Alexandre Vassalotti | ca2d610 | 2008-06-12 18:26:05 +0000 | [diff] [blame] | 1 | #include "Python.h" |
| 2 | #include "structmember.h" |
| 3 | |
| 4 | PyDoc_STRVAR(pickle_module_doc, |
| 5 | "Optimized C implementation for the Python pickle module."); |
| 6 | |
| 7 | /* Bump this when new opcodes are added to the pickle protocol. */ |
| 8 | enum { |
| 9 | HIGHEST_PROTOCOL = 3, |
| 10 | DEFAULT_PROTOCOL = 3 |
| 11 | }; |
| 12 | |
| 13 | |
| 14 | /* Pickle opcodes. These must be kept updated with pickle.py. |
| 15 | Extensive docs are in pickletools.py. */ |
| 16 | enum opcode { |
| 17 | MARK = '(', |
| 18 | STOP = '.', |
| 19 | POP = '0', |
| 20 | POP_MARK = '1', |
| 21 | DUP = '2', |
| 22 | FLOAT = 'F', |
| 23 | INT = 'I', |
| 24 | BININT = 'J', |
| 25 | BININT1 = 'K', |
| 26 | LONG = 'L', |
| 27 | BININT2 = 'M', |
| 28 | NONE = 'N', |
| 29 | PERSID = 'P', |
| 30 | BINPERSID = 'Q', |
| 31 | REDUCE = 'R', |
| 32 | STRING = 'S', |
| 33 | BINSTRING = 'T', |
| 34 | SHORT_BINSTRING = 'U', |
| 35 | UNICODE = 'V', |
| 36 | BINUNICODE = 'X', |
| 37 | APPEND = 'a', |
| 38 | BUILD = 'b', |
| 39 | GLOBAL = 'c', |
| 40 | DICT = 'd', |
| 41 | EMPTY_DICT = '}', |
| 42 | APPENDS = 'e', |
| 43 | GET = 'g', |
| 44 | BINGET = 'h', |
| 45 | INST = 'i', |
| 46 | LONG_BINGET = 'j', |
| 47 | LIST = 'l', |
| 48 | EMPTY_LIST = ']', |
| 49 | OBJ = 'o', |
| 50 | PUT = 'p', |
| 51 | BINPUT = 'q', |
| 52 | LONG_BINPUT = 'r', |
| 53 | SETITEM = 's', |
| 54 | TUPLE = 't', |
| 55 | EMPTY_TUPLE = ')', |
| 56 | SETITEMS = 'u', |
| 57 | BINFLOAT = 'G', |
| 58 | |
| 59 | /* Protocol 2. */ |
| 60 | PROTO = '\x80', |
| 61 | NEWOBJ = '\x81', |
| 62 | EXT1 = '\x82', |
| 63 | EXT2 = '\x83', |
| 64 | EXT4 = '\x84', |
| 65 | TUPLE1 = '\x85', |
| 66 | TUPLE2 = '\x86', |
| 67 | TUPLE3 = '\x87', |
| 68 | NEWTRUE = '\x88', |
| 69 | NEWFALSE = '\x89', |
| 70 | LONG1 = '\x8a', |
| 71 | LONG4 = '\x8b', |
| 72 | |
| 73 | /* Protocol 3 (Python 3.x) */ |
| 74 | BINBYTES = 'B', |
| 75 | SHORT_BINBYTES = 'C', |
| 76 | }; |
| 77 | |
| 78 | /* These aren't opcodes -- they're ways to pickle bools before protocol 2 |
| 79 | * so that unpicklers written before bools were introduced unpickle them |
| 80 | * as ints, but unpicklers after can recognize that bools were intended. |
| 81 | * Note that protocol 2 added direct ways to pickle bools. |
| 82 | */ |
| 83 | #undef TRUE |
| 84 | #define TRUE "I01\n" |
| 85 | #undef FALSE |
| 86 | #define FALSE "I00\n" |
| 87 | |
| 88 | enum { |
| 89 | /* Keep in synch with pickle.Pickler._BATCHSIZE. This is how many elements |
| 90 | batch_list/dict() pumps out before doing APPENDS/SETITEMS. Nothing will |
| 91 | break if this gets out of synch with pickle.py, but it's unclear that would |
| 92 | help anything either. */ |
| 93 | BATCHSIZE = 1000, |
| 94 | |
| 95 | /* Nesting limit until Pickler, when running in "fast mode", starts |
| 96 | checking for self-referential data-structures. */ |
