size - 1; |
867 | memset(self->mt_table, 0, sizeof(PyMemoEntry) * new_size); |
868 | |
869 | /* Copy entries from the old table. */ |
870 | to_process = self->mt_used; |
871 | for (oldentry = oldtable; to_process > 0; oldentry++) { |
872 | if (oldentry->me_key != NULL) { |
873 | to_process--; |
874 | /* newentry is a pointer to a chunk of the new |
875 | mt_table, so we're setting the key:value pair |
876 | in-place. */ |
877 | newentry = _PyMemoTable_Lookup(self, oldentry->me_key); |
878 | newentry->me_key = oldentry->me_key; |
879 | newentry->me_value = oldentry->me_value; |
880 | } |
881 | } |
882 | |
883 | /* Deallocate the old table. */ |
884 | PyMem_Free(oldtable); |
885 | return 0; |
886 | } |
887 | |
888 | /* Returns NULL on failure, a pointer to the value otherwise. */ |
889 | static Py_ssize_t * |
890 | PyMemoTable_Get(PyMemoTable *self, PyObject *key) |
891 | { |
892 | PyMemoEntry *entry = _PyMemoTable_Lookup(self, key); |
893 | if (entry->me_key == NULL) |
894 | return NULL; |
895 | return &entry->me_value; |
896 | } |
897 | |
898 | /* Returns -1 on failure, 0 on success. */ |
899 | static int |
900 | PyMemoTable_Set(PyMemoTable *self, PyObject *key, Py_ssize_t value) |
901 | { |
902 | PyMemoEntry *entry; |
903 | |
904 | assert(key != NULL); |
905 | |
906 | entry = _PyMemoTable_Lookup(self, key); |
907 | if (entry->me_key != NULL) { |
908 | entry->me_value = value; |
909 | return 0; |
910 | } |
911 | Py_INCREF(key); |
912 | entry->me_key = key; |
913 | entry->me_value = value; |
914 | self->mt_used++; |
915 | |
916 | /* If we added a key, we can safely resize. Otherwise just return! |
917 | * If used >= 2/3 size, adjust size. Normally, this quaduples the size. |
918 | * |
919 | * Quadrupling the size improves average table sparseness |
920 | * (reducing collisions) at the cost of some memory. It also halves |
921 | * the number of expensive resize operations in a growing memo table. |
922 | * |
923 | * Very large memo tables (over 50K items) use doubling instead. |
924 | * This may help applications with severe memory constraints. |
925 | */ |
926 | if (SIZE_MAX / 3 >= self->mt_used && self->mt_used * 3 < self->mt_allocated * 2) { |
927 | return 0; |
928 | } |
929 | // self->mt_used is always < PY_SSIZE_T_MAX, so this can't overflow. |
930 | size_t desired_size = (self->mt_used > 50000 ? 2 : 4) * self->mt_used; |
931 | return _PyMemoTable_ResizeTable(self, desired_size); |
932 | } |
933 | |
934 | #undef MT_MINSIZE |
935 | #undef PERTURB_SHIFT |
936 | |
937 | /*************************************************************************/ |
938 | |
939 | |
940 | static int |
941 | _Pickler_ClearBuffer(PicklerObject *self) |
942 | { |
943 | Py_XSETREF(self->output_buffer, |
944 | PyBytes_FromStringAndSize(NULL, self->max_output_len)); |
945 | if (self->output_buffer == NULL) |
946 | return -1; |
947 | self->output_len = 0; |
948 | self->frame_start = -1; |
949 | return 0; |
950 | } |
951 | |
952 | static void |
953 | _write_size64(char *out, size_t value) |
954 | { |
955 | size_t i; |
956 | |
957 | Py_BUILD_ASSERT(sizeof(size_t) <= 8); |
958 | |
959 | for (i = 0; i < sizeof(size_t); i++) { |
960 | out[i] = (unsigned char)((value >> (8 * i)) & 0xff); |
961 | } |
962 | for (i = sizeof(size_t); i < 8; i++) { |
963 | out[i] = 0; |
964 | } |
965 | } |
966 | |
967 | static int |
968 | _Pickler_CommitFrame(PicklerObject *self) |
969 | { |
970 | size_t frame_len; |
971 | char *qdata; |
972 | |
973 | if (!self->framing || self->frame_start == -1) |
974 | return 0; |
975 | frame_len = self->output_len - self->frame_start - FRAME_HEADER_SIZE; |
976 | qdata = PyBytes_AS_STRING(self->output_buffer) + self->frame_start; |
977 | if (frame_len >= FRAME_SIZE_MIN) { |
978 | qdata[0] = FRAME; |
979 | _write_size64(qdata + 1, frame_len); |
980 | } |
981 | else { |
982 | memmove(qdata, qdata + FRAME_HEADER_SIZE, frame_len); |
983 | self->output_len -= FRAME_HEADER_SIZE; |
984 | } |
985 | self->frame_start = -1; |
986 | return 0; |
987 | } |
988 | |
989 | static PyObject * |
990 | _Pickler_GetString(PicklerObject *self) |
991 | { |
992 | PyObject *output_buffer = self->output_buffer; |
993 | |
994 | assert(self->output_buffer != NULL); |
995 | |
996 | if (_Pickler_CommitFrame(self)) |
997 | return NULL; |
998 | |
999 | self->output_buffer = NULL; |
1000 | /* Resize down to exact size */ |
1001 | if (_PyBytes_Resize(&output_buffer, self->output_len) < 0) |
1002 | return NULL; |
1003 | return output_buffer; |
1004 | } |
1005 | |
1006 | static int |
1007 | _Pickler_FlushToFile(PicklerObject *self) |
1008 | { |
1009 | PyObject *output, *result; |
1010 | |
1011 | assert(self->write != NULL); |
1012 | |
1013 | /* This will commit the frame first */ |
1014 | output = _Pickler_GetString(self); |
1015 | if (output == NULL) |
1016 | return -1; |
1017 | |
1018 | result = _Pickle_FastCall(self->write, output); |
1019 | Py_XDECREF(result); |
1020 | return (result == NULL) ? -1 : 0; |
1021 | } |
1022 | |
1023 | static int |
1024 | _Pickler_OpcodeBoundary(PicklerObject *self) |
1025 | { |
1026 | Py_ssize_t frame_len; |
1027 | |
1028 | if (!self->framing || self->frame_start == -1) { |
1029 | return 0; |
1030 | } |
1031 | frame_len = self->output_len - self->frame_start - FRAME_HEADER_SIZE; |
1032 | if (frame_len >= FRAME_SIZE_TARGET) { |
1033 | if(_Pickler_CommitFrame(self)) { |
1034 | return -1; |
1035 | } |
1036 | /* Flush the content of the committed frame to the underlying |
1037 | * file and reuse the pickler buffer for the next frame so as |
1038 | * to limit memory usage when dumping large complex objects to |
1039 | * a file. |
1040 | * |
1041 | * self->write is NULL when called via dumps. |
1042 | */ |
1043 | if (self->write != NULL) { |
1044 | if (_Pickler_FlushToFile(self) < 0) { |
1045 | return -1; |
1046 | } |
1047 | if (_Pickler_ClearBuffer(self) < 0) { |
1048 | return -1; |
1049 | } |
1050 | } |
1051 | } |
1052 | return 0; |
1053 | } |
1054 | |
1055 | static Py_ssize_t |
1056 | _Pickler_Write(PicklerObject *self, const char *s, Py_ssize_t data_len) |
1057 | { |
1058 | Py_ssize_t i, n, required; |
1059 | char *buffer; |
1060 | int need_new_frame; |
1061 | |
1062 | assert(s != NULL); |
1063 | need_new_frame = (self->framing && self->frame_start == -1); |
1064 | |
1065 | if (need_new_frame) |
1066 | n = data_len + FRAME_HEADER_SIZE; |
1067 | else |
1068 | n = data_len; |
1069 | |
1070 | required = self->output_len + n; |
1071 | if (required > self->max_output_len) { |
1072 | /* Make place in buffer for the pickle chunk */ |
1073 | if (self->output_len >= PY_SSIZE_T_MAX / 2 - n) { |
1074 | PyErr_NoMemory(); |
1075 | return -1; |
1076 | } |
1077 | self->max_output_len = (self->output_len + n) / 2 * 3; |
1078 | if (_PyBytes_Resize(&self->output_buffer, self->max_output_len) < 0) |
1079 | return -1; |
1080 | } |
1081 | buffer = PyBytes_AS_STRING(self->output_buffer); |
1082 | if (need_new_frame) { |
1083 | /* Setup new frame */ |
1084 | Py_ssize_t frame_start = self->output_len; |
1085 | self->frame_start = frame_start; |
1086 | for (i = 0; i < FRAME_HEADER_SIZE; i++) { |
1087 | /* Write an invalid value, for debugging */ |
1088 | buffer[frame_start + i] = 0xFE; |
1089 | } |
1090 | self->output_len += FRAME_HEADER_SIZE; |
1091 | } |
1092 | if (data_len < 8) { |
1093 | /* This is faster than memcpy when the string is short. */ |
1094 | for (i = 0; i < data_len; i++) { |
1095 | buffer[self->output_len + i] = s[i]; |
1096 | } |
1097 | } |
1098 | else { |
1099 | memcpy(buffer + self->output_len, s, data_len); |
1100 | } |
1101 | self->output_len += data_len; |
1102 | return data_len; |
1103 | } |
1104 | |
1105 | static PicklerObject * |
1106 | _Pickler_New(void) |
1107 | { |
1108 | PicklerObject *self; |
1109 | |
1110 | self = PyObject_GC_New(PicklerObject, &Pickler_Type); |
1111 | if (self == NULL) |
1112 | return NULL; |
1113 | |
1114 | self->pers_func = NULL; |
1115 | self->dispatch_table = NULL; |
1116 | self->buffer_callback = NULL; |
1117 | self->write = NULL; |
1118 | self->proto = 0; |
1119 | self->bin = 0; |
1120 | self->framing = 0; |
1121 | self->frame_start = -1; |
1122 | self->fast = 0; |
1123 | self->fast_nesting = 0; |
1124 | self->fix_imports = 0; |
1125 | self->fast_memo = NULL; |
1126 | self->max_output_len = WRITE_BUF_SIZE; |
1127 | self->output_len = 0; |
1128 | self->reducer_override = NULL; |
1129 | |
1130 | self->memo = PyMemoTable_New(); |
1131 | self->output_buffer = PyBytes_FromStringAndSize(NULL, |
1132 | self->max_output_len); |
1133 | |
1134 | if (self->memo == NULL || self->output_buffer == NULL) { |
1135 | Py_DECREF(self); |
1136 | return NULL; |
1137 | } |
1138 | |
1139 | PyObject_GC_Track(self); |
1140 | return self; |
1141 | } |
1142 | |
1143 | static int |
1144 | _Pickler_SetProtocol(PicklerObject *self, PyObject *protocol, int fix_imports) |
1145 | { |
1146 | long proto; |
1147 | |
1148 | if (protocol == Py_None) { |
1149 | proto = DEFAULT_PROTOCOL; |
1150 | } |
1151 | else { |
1152 | proto = PyLong_AsLong(protocol); |
1153 | if (proto < 0) { |
1154 | if (proto == -1 && PyErr_Occurred()) |
1155 | return -1; |
1156 | proto = HIGHEST_PROTOCOL; |
1157 | } |
1158 | else if (proto > HIGHEST_PROTOCOL) { |
1159 | PyErr_Format(PyExc_ValueError, "pickle protocol must be <= %d" , |
1160 | HIGHEST_PROTOCOL); |
1161 | return -1; |
1162 | } |
1163 | } |
1164 | self->proto = (int)proto; |
1165 | self->bin = proto > 0; |
1166 | self->fix_imports = fix_imports && proto < 3; |
1167 | return 0; |
1168 | } |
1169 | |
1170 | /* Returns -1 (with an exception set) on failure, 0 on success. This may |
1171 | be called once on a freshly created Pickler. */ |
1172 | static int |
1173 | _Pickler_SetOutputStream(PicklerObject *self, PyObject *file) |
1174 | { |
1175 | _Py_IDENTIFIER(write); |
1176 | assert(file != NULL); |
1177 | if (_PyObject_LookupAttrId(file, &PyId_write, &self->write) < 0) { |
1178 | return -1; |
1179 | } |
1180 | if (self->write == NULL) { |
1181 | PyErr_SetString(PyExc_TypeError, |
1182 | "file must have a 'write' attribute" ); |
1183 | return -1; |
1184 | } |
1185 | |
1186 | return 0; |
1187 | } |
1188 | |
1189 | static int |
1190 | _Pickler_SetBufferCallback(PicklerObject *self, PyObject *buffer_callback) |
1191 | { |
1192 | if (buffer_callback == Py_None) { |
1193 | buffer_callback = NULL; |
1194 | } |
1195 | if (buffer_callback != NULL && self->proto < 5) { |
1196 | PyErr_SetString(PyExc_ValueError, |
1197 | "buffer_callback needs protocol >= 5" ); |
1198 | return -1; |
1199 | } |
1200 | |
1201 | Py_XINCREF(buffer_callback); |
1202 | self->buffer_callback = buffer_callback; |
1203 | return 0; |
1204 | } |
1205 | |
1206 | /* Returns the size of the input on success, -1 on failure. This takes its |
1207 | own reference to `input`. */ |
1208 | static Py_ssize_t |
1209 | _Unpickler_SetStringInput(UnpicklerObject *self, PyObject *input) |
1210 | { |
1211 | if (self->buffer.buf != NULL) |
1212 | PyBuffer_Release(&self->buffer); |
1213 | if (PyObject_GetBuffer(input, &self->buffer, PyBUF_CONTIG_RO) < 0) |
1214 | return -1; |
1215 | self->input_buffer = self->buffer.buf; |
1216 | self->input_len = self->buffer.len; |
1217 | self->next_read_idx = 0; |
1218 | self->prefetched_idx = self->input_len; |
1219 | return self->input_len; |
1220 | } |
1221 | |
1222 | static int |
1223 | bad_readline(void) |
1224 | { |
1225 | PickleState *st = _Pickle_GetGlobalState(); |
1226 | PyErr_SetString(st->UnpicklingError, "pickle data was truncated" ); |
1227 | return -1; |
1228 | } |
1229 | |
1230 | /* Skip any consumed data that was only prefetched using peek() */ |
1231 | static int |
1232 | _Unpickler_SkipConsumed(UnpicklerObject *self) |
1233 | { |
1234 | Py_ssize_t consumed; |
1235 | PyObject *r; |
1236 | |
1237 | consumed = self->next_read_idx - self->prefetched_idx; |
1238 | if (consumed <= 0) |
1239 | return 0; |
1240 | |
1241 | assert(self->peek); /* otherwise we did something wrong */ |
1242 | /* This makes a useless copy... */ |
1243 | r = PyObject_CallFunction(self->read, "n" , consumed); |
1244 | if (r == NULL) |
1245 | return -1; |
1246 | Py_DECREF(r); |
1247 | |
1248 | self->prefetched_idx = self->next_read_idx; |
1249 | return 0; |
1250 | } |
1251 | |
1252 | static const Py_ssize_t READ_WHOLE_LINE = -1; |
1253 | |
1254 | /* If reading from a file, we need to only pull the bytes we need, since there |
1255 | may be multiple pickle objects arranged contiguously in the same input |
1256 | buffer. |
1257 | |
1258 | If `n` is READ_WHOLE_LINE, read a whole line. Otherwise, read up to `n` |
1259 | bytes from the input stream/buffer. |
1260 | |
1261 | Update the unpickler's input buffer with the newly-read data. Returns -1 on |
1262 | failure; on success, returns the number of bytes read from the file. |
1263 | |
1264 | On success, self->input_len will be 0; this is intentional so that when |
1265 | unpickling from a file, the "we've run out of data" code paths will trigger, |
1266 | causing the Unpickler to go back to the file for more data. Use the returned |
1267 | size to tell you how much data you can process. */ |
1268 | static Py_ssize_t |
1269 | _Unpickler_ReadFromFile(UnpicklerObject *self, Py_ssize_t n) |
1270 | { |
1271 | PyObject *data; |
1272 | Py_ssize_t read_size; |
1273 | |
1274 | assert(self->read != NULL); |
1275 | |
1276 | if (_Unpickler_SkipConsumed(self) < 0) |
1277 | return -1; |
1278 | |
1279 | if (n == READ_WHOLE_LINE) { |
1280 | data = PyObject_CallNoArgs(self->readline); |
1281 | } |
1282 | else { |
1283 | PyObject *len; |
1284 | /* Prefetch some data without advancing the file pointer, if possible */ |
1285 | if (self->peek && n < PREFETCH) { |
1286 | len = PyLong_FromSsize_t(PREFETCH); |
1287 | if (len == NULL) |
1288 | return -1; |
1289 | data = _Pickle_FastCall(self->peek, len); |
1290 | if (data == NULL) { |
1291 | if (!PyErr_ExceptionMatches(PyExc_NotImplementedError)) |
1292 | return -1; |
1293 | /* peek() is probably not supported by the given file object */ |
1294 | PyErr_Clear(); |
1295 | Py_CLEAR(self->peek); |
1296 | } |
1297 | else { |
1298 | read_size = _Unpickler_SetStringInput(self, data); |
1299 | Py_DECREF(data); |
1300 | self->prefetched_idx = 0; |
1301 | if (n <= read_size) |
1302 | return n; |
1303 | } |
1304 | } |
1305 | len = PyLong_FromSsize_t(n); |
1306 | if (len == NULL) |
1307 | return -1; |
1308 | data = _Pickle_FastCall(self->read, len); |
1309 | } |
1310 | if (data == NULL) |
1311 | return -1; |
1312 | |
1313 | read_size = _Unpickler_SetStringInput(self, data); |
1314 | Py_DECREF(data); |
1315 | return read_size; |
1316 | } |
1317 | |
1318 | /* Don't call it directly: use _Unpickler_Read() */ |
1319 | static Py_ssize_t |
1320 | _Unpickler_ReadImpl(UnpicklerObject *self, char **s, Py_ssize_t n) |
1321 | { |
1322 | Py_ssize_t num_read; |
1323 | |
1324 | *s = NULL; |
1325 | if (self->next_read_idx > PY_SSIZE_T_MAX - n) { |
1326 | PickleState *st = _Pickle_GetGlobalState(); |
1327 | PyErr_SetString(st->UnpicklingError, |
1328 | "read would overflow (invalid bytecode)" ); |
1329 | return -1; |
1330 | } |
1331 | |
1332 | /* This case is handled by the _Unpickler_Read() macro for efficiency */ |
1333 | assert(self->next_read_idx + n > self->input_len); |
1334 | |
1335 | if (!self->read) |
1336 | return bad_readline(); |
1337 | |
1338 | /* Extend the buffer to satisfy desired size */ |
1339 | num_read = _Unpickler_ReadFromFile(self, n); |
1340 | if (num_read < 0) |
1341 | return -1; |
1342 | if (num_read < n) |
1343 | return bad_readline(); |
1344 | *s = self->input_buffer; |
1345 | self->next_read_idx = n; |
1346 | return n; |
1347 | } |
1348 | |
1349 | /* Read `n` bytes from the unpickler's data source, storing the result in `buf`. |
1350 | * |
1351 | * This should only be used for non-small data reads where potentially |
1352 | * avoiding a copy is beneficial. This method does not try to prefetch |
1353 | * more data into the input buffer. |
1354 | * |
1355 | * _Unpickler_Read() is recommended in most cases. |
1356 | */ |
1357 | static Py_ssize_t |
1358 | _Unpickler_ReadInto(UnpicklerObject *self, char *buf, Py_ssize_t n) |
1359 | { |
1360 | assert(n != READ_WHOLE_LINE); |
1361 | |
1362 | /* Read from available buffer data, if any */ |
1363 | Py_ssize_t in_buffer = self->input_len - self->next_read_idx; |
1364 | if (in_buffer > 0) { |
1365 | Py_ssize_t to_read = Py_MIN(in_buffer, n); |
1366 | memcpy(buf, self->input_buffer + self->next_read_idx, to_read); |
1367 | self->next_read_idx += to_read; |
1368 | buf += to_read; |
1369 | n -= to_read; |
1370 | if (n == 0) { |
1371 | /* Entire read was satisfied from buffer */ |
1372 | return n; |
1373 | } |
1374 | } |
1375 | |
1376 | /* Read from file */ |
1377 | if (!self->read) { |
1378 | /* We're unpickling memory, this means the input is truncated */ |
1379 | return bad_readline(); |
1380 | } |
1381 | if (_Unpickler_SkipConsumed(self) < 0) { |
1382 | return -1; |
1383 | } |
1384 | |
1385 | if (!self->readinto) { |
1386 | /* readinto() not supported on file-like object, fall back to read() |
1387 | * and copy into destination buffer (bpo-39681) */ |
1388 | PyObject* len = PyLong_FromSsize_t(n); |
1389 | if (len == NULL) { |
1390 | return -1; |
1391 | } |
1392 | PyObject* data = _Pickle_FastCall(self->read, len); |
1393 | if (data == NULL) { |
1394 | return -1; |
1395 | } |
1396 | if (!PyBytes_Check(data)) { |
1397 | PyErr_Format(PyExc_ValueError, |
1398 | "read() returned non-bytes object (%R)" , |
1399 | Py_TYPE(data)); |
1400 | Py_DECREF(data); |
1401 | return -1; |
1402 | } |
1403 | Py_ssize_t read_size = PyBytes_GET_SIZE(data); |
1404 | if (read_size < n) { |
1405 | Py_DECREF(data); |
1406 | return bad_readline(); |
1407 | } |
1408 | memcpy(buf, PyBytes_AS_STRING(data), n); |
1409 | Py_DECREF(data); |
1410 | return n; |
1411 | } |
1412 | |
1413 | /* Call readinto() into user buffer */ |
1414 | PyObject *buf_obj = PyMemoryView_FromMemory(buf, n, PyBUF_WRITE); |
1415 | if (buf_obj == NULL) { |
1416 | return -1; |
1417 | } |
1418 | PyObject *read_size_obj = _Pickle_FastCall(self->readinto, buf_obj); |
1419 | if (read_size_obj == NULL) { |
1420 | return -1; |
1421 | } |
1422 | Py_ssize_t read_size = PyLong_AsSsize_t(read_size_obj); |
1423 | Py_DECREF(read_size_obj); |
1424 | |
1425 | if (read_size < 0) { |
1426 | if (!PyErr_Occurred()) { |
1427 | PyErr_SetString(PyExc_ValueError, |
1428 | "readinto() returned negative size" ); |
1429 | } |
1430 | return -1; |
1431 | } |
1432 | if (read_size < n) { |
1433 | return bad_readline(); |
1434 | } |
1435 | return n; |
1436 | } |
1437 | |
1438 | /* Read `n` bytes from the unpickler's data source, storing the result in `*s`. |
1439 | |
1440 | This should be used for all data reads, rather than accessing the unpickler's |
1441 | input buffer directly. This method deals correctly with reading from input |
1442 | streams, which the input buffer doesn't deal with. |
1443 | |
1444 | Note that when reading from a file-like object, self->next_read_idx won't |
1445 | be updated (it should remain at 0 for the entire unpickling process). You |
1446 | should use this function's return value to know how many bytes you can |
1447 | consume. |
1448 | |
1449 | Returns -1 (with an exception set) on failure. On success, return the |
1450 | number of chars read. */ |
1451 | #define _Unpickler_Read(self, s, n) \ |
1452 | (((n) <= (self)->input_len - (self)->next_read_idx) \ |
1453 | ? (*(s) = (self)->input_buffer + (self)->next_read_idx, \ |
1454 | (self)->next_read_idx += (n), \ |
1455 | (n)) \ |
1456 | : _Unpickler_ReadImpl(self, (s), (n))) |
1457 | |
1458 | static Py_ssize_t |
1459 | _Unpickler_CopyLine(UnpicklerObject *self, char *line, Py_ssize_t len, |
1460 | char **result) |
1461 | { |
1462 | char *input_line = PyMem_Realloc(self->input_line, len + 1); |
1463 | if (input_line == NULL) { |
1464 | PyErr_NoMemory(); |
1465 | return -1; |
1466 | } |
1467 | |
1468 | memcpy(input_line, line, len); |
1469 | input_line[len] = '\0'; |
1470 | self->input_line = input_line; |
1471 | *result = self->input_line; |
1472 | return len; |
1473 | } |
1474 | |
1475 | /* Read a line from the input stream/buffer. If we run off the end of the input |
1476 | before hitting \n, raise an error. |
1477 | |
1478 | Returns the number of chars read, or -1 on failure. */ |
1479 | static Py_ssize_t |
1480 | _Unpickler_Readline(UnpicklerObject *self, char **result) |
1481 | { |
1482 | Py_ssize_t i, num_read; |
1483 | |
1484 | for (i = self->next_read_idx; i < self->input_len; i++) { |
1485 | if (self->input_buffer[i] == '\n') { |
1486 | char *line_start = self->input_buffer + self->next_read_idx; |
1487 | num_read = i - self->next_read_idx + 1; |
1488 | self->next_read_idx = i + 1; |
1489 | return _Unpickler_CopyLine(self, line_start, num_read, result); |
1490 | } |
1491 | } |
1492 | if (!self->read) |
1493 | return bad_readline(); |
1494 | |
1495 | num_read = _Unpickler_ReadFromFile(self, READ_WHOLE_LINE); |
1496 | if (num_read < 0) |
1497 | return -1; |
1498 | if (num_read == 0 || self->input_buffer[num_read - 1] != '\n') |
1499 | return bad_readline(); |
1500 | self->next_read_idx = num_read; |
1501 | return _Unpickler_CopyLine(self, self->input_buffer, num_read, result); |
1502 | } |
1503 | |
1504 | /* Returns -1 (with an exception set) on failure, 0 on success. The memo array |
1505 | will be modified in place. */ |
1506 | static int |
1507 | _Unpickler_ResizeMemoList(UnpicklerObject *self, size_t new_size) |
1508 | { |
1509 | size_t i; |
1510 | |
1511 | assert(new_size > self->memo_size); |
1512 | |
1513 | PyObject **memo_new = self->memo; |
1514 | PyMem_RESIZE(memo_new, PyObject *, new_size); |
1515 | if (memo_new == NULL) { |
1516 | PyErr_NoMemory(); |
1517 | return -1; |
1518 | } |
1519 | self->memo = memo_new; |
1520 | for (i = self->memo_size; i < new_size; i++) |
1521 | self->memo[i] = NULL; |
1522 | self->memo_size = new_size; |
1523 | return 0; |
1524 | } |
1525 | |
1526 | /* Returns NULL if idx is out of bounds. */ |
1527 | static PyObject * |
1528 | _Unpickler_MemoGet(UnpicklerObject *self, size_t idx) |
1529 | { |
1530 | if (idx >= self->memo_size) |
1531 | return NULL; |
1532 | |
1533 | return self->memo[idx]; |
1534 | } |
1535 | |
1536 | /* Returns -1 (with an exception set) on failure, 0 on success. |
1537 | This takes its own reference to `value`. */ |
1538 | static int |
1539 | _Unpickler_MemoPut(UnpicklerObject *self, size_t idx, PyObject *value) |
1540 | { |
1541 | PyObject *old_item; |
1542 | |
1543 | if (idx >= self->memo_size) { |
1544 | if (_Unpickler_ResizeMemoList(self, idx * 2) < 0) |
1545 | return -1; |
1546 | assert(idx < self->memo_size); |
1547 | } |
1548 | Py_INCREF(value); |
1549 | old_item = self->memo[idx]; |
1550 | self->memo[idx] = value; |
1551 | if (old_item != NULL) { |
1552 | Py_DECREF(old_item); |
1553 | } |
1554 | else { |
1555 | self->memo_len++; |
1556 | } |
1557 | return 0; |
1558 | } |
1559 | |
1560 | static PyObject ** |
1561 | _Unpickler_NewMemo(Py_ssize_t new_size) |
1562 | { |
1563 | PyObject **memo = PyMem_NEW(PyObject *, new_size); |
1564 | if (memo == NULL) { |
1565 | PyErr_NoMemory(); |
1566 | return NULL; |
1567 | } |
1568 | memset(memo, 0, new_size * sizeof(PyObject *)); |
1569 | return memo; |
1570 | } |
1571 | |
1572 | /* Free the unpickler's memo, taking care to decref any items left in it. */ |
1573 | static void |
1574 | _Unpickler_MemoCleanup(UnpicklerObject *self) |
1575 | { |
1576 | Py_ssize_t i; |
1577 | PyObject **memo = self->memo; |
1578 | |
1579 | if (self->memo == NULL) |
1580 | return; |
1581 | self->memo = NULL; |
1582 | i = self->memo_size; |
1583 | while (--i >= 0) { |
1584 | Py_XDECREF(memo[i]); |
1585 | } |
1586 | PyMem_Free(memo); |
1587 | } |
1588 | |
1589 | static UnpicklerObject * |
1590 | _Unpickler_New(void) |
1591 | { |
1592 | UnpicklerObject *self; |
1593 | |
1594 | self = PyObject_GC_New(UnpicklerObject, &Unpickler_Type); |
1595 | if (self == NULL) |
1596 | return NULL; |
1597 | |
1598 | self->pers_func = NULL; |
1599 | self->input_buffer = NULL; |
1600 | self->input_line = NULL; |
1601 | self->input_len = 0; |
1602 | self->next_read_idx = 0; |
1603 | self->prefetched_idx = 0; |
1604 | self->read = NULL; |
1605 | self->readinto = NULL; |
1606 | self->readline = NULL; |
1607 | self->peek = NULL; |
1608 | self->buffers = NULL; |
1609 | self->encoding = NULL; |
1610 | self->errors = NULL; |
1611 | self->marks = NULL; |
1612 | self->num_marks = 0; |
1613 | self->marks_size = 0; |
1614 | self->proto = 0; |
1615 | self->fix_imports = 0; |
1616 | memset(&self->buffer, 0, sizeof(Py_buffer)); |
1617 | self->memo_size = 32; |
1618 | self->memo_len = 0; |
1619 | self->memo = _Unpickler_NewMemo(self->memo_size); |
1620 | self->stack = (Pdata *)Pdata_New(); |
1621 | |
1622 | if (self->memo == NULL || self->stack == NULL) { |
1623 | Py_DECREF(self); |
1624 | return NULL; |
1625 | } |
1626 | |
1627 | PyObject_GC_Track(self); |
1628 | return self; |
1629 | } |
1630 | |
1631 | /* Returns -1 (with an exception set) on failure, 0 on success. This may |
1632 | be called once on a freshly created Unpickler. */ |
1633 | static int |
1634 | _Unpickler_SetInputStream(UnpicklerObject *self, PyObject *file) |
1635 | { |
1636 | _Py_IDENTIFIER(peek); |
1637 | _Py_IDENTIFIER(read); |
1638 | _Py_IDENTIFIER(readinto); |
1639 | _Py_IDENTIFIER(readline); |
1640 | |
1641 | /* Optional file methods */ |
1642 | if (_PyObject_LookupAttrId(file, &PyId_peek, &self->peek) < 0) { |
1643 | return -1; |
1644 | } |
1645 | if (_PyObject_LookupAttrId(file, &PyId_readinto, &self->readinto) < 0) { |
1646 | return -1; |
1647 | } |
1648 | (void)_PyObject_LookupAttrId(file, &PyId_read, &self->read); |
1649 | (void)_PyObject_LookupAttrId(file, &PyId_readline, &self->readline); |
1650 | if (!self->readline || !self->read) { |
1651 | if (!PyErr_Occurred()) { |
1652 | PyErr_SetString(PyExc_TypeError, |
1653 | "file must have 'read' and 'readline' attributes" ); |
1654 | } |
1655 | Py_CLEAR(self->read); |
1656 | Py_CLEAR(self->readinto); |
1657 | Py_CLEAR(self->readline); |
1658 | Py_CLEAR(self->peek); |
1659 | return -1; |
1660 | } |
1661 | return 0; |
1662 | } |
1663 | |
1664 | /* Returns -1 (with an exception set) on failure, 0 on success. This may |
1665 | be called once on a freshly created Unpickler. */ |
1666 | static int |
1667 | _Unpickler_SetInputEncoding(UnpicklerObject *self, |
1668 | const char *encoding, |
1669 | const char *errors) |
1670 | { |
1671 | if (encoding == NULL) |
1672 | encoding = "ASCII" ; |
1673 | if (errors == NULL) |
1674 | errors = "strict" ; |
1675 | |
1676 | self->encoding = _PyMem_Strdup(encoding); |
1677 | self->errors = _PyMem_Strdup(errors); |
1678 | if (self->encoding == NULL || self->errors == NULL) { |
1679 | PyErr_NoMemory(); |
1680 | return -1; |
1681 | } |
1682 | return 0; |
1683 | } |
1684 | |
1685 | /* Returns -1 (with an exception set) on failure, 0 on success. This may |
1686 | be called once on a freshly created Unpickler. */ |
1687 | static int |
1688 | _Unpickler_SetBuffers(UnpicklerObject *self, PyObject *buffers) |
1689 | { |
1690 | if (buffers == NULL || buffers == Py_None) { |
1691 | self->buffers = NULL; |
1692 | } |
1693 | else { |
1694 | self->buffers = PyObject_GetIter(buffers); |
1695 | if (self->buffers == NULL) { |
1696 | return -1; |
1697 | } |
1698 | } |
1699 | return 0; |
1700 | } |
1701 | |
1702 | /* Generate a GET opcode for an object stored in the memo. */ |
1703 | static int |
1704 | memo_get(PicklerObject *self, PyObject *key) |
1705 | { |
1706 | Py_ssize_t *value; |
1707 | char pdata[30]; |
1708 | Py_ssize_t len; |
1709 | |
1710 | value = PyMemoTable_Get(self->memo, key); |
1711 | if (value == NULL) { |
1712 | PyErr_SetObject(PyExc_KeyError, key); |
1713 | return -1; |
1714 | } |
1715 | |
1716 | if (!self->bin) { |
1717 | pdata[0] = GET; |
1718 | PyOS_snprintf(pdata + 1, sizeof(pdata) - 1, |
1719 | "%zd\n" , *value); |
1720 | len = strlen(pdata); |
1721 | } |
1722 | else { |
1723 | if (*value < 256) { |
1724 | pdata[0] = BINGET; |
1725 | pdata[1] = (unsigned char)(*value & 0xff); |
1726 | len = 2; |
1727 | } |
1728 | else if ((size_t)*value <= 0xffffffffUL) { |
1729 | pdata[0] = LONG_BINGET; |
1730 | pdata[1] = (unsigned char)(*value & 0xff); |
1731 | pdata[2] = (unsigned char)((*value >> 8) & 0xff); |
1732 | pdata[3] = (unsigned char)((*value >> 16) & 0xff); |
1733 | pdata[4] = (unsigned char)((*value >> 24) & 0xff); |
1734 | len = 5; |
1735 | } |
1736 | else { /* unlikely */ |
1737 | PickleState *st = _Pickle_GetGlobalState(); |
1738 | PyErr_SetString(st->PicklingError, |
1739 | "memo id too large for LONG_BINGET" ); |
1740 | return -1; |
1741 | } |
1742 | } |
1743 | |
1744 | if (_Pickler_Write(self, pdata, len) < 0) |
1745 | return -1; |
1746 | |
1747 | return 0; |
1748 | } |
1749 | |
1750 | /* Store an object in the memo, assign it a new unique ID based on the number |
1751 | of objects currently stored in the memo and generate a PUT opcode. */ |
1752 | static int |
1753 | memo_put(PicklerObject *self, PyObject *obj) |
1754 | { |
1755 | char pdata[30]; |
1756 | Py_ssize_t len; |
1757 | Py_ssize_t idx; |
1758 | |
1759 | const char memoize_op = MEMOIZE; |
1760 | |
1761 | if (self->fast) |
1762 | return 0; |
1763 | |
1764 | idx = PyMemoTable_Size(self->memo); |
1765 | if (PyMemoTable_Set(self->memo, obj, idx) < 0) |
1766 | return -1; |
1767 | |
1768 | if (self->proto >= 4) { |
1769 | if (_Pickler_Write(self, &memoize_op, 1) < 0) |
1770 | return -1; |
1771 | return 0; |
1772 | } |
1773 | else if (!self->bin) { |
1774 | pdata[0] = PUT; |
1775 | PyOS_snprintf(pdata + 1, sizeof(pdata) - 1, |
1776 | "%zd\n" , idx); |
1777 | len = strlen(pdata); |
1778 | } |
1779 | else { |
1780 | if (idx < 256) { |
1781 | pdata[0] = BINPUT; |
1782 | pdata[1] = (unsigned char)idx; |
1783 | len = 2; |
1784 | } |
1785 | else if ((size_t)idx <= 0xffffffffUL) { |
1786 | pdata[0] = LONG_BINPUT; |
1787 | pdata[1] = (unsigned char)(idx & 0xff); |
1788 | pdata[2] = (unsigned char)((idx >> 8) & 0xff); |
1789 | pdata[3] = (unsigned char)((idx >> 16) & 0xff); |
1790 | pdata[4] = (unsigned char)((idx >> 24) & 0xff); |
1791 | len = 5; |
1792 | } |
1793 | else { /* unlikely */ |
1794 | PickleState *st = _Pickle_GetGlobalState(); |
1795 | PyErr_SetString(st->PicklingError, |
1796 | "memo id too large for LONG_BINPUT" ); |
1797 | return -1; |
1798 | } |
1799 | } |
1800 | if (_Pickler_Write(self, pdata, len) < 0) |
1801 | return -1; |
1802 | |
1803 | return 0; |
1804 | } |
1805 | |
1806 | static PyObject * |
1807 | get_dotted_path(PyObject *obj, PyObject *name) |
1808 | { |
1809 | _Py_static_string(PyId_dot, "." ); |
1810 | PyObject *dotted_path; |
1811 | Py_ssize_t i, n; |
1812 | |
1813 | dotted_path = PyUnicode_Split(name, _PyUnicode_FromId(&PyId_dot), -1); |
1814 | if (dotted_path == NULL) |
1815 | return NULL; |
1816 | n = PyList_GET_SIZE(dotted_path); |
1817 | assert(n >= 1); |
1818 | for (i = 0; i < n; i++) { |
1819 | PyObject *subpath = PyList_GET_ITEM(dotted_path, i); |
1820 | if (_PyUnicode_EqualToASCIIString(subpath, "<locals>" )) { |
1821 | if (obj == NULL) |
1822 | PyErr_Format(PyExc_AttributeError, |
1823 | "Can't pickle local object %R" , name); |
1824 | else |
1825 | PyErr_Format(PyExc_AttributeError, |
1826 | "Can't pickle local attribute %R on %R" , name, obj); |
1827 | Py_DECREF(dotted_path); |
1828 | return NULL; |
1829 | } |
1830 | } |
1831 | return dotted_path; |
1832 | } |
1833 | |
1834 | static PyObject * |
1835 | get_deep_attribute(PyObject *obj, PyObject *names, PyObject **pparent) |
1836 | { |
1837 | Py_ssize_t i, n; |
1838 | PyObject *parent = NULL; |
1839 | |
1840 | assert(PyList_CheckExact(names)); |
1841 | Py_INCREF(obj); |
1842 | n = PyList_GET_SIZE(names); |
1843 | for (i = 0; i < n; i++) { |
1844 | PyObject *name = PyList_GET_ITEM(names, i); |
1845 | Py_XDECREF(parent); |
1846 | parent = obj; |
1847 | (void)_PyObject_LookupAttr(parent, name, &obj); |
1848 | if (obj == NULL) { |
1849 | Py_DECREF(parent); |
1850 | return NULL; |
1851 | } |
1852 | } |
1853 | if (pparent != NULL) |
1854 | *pparent = parent; |
1855 | else |
1856 | Py_XDECREF(parent); |
1857 | return obj; |
1858 | } |
1859 | |
1860 | |
1861 | static PyObject * |
1862 | getattribute(PyObject *obj, PyObject *name, int allow_qualname) |
1863 | { |
1864 | PyObject *dotted_path, *attr; |
1865 | |
1866 | if (allow_qualname) { |
1867 | dotted_path = get_dotted_path(obj, name); |
1868 | if (dotted_path == NULL) |
1869 | return NULL; |
1870 | attr = get_deep_attribute(obj, dotted_path, NULL); |
1871 | Py_DECREF(dotted_path); |
1872 | } |
1873 | else { |
1874 | (void)_PyObject_LookupAttr(obj, name, &attr); |
1875 | } |
1876 | if (attr == NULL && !PyErr_Occurred()) { |
1877 | PyErr_Format(PyExc_AttributeError, |
1878 | "Can't get attribute %R on %R" , name, obj); |
1879 | } |
1880 | return attr; |
1881 | } |
1882 | |
1883 | static int |
1884 | _checkmodule(PyObject *module_name, PyObject *module, |
1885 | PyObject *global, PyObject *dotted_path) |
1886 | { |
1887 | if (module == Py_None) { |
1888 | return -1; |
1889 | } |
1890 | if (PyUnicode_Check(module_name) && |
1891 | _PyUnicode_EqualToASCIIString(module_name, "__main__" )) { |
1892 | return -1; |
1893 | } |
1894 | |
1895 | PyObject *candidate = get_deep_attribute(module, dotted_path, NULL); |
1896 | if (candidate == NULL) { |
1897 | return -1; |
1898 | } |
1899 | if (candidate != global) { |
1900 | Py_DECREF(candidate); |
1901 | return -1; |
1902 | } |
1903 | Py_DECREF(candidate); |
1904 | return 0; |
1905 | } |
1906 | |
1907 | static PyObject * |
1908 | whichmodule(PyObject *global, PyObject *dotted_path) |
1909 | { |
1910 | PyObject *module_name; |
1911 | PyObject *module = NULL; |
1912 | Py_ssize_t i; |
1913 | PyObject *modules; |
1914 | _Py_IDENTIFIER(__module__); |
1915 | _Py_IDENTIFIER(modules); |
1916 | _Py_IDENTIFIER(__main__); |
1917 | |
1918 | if (_PyObject_LookupAttrId(global, &PyId___module__, &module_name) < 0) { |
1919 | return NULL; |
1920 | } |
1921 | if (module_name) { |
1922 | /* In some rare cases (e.g., bound methods of extension types), |
1923 | __module__ can be None. If it is so, then search sys.modules for |
1924 | the module of global. */ |
1925 | if (module_name != Py_None) |
1926 | return module_name; |
1927 | Py_CLEAR(module_name); |
1928 | } |
1929 | assert(module_name == NULL); |
1930 | |
1931 | /* Fallback on walking sys.modules */ |
1932 | modules = _PySys_GetObjectId(&PyId_modules); |
1933 | if (modules == NULL) { |
1934 | PyErr_SetString(PyExc_RuntimeError, "unable to get sys.modules" ); |
1935 | return NULL; |
1936 | } |
1937 | if (PyDict_CheckExact(modules)) { |
1938 | i = 0; |
1939 | while (PyDict_Next(modules, &i, &module_name, &module)) { |
1940 | if (_checkmodule(module_name, module, global, dotted_path) == 0) { |
1941 | Py_INCREF(module_name); |
1942 | return module_name; |
1943 | } |
1944 | if (PyErr_Occurred()) { |
1945 | return NULL; |
1946 | } |
1947 | } |
1948 | } |
1949 | else { |
1950 | PyObject *iterator = PyObject_GetIter(modules); |
1951 | if (iterator == NULL) { |
1952 | return NULL; |
1953 | } |
1954 | while ((module_name = PyIter_Next(iterator))) { |
1955 | module = PyObject_GetItem(modules, module_name); |
1956 | if (module == NULL) { |
1957 | Py_DECREF(module_name); |
1958 | Py_DECREF(iterator); |
1959 | return NULL; |
1960 | } |
1961 | if (_checkmodule(module_name, module, global, dotted_path) == 0) { |
1962 | Py_DECREF(module); |
1963 | Py_DECREF(iterator); |
1964 | return module_name; |
1965 | } |
1966 | Py_DECREF(module); |
1967 | Py_DECREF(module_name); |
1968 | if (PyErr_Occurred()) { |
1969 | Py_DECREF(iterator); |
1970 | return NULL; |
1971 | } |
1972 | } |
1973 | Py_DECREF(iterator); |
1974 | } |
1975 | |
1976 | /* If no module is found, use __main__. */ |
1977 | module_name = _PyUnicode_FromId(&PyId___main__); |
1978 | Py_XINCREF(module_name); |
1979 | return module_name; |
1980 | } |
1981 | |
1982 | /* fast_save_enter() and fast_save_leave() are guards against recursive |
1983 | objects when Pickler is used with the "fast mode" (i.e., with object |
1984 | memoization disabled). If the nesting of a list or dict object exceed |
1985 | FAST_NESTING_LIMIT, these guards will start keeping an internal |
1986 | reference to the seen list or dict objects and check whether these objects |
1987 | are recursive. These are not strictly necessary, since save() has a |
1988 | hard-coded recursion limit, but they give a nicer error message than the |
1989 | typical RuntimeError. */ |
1990 | static int |
1991 | fast_save_enter(PicklerObject *self, PyObject *obj) |
1992 | { |
1993 | /* if fast_nesting < 0, we're doing an error exit. */ |
1994 | if (++self->fast_nesting >= FAST_NESTING_LIMIT) { |
1995 | PyObject *key = NULL; |
1996 | if (self->fast_memo == NULL) { |
1997 | self->fast_memo = PyDict_New(); |
1998 | if (self->fast_memo == NULL) { |
1999 | self->fast_nesting = -1; |
2000 | return 0; |
2001 | } |
2002 | } |
2003 | key = PyLong_FromVoidPtr(obj); |
2004 | if (key == NULL) { |
2005 | self->fast_nesting = -1; |
2006 | return 0; |
2007 | } |
2008 | int r = PyDict_Contains(self->fast_memo, key); |
2009 | if (r > 0) { |
2010 | PyErr_Format(PyExc_ValueError, |
2011 | "fast mode: can't pickle cyclic objects " |
2012 | "including object type %.200s at %p" , |
2013 | Py_TYPE(obj)->tp_name, obj); |
2014 | } |
2015 | else if (r == 0) { |
2016 | r = PyDict_SetItem(self->fast_memo, key, Py_None); |
2017 | } |
2018 | Py_DECREF(key); |
2019 | if (r != 0) { |
2020 | self->fast_nesting = -1; |
2021 | return 0; |
2022 | } |
2023 | } |
2024 | return 1; |
2025 | } |
2026 | |
2027 | static int |
2028 | fast_save_leave(PicklerObject *self, PyObject *obj) |
2029 | { |
2030 | if (self->fast_nesting-- >= FAST_NESTING_LIMIT) { |
2031 | PyObject *key = PyLong_FromVoidPtr(obj); |
2032 | if (key == NULL) |
2033 | return 0; |
2034 | if (PyDict_DelItem(self->fast_memo, key) < 0) { |
2035 | Py_DECREF(key); |
2036 | return 0; |
2037 | } |
2038 | Py_DECREF(key); |
2039 | } |
2040 | return 1; |
2041 | } |
2042 | |
2043 | static int |
2044 | save_none(PicklerObject *self, PyObject *obj) |
2045 | { |
2046 | const char none_op = NONE; |
2047 | if (_Pickler_Write(self, &none_op, 1) < 0) |
2048 | return -1; |
2049 | |
2050 | return 0; |
2051 | } |
2052 | |
2053 | static int |
2054 | save_bool(PicklerObject *self, PyObject *obj) |
2055 | { |
2056 | if (self->proto >= 2) { |
2057 | const char bool_op = (obj == Py_True) ? NEWTRUE : NEWFALSE; |
2058 | if (_Pickler_Write(self, &bool_op, 1) < 0) |
2059 | return -1; |
2060 | } |
2061 | else { |
2062 | /* These aren't opcodes -- they're ways to pickle bools before protocol 2 |
2063 | * so that unpicklers written before bools were introduced unpickle them |
2064 | * as ints, but unpicklers after can recognize that bools were intended. |
2065 | * Note that protocol 2 added direct ways to pickle bools. |
2066 | */ |
2067 | const char *bool_str = (obj == Py_True) ? "I01\n" : "I00\n" ; |
2068 | if (_Pickler_Write(self, bool_str, strlen(bool_str)) < 0) |
2069 | return -1; |
2070 | } |
2071 | return 0; |
2072 | } |
2073 | |
2074 | static int |
2075 | save_long(PicklerObject *self, PyObject *obj) |
2076 | { |
2077 | PyObject *repr = NULL; |
2078 | Py_ssize_t size; |
2079 | long val; |
2080 | int overflow; |
2081 | int status = 0; |
2082 | |
2083 | val= PyLong_AsLongAndOverflow(obj, &overflow); |
2084 | if (!overflow && (sizeof(long) <= 4 || |
2085 | (val <= 0x7fffffffL && val >= (-0x7fffffffL - 1)))) |
2086 | { |
2087 | /* result fits in a signed 4-byte integer. |
2088 | |
2089 | Note: we can't use -0x80000000L in the above condition because some |
2090 | compilers (e.g., MSVC) will promote 0x80000000L to an unsigned type |
2091 | before applying the unary minus when sizeof(long) <= 4. The |
2092 | resulting value stays unsigned which is commonly not what we want, |
2093 | so MSVC happily warns us about it. However, that result would have |
2094 | been fine because we guard for sizeof(long) <= 4 which turns the |
2095 | condition true in that particular case. */ |
2096 | char pdata[32]; |
2097 | Py_ssize_t len = 0; |
2098 | |
2099 | if (self->bin) { |
2100 | pdata[1] = (unsigned char)(val & 0xff); |
2101 | pdata[2] = (unsigned char)((val >> 8) & 0xff); |
2102 | pdata[3] = (unsigned char)((val >> 16) & 0xff); |
2103 | pdata[4] = (unsigned char)((val >> 24) & 0xff); |
2104 | |
2105 | if ((pdata[4] != 0) || (pdata[3] != 0)) { |
2106 | pdata[0] = BININT; |
2107 | len = 5; |
2108 | } |
2109 | else if (pdata[2] != 0) { |
2110 | pdata[0] = BININT2; |
2111 | len = 3; |
2112 | } |
2113 | else { |
2114 | pdata[0] = BININT1; |
2115 | len = 2; |
2116 | } |
2117 | } |
2118 | else { |
2119 | sprintf(pdata, "%c%ld\n" , INT, val); |
2120 | len = strlen(pdata); |
2121 | } |
2122 | if (_Pickler_Write(self, pdata, len) < 0) |
2123 | return -1; |
2124 | |
2125 | return 0; |
2126 | } |
2127 | assert(!PyErr_Occurred()); |
2128 | |
2129 | if (self->proto >= 2) { |
2130 | /* Linear-time pickling. */ |
2131 | size_t nbits; |
2132 | size_t nbytes; |
2133 | unsigned char *pdata; |
2134 | char [5]; |
2135 | int i; |
2136 | int sign = _PyLong_Sign(obj); |
2137 | |
2138 | if (sign == 0) { |
2139 | header[0] = LONG1; |
2140 | header[1] = 0; /* It's 0 -- an empty bytestring. */ |
2141 | if (_Pickler_Write(self, header, 2) < 0) |
2142 | goto error; |
2143 | return 0; |
2144 | } |
2145 | nbits = _PyLong_NumBits(obj); |
2146 | if (nbits == (size_t)-1 && PyErr_Occurred()) |
2147 | goto error; |
2148 | /* How many bytes do we need? There are nbits >> 3 full |
2149 | * bytes of data, and nbits & 7 leftover bits. If there |
2150 | * are any leftover bits, then we clearly need another |
2151 | * byte. What's not so obvious is that we *probably* |
2152 | * need another byte even if there aren't any leftovers: |
2153 | * the most-significant bit of the most-significant byte |
2154 | * acts like a sign bit, and it's usually got a sense |
2155 | * opposite of the one we need. The exception is ints |
2156 | * of the form -(2**(8*j-1)) for j > 0. Such an int is |
2157 | * its own 256's-complement, so has the right sign bit |
2158 | * even without the extra byte. That's a pain to check |
2159 | * for in advance, though, so we always grab an extra |
2160 | * byte at the start, and cut it back later if possible. |
2161 | */ |
2162 | nbytes = (nbits >> 3) + 1; |
2163 | if (nbytes > 0x7fffffffL) { |
2164 | PyErr_SetString(PyExc_OverflowError, |
2165 | "int too large to pickle" ); |
2166 | goto error; |
2167 | } |
2168 | repr = PyBytes_FromStringAndSize(NULL, (Py_ssize_t)nbytes); |
2169 | if (repr == NULL) |
2170 | goto error; |
2171 | pdata = (unsigned char *)PyBytes_AS_STRING(repr); |
2172 | i = _PyLong_AsByteArray((PyLongObject *)obj, |
2173 | pdata, nbytes, |
2174 | 1 /* little endian */ , 1 /* signed */ ); |
2175 | if (i < 0) |
2176 | goto error; |
2177 | /* If the int is negative, this may be a byte more than |
2178 | * needed. This is so iff the MSB is all redundant sign |
2179 | * bits. |
2180 | */ |
2181 | if (sign < 0 && |
2182 | nbytes > 1 && |
2183 | pdata[nbytes - 1] == 0xff && |
2184 | (pdata[nbytes - 2] & 0x80) != 0) { |
2185 | nbytes--; |
2186 | } |
2187 | |
2188 | if (nbytes < 256) { |
2189 | header[0] = LONG1; |
2190 | header[1] = (unsigned char)nbytes; |
2191 | size = 2; |
2192 | } |
2193 | else { |
2194 | header[0] = LONG4; |
2195 | size = (Py_ssize_t) nbytes; |
2196 | for (i = 1; i < 5; i++) { |
2197 | header[i] = (unsigned char)(size & 0xff); |
2198 | size >>= 8; |
2199 | } |
2200 | size = 5; |
2201 | } |
2202 | if (_Pickler_Write(self, header, size) < 0 || |
2203 | _Pickler_Write(self, (char *)pdata, (int)nbytes) < 0) |
2204 | goto error; |
2205 | } |
2206 | else { |
2207 | const char long_op = LONG; |
2208 | const char *string; |
2209 | |
2210 | /* proto < 2: write the repr and newline. This is quadratic-time (in |
2211 | the number of digits), in both directions. We add a trailing 'L' |
2212 | to the repr, for compatibility with Python 2.x. */ |
2213 | |
2214 | repr = PyObject_Repr(obj); |
2215 | if (repr == NULL) |
2216 | goto error; |
2217 | |
2218 | string = PyUnicode_AsUTF8AndSize(repr, &size); |
2219 | if (string == NULL) |
2220 | goto error; |
2221 | |
2222 | if (_Pickler_Write(self, &long_op, 1) < 0 || |
2223 | _Pickler_Write(self, string, size) < 0 || |
2224 | _Pickler_Write(self, "L\n" , 2) < 0) |
2225 | goto error; |
2226 | } |
2227 | |
2228 | if (0) { |
2229 | error: |
2230 | status = -1; |
2231 | } |
2232 | Py_XDECREF(repr); |
2233 | |
2234 | return status; |
2235 | } |
2236 | |
2237 | static int |
2238 | save_float(PicklerObject *self, PyObject *obj) |
2239 | { |
2240 | double x = PyFloat_AS_DOUBLE((PyFloatObject *)obj); |
2241 | |
2242 | if (self->bin) { |
2243 | char pdata[9]; |
2244 | pdata[0] = BINFLOAT; |
2245 | if (_PyFloat_Pack8(x, (unsigned char *)&pdata[1], 0) < 0) |
2246 | return -1; |
2247 | if (_Pickler_Write(self, pdata, 9) < 0) |
2248 | return -1; |
2249 | } |
2250 | else { |
2251 | int result = -1; |
2252 | char *buf = NULL; |
2253 | char op = FLOAT; |
2254 | |
2255 | if (_Pickler_Write(self, &op, 1) < 0) |
2256 | goto done; |
2257 | |
2258 | buf = PyOS_double_to_string(x, 'r', 0, Py_DTSF_ADD_DOT_0, NULL); |
2259 | if (!buf) { |
2260 | PyErr_NoMemory(); |
2261 | goto done; |
2262 | } |
2263 | |
2264 | if (_Pickler_Write(self, buf, strlen(buf)) < 0) |
2265 | goto done; |
2266 | |
2267 | if (_Pickler_Write(self, "\n" , 1) < 0) |
2268 | goto done; |
2269 | |
2270 | result = 0; |
2271 | done: |
2272 | PyMem_Free(buf); |
2273 | return result; |
2274 | } |
2275 | |
2276 | return 0; |
2277 | } |
2278 | |
2279 | /* Perform direct write of the header and payload of the binary object. |
2280 | |
2281 | The large contiguous data is written directly into the underlying file |
2282 | object, bypassing the output_buffer of the Pickler. We intentionally |
2283 | do not insert a protocol 4 frame opcode to make it possible to optimize |
2284 | file.read calls in the loader. |
2285 | */ |
2286 | static int |
2287 | _Pickler_write_bytes(PicklerObject *self, |
2288 | const char *, Py_ssize_t , |
2289 | const char *data, Py_ssize_t data_size, |
2290 | PyObject *payload) |
2291 | { |
2292 | int bypass_buffer = (data_size >= FRAME_SIZE_TARGET); |
2293 | int framing = self->framing; |
2294 | |
2295 | if (bypass_buffer) { |
2296 | assert(self->output_buffer != NULL); |
2297 | /* Commit the previous frame. */ |
2298 | if (_Pickler_CommitFrame(self)) { |
2299 | return -1; |
2300 | } |
2301 | /* Disable framing temporarily */ |
2302 | self->framing = 0; |
2303 | } |
2304 | |
2305 | if (_Pickler_Write(self, header, header_size) < 0) { |
2306 | return -1; |
2307 | } |
2308 | |
2309 | if (bypass_buffer && self->write != NULL) { |
2310 | /* Bypass the in-memory buffer to directly stream large data |
2311 | into the underlying file object. */ |
2312 | PyObject *result, *mem = NULL; |
2313 | /* Dump the output buffer to the file. */ |
2314 | if (_Pickler_FlushToFile(self) < 0) { |
2315 | return -1; |
2316 | } |
2317 | |
2318 | /* Stream write the payload into the file without going through the |
2319 | output buffer. */ |
2320 | if (payload == NULL) { |
2321 | /* TODO: It would be better to use a memoryview with a linked |
2322 | original string if this is possible. */ |
2323 | payload = mem = PyBytes_FromStringAndSize(data, data_size); |
2324 | if (payload == NULL) { |
2325 | return -1; |
2326 | } |
2327 | } |
2328 | result = PyObject_CallOneArg(self->write, payload); |
2329 | Py_XDECREF(mem); |
2330 | if (result == NULL) { |
2331 | return -1; |
2332 | } |
2333 | Py_DECREF(result); |
2334 | |
2335 | /* Reinitialize the buffer for subsequent calls to _Pickler_Write. */ |
2336 | if (_Pickler_ClearBuffer(self) < 0) { |
2337 | return -1; |
2338 | } |
2339 | } |
2340 | else { |
2341 | if (_Pickler_Write(self, data, data_size) < 0) { |
2342 | return -1; |
2343 | } |
2344 | } |
2345 | |
2346 | /* Re-enable framing for subsequent calls to _Pickler_Write. */ |
2347 | self->framing = framing; |
2348 | |
2349 | return 0; |
2350 | } |
2351 | |
2352 | static int |
2353 | _save_bytes_data(PicklerObject *self, PyObject *obj, const char *data, |
2354 | Py_ssize_t size) |
2355 | { |
2356 | assert(self->proto >= 3); |
2357 | |
2358 | char [9]; |
2359 | Py_ssize_t len; |
2360 | |
2361 | if (size < 0) |
2362 | return -1; |
2363 | |
2364 | if (size <= 0xff) { |
2365 | header[0] = SHORT_BINBYTES; |
2366 | header[1] = (unsigned char)size; |
2367 | len = 2; |
2368 | } |
2369 | else if ((size_t)size <= 0xffffffffUL) { |
2370 | header[0] = BINBYTES; |
2371 | header[1] = (unsigned char)(size & 0xff); |
2372 | header[2] = (unsigned char)((size >> 8) & 0xff); |
2373 | header[3] = (unsigned char)((size >> 16) & 0xff); |
2374 | header[4] = (unsigned char)((size >> 24) & 0xff); |
2375 | len = 5; |
2376 | } |
2377 | else if (self->proto >= 4) { |
2378 | header[0] = BINBYTES8; |
2379 | _write_size64(header + 1, size); |
2380 | len = 9; |
2381 | } |
2382 | else { |
2383 | PyErr_SetString(PyExc_OverflowError, |
2384 | "serializing a bytes object larger than 4 GiB " |
2385 | "requires pickle protocol 4 or higher" ); |
2386 | return -1; |
2387 | } |
2388 | |
2389 | if (_Pickler_write_bytes(self, header, len, data, size, obj) < 0) { |
2390 | return -1; |
2391 | } |
2392 | |
2393 | if (memo_put(self, obj) < 0) { |
2394 | return -1; |
2395 | } |
2396 | |
2397 | return 0; |
2398 | } |
2399 | |
2400 | static int |
2401 | save_bytes(PicklerObject *self, PyObject *obj) |
2402 | { |
2403 | if (self->proto < 3) { |
2404 | /* Older pickle protocols do not have an opcode for pickling bytes |
2405 | objects. Therefore, we need to fake the copy protocol (i.e., |
2406 | the __reduce__ method) to permit bytes object unpickling. |
2407 | |
2408 | Here we use a hack to be compatible with Python 2. Since in Python |
2409 | 2 'bytes' is just an alias for 'str' (which has different |
2410 | parameters than the actual bytes object), we use codecs.encode |
2411 | to create the appropriate 'str' object when unpickled using |
2412 | Python 2 *and* the appropriate 'bytes' object when unpickled |
2413 | using Python 3. Again this is a hack and we don't need to do this |
2414 | with newer protocols. */ |
2415 | PyObject *reduce_value; |
2416 | int status; |
2417 | |
2418 | if (PyBytes_GET_SIZE(obj) == 0) { |
2419 | reduce_value = Py_BuildValue("(O())" , (PyObject*)&PyBytes_Type); |
2420 | } |
2421 | else { |
2422 | PickleState *st = _Pickle_GetGlobalState(); |
2423 | PyObject *unicode_str = |
2424 | PyUnicode_DecodeLatin1(PyBytes_AS_STRING(obj), |
2425 | PyBytes_GET_SIZE(obj), |
2426 | "strict" ); |
2427 | _Py_IDENTIFIER(latin1); |
2428 | |
2429 | if (unicode_str == NULL) |
2430 | return -1; |
2431 | reduce_value = Py_BuildValue("(O(OO))" , |
2432 | st->codecs_encode, unicode_str, |
2433 | _PyUnicode_FromId(&PyId_latin1)); |
2434 | Py_DECREF(unicode_str); |
2435 | } |
2436 | |
2437 | if (reduce_value == NULL) |
2438 | return -1; |
2439 | |
2440 | /* save_reduce() will memoize the object automatically. */ |
2441 | status = save_reduce(self, reduce_value, obj); |
2442 | Py_DECREF(reduce_value); |
2443 | return status; |
2444 | } |
2445 | else { |
2446 | return _save_bytes_data(self, obj, PyBytes_AS_STRING(obj), |
2447 | PyBytes_GET_SIZE(obj)); |
2448 | } |
2449 | } |
2450 | |
2451 | static int |
2452 | _save_bytearray_data(PicklerObject *self, PyObject *obj, const char *data, |
2453 | Py_ssize_t size) |
2454 | { |
2455 | assert(self->proto >= 5); |
2456 | |
2457 | char [9]; |
2458 | Py_ssize_t len; |
2459 | |
2460 | if (size < 0) |
2461 | return -1; |
2462 | |
2463 | header[0] = BYTEARRAY8; |
2464 | _write_size64(header + 1, size); |
2465 | len = 9; |
2466 | |
2467 | if (_Pickler_write_bytes(self, header, len, data, size, obj) < 0) { |
2468 | return -1; |
2469 | } |
2470 | |
2471 | if (memo_put(self, obj) < 0) { |
2472 | return -1; |
2473 | } |
2474 | |
2475 | return 0; |
2476 | } |
2477 | |
2478 | static int |
2479 | save_bytearray(PicklerObject *self, PyObject *obj) |
2480 | { |
2481 | if (self->proto < 5) { |
2482 | /* Older pickle protocols do not have an opcode for pickling |
2483 | * bytearrays. */ |
2484 | PyObject *reduce_value = NULL; |
2485 | int status; |
2486 | |
2487 | if (PyByteArray_GET_SIZE(obj) == 0) { |
2488 | reduce_value = Py_BuildValue("(O())" , |
2489 | (PyObject *) &PyByteArray_Type); |
2490 | } |
2491 | else { |
2492 | PyObject *bytes_obj = PyBytes_FromObject(obj); |
2493 | if (bytes_obj != NULL) { |
2494 | reduce_value = Py_BuildValue("(O(O))" , |
2495 | (PyObject *) &PyByteArray_Type, |
2496 | bytes_obj); |
2497 | Py_DECREF(bytes_obj); |
2498 | } |
2499 | } |
2500 | if (reduce_value == NULL) |
2501 | return -1; |
2502 | |
2503 | /* save_reduce() will memoize the object automatically. */ |
2504 | status = save_reduce(self, reduce_value, obj); |
2505 | Py_DECREF(reduce_value); |
2506 | return status; |
2507 | } |
2508 | else { |
2509 | return _save_bytearray_data(self, obj, PyByteArray_AS_STRING(obj), |
2510 | PyByteArray_GET_SIZE(obj)); |
2511 | } |
2512 | } |
2513 | |
2514 | static int |
2515 | save_picklebuffer(PicklerObject *self, PyObject *obj) |
2516 | { |
2517 | if (self->proto < 5) { |
2518 | PickleState *st = _Pickle_GetGlobalState(); |
2519 | PyErr_SetString(st->PicklingError, |
2520 | "PickleBuffer can only pickled with protocol >= 5" ); |
2521 | return -1; |
2522 | } |
2523 | const Py_buffer* view = PyPickleBuffer_GetBuffer(obj); |
2524 | if (view == NULL) { |
2525 | return -1; |
2526 | } |
2527 | if (view->suboffsets != NULL || !PyBuffer_IsContiguous(view, 'A')) { |
2528 | PickleState *st = _Pickle_GetGlobalState(); |
2529 | PyErr_SetString(st->PicklingError, |
2530 | "PickleBuffer can not be pickled when " |
2531 | "pointing to a non-contiguous buffer" ); |
2532 | return -1; |
2533 | } |
2534 | int in_band = 1; |
2535 | if (self->buffer_callback != NULL) { |
2536 | PyObject *ret = PyObject_CallOneArg(self->buffer_callback, obj); |
2537 | if (ret == NULL) { |
2538 | return -1; |
2539 | } |
2540 | in_band = PyObject_IsTrue(ret); |
2541 | Py_DECREF(ret); |
2542 | if (in_band == -1) { |
2543 | return -1; |
2544 | } |
2545 | } |
2546 | if (in_band) { |
2547 | /* Write data in-band */ |
2548 | if (view->readonly) { |
2549 | return _save_bytes_data(self, obj, (const char*) view->buf, |
2550 | view->len); |
2551 | } |
2552 | else { |
2553 | return _save_bytearray_data(self, obj, (const char*) view->buf, |
2554 | view->len); |
2555 | } |
2556 | } |
2557 | else { |
2558 | /* Write data out-of-band */ |
2559 | const char next_buffer_op = NEXT_BUFFER; |
2560 | if (_Pickler_Write(self, &next_buffer_op, 1) < 0) { |
2561 | return -1; |
2562 | } |
2563 | if (view->readonly) { |
2564 | const char readonly_buffer_op = READONLY_BUFFER; |
2565 | if (_Pickler_Write(self, &readonly_buffer_op, 1) < 0) { |
2566 | return -1; |
2567 | } |
2568 | } |
2569 | } |
2570 | return 0; |
2571 | } |
2572 | |
2573 | /* A copy of PyUnicode_EncodeRawUnicodeEscape() that also translates |
2574 | backslash and newline characters to \uXXXX escapes. */ |
2575 | static PyObject * |
2576 | raw_unicode_escape(PyObject *obj) |
2577 | { |
2578 | char *p; |
2579 | Py_ssize_t i, size; |
2580 | const void *data; |
2581 | unsigned int kind; |
2582 | _PyBytesWriter writer; |
2583 | |
2584 | if (PyUnicode_READY(obj)) |
2585 | return NULL; |
2586 | |
2587 | _PyBytesWriter_Init(&writer); |
2588 | |
2589 | size = PyUnicode_GET_LENGTH(obj); |
2590 | data = PyUnicode_DATA(obj); |
2591 | kind = PyUnicode_KIND(obj); |
2592 | |
2593 | p = _PyBytesWriter_Alloc(&writer, size); |
2594 | if (p == NULL) |
2595 | goto error; |
2596 | writer.overallocate = 1; |
2597 | |
2598 | for (i=0; i < size; i++) { |
2599 | Py_UCS4 ch = PyUnicode_READ(kind, data, i); |
2600 | /* Map 32-bit characters to '\Uxxxxxxxx' */ |
2601 | if (ch >= 0x10000) { |
2602 | /* -1: subtract 1 preallocated byte */ |
2603 | p = _PyBytesWriter_Prepare(&writer, p, 10-1); |
2604 | if (p == NULL) |
2605 | goto error; |
2606 | |
2607 | *p++ = '\\'; |
2608 | *p++ = 'U'; |
2609 | *p++ = Py_hexdigits[(ch >> 28) & 0xf]; |
2610 | *p++ = Py_hexdigits[(ch >> 24) & 0xf]; |
2611 | *p++ = Py_hexdigits[(ch >> 20) & 0xf]; |
2612 | *p++ = Py_hexdigits[(ch >> 16) & 0xf]; |
2613 | *p++ = Py_hexdigits[(ch >> 12) & 0xf]; |
2614 | *p++ = Py_hexdigits[(ch >> 8) & 0xf]; |
2615 | *p++ = Py_hexdigits[(ch >> 4) & 0xf]; |
2616 | *p++ = Py_hexdigits[ch & 15]; |
2617 | } |
2618 | /* Map 16-bit characters, '\\' and '\n' to '\uxxxx' */ |
2619 | else if (ch >= 256 || |
2620 | ch == '\\' || ch == 0 || ch == '\n' || ch == '\r' || |
2621 | ch == 0x1a) |
2622 | { |
2623 | /* -1: subtract 1 preallocated byte */ |
2624 | p = _PyBytesWriter_Prepare(&writer, p, 6-1); |