| 97 | FAST_NESTING_LIMIT = 50, |
| 98 | |
| 99 | /* Size of the write buffer of Pickler. Higher values will reduce the |
| 100 | number of calls to the write() method of the output stream. */ |
| 101 | WRITE_BUF_SIZE = 256, |
| 102 | }; |
| 103 | |
| 104 | /* Exception classes for pickle. These should override the ones defined in |
| 105 | pickle.py, when the C-optimized Pickler and Unpickler are used. */ |
| 106 | static PyObject *PickleError; |
| 107 | static PyObject *PicklingError; |
| 108 | static PyObject *UnpicklingError; |
| 109 | |
| 110 | /* copyreg.dispatch_table, {type_object: pickling_function} */ |
| 111 | static PyObject *dispatch_table; |
| 112 | /* For EXT[124] opcodes. */ |
| 113 | /* copyreg._extension_registry, {(module_name, function_name): code} */ |
| 114 | static PyObject *extension_registry; |
| 115 | /* copyreg._inverted_registry, {code: (module_name, function_name)} */ |
| 116 | static PyObject *inverted_registry; |
| 117 | /* copyreg._extension_cache, {code: object} */ |
| 118 | static PyObject *extension_cache; |
| 119 | |
| 120 | /* XXX: Are these really nescessary? */ |
| 121 | /* As the name says, an empty tuple. */ |
| 122 | static PyObject *empty_tuple; |
| 123 | /* For looking up name pairs in copyreg._extension_registry. */ |
| 124 | static PyObject *two_tuple; |
| 125 | |
| 126 | static int |
| 127 | stack_underflow(void) |
| 128 | { |
| 129 | PyErr_SetString(UnpicklingError, "unpickling stack underflow"); |
| 130 | return -1; |
| 131 | } |
| 132 | |
| 133 | /* Internal data type used as the unpickling stack. */ |
| 134 | typedef struct { |
| 135 | PyObject_HEAD |
| 136 | int length; /* number of initial slots in data currently used */ |
| 137 | int size; /* number of slots in data allocated */ |
| 138 | PyObject **data; |
| 139 | } Pdata; |
| 140 | |
| 141 | static void |
| 142 | Pdata_dealloc(Pdata *self) |
| 143 | { |
| 144 | int i; |
| 145 | PyObject **p; |
| 146 | |
| 147 | for (i = self->length, p = self->data; --i >= 0; p++) { |
| 148 | Py_DECREF(*p); |
| 149 | } |
| 150 | if (self->data) |
| 151 | PyMem_Free(self->data); |
| 152 | PyObject_Del(self); |
| 153 | } |
| 154 | |
| 155 | static PyTypeObject Pdata_Type = { |
| 156 | PyVarObject_HEAD_INIT(NULL, 0) |
| 157 | "_pickle.Pdata", /*tp_name*/ |
| 158 | sizeof(Pdata), /*tp_basicsize*/ |
| 159 | 0, /*tp_itemsize*/ |
| 160 | (destructor)Pdata_dealloc, /*tp_dealloc*/ |
| 161 | }; |
| 162 | |
| 163 | static PyObject * |
| 164 | Pdata_New(void) |
| 165 | { |
| 166 | Pdata *self; |
| 167 | |
| 168 | if (!(self = PyObject_New(Pdata, &Pdata_Type))) |
| 169 | return NULL; |
| 170 | self->size = 8; |
| 171 | self->length = 0; |
| 172 | self->data = PyMem_Malloc(self->size * sizeof(PyObject *)); |
| 173 | if (self->data) |
| 174 | return (PyObject *)self; |
| 175 | Py_DECREF(self); |
| 176 | return PyErr_NoMemory(); |
| 177 | } |
| 178 | |
| 179 | |