2625 | if (p == NULL) |
2626 | goto error; |
2627 | |
2628 | *p++ = '\\'; |
2629 | *p++ = 'u'; |
2630 | *p++ = Py_hexdigits[(ch >> 12) & 0xf]; |
2631 | *p++ = Py_hexdigits[(ch >> 8) & 0xf]; |
2632 | *p++ = Py_hexdigits[(ch >> 4) & 0xf]; |
2633 | *p++ = Py_hexdigits[ch & 15]; |
2634 | } |
2635 | /* Copy everything else as-is */ |
2636 | else |
2637 | *p++ = (char) ch; |
2638 | } |
2639 | |
2640 | return _PyBytesWriter_Finish(&writer, p); |
2641 | |
2642 | error: |
2643 | _PyBytesWriter_Dealloc(&writer); |
2644 | return NULL; |
2645 | } |
2646 | |
2647 | static int |
2648 | write_unicode_binary(PicklerObject *self, PyObject *obj) |
2649 | { |
2650 | char [9]; |
2651 | Py_ssize_t len; |
2652 | PyObject *encoded = NULL; |
2653 | Py_ssize_t size; |
2654 | const char *data; |
2655 | |
2656 | if (PyUnicode_READY(obj)) |
2657 | return -1; |
2658 | |
2659 | data = PyUnicode_AsUTF8AndSize(obj, &size); |
2660 | if (data == NULL) { |
2661 | /* Issue #8383: for strings with lone surrogates, fallback on the |
2662 | "surrogatepass" error handler. */ |
2663 | PyErr_Clear(); |
2664 | encoded = PyUnicode_AsEncodedString(obj, "utf-8" , "surrogatepass" ); |
2665 | if (encoded == NULL) |
2666 | return -1; |
2667 | |
2668 | data = PyBytes_AS_STRING(encoded); |
2669 | size = PyBytes_GET_SIZE(encoded); |
2670 | } |
2671 | |
2672 | assert(size >= 0); |
2673 | if (size <= 0xff && self->proto >= 4) { |
2674 | header[0] = SHORT_BINUNICODE; |
2675 | header[1] = (unsigned char)(size & 0xff); |
2676 | len = 2; |
2677 | } |
2678 | else if ((size_t)size <= 0xffffffffUL) { |
2679 | header[0] = BINUNICODE; |
2680 | header[1] = (unsigned char)(size & 0xff); |
2681 | header[2] = (unsigned char)((size >> 8) & 0xff); |
2682 | header[3] = (unsigned char)((size >> 16) & 0xff); |
2683 | header[4] = (unsigned char)((size >> 24) & 0xff); |
2684 | len = 5; |
2685 | } |
2686 | else if (self->proto >= 4) { |
2687 | header[0] = BINUNICODE8; |
2688 | _write_size64(header + 1, size); |
2689 | len = 9; |
2690 | } |
2691 | else { |
2692 | PyErr_SetString(PyExc_OverflowError, |
2693 | "serializing a string larger than 4 GiB " |
2694 | "requires pickle protocol 4 or higher" ); |
2695 | Py_XDECREF(encoded); |
2696 | return -1; |
2697 | } |
2698 | |
2699 | if (_Pickler_write_bytes(self, header, len, data, size, encoded) < 0) { |
2700 | Py_XDECREF(encoded); |
2701 | return -1; |
2702 | } |
2703 | Py_XDECREF(encoded); |
2704 | return 0; |
2705 | } |
2706 | |
2707 | static int |
2708 | save_unicode(PicklerObject *self, PyObject *obj) |
2709 | { |
2710 | if (self->bin) { |
2711 | if (write_unicode_binary(self, obj) < 0) |
2712 | return -1; |
2713 | } |
2714 | else { |
2715 | PyObject *encoded; |
2716 | Py_ssize_t size; |
2717 | const char unicode_op = UNICODE; |
2718 | |
2719 | encoded = raw_unicode_escape(obj); |
2720 | if (encoded == NULL) |
2721 | return -1; |
2722 | |
2723 | if (_Pickler_Write(self, &unicode_op, 1) < 0) { |
2724 | Py_DECREF(encoded); |
2725 | return -1; |
2726 | } |
2727 | |
2728 | size = PyBytes_GET_SIZE(encoded); |
2729 | if (_Pickler_Write(self, PyBytes_AS_STRING(encoded), size) < 0) { |
2730 | Py_DECREF(encoded); |
2731 | return -1; |
2732 | } |
2733 | Py_DECREF(encoded); |
2734 | |
2735 | if (_Pickler_Write(self, "\n" , 1) < 0) |
2736 | return -1; |
2737 | } |
2738 | if (memo_put(self, obj) < 0) |
2739 | return -1; |
2740 | |
2741 | return 0; |
2742 | } |
2743 | |
2744 | /* A helper for save_tuple. Push the len elements in tuple t on the stack. */ |
2745 | static int |
2746 | store_tuple_elements(PicklerObject *self, PyObject *t, Py_ssize_t len) |
2747 | { |
2748 | Py_ssize_t i; |
2749 | |
2750 | assert(PyTuple_Size(t) == len); |
2751 | |
2752 | for (i = 0; i < len; i++) { |
2753 | PyObject *element = PyTuple_GET_ITEM(t, i); |
2754 | |
2755 | if (element == NULL) |
2756 | return -1; |
2757 | if (save(self, element, 0) < 0) |
2758 | return -1; |
2759 | } |
2760 | |
2761 | return 0; |
2762 | } |
2763 | |
2764 | /* Tuples are ubiquitous in the pickle protocols, so many techniques are |
2765 | * used across protocols to minimize the space needed to pickle them. |
2766 | * Tuples are also the only builtin immutable type that can be recursive |
2767 | * (a tuple can be reached from itself), and that requires some subtle |
2768 | * magic so that it works in all cases. IOW, this is a long routine. |
2769 | */ |
2770 | static int |
2771 | save_tuple(PicklerObject *self, PyObject *obj) |
2772 | { |
2773 | Py_ssize_t len, i; |
2774 | |
2775 | const char mark_op = MARK; |
2776 | const char tuple_op = TUPLE; |
2777 | const char pop_op = POP; |
2778 | const char pop_mark_op = POP_MARK; |
2779 | const char len2opcode[] = {EMPTY_TUPLE, TUPLE1, TUPLE2, TUPLE3}; |
2780 | |
2781 | if ((len = PyTuple_Size(obj)) < 0) |
2782 | return -1; |
2783 | |
2784 | if (len == 0) { |
2785 | char pdata[2]; |
2786 | |
2787 | if (self->proto) { |
2788 | pdata[0] = EMPTY_TUPLE; |
2789 | len = 1; |
2790 | } |
2791 | else { |
2792 | pdata[0] = MARK; |
2793 | pdata[1] = TUPLE; |
2794 | len = 2; |
2795 | } |
2796 | if (_Pickler_Write(self, pdata, len) < 0) |
2797 | return -1; |
2798 | return 0; |
2799 | } |
2800 | |
2801 | /* The tuple isn't in the memo now. If it shows up there after |
2802 | * saving the tuple elements, the tuple must be recursive, in |
2803 | * which case we'll pop everything we put on the stack, and fetch |
2804 | * its value from the memo. |
2805 | */ |
2806 | if (len <= 3 && self->proto >= 2) { |
2807 | /* Use TUPLE{1,2,3} opcodes. */ |
2808 | if (store_tuple_elements(self, obj, len) < 0) |
2809 | return -1; |
2810 | |
2811 | if (PyMemoTable_Get(self->memo, obj)) { |
2812 | /* pop the len elements */ |
2813 | for (i = 0; i < len; i++) |
2814 | if (_Pickler_Write(self, &pop_op, 1) < 0) |
2815 | return -1; |
2816 | /* fetch from memo */ |
2817 | if (memo_get(self, obj) < 0) |
2818 | return -1; |
2819 | |
2820 | return 0; |
2821 | } |
2822 | else { /* Not recursive. */ |
2823 | if (_Pickler_Write(self, len2opcode + len, 1) < 0) |
2824 | return -1; |
2825 | } |
2826 | goto memoize; |
2827 | } |
2828 | |
2829 | /* proto < 2 and len > 0, or proto >= 2 and len > 3. |
2830 | * Generate MARK e1 e2 ... TUPLE |
2831 | */ |
2832 | if (_Pickler_Write(self, &mark_op, 1) < 0) |
2833 | return -1; |
2834 | |
2835 | if (store_tuple_elements(self, obj, len) < 0) |
2836 | return -1; |
2837 | |
2838 | if (PyMemoTable_Get(self->memo, obj)) { |
2839 | /* pop the stack stuff we pushed */ |
2840 | if (self->bin) { |
2841 | if (_Pickler_Write(self, &pop_mark_op, 1) < 0) |
2842 | return -1; |
2843 | } |
2844 | else { |
2845 | /* Note that we pop one more than len, to remove |
2846 | * the MARK too. |
2847 | */ |
2848 | for (i = 0; i <= len; i++) |
2849 | if (_Pickler_Write(self, &pop_op, 1) < 0) |
2850 | return -1; |
2851 | } |
2852 | /* fetch from memo */ |
2853 | if (memo_get(self, obj) < 0) |
2854 | return -1; |
2855 | |
2856 | return 0; |
2857 | } |
2858 | else { /* Not recursive. */ |
2859 | if (_Pickler_Write(self, &tuple_op, 1) < 0) |
2860 | return -1; |
2861 | } |
2862 | |
2863 | memoize: |
2864 | if (memo_put(self, obj) < 0) |
2865 | return -1; |
2866 | |
2867 | return 0; |
2868 | } |
2869 | |
2870 | /* iter is an iterator giving items, and we batch up chunks of |
2871 | * MARK item item ... item APPENDS |
2872 | * opcode sequences. Calling code should have arranged to first create an |
2873 | * empty list, or list-like object, for the APPENDS to operate on. |
2874 | * Returns 0 on success, <0 on error. |
2875 | */ |
2876 | static int |
2877 | batch_list(PicklerObject *self, PyObject *iter) |
2878 | { |
2879 | PyObject *obj = NULL; |
2880 | PyObject *firstitem = NULL; |
2881 | int i, n; |
2882 | |
2883 | const char mark_op = MARK; |
2884 | const char append_op = APPEND; |
2885 | const char appends_op = APPENDS; |
2886 | |
2887 | assert(iter != NULL); |
2888 | |
2889 | /* XXX: I think this function could be made faster by avoiding the |
2890 | iterator interface and fetching objects directly from list using |
2891 | PyList_GET_ITEM. |
2892 | */ |
2893 | |
2894 | if (self->proto == 0) { |
2895 | /* APPENDS isn't available; do one at a time. */ |
2896 | for (;;) { |
2897 | obj = PyIter_Next(iter); |
2898 | if (obj == NULL) { |
2899 | if (PyErr_Occurred()) |
2900 | return -1; |
2901 | break; |
2902 | } |
2903 | i = save(self, obj, 0); |
2904 | Py_DECREF(obj); |
2905 | if (i < 0) |
2906 | return -1; |
2907 | if (_Pickler_Write(self, &append_op, 1) < 0) |
2908 | return -1; |
2909 | } |
2910 | return 0; |
2911 | } |
2912 | |
2913 | /* proto > 0: write in batches of BATCHSIZE. */ |
2914 | do { |
2915 | /* Get first item */ |
2916 | firstitem = PyIter_Next(iter); |
2917 | if (firstitem == NULL) { |
2918 | if (PyErr_Occurred()) |
2919 | goto error; |
2920 | |
2921 | /* nothing more to add */ |
2922 | break; |
2923 | } |
2924 | |
2925 | /* Try to get a second item */ |
2926 | obj = PyIter_Next(iter); |
2927 | if (obj == NULL) { |
2928 | if (PyErr_Occurred()) |
2929 | goto error; |
2930 | |
2931 | /* Only one item to write */ |
2932 | if (save(self, firstitem, 0) < 0) |
2933 | goto error; |
2934 | if (_Pickler_Write(self, &append_op, 1) < 0) |
2935 | goto error; |
2936 | Py_CLEAR(firstitem); |
2937 | break; |
2938 | } |
2939 | |
2940 | /* More than one item to write */ |
2941 | |
2942 | /* Pump out MARK, items, APPENDS. */ |
2943 | if (_Pickler_Write(self, &mark_op, 1) < 0) |
2944 | goto error; |
2945 | |
2946 | if (save(self, firstitem, 0) < 0) |
2947 | goto error; |
2948 | Py_CLEAR(firstitem); |
2949 | n = 1; |
2950 | |
2951 | /* Fetch and save up to BATCHSIZE items */ |
2952 | while (obj) { |
2953 | if (save(self, obj, 0) < 0) |
2954 | goto error; |
2955 | Py_CLEAR(obj); |
2956 | n += 1; |
2957 | |
2958 | if (n == BATCHSIZE) |
2959 | break; |
2960 | |
2961 | obj = PyIter_Next(iter); |
2962 | if (obj == NULL) { |
2963 | if (PyErr_Occurred()) |
2964 | goto error; |
2965 | break; |
2966 | } |
2967 | } |
2968 | |
2969 | if (_Pickler_Write(self, &appends_op, 1) < 0) |
2970 | goto error; |
2971 | |
2972 | } while (n == BATCHSIZE); |
2973 | return 0; |
2974 | |
2975 | error: |
2976 | Py_XDECREF(firstitem); |
2977 | Py_XDECREF(obj); |
2978 | return -1; |
2979 | } |
2980 | |
2981 | /* This is a variant of batch_list() above, specialized for lists (with no |
2982 | * support for list subclasses). Like batch_list(), we batch up chunks of |
2983 | * MARK item item ... item APPENDS |
2984 | * opcode sequences. Calling code should have arranged to first create an |
2985 | * empty list, or list-like object, for the APPENDS to operate on. |
2986 | * Returns 0 on success, -1 on error. |
2987 | * |
2988 | * This version is considerably faster than batch_list(), if less general. |
2989 | * |
2990 | * Note that this only works for protocols > 0. |
2991 | */ |
2992 | static int |
2993 | batch_list_exact(PicklerObject *self, PyObject *obj) |
2994 | { |
2995 | PyObject *item = NULL; |
2996 | Py_ssize_t this_batch, total; |
2997 | |
2998 | const char append_op = APPEND; |
2999 | const char appends_op = APPENDS; |
3000 | const char mark_op = MARK; |
3001 | |
3002 | assert(obj != NULL); |
3003 | assert(self->proto > 0); |
3004 | assert(PyList_CheckExact(obj)); |
3005 | |
3006 | if (PyList_GET_SIZE(obj) == 1) { |
3007 | item = PyList_GET_ITEM(obj, 0); |
3008 | if (save(self, item, 0) < 0) |
3009 | return -1; |
3010 | if (_Pickler_Write(self, &append_op, 1) < 0) |
3011 | return -1; |
3012 | return 0; |
3013 | } |
3014 | |
3015 | /* Write in batches of BATCHSIZE. */ |
3016 | total = 0; |
3017 | do { |
3018 | this_batch = 0; |
3019 | if (_Pickler_Write(self, &mark_op, 1) < 0) |
3020 | return -1; |
3021 | while (total < PyList_GET_SIZE(obj)) { |
3022 | item = PyList_GET_ITEM(obj, total); |
3023 | if (save(self, item, 0) < 0) |
3024 | return -1; |
3025 | total++; |
3026 | if (++this_batch == BATCHSIZE) |
3027 | break; |
3028 | } |
3029 | if (_Pickler_Write(self, &appends_op, 1) < 0) |
3030 | return -1; |
3031 | |
3032 | } while (total < PyList_GET_SIZE(obj)); |
3033 | |
3034 | return 0; |
3035 | } |
3036 | |
3037 | static int |
3038 | save_list(PicklerObject *self, PyObject *obj) |
3039 | { |
3040 | char [3]; |
3041 | Py_ssize_t len; |
3042 | int status = 0; |
3043 | |
3044 | if (self->fast && !fast_save_enter(self, obj)) |
3045 | goto error; |
3046 | |
3047 | /* Create an empty list. */ |
3048 | if (self->bin) { |
3049 | header[0] = EMPTY_LIST; |
3050 | len = 1; |
3051 | } |
3052 | else { |
3053 | header[0] = MARK; |
3054 | header[1] = LIST; |
3055 | len = 2; |
3056 | } |
3057 | |
3058 | if (_Pickler_Write(self, header, len) < 0) |
3059 | goto error; |
3060 | |
3061 | /* Get list length, and bow out early if empty. */ |
3062 | if ((len = PyList_Size(obj)) < 0) |
3063 | goto error; |
3064 | |
3065 | if (memo_put(self, obj) < 0) |
3066 | goto error; |
3067 | |
3068 | if (len != 0) { |
3069 | /* Materialize the list elements. */ |
3070 | if (PyList_CheckExact(obj) && self->proto > 0) { |
3071 | if (Py_EnterRecursiveCall(" while pickling an object" )) |
3072 | goto error; |
3073 | status = batch_list_exact(self, obj); |
3074 | Py_LeaveRecursiveCall(); |
3075 | } else { |
3076 | PyObject *iter = PyObject_GetIter(obj); |
3077 | if (iter == NULL) |
3078 | goto error; |
3079 | |
3080 | if (Py_EnterRecursiveCall(" while pickling an object" )) { |
3081 | Py_DECREF(iter); |
3082 | goto error; |
3083 | } |
3084 | status = batch_list(self, iter); |
3085 | Py_LeaveRecursiveCall(); |
3086 | Py_DECREF(iter); |
3087 | } |
3088 | } |
3089 | if (0) { |
3090 | error: |
3091 | status = -1; |
3092 | } |
3093 | |
3094 | if (self->fast && !fast_save_leave(self, obj)) |
3095 | status = -1; |
3096 | |
3097 | return status; |
3098 | } |
3099 | |
3100 | /* iter is an iterator giving (key, value) pairs, and we batch up chunks of |
3101 | * MARK key value ... key value SETITEMS |
3102 | * opcode sequences. Calling code should have arranged to first create an |
3103 | * empty dict, or dict-like object, for the SETITEMS to operate on. |
3104 | * Returns 0 on success, <0 on error. |
3105 | * |
3106 | * This is very much like batch_list(). The difference between saving |
3107 | * elements directly, and picking apart two-tuples, is so long-winded at |
3108 | * the C level, though, that attempts to combine these routines were too |
3109 | * ugly to bear. |
3110 | */ |
3111 | static int |
3112 | batch_dict(PicklerObject *self, PyObject *iter) |
3113 | { |
3114 | PyObject *obj = NULL; |
3115 | PyObject *firstitem = NULL; |
3116 | int i, n; |
3117 | |
3118 | const char mark_op = MARK; |
3119 | const char setitem_op = SETITEM; |
3120 | const char setitems_op = SETITEMS; |
3121 | |
3122 | assert(iter != NULL); |
3123 | |
3124 | if (self->proto == 0) { |
3125 | /* SETITEMS isn't available; do one at a time. */ |
3126 | for (;;) { |
3127 | obj = PyIter_Next(iter); |
3128 | if (obj == NULL) { |
3129 | if (PyErr_Occurred()) |
3130 | return -1; |
3131 | break; |
3132 | } |
3133 | if (!PyTuple_Check(obj) || PyTuple_Size(obj) != 2) { |
3134 | PyErr_SetString(PyExc_TypeError, "dict items " |
3135 | "iterator must return 2-tuples" ); |
3136 | return -1; |
3137 | } |
3138 | i = save(self, PyTuple_GET_ITEM(obj, 0), 0); |
3139 | if (i >= 0) |
3140 | i = save(self, PyTuple_GET_ITEM(obj, 1), 0); |
3141 | Py_DECREF(obj); |
3142 | if (i < 0) |
3143 | return -1; |
3144 | if (_Pickler_Write(self, &setitem_op, 1) < 0) |
3145 | return -1; |
3146 | } |
3147 | return 0; |
3148 | } |
3149 | |
3150 | /* proto > 0: write in batches of BATCHSIZE. */ |
3151 | do { |
3152 | /* Get first item */ |
3153 | firstitem = PyIter_Next(iter); |
3154 | if (firstitem == NULL) { |
3155 | if (PyErr_Occurred()) |
3156 | goto error; |
3157 | |
3158 | /* nothing more to add */ |
3159 | break; |
3160 | } |
3161 | if (!PyTuple_Check(firstitem) || PyTuple_Size(firstitem) != 2) { |
3162 | PyErr_SetString(PyExc_TypeError, "dict items " |
3163 | "iterator must return 2-tuples" ); |
3164 | goto error; |
3165 | } |
3166 | |
3167 | /* Try to get a second item */ |
3168 | obj = PyIter_Next(iter); |
3169 | if (obj == NULL) { |
3170 | if (PyErr_Occurred()) |
3171 | goto error; |
3172 | |
3173 | /* Only one item to write */ |
3174 | if (save(self, PyTuple_GET_ITEM(firstitem, 0), 0) < 0) |
3175 | goto error; |
3176 | if (save(self, PyTuple_GET_ITEM(firstitem, 1), 0) < 0) |
3177 | goto error; |
3178 | if (_Pickler_Write(self, &setitem_op, 1) < 0) |
3179 | goto error; |
3180 | Py_CLEAR(firstitem); |
3181 | break; |
3182 | } |
3183 | |
3184 | /* More than one item to write */ |
3185 | |
3186 | /* Pump out MARK, items, SETITEMS. */ |
3187 | if (_Pickler_Write(self, &mark_op, 1) < 0) |
3188 | goto error; |
3189 | |
3190 | if (save(self, PyTuple_GET_ITEM(firstitem, 0), 0) < 0) |
3191 | goto error; |
3192 | if (save(self, PyTuple_GET_ITEM(firstitem, 1), 0) < 0) |
3193 | goto error; |
3194 | Py_CLEAR(firstitem); |
3195 | n = 1; |
3196 | |
3197 | /* Fetch and save up to BATCHSIZE items */ |
3198 | while (obj) { |
3199 | if (!PyTuple_Check(obj) || PyTuple_Size(obj) != 2) { |
3200 | PyErr_SetString(PyExc_TypeError, "dict items " |
3201 | "iterator must return 2-tuples" ); |
3202 | goto error; |
3203 | } |
3204 | if (save(self, PyTuple_GET_ITEM(obj, 0), 0) < 0 || |
3205 | save(self, PyTuple_GET_ITEM(obj, 1), 0) < 0) |
3206 | goto error; |
3207 | Py_CLEAR(obj); |
3208 | n += 1; |
3209 | |
3210 | if (n == BATCHSIZE) |
3211 | break; |
3212 | |
3213 | obj = PyIter_Next(iter); |
3214 | if (obj == NULL) { |
3215 | if (PyErr_Occurred()) |
3216 | goto error; |
3217 | break; |
3218 | } |
3219 | } |
3220 | |
3221 | if (_Pickler_Write(self, &setitems_op, 1) < 0) |
3222 | goto error; |
3223 | |
3224 | } while (n == BATCHSIZE); |
3225 | return 0; |
3226 | |
3227 | error: |
3228 | Py_XDECREF(firstitem); |
3229 | Py_XDECREF(obj); |
3230 | return -1; |
3231 | } |
3232 | |
3233 | /* This is a variant of batch_dict() above that specializes for dicts, with no |
3234 | * support for dict subclasses. Like batch_dict(), we batch up chunks of |
3235 | * MARK key value ... key value SETITEMS |
3236 | * opcode sequences. Calling code should have arranged to first create an |
3237 | * empty dict, or dict-like object, for the SETITEMS to operate on. |
3238 | * Returns 0 on success, -1 on error. |
3239 | * |
3240 | * Note that this currently doesn't work for protocol 0. |
3241 | */ |
3242 | static int |
3243 | batch_dict_exact(PicklerObject *self, PyObject *obj) |
3244 | { |
3245 | PyObject *key = NULL, *value = NULL; |
3246 | int i; |
3247 | Py_ssize_t dict_size, ppos = 0; |
3248 | |
3249 | const char mark_op = MARK; |
3250 | const char setitem_op = SETITEM; |
3251 | const char setitems_op = SETITEMS; |
3252 | |
3253 | assert(obj != NULL && PyDict_CheckExact(obj)); |
3254 | assert(self->proto > 0); |
3255 | |
3256 | dict_size = PyDict_GET_SIZE(obj); |
3257 | |
3258 | /* Special-case len(d) == 1 to save space. */ |
3259 | if (dict_size == 1) { |
3260 | PyDict_Next(obj, &ppos, &key, &value); |
3261 | if (save(self, key, 0) < 0) |
3262 | return -1; |
3263 | if (save(self, value, 0) < 0) |
3264 | return -1; |
3265 | if (_Pickler_Write(self, &setitem_op, 1) < 0) |
3266 | return -1; |
3267 | return 0; |
3268 | } |
3269 | |
3270 | /* Write in batches of BATCHSIZE. */ |
3271 | do { |
3272 | i = 0; |
3273 | if (_Pickler_Write(self, &mark_op, 1) < 0) |
3274 | return -1; |
3275 | while (PyDict_Next(obj, &ppos, &key, &value)) { |
3276 | if (save(self, key, 0) < 0) |
3277 | return -1; |
3278 | if (save(self, value, 0) < 0) |
3279 | return -1; |
3280 | if (++i == BATCHSIZE) |
3281 | break; |
3282 | } |
3283 | if (_Pickler_Write(self, &setitems_op, 1) < 0) |
3284 | return -1; |
3285 | if (PyDict_GET_SIZE(obj) != dict_size) { |
3286 | PyErr_Format( |
3287 | PyExc_RuntimeError, |
3288 | "dictionary changed size during iteration" ); |
3289 | return -1; |
3290 | } |
3291 | |
3292 | } while (i == BATCHSIZE); |
3293 | return 0; |
3294 | } |
3295 | |
3296 | static int |
3297 | save_dict(PicklerObject *self, PyObject *obj) |
3298 | { |
3299 | PyObject *items, *iter; |
3300 | char [3]; |
3301 | Py_ssize_t len; |
3302 | int status = 0; |
3303 | assert(PyDict_Check(obj)); |
3304 | |
3305 | if (self->fast && !fast_save_enter(self, obj)) |
3306 | goto error; |
3307 | |
3308 | /* Create an empty dict. */ |
3309 | if (self->bin) { |
3310 | header[0] = EMPTY_DICT; |
3311 | len = 1; |
3312 | } |
3313 | else { |
3314 | header[0] = MARK; |
3315 | header[1] = DICT; |
3316 | len = 2; |
3317 | } |
3318 | |
3319 | if (_Pickler_Write(self, header, len) < 0) |
3320 | goto error; |
3321 | |
3322 | if (memo_put(self, obj) < 0) |
3323 | goto error; |
3324 | |
3325 | if (PyDict_GET_SIZE(obj)) { |
3326 | /* Save the dict items. */ |
3327 | if (PyDict_CheckExact(obj) && self->proto > 0) { |
3328 | /* We can take certain shortcuts if we know this is a dict and |
3329 | not a dict subclass. */ |
3330 | if (Py_EnterRecursiveCall(" while pickling an object" )) |
3331 | goto error; |
3332 | status = batch_dict_exact(self, obj); |
3333 | Py_LeaveRecursiveCall(); |
3334 | } else { |
3335 | _Py_IDENTIFIER(items); |
3336 | |
3337 | items = _PyObject_CallMethodIdNoArgs(obj, &PyId_items); |
3338 | if (items == NULL) |
3339 | goto error; |
3340 | iter = PyObject_GetIter(items); |
3341 | Py_DECREF(items); |
3342 | if (iter == NULL) |
3343 | goto error; |
3344 | if (Py_EnterRecursiveCall(" while pickling an object" )) { |
3345 | Py_DECREF(iter); |
3346 | goto error; |
3347 | } |
3348 | status = batch_dict(self, iter); |
3349 | Py_LeaveRecursiveCall(); |
3350 | Py_DECREF(iter); |
3351 | } |
3352 | } |
3353 | |
3354 | if (0) { |
3355 | error: |
3356 | status = -1; |
3357 | } |
3358 | |
3359 | if (self->fast && !fast_save_leave(self, obj)) |
3360 | status = -1; |
3361 | |
3362 | return status; |
3363 | } |
3364 | |
3365 | static int |
3366 | save_set(PicklerObject *self, PyObject *obj) |
3367 | { |
3368 | PyObject *item; |
3369 | int i; |
3370 | Py_ssize_t set_size, ppos = 0; |
3371 | Py_hash_t hash; |
3372 | |
3373 | const char empty_set_op = EMPTY_SET; |
3374 | const char mark_op = MARK; |
3375 | const char additems_op = ADDITEMS; |
3376 | |
3377 | if (self->proto < 4) { |
3378 | PyObject *items; |
3379 | PyObject *reduce_value; |
3380 | int status; |
3381 | |
3382 | items = PySequence_List(obj); |
3383 | if (items == NULL) { |
3384 | return -1; |
3385 | } |
3386 | reduce_value = Py_BuildValue("(O(O))" , (PyObject*)&PySet_Type, items); |
3387 | Py_DECREF(items); |
3388 | if (reduce_value == NULL) { |
3389 | return -1; |
3390 | } |
3391 | /* save_reduce() will memoize the object automatically. */ |
3392 | status = save_reduce(self, reduce_value, obj); |
3393 | Py_DECREF(reduce_value); |
3394 | return status; |
3395 | } |
3396 | |
3397 | if (_Pickler_Write(self, &empty_set_op, 1) < 0) |
3398 | return -1; |
3399 | |
3400 | if (memo_put(self, obj) < 0) |
3401 | return -1; |
3402 | |
3403 | set_size = PySet_GET_SIZE(obj); |
3404 | if (set_size == 0) |
3405 | return 0; /* nothing to do */ |
3406 | |
3407 | /* Write in batches of BATCHSIZE. */ |
3408 | do { |
3409 | i = 0; |
3410 | if (_Pickler_Write(self, &mark_op, 1) < 0) |
3411 | return -1; |
3412 | while (_PySet_NextEntry(obj, &ppos, &item, &hash)) { |
3413 | if (save(self, item, 0) < 0) |
3414 | return -1; |
3415 | if (++i == BATCHSIZE) |
3416 | break; |
3417 | } |
3418 | if (_Pickler_Write(self, &additems_op, 1) < 0) |
3419 | return -1; |
3420 | if (PySet_GET_SIZE(obj) != set_size) { |
3421 | PyErr_Format( |
3422 | PyExc_RuntimeError, |
3423 | "set changed size during iteration" ); |
3424 | return -1; |
3425 | } |
3426 | } while (i == BATCHSIZE); |
3427 | |
3428 | return 0; |
3429 | } |
3430 | |
3431 | static int |
3432 | save_frozenset(PicklerObject *self, PyObject *obj) |
3433 | { |
3434 | PyObject *iter; |
3435 | |
3436 | const char mark_op = MARK; |
3437 | const char frozenset_op = FROZENSET; |
3438 | |
3439 | if (self->fast && !fast_save_enter(self, obj)) |
3440 | return -1; |
3441 | |
3442 | if (self->proto < 4) { |
3443 | PyObject *items; |
3444 | PyObject *reduce_value; |
3445 | int status; |
3446 | |
3447 | items = PySequence_List(obj); |
3448 | if (items == NULL) { |
3449 | return -1; |
3450 | } |
3451 | reduce_value = Py_BuildValue("(O(O))" , (PyObject*)&PyFrozenSet_Type, |
3452 | items); |
3453 | Py_DECREF(items); |
3454 | if (reduce_value == NULL) { |
3455 | return -1; |
3456 | } |
3457 | /* save_reduce() will memoize the object automatically. */ |
3458 | status = save_reduce(self, reduce_value, obj); |
3459 | Py_DECREF(reduce_value); |
3460 | return status; |
3461 | } |
3462 | |
3463 | if (_Pickler_Write(self, &mark_op, 1) < 0) |
3464 | return -1; |
3465 | |
3466 | iter = PyObject_GetIter(obj); |
3467 | if (iter == NULL) { |
3468 | return -1; |
3469 | } |
3470 | for (;;) { |
3471 | PyObject *item; |
3472 | |
3473 | item = PyIter_Next(iter); |
3474 | if (item == NULL) { |
3475 | if (PyErr_Occurred()) { |
3476 | Py_DECREF(iter); |
3477 | return -1; |
3478 | } |
3479 | break; |
3480 | } |
3481 | if (save(self, item, 0) < 0) { |
3482 | Py_DECREF(item); |
3483 | Py_DECREF(iter); |
3484 | return -1; |
3485 | } |
3486 | Py_DECREF(item); |
3487 | } |
3488 | Py_DECREF(iter); |
3489 | |
3490 | /* If the object is already in the memo, this means it is |
3491 | recursive. In this case, throw away everything we put on the |
3492 | stack, and fetch the object back from the memo. */ |
3493 | if (PyMemoTable_Get(self->memo, obj)) { |
3494 | const char pop_mark_op = POP_MARK; |
3495 | |
3496 | if (_Pickler_Write(self, &pop_mark_op, 1) < 0) |
3497 | return -1; |
3498 | if (memo_get(self, obj) < 0) |
3499 | return -1; |
3500 | return 0; |
3501 | } |
3502 | |
3503 | if (_Pickler_Write(self, &frozenset_op, 1) < 0) |
3504 | return -1; |
3505 | if (memo_put(self, obj) < 0) |
3506 | return -1; |
3507 | |
3508 | return 0; |
3509 | } |
3510 | |
3511 | static int |
3512 | fix_imports(PyObject **module_name, PyObject **global_name) |
3513 | { |
3514 | PyObject *key; |
3515 | PyObject *item; |
3516 | PickleState *st = _Pickle_GetGlobalState(); |
3517 | |
3518 | key = PyTuple_Pack(2, *module_name, *global_name); |
3519 | if (key == NULL) |
3520 | return -1; |
3521 | item = PyDict_GetItemWithError(st->name_mapping_3to2, key); |
3522 | Py_DECREF(key); |
3523 | if (item) { |
3524 | PyObject *fixed_module_name; |
3525 | PyObject *fixed_global_name; |
3526 | |
3527 | if (!PyTuple_Check(item) || PyTuple_GET_SIZE(item) != 2) { |
3528 | PyErr_Format(PyExc_RuntimeError, |
3529 | "_compat_pickle.REVERSE_NAME_MAPPING values " |
3530 | "should be 2-tuples, not %.200s" , |
3531 | Py_TYPE(item)->tp_name); |
3532 | return -1; |
3533 | } |
3534 | fixed_module_name = PyTuple_GET_ITEM(item, 0); |
3535 | fixed_global_name = PyTuple_GET_ITEM(item, 1); |
3536 | if (!PyUnicode_Check(fixed_module_name) || |
3537 | !PyUnicode_Check(fixed_global_name)) { |
3538 | PyErr_Format(PyExc_RuntimeError, |
3539 | "_compat_pickle.REVERSE_NAME_MAPPING values " |
3540 | "should be pairs of str, not (%.200s, %.200s)" , |
3541 | Py_TYPE(fixed_module_name)->tp_name, |
3542 | Py_TYPE(fixed_global_name)->tp_name); |
3543 | return -1; |
3544 | } |
3545 | |
3546 | Py_CLEAR(*module_name); |
3547 | Py_CLEAR(*global_name); |
3548 | Py_INCREF(fixed_module_name); |
3549 | Py_INCREF(fixed_global_name); |
3550 | *module_name = fixed_module_name; |
3551 | *global_name = fixed_global_name; |
3552 | return 0; |
3553 | } |
3554 | else if (PyErr_Occurred()) { |
3555 | return -1; |
3556 | } |
3557 | |
3558 | item = PyDict_GetItemWithError(st->import_mapping_3to2, *module_name); |
3559 | if (item) { |
3560 | if (!PyUnicode_Check(item)) { |
3561 | PyErr_Format(PyExc_RuntimeError, |
3562 | "_compat_pickle.REVERSE_IMPORT_MAPPING values " |
3563 | "should be strings, not %.200s" , |
3564 | Py_TYPE(item)->tp_name); |
3565 | return -1; |
3566 | } |
3567 | Py_INCREF(item); |
3568 | Py_XSETREF(*module_name, item); |
3569 | } |
3570 | else if (PyErr_Occurred()) { |
3571 | return -1; |
3572 | } |
3573 | |
3574 | return 0; |
3575 | } |
3576 | |
3577 | static int |
3578 | save_global(PicklerObject *self, PyObject *obj, PyObject *name) |
3579 | { |
3580 | PyObject *global_name = NULL; |
3581 | PyObject *module_name = NULL; |
3582 | PyObject *module = NULL; |
3583 | PyObject *parent = NULL; |
3584 | PyObject *dotted_path = NULL; |
3585 | PyObject *lastname = NULL; |
3586 | PyObject *cls; |