| 180 | /* Retain only the initial clearto items. If clearto >= the current |
| 181 | * number of items, this is a (non-erroneous) NOP. |
| 182 | */ |
| 183 | static int |
| 184 | Pdata_clear(Pdata *self, int clearto) |
| 185 | { |
| 186 | int i; |
| 187 | PyObject **p; |
| 188 | |
| 189 | if (clearto < 0) |
| 190 | return stack_underflow(); |
| 191 | if (clearto >= self->length) |
| 192 | return 0; |
| 193 | |
| 194 | for (i = self->length, p = self->data + clearto; --i >= clearto; p++) { |
| 195 | Py_CLEAR(*p); |
| 196 | } |
| 197 | self->length = clearto; |
| 198 | |
| 199 | return 0; |
| 200 | } |
| 201 | |
| 202 | static int |
| 203 | Pdata_grow(Pdata *self) |
| 204 | { |
| 205 | int bigger; |
| 206 | size_t nbytes; |
| 207 | PyObject **tmp; |
| 208 | |
| 209 | bigger = (self->size << 1) + 1; |
| 210 | if (bigger <= 0) /* was 0, or new value overflows */ |
| 211 | goto nomemory; |
| 212 | if ((int)(size_t)bigger != bigger) |
| 213 | goto nomemory; |
| 214 | nbytes = (size_t)bigger * sizeof(PyObject *); |
| 215 | if (nbytes / sizeof(PyObject *) != (size_t)bigger) |
| 216 | goto nomemory; |
| 217 | tmp = PyMem_Realloc(self->data, nbytes); |
| 218 | if (tmp == NULL) |
| 219 | goto nomemory; |
| 220 | self->data = tmp; |
| 221 | self->size = bigger; |
| 222 | return 0; |
| 223 | |
| 224 | nomemory: |
| 225 | PyErr_NoMemory(); |
| 226 | return -1; |
| 227 | } |
| 228 | |
| 229 | /* D is a Pdata*. Pop the topmost element and store it into V, which |
| 230 | * must be an lvalue holding PyObject*. On stack underflow, UnpicklingError |
| 231 | * is raised and V is set to NULL. |
| 232 | */ |
| 233 | static PyObject * |
| 234 | Pdata_pop(Pdata *self) |
| 235 | { |
| 236 | if (self->length == 0) { |
| 237 | PyErr_SetString(UnpicklingError, "bad pickle data"); |
| 238 | return NULL; |
| 239 | } |
| 240 | return self->data[--(self->length)]; |
| 241 | } |
| 242 | #define PDATA_POP(D, V) do { (V) = Pdata_pop((D)); } while (0) |
| 243 | |
| 244 | static int |
| 245 | Pdata_push(Pdata *self, PyObject *obj) |
| 246 | { |
| 247 | if (self->length == self->size && Pdata_grow(self) < 0) { |
| 248 | return -1; |
| 249 | } |
| 250 | self->data[self->length++] = obj; |
| 251 | return 0; |
| 252 | } |
| 253 | |
| 254 | /* Push an object on stack, transferring its ownership to the stack. */ |
| 255 | #define PDATA_PUSH(D, O, ER) do { \ |
| 256 | if (Pdata_push((D), (O)) < 0) return (ER); } while(0) |
| 257 | |
| 258 | /* Push an object on stack, adding a new reference to the object. */ |
| 259 | #define PDATA_APPEND(D, O, ER) do { \ |
| 260 | Py_INCREF((O)); \ |
| 261 | if (Pdata_push((D), (O)) < 0) return (ER); } while(0) |
| 262 | |
| 263 | static PyObject * |
| 264 | Pdata_poptuple(Pdata *self, Py_ssize_t start) |
| 265 | { |
| 266 | PyObject *tuple; |
| 267 | Py_ssize_t len, i, j; |
| 268 | |
| 269 | len = self->length - start; |
| 270 | tuple = PyTuple_New(len); |
| 271 | if (tuple == NULL) |
| 272 | return NULL; |
| 273 | for (i = start, j = 0; j < len; i++, j++) |
| 274 | PyTuple_SET_ITEM(tuple, j, self->data[i]); |
| 275 | |
| 276 | self->length = start; |
| 277 | return tuple; |
| 278 | } |
| 279 | |
| 280 | static PyObject * |
| 281 | Pdata_poplist(Pdata *self, Py_ssize_t start) |
| 282 | { |
| 283 | PyObject *list; |
| 284 | Py_ssize_t len, i, j; |
| 285 | |
| 286 | len = self->length - start; |
| 287 | list = PyList_New(len); |
| 288 | if (list == NULL) |
| 289 | return NULL; |
| 290 | for (i = start, j = 0; j < len; i++, j++) |
| 291 | PyList_SET_ITEM(list, j, self->data[i]); |
| 292 | |
| 293 | self->length = start; |
| 294 | return list; |
| 295 | } |
| 296 | |
| 297 | typedef struct PicklerObject { |
| 298 | PyObject_HEAD |
| 299 | PyObject *write; /* write() method of the output stream */ |
| 300 | PyObject *memo; /* Memo dictionary, keep track of the seen |
| 301 | objects to support self-referential objects |
| 302 | pickling. */ |
| 303 | PyObject *pers_func; /* persistent_id() method, can be NULL */ |
| 304 | PyObject *arg; |
| 305 | int proto; /* Pickle protocol number, >= 0 */ |
| 306 | int bin; /* Boolean, true if proto > 0 */ |
| 307 | int nesting; /* Current nesting level, this is to guard |
| 308 | save() from going into infinite recursion |
| 309 | and segfaulting. */ |
| 310 | int buf_size; /* Size of the current buffered pickle data */ |
| 311 | char *write_buf; /* Write buffer, this is to avoid calling the |
| 312 | write() method of the output stream too |
| 313 | often. */ |
| 314 | int fast; /* Enable fast mode if set to a true value. |
| 315 | The fast mode disable the usage of memo, |
| 316 | therefore speeding the pickling process by |
| 317 | not generating superfluous PUT opcodes. It |
| 318 | should not be used if with self-referential |
| 319 | objects. */ |
| 320 | int fast_nesting; |
| 321 | PyObject *fast_memo; |
| 322 | } PicklerObject; |
| 323 | |
| 324 | typedef struct UnpicklerObject { |
| 325 | PyObject_HEAD |
| 326 | Pdata *stack; /* Pickle data stack, store unpickled objects. */ |
| 327 | PyObject *readline; /* readline() method of the output stream */ |
| 328 | PyObject *read; /* read() method of the output stream */ |
| 329 | PyObject *memo; /* Memo dictionary, provide the objects stored |
| 330 | using the PUT opcodes. */ |
| 331 | PyObject *arg; |
| 332 | PyObject *pers_func; /* persistent_load() method, can be NULL. */ |
| 333 | PyObject *last_string; /* Reference to the last string read by the |
| 334 | readline() method. */ |
| 335 | char *buffer; /* Reading buffer. */ |
| 336 | char *encoding; /* Name of the encoding to be used for |
| 337 | decoding strings pickled using Python |
| 338 | 2.x. The default value is "ASCII" */ |
| 339 | char *errors; /* Name of errors handling scheme to used when |
| 340 | decoding strings. The default value is |
| 341 | "strict". */ |