3587 | PickleState *st = _Pickle_GetGlobalState(); |
3588 | int status = 0; |
3589 | _Py_IDENTIFIER(__name__); |
3590 | _Py_IDENTIFIER(__qualname__); |
3591 | |
3592 | const char global_op = GLOBAL; |
3593 | |
3594 | if (name) { |
3595 | Py_INCREF(name); |
3596 | global_name = name; |
3597 | } |
3598 | else { |
3599 | if (_PyObject_LookupAttrId(obj, &PyId___qualname__, &global_name) < 0) |
3600 | goto error; |
3601 | if (global_name == NULL) { |
3602 | global_name = _PyObject_GetAttrId(obj, &PyId___name__); |
3603 | if (global_name == NULL) |
3604 | goto error; |
3605 | } |
3606 | } |
3607 | |
3608 | dotted_path = get_dotted_path(module, global_name); |
3609 | if (dotted_path == NULL) |
3610 | goto error; |
3611 | module_name = whichmodule(obj, dotted_path); |
3612 | if (module_name == NULL) |
3613 | goto error; |
3614 | |
3615 | /* XXX: Change to use the import C API directly with level=0 to disallow |
3616 | relative imports. |
3617 | |
3618 | XXX: PyImport_ImportModuleLevel could be used. However, this bypasses |
3619 | builtins.__import__. Therefore, _pickle, unlike pickle.py, will ignore |
3620 | custom import functions (IMHO, this would be a nice security |
3621 | feature). The import C API would need to be extended to support the |
3622 | extra parameters of __import__ to fix that. */ |
3623 | module = PyImport_Import(module_name); |
3624 | if (module == NULL) { |
3625 | PyErr_Format(st->PicklingError, |
3626 | "Can't pickle %R: import of module %R failed" , |
3627 | obj, module_name); |
3628 | goto error; |
3629 | } |
3630 | lastname = PyList_GET_ITEM(dotted_path, PyList_GET_SIZE(dotted_path)-1); |
3631 | Py_INCREF(lastname); |
3632 | cls = get_deep_attribute(module, dotted_path, &parent); |
3633 | Py_CLEAR(dotted_path); |
3634 | if (cls == NULL) { |
3635 | PyErr_Format(st->PicklingError, |
3636 | "Can't pickle %R: attribute lookup %S on %S failed" , |
3637 | obj, global_name, module_name); |
3638 | goto error; |
3639 | } |
3640 | if (cls != obj) { |
3641 | Py_DECREF(cls); |
3642 | PyErr_Format(st->PicklingError, |
3643 | "Can't pickle %R: it's not the same object as %S.%S" , |
3644 | obj, module_name, global_name); |
3645 | goto error; |
3646 | } |
3647 | Py_DECREF(cls); |
3648 | |
3649 | if (self->proto >= 2) { |
3650 | /* See whether this is in the extension registry, and if |
3651 | * so generate an EXT opcode. |
3652 | */ |
3653 | PyObject *extension_key; |
3654 | PyObject *code_obj; /* extension code as Python object */ |
3655 | long code; /* extension code as C value */ |
3656 | char pdata[5]; |
3657 | Py_ssize_t n; |
3658 | |
3659 | extension_key = PyTuple_Pack(2, module_name, global_name); |
3660 | if (extension_key == NULL) { |
3661 | goto error; |
3662 | } |
3663 | code_obj = PyDict_GetItemWithError(st->extension_registry, |
3664 | extension_key); |
3665 | Py_DECREF(extension_key); |
3666 | /* The object is not registered in the extension registry. |
3667 | This is the most likely code path. */ |
3668 | if (code_obj == NULL) { |
3669 | if (PyErr_Occurred()) { |
3670 | goto error; |
3671 | } |
3672 | goto gen_global; |
3673 | } |
3674 | |
3675 | /* XXX: pickle.py doesn't check neither the type, nor the range |
3676 | of the value returned by the extension_registry. It should for |
3677 | consistency. */ |
3678 | |
3679 | /* Verify code_obj has the right type and value. */ |
3680 | if (!PyLong_Check(code_obj)) { |
3681 | PyErr_Format(st->PicklingError, |
3682 | "Can't pickle %R: extension code %R isn't an integer" , |
3683 | obj, code_obj); |
3684 | goto error; |
3685 | } |
3686 | code = PyLong_AS_LONG(code_obj); |
3687 | if (code <= 0 || code > 0x7fffffffL) { |
3688 | if (!PyErr_Occurred()) |
3689 | PyErr_Format(st->PicklingError, "Can't pickle %R: extension " |
3690 | "code %ld is out of range" , obj, code); |
3691 | goto error; |
3692 | } |
3693 | |
3694 | /* Generate an EXT opcode. */ |
3695 | if (code <= 0xff) { |
3696 | pdata[0] = EXT1; |
3697 | pdata[1] = (unsigned char)code; |
3698 | n = 2; |
3699 | } |
3700 | else if (code <= 0xffff) { |
3701 | pdata[0] = EXT2; |
3702 | pdata[1] = (unsigned char)(code & 0xff); |
3703 | pdata[2] = (unsigned char)((code >> 8) & 0xff); |
3704 | n = 3; |
3705 | } |
3706 | else { |
3707 | pdata[0] = EXT4; |
3708 | pdata[1] = (unsigned char)(code & 0xff); |
3709 | pdata[2] = (unsigned char)((code >> 8) & 0xff); |
3710 | pdata[3] = (unsigned char)((code >> 16) & 0xff); |
3711 | pdata[4] = (unsigned char)((code >> 24) & 0xff); |
3712 | n = 5; |
3713 | } |
3714 | |
3715 | if (_Pickler_Write(self, pdata, n) < 0) |
3716 | goto error; |
3717 | } |
3718 | else { |
3719 | gen_global: |
3720 | if (parent == module) { |
3721 | Py_INCREF(lastname); |
3722 | Py_DECREF(global_name); |
3723 | global_name = lastname; |
3724 | } |
3725 | if (self->proto >= 4) { |
3726 | const char stack_global_op = STACK_GLOBAL; |
3727 | |
3728 | if (save(self, module_name, 0) < 0) |
3729 | goto error; |
3730 | if (save(self, global_name, 0) < 0) |
3731 | goto error; |
3732 | |
3733 | if (_Pickler_Write(self, &stack_global_op, 1) < 0) |
3734 | goto error; |
3735 | } |
3736 | else if (parent != module) { |
3737 | PickleState *st = _Pickle_GetGlobalState(); |
3738 | PyObject *reduce_value = Py_BuildValue("(O(OO))" , |
3739 | st->getattr, parent, lastname); |
3740 | if (reduce_value == NULL) |
3741 | goto error; |
3742 | status = save_reduce(self, reduce_value, NULL); |
3743 | Py_DECREF(reduce_value); |
3744 | if (status < 0) |
3745 | goto error; |
3746 | } |
3747 | else { |
3748 | /* Generate a normal global opcode if we are using a pickle |
3749 | protocol < 4, or if the object is not registered in the |
3750 | extension registry. */ |
3751 | PyObject *encoded; |
3752 | PyObject *(*unicode_encoder)(PyObject *); |
3753 | |
3754 | if (_Pickler_Write(self, &global_op, 1) < 0) |
3755 | goto error; |
3756 | |
3757 | /* For protocol < 3 and if the user didn't request against doing |
3758 | so, we convert module names to the old 2.x module names. */ |
3759 | if (self->proto < 3 && self->fix_imports) { |
3760 | if (fix_imports(&module_name, &global_name) < 0) { |
3761 | goto error; |
3762 | } |
3763 | } |
3764 | |
3765 | /* Since Python 3.0 now supports non-ASCII identifiers, we encode |
3766 | both the module name and the global name using UTF-8. We do so |
3767 | only when we are using the pickle protocol newer than version |
3768 | 3. This is to ensure compatibility with older Unpickler running |
3769 | on Python 2.x. */ |
3770 | if (self->proto == 3) { |
3771 | unicode_encoder = PyUnicode_AsUTF8String; |
3772 | } |
3773 | else { |
3774 | unicode_encoder = PyUnicode_AsASCIIString; |
3775 | } |
3776 | encoded = unicode_encoder(module_name); |
3777 | if (encoded == NULL) { |
3778 | if (PyErr_ExceptionMatches(PyExc_UnicodeEncodeError)) |
3779 | PyErr_Format(st->PicklingError, |
3780 | "can't pickle module identifier '%S' using " |
3781 | "pickle protocol %i" , |
3782 | module_name, self->proto); |
3783 | goto error; |
3784 | } |
3785 | if (_Pickler_Write(self, PyBytes_AS_STRING(encoded), |
3786 | PyBytes_GET_SIZE(encoded)) < 0) { |
3787 | Py_DECREF(encoded); |
3788 | goto error; |
3789 | } |
3790 | Py_DECREF(encoded); |
3791 | if(_Pickler_Write(self, "\n" , 1) < 0) |
3792 | goto error; |
3793 | |
3794 | /* Save the name of the module. */ |
3795 | encoded = unicode_encoder(global_name); |
3796 | if (encoded == NULL) { |
3797 | if (PyErr_ExceptionMatches(PyExc_UnicodeEncodeError)) |
3798 | PyErr_Format(st->PicklingError, |
3799 | "can't pickle global identifier '%S' using " |
3800 | "pickle protocol %i" , |
3801 | global_name, self->proto); |
3802 | goto error; |
3803 | } |
3804 | if (_Pickler_Write(self, PyBytes_AS_STRING(encoded), |
3805 | PyBytes_GET_SIZE(encoded)) < 0) { |
3806 | Py_DECREF(encoded); |
3807 | goto error; |
3808 | } |
3809 | Py_DECREF(encoded); |
3810 | if (_Pickler_Write(self, "\n" , 1) < 0) |
3811 | goto error; |
3812 | } |
3813 | /* Memoize the object. */ |
3814 | if (memo_put(self, obj) < 0) |
3815 | goto error; |
3816 | } |
3817 | |
3818 | if (0) { |
3819 | error: |
3820 | status = -1; |
3821 | } |
3822 | Py_XDECREF(module_name); |
3823 | Py_XDECREF(global_name); |
3824 | Py_XDECREF(module); |
3825 | Py_XDECREF(parent); |
3826 | Py_XDECREF(dotted_path); |
3827 | Py_XDECREF(lastname); |
3828 | |
3829 | return status; |
3830 | } |
3831 | |
3832 | static int |
3833 | save_singleton_type(PicklerObject *self, PyObject *obj, PyObject *singleton) |
3834 | { |
3835 | PyObject *reduce_value; |
3836 | int status; |
3837 | |
3838 | reduce_value = Py_BuildValue("O(O)" , &PyType_Type, singleton); |
3839 | if (reduce_value == NULL) { |
3840 | return -1; |
3841 | } |
3842 | status = save_reduce(self, reduce_value, obj); |
3843 | Py_DECREF(reduce_value); |
3844 | return status; |
3845 | } |
3846 | |
3847 | static int |
3848 | save_type(PicklerObject *self, PyObject *obj) |
3849 | { |
3850 | if (obj == (PyObject *)&_PyNone_Type) { |
3851 | return save_singleton_type(self, obj, Py_None); |
3852 | } |
3853 | else if (obj == (PyObject *)&PyEllipsis_Type) { |
3854 | return save_singleton_type(self, obj, Py_Ellipsis); |
3855 | } |
3856 | else if (obj == (PyObject *)&_PyNotImplemented_Type) { |
3857 | return save_singleton_type(self, obj, Py_NotImplemented); |
3858 | } |
3859 | return save_global(self, obj, NULL); |
3860 | } |
3861 | |
3862 | static int |
3863 | save_pers(PicklerObject *self, PyObject *obj) |
3864 | { |
3865 | PyObject *pid = NULL; |
3866 | int status = 0; |
3867 | |
3868 | const char persid_op = PERSID; |
3869 | const char binpersid_op = BINPERSID; |
3870 | |
3871 | pid = call_method(self->pers_func, self->pers_func_self, obj); |
3872 | if (pid == NULL) |
3873 | return -1; |
3874 | |
3875 | if (pid != Py_None) { |
3876 | if (self->bin) { |
3877 | if (save(self, pid, 1) < 0 || |
3878 | _Pickler_Write(self, &binpersid_op, 1) < 0) |
3879 | goto error; |
3880 | } |
3881 | else { |
3882 | PyObject *pid_str; |
3883 | |
3884 | pid_str = PyObject_Str(pid); |
3885 | if (pid_str == NULL) |
3886 | goto error; |
3887 | |
3888 | /* XXX: Should it check whether the pid contains embedded |
3889 | newlines? */ |
3890 | if (!PyUnicode_IS_ASCII(pid_str)) { |
3891 | PyErr_SetString(_Pickle_GetGlobalState()->PicklingError, |
3892 | "persistent IDs in protocol 0 must be " |
3893 | "ASCII strings" ); |
3894 | Py_DECREF(pid_str); |
3895 | goto error; |
3896 | } |
3897 | |
3898 | if (_Pickler_Write(self, &persid_op, 1) < 0 || |
3899 | _Pickler_Write(self, PyUnicode_DATA(pid_str), |
3900 | PyUnicode_GET_LENGTH(pid_str)) < 0 || |
3901 | _Pickler_Write(self, "\n" , 1) < 0) { |
3902 | Py_DECREF(pid_str); |
3903 | goto error; |
3904 | } |
3905 | Py_DECREF(pid_str); |
3906 | } |
3907 | status = 1; |
3908 | } |
3909 | |
3910 | if (0) { |
3911 | error: |
3912 | status = -1; |
3913 | } |
3914 | Py_XDECREF(pid); |
3915 | |
3916 | return status; |
3917 | } |
3918 | |
3919 | static PyObject * |
3920 | get_class(PyObject *obj) |
3921 | { |
3922 | PyObject *cls; |
3923 | _Py_IDENTIFIER(__class__); |
3924 | |
3925 | if (_PyObject_LookupAttrId(obj, &PyId___class__, &cls) == 0) { |
3926 | cls = (PyObject *) Py_TYPE(obj); |
3927 | Py_INCREF(cls); |
3928 | } |
3929 | return cls; |
3930 | } |
3931 | |
3932 | /* We're saving obj, and args is the 2-thru-5 tuple returned by the |
3933 | * appropriate __reduce__ method for obj. |
3934 | */ |
3935 | static int |
3936 | save_reduce(PicklerObject *self, PyObject *args, PyObject *obj) |
3937 | { |
3938 | PyObject *callable; |
3939 | PyObject *argtup; |
3940 | PyObject *state = NULL; |
3941 | PyObject *listitems = Py_None; |
3942 | PyObject *dictitems = Py_None; |
3943 | PyObject *state_setter = Py_None; |
3944 | PickleState *st = _Pickle_GetGlobalState(); |
3945 | Py_ssize_t size; |
3946 | int use_newobj = 0, use_newobj_ex = 0; |
3947 | |
3948 | const char reduce_op = REDUCE; |
3949 | const char build_op = BUILD; |
3950 | const char newobj_op = NEWOBJ; |
3951 | const char newobj_ex_op = NEWOBJ_EX; |
3952 | |
3953 | size = PyTuple_Size(args); |
3954 | if (size < 2 || size > 6) { |
3955 | PyErr_SetString(st->PicklingError, "tuple returned by " |
3956 | "__reduce__ must contain 2 through 6 elements" ); |
3957 | return -1; |
3958 | } |
3959 | |
3960 | if (!PyArg_UnpackTuple(args, "save_reduce" , 2, 6, |
3961 | &callable, &argtup, &state, &listitems, &dictitems, |
3962 | &state_setter)) |
3963 | return -1; |
3964 | |
3965 | if (!PyCallable_Check(callable)) { |
3966 | PyErr_SetString(st->PicklingError, "first item of the tuple " |
3967 | "returned by __reduce__ must be callable" ); |
3968 | return -1; |
3969 | } |
3970 | if (!PyTuple_Check(argtup)) { |
3971 | PyErr_SetString(st->PicklingError, "second item of the tuple " |
3972 | "returned by __reduce__ must be a tuple" ); |
3973 | return -1; |
3974 | } |
3975 | |
3976 | if (state == Py_None) |
3977 | state = NULL; |
3978 | |
3979 | if (listitems == Py_None) |
3980 | listitems = NULL; |
3981 | else if (!PyIter_Check(listitems)) { |
3982 | PyErr_Format(st->PicklingError, "fourth element of the tuple " |
3983 | "returned by __reduce__ must be an iterator, not %s" , |
3984 | Py_TYPE(listitems)->tp_name); |
3985 | return -1; |
3986 | } |
3987 | |
3988 | if (dictitems == Py_None) |
3989 | dictitems = NULL; |
3990 | else if (!PyIter_Check(dictitems)) { |
3991 | PyErr_Format(st->PicklingError, "fifth element of the tuple " |
3992 | "returned by __reduce__ must be an iterator, not %s" , |
3993 | Py_TYPE(dictitems)->tp_name); |
3994 | return -1; |
3995 | } |
3996 | |
3997 | if (state_setter == Py_None) |
3998 | state_setter = NULL; |
3999 | else if (!PyCallable_Check(state_setter)) { |
4000 | PyErr_Format(st->PicklingError, "sixth element of the tuple " |
4001 | "returned by __reduce__ must be a function, not %s" , |
4002 | Py_TYPE(state_setter)->tp_name); |
4003 | return -1; |
4004 | } |
4005 | |
4006 | if (self->proto >= 2) { |
4007 | PyObject *name; |
4008 | _Py_IDENTIFIER(__name__); |
4009 | |
4010 | if (_PyObject_LookupAttrId(callable, &PyId___name__, &name) < 0) { |
4011 | return -1; |
4012 | } |
4013 | if (name != NULL && PyUnicode_Check(name)) { |
4014 | _Py_IDENTIFIER(__newobj_ex__); |
4015 | use_newobj_ex = _PyUnicode_EqualToASCIIId( |
4016 | name, &PyId___newobj_ex__); |
4017 | if (!use_newobj_ex) { |
4018 | _Py_IDENTIFIER(__newobj__); |
4019 | use_newobj = _PyUnicode_EqualToASCIIId(name, &PyId___newobj__); |
4020 | } |
4021 | } |
4022 | Py_XDECREF(name); |
4023 | } |
4024 | |
4025 | if (use_newobj_ex) { |
4026 | PyObject *cls; |
4027 | PyObject *args; |
4028 | PyObject *kwargs; |
4029 | |
4030 | if (PyTuple_GET_SIZE(argtup) != 3) { |
4031 | PyErr_Format(st->PicklingError, |
4032 | "length of the NEWOBJ_EX argument tuple must be " |
4033 | "exactly 3, not %zd" , PyTuple_GET_SIZE(argtup)); |
4034 | return -1; |
4035 | } |
4036 | |
4037 | cls = PyTuple_GET_ITEM(argtup, 0); |
4038 | if (!PyType_Check(cls)) { |
4039 | PyErr_Format(st->PicklingError, |
4040 | "first item from NEWOBJ_EX argument tuple must " |
4041 | "be a class, not %.200s" , Py_TYPE(cls)->tp_name); |
4042 | return -1; |
4043 | } |
4044 | args = PyTuple_GET_ITEM(argtup, 1); |
4045 | if (!PyTuple_Check(args)) { |
4046 | PyErr_Format(st->PicklingError, |
4047 | "second item from NEWOBJ_EX argument tuple must " |
4048 | "be a tuple, not %.200s" , Py_TYPE(args)->tp_name); |
4049 | return -1; |
4050 | } |
4051 | kwargs = PyTuple_GET_ITEM(argtup, 2); |
4052 | if (!PyDict_Check(kwargs)) { |
4053 | PyErr_Format(st->PicklingError, |
4054 | "third item from NEWOBJ_EX argument tuple must " |
4055 | "be a dict, not %.200s" , Py_TYPE(kwargs)->tp_name); |
4056 | return -1; |
4057 | } |
4058 | |
4059 | if (self->proto >= 4) { |
4060 | if (save(self, cls, 0) < 0 || |
4061 | save(self, args, 0) < 0 || |
4062 | save(self, kwargs, 0) < 0 || |
4063 | _Pickler_Write(self, &newobj_ex_op, 1) < 0) { |
4064 | return -1; |
4065 | } |
4066 | } |
4067 | else { |
4068 | PyObject *newargs; |
4069 | PyObject *cls_new; |
4070 | Py_ssize_t i; |
4071 | _Py_IDENTIFIER(__new__); |
4072 | |
4073 | newargs = PyTuple_New(PyTuple_GET_SIZE(args) + 2); |
4074 | if (newargs == NULL) |
4075 | return -1; |
4076 | |
4077 | cls_new = _PyObject_GetAttrId(cls, &PyId___new__); |
4078 | if (cls_new == NULL) { |
4079 | Py_DECREF(newargs); |
4080 | return -1; |
4081 | } |
4082 | PyTuple_SET_ITEM(newargs, 0, cls_new); |
4083 | Py_INCREF(cls); |
4084 | PyTuple_SET_ITEM(newargs, 1, cls); |
4085 | for (i = 0; i < PyTuple_GET_SIZE(args); i++) { |
4086 | PyObject *item = PyTuple_GET_ITEM(args, i); |
4087 | Py_INCREF(item); |
4088 | PyTuple_SET_ITEM(newargs, i + 2, item); |
4089 | } |
4090 | |
4091 | callable = PyObject_Call(st->partial, newargs, kwargs); |
4092 | Py_DECREF(newargs); |
4093 | if (callable == NULL) |
4094 | return -1; |
4095 | |
4096 | newargs = PyTuple_New(0); |
4097 | if (newargs == NULL) { |
4098 | Py_DECREF(callable); |
4099 | return -1; |
4100 | } |
4101 | |
4102 | if (save(self, callable, 0) < 0 || |
4103 | save(self, newargs, 0) < 0 || |
4104 | _Pickler_Write(self, &reduce_op, 1) < 0) { |
4105 | Py_DECREF(newargs); |
4106 | Py_DECREF(callable); |
4107 | return -1; |
4108 | } |
4109 | Py_DECREF(newargs); |
4110 | Py_DECREF(callable); |
4111 | } |
4112 | } |
4113 | else if (use_newobj) { |
4114 | PyObject *cls; |
4115 | PyObject *newargtup; |
4116 | PyObject *obj_class; |
4117 | int p; |
4118 | |
4119 | /* Sanity checks. */ |
4120 | if (PyTuple_GET_SIZE(argtup) < 1) { |
4121 | PyErr_SetString(st->PicklingError, "__newobj__ arglist is empty" ); |
4122 | return -1; |
4123 | } |
4124 | |
4125 | cls = PyTuple_GET_ITEM(argtup, 0); |
4126 | if (!PyType_Check(cls)) { |
4127 | PyErr_SetString(st->PicklingError, "args[0] from " |
4128 | "__newobj__ args is not a type" ); |
4129 | return -1; |
4130 | } |
4131 | |
4132 | if (obj != NULL) { |
4133 | obj_class = get_class(obj); |
4134 | if (obj_class == NULL) { |
4135 | return -1; |
4136 | } |
4137 | p = obj_class != cls; |
4138 | Py_DECREF(obj_class); |
4139 | if (p) { |
4140 | PyErr_SetString(st->PicklingError, "args[0] from " |
4141 | "__newobj__ args has the wrong class" ); |
4142 | return -1; |
4143 | } |
4144 | } |
4145 | /* XXX: These calls save() are prone to infinite recursion. Imagine |
4146 | what happen if the value returned by the __reduce__() method of |
4147 | some extension type contains another object of the same type. Ouch! |
4148 | |
4149 | Here is a quick example, that I ran into, to illustrate what I |
4150 | mean: |
4151 | |
4152 | >>> import pickle, copyreg |
4153 | >>> copyreg.dispatch_table.pop(complex) |
4154 | >>> pickle.dumps(1+2j) |
4155 | Traceback (most recent call last): |
4156 | ... |
4157 | RecursionError: maximum recursion depth exceeded |
4158 | |
4159 | Removing the complex class from copyreg.dispatch_table made the |
4160 | __reduce_ex__() method emit another complex object: |
4161 | |
4162 | >>> (1+1j).__reduce_ex__(2) |
4163 | (<function __newobj__ at 0xb7b71c3c>, |
4164 | (<class 'complex'>, (1+1j)), None, None, None) |
4165 | |
4166 | Thus when save() was called on newargstup (the 2nd item) recursion |
4167 | ensued. Of course, the bug was in the complex class which had a |
4168 | broken __getnewargs__() that emitted another complex object. But, |
4169 | the point, here, is it is quite easy to end up with a broken reduce |
4170 | function. */ |
4171 | |
4172 | /* Save the class and its __new__ arguments. */ |
4173 | if (save(self, cls, 0) < 0) |
4174 | return -1; |
4175 | |
4176 | newargtup = PyTuple_GetSlice(argtup, 1, PyTuple_GET_SIZE(argtup)); |
4177 | if (newargtup == NULL) |
4178 | return -1; |
4179 | |
4180 | p = save(self, newargtup, 0); |
4181 | Py_DECREF(newargtup); |
4182 | if (p < 0) |
4183 | return -1; |
4184 | |
4185 | /* Add NEWOBJ opcode. */ |
4186 | if (_Pickler_Write(self, &newobj_op, 1) < 0) |
4187 | return -1; |
4188 | } |
4189 | else { /* Not using NEWOBJ. */ |
4190 | if (save(self, callable, 0) < 0 || |
4191 | save(self, argtup, 0) < 0 || |
4192 | _Pickler_Write(self, &reduce_op, 1) < 0) |
4193 | return -1; |
4194 | } |
4195 | |
4196 | /* obj can be NULL when save_reduce() is used directly. A NULL obj means |
4197 | the caller do not want to memoize the object. Not particularly useful, |
4198 | but that is to mimic the behavior save_reduce() in pickle.py when |
4199 | obj is None. */ |
4200 | if (obj != NULL) { |
4201 | /* If the object is already in the memo, this means it is |
4202 | recursive. In this case, throw away everything we put on the |
4203 | stack, and fetch the object back from the memo. */ |
4204 | if (PyMemoTable_Get(self->memo, obj)) { |
4205 | const char pop_op = POP; |
4206 | |
4207 | if (_Pickler_Write(self, &pop_op, 1) < 0) |
4208 | return -1; |
4209 | if (memo_get(self, obj) < 0) |
4210 | return -1; |
4211 | |
4212 | return 0; |
4213 | } |
4214 | else if (memo_put(self, obj) < 0) |
4215 | return -1; |
4216 | } |
4217 | |
4218 | if (listitems && batch_list(self, listitems) < 0) |
4219 | return -1; |
4220 | |
4221 | if (dictitems && batch_dict(self, dictitems) < 0) |
4222 | return -1; |
4223 | |
4224 | if (state) { |
4225 | if (state_setter == NULL) { |
4226 | if (save(self, state, 0) < 0 || |
4227 | _Pickler_Write(self, &build_op, 1) < 0) |
4228 | return -1; |
4229 | } |
4230 | else { |
4231 | |
4232 | /* If a state_setter is specified, call it instead of load_build to |
4233 | * update obj's with its previous state. |
4234 | * The first 4 save/write instructions push state_setter and its |
4235 | * tuple of expected arguments (obj, state) onto the stack. The |
4236 | * REDUCE opcode triggers the state_setter(obj, state) function |
4237 | * call. Finally, because state-updating routines only do in-place |
4238 | * modification, the whole operation has to be stack-transparent. |
4239 | * Thus, we finally pop the call's output from the stack.*/ |
4240 | |
4241 | const char tupletwo_op = TUPLE2; |
4242 | const char pop_op = POP; |
4243 | if (save(self, state_setter, 0) < 0 || |
4244 | save(self, obj, 0) < 0 || save(self, state, 0) < 0 || |
4245 | _Pickler_Write(self, &tupletwo_op, 1) < 0 || |
4246 | _Pickler_Write(self, &reduce_op, 1) < 0 || |
4247 | _Pickler_Write(self, &pop_op, 1) < 0) |
4248 | return -1; |
4249 | } |
4250 | } |
4251 | return 0; |
4252 | } |
4253 | |
4254 | static int |
4255 | save(PicklerObject *self, PyObject *obj, int pers_save) |
4256 | { |
4257 | PyTypeObject *type; |
4258 | PyObject *reduce_func = NULL; |
4259 | PyObject *reduce_value = NULL; |
4260 | int status = 0; |
4261 | |
4262 | if (_Pickler_OpcodeBoundary(self) < 0) |
4263 | return -1; |
4264 | |
4265 | /* The extra pers_save argument is necessary to avoid calling save_pers() |
4266 | on its returned object. */ |
4267 | if (!pers_save && self->pers_func) { |
4268 | /* save_pers() returns: |
4269 | -1 to signal an error; |
4270 | 0 if it did nothing successfully; |
4271 | 1 if a persistent id was saved. |
4272 | */ |
4273 | if ((status = save_pers(self, obj)) != 0) |
4274 | return status; |
4275 | } |
4276 | |
4277 | type = Py_TYPE(obj); |
4278 | |
4279 | /* The old cPickle had an optimization that used switch-case statement |
4280 | dispatching on the first letter of the type name. This has was removed |
4281 | since benchmarks shown that this optimization was actually slowing |
4282 | things down. */ |
4283 | |
4284 | /* Atom types; these aren't memoized, so don't check the memo. */ |
4285 | |
4286 | if (obj == Py_None) { |
4287 | return save_none(self, obj); |
4288 | } |
4289 | else if (obj == Py_False || obj == Py_True) { |
4290 | return save_bool(self, obj); |
4291 | } |
4292 | else if (type == &PyLong_Type) { |
4293 | return save_long(self, obj); |
4294 | } |
4295 | else if (type == &PyFloat_Type) { |
4296 | return save_float(self, obj); |
4297 | } |
4298 | |
4299 | /* Check the memo to see if it has the object. If so, generate |
4300 | a GET (or BINGET) opcode, instead of pickling the object |
4301 | once again. */ |
4302 | if (PyMemoTable_Get(self->memo, obj)) { |
4303 | return memo_get(self, obj); |
4304 | } |
4305 | |
4306 | if (type == &PyBytes_Type) { |
4307 | return save_bytes(self, obj); |
4308 | } |
4309 | else if (type == &PyUnicode_Type) { |
4310 | return save_unicode(self, obj); |
4311 | } |
4312 | |
4313 | /* We're only calling Py_EnterRecursiveCall here so that atomic |
4314 | types above are pickled faster. */ |
4315 | if (Py_EnterRecursiveCall(" while pickling an object" )) { |
4316 | return -1; |
4317 | } |
4318 | |
4319 | if (type == &PyDict_Type) { |
4320 | status = save_dict(self, obj); |
4321 | goto done; |
4322 | } |
4323 | else if (type == &PySet_Type) { |
4324 | status = save_set(self, obj); |
4325 | goto done; |
4326 | } |
4327 | else if (type == &PyFrozenSet_Type) { |
4328 | status = save_frozenset(self, obj); |
4329 | goto done; |
4330 | } |
4331 | else if (type == &PyList_Type) { |
4332 | status = save_list(self, obj); |
4333 | goto done; |
4334 | } |
4335 | else if (type == &PyTuple_Type) { |
4336 | status = save_tuple(self, obj); |
4337 | goto done; |
4338 | } |
4339 | else if (type == &PyByteArray_Type) { |
4340 | status = save_bytearray(self, obj); |
4341 | goto done; |
4342 | } |
4343 | else if (type == &PyPickleBuffer_Type) { |
4344 | status = save_picklebuffer(self, obj); |
4345 | goto done; |
4346 | } |
4347 | |
4348 | /* Now, check reducer_override. If it returns NotImplemented, |
4349 | * fallback to save_type or save_global, and then perhaps to the |
4350 | * regular reduction mechanism. |
4351 | */ |
4352 | if (self->reducer_override != NULL) { |
4353 | reduce_value = PyObject_CallOneArg(self->reducer_override, obj); |
4354 | if (reduce_value == NULL) { |
4355 | goto error; |
4356 | } |
4357 | if (reduce_value != Py_NotImplemented) { |
4358 | goto reduce; |
4359 | } |
4360 | Py_DECREF(reduce_value); |
4361 | reduce_value = NULL; |
4362 | } |
4363 | |
4364 | if (type == &PyType_Type) { |
4365 | status = save_type(self, obj); |
4366 | goto done; |
4367 | } |
4368 | else if (type == &PyFunction_Type) { |
4369 | status = save_global(self, obj, NULL); |
4370 | goto done; |
4371 | } |
4372 | |
4373 | /* XXX: This part needs some unit tests. */ |
4374 | |
4375 | /* Get a reduction callable, and call it. This may come from |
4376 | * self.dispatch_table, copyreg.dispatch_table, the object's |
4377 | * __reduce_ex__ method, or the object's __reduce__ method. |
4378 | */ |
4379 | if (self->dispatch_table == NULL) { |
4380 | PickleState *st = _Pickle_GetGlobalState(); |
4381 | reduce_func = PyDict_GetItemWithError(st->dispatch_table, |
4382 | (PyObject *)type); |
4383 | if (reduce_func == NULL) { |
4384 | if (PyErr_Occurred()) { |
4385 | goto error; |
4386 | } |
4387 | } else { |
4388 | /* PyDict_GetItemWithError() returns a borrowed reference. |
4389 | Increase the reference count to be consistent with |
4390 | PyObject_GetItem and _PyObject_GetAttrId used below. */ |
4391 | Py_INCREF(reduce_func); |
4392 | } |
4393 | } else { |
4394 | reduce_func = PyObject_GetItem(self->dispatch_table, |
4395 | (PyObject *)type); |
4396 | if (reduce_func == NULL) { |
4397 | if (PyErr_ExceptionMatches(PyExc_KeyError)) |
4398 | PyErr_Clear(); |
4399 | else |
4400 | goto error; |
4401 | } |
4402 | } |
4403 | if (reduce_func != NULL) { |
4404 | Py_INCREF(obj); |
4405 | reduce_value = _Pickle_FastCall(reduce_func, obj); |
4406 | } |
4407 | else if (PyType_IsSubtype(type, &PyType_Type)) { |
4408 | status = save_global(self, obj, NULL); |
4409 | goto done; |
4410 | } |
4411 | else { |
4412 | _Py_IDENTIFIER(__reduce__); |
4413 | _Py_IDENTIFIER(__reduce_ex__); |
4414 | |
4415 | /* XXX: If the __reduce__ method is defined, __reduce_ex__ is |
4416 | automatically defined as __reduce__. While this is convenient, this |
4417 | make it impossible to know which method was actually called. Of |
4418 | course, this is not a big deal. But still, it would be nice to let |
4419 | the user know which method was called when something go |
4420 | wrong. Incidentally, this means if __reduce_ex__ is not defined, we |
4421 | don't actually have to check for a __reduce__ method. */ |
4422 | |
4423 | /* Check for a __reduce_ex__ method. */ |
4424 | if (_PyObject_LookupAttrId(obj, &PyId___reduce_ex__, &reduce_func) < 0) { |
4425 | goto error; |
4426 | } |
4427 | if (reduce_func != NULL) { |
4428 | PyObject *proto; |
4429 | proto = PyLong_FromLong(self->proto); |