| 342 | int *marks; /* Mark stack, used for unpickling container |
| 343 | objects. */ |
| 344 | Py_ssize_t num_marks; /* Number of marks in the mark stack. */ |
| 345 | Py_ssize_t marks_size; /* Current allocated size of the mark stack. */ |
| 346 | } UnpicklerObject; |
| 347 | |
| 348 | /* Forward declarations */ |
| 349 | static int save(PicklerObject *, PyObject *, int); |
| 350 | static int save_reduce(PicklerObject *, PyObject *, PyObject *); |
| 351 | static PyTypeObject Pickler_Type; |
| 352 | static PyTypeObject Unpickler_Type; |
| 353 | |
| 354 | |
| 355 | /* Helpers for creating the argument tuple passed to functions. This has the |
| 356 | performance advantage of calling PyTuple_New() only once. */ |
| 357 | |
| 358 | #define ARG_TUP(self, obj) do { \ |
| 359 | if ((self)->arg || ((self)->arg=PyTuple_New(1))) { \ |
| 360 | Py_XDECREF(PyTuple_GET_ITEM((self)->arg, 0)); \ |
| 361 | PyTuple_SET_ITEM((self)->arg, 0, (obj)); \ |
| 362 | } \ |
| 363 | else { \ |
| 364 | Py_DECREF((obj)); \ |
| 365 | } \ |
| 366 | } while (0) |
| 367 | |
| 368 | #define FREE_ARG_TUP(self) do { \ |
| 369 | if ((self)->arg->ob_refcnt > 1) \ |
| 370 | Py_CLEAR((self)->arg); \ |
| 371 | } while (0) |
| 372 | |
| 373 | /* A temporary cleaner API for fast single argument function call. |
| 374 | |
| 375 | XXX: Does caching the argument tuple provides any real performance benefits? |
| 376 | |
| 377 | A quick benchmark, on a 2.0GHz Athlon64 3200+ running Linux 2.6.24 with |
| 378 | glibc 2.7, tells me that it takes roughly 20,000,000 PyTuple_New(1) calls |
| 379 | when the tuple is retrieved from the freelist (i.e, call PyTuple_New() then |
| 380 | immediately DECREF it) and 1,200,000 calls when allocating brand new tuples |
| 381 | (i.e, call PyTuple_New() and store the returned value in an array), to save |
| 382 | one second (wall clock time). Either ways, the loading time a pickle stream |
| 383 | large enough to generate this number of calls would be massively |
| 384 | overwhelmed by other factors, like I/O throughput, the GC traversal and |
| 385 | object allocation overhead. So, I really doubt these functions provide any |
| 386 | real benefits. |
| 387 | |
| 388 | On the other hand, oprofile reports that pickle spends a lot of time in |
| 389 | these functions. But, that is probably more related to the function call |
| 390 | overhead, than the argument tuple allocation. |
| 391 | |
| 392 | XXX: And, what is the reference behavior of these? Steal, borrow? At first |
| 393 | glance, it seems to steal the reference of 'arg' and borrow the reference |
| 394 | of 'func'. |
| 395 | */ |
| 396 | static PyObject * |
| 397 | pickler_call(PicklerObject *self, PyObject *func, PyObject *arg) |
| 398 | { |
| 399 | PyObject *result = NULL; |
| 400 | |
| 401 | ARG_TUP(self, arg); |
| 402 | if (self->arg) { |
| 403 | result = PyObject_Call(func, self->arg, NULL); |
| 404 | FREE_ARG_TUP(self); |
| 405 | } |
| 406 | return result; |
| 407 | } |
| 408 | |
| 409 | static PyObject * |
| 410 | unpickler_call(UnpicklerObject *self, PyObject *func, PyObject *arg) |
| 411 | { |
| 412 | PyObject *result = NULL; |
| 413 | |
| 414 | ARG_TUP(self, arg); |
| 415 | if (self->arg) { |
| 416 | result = PyObject_Call(func, self->arg, NULL); |
| 417 | FREE_ARG_TUP(self); |
| 418 | } |
| 419 | return result; |
| 420 | } |
| 421 | |
| 422 | static Py_ssize_t |
| 423 | pickler_write(PicklerObject *self, const char *s, Py_ssize_t n) |
| 424 | { |
| 425 | PyObject *data, *result; |
| 426 | |
| 427 | if (s == NULL) { |
| 428 | if (!(self->buf_size)) |
| 429 | return 0; |
| 430 | data = PyBytes_FromStringAndSize(self->write_buf, self->buf_size); |
| 431 | if (data == NULL) |
| 432 | return -1; |
| 433 | } |
| 434 | else { |
| 435 | if (self->buf_size && (n + self->buf_size) > WRITE_BUF_SIZE) { |
| 436 | if (pickler_write(self, NULL, 0) < 0) |
| 437 | return -1; |
| 438 | } |
| 439 | |
| 440 | if (n > WRITE_BUF_SIZE) { |
| 441 | if (!(data = PyBytes_FromStringAndSize(s, n))) |
| 442 | return -1; |
| 443 | } |
| 444 | else { |
| 445 | memcpy(self->write_buf + self->buf_size, s, n); |
| 446 | self->buf_size += n; |
| 447 | return n; |
| 448 | } |
| 449 | } |
| 450 | |
| 451 | /* object with write method */ |
| 452 | result = pickler_call(self, self->write, data); |
| 453 | if (result == NULL) |
| 454 | return -1; |
| 455 | |
| 456 | Py_DECREF(result); |
| 457 | self->buf_size = 0; |
| 458 | return n; |
| 459 | } |
| 460 | |
| 461 | /* XXX: These read/readline functions ought to be optimized. Buffered I/O |
| 462 | might help a lot, especially with the new (but much slower) io library. |
| 463 | On the other hand, the added complexity might not worth it. |
| 464 | */ |
| 465 | |
| 466 | /* Read at least n characters from the input stream and set s to the current |
| 467 | reading position. */ |
| 468 | static Py_ssize_t |
| 469 | unpickler_read(UnpicklerObject *self, char **s, Py_ssize_t n) |
| 470 | { |
| 471 | PyObject *len; |
| 472 | PyObject *data; |
| 473 | |
| 474 | len = PyLong_FromSsize_t(n); |
| 475 | if (len == NULL) |
| 476 | return -1; |
| 477 | |
| 478 | data = unpickler_call(self, self->read, len); |
| 479 | if (data == NULL) |
| 480 | return -1; |
| 481 | |
| 482 | /* XXX: Should bytearray be supported too? */ |
| 483 | if (!PyBytes_Check(data)) { |
| 484 | PyErr_SetString(PyExc_ValueError, |
| 485 | "read() from the underlying stream did not" |
| 486 | "return bytes"); |
| 487 | return -1; |
| 488 | } |
| 489 | |
| 490 | Py_XDECREF(self->last_string); |
| 491 | self->last_string = data; |
| 492 | |
| 493 | if (!(*s = PyBytes_AS_STRING(data))) |
| 494 | return -1; |
| 495 | |
| 496 | return n; |
| 497 | } |
| 498 | |
| 499 | static Py_ssize_t |
| 500 | unpickler_readline(UnpicklerObject *self, char **s) |
| 501 | { |
| 502 | PyObject *data; |