4430 | if (proto != NULL) { |
4431 | reduce_value = _Pickle_FastCall(reduce_func, proto); |
4432 | } |
4433 | } |
4434 | else { |
4435 | /* Check for a __reduce__ method. */ |
4436 | if (_PyObject_LookupAttrId(obj, &PyId___reduce__, &reduce_func) < 0) { |
4437 | goto error; |
4438 | } |
4439 | if (reduce_func != NULL) { |
4440 | reduce_value = PyObject_CallNoArgs(reduce_func); |
4441 | } |
4442 | else { |
4443 | PickleState *st = _Pickle_GetGlobalState(); |
4444 | PyErr_Format(st->PicklingError, |
4445 | "can't pickle '%.200s' object: %R" , |
4446 | type->tp_name, obj); |
4447 | goto error; |
4448 | } |
4449 | } |
4450 | } |
4451 | |
4452 | if (reduce_value == NULL) |
4453 | goto error; |
4454 | |
4455 | reduce: |
4456 | if (PyUnicode_Check(reduce_value)) { |
4457 | status = save_global(self, obj, reduce_value); |
4458 | goto done; |
4459 | } |
4460 | |
4461 | if (!PyTuple_Check(reduce_value)) { |
4462 | PickleState *st = _Pickle_GetGlobalState(); |
4463 | PyErr_SetString(st->PicklingError, |
4464 | "__reduce__ must return a string or tuple" ); |
4465 | goto error; |
4466 | } |
4467 | |
4468 | status = save_reduce(self, reduce_value, obj); |
4469 | |
4470 | if (0) { |
4471 | error: |
4472 | status = -1; |
4473 | } |
4474 | done: |
4475 | |
4476 | Py_LeaveRecursiveCall(); |
4477 | Py_XDECREF(reduce_func); |
4478 | Py_XDECREF(reduce_value); |
4479 | |
4480 | return status; |
4481 | } |
4482 | |
4483 | static int |
4484 | dump(PicklerObject *self, PyObject *obj) |
4485 | { |
4486 | const char stop_op = STOP; |
4487 | int status = -1; |
4488 | PyObject *tmp; |
4489 | _Py_IDENTIFIER(reducer_override); |
4490 | |
4491 | if (_PyObject_LookupAttrId((PyObject *)self, &PyId_reducer_override, |
4492 | &tmp) < 0) { |
4493 | goto error; |
4494 | } |
4495 | /* Cache the reducer_override method, if it exists. */ |
4496 | if (tmp != NULL) { |
4497 | Py_XSETREF(self->reducer_override, tmp); |
4498 | } |
4499 | else { |
4500 | Py_CLEAR(self->reducer_override); |
4501 | } |
4502 | |
4503 | if (self->proto >= 2) { |
4504 | char [2]; |
4505 | |
4506 | header[0] = PROTO; |
4507 | assert(self->proto >= 0 && self->proto < 256); |
4508 | header[1] = (unsigned char)self->proto; |
4509 | if (_Pickler_Write(self, header, 2) < 0) |
4510 | goto error; |
4511 | if (self->proto >= 4) |
4512 | self->framing = 1; |
4513 | } |
4514 | |
4515 | if (save(self, obj, 0) < 0 || |
4516 | _Pickler_Write(self, &stop_op, 1) < 0 || |
4517 | _Pickler_CommitFrame(self) < 0) |
4518 | goto error; |
4519 | |
4520 | // Success |
4521 | status = 0; |
4522 | |
4523 | error: |
4524 | self->framing = 0; |
4525 | |
4526 | /* Break the reference cycle we generated at the beginning this function |
4527 | * call when setting the reducer_override attribute of the Pickler instance |
4528 | * to a bound method of the same instance. This is important as the Pickler |
4529 | * instance holds a reference to each object it has pickled (through its |
4530 | * memo): thus, these objects won't be garbage-collected as long as the |
4531 | * Pickler itself is not collected. */ |
4532 | Py_CLEAR(self->reducer_override); |
4533 | return status; |
4534 | } |
4535 | |
4536 | /*[clinic input] |
4537 | |
4538 | _pickle.Pickler.clear_memo |
4539 | |
4540 | Clears the pickler's "memo". |
4541 | |
4542 | The memo is the data structure that remembers which objects the |
4543 | pickler has already seen, so that shared or recursive objects are |
4544 | pickled by reference and not by value. This method is useful when |
4545 | re-using picklers. |
4546 | [clinic start generated code]*/ |
4547 | |
4548 | static PyObject * |
4549 | _pickle_Pickler_clear_memo_impl(PicklerObject *self) |
4550 | /*[clinic end generated code: output=8665c8658aaa094b input=01bdad52f3d93e56]*/ |
4551 | { |
4552 | if (self->memo) |
4553 | PyMemoTable_Clear(self->memo); |
4554 | |
4555 | Py_RETURN_NONE; |
4556 | } |
4557 | |
4558 | /*[clinic input] |
4559 | |
4560 | _pickle.Pickler.dump |
4561 | |
4562 | obj: object |
4563 | / |
4564 | |
4565 | Write a pickled representation of the given object to the open file. |
4566 | [clinic start generated code]*/ |
4567 | |
4568 | static PyObject * |
4569 | _pickle_Pickler_dump(PicklerObject *self, PyObject *obj) |
4570 | /*[clinic end generated code: output=87ecad1261e02ac7 input=552eb1c0f52260d9]*/ |
4571 | { |
4572 | /* Check whether the Pickler was initialized correctly (issue3664). |
4573 | Developers often forget to call __init__() in their subclasses, which |
4574 | would trigger a segfault without this check. */ |
4575 | if (self->write == NULL) { |
4576 | PickleState *st = _Pickle_GetGlobalState(); |
4577 | PyErr_Format(st->PicklingError, |
4578 | "Pickler.__init__() was not called by %s.__init__()" , |
4579 | Py_TYPE(self)->tp_name); |
4580 | return NULL; |
4581 | } |
4582 | |
4583 | if (_Pickler_ClearBuffer(self) < 0) |
4584 | return NULL; |
4585 | |
4586 | if (dump(self, obj) < 0) |
4587 | return NULL; |
4588 | |
4589 | if (_Pickler_FlushToFile(self) < 0) |
4590 | return NULL; |
4591 | |
4592 | Py_RETURN_NONE; |
4593 | } |
4594 | |
4595 | /*[clinic input] |
4596 | |
4597 | _pickle.Pickler.__sizeof__ -> Py_ssize_t |
4598 | |
4599 | Returns size in memory, in bytes. |
4600 | [clinic start generated code]*/ |
4601 | |
4602 | static Py_ssize_t |
4603 | _pickle_Pickler___sizeof___impl(PicklerObject *self) |
4604 | /*[clinic end generated code: output=106edb3123f332e1 input=8cbbec9bd5540d42]*/ |
4605 | { |
4606 | Py_ssize_t res, s; |
4607 | |
4608 | res = _PyObject_SIZE(Py_TYPE(self)); |
4609 | if (self->memo != NULL) { |
4610 | res += sizeof(PyMemoTable); |
4611 | res += self->memo->mt_allocated * sizeof(PyMemoEntry); |
4612 | } |
4613 | if (self->output_buffer != NULL) { |
4614 | s = _PySys_GetSizeOf(self->output_buffer); |
4615 | if (s == -1) |
4616 | return -1; |
4617 | res += s; |
4618 | } |
4619 | return res; |
4620 | } |
4621 | |
4622 | static struct PyMethodDef Pickler_methods[] = { |
4623 | _PICKLE_PICKLER_DUMP_METHODDEF |
4624 | _PICKLE_PICKLER_CLEAR_MEMO_METHODDEF |
4625 | _PICKLE_PICKLER___SIZEOF___METHODDEF |
4626 | {NULL, NULL} /* sentinel */ |
4627 | }; |
4628 | |
4629 | static void |
4630 | Pickler_dealloc(PicklerObject *self) |
4631 | { |
4632 | PyObject_GC_UnTrack(self); |
4633 | |
4634 | Py_XDECREF(self->output_buffer); |
4635 | Py_XDECREF(self->write); |
4636 | Py_XDECREF(self->pers_func); |
4637 | Py_XDECREF(self->dispatch_table); |
4638 | Py_XDECREF(self->fast_memo); |
4639 | Py_XDECREF(self->reducer_override); |
4640 | Py_XDECREF(self->buffer_callback); |
4641 | |
4642 | PyMemoTable_Del(self->memo); |
4643 | |
4644 | Py_TYPE(self)->tp_free((PyObject *)self); |
4645 | } |
4646 | |
4647 | static int |
4648 | Pickler_traverse(PicklerObject *self, visitproc visit, void *arg) |
4649 | { |
4650 | Py_VISIT(self->write); |
4651 | Py_VISIT(self->pers_func); |
4652 | Py_VISIT(self->dispatch_table); |
4653 | Py_VISIT(self->fast_memo); |
4654 | Py_VISIT(self->reducer_override); |
4655 | Py_VISIT(self->buffer_callback); |
4656 | return 0; |
4657 | } |
4658 | |
4659 | static int |
4660 | Pickler_clear(PicklerObject *self) |
4661 | { |
4662 | Py_CLEAR(self->output_buffer); |
4663 | Py_CLEAR(self->write); |
4664 | Py_CLEAR(self->pers_func); |
4665 | Py_CLEAR(self->dispatch_table); |
4666 | Py_CLEAR(self->fast_memo); |
4667 | Py_CLEAR(self->reducer_override); |
4668 | Py_CLEAR(self->buffer_callback); |
4669 | |
4670 | if (self->memo != NULL) { |
4671 | PyMemoTable *memo = self->memo; |
4672 | self->memo = NULL; |
4673 | PyMemoTable_Del(memo); |
4674 | } |
4675 | return 0; |
4676 | } |
4677 | |
4678 | |
4679 | /*[clinic input] |
4680 | |
4681 | _pickle.Pickler.__init__ |
4682 | |
4683 | file: object |
4684 | protocol: object = None |
4685 | fix_imports: bool = True |
4686 | buffer_callback: object = None |
4687 | |
4688 | This takes a binary file for writing a pickle data stream. |
4689 | |
4690 | The optional *protocol* argument tells the pickler to use the given |
4691 | protocol; supported protocols are 0, 1, 2, 3, 4 and 5. The default |
4692 | protocol is 4. It was introduced in Python 3.4, and is incompatible |
4693 | with previous versions. |
4694 | |
4695 | Specifying a negative protocol version selects the highest protocol |
4696 | version supported. The higher the protocol used, the more recent the |
4697 | version of Python needed to read the pickle produced. |
4698 | |
4699 | The *file* argument must have a write() method that accepts a single |
4700 | bytes argument. It can thus be a file object opened for binary |
4701 | writing, an io.BytesIO instance, or any other custom object that meets |
4702 | this interface. |
4703 | |
4704 | If *fix_imports* is True and protocol is less than 3, pickle will try |
4705 | to map the new Python 3 names to the old module names used in Python |
4706 | 2, so that the pickle data stream is readable with Python 2. |
4707 | |
4708 | If *buffer_callback* is None (the default), buffer views are |
4709 | serialized into *file* as part of the pickle stream. |
4710 | |
4711 | If *buffer_callback* is not None, then it can be called any number |
4712 | of times with a buffer view. If the callback returns a false value |
4713 | (such as None), the given buffer is out-of-band; otherwise the |
4714 | buffer is serialized in-band, i.e. inside the pickle stream. |
4715 | |
4716 | It is an error if *buffer_callback* is not None and *protocol* |
4717 | is None or smaller than 5. |
4718 | |
4719 | [clinic start generated code]*/ |
4720 | |
4721 | static int |
4722 | _pickle_Pickler___init___impl(PicklerObject *self, PyObject *file, |
4723 | PyObject *protocol, int fix_imports, |
4724 | PyObject *buffer_callback) |
4725 | /*[clinic end generated code: output=0abedc50590d259b input=a7c969699bf5dad3]*/ |
4726 | { |
4727 | _Py_IDENTIFIER(persistent_id); |
4728 | _Py_IDENTIFIER(dispatch_table); |
4729 | |
4730 | /* In case of multiple __init__() calls, clear previous content. */ |
4731 | if (self->write != NULL) |
4732 | (void)Pickler_clear(self); |
4733 | |
4734 | if (_Pickler_SetProtocol(self, protocol, fix_imports) < 0) |
4735 | return -1; |
4736 | |
4737 | if (_Pickler_SetOutputStream(self, file) < 0) |
4738 | return -1; |
4739 | |
4740 | if (_Pickler_SetBufferCallback(self, buffer_callback) < 0) |
4741 | return -1; |
4742 | |
4743 | /* memo and output_buffer may have already been created in _Pickler_New */ |
4744 | if (self->memo == NULL) { |
4745 | self->memo = PyMemoTable_New(); |
4746 | if (self->memo == NULL) |
4747 | return -1; |
4748 | } |
4749 | self->output_len = 0; |
4750 | if (self->output_buffer == NULL) { |
4751 | self->max_output_len = WRITE_BUF_SIZE; |
4752 | self->output_buffer = PyBytes_FromStringAndSize(NULL, |
4753 | self->max_output_len); |
4754 | if (self->output_buffer == NULL) |
4755 | return -1; |
4756 | } |
4757 | |
4758 | self->fast = 0; |
4759 | self->fast_nesting = 0; |
4760 | self->fast_memo = NULL; |
4761 | |
4762 | if (init_method_ref((PyObject *)self, &PyId_persistent_id, |
4763 | &self->pers_func, &self->pers_func_self) < 0) |
4764 | { |
4765 | return -1; |
4766 | } |
4767 | |
4768 | if (_PyObject_LookupAttrId((PyObject *)self, |
4769 | &PyId_dispatch_table, &self->dispatch_table) < 0) { |
4770 | return -1; |
4771 | } |
4772 | |
4773 | return 0; |
4774 | } |
4775 | |
4776 | |
4777 | /* Define a proxy object for the Pickler's internal memo object. This is to |
4778 | * avoid breaking code like: |
4779 | * pickler.memo.clear() |
4780 | * and |
4781 | * pickler.memo = saved_memo |
4782 | * Is this a good idea? Not really, but we don't want to break code that uses |
4783 | * it. Note that we don't implement the entire mapping API here. This is |
4784 | * intentional, as these should be treated as black-box implementation details. |
4785 | */ |
4786 | |
4787 | /*[clinic input] |
4788 | _pickle.PicklerMemoProxy.clear |
4789 | |
4790 | Remove all items from memo. |
4791 | [clinic start generated code]*/ |
4792 | |
4793 | static PyObject * |
4794 | _pickle_PicklerMemoProxy_clear_impl(PicklerMemoProxyObject *self) |
4795 | /*[clinic end generated code: output=5fb9370d48ae8b05 input=ccc186dacd0f1405]*/ |
4796 | { |
4797 | if (self->pickler->memo) |
4798 | PyMemoTable_Clear(self->pickler->memo); |
4799 | Py_RETURN_NONE; |
4800 | } |
4801 | |
4802 | /*[clinic input] |
4803 | _pickle.PicklerMemoProxy.copy |
4804 | |
4805 | Copy the memo to a new object. |
4806 | [clinic start generated code]*/ |
4807 | |
4808 | static PyObject * |
4809 | _pickle_PicklerMemoProxy_copy_impl(PicklerMemoProxyObject *self) |
4810 | /*[clinic end generated code: output=bb83a919d29225ef input=b73043485ac30b36]*/ |
4811 | { |
4812 | PyMemoTable *memo; |
4813 | PyObject *new_memo = PyDict_New(); |
4814 | if (new_memo == NULL) |
4815 | return NULL; |
4816 | |
4817 | memo = self->pickler->memo; |
4818 | for (size_t i = 0; i < memo->mt_allocated; ++i) { |
4819 | PyMemoEntry entry = memo->mt_table[i]; |
4820 | if (entry.me_key != NULL) { |
4821 | int status; |
4822 | PyObject *key, *value; |
4823 | |
4824 | key = PyLong_FromVoidPtr(entry.me_key); |
4825 | value = Py_BuildValue("nO" , entry.me_value, entry.me_key); |
4826 | |
4827 | if (key == NULL || value == NULL) { |
4828 | Py_XDECREF(key); |
4829 | Py_XDECREF(value); |
4830 | goto error; |
4831 | } |
4832 | status = PyDict_SetItem(new_memo, key, value); |
4833 | Py_DECREF(key); |
4834 | Py_DECREF(value); |
4835 | if (status < 0) |
4836 | goto error; |
4837 | } |
4838 | } |
4839 | return new_memo; |
4840 | |
4841 | error: |
4842 | Py_XDECREF(new_memo); |
4843 | return NULL; |
4844 | } |
4845 | |
4846 | /*[clinic input] |
4847 | _pickle.PicklerMemoProxy.__reduce__ |
4848 | |
4849 | Implement pickle support. |
4850 | [clinic start generated code]*/ |
4851 | |
4852 | static PyObject * |
4853 | _pickle_PicklerMemoProxy___reduce___impl(PicklerMemoProxyObject *self) |
4854 | /*[clinic end generated code: output=bebba1168863ab1d input=2f7c540e24b7aae4]*/ |
4855 | { |
4856 | PyObject *reduce_value, *dict_args; |
4857 | PyObject *contents = _pickle_PicklerMemoProxy_copy_impl(self); |
4858 | if (contents == NULL) |
4859 | return NULL; |
4860 | |
4861 | reduce_value = PyTuple_New(2); |
4862 | if (reduce_value == NULL) { |
4863 | Py_DECREF(contents); |
4864 | return NULL; |
4865 | } |
4866 | dict_args = PyTuple_New(1); |
4867 | if (dict_args == NULL) { |
4868 | Py_DECREF(contents); |
4869 | Py_DECREF(reduce_value); |
4870 | return NULL; |
4871 | } |
4872 | PyTuple_SET_ITEM(dict_args, 0, contents); |
4873 | Py_INCREF((PyObject *)&PyDict_Type); |
4874 | PyTuple_SET_ITEM(reduce_value, 0, (PyObject *)&PyDict_Type); |
4875 | PyTuple_SET_ITEM(reduce_value, 1, dict_args); |
4876 | return reduce_value; |
4877 | } |
4878 | |
4879 | static PyMethodDef picklerproxy_methods[] = { |
4880 | _PICKLE_PICKLERMEMOPROXY_CLEAR_METHODDEF |
4881 | _PICKLE_PICKLERMEMOPROXY_COPY_METHODDEF |
4882 | _PICKLE_PICKLERMEMOPROXY___REDUCE___METHODDEF |
4883 | {NULL, NULL} /* sentinel */ |
4884 | }; |
4885 | |
4886 | static void |
4887 | PicklerMemoProxy_dealloc(PicklerMemoProxyObject *self) |
4888 | { |
4889 | PyObject_GC_UnTrack(self); |
4890 | Py_XDECREF(self->pickler); |
4891 | PyObject_GC_Del((PyObject *)self); |
4892 | } |
4893 | |
4894 | static int |
4895 | PicklerMemoProxy_traverse(PicklerMemoProxyObject *self, |
4896 | visitproc visit, void *arg) |
4897 | { |
4898 | Py_VISIT(self->pickler); |
4899 | return 0; |
4900 | } |
4901 | |
4902 | static int |
4903 | PicklerMemoProxy_clear(PicklerMemoProxyObject *self) |
4904 | { |
4905 | Py_CLEAR(self->pickler); |
4906 | return 0; |
4907 | } |
4908 | |
4909 | static PyTypeObject PicklerMemoProxyType = { |
4910 | PyVarObject_HEAD_INIT(NULL, 0) |
4911 | "_pickle.PicklerMemoProxy" , /*tp_name*/ |
4912 | sizeof(PicklerMemoProxyObject), /*tp_basicsize*/ |
4913 | 0, |
4914 | (destructor)PicklerMemoProxy_dealloc, /* tp_dealloc */ |
4915 | 0, /* tp_vectorcall_offset */ |
4916 | 0, /* tp_getattr */ |
4917 | 0, /* tp_setattr */ |
4918 | 0, /* tp_as_async */ |
4919 | 0, /* tp_repr */ |
4920 | 0, /* tp_as_number */ |
4921 | 0, /* tp_as_sequence */ |
4922 | 0, /* tp_as_mapping */ |
4923 | PyObject_HashNotImplemented, /* tp_hash */ |
4924 | 0, /* tp_call */ |
4925 | 0, /* tp_str */ |
4926 | PyObject_GenericGetAttr, /* tp_getattro */ |
4927 | PyObject_GenericSetAttr, /* tp_setattro */ |
4928 | 0, /* tp_as_buffer */ |
4929 | Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE | Py_TPFLAGS_HAVE_GC, |
4930 | 0, /* tp_doc */ |
4931 | (traverseproc)PicklerMemoProxy_traverse, /* tp_traverse */ |
4932 | (inquiry)PicklerMemoProxy_clear, /* tp_clear */ |
4933 | 0, /* tp_richcompare */ |
4934 | 0, /* tp_weaklistoffset */ |
4935 | 0, /* tp_iter */ |
4936 | 0, /* tp_iternext */ |
4937 | picklerproxy_methods, /* tp_methods */ |
4938 | }; |
4939 | |
4940 | static PyObject * |
4941 | PicklerMemoProxy_New(PicklerObject *pickler) |
4942 | { |
4943 | PicklerMemoProxyObject *self; |
4944 | |
4945 | self = PyObject_GC_New(PicklerMemoProxyObject, &PicklerMemoProxyType); |
4946 | if (self == NULL) |
4947 | return NULL; |
4948 | Py_INCREF(pickler); |
4949 | self->pickler = pickler; |
4950 | PyObject_GC_Track(self); |
4951 | return (PyObject *)self; |
4952 | } |
4953 | |
4954 | /*****************************************************************************/ |
4955 | |
4956 | static PyObject * |
4957 | Pickler_get_memo(PicklerObject *self, void *Py_UNUSED(ignored)) |
4958 | { |
4959 | return PicklerMemoProxy_New(self); |
4960 | } |
4961 | |
4962 | static int |
4963 | Pickler_set_memo(PicklerObject *self, PyObject *obj, void *Py_UNUSED(ignored)) |
4964 | { |
4965 | PyMemoTable *new_memo = NULL; |
4966 | |
4967 | if (obj == NULL) { |
4968 | PyErr_SetString(PyExc_TypeError, |
4969 | "attribute deletion is not supported" ); |
4970 | return -1; |
4971 | } |
4972 | |
4973 | if (Py_IS_TYPE(obj, &PicklerMemoProxyType)) { |
4974 | PicklerObject *pickler = |
4975 | ((PicklerMemoProxyObject *)obj)->pickler; |
4976 | |
4977 | new_memo = PyMemoTable_Copy(pickler->memo); |
4978 | if (new_memo == NULL) |
4979 | return -1; |
4980 | } |
4981 | else if (PyDict_Check(obj)) { |
4982 | Py_ssize_t i = 0; |
4983 | PyObject *key, *value; |
4984 | |
4985 | new_memo = PyMemoTable_New(); |
4986 | if (new_memo == NULL) |
4987 | return -1; |
4988 | |
4989 | while (PyDict_Next(obj, &i, &key, &value)) { |
4990 | Py_ssize_t memo_id; |
4991 | PyObject *memo_obj; |
4992 | |
4993 | if (!PyTuple_Check(value) || PyTuple_GET_SIZE(value) != 2) { |
4994 | PyErr_SetString(PyExc_TypeError, |
4995 | "'memo' values must be 2-item tuples" ); |
4996 | goto error; |
4997 | } |
4998 | memo_id = PyLong_AsSsize_t(PyTuple_GET_ITEM(value, 0)); |
4999 | if (memo_id == -1 && PyErr_Occurred()) |
5000 | goto error; |
5001 | memo_obj = PyTuple_GET_ITEM(value, 1); |
5002 | if (PyMemoTable_Set(new_memo, memo_obj, memo_id) < 0) |
5003 | goto error; |
5004 | } |
5005 | } |
5006 | else { |
5007 | PyErr_Format(PyExc_TypeError, |
5008 | "'memo' attribute must be a PicklerMemoProxy object " |
5009 | "or dict, not %.200s" , Py_TYPE(obj)->tp_name); |
5010 | return -1; |
5011 | } |
5012 | |
5013 | PyMemoTable_Del(self->memo); |
5014 | self->memo = new_memo; |
5015 | |
5016 | return 0; |
5017 | |
5018 | error: |
5019 | if (new_memo) |
5020 | PyMemoTable_Del(new_memo); |
5021 | return -1; |
5022 | } |
5023 | |
5024 | static PyObject * |
5025 | Pickler_get_persid(PicklerObject *self, void *Py_UNUSED(ignored)) |
5026 | { |
5027 | if (self->pers_func == NULL) { |
5028 | PyErr_SetString(PyExc_AttributeError, "persistent_id" ); |
5029 | return NULL; |
5030 | } |
5031 | return reconstruct_method(self->pers_func, self->pers_func_self); |
5032 | } |
5033 | |
5034 | static int |
5035 | Pickler_set_persid(PicklerObject *self, PyObject *value, void *Py_UNUSED(ignored)) |
5036 | { |
5037 | if (value == NULL) { |
5038 | PyErr_SetString(PyExc_TypeError, |
5039 | "attribute deletion is not supported" ); |
5040 | return -1; |
5041 | } |
5042 | if (!PyCallable_Check(value)) { |
5043 | PyErr_SetString(PyExc_TypeError, |
5044 | "persistent_id must be a callable taking one argument" ); |
5045 | return -1; |
5046 | } |
5047 | |
5048 | self->pers_func_self = NULL; |
5049 | Py_INCREF(value); |
5050 | Py_XSETREF(self->pers_func, value); |
5051 | |
5052 | return 0; |
5053 | } |
5054 | |
5055 | static PyMemberDef Pickler_members[] = { |
5056 | {"bin" , T_INT, offsetof(PicklerObject, bin)}, |
5057 | {"fast" , T_INT, offsetof(PicklerObject, fast)}, |
5058 | {"dispatch_table" , T_OBJECT_EX, offsetof(PicklerObject, dispatch_table)}, |
5059 | {NULL} |
5060 | }; |
5061 | |
5062 | static PyGetSetDef Pickler_getsets[] = { |
5063 | {"memo" , (getter)Pickler_get_memo, |
5064 | (setter)Pickler_set_memo}, |
5065 | {"persistent_id" , (getter)Pickler_get_persid, |
5066 | (setter)Pickler_set_persid}, |
5067 | {NULL} |
5068 | }; |
5069 | |
5070 | static PyTypeObject Pickler_Type = { |
5071 | PyVarObject_HEAD_INIT(NULL, 0) |
5072 | "_pickle.Pickler" , /*tp_name*/ |
5073 | sizeof(PicklerObject), /*tp_basicsize*/ |
5074 | 0, /*tp_itemsize*/ |
5075 | (destructor)Pickler_dealloc, /*tp_dealloc*/ |
5076 | 0, /*tp_vectorcall_offset*/ |
5077 | 0, /*tp_getattr*/ |
5078 | 0, /*tp_setattr*/ |
5079 | 0, /*tp_as_async*/ |
5080 | 0, /*tp_repr*/ |
5081 | 0, /*tp_as_number*/ |
5082 | 0, /*tp_as_sequence*/ |
5083 | 0, /*tp_as_mapping*/ |
5084 | 0, /*tp_hash*/ |
5085 | 0, /*tp_call*/ |
5086 | 0, /*tp_str*/ |
5087 | 0, /*tp_getattro*/ |
5088 | 0, /*tp_setattro*/ |
5089 | 0, /*tp_as_buffer*/ |
5090 | Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE | Py_TPFLAGS_HAVE_GC, |
5091 | _pickle_Pickler___init____doc__, /*tp_doc*/ |
5092 | (traverseproc)Pickler_traverse, /*tp_traverse*/ |
5093 | (inquiry)Pickler_clear, /*tp_clear*/ |
5094 | 0, /*tp_richcompare*/ |
5095 | 0, /*tp_weaklistoffset*/ |
5096 | 0, /*tp_iter*/ |
5097 | 0, /*tp_iternext*/ |
5098 | Pickler_methods, /*tp_methods*/ |
5099 | Pickler_members, /*tp_members*/ |
5100 | Pickler_getsets, /*tp_getset*/ |
5101 | 0, /*tp_base*/ |
5102 | 0, /*tp_dict*/ |
5103 | 0, /*tp_descr_get*/ |
5104 | 0, /*tp_descr_set*/ |
5105 | 0, /*tp_dictoffset*/ |
5106 | _pickle_Pickler___init__, /*tp_init*/ |
5107 | PyType_GenericAlloc, /*tp_alloc*/ |
5108 | PyType_GenericNew, /*tp_new*/ |
5109 | PyObject_GC_Del, /*tp_free*/ |
5110 | 0, /*tp_is_gc*/ |
5111 | }; |
5112 | |
5113 | /* Temporary helper for calling self.find_class(). |
5114 | |
5115 | XXX: It would be nice to able to avoid Python function call overhead, by |
5116 | using directly the C version of find_class(), when find_class() is not |
5117 | overridden by a subclass. Although, this could become rather hackish. A |
5118 | simpler optimization would be to call the C function when self is not a |
5119 | subclass instance. */ |
5120 | static PyObject * |
5121 | find_class(UnpicklerObject *self, PyObject *module_name, PyObject *global_name) |
5122 | { |
5123 | _Py_IDENTIFIER(find_class); |
5124 | |
5125 | return _PyObject_CallMethodIdObjArgs((PyObject *)self, &PyId_find_class, |
5126 | module_name, global_name, NULL); |
5127 | } |
5128 | |
5129 | static Py_ssize_t |
5130 | marker(UnpicklerObject *self) |
5131 | { |
5132 | Py_ssize_t mark; |
5133 | |
5134 | if (self->num_marks < 1) { |
5135 | PickleState *st = _Pickle_GetGlobalState(); |
5136 | PyErr_SetString(st->UnpicklingError, "could not find MARK" ); |
5137 | return -1; |
5138 | } |
5139 | |
5140 | mark = self->marks[--self->num_marks]; |
5141 | self->stack->mark_set = self->num_marks != 0; |
5142 | self->stack->fence = self->num_marks ? |
5143 | self->marks[self->num_marks - 1] : 0; |
5144 | return mark; |
5145 | } |
5146 | |
5147 | static int |
5148 | load_none(UnpicklerObject *self) |
5149 | { |
5150 | PDATA_APPEND(self->stack, Py_None, -1); |
5151 | return 0; |
5152 | } |
5153 | |
5154 | static int |
5155 | load_int(UnpicklerObject *self) |
5156 | { |
5157 | PyObject *value; |
5158 | char *endptr, *s; |
5159 | Py_ssize_t len; |
5160 | long x; |
5161 | |
5162 | if ((len = _Unpickler_Readline(self, &s)) < 0) |
5163 | return -1; |
5164 | if (len < 2) |
5165 | return bad_readline(); |
5166 | |
5167 | errno = 0; |
5168 | /* XXX: Should the base argument of strtol() be explicitly set to 10? |
5169 | XXX(avassalotti): Should this uses PyOS_strtol()? */ |
5170 | x = strtol(s, &endptr, 0); |
5171 | |
5172 | if (errno || (*endptr != '\n' && *endptr != '\0')) { |
5173 | /* Hm, maybe we've got something long. Let's try reading |
5174 | * it as a Python int object. */ |
5175 | errno = 0; |
5176 | /* XXX: Same thing about the base here. */ |
5177 | value = PyLong_FromString(s, NULL, 0); |
5178 | if (value == NULL) { |
5179 | PyErr_SetString(PyExc_ValueError, |
5180 | "could not convert string to int" ); |
5181 | return -1; |
5182 | } |
5183 | } |
5184 | else { |
5185 | if (len == 3 && (x == 0 || x == 1)) { |
5186 | if ((value = PyBool_FromLong(x)) == NULL) |
5187 | return -1; |
5188 | } |
5189 | else { |
5190 | if ((value = PyLong_FromLong(x)) == NULL) |
5191 | return -1; |
5192 | } |
5193 | } |
5194 | |
5195 | PDATA_PUSH(self->stack, value, -1); |
5196 | return 0; |
5197 | } |
5198 | |
5199 | static int |
5200 | load_bool(UnpicklerObject *self, PyObject *boolean) |
5201 | { |
5202 | assert(boolean == Py_True || boolean == Py_False); |
5203 | PDATA_APPEND(self->stack, boolean, -1); |
5204 | return 0; |
5205 | } |
5206 | |
5207 | /* s contains x bytes of an unsigned little-endian integer. Return its value |
5208 | * as a C Py_ssize_t, or -1 if it's higher than PY_SSIZE_T_MAX. |
5209 | */ |
5210 | static Py_ssize_t |
5211 | calc_binsize(char *bytes, int nbytes) |
5212 | { |
5213 | unsigned char *s = (unsigned char *)bytes; |
5214 | int i; |
5215 | size_t x = 0; |
5216 | |
5217 | if (nbytes > (int)sizeof(size_t)) { |
5218 | /* Check for integer overflow. BINBYTES8 and BINUNICODE8 opcodes |
5219 | * have 64-bit size that can't be represented on 32-bit platform. |
5220 | */ |
5221 | for (i = (int)sizeof(size_t); i < nbytes; i++) { |
5222 | if (s[i]) |
5223 | return -1; |
5224 | } |
5225 | nbytes = (int)sizeof(size_t); |
5226 | } |
5227 | for (i = 0; i < nbytes; i++) { |
5228 | x |= (size_t) s[i] << (8 * i); |
5229 | } |
5230 | |
5231 | if (x > PY_SSIZE_T_MAX) |
5232 | return -1; |
5233 | else |
5234 | return (Py_ssize_t) x; |
5235 | } |
5236 | |
5237 | /* s contains x bytes of a little-endian integer. Return its value as a |
5238 | * C int. Obscure: when x is 1 or 2, this is an unsigned little-endian |
5239 | * int, but when x is 4 it's a signed one. This is a historical source |
5240 | * of x-platform bugs. |
5241 | */ |
5242 | static long |
5243 | calc_binint(char *bytes, int nbytes) |
5244 | { |
5245 | unsigned char *s = (unsigned char *)bytes; |
5246 | Py_ssize_t i; |
5247 | long x = 0; |
5248 | |
5249 | for (i = 0; i < nbytes; i++) { |
5250 | x |= (long)s[i] << (8 * i); |
5251 | } |
5252 | |
5253 | /* Unlike BININT1 and BININT2, BININT (more accurately BININT4) |
5254 | * is signed, so on a box with longs bigger than 4 bytes we need |
5255 | * to extend a BININT's sign bit to the full width. |
5256 | */ |
5257 | if (SIZEOF_LONG > 4 && nbytes == 4) { |
5258 | x |= -(x & (1L << 31)); |
5259 | } |
5260 | |
5261 | return x; |
5262 | } |
5263 | |
5264 | static int |
5265 | load_binintx(UnpicklerObject *self, char *s, int size) |
5266 | { |
5267 | PyObject *value; |
5268 | long x; |
5269 | |
5270 | x = calc_binint(s, size); |
5271 | |
5272 | if ((value = PyLong_FromLong(x)) == NULL) |
5273 | return -1; |
5274 | |
5275 | PDATA_PUSH(self->stack, value, -1); |
5276 | return 0; |
5277 | } |
5278 | |
5279 | static int |
5280 | load_binint(UnpicklerObject *self) |
5281 | { |
5282 | char *s; |
5283 | |
5284 | if (_Unpickler_Read(self, &s, 4) < 0) |
5285 | return -1; |
5286 | |
5287 | return load_binintx(self, s, 4); |
5288 | } |
5289 | |
5290 | static int |
5291 | load_binint1(UnpicklerObject *self) |
5292 | { |
5293 | char *s; |
5294 | |
5295 | if (_Unpickler_Read(self, &s, 1) < 0) |
5296 | return -1; |
5297 | |
5298 | return load_binintx(self, s, 1); |
5299 | } |
5300 | |
5301 | static int |
5302 | load_binint2(UnpicklerObject *self) |
5303 | { |
5304 | char *s; |
5305 | |
5306 | if (_Unpickler_Read(self, &s, 2) < 0) |
5307 | return -1; |
5308 | |
5309 | return load_binintx(self, s, 2); |
5310 | } |
5311 | |
5312 | static int |
5313 | load_long(UnpicklerObject *self) |
5314 | { |