| 503 | |
| 504 | data = PyObject_CallObject(self->readline, empty_tuple); |
| 505 | if (data == NULL) |
| 506 | return -1; |
| 507 | |
| 508 | /* XXX: Should bytearray be supported too? */ |
| 509 | if (!PyBytes_Check(data)) { |
| 510 | PyErr_SetString(PyExc_ValueError, |
| 511 | "readline() from the underlying stream did not" |
| 512 | "return bytes"); |
| 513 | return -1; |
| 514 | } |
| 515 | |
| 516 | Py_XDECREF(self->last_string); |
| 517 | self->last_string = data; |
| 518 | |
| 519 | if (!(*s = PyBytes_AS_STRING(data))) |
| 520 | return -1; |
| 521 | |
| 522 | return PyBytes_GET_SIZE(data); |
| 523 | } |
| 524 | |
| 525 | /* Generate a GET opcode for an object stored in the memo. The 'key' argument |
| 526 | should be the address of the object as returned by PyLong_FromVoidPtr(). */ |
| 527 | static int |
| 528 | memo_get(PicklerObject *self, PyObject *key) |
| 529 | { |
| 530 | PyObject *value; |
| 531 | PyObject *memo_id; |
| 532 | long x; |
| 533 | char pdata[30]; |
| 534 | int len; |
| 535 | |
| 536 | value = PyDict_GetItemWithError(self->memo, key); |
| 537 | if (value == NULL) { |
| 538 | if (!PyErr_Occurred()) |
| 539 | PyErr_SetObject(PyExc_KeyError, key); |
| 540 | return -1; |
| 541 | } |
| 542 | |
| 543 | memo_id = PyTuple_GetItem(value, 0); |
| 544 | if (memo_id == NULL) |
| 545 | return -1; |
| 546 | |
| 547 | if (!PyLong_Check(memo_id)) { |
| 548 | PyErr_SetString(PicklingError, "memo id must be an integer"); |
| 549 | return -1; |
| 550 | } |
| 551 | x = PyLong_AsLong(memo_id); |
| 552 | if (x == -1 && PyErr_Occurred()) |
| 553 | return -1; |
| 554 | |
| 555 | if (!self->bin) { |
| 556 | pdata[0] = GET; |
| 557 | PyOS_snprintf(pdata + 1, sizeof(pdata) - 1, "%ld\n", x); |
| 558 | len = (int)strlen(pdata); |
| 559 | } |
| 560 | else { |
| 561 | if (x < 256) { |
| 562 | pdata[0] = BINGET; |
| 563 | pdata[1] = (unsigned char)(x & 0xff); |
| 564 | len = 2; |
| 565 | } |
| 566 | else if (x <= 0xffffffffL) { |
| 567 | pdata[0] = LONG_BINGET; |
| 568 | pdata[1] = (unsigned char)(x & 0xff); |
| 569 | pdata[2] = (unsigned char)((x >> 8) & 0xff); |
| 570 | pdata[3] = (unsigned char)((x >> 16) & 0xff); |
| 571 | pdata[4] = (unsigned char)((x >> 24) & 0xff); |
| 572 | len = 5; |
| 573 | } |
| 574 | else { /* unlikely */ |
| 575 | PyErr_SetString(PicklingError, |
| 576 | "memo id too large for LONG_BINGET"); |
| 577 | return -1; |
| 578 | } |
| 579 | } |
| 580 | |
| 581 | if (pickler_write(self, pdata, len) < 0) |
| 582 | return -1; |
| 583 | |
| 584 | return 0; |
| 585 | } |
| 586 | |
| 587 | /* Store an object in the memo, assign it a new unique ID based on the number |
| 588 | of objects currently stored in the memo and generate a PUT opcode. */ |
| 589 | static int |
| 590 | memo_put(PicklerObject *self, PyObject *obj) |
| 591 | { |
| 592 | PyObject *key = NULL; |
| 593 | PyObject *memo_id = NULL; |
| 594 | PyObject *tuple = NULL; |