5315 | PyObject *value; |
5316 | char *s = NULL; |
5317 | Py_ssize_t len; |
5318 | |
5319 | if ((len = _Unpickler_Readline(self, &s)) < 0) |
5320 | return -1; |
5321 | if (len < 2) |
5322 | return bad_readline(); |
5323 | |
5324 | /* s[len-2] will usually be 'L' (and s[len-1] is '\n'); we need to remove |
5325 | the 'L' before calling PyLong_FromString. In order to maintain |
5326 | compatibility with Python 3.0.0, we don't actually *require* |
5327 | the 'L' to be present. */ |
5328 | if (s[len-2] == 'L') |
5329 | s[len-2] = '\0'; |
5330 | /* XXX: Should the base argument explicitly set to 10? */ |
5331 | value = PyLong_FromString(s, NULL, 0); |
5332 | if (value == NULL) |
5333 | return -1; |
5334 | |
5335 | PDATA_PUSH(self->stack, value, -1); |
5336 | return 0; |
5337 | } |
5338 | |
5339 | /* 'size' bytes contain the # of bytes of little-endian 256's-complement |
5340 | * data following. |
5341 | */ |
5342 | static int |
5343 | load_counted_long(UnpicklerObject *self, int size) |
5344 | { |
5345 | PyObject *value; |
5346 | char *nbytes; |
5347 | char *pdata; |
5348 | |
5349 | assert(size == 1 || size == 4); |
5350 | if (_Unpickler_Read(self, &nbytes, size) < 0) |
5351 | return -1; |
5352 | |
5353 | size = calc_binint(nbytes, size); |
5354 | if (size < 0) { |
5355 | PickleState *st = _Pickle_GetGlobalState(); |
5356 | /* Corrupt or hostile pickle -- we never write one like this */ |
5357 | PyErr_SetString(st->UnpicklingError, |
5358 | "LONG pickle has negative byte count" ); |
5359 | return -1; |
5360 | } |
5361 | |
5362 | if (size == 0) |
5363 | value = PyLong_FromLong(0L); |
5364 | else { |
5365 | /* Read the raw little-endian bytes and convert. */ |
5366 | if (_Unpickler_Read(self, &pdata, size) < 0) |
5367 | return -1; |
5368 | value = _PyLong_FromByteArray((unsigned char *)pdata, (size_t)size, |
5369 | 1 /* little endian */ , 1 /* signed */ ); |
5370 | } |
5371 | if (value == NULL) |
5372 | return -1; |
5373 | PDATA_PUSH(self->stack, value, -1); |
5374 | return 0; |
5375 | } |
5376 | |
5377 | static int |
5378 | load_float(UnpicklerObject *self) |
5379 | { |
5380 | PyObject *value; |
5381 | char *endptr, *s; |
5382 | Py_ssize_t len; |
5383 | double d; |
5384 | |
5385 | if ((len = _Unpickler_Readline(self, &s)) < 0) |
5386 | return -1; |
5387 | if (len < 2) |
5388 | return bad_readline(); |
5389 | |
5390 | errno = 0; |
5391 | d = PyOS_string_to_double(s, &endptr, PyExc_OverflowError); |
5392 | if (d == -1.0 && PyErr_Occurred()) |
5393 | return -1; |
5394 | if ((endptr[0] != '\n') && (endptr[0] != '\0')) { |
5395 | PyErr_SetString(PyExc_ValueError, "could not convert string to float" ); |
5396 | return -1; |
5397 | } |
5398 | value = PyFloat_FromDouble(d); |
5399 | if (value == NULL) |
5400 | return -1; |
5401 | |
5402 | PDATA_PUSH(self->stack, value, -1); |
5403 | return 0; |
5404 | } |
5405 | |
5406 | static int |
5407 | load_binfloat(UnpicklerObject *self) |
5408 | { |
5409 | PyObject *value; |
5410 | double x; |
5411 | char *s; |
5412 | |
5413 | if (_Unpickler_Read(self, &s, 8) < 0) |
5414 | return -1; |
5415 | |
5416 | x = _PyFloat_Unpack8((unsigned char *)s, 0); |
5417 | if (x == -1.0 && PyErr_Occurred()) |
5418 | return -1; |
5419 | |
5420 | if ((value = PyFloat_FromDouble(x)) == NULL) |
5421 | return -1; |
5422 | |
5423 | PDATA_PUSH(self->stack, value, -1); |
5424 | return 0; |
5425 | } |
5426 | |
5427 | static int |
5428 | load_string(UnpicklerObject *self) |
5429 | { |
5430 | PyObject *bytes; |
5431 | PyObject *obj; |
5432 | Py_ssize_t len; |
5433 | char *s, *p; |
5434 | |
5435 | if ((len = _Unpickler_Readline(self, &s)) < 0) |
5436 | return -1; |
5437 | /* Strip the newline */ |
5438 | len--; |
5439 | /* Strip outermost quotes */ |
5440 | if (len >= 2 && s[0] == s[len - 1] && (s[0] == '\'' || s[0] == '"')) { |
5441 | p = s + 1; |
5442 | len -= 2; |
5443 | } |
5444 | else { |
5445 | PickleState *st = _Pickle_GetGlobalState(); |
5446 | PyErr_SetString(st->UnpicklingError, |
5447 | "the STRING opcode argument must be quoted" ); |
5448 | return -1; |
5449 | } |
5450 | assert(len >= 0); |
5451 | |
5452 | /* Use the PyBytes API to decode the string, since that is what is used |
5453 | to encode, and then coerce the result to Unicode. */ |
5454 | bytes = PyBytes_DecodeEscape(p, len, NULL, 0, NULL); |
5455 | if (bytes == NULL) |
5456 | return -1; |
5457 | |
5458 | /* Leave the Python 2.x strings as bytes if the *encoding* given to the |
5459 | Unpickler was 'bytes'. Otherwise, convert them to unicode. */ |
5460 | if (strcmp(self->encoding, "bytes" ) == 0) { |
5461 | obj = bytes; |
5462 | } |
5463 | else { |
5464 | obj = PyUnicode_FromEncodedObject(bytes, self->encoding, self->errors); |
5465 | Py_DECREF(bytes); |
5466 | if (obj == NULL) { |
5467 | return -1; |
5468 | } |
5469 | } |
5470 | |
5471 | PDATA_PUSH(self->stack, obj, -1); |
5472 | return 0; |
5473 | } |
5474 | |
5475 | static int |
5476 | load_counted_binstring(UnpicklerObject *self, int nbytes) |
5477 | { |
5478 | PyObject *obj; |
5479 | Py_ssize_t size; |
5480 | char *s; |
5481 | |
5482 | if (_Unpickler_Read(self, &s, nbytes) < 0) |
5483 | return -1; |
5484 | |
5485 | size = calc_binsize(s, nbytes); |
5486 | if (size < 0) { |
5487 | PickleState *st = _Pickle_GetGlobalState(); |
5488 | PyErr_Format(st->UnpicklingError, |
5489 | "BINSTRING exceeds system's maximum size of %zd bytes" , |
5490 | PY_SSIZE_T_MAX); |
5491 | return -1; |
5492 | } |
5493 | |
5494 | if (_Unpickler_Read(self, &s, size) < 0) |
5495 | return -1; |
5496 | |
5497 | /* Convert Python 2.x strings to bytes if the *encoding* given to the |
5498 | Unpickler was 'bytes'. Otherwise, convert them to unicode. */ |
5499 | if (strcmp(self->encoding, "bytes" ) == 0) { |
5500 | obj = PyBytes_FromStringAndSize(s, size); |
5501 | } |
5502 | else { |
5503 | obj = PyUnicode_Decode(s, size, self->encoding, self->errors); |
5504 | } |
5505 | if (obj == NULL) { |
5506 | return -1; |
5507 | } |
5508 | |
5509 | PDATA_PUSH(self->stack, obj, -1); |
5510 | return 0; |
5511 | } |
5512 | |
5513 | static int |
5514 | load_counted_binbytes(UnpicklerObject *self, int nbytes) |
5515 | { |
5516 | PyObject *bytes; |
5517 | Py_ssize_t size; |
5518 | char *s; |
5519 | |
5520 | if (_Unpickler_Read(self, &s, nbytes) < 0) |
5521 | return -1; |
5522 | |
5523 | size = calc_binsize(s, nbytes); |
5524 | if (size < 0) { |
5525 | PyErr_Format(PyExc_OverflowError, |
5526 | "BINBYTES exceeds system's maximum size of %zd bytes" , |
5527 | PY_SSIZE_T_MAX); |
5528 | return -1; |
5529 | } |
5530 | |
5531 | bytes = PyBytes_FromStringAndSize(NULL, size); |
5532 | if (bytes == NULL) |
5533 | return -1; |
5534 | if (_Unpickler_ReadInto(self, PyBytes_AS_STRING(bytes), size) < 0) { |
5535 | Py_DECREF(bytes); |
5536 | return -1; |
5537 | } |
5538 | |
5539 | PDATA_PUSH(self->stack, bytes, -1); |
5540 | return 0; |
5541 | } |
5542 | |
5543 | static int |
5544 | load_counted_bytearray(UnpicklerObject *self) |
5545 | { |
5546 | PyObject *bytearray; |
5547 | Py_ssize_t size; |
5548 | char *s; |
5549 | |
5550 | if (_Unpickler_Read(self, &s, 8) < 0) { |
5551 | return -1; |
5552 | } |
5553 | |
5554 | size = calc_binsize(s, 8); |
5555 | if (size < 0) { |
5556 | PyErr_Format(PyExc_OverflowError, |
5557 | "BYTEARRAY8 exceeds system's maximum size of %zd bytes" , |
5558 | PY_SSIZE_T_MAX); |
5559 | return -1; |
5560 | } |
5561 | |
5562 | bytearray = PyByteArray_FromStringAndSize(NULL, size); |
5563 | if (bytearray == NULL) { |
5564 | return -1; |
5565 | } |
5566 | if (_Unpickler_ReadInto(self, PyByteArray_AS_STRING(bytearray), size) < 0) { |
5567 | Py_DECREF(bytearray); |
5568 | return -1; |
5569 | } |
5570 | |
5571 | PDATA_PUSH(self->stack, bytearray, -1); |
5572 | return 0; |
5573 | } |
5574 | |
5575 | static int |
5576 | load_next_buffer(UnpicklerObject *self) |
5577 | { |
5578 | if (self->buffers == NULL) { |
5579 | PickleState *st = _Pickle_GetGlobalState(); |
5580 | PyErr_SetString(st->UnpicklingError, |
5581 | "pickle stream refers to out-of-band data " |
5582 | "but no *buffers* argument was given" ); |
5583 | return -1; |
5584 | } |
5585 | PyObject *buf = PyIter_Next(self->buffers); |
5586 | if (buf == NULL) { |
5587 | if (!PyErr_Occurred()) { |
5588 | PickleState *st = _Pickle_GetGlobalState(); |
5589 | PyErr_SetString(st->UnpicklingError, |
5590 | "not enough out-of-band buffers" ); |
5591 | } |
5592 | return -1; |
5593 | } |
5594 | |
5595 | PDATA_PUSH(self->stack, buf, -1); |
5596 | return 0; |
5597 | } |
5598 | |
5599 | static int |
5600 | load_readonly_buffer(UnpicklerObject *self) |
5601 | { |
5602 | Py_ssize_t len = Py_SIZE(self->stack); |
5603 | if (len <= self->stack->fence) { |
5604 | return Pdata_stack_underflow(self->stack); |
5605 | } |
5606 | |
5607 | PyObject *obj = self->stack->data[len - 1]; |
5608 | PyObject *view = PyMemoryView_FromObject(obj); |
5609 | if (view == NULL) { |
5610 | return -1; |
5611 | } |
5612 | if (!PyMemoryView_GET_BUFFER(view)->readonly) { |
5613 | /* Original object is writable */ |
5614 | PyMemoryView_GET_BUFFER(view)->readonly = 1; |
5615 | self->stack->data[len - 1] = view; |
5616 | Py_DECREF(obj); |
5617 | } |
5618 | else { |
5619 | /* Original object is read-only, no need to replace it */ |
5620 | Py_DECREF(view); |
5621 | } |
5622 | return 0; |
5623 | } |
5624 | |
5625 | static int |
5626 | load_unicode(UnpicklerObject *self) |
5627 | { |
5628 | PyObject *str; |
5629 | Py_ssize_t len; |
5630 | char *s = NULL; |
5631 | |
5632 | if ((len = _Unpickler_Readline(self, &s)) < 0) |
5633 | return -1; |
5634 | if (len < 1) |
5635 | return bad_readline(); |
5636 | |
5637 | str = PyUnicode_DecodeRawUnicodeEscape(s, len - 1, NULL); |
5638 | if (str == NULL) |
5639 | return -1; |
5640 | |
5641 | PDATA_PUSH(self->stack, str, -1); |
5642 | return 0; |
5643 | } |
5644 | |
5645 | static int |
5646 | load_counted_binunicode(UnpicklerObject *self, int nbytes) |
5647 | { |
5648 | PyObject *str; |
5649 | Py_ssize_t size; |
5650 | char *s; |
5651 | |
5652 | if (_Unpickler_Read(self, &s, nbytes) < 0) |
5653 | return -1; |
5654 | |
5655 | size = calc_binsize(s, nbytes); |
5656 | if (size < 0) { |
5657 | PyErr_Format(PyExc_OverflowError, |
5658 | "BINUNICODE exceeds system's maximum size of %zd bytes" , |
5659 | PY_SSIZE_T_MAX); |
5660 | return -1; |
5661 | } |
5662 | |
5663 | if (_Unpickler_Read(self, &s, size) < 0) |
5664 | return -1; |
5665 | |
5666 | str = PyUnicode_DecodeUTF8(s, size, "surrogatepass" ); |
5667 | if (str == NULL) |
5668 | return -1; |
5669 | |
5670 | PDATA_PUSH(self->stack, str, -1); |
5671 | return 0; |
5672 | } |
5673 | |
5674 | static int |
5675 | load_counted_tuple(UnpicklerObject *self, Py_ssize_t len) |
5676 | { |
5677 | PyObject *tuple; |
5678 | |
5679 | if (Py_SIZE(self->stack) < len) |
5680 | return Pdata_stack_underflow(self->stack); |
5681 | |
5682 | tuple = Pdata_poptuple(self->stack, Py_SIZE(self->stack) - len); |
5683 | if (tuple == NULL) |
5684 | return -1; |
5685 | PDATA_PUSH(self->stack, tuple, -1); |
5686 | return 0; |
5687 | } |
5688 | |
5689 | static int |
5690 | load_tuple(UnpicklerObject *self) |
5691 | { |
5692 | Py_ssize_t i; |
5693 | |
5694 | if ((i = marker(self)) < 0) |
5695 | return -1; |
5696 | |
5697 | return load_counted_tuple(self, Py_SIZE(self->stack) - i); |
5698 | } |
5699 | |
5700 | static int |
5701 | load_empty_list(UnpicklerObject *self) |
5702 | { |
5703 | PyObject *list; |
5704 | |
5705 | if ((list = PyList_New(0)) == NULL) |
5706 | return -1; |
5707 | PDATA_PUSH(self->stack, list, -1); |
5708 | return 0; |
5709 | } |
5710 | |
5711 | static int |
5712 | load_empty_dict(UnpicklerObject *self) |
5713 | { |
5714 | PyObject *dict; |
5715 | |
5716 | if ((dict = PyDict_New()) == NULL) |
5717 | return -1; |
5718 | PDATA_PUSH(self->stack, dict, -1); |
5719 | return 0; |
5720 | } |
5721 | |
5722 | static int |
5723 | load_empty_set(UnpicklerObject *self) |
5724 | { |
5725 | PyObject *set; |
5726 | |
5727 | if ((set = PySet_New(NULL)) == NULL) |
5728 | return -1; |
5729 | PDATA_PUSH(self->stack, set, -1); |
5730 | return 0; |
5731 | } |
5732 | |
5733 | static int |
5734 | load_list(UnpicklerObject *self) |
5735 | { |
5736 | PyObject *list; |
5737 | Py_ssize_t i; |
5738 | |
5739 | if ((i = marker(self)) < 0) |
5740 | return -1; |
5741 | |
5742 | list = Pdata_poplist(self->stack, i); |
5743 | if (list == NULL) |
5744 | return -1; |
5745 | PDATA_PUSH(self->stack, list, -1); |
5746 | return 0; |
5747 | } |
5748 | |
5749 | static int |
5750 | load_dict(UnpicklerObject *self) |
5751 | { |
5752 | PyObject *dict, *key, *value; |
5753 | Py_ssize_t i, j, k; |
5754 | |
5755 | if ((i = marker(self)) < 0) |
5756 | return -1; |
5757 | j = Py_SIZE(self->stack); |
5758 | |
5759 | if ((dict = PyDict_New()) == NULL) |
5760 | return -1; |
5761 | |
5762 | if ((j - i) % 2 != 0) { |
5763 | PickleState *st = _Pickle_GetGlobalState(); |
5764 | PyErr_SetString(st->UnpicklingError, "odd number of items for DICT" ); |
5765 | Py_DECREF(dict); |
5766 | return -1; |
5767 | } |
5768 | |
5769 | for (k = i + 1; k < j; k += 2) { |
5770 | key = self->stack->data[k - 1]; |
5771 | value = self->stack->data[k]; |
5772 | if (PyDict_SetItem(dict, key, value) < 0) { |
5773 | Py_DECREF(dict); |
5774 | return -1; |
5775 | } |
5776 | } |
5777 | Pdata_clear(self->stack, i); |
5778 | PDATA_PUSH(self->stack, dict, -1); |
5779 | return 0; |
5780 | } |
5781 | |
5782 | static int |
5783 | load_frozenset(UnpicklerObject *self) |
5784 | { |
5785 | PyObject *items; |
5786 | PyObject *frozenset; |
5787 | Py_ssize_t i; |
5788 | |
5789 | if ((i = marker(self)) < 0) |
5790 | return -1; |
5791 | |
5792 | items = Pdata_poptuple(self->stack, i); |
5793 | if (items == NULL) |
5794 | return -1; |
5795 | |
5796 | frozenset = PyFrozenSet_New(items); |
5797 | Py_DECREF(items); |
5798 | if (frozenset == NULL) |
5799 | return -1; |
5800 | |
5801 | PDATA_PUSH(self->stack, frozenset, -1); |
5802 | return 0; |
5803 | } |
5804 | |
5805 | static PyObject * |
5806 | instantiate(PyObject *cls, PyObject *args) |
5807 | { |
5808 | /* Caller must assure args are a tuple. Normally, args come from |
5809 | Pdata_poptuple which packs objects from the top of the stack |
5810 | into a newly created tuple. */ |
5811 | assert(PyTuple_Check(args)); |
5812 | if (!PyTuple_GET_SIZE(args) && PyType_Check(cls)) { |
5813 | _Py_IDENTIFIER(__getinitargs__); |
5814 | _Py_IDENTIFIER(__new__); |
5815 | PyObject *func; |
5816 | if (_PyObject_LookupAttrId(cls, &PyId___getinitargs__, &func) < 0) { |
5817 | return NULL; |
5818 | } |
5819 | if (func == NULL) { |
5820 | return _PyObject_CallMethodIdOneArg(cls, &PyId___new__, cls); |
5821 | } |
5822 | Py_DECREF(func); |
5823 | } |
5824 | return PyObject_CallObject(cls, args); |
5825 | } |
5826 | |
5827 | static int |
5828 | load_obj(UnpicklerObject *self) |
5829 | { |
5830 | PyObject *cls, *args, *obj = NULL; |
5831 | Py_ssize_t i; |
5832 | |
5833 | if ((i = marker(self)) < 0) |
5834 | return -1; |
5835 | |
5836 | if (Py_SIZE(self->stack) - i < 1) |
5837 | return Pdata_stack_underflow(self->stack); |
5838 | |
5839 | args = Pdata_poptuple(self->stack, i + 1); |
5840 | if (args == NULL) |
5841 | return -1; |
5842 | |
5843 | PDATA_POP(self->stack, cls); |
5844 | if (cls) { |
5845 | obj = instantiate(cls, args); |
5846 | Py_DECREF(cls); |
5847 | } |
5848 | Py_DECREF(args); |
5849 | if (obj == NULL) |
5850 | return -1; |
5851 | |
5852 | PDATA_PUSH(self->stack, obj, -1); |
5853 | return 0; |
5854 | } |
5855 | |
5856 | static int |
5857 | load_inst(UnpicklerObject *self) |
5858 | { |
5859 | PyObject *cls = NULL; |
5860 | PyObject *args = NULL; |
5861 | PyObject *obj = NULL; |
5862 | PyObject *module_name; |
5863 | PyObject *class_name; |
5864 | Py_ssize_t len; |
5865 | Py_ssize_t i; |
5866 | char *s; |
5867 | |
5868 | if ((i = marker(self)) < 0) |
5869 | return -1; |
5870 | if ((len = _Unpickler_Readline(self, &s)) < 0) |
5871 | return -1; |
5872 | if (len < 2) |
5873 | return bad_readline(); |
5874 | |
5875 | /* Here it is safe to use PyUnicode_DecodeASCII(), even though non-ASCII |
5876 | identifiers are permitted in Python 3.0, since the INST opcode is only |
5877 | supported by older protocols on Python 2.x. */ |
5878 | module_name = PyUnicode_DecodeASCII(s, len - 1, "strict" ); |
5879 | if (module_name == NULL) |
5880 | return -1; |
5881 | |
5882 | if ((len = _Unpickler_Readline(self, &s)) >= 0) { |
5883 | if (len < 2) { |
5884 | Py_DECREF(module_name); |
5885 | return bad_readline(); |
5886 | } |
5887 | class_name = PyUnicode_DecodeASCII(s, len - 1, "strict" ); |
5888 | if (class_name != NULL) { |
5889 | cls = find_class(self, module_name, class_name); |
5890 | Py_DECREF(class_name); |
5891 | } |
5892 | } |
5893 | Py_DECREF(module_name); |
5894 | |
5895 | if (cls == NULL) |
5896 | return -1; |
5897 | |
5898 | if ((args = Pdata_poptuple(self->stack, i)) != NULL) { |
5899 | obj = instantiate(cls, args); |
5900 | Py_DECREF(args); |
5901 | } |
5902 | Py_DECREF(cls); |
5903 | |
5904 | if (obj == NULL) |
5905 | return -1; |
5906 | |
5907 | PDATA_PUSH(self->stack, obj, -1); |
5908 | return 0; |
5909 | } |
5910 | |
5911 | static void |
5912 | newobj_unpickling_error(const char * msg, int use_kwargs, PyObject *arg) |
5913 | { |
5914 | PickleState *st = _Pickle_GetGlobalState(); |
5915 | PyErr_Format(st->UnpicklingError, msg, |
5916 | use_kwargs ? "NEWOBJ_EX" : "NEWOBJ" , |
5917 | Py_TYPE(arg)->tp_name); |
5918 | } |
5919 | |
5920 | static int |
5921 | load_newobj(UnpicklerObject *self, int use_kwargs) |
5922 | { |
5923 | PyObject *cls, *args, *kwargs = NULL; |
5924 | PyObject *obj; |
5925 | |
5926 | /* Stack is ... cls args [kwargs], and we want to call |
5927 | * cls.__new__(cls, *args, **kwargs). |
5928 | */ |
5929 | if (use_kwargs) { |
5930 | PDATA_POP(self->stack, kwargs); |
5931 | if (kwargs == NULL) { |
5932 | return -1; |
5933 | } |
5934 | } |
5935 | PDATA_POP(self->stack, args); |
5936 | if (args == NULL) { |
5937 | Py_XDECREF(kwargs); |
5938 | return -1; |
5939 | } |
5940 | PDATA_POP(self->stack, cls); |
5941 | if (cls == NULL) { |
5942 | Py_XDECREF(kwargs); |
5943 | Py_DECREF(args); |
5944 | return -1; |
5945 | } |
5946 | |
5947 | if (!PyType_Check(cls)) { |
5948 | newobj_unpickling_error("%s class argument must be a type, not %.200s" , |
5949 | use_kwargs, cls); |
5950 | goto error; |
5951 | } |
5952 | if (((PyTypeObject *)cls)->tp_new == NULL) { |
5953 | newobj_unpickling_error("%s class argument '%.200s' doesn't have __new__" , |
5954 | use_kwargs, cls); |
5955 | goto error; |
5956 | } |
5957 | if (!PyTuple_Check(args)) { |
5958 | newobj_unpickling_error("%s args argument must be a tuple, not %.200s" , |
5959 | use_kwargs, args); |
5960 | goto error; |
5961 | } |
5962 | if (use_kwargs && !PyDict_Check(kwargs)) { |
5963 | newobj_unpickling_error("%s kwargs argument must be a dict, not %.200s" , |
5964 | use_kwargs, kwargs); |
5965 | goto error; |
5966 | } |
5967 | |
5968 | obj = ((PyTypeObject *)cls)->tp_new((PyTypeObject *)cls, args, kwargs); |
5969 | if (obj == NULL) { |
5970 | goto error; |
5971 | } |
5972 | Py_XDECREF(kwargs); |
5973 | Py_DECREF(args); |
5974 | Py_DECREF(cls); |
5975 | PDATA_PUSH(self->stack, obj, -1); |
5976 | return 0; |
5977 | |
5978 | error: |
5979 | Py_XDECREF(kwargs); |
5980 | Py_DECREF(args); |
5981 | Py_DECREF(cls); |
5982 | return -1; |
5983 | } |
5984 | |
5985 | static int |
5986 | load_global(UnpicklerObject *self) |
5987 | { |
5988 | PyObject *global = NULL; |
5989 | PyObject *module_name; |
5990 | PyObject *global_name; |
5991 | Py_ssize_t len; |
5992 | char *s; |
5993 | |
5994 | if ((len = _Unpickler_Readline(self, &s)) < 0) |
5995 | return -1; |
5996 | if (len < 2) |
5997 | return bad_readline(); |
5998 | module_name = PyUnicode_DecodeUTF8(s, len - 1, "strict" ); |
5999 | if (!module_name) |
6000 | return -1; |
6001 | |
6002 | if ((len = _Unpickler_Readline(self, &s)) >= 0) { |
6003 | if (len < 2) { |
6004 | Py_DECREF(module_name); |
6005 | return bad_readline(); |
6006 | } |
6007 | global_name = PyUnicode_DecodeUTF8(s, len - 1, "strict" ); |
6008 | if (global_name) { |
6009 | global = find_class(self, module_name, global_name); |
6010 | Py_DECREF(global_name); |
6011 | } |
6012 | } |
6013 | Py_DECREF(module_name); |
6014 | |
6015 | if (global == NULL) |
6016 | return -1; |
6017 | PDATA_PUSH(self->stack, global, -1); |
6018 | return 0; |
6019 | } |
6020 | |
6021 | static int |
6022 | load_stack_global(UnpicklerObject *self) |
6023 | { |
6024 | PyObject *global; |
6025 | PyObject *module_name; |
6026 | PyObject *global_name; |
6027 | |
6028 | PDATA_POP(self->stack, global_name); |
6029 | PDATA_POP(self->stack, module_name); |
6030 | if (module_name == NULL || !PyUnicode_CheckExact(module_name) || |
6031 | global_name == NULL || !PyUnicode_CheckExact(global_name)) { |
6032 | PickleState *st = _Pickle_GetGlobalState(); |
6033 | PyErr_SetString(st->UnpicklingError, "STACK_GLOBAL requires str" ); |
6034 | Py_XDECREF(global_name); |
6035 | Py_XDECREF(module_name); |
6036 | return -1; |
6037 | } |
6038 | global = find_class(self, module_name, global_name); |
6039 | Py_DECREF(global_name); |
6040 | Py_DECREF(module_name); |
6041 | if (global == NULL) |
6042 | return -1; |
6043 | PDATA_PUSH(self->stack, global, -1); |
6044 | return 0; |
6045 | } |
6046 | |
6047 | static int |
6048 | load_persid(UnpicklerObject *self) |
6049 | { |
6050 | PyObject *pid, *obj; |
6051 | Py_ssize_t len; |
6052 | char *s; |
6053 | |
6054 | if (self->pers_func) { |
6055 | if ((len = _Unpickler_Readline(self, &s)) < 0) |
6056 | return -1; |
6057 | if (len < 1) |
6058 | return bad_readline(); |
6059 | |
6060 | pid = PyUnicode_DecodeASCII(s, len - 1, "strict" ); |
6061 | if (pid == NULL) { |
6062 | if (PyErr_ExceptionMatches(PyExc_UnicodeDecodeError)) { |
6063 | PyErr_SetString(_Pickle_GetGlobalState()->UnpicklingError, |
6064 | "persistent IDs in protocol 0 must be " |
6065 | "ASCII strings" ); |
6066 | } |
6067 | return -1; |
6068 | } |
6069 | |
6070 | obj = call_method(self->pers_func, self->pers_func_self, pid); |
6071 | Py_DECREF(pid); |
6072 | if (obj == NULL) |
6073 | return -1; |
6074 | |
6075 | PDATA_PUSH(self->stack, obj, -1); |
6076 | return 0; |
6077 | } |
6078 | else { |
6079 | PickleState *st = _Pickle_GetGlobalState(); |
6080 | PyErr_SetString(st->UnpicklingError, |
6081 | "A load persistent id instruction was encountered,\n" |
6082 | "but no persistent_load function was specified." ); |
6083 | return -1; |
6084 | } |
6085 | } |
6086 | |
6087 | static int |
6088 | load_binpersid(UnpicklerObject *self) |
6089 | { |
6090 | PyObject *pid, *obj; |
6091 | |
6092 | if (self->pers_func) { |
6093 | PDATA_POP(self->stack, pid); |
6094 | if (pid == NULL) |
6095 | return -1; |
6096 | |
6097 | obj = call_method(self->pers_func, self->pers_func_self, pid); |
6098 | Py_DECREF(pid); |
6099 | if (obj == NULL) |
6100 | return -1; |
6101 | |
6102 | PDATA_PUSH(self->stack, obj, -1); |
6103 | return 0; |
6104 | } |
6105 | else { |
6106 | PickleState *st = _Pickle_GetGlobalState(); |
6107 | PyErr_SetString(st->UnpicklingError, |
6108 | "A load persistent id instruction was encountered,\n" |
6109 | "but no persistent_load function was specified." ); |
6110 | return -1; |
6111 | } |
6112 | } |
6113 | |
6114 | static int |
6115 | load_pop(UnpicklerObject *self) |
6116 | { |
6117 | Py_ssize_t len = Py_SIZE(self->stack); |
6118 | |
6119 | /* Note that we split the (pickle.py) stack into two stacks, |
6120 | * an object stack and a mark stack. We have to be clever and |
6121 | * pop the right one. We do this by looking at the top of the |
6122 | * mark stack first, and only signalling a stack underflow if |
6123 | * the object stack is empty and the mark stack doesn't match |
6124 | * our expectations. |
6125 | */ |
6126 | if (self->num_marks > 0 && self->marks[self->num_marks - 1] == len) { |
6127 | self->num_marks--; |
6128 | self->stack->mark_set = self->num_marks != 0; |
6129 | self->stack->fence = self->num_marks ? |
6130 | self->marks[self->num_marks - 1] : 0; |
6131 | } else if (len <= self->stack->fence) |
6132 | return Pdata_stack_underflow(self->stack); |
6133 | else { |
6134 | len--; |
6135 | Py_DECREF(self->stack->data[len]); |
6136 | Py_SET_SIZE(self->stack, len); |
6137 | } |
6138 | return 0; |
6139 | } |
6140 | |
6141 | static int |
6142 | load_pop_mark(UnpicklerObject *self) |
6143 | { |
6144 | Py_ssize_t i; |
6145 | |
6146 | if ((i = marker(self)) < 0) |
6147 | return -1; |
6148 | |
6149 | Pdata_clear(self->stack, i); |
6150 | |
6151 | return 0; |
6152 | } |
6153 | |
6154 | static int |
6155 | load_dup(UnpicklerObject *self) |
6156 | { |
6157 | PyObject *last; |
6158 | Py_ssize_t len = Py_SIZE(self->stack); |
6159 | |
6160 | if (len <= self->stack->fence) |
6161 | return Pdata_stack_underflow(self->stack); |
6162 | last = self->stack->data[len - 1]; |
6163 | PDATA_APPEND(self->stack, last, -1); |
6164 | return 0; |
6165 | } |
6166 | |
6167 | static int |
6168 | load_get(UnpicklerObject *self) |
6169 | { |
6170 | PyObject *key, *value; |
6171 | Py_ssize_t idx; |
6172 | Py_ssize_t len; |
6173 | char *s; |
6174 | |
6175 | if ((len = _Unpickler_Readline(self, &s)) < 0) |
6176 | return -1; |
6177 | if (len < 2) |
6178 | return bad_readline(); |
6179 | |
6180 | key = PyLong_FromString(s, NULL, 10); |
6181 | if (key == NULL) |
6182 | return -1; |
6183 | idx = PyLong_AsSsize_t(key); |
6184 | if (idx == -1 && PyErr_Occurred()) { |
6185 | Py_DECREF(key); |
6186 | return -1; |
6187 | } |
6188 | |
6189 | value = _Unpickler_MemoGet(self, idx); |
6190 | if (value == NULL) { |
6191 | if (!PyErr_Occurred()) { |
6192 | PickleState *st = _Pickle_GetGlobalState(); |
6193 | PyErr_Format(st->UnpicklingError, "Memo value not found at index %ld" , idx); |
6194 | } |
6195 | Py_DECREF(key); |
6196 | return -1; |
6197 | } |
6198 | Py_DECREF(key); |
6199 | |
6200 | PDATA_APPEND(self->stack, value, -1); |
6201 | return 0; |
6202 | } |
6203 | |
6204 | static int |
6205 | load_binget(UnpicklerObject *self) |
6206 | { |
6207 | PyObject *value; |
6208 | Py_ssize_t idx; |
6209 | char *s; |
6210 | |
6211 | if (_Unpickler_Read(self, &s, 1) < 0) |
6212 | return -1; |
6213 | |
6214 | idx = Py_CHARMASK(s[0]); |
6215 | |
6216 | value = _Unpickler_MemoGet(self, idx); |
6217 | if (value == NULL) { |
6218 | PyObject *key = PyLong_FromSsize_t(idx); |
6219 | if (key != NULL) { |
6220 | PickleState *st = _Pickle_GetGlobalState(); |
6221 | PyErr_Format(st->UnpicklingError, "Memo value not found at index %ld" , idx); |
6222 | Py_DECREF(key); |
6223 | } |
6224 | return -1; |
6225 | } |
6226 | |
6227 | PDATA_APPEND(self->stack, value, -1); |
6228 | return 0; |
6229 | } |
6230 | |
6231 | static int |
6232 | load_long_binget(UnpicklerObject *self) |
6233 | { |
6234 | PyObject *value; |
6235 | Py_ssize_t idx; |
6236 | char *s; |
6237 | |
6238 | if (_Unpickler_Read(self, &s, 4) < 0) |
6239 | return -1; |
6240 | |
6241 | idx = calc_binsize(s, 4); |
6242 | |
6243 | value = _Unpickler_MemoGet(self, idx); |
6244 | if (value == NULL) { |
6245 | PyObject *key = PyLong_FromSsize_t(idx); |
6246 | if (key != NULL) { |
6247 | PickleState *st = _Pickle_GetGlobalState(); |
6248 | PyErr_Format(st->UnpicklingError, "Memo value not found at index %ld" , idx); |
6249 | Py_DECREF(key); |
6250 | } |
6251 | return -1; |
6252 | } |
6253 | |
6254 | PDATA_APPEND(self->stack, value, -1); |
6255 | return 0; |
6256 | } |
6257 | |
6258 | /* Push an object from the extension registry (EXT[124]). nbytes is |
6259 | * the number of bytes following the opcode, holding the index (code) value. |
6260 | */ |
6261 | static int |
6262 | load_extension(UnpicklerObject *self, int nbytes) |
6263 | { |
6264 | char *codebytes; /* the nbytes bytes after the opcode */ |
6265 | long code; /* calc_binint returns long */ |
6266 | PyObject *py_code; /* code as a Python int */ |
6267 | PyObject *obj; /* the object to push */ |
6268 | PyObject *pair; /* (module_name, class_name) */ |
6269 | PyObject *module_name, *class_name; |
6270 | PickleState *st = _Pickle_GetGlobalState(); |
6271 | |
6272 | assert(nbytes == 1 || nbytes == 2 || nbytes == 4); |
6273 | if (_Unpickler_Read(self, &codebytes, nbytes) < 0) |
6274 | return -1; |
6275 | code = calc_binint(codebytes, nbytes); |
6276 | if (code <= 0) { /* note that 0 is forbidden */ |
6277 | /* Corrupt or hostile pickle. */ |
6278 | PyErr_SetString(st->UnpicklingError, "EXT specifies code <= 0" ); |
6279 | return -1; |
6280 | } |
6281 | |
6282 | /* Look for the code in the cache. */ |
6283 | py_code = PyLong_FromLong(code); |
6284 | if (py_code == NULL) |
6285 | return -1; |
6286 | obj = PyDict_GetItemWithError(st->extension_cache, py_code); |
6287 | if (obj != NULL) { |
6288 | /* Bingo. */ |
6289 | Py_DECREF(py_code); |
6290 | PDATA_APPEND(self->stack, obj, -1); |
6291 | return 0; |
6292 | } |
6293 | if (PyErr_Occurred()) { |
6294 | Py_DECREF(py_code); |
6295 | return -1; |
6296 | } |
6297 | |
6298 | /* Look up the (module_name, class_name) pair. */ |
6299 | pair = PyDict_GetItemWithError(st->inverted_registry, py_code); |
6300 | if (pair == NULL) { |
6301 | Py_DECREF(py_code); |
6302 | if (!PyErr_Occurred()) { |
6303 | PyErr_Format(PyExc_ValueError, "unregistered extension " |
6304 | "code %ld" , code); |
6305 | } |
6306 | return -1; |
6307 | } |
6308 | /* Since the extension registry is manipulable via Python code, |
6309 | * confirm that pair is really a 2-tuple of strings. |
6310 | */ |
6311 | if (!PyTuple_Check(pair) || PyTuple_Size(pair) != 2) { |
6312 | goto error; |
6313 | } |
6314 | |
6315 | module_name = PyTuple_GET_ITEM(pair, 0); |
6316 | if (!PyUnicode_Check(module_name)) { |
6317 | goto error; |
6318 | } |
6319 | |
6320 | class_name = PyTuple_GET_ITEM(pair, 1); |
6321 | if (!PyUnicode_Check(class_name)) { |
6322 | goto error; |
6323 | } |
6324 | |
6325 | /* Load the object. */ |
6326 | obj = find_class(self, module_name, class_name); |
6327 | if (obj == NULL) { |
6328 | Py_DECREF(py_code); |
6329 | return -1; |
6330 | } |
6331 | /* Cache code -> obj. */ |
6332 | code = PyDict_SetItem(st->extension_cache, py_code, obj); |
6333 | Py_DECREF(py_code); |
6334 | if (code < 0) { |
6335 | Py_DECREF(obj); |
6336 | return -1; |
6337 | } |
6338 | PDATA_PUSH(self->stack, obj, -1); |
6339 | return 0; |
6340 | |
6341 | error: |
6342 | Py_DECREF(py_code); |
6343 | PyErr_Format(PyExc_ValueError, "_inverted_registry[%ld] " |
6344 | "isn't a 2-tuple of strings" , code); |
6345 | return -1; |
6346 | } |
6347 | |
6348 | static int |
6349 | load_put(UnpicklerObject *self) |
6350 | { |
6351 | PyObject *key, *value; |
6352 | Py_ssize_t idx; |
6353 | Py_ssize_t len; |
6354 | char *s = NULL; |
6355 | |
6356 | if ((len = _Unpickler_Readline(self, &s)) < 0) |
6357 | return -1; |
6358 | if (len < 2) |
6359 | return bad_readline(); |
6360 | if (Py_SIZE(self->stack) <= self->stack->fence) |
6361 | return Pdata_stack_underflow(self->stack); |
6362 | value = self->stack->data[Py_SIZE(self->stack) - 1]; |
6363 | |
6364 | key = PyLong_FromString(s, NULL, 10); |
6365 | if (key == NULL) |
6366 | return -1; |
6367 | idx = PyLong_AsSsize_t(key); |
6368 | Py_DECREF(key); |
6369 | if (idx < 0) { |
6370 | if (!PyErr_Occurred()) |
6371 | PyErr_SetString(PyExc_ValueError, |
6372 | "negative PUT argument" ); |
6373 | return -1; |
6374 | } |
6375 | |
6376 | return _Unpickler_MemoPut(self, idx, value); |
6377 | } |
6378 | |
6379 | static int |
6380 | load_binput(UnpicklerObject *self) |
6381 | { |
6382 | PyObject *value; |
6383 | Py_ssize_t idx; |
6384 | char *s; |
6385 | |
6386 | if (_Unpickler_Read(self, &s, 1) < 0) |
6387 | return -1; |
6388 | |
6389 | if (Py_SIZE(self->stack) <= self->stack->fence) |
6390 | return Pdata_stack_underflow(self->stack); |
6391 | value = self->stack->data[Py_SIZE(self->stack) - 1]; |
6392 | |
6393 | idx = Py_CHARMASK(s[0]); |
6394 | |
6395 | return _Unpickler_MemoPut(self, idx, value); |
6396 | } |
6397 | |
6398 | static int |
6399 | load_long_binput(UnpicklerObject *self) |
6400 | { |
6401 | PyObject *value; |
6402 | Py_ssize_t idx; |
6403 | char *s; |
6404 | |
6405 | if (_Unpickler_Read(self, &s, 4) < 0) |
6406 | return -1; |
6407 | |
6408 | if (Py_SIZE(self->stack) <= self->stack->fence) |
6409 | return Pdata_stack_underflow(self->stack); |
6410 | value = self->stack->data[Py_SIZE(self->stack) - 1]; |
6411 | |
6412 | idx = calc_binsize(s, 4); |
6413 | if (idx < 0) { |
6414 | PyErr_SetString(PyExc_ValueError, |
6415 | "negative LONG_BINPUT argument" ); |
6416 | return -1; |
6417 | } |
6418 | |
6419 | return _Unpickler_MemoPut(self, idx, value); |
6420 | } |
6421 | |
6422 | static int |
6423 | load_memoize(UnpicklerObject *self) |
6424 | { |
6425 | PyObject *value; |
6426 | |
6427 | if (Py_SIZE(self->stack) <= self->stack->fence) |
6428 | return Pdata_stack_underflow(self->stack); |
6429 | value = self->stack->data[Py_SIZE(self->stack) - 1]; |
6430 | |
6431 | return _Unpickler_MemoPut(self, self->memo_len, value); |
6432 | } |
6433 | |
6434 | static int |
6435 | do_append(UnpicklerObject *self, Py_ssize_t x) |
6436 | { |
6437 | PyObject *value; |
6438 | PyObject *slice; |
6439 | PyObject *list; |
6440 | PyObject *result; |
6441 | Py_ssize_t len, i; |
6442 | |
6443 | len = Py_SIZE(self->stack); |
6444 | if (x > len || x <= self->stack->fence) |
6445 | return Pdata_stack_underflow(self->stack); |
6446 | if (len == x) /* nothing to do */ |
6447 | return 0; |
6448 | |
6449 | list = self->stack->data[x - 1]; |
6450 | |
6451 | if (PyList_CheckExact(list)) { |
6452 | Py_ssize_t list_len; |
6453 | int ret; |
6454 | |
6455 | slice = Pdata_poplist(self->stack, x); |
6456 | if (!slice) |
6457 | return -1; |
6458 | list_len = PyList_GET_SIZE(list); |
6459 | ret = PyList_SetSlice(list, list_len, list_len, slice); |
6460 | Py_DECREF(slice); |
6461 | return ret; |
6462 | } |
6463 | else { |
6464 | PyObject *extend_func; |
6465 | _Py_IDENTIFIER(extend); |
6466 | |
6467 | if (_PyObject_LookupAttrId(list, &PyId_extend, &extend_func) < 0) { |
6468 | return -1; |
6469 | } |
6470 | if (extend_func != NULL) { |
6471 | slice = Pdata_poplist(self->stack, x); |
6472 | if (!slice) { |
6473 | Py_DECREF(extend_func); |
6474 | return -1; |
6475 | } |
6476 | result = _Pickle_FastCall(extend_func, slice); |
6477 | Py_DECREF(extend_func); |
6478 | if (result == NULL) |
6479 | return -1; |
6480 | Py_DECREF(result); |
6481 | } |
6482 | else { |
6483 | PyObject *append_func; |
6484 | _Py_IDENTIFIER(append); |
6485 | |
6486 | /* Even if the PEP 307 requires extend() and append() methods, |
6487 | fall back on append() if the object has no extend() method |
6488 | for backward compatibility. */ |
6489 | append_func = _PyObject_GetAttrId(list, &PyId_append); |
6490 | if (append_func == NULL) |
6491 | return -1; |
6492 | for (i = x; i < len; i++) { |
6493 | value = self->stack->data[i]; |
6494 | result = _Pickle_FastCall(append_func, value); |
6495 | if (result == NULL) { |
6496 | Pdata_clear(self->stack, i + 1); |
6497 | Py_SET_SIZE(self->stack, x); |
6498 | Py_DECREF(append_func); |
6499 | return -1; |
6500 | } |
6501 | Py_DECREF(result); |
6502 | } |
6503 | Py_SET_SIZE(self->stack, x); |
6504 | Py_DECREF(append_func); |
6505 | } |
6506 | } |
6507 | |
6508 | return 0; |
6509 | } |
6510 | |
6511 | static int |
6512 | load_append(UnpicklerObject *self) |
6513 | { |
6514 | if (Py_SIZE(self->stack) - 1 <= self->stack->fence) |
6515 | return Pdata_stack_underflow(self->stack); |
6516 | return do_append(self, Py_SIZE(self->stack) - 1); |
6517 | } |
6518 | |
6519 | static int |
6520 | load_appends(UnpicklerObject *self) |
6521 | { |
6522 | Py_ssize_t i = marker(self); |
6523 | if (i < 0) |
6524 | return -1; |
6525 | return do_append(self, i); |
6526 | } |
6527 | |
6528 | static int |
6529 | do_setitems(UnpicklerObject *self, Py_ssize_t x) |
6530 | { |
6531 | PyObject *value, *key; |
6532 | PyObject *dict; |
6533 | Py_ssize_t len, i; |
6534 | int status = 0; |
6535 | |
6536 | len = Py_SIZE(self->stack); |
6537 | if (x > len || x <= self->stack->fence) |
6538 | return Pdata_stack_underflow(self->stack); |
6539 | if (len == x) /* nothing to do */ |
6540 | return 0; |
6541 | if ((len - x) % 2 != 0) { |
6542 | PickleState *st = _Pickle_GetGlobalState(); |
6543 | /* Corrupt or hostile pickle -- we never write one like this. */ |
6544 | PyErr_SetString(st->UnpicklingError, |
6545 | "odd number of items for SETITEMS" ); |
6546 | return -1; |
6547 | } |
6548 | |
6549 | /* Here, dict does not actually need to be a PyDict; it could be anything |
6550 | that supports the __setitem__ attribute. */ |
6551 | dict = self->stack->data[x - 1]; |
6552 | |
6553 | for (i = x + 1; i < len; i += 2) { |
6554 | key = self->stack->data[i - 1]; |
6555 | value = self->stack->data[i]; |
6556 | if (PyObject_SetItem(dict, key, value) < 0) { |
6557 | status = -1; |
6558 | break; |
6559 | } |
6560 | } |
6561 | |
6562 | Pdata_clear(self->stack, x); |
6563 | return status; |
6564 | } |
6565 | |
6566 | static int |
6567 | load_setitem(UnpicklerObject *self) |
6568 | { |
6569 | return do_setitems(self, Py_SIZE(self->stack) - 2); |
6570 | } |
6571 | |
6572 | static int |
6573 | load_setitems(UnpicklerObject *self) |
6574 | { |
6575 | Py_ssize_t i = marker(self); |
6576 | if (i < 0) |
6577 | return -1; |
6578 | return do_setitems(self, i); |
6579 | } |
6580 | |
6581 | static int |
6582 | load_additems(UnpicklerObject *self) |
6583 | { |
6584 | PyObject *set; |
6585 | Py_ssize_t mark, len, i; |
6586 | |
6587 | mark = marker(self); |
6588 | if (mark < 0) |
6589 | return -1; |
6590 | len = Py_SIZE(self->stack); |
6591 | if (mark > len || mark <= self->stack->fence) |
6592 | return Pdata_stack_underflow(self->stack); |
6593 | if (len == mark) /* nothing to do */ |
6594 | return 0; |
6595 | |
6596 | set = self->stack->data[mark - 1]; |
6597 | |
6598 | if (PySet_Check(set)) { |
6599 | PyObject *items; |
6600 | int status; |
6601 | |
6602 | items = Pdata_poptuple(self->stack, mark); |
6603 | if (items == NULL) |
6604 | return -1; |
6605 | |
6606 | status = _PySet_Update(set, items); |
6607 | Py_DECREF(items); |
6608 | return status; |
6609 | } |
6610 | else { |
6611 | PyObject *add_func; |
6612 | _Py_IDENTIFIER(add); |
6613 | |
6614 | add_func = _PyObject_GetAttrId(set, &PyId_add); |
6615 | if (add_func == NULL) |
6616 | return -1; |
6617 | for (i = mark; i < len; i++) { |
6618 | PyObject *result; |
6619 | PyObject *item; |
6620 | |
6621 | item = self->stack->data[i]; |
6622 | result = _Pickle_FastCall(add_func, item); |
6623 | if (result == NULL) { |
6624 | Pdata_clear(self->stack, i + 1); |
6625 | Py_SET_SIZE(self->stack, mark); |
6626 | return -1; |
6627 | } |
6628 | Py_DECREF(result); |
6629 | } |
6630 | Py_SET_SIZE(self->stack, mark); |
6631 | } |
6632 | |
6633 | return 0; |
6634 | } |
6635 | |
6636 | static int |
6637 | load_build(UnpicklerObject *self) |
6638 | { |
6639 | PyObject *state, *inst, *slotstate; |
6640 | PyObject *setstate; |
6641 | int status = 0; |
6642 | _Py_IDENTIFIER(__setstate__); |
6643 | |
6644 | /* Stack is ... instance, state. We want to leave instance at |
6645 | * the stack top, possibly mutated via instance.__setstate__(state). |
6646 | */ |
6647 | if (Py_SIZE(self->stack) - 2 < self->stack->fence) |
6648 | return Pdata_stack_underflow(self->stack); |
6649 | |
6650 | PDATA_POP(self->stack, state); |
6651 | if (state == NULL) |
6652 | return -1; |
6653 | |
6654 | inst = self->stack->data[Py_SIZE(self->stack) - 1]; |
6655 | |
6656 | if (_PyObject_LookupAttrId(inst, &PyId___setstate__, &setstate) < 0) { |
6657 | Py_DECREF(state); |
6658 | return -1; |
6659 | } |
6660 | if (setstate != NULL) { |
6661 | PyObject *result; |
6662 | |
6663 | /* The explicit __setstate__ is responsible for everything. */ |
6664 | result = _Pickle_FastCall(setstate, state); |
6665 | Py_DECREF(setstate); |
6666 | if (result == NULL) |
6667 | return -1; |
6668 | Py_DECREF(result); |
6669 | return 0; |
6670 | } |
6671 | |
6672 | /* A default __setstate__. First see whether state embeds a |
6673 | * slot state dict too (a proto 2 addition). |
6674 | */ |
6675 | if (PyTuple_Check(state) && PyTuple_GET_SIZE(state) == 2) { |
6676 | PyObject *tmp = state; |
6677 | |
6678 | state = PyTuple_GET_ITEM(tmp, 0); |
6679 | slotstate = PyTuple_GET_ITEM(tmp, 1); |
6680 | Py_INCREF(state); |
6681 | Py_INCREF(slotstate); |
6682 | Py_DECREF(tmp); |
6683 | } |
6684 | else |
6685 | slotstate = NULL; |
6686 | |
6687 | /* Set inst.__dict__ from the state dict (if any). */ |
6688 | if (state != Py_None) { |
6689 | PyObject *dict; |
6690 | PyObject *d_key, *d_value; |
6691 | Py_ssize_t i; |
6692 | _Py_IDENTIFIER(__dict__); |
6693 | |
6694 | if (!PyDict_Check(state)) { |
6695 | PickleState *st = _Pickle_GetGlobalState(); |
6696 | PyErr_SetString(st->UnpicklingError, "state is not a dictionary" ); |
6697 | goto error; |
6698 | } |
6699 | dict = _PyObject_GetAttrId(inst, &PyId___dict__); |
6700 | if (dict == NULL) |
6701 | goto error; |
6702 | |
6703 | i = 0; |
6704 | while (PyDict_Next(state, &i, &d_key, &d_value)) { |
6705 | /* normally the keys for instance attributes are |
6706 | interned. we should try to do that here. */ |
6707 | Py_INCREF(d_key); |
6708 | if (PyUnicode_CheckExact(d_key)) |
6709 | PyUnicode_InternInPlace(&d_key); |
6710 | if (PyObject_SetItem(dict, d_key, d_value) < 0) { |
6711 | Py_DECREF(d_key); |
6712 | goto error; |
6713 | } |
6714 | Py_DECREF(d_key); |
6715 | } |
6716 | Py_DECREF(dict); |
6717 | } |
6718 | |
6719 | /* Also set instance attributes from the slotstate dict (if any). */ |
6720 | if (slotstate != NULL) { |
6721 | PyObject *d_key, *d_value; |
6722 | Py_ssize_t i; |
6723 | |
6724 | if (!PyDict_Check(slotstate)) { |
6725 | PickleState *st = _Pickle_GetGlobalState(); |
6726 | PyErr_SetString(st->UnpicklingError, |
6727 | "slot state is not a dictionary" ); |
6728 | goto error; |
6729 | } |
6730 | i = 0; |
6731 | while (PyDict_Next(slotstate, &i, &d_key, &d_value)) { |
6732 | if (PyObject_SetAttr(inst, d_key, d_value) < 0) |
6733 | goto error; |
6734 | } |
6735 | } |
6736 | |
6737 | if (0) { |
6738 | error: |
6739 | status = -1; |
6740 | } |
6741 | |
6742 | Py_DECREF(state); |
6743 | Py_XDECREF(slotstate); |
6744 | return status; |
6745 | } |
6746 | |
6747 | static int |
6748 | load_mark(UnpicklerObject *self) |
6749 | { |
6750 | |
6751 | /* Note that we split the (pickle.py) stack into two stacks, an |
6752 | * object stack and a mark stack. Here we push a mark onto the |
6753 | * mark stack. |
6754 | */ |
6755 | |
6756 | if (self->num_marks >= self->marks_size) { |
6757 | size_t alloc = ((size_t)self->num_marks << 1) + 20; |
6758 | Py_ssize_t *marks_new = self->marks; |
6759 | PyMem_RESIZE(marks_new, Py_ssize_t, alloc); |
6760 | if (marks_new == NULL) { |
6761 | PyErr_NoMemory(); |
6762 | return -1; |
6763 | } |
6764 | self->marks = marks_new; |
6765 | self->marks_size = (Py_ssize_t)alloc; |
6766 | } |
6767 | |
6768 | self->stack->mark_set = 1; |
6769 | self->marks[self->num_marks++] = self->stack->fence = Py_SIZE(self->stack); |
6770 | |
6771 | return 0; |
6772 | } |
6773 | |
6774 | static int |
6775 | load_reduce(UnpicklerObject *self) |
6776 | { |
6777 | PyObject *callable = NULL; |
6778 | PyObject *argtup = NULL; |
6779 | PyObject *obj = NULL; |
6780 | |
6781 | PDATA_POP(self->stack, argtup); |
6782 | if (argtup == NULL) |
6783 | return -1; |
6784 | PDATA_POP(self->stack, callable); |
6785 | if (callable) { |
6786 | obj = PyObject_CallObject(callable, argtup); |
6787 | Py_DECREF(callable); |
6788 | } |
6789 | Py_DECREF(argtup); |
6790 | |
6791 | if (obj == NULL) |
6792 | return -1; |
6793 | |
6794 | PDATA_PUSH(self->stack, obj, -1); |
6795 | return 0; |
6796 | } |
6797 | |
6798 | /* Just raises an error if we don't know the protocol specified. PROTO |
6799 | * is the first opcode for protocols >= 2. |
6800 | */ |
6801 | static int |
6802 | load_proto(UnpicklerObject *self) |
6803 | { |
6804 | char *s; |
6805 | int i; |
6806 | |
6807 | if (_Unpickler_Read(self, &s, 1) < 0) |
6808 | return -1; |
6809 | |
6810 | i = (unsigned char)s[0]; |
6811 | if (i <= HIGHEST_PROTOCOL) { |
6812 | self->proto = i; |
6813 | return 0; |
6814 | } |
6815 | |
6816 | PyErr_Format(PyExc_ValueError, "unsupported pickle protocol: %d" , i); |
6817 | return -1; |
6818 | } |
6819 | |
6820 | static int |
6821 | load_frame(UnpicklerObject *self) |
6822 | { |
6823 | char *s; |
6824 | Py_ssize_t frame_len; |
6825 | |
6826 | if (_Unpickler_Read(self, &s, 8) < 0) |
6827 | return -1; |
6828 | |
6829 | frame_len = calc_binsize(s, 8); |
6830 | if (frame_len < 0) { |
6831 | PyErr_Format(PyExc_OverflowError, |
6832 | "FRAME length exceeds system's maximum of %zd bytes" , |
6833 | PY_SSIZE_T_MAX); |
6834 | return -1; |
6835 | } |
6836 | |
6837 | if (_Unpickler_Read(self, &s, frame_len) < 0) |
6838 | return -1; |
6839 | |
6840 | /* Rewind to start of frame */ |
6841 | self->next_read_idx -= frame_len; |
6842 | return 0; |
6843 | } |
6844 | |
6845 | static PyObject * |
6846 | load(UnpicklerObject *self) |
6847 | { |
6848 | PyObject *value = NULL; |
6849 | char *s = NULL; |
6850 | |
6851 | self->num_marks = 0; |
6852 | self->stack->mark_set = 0; |
6853 | self->stack->fence = 0; |
6854 | self->proto = 0; |
6855 | if (Py_SIZE(self->stack)) |
6856 | Pdata_clear(self->stack, 0); |
6857 | |
6858 | /* Convenient macros for the dispatch while-switch loop just below. */ |
6859 | #define OP(opcode, load_func) \ |
6860 | case opcode: if (load_func(self) < 0) break; continue; |
6861 | |
6862 | #define OP_ARG(opcode, load_func, arg) \ |
6863 | case opcode: if (load_func(self, (arg)) < 0) break; continue; |
6864 | |
6865 | while (1) { |
6866 | if (_Unpickler_Read(self, &s, 1) < 0) { |
6867 | PickleState *st = _Pickle_GetGlobalState(); |
6868 | if (PyErr_ExceptionMatches(st->UnpicklingError)) { |
6869 | PyErr_Format(PyExc_EOFError, "Ran out of input" ); |
6870 | } |
6871 | return NULL; |
6872 | } |
6873 | |
6874 | switch ((enum opcode)s[0]) { |
6875 | OP(NONE, load_none) |
6876 | OP(BININT, load_binint) |
6877 | OP(BININT1, load_binint1) |
6878 | OP(BININT2, load_binint2) |
6879 | OP(INT, load_int) |
6880 | OP(LONG, load_long) |
6881 | OP_ARG(LONG1, load_counted_long, 1) |
6882 | OP_ARG(LONG4, load_counted_long, 4) |
6883 | OP(FLOAT, load_float) |
6884 | OP(BINFLOAT, load_binfloat) |
6885 | OP_ARG(SHORT_BINBYTES, load_counted_binbytes, 1) |
6886 | OP_ARG(BINBYTES, load_counted_binbytes, 4) |
6887 | OP_ARG(BINBYTES8, load_counted_binbytes, 8) |
6888 | OP(BYTEARRAY8, load_counted_bytearray) |
6889 | OP(NEXT_BUFFER, load_next_buffer) |
6890 | OP(READONLY_BUFFER, load_readonly_buffer) |
6891 | OP_ARG(SHORT_BINSTRING, load_counted_binstring, 1) |
6892 | OP_ARG(BINSTRING, load_counted_binstring, 4) |
6893 | OP(STRING, load_string) |
6894 | OP(UNICODE, load_unicode) |
6895 | OP_ARG(SHORT_BINUNICODE, load_counted_binunicode, 1) |
6896 | OP_ARG(BINUNICODE, load_counted_binunicode, 4) |
6897 | OP_ARG(BINUNICODE8, load_counted_binunicode, 8) |
6898 | OP_ARG(EMPTY_TUPLE, load_counted_tuple, 0) |
6899 | OP_ARG(TUPLE1, load_counted_tuple, 1) |
6900 | OP_ARG(TUPLE2, load_counted_tuple, 2) |
6901 | OP_ARG(TUPLE3, load_counted_tuple, 3) |
6902 | OP(TUPLE, load_tuple) |
6903 | OP(EMPTY_LIST, load_empty_list) |
6904 | OP(LIST, load_list) |
6905 | OP(EMPTY_DICT, load_empty_dict) |
6906 | OP(DICT, load_dict) |
6907 | OP(EMPTY_SET, load_empty_set) |
6908 | OP(ADDITEMS, load_additems) |
6909 | OP(FROZENSET, load_frozenset) |
6910 | OP(OBJ, load_obj) |
6911 | OP(INST, load_inst) |
6912 | OP_ARG(NEWOBJ, load_newobj, 0) |
6913 | OP_ARG(NEWOBJ_EX, load_newobj, 1) |
6914 | OP(GLOBAL, load_global) |
6915 | OP(STACK_GLOBAL, load_stack_global) |
6916 | OP(APPEND, load_append) |
6917 | OP(APPENDS, load_appends) |
6918 | OP(BUILD, load_build) |
6919 | OP(DUP, load_dup) |
6920 | OP(BINGET, load_binget) |
6921 | OP(LONG_BINGET, load_long_binget) |
6922 | OP(GET, load_get) |
6923 | OP(MARK, load_mark) |
6924 | OP(BINPUT, load_binput) |
6925 | OP(LONG_BINPUT, load_long_binput) |
6926 | OP(PUT, load_put) |
6927 | OP(MEMOIZE, load_memoize) |
6928 | OP(POP, load_pop) |
6929 | OP(POP_MARK, load_pop_mark) |
6930 | OP(SETITEM, load_setitem) |
6931 | OP(SETITEMS, load_setitems) |
6932 | OP(PERSID, load_persid) |
6933 | OP(BINPERSID, load_binpersid) |
6934 | OP(REDUCE, load_reduce) |
6935 | OP(PROTO, load_proto) |
6936 | OP(FRAME, load_frame) |
6937 | OP_ARG(EXT1, load_extension, 1) |
6938 | OP_ARG(EXT2, load_extension, 2) |
6939 | OP_ARG(EXT4, load_extension, 4) |
6940 | OP_ARG(NEWTRUE, load_bool, Py_True) |
6941 | OP_ARG(NEWFALSE, load_bool, Py_False) |
6942 | |
6943 | case STOP: |
6944 | break; |
6945 | |
6946 | default: |
6947 | { |
6948 | PickleState *st = _Pickle_GetGlobalState(); |
6949 | unsigned char c = (unsigned char) *s; |
6950 | if (0x20 <= c && c <= 0x7e && c != '\'' && c != '\\') { |
6951 | PyErr_Format(st->UnpicklingError, |
6952 | "invalid load key, '%c'." , c); |
6953 | } |
6954 | else { |
6955 | PyErr_Format(st->UnpicklingError, |
6956 | "invalid load key, '\\x%02x'." , c); |
6957 | } |
6958 | return NULL; |
6959 | } |
6960 | } |
6961 | |
6962 | break; /* and we are done! */ |
6963 | } |
6964 | |
6965 | if (PyErr_Occurred()) { |
6966 | return NULL; |
6967 | } |
6968 | |
6969 | if (_Unpickler_SkipConsumed(self) < 0) |
6970 | return NULL; |
6971 | |
6972 | PDATA_POP(self->stack, value); |
6973 | return value; |
6974 | } |
6975 | |
6976 | /*[clinic input] |
6977 | |
6978 | _pickle.Unpickler.load |
6979 | |
6980 | Load a pickle. |
6981 | |
6982 | Read a pickled object representation from the open file object given |
6983 | in the constructor, and return the reconstituted object hierarchy |
6984 | specified therein. |
6985 | [clinic start generated code]*/ |
6986 | |
6987 | static PyObject * |
6988 | _pickle_Unpickler_load_impl(UnpicklerObject *self) |
6989 | /*[clinic end generated code: output=fdcc488aad675b14 input=acbb91a42fa9b7b9]*/ |
6990 | { |
6991 | UnpicklerObject *unpickler = (UnpicklerObject*)self; |
6992 | |
6993 | /* Check whether the Unpickler was initialized correctly. This prevents |
6994 | segfaulting if a subclass overridden __init__ with a function that does |
6995 | not call Unpickler.__init__(). Here, we simply ensure that self->read |
6996 | is not NULL. */ |
6997 | if (unpickler->read == NULL) { |
6998 | PickleState *st = _Pickle_GetGlobalState(); |
6999 | PyErr_Format(st->UnpicklingError, |
7000 | "Unpickler.__init__() was not called by %s.__init__()" , |
7001 | Py_TYPE(unpickler)->tp_name); |
7002 | return NULL; |
7003 | } |
7004 | |
7005 | return load(unpickler); |
7006 | } |
7007 | |
7008 | /* The name of find_class() is misleading. In newer pickle protocols, this |
7009 | function is used for loading any global (i.e., functions), not just |
7010 | classes. The name is kept only for backward compatibility. */ |
7011 | |
7012 | /*[clinic input] |
7013 | |
7014 | _pickle.Unpickler.find_class |
7015 | |
7016 | module_name: object |
7017 | global_name: object |
7018 | / |
7019 | |
7020 | Return an object from a specified module. |
7021 | |
7022 | If necessary, the module will be imported. Subclasses may override |
7023 | this method (e.g. to restrict unpickling of arbitrary classes and |
7024 | functions). |
7025 | |
7026 | This method is called whenever a class or a function object is |
7027 | needed. Both arguments passed are str objects. |
7028 | [clinic start generated code]*/ |
7029 | |
7030 | static PyObject * |
7031 | _pickle_Unpickler_find_class_impl(UnpicklerObject *self, |
7032 | PyObject *module_name, |
7033 | PyObject *global_name) |
7034 | /*[clinic end generated code: output=becc08d7f9ed41e3 input=e2e6a865de093ef4]*/ |
7035 | { |
7036 | PyObject *global; |
7037 | PyObject *module; |
7038 | |
7039 | if (PySys_Audit("pickle.find_class" , "OO" , |
7040 | module_name, global_name) < 0) { |
7041 | return NULL; |
7042 | } |
7043 | |
7044 | /* Try to map the old names used in Python 2.x to the new ones used in |
7045 | Python 3.x. We do this only with old pickle protocols and when the |
7046 | user has not disabled the feature. */ |
7047 | if (self->proto < 3 && self->fix_imports) { |
7048 | PyObject *key; |
7049 | PyObject *item; |
7050 | PickleState *st = _Pickle_GetGlobalState(); |
7051 | |
7052 | /* Check if the global (i.e., a function or a class) was renamed |
7053 | or moved to another module. */ |
7054 | key = PyTuple_Pack(2, module_name, global_name); |
7055 | if (key == NULL) |
7056 | return NULL; |
7057 | item = PyDict_GetItemWithError(st->name_mapping_2to3, key); |
7058 | Py_DECREF(key); |
7059 | if (item) { |
7060 | if (!PyTuple_Check(item) || PyTuple_GET_SIZE(item) != 2) { |
7061 | PyErr_Format(PyExc_RuntimeError, |
7062 | "_compat_pickle.NAME_MAPPING values should be " |
7063 | "2-tuples, not %.200s" , Py_TYPE(item)->tp_name); |
7064 | return NULL; |
7065 | } |
7066 | module_name = PyTuple_GET_ITEM(item, 0); |
7067 | global_name = PyTuple_GET_ITEM(item, 1); |
7068 | if (!PyUnicode_Check(module_name) || |
7069 | !PyUnicode_Check(global_name)) { |
7070 | PyErr_Format(PyExc_RuntimeError, |
7071 | "_compat_pickle.NAME_MAPPING values should be " |
7072 | "pairs of str, not (%.200s, %.200s)" , |
7073 | Py_TYPE(module_name)->tp_name, |
7074 | Py_TYPE(global_name)->tp_name); |
7075 | return NULL; |
7076 | } |
7077 | } |
7078 | else if (PyErr_Occurred()) { |
7079 | return NULL; |
7080 | } |
7081 | else { |
7082 | /* Check if the module was renamed. */ |
7083 | item = PyDict_GetItemWithError(st->import_mapping_2to3, module_name); |
7084 | if (item) { |
7085 | if (!PyUnicode_Check(item)) { |
7086 | PyErr_Format(PyExc_RuntimeError, |
7087 | "_compat_pickle.IMPORT_MAPPING values should be " |
7088 | "strings, not %.200s" , Py_TYPE(item)->tp_name); |
7089 | return NULL; |
7090 | } |
7091 | module_name = item; |
7092 | } |
7093 | else if (PyErr_Occurred()) { |
7094 | return NULL; |
7095 | } |
7096 | } |
7097 | } |
7098 | |
7099 | /* |
7100 | * we don't use PyImport_GetModule here, because it can return partially- |
7101 | * initialised modules, which then cause the getattribute to fail. |
7102 | */ |
7103 | module = PyImport_Import(module_name); |
7104 | if (module == NULL) { |
7105 | return NULL; |
7106 | } |
7107 | global = getattribute(module, global_name, self->proto >= 4); |
7108 | Py_DECREF(module); |
7109 | return global; |
7110 | } |
7111 | |
7112 | /*[clinic input] |
7113 | |
7114 | _pickle.Unpickler.__sizeof__ -> Py_ssize_t |
7115 | |
7116 | Returns size in memory, in bytes. |
7117 | [clinic start generated code]*/ |
7118 | |
7119 | static Py_ssize_t |
7120 | _pickle_Unpickler___sizeof___impl(UnpicklerObject *self) |
7121 | /*[clinic end generated code: output=119d9d03ad4c7651 input=13333471fdeedf5e]*/ |
7122 | { |
7123 | Py_ssize_t res; |
7124 | |
7125 | res = _PyObject_SIZE(Py_TYPE(self)); |
7126 | if (self->memo != NULL) |
7127 | res += self->memo_size * sizeof(PyObject *); |
7128 | if (self->marks != NULL) |
7129 | res += self->marks_size * sizeof(Py_ssize_t); |
7130 | if (self->input_line != NULL) |
7131 | res += strlen(self->input_line) + 1; |
7132 | if (self->encoding != NULL) |
7133 | res += strlen(self->encoding) + 1; |
7134 | if (self->errors != NULL) |
7135 | res += strlen(self->errors) + 1; |
7136 | return res; |
7137 | } |
7138 | |
7139 | static struct PyMethodDef Unpickler_methods[] = { |
7140 | _PICKLE_UNPICKLER_LOAD_METHODDEF |
7141 | _PICKLE_UNPICKLER_FIND_CLASS_METHODDEF |
7142 | _PICKLE_UNPICKLER___SIZEOF___METHODDEF |
7143 | {NULL, NULL} /* sentinel */ |
7144 | }; |
7145 | |
7146 | static void |
7147 | Unpickler_dealloc(UnpicklerObject *self) |
7148 | { |
7149 | PyObject_GC_UnTrack((PyObject *)self); |
7150 | Py_XDECREF(self->readline); |
7151 | Py_XDECREF(self->readinto); |
7152 | Py_XDECREF(self->read); |
7153 | Py_XDECREF(self->peek); |
7154 | Py_XDECREF(self->stack); |
7155 | Py_XDECREF(self->pers_func); |
7156 | Py_XDECREF(self->buffers); |
7157 | if (self->buffer.buf != NULL) { |
7158 | PyBuffer_Release(&self->buffer); |
7159 | self->buffer.buf = NULL; |
7160 | } |
7161 | |
7162 | _Unpickler_MemoCleanup(self); |
7163 | PyMem_Free(self->marks); |
7164 | PyMem_Free(self->input_line); |
7165 | PyMem_Free(self->encoding); |
7166 | PyMem_Free(self->errors); |
7167 | |
7168 | Py_TYPE(self)->tp_free((PyObject *)self); |
7169 | } |
7170 | |
7171 | static int |
7172 | Unpickler_traverse(UnpicklerObject *self, visitproc visit, void *arg) |
7173 | { |
7174 | Py_VISIT(self->readline); |
7175 | Py_VISIT(self->readinto); |
7176 | Py_VISIT(self->read); |
7177 | Py_VISIT(self->peek); |
7178 | Py_VISIT(self->stack); |
7179 | Py_VISIT(self->pers_func); |
7180 | Py_VISIT(self->buffers); |
7181 | return 0; |
7182 | } |
7183 | |
7184 | static int |
7185 | Unpickler_clear(UnpicklerObject *self) |
7186 | { |
7187 | Py_CLEAR(self->readline); |
7188 | Py_CLEAR(self->readinto); |
7189 | Py_CLEAR(self->read); |
7190 | Py_CLEAR(self->peek); |
7191 | Py_CLEAR(self->stack); |
7192 | Py_CLEAR(self->pers_func); |
7193 | Py_CLEAR(self->buffers); |
7194 | if (self->buffer.buf != NULL) { |
7195 | PyBuffer_Release(&self->buffer); |
7196 | self->buffer.buf = NULL; |
7197 | } |
7198 | |
7199 | _Unpickler_MemoCleanup(self); |
7200 | PyMem_Free(self->marks); |
7201 | self->marks = NULL; |
7202 | PyMem_Free(self->input_line); |
7203 | self->input_line = NULL; |
7204 | PyMem_Free(self->encoding); |
7205 | self->encoding = NULL; |
7206 | PyMem_Free(self->errors); |
7207 | self->errors = NULL; |
7208 | |
7209 | return 0; |
7210 | } |
7211 | |
7212 | /*[clinic input] |
7213 | |
7214 | _pickle.Unpickler.__init__ |
7215 | |
7216 | file: object |
7217 | * |
7218 | fix_imports: bool = True |
7219 | encoding: str = 'ASCII' |
7220 | errors: str = 'strict' |
7221 | buffers: object(c_default="NULL") = () |
7222 | |
7223 | This takes a binary file for reading a pickle data stream. |
7224 | |
7225 | The protocol version of the pickle is detected automatically, so no |
7226 | protocol argument is needed. Bytes past the pickled object's |
7227 | representation are ignored. |
7228 | |
7229 | The argument *file* must have two methods, a read() method that takes |
7230 | an integer argument, and a readline() method that requires no |
7231 | arguments. Both methods should return bytes. Thus *file* can be a |
7232 | binary file object opened for reading, an io.BytesIO object, or any |
7233 | other custom object that meets this interface. |
7234 | |
7235 | Optional keyword arguments are *fix_imports*, *encoding* and *errors*, |
7236 | which are used to control compatibility support for pickle stream |
7237 | generated by Python 2. If *fix_imports* is True, pickle will try to |
7238 | map the old Python 2 names to the new names used in Python 3. The |
7239 | *encoding* and *errors* tell pickle how to decode 8-bit string |
7240 | instances pickled by Python 2; these default to 'ASCII' and 'strict', |
7241 | respectively. The *encoding* can be 'bytes' to read these 8-bit |
7242 | string instances as bytes objects. |
7243 | [clinic start generated code]*/ |
7244 | |
7245 | static int |
7246 | _pickle_Unpickler___init___impl(UnpicklerObject *self, PyObject *file, |
7247 | int fix_imports, const char *encoding, |
7248 | const char *errors, PyObject *buffers) |
7249 | /*[clinic end generated code: output=09f0192649ea3f85 input=ca4c1faea9553121]*/ |
7250 | { |
7251 | _Py_IDENTIFIER(persistent_load); |
7252 | |
7253 | /* In case of multiple __init__() calls, clear previous content. */ |
7254 | if (self->read != NULL) |
7255 | (void)Unpickler_clear(self); |
7256 | |
7257 | if (_Unpickler_SetInputStream(self, file) < 0) |
7258 | return -1; |
7259 | |
7260 | if (_Unpickler_SetInputEncoding(self, encoding, errors) < 0) |
7261 | return -1; |
7262 | |
7263 | if (_Unpickler_SetBuffers(self, buffers) < 0) |
7264 | return -1; |
7265 | |
7266 | self->fix_imports = fix_imports; |
7267 | |
7268 | if (init_method_ref((PyObject *)self, &PyId_persistent_load, |
7269 | &self->pers_func, &self->pers_func_self) < 0) |
7270 | { |
7271 | return -1; |
7272 | } |
7273 | |
7274 | self->stack = (Pdata *)Pdata_New(); |
7275 | if (self->stack == NULL) |
7276 | return -1; |
7277 | |
7278 | self->memo_size = 32; |
7279 | self->memo = _Unpickler_NewMemo(self->memo_size); |
7280 | if (self->memo == NULL) |
7281 | return -1; |
7282 | |
7283 | self->proto = 0; |
7284 | |
7285 | return 0; |
7286 | } |
7287 | |
7288 | |
7289 | /* Define a proxy object for the Unpickler's internal memo object. This is to |
7290 | * avoid breaking code like: |
7291 | * unpickler.memo.clear() |
7292 | * and |
7293 | * unpickler.memo = saved_memo |
7294 | * Is this a good idea? Not really, but we don't want to break code that uses |
7295 | * it. Note that we don't implement the entire mapping API here. This is |
7296 | * intentional, as these should be treated as black-box implementation details. |
7297 | * |
7298 | * We do, however, have to implement pickling/unpickling support because of |
7299 | * real-world code like cvs2svn. |
7300 | */ |
7301 | |
7302 | /*[clinic input] |
7303 | _pickle.UnpicklerMemoProxy.clear |
7304 | |
7305 | Remove all items from memo. |
7306 | [clinic start generated code]*/ |
7307 | |
7308 | static PyObject * |
7309 | _pickle_UnpicklerMemoProxy_clear_impl(UnpicklerMemoProxyObject *self) |
7310 | /*[clinic end generated code: output=d20cd43f4ba1fb1f input=b1df7c52e7afd9bd]*/ |
7311 | { |
7312 | _Unpickler_MemoCleanup(self->unpickler); |
7313 | self->unpickler->memo = _Unpickler_NewMemo(self->unpickler->memo_size); |
7314 | if (self->unpickler->memo == NULL) |
7315 | return NULL; |
7316 | Py_RETURN_NONE; |
7317 | } |
7318 | |
7319 | /*[clinic input] |
7320 | _pickle.UnpicklerMemoProxy.copy |
7321 | |
7322 | Copy the memo to a new object. |
7323 | [clinic start generated code]*/ |
7324 | |
7325 | static PyObject * |
7326 | _pickle_UnpicklerMemoProxy_copy_impl(UnpicklerMemoProxyObject *self) |
7327 | /*[clinic end generated code: output=e12af7e9bc1e4c77 input=97769247ce032c1d]*/ |
7328 | { |
7329 | size_t i; |
7330 | PyObject *new_memo = PyDict_New(); |
7331 | if (new_memo == NULL) |
7332 | return NULL; |
7333 | |
7334 | for (i = 0; i < self->unpickler->memo_size; i++) { |
7335 | int status; |
7336 | PyObject *key, *value; |
7337 | |
7338 | value = self->unpickler->memo[i]; |
7339 | if (value == NULL) |
7340 | continue; |
7341 | |
7342 | key = PyLong_FromSsize_t(i); |
7343 | if (key == NULL) |
7344 | goto error; |
7345 | status = PyDict_SetItem(new_memo, key, value); |
7346 | Py_DECREF(key); |
7347 | if (status < 0) |
7348 | goto error; |
7349 | } |
7350 | return new_memo; |
7351 | |
7352 | error: |
7353 | Py_DECREF(new_memo); |
7354 | return NULL; |
7355 | } |
7356 | |
7357 | /*[clinic input] |
7358 | _pickle.UnpicklerMemoProxy.__reduce__ |
7359 | |
7360 | Implement pickling support. |
7361 | [clinic start generated code]*/ |
7362 | |
7363 | static PyObject * |
7364 | _pickle_UnpicklerMemoProxy___reduce___impl(UnpicklerMemoProxyObject *self) |
7365 | /*[clinic end generated code: output=6da34ac048d94cca input=6920862413407199]*/ |
7366 | { |
7367 | PyObject *reduce_value; |
7368 | PyObject *constructor_args; |
7369 | PyObject *contents = _pickle_UnpicklerMemoProxy_copy_impl(self); |
7370 | if (contents == NULL) |
7371 | return NULL; |
7372 | |
7373 | reduce_value = PyTuple_New(2); |
7374 | if (reduce_value == NULL) { |
7375 | Py_DECREF(contents); |
7376 | return NULL; |
7377 | } |
7378 | constructor_args = PyTuple_New(1); |
7379 | if (constructor_args == NULL) { |
7380 | Py_DECREF(contents); |
7381 | Py_DECREF(reduce_value); |
7382 | return NULL; |
7383 | } |
7384 | PyTuple_SET_ITEM(constructor_args, 0, contents); |
7385 | Py_INCREF((PyObject *)&PyDict_Type); |
7386 | PyTuple_SET_ITEM(reduce_value, 0, (PyObject *)&PyDict_Type); |
7387 | PyTuple_SET_ITEM(reduce_value, 1, constructor_args); |
7388 | return reduce_value; |
7389 | } |
7390 | |
7391 | static PyMethodDef unpicklerproxy_methods[] = { |
7392 | _PICKLE_UNPICKLERMEMOPROXY_CLEAR_METHODDEF |
7393 | _PICKLE_UNPICKLERMEMOPROXY_COPY_METHODDEF |
7394 | _PICKLE_UNPICKLERMEMOPROXY___REDUCE___METHODDEF |
7395 | {NULL, NULL} /* sentinel */ |
7396 | }; |
7397 | |
7398 | static void |
7399 | UnpicklerMemoProxy_dealloc(UnpicklerMemoProxyObject *self) |
7400 | { |
7401 | PyObject_GC_UnTrack(self); |
7402 | Py_XDECREF(self->unpickler); |
7403 | PyObject_GC_Del((PyObject *)self); |
7404 | } |
7405 | |
7406 | static int |
7407 | UnpicklerMemoProxy_traverse(UnpicklerMemoProxyObject *self, |
7408 | visitproc visit, void *arg) |
7409 | { |
7410 | Py_VISIT(self->unpickler); |
7411 | return 0; |
7412 | } |
7413 | |
7414 | static int |
7415 | UnpicklerMemoProxy_clear(UnpicklerMemoProxyObject *self) |
7416 | { |
7417 | Py_CLEAR(self->unpickler); |
7418 | return 0; |
7419 | } |
7420 | |
7421 | static PyTypeObject UnpicklerMemoProxyType = { |
7422 | PyVarObject_HEAD_INIT(NULL, 0) |
7423 | "_pickle.UnpicklerMemoProxy" , /*tp_name*/ |
7424 | sizeof(UnpicklerMemoProxyObject), /*tp_basicsize*/ |
7425 | 0, |
7426 | (destructor)UnpicklerMemoProxy_dealloc, /* tp_dealloc */ |
7427 | 0, /* tp_vectorcall_offset */ |
7428 | 0, /* tp_getattr */ |
7429 | 0, /* tp_setattr */ |
7430 | 0, /* tp_as_async */ |
7431 | 0, /* tp_repr */ |
7432 | 0, /* tp_as_number */ |
7433 | 0, /* tp_as_sequence */ |
7434 | 0, /* tp_as_mapping */ |
7435 | PyObject_HashNotImplemented, /* tp_hash */ |
7436 | 0, /* tp_call */ |
7437 | 0, /* tp_str */ |
7438 | PyObject_GenericGetAttr, /* tp_getattro */ |
7439 | PyObject_GenericSetAttr, /* tp_setattro */ |
7440 | 0, /* tp_as_buffer */ |
7441 | Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE | Py_TPFLAGS_HAVE_GC, |
7442 | 0, /* tp_doc */ |
7443 | (traverseproc)UnpicklerMemoProxy_traverse, /* tp_traverse */ |
7444 | (inquiry)UnpicklerMemoProxy_clear, /* tp_clear */ |
7445 | 0, /* tp_richcompare */ |
7446 | 0, /* tp_weaklistoffset */ |
7447 | 0, /* tp_iter */ |
7448 | 0, /* tp_iternext */ |
7449 | unpicklerproxy_methods, /* tp_methods */ |
7450 | }; |
7451 | |
7452 | static PyObject * |
7453 | UnpicklerMemoProxy_New(UnpicklerObject *unpickler) |
7454 | { |
7455 | UnpicklerMemoProxyObject *self; |
7456 | |
7457 | self = PyObject_GC_New(UnpicklerMemoProxyObject, |
7458 | &UnpicklerMemoProxyType); |
7459 | if (self == NULL) |
7460 | return NULL; |
7461 | Py_INCREF(unpickler); |
7462 | self->unpickler = unpickler; |
7463 | PyObject_GC_Track(self); |
7464 | return (PyObject *)self; |
7465 | } |
7466 | |
7467 | /*****************************************************************************/ |
7468 | |
7469 | |
7470 | static PyObject * |
7471 | Unpickler_get_memo(UnpicklerObject *self, void *Py_UNUSED(ignored)) |
7472 | { |
7473 | return UnpicklerMemoProxy_New(self); |
7474 | } |
7475 | |
7476 | static int |
7477 | Unpickler_set_memo(UnpicklerObject *self, PyObject *obj, void *Py_UNUSED(ignored)) |
7478 | { |
7479 | PyObject **new_memo; |
7480 | size_t new_memo_size = 0; |
7481 | |
7482 | if (obj == NULL) { |
7483 | PyErr_SetString(PyExc_TypeError, |
7484 | "attribute deletion is not supported" ); |
7485 | return -1; |
7486 | } |
7487 | |
7488 | if (Py_IS_TYPE(obj, &UnpicklerMemoProxyType)) { |
7489 | UnpicklerObject *unpickler = |
7490 | ((UnpicklerMemoProxyObject *)obj)->unpickler; |
7491 | |
7492 | new_memo_size = unpickler->memo_size; |
7493 | new_memo = _Unpickler_NewMemo(new_memo_size); |
7494 | if (new_memo == NULL) |
7495 | return -1; |
7496 | |
7497 | for (size_t i = 0; i < new_memo_size; i++) { |
7498 | Py_XINCREF(unpickler->memo[i]); |
7499 | new_memo[i] = unpickler->memo[i]; |
7500 | } |
7501 | } |
7502 | else if (PyDict_Check(obj)) { |
7503 | Py_ssize_t i = 0; |
7504 | PyObject *key, *value; |
7505 | |
7506 | new_memo_size = PyDict_GET_SIZE(obj); |
7507 | new_memo = _Unpickler_NewMemo(new_memo_size); |
7508 | if (new_memo == NULL) |
7509 | return -1; |
7510 | |
7511 | while (PyDict_Next(obj, &i, &key, &value)) { |
7512 | Py_ssize_t idx; |
7513 | if (!PyLong_Check(key)) { |
7514 | PyErr_SetString(PyExc_TypeError, |
7515 | "memo key must be integers" ); |
7516 | goto error; |
7517 | } |
7518 | idx = PyLong_AsSsize_t(key); |
7519 | if (idx == -1 && PyErr_Occurred()) |
7520 | goto error; |
7521 | if (idx < 0) { |
7522 | PyErr_SetString(PyExc_ValueError, |
7523 | "memo key must be positive integers." ); |
7524 | goto error; |
7525 | } |
7526 | if (_Unpickler_MemoPut(self, idx, value) < 0) |
7527 | goto error; |
7528 | } |
7529 | } |
7530 | else { |
7531 | PyErr_Format(PyExc_TypeError, |
7532 | "'memo' attribute must be an UnpicklerMemoProxy object " |
7533 | "or dict, not %.200s" , Py_TYPE(obj)->tp_name); |
7534 | return -1; |
7535 | } |
7536 | |
7537 | _Unpickler_MemoCleanup(self); |
7538 | self->memo_size = new_memo_size; |
7539 | self->memo = new_memo; |
7540 | |
7541 | return 0; |
7542 | |
7543 | error: |
7544 | if (new_memo_size) { |
7545 | for (size_t i = new_memo_size - 1; i != SIZE_MAX; i--) { |
7546 | Py_XDECREF(new_memo[i]); |
7547 | } |
7548 | PyMem_Free(new_memo); |
7549 | } |
7550 | return -1; |
7551 | } |
7552 | |
7553 | static PyObject * |
7554 | Unpickler_get_persload(UnpicklerObject *self, void *Py_UNUSED(ignored)) |
7555 | { |
7556 | if (self->pers_func == NULL) { |
7557 | PyErr_SetString(PyExc_AttributeError, "persistent_load" ); |
7558 | return NULL; |
7559 | } |
7560 | return reconstruct_method(self->pers_func, self->pers_func_self); |
7561 | } |
7562 | |
7563 | static int |
7564 | Unpickler_set_persload(UnpicklerObject *self, PyObject *value, void *Py_UNUSED(ignored)) |
7565 | { |
7566 | if (value == NULL) { |
7567 | PyErr_SetString(PyExc_TypeError, |
7568 | "attribute deletion is not supported" ); |
7569 | return -1; |
7570 | } |
7571 | if (!PyCallable_Check(value)) { |
7572 | PyErr_SetString(PyExc_TypeError, |
7573 | "persistent_load must be a callable taking " |
7574 | "one argument" ); |
7575 | return -1; |
7576 | } |
7577 | |
7578 | self->pers_func_self = NULL; |
7579 | Py_INCREF(value); |
7580 | Py_XSETREF(self->pers_func, value); |
7581 | |
7582 | return 0; |
7583 | } |
7584 | |
7585 | static PyGetSetDef Unpickler_getsets[] = { |
7586 | {"memo" , (getter)Unpickler_get_memo, (setter)Unpickler_set_memo}, |
7587 | {"persistent_load" , (getter)Unpickler_get_persload, |
7588 | (setter)Unpickler_set_persload}, |
7589 | {NULL} |
7590 | }; |
7591 | |
7592 | static PyTypeObject Unpickler_Type = { |
7593 | PyVarObject_HEAD_INIT(NULL, 0) |
7594 | "_pickle.Unpickler" , /*tp_name*/ |
7595 | sizeof(UnpicklerObject), /*tp_basicsize*/ |
7596 | 0, /*tp_itemsize*/ |
7597 | (destructor)Unpickler_dealloc, /*tp_dealloc*/ |
7598 | 0, /*tp_vectorcall_offset*/ |
7599 | 0, /*tp_getattr*/ |
7600 | 0, /*tp_setattr*/ |
7601 | 0, /*tp_as_async*/ |
7602 | 0, /*tp_repr*/ |
7603 | 0, /*tp_as_number*/ |
7604 | 0, /*tp_as_sequence*/ |
7605 | 0, /*tp_as_mapping*/ |
7606 | 0, /*tp_hash*/ |
7607 | 0, /*tp_call*/ |
7608 | 0, /*tp_str*/ |
7609 | 0, /*tp_getattro*/ |
7610 | 0, /*tp_setattro*/ |
7611 | 0, /*tp_as_buffer*/ |
7612 | Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE | Py_TPFLAGS_HAVE_GC, |
7613 | _pickle_Unpickler___init____doc__, /*tp_doc*/ |
7614 | (traverseproc)Unpickler_traverse, /*tp_traverse*/ |
7615 | (inquiry)Unpickler_clear, /*tp_clear*/ |
7616 | 0, /*tp_richcompare*/ |
7617 | 0, /*tp_weaklistoffset*/ |
7618 | 0, /*tp_iter*/ |
7619 | 0, /*tp_iternext*/ |
7620 | Unpickler_methods, /*tp_methods*/ |
7621 | 0, /*tp_members*/ |
7622 | Unpickler_getsets, /*tp_getset*/ |
7623 | 0, /*tp_base*/ |
7624 | 0, /*tp_dict*/ |
7625 | 0, /*tp_descr_get*/ |
7626 | 0, /*tp_descr_set*/ |
7627 | 0, /*tp_dictoffset*/ |
7628 | _pickle_Unpickler___init__, /*tp_init*/ |
7629 | PyType_GenericAlloc, /*tp_alloc*/ |
7630 | PyType_GenericNew, /*tp_new*/ |
7631 | PyObject_GC_Del, /*tp_free*/ |
7632 | 0, /*tp_is_gc*/ |
7633 | }; |
7634 | |
7635 | /*[clinic input] |
7636 | |
7637 | _pickle.dump |
7638 | |
7639 | obj: object |
7640 | file: object |
7641 | protocol: object = None |
7642 | * |
7643 | fix_imports: bool = True |
7644 | buffer_callback: object = None |
7645 | |
7646 | Write a pickled representation of obj to the open file object file. |
7647 | |
7648 | This is equivalent to ``Pickler(file, protocol).dump(obj)``, but may |
7649 | be more efficient. |
7650 | |
7651 | The optional *protocol* argument tells the pickler to use the given |
7652 | protocol; supported protocols are 0, 1, 2, 3, 4 and 5. The default |
7653 | protocol is 4. It was introduced in Python 3.4, and is incompatible |
7654 | with previous versions. |
7655 | |
7656 | Specifying a negative protocol version selects the highest protocol |
7657 | version supported. The higher the protocol used, the more recent the |
7658 | version of Python needed to read the pickle produced. |
7659 | |
7660 | The *file* argument must have a write() method that accepts a single |
7661 | bytes argument. It can thus be a file object opened for binary |
7662 | writing, an io.BytesIO instance, or any other custom object that meets |
7663 | this interface. |
7664 | |
7665 | If *fix_imports* is True and protocol is less than 3, pickle will try |
7666 | to map the new Python 3 names to the old module names used in Python |
7667 | 2, so that the pickle data stream is readable with Python 2. |
7668 | |
7669 | If *buffer_callback* is None (the default), buffer views are serialized |
7670 | into *file* as part of the pickle stream. It is an error if |
7671 | *buffer_callback* is not None and *protocol* is None or smaller than 5. |
7672 | |
7673 | [clinic start generated code]*/ |
7674 | |
7675 | static PyObject * |
7676 | _pickle_dump_impl(PyObject *module, PyObject *obj, PyObject *file, |
7677 | PyObject *protocol, int fix_imports, |
7678 | PyObject *buffer_callback) |
7679 | /*[clinic end generated code: output=706186dba996490c input=5ed6653da99cd97c]*/ |
7680 | { |
7681 | PicklerObject *pickler = _Pickler_New(); |
7682 | |
7683 | if (pickler == NULL) |
7684 | return NULL; |
7685 | |
7686 | if (_Pickler_SetProtocol(pickler, protocol, fix_imports) < 0) |
7687 | goto error; |
7688 | |
7689 | if (_Pickler_SetOutputStream(pickler, file) < 0) |
7690 | goto error; |
7691 | |
7692 | if (_Pickler_SetBufferCallback(pickler, buffer_callback) < 0) |
7693 | goto error; |
7694 | |
7695 | if (dump(pickler, obj) < 0) |
7696 | goto error; |
7697 | |
7698 | if (_Pickler_FlushToFile(pickler) < 0) |
7699 | goto error; |
7700 | |
7701 | Py_DECREF(pickler); |
7702 | Py_RETURN_NONE; |
7703 | |
7704 | error: |
7705 | Py_XDECREF(pickler); |
7706 | return NULL; |
7707 | } |
7708 | |
7709 | /*[clinic input] |
7710 | |
7711 | _pickle.dumps |
7712 | |
7713 | obj: object |
7714 | protocol: object = None |
7715 | * |
7716 | fix_imports: bool = True |
7717 | buffer_callback: object = None |
7718 | |
7719 | Return the pickled representation of the object as a bytes object. |
7720 | |
7721 | The optional *protocol* argument tells the pickler to use the given |
7722 | protocol; supported protocols are 0, 1, 2, 3, 4 and 5. The default |
7723 | protocol is 4. It was introduced in Python 3.4, and is incompatible |
7724 | with previous versions. |
7725 | |
7726 | Specifying a negative protocol version selects the highest protocol |
7727 | version supported. The higher the protocol used, the more recent the |
7728 | version of Python needed to read the pickle produced. |
7729 | |
7730 | If *fix_imports* is True and *protocol* is less than 3, pickle will |
7731 | try to map the new Python 3 names to the old module names used in |
7732 | Python 2, so that the pickle data stream is readable with Python 2. |
7733 | |
7734 | If *buffer_callback* is None (the default), buffer views are serialized |
7735 | into *file* as part of the pickle stream. It is an error if |
7736 | *buffer_callback* is not None and *protocol* is None or smaller than 5. |
7737 | |
7738 | [clinic start generated code]*/ |
7739 | |
7740 | static PyObject * |
7741 | _pickle_dumps_impl(PyObject *module, PyObject *obj, PyObject *protocol, |
7742 | int fix_imports, PyObject *buffer_callback) |
7743 | /*[clinic end generated code: output=fbab0093a5580fdf input=e543272436c6f987]*/ |
7744 | { |
7745 | PyObject *result; |
7746 | PicklerObject *pickler = _Pickler_New(); |
7747 | |
7748 | if (pickler == NULL) |
7749 | return NULL; |
7750 | |
7751 | if (_Pickler_SetProtocol(pickler, protocol, fix_imports) < 0) |
7752 | goto error; |
7753 | |
7754 | if (_Pickler_SetBufferCallback(pickler, buffer_callback) < 0) |
7755 | goto error; |
7756 | |
7757 | if (dump(pickler, obj) < 0) |
7758 | goto error; |
7759 | |
7760 | result = _Pickler_GetString(pickler); |
7761 | Py_DECREF(pickler); |
7762 | return result; |
7763 | |
7764 | error: |
7765 | Py_XDECREF(pickler); |
7766 | return NULL; |
7767 | } |
7768 | |
7769 | /*[clinic input] |
7770 | |
7771 | _pickle.load |
7772 | |
7773 | file: object |
7774 | * |
7775 | fix_imports: bool = True |
7776 | encoding: str = 'ASCII' |
7777 | errors: str = 'strict' |
7778 | buffers: object(c_default="NULL") = () |
7779 | |
7780 | Read and return an object from the pickle data stored in a file. |
7781 | |
7782 | This is equivalent to ``Unpickler(file).load()``, but may be more |
7783 | efficient. |
7784 | |
7785 | The protocol version of the pickle is detected automatically, so no |
7786 | protocol argument is needed. Bytes past the pickled object's |
7787 | representation are ignored. |
7788 | |
7789 | The argument *file* must have two methods, a read() method that takes |
7790 | an integer argument, and a readline() method that requires no |
7791 | arguments. Both methods should return bytes. Thus *file* can be a |
7792 | binary file object opened for reading, an io.BytesIO object, or any |
7793 | other custom object that meets this interface. |
7794 | |
7795 | Optional keyword arguments are *fix_imports*, *encoding* and *errors*, |
7796 | which are used to control compatibility support for pickle stream |
7797 | generated by Python 2. If *fix_imports* is True, pickle will try to |
7798 | map the old Python 2 names to the new names used in Python 3. The |
7799 | *encoding* and *errors* tell pickle how to decode 8-bit string |
7800 | instances pickled by Python 2; these default to 'ASCII' and 'strict', |
7801 | respectively. The *encoding* can be 'bytes' to read these 8-bit |
7802 | string instances as bytes objects. |
7803 | [clinic start generated code]*/ |
7804 | |
7805 | static PyObject * |
7806 | _pickle_load_impl(PyObject *module, PyObject *file, int fix_imports, |
7807 | const char *encoding, const char *errors, |
7808 | PyObject *buffers) |
7809 | /*[clinic end generated code: output=250452d141c23e76 input=46c7c31c92f4f371]*/ |
7810 | { |
7811 | PyObject *result; |
7812 | UnpicklerObject *unpickler = _Unpickler_New(); |
7813 | |
7814 | if (unpickler == NULL) |
7815 | return NULL; |
7816 | |
7817 | if (_Unpickler_SetInputStream(unpickler, file) < 0) |
7818 | goto error; |
7819 | |
7820 | if (_Unpickler_SetInputEncoding(unpickler, encoding, errors) < 0) |
7821 | goto error; |
7822 | |
7823 | if (_Unpickler_SetBuffers(unpickler, buffers) < 0) |
7824 | goto error; |
7825 | |
7826 | unpickler->fix_imports = fix_imports; |
7827 | |
7828 | result = load(unpickler); |
7829 | Py_DECREF(unpickler); |
7830 | return result; |
7831 | |
7832 | error: |
7833 | Py_XDECREF(unpickler); |
7834 | return NULL; |
7835 | } |
7836 | |
7837 | /*[clinic input] |
7838 | |
7839 | _pickle.loads |
7840 | |
7841 | data: object |
7842 | / |
7843 | * |
7844 | fix_imports: bool = True |
7845 | encoding: str = 'ASCII' |
7846 | errors: str = 'strict' |
7847 | buffers: object(c_default="NULL") = () |
7848 | |
7849 | Read and return an object from the given pickle data. |
7850 | |
7851 | The protocol version of the pickle is detected automatically, so no |
7852 | protocol argument is needed. Bytes past the pickled object's |
7853 | representation are ignored. |
7854 | |
7855 | Optional keyword arguments are *fix_imports*, *encoding* and *errors*, |
7856 | which are used to control compatibility support for pickle stream |
7857 | generated by Python 2. If *fix_imports* is True, pickle will try to |
7858 | map the old Python 2 names to the new names used in Python 3. The |
7859 | *encoding* and *errors* tell pickle how to decode 8-bit string |
7860 | instances pickled by Python 2; these default to 'ASCII' and 'strict', |
7861 | respectively. The *encoding* can be 'bytes' to read these 8-bit |
7862 | string instances as bytes objects. |
7863 | [clinic start generated code]*/ |
7864 | |
7865 | static PyObject * |
7866 | _pickle_loads_impl(PyObject *module, PyObject *data, int fix_imports, |
7867 | const char *encoding, const char *errors, |
7868 | PyObject *buffers) |
7869 | /*[clinic end generated code: output=82ac1e6b588e6d02 input=b3615540d0535087]*/ |
7870 | { |
7871 | PyObject *result; |
7872 | UnpicklerObject *unpickler = _Unpickler_New(); |
7873 | |
7874 | if (unpickler == NULL) |
7875 | return NULL; |
7876 | |
7877 | if (_Unpickler_SetStringInput(unpickler, data) < 0) |
7878 | goto error; |
7879 | |
7880 | if (_Unpickler_SetInputEncoding(unpickler, encoding, errors) < 0) |
7881 | goto error; |
7882 | |
7883 | if (_Unpickler_SetBuffers(unpickler, buffers) < 0) |
7884 | goto error; |
7885 | |
7886 | unpickler->fix_imports = fix_imports; |
7887 | |
7888 | result = load(unpickler); |
7889 | Py_DECREF(unpickler); |
7890 | return result; |
7891 | |
7892 | error: |
7893 | Py_XDECREF(unpickler); |
7894 | return NULL; |
7895 | } |
7896 | |
7897 | static struct PyMethodDef pickle_methods[] = { |
7898 | _PICKLE_DUMP_METHODDEF |
7899 | _PICKLE_DUMPS_METHODDEF |
7900 | _PICKLE_LOAD_METHODDEF |
7901 | _PICKLE_LOADS_METHODDEF |
7902 | {NULL, NULL} /* sentinel */ |
7903 | }; |
7904 | |
7905 | static int |
7906 | pickle_clear(PyObject *m) |
7907 | { |
7908 | _Pickle_ClearState(_Pickle_GetState(m)); |
7909 | return 0; |
7910 | } |
7911 | |
7912 | static void |
7913 | pickle_free(PyObject *m) |
7914 | { |
7915 | _Pickle_ClearState(_Pickle_GetState(m)); |
7916 | } |
7917 | |
7918 | static int |
7919 | pickle_traverse(PyObject *m, visitproc visit, void *arg) |
7920 | { |
7921 | PickleState *st = _Pickle_GetState(m); |
7922 | Py_VISIT(st->PickleError); |
7923 | Py_VISIT(st->PicklingError); |
7924 | Py_VISIT(st->UnpicklingError); |
7925 | Py_VISIT(st->dispatch_table); |
7926 | Py_VISIT(st->extension_registry); |
7927 | Py_VISIT(st->extension_cache); |
7928 | Py_VISIT(st->inverted_registry); |
7929 | Py_VISIT(st->name_mapping_2to3); |
7930 | Py_VISIT(st->import_mapping_2to3); |
7931 | Py_VISIT(st->name_mapping_3to2); |
7932 | Py_VISIT(st->import_mapping_3to2); |
7933 | Py_VISIT(st->codecs_encode); |
7934 | Py_VISIT(st->getattr); |
7935 | Py_VISIT(st->partial); |
7936 | return 0; |
7937 | } |
7938 | |
7939 | static struct PyModuleDef _picklemodule = { |
7940 | PyModuleDef_HEAD_INIT, |
7941 | "_pickle" , /* m_name */ |
7942 | pickle_module_doc, /* m_doc */ |
7943 | sizeof(PickleState), /* m_size */ |
7944 | pickle_methods, /* m_methods */ |
7945 | NULL, /* m_reload */ |
7946 | pickle_traverse, /* m_traverse */ |
7947 | pickle_clear, /* m_clear */ |
7948 | (freefunc)pickle_free /* m_free */ |
7949 | }; |
7950 | |
7951 | PyMODINIT_FUNC |
7952 | PyInit__pickle(void) |
7953 | { |
7954 | PyObject *m; |
7955 | PickleState *st; |
7956 | |
7957 | m = PyState_FindModule(&_picklemodule); |
7958 | if (m) { |
7959 | Py_INCREF(m); |
7960 | return m; |
7961 | } |
7962 | |
7963 | if (PyType_Ready(&Pdata_Type) < 0) |
7964 | return NULL; |
7965 | if (PyType_Ready(&PicklerMemoProxyType) < 0) |
7966 | return NULL; |
7967 | if (PyType_Ready(&UnpicklerMemoProxyType) < 0) |
7968 | return NULL; |
7969 | |
7970 | /* Create the module and add the functions. */ |
7971 | m = PyModule_Create(&_picklemodule); |
7972 | if (m == NULL) |
7973 | return NULL; |
7974 | |
7975 | /* Add types */ |
7976 | if (PyModule_AddType(m, &Pickler_Type) < 0) { |
7977 | return NULL; |
7978 | } |
7979 | if (PyModule_AddType(m, &Unpickler_Type) < 0) { |
7980 | return NULL; |
7981 | } |
7982 | if (PyModule_AddType(m, &PyPickleBuffer_Type) < 0) { |
7983 | return NULL; |
7984 | } |
7985 | |
7986 | st = _Pickle_GetState(m); |
7987 | |
7988 | /* Initialize the exceptions. */ |
7989 | st->PickleError = PyErr_NewException("_pickle.PickleError" , NULL, NULL); |
7990 | if (st->PickleError == NULL) |
7991 | return NULL; |
7992 | st->PicklingError = \ |
7993 | PyErr_NewException("_pickle.PicklingError" , st->PickleError, NULL); |
7994 | if (st->PicklingError == NULL) |
7995 | return NULL; |
7996 | st->UnpicklingError = \ |
7997 | PyErr_NewException("_pickle.UnpicklingError" , st->PickleError, NULL); |
7998 | if (st->UnpicklingError == NULL) |
7999 | return NULL; |
8000 | |
8001 | Py_INCREF(st->PickleError); |
8002 | if (PyModule_AddObject(m, "PickleError" , st->PickleError) < 0) |
8003 | return NULL; |
8004 | Py_INCREF(st->PicklingError); |
8005 | if (PyModule_AddObject(m, "PicklingError" , st->PicklingError) < 0) |
8006 | return NULL; |
8007 | Py_INCREF(st->UnpicklingError); |
8008 | if (PyModule_AddObject(m, "UnpicklingError" , st->UnpicklingError) < 0) |
8009 | return NULL; |
8010 | |
8011 | if (_Pickle_InitState(st) < 0) |
8012 | return NULL; |
8013 | |
8014 | return m; |
8015 | } |
8016 | |