| 595 | long x; |
| 596 | char pdata[30]; |
| 597 | int len; |
| 598 | int status = 0; |
| 599 | |
| 600 | if (self->fast) |
| 601 | return 0; |
| 602 | |
| 603 | key = PyLong_FromVoidPtr(obj); |
| 604 | if (key == NULL) |
| 605 | goto error; |
| 606 | if ((x = PyDict_Size(self->memo)) < 0) |
| 607 | goto error; |
| 608 | memo_id = PyLong_FromLong(x); |
| 609 | if (memo_id == NULL) |
| 610 | goto error; |
| 611 | tuple = PyTuple_New(2); |
| 612 | if (tuple == NULL) |
| 613 | goto error; |
| 614 | |
| 615 | Py_INCREF(memo_id); |
| 616 | PyTuple_SET_ITEM(tuple, 0, memo_id); |
| 617 | Py_INCREF(obj); |
| 618 | PyTuple_SET_ITEM(tuple, 1, obj); |
| 619 | if (PyDict_SetItem(self->memo, key, tuple) < 0) |
| 620 | goto error; |
| 621 | |
| 622 | if (!self->bin) { |
| 623 | pdata[0] = PUT; |
| 624 | PyOS_snprintf(pdata + 1, sizeof(pdata) - 1, "%ld\n", x); |
| 625 | len = strlen(pdata); |
| 626 | } |
| 627 | else { |
| 628 | if (x < 256) { |
| 629 | pdata[0] = BINPUT; |
Alexandre Vassalotti | 7634ff5 | 2008-06-13 02:16:06 +0000 | [diff] [blame] | 630 | pdata[1] = (unsigned char)x; |
Alexandre Vassalotti | ca2d610 | 2008-06-12 18:26:05 +0000 | [diff] [blame] | 631 | len = 2; |
| 632 | } |
| 633 | else if (x <= 0xffffffffL) { |
| 634 | pdata[0] = LONG_BINPUT; |
| 635 | pdata[1] = (unsigned char)(x & 0xff); |
| 636 | pdata[2] = (unsigned char)((x >> 8) & 0xff); |
| 637 | pdata[3] = (unsigned char)((x >> 16) & 0xff); |
| 638 | pdata[4] = (unsigned char)((x >> 24) & 0xff); |
| 639 | len = 5; |
| 640 | } |
| 641 | else { /* unlikely */ |
| 642 | PyErr_SetString(PicklingError, |
| 643 | "memo id too large for LONG_BINPUT"); |
| 644 | return -1; |
| 645 | } |
| 646 | } |
| 647 | |
| 648 | if (pickler_write(self, pdata, len) < 0) |
| 649 | goto error; |
| 650 | |
| 651 | if (0) { |
| 652 | error: |
| 653 | status = -1; |
| 654 | } |
| 655 | |
| 656 | Py_XDECREF(key); |
| 657 | Py_XDECREF(memo_id); |
| 658 | Py_XDECREF(tuple); |
| 659 | |
| 660 | return status; |
| 661 | } |
| 662 | |
| 663 | static PyObject * |
| 664 | whichmodule(PyObject *global, PyObject *global_name) |
| 665 | { |
| 666 | Py_ssize_t i, j; |
| 667 | static PyObject *module_str = NULL; |
| 668 | static PyObject *main_str = NULL; |
| 669 | PyObject *module_name; |
| 670 | PyObject *modules_dict; |
| 671 | PyObject *module; |
| 672 | PyObject *obj; |
| 673 | |
| 674 | if (module_str == NULL) { |
| 675 | module_str = PyUnicode_InternFromString("__module__"); |
| 676 | if (module_str == NULL) |
| 677 | return NULL; |
| 678 | main_str = PyUnicode_InternFromString("__main__"); |
| 679 | if (main_str == NULL) |
| 680 | return NULL; |
| 681 | } |
| 682 | |
| 683 | module_name = PyObject_GetAttr(global, module_str); |
| 684 | |
| 685 | /* In some rare cases (e.g., random.getrandbits), __module__ can be |
| 686 | None. If it is so, then search sys.modules for the module of |
| 687 | global. */ |
| 688 | if (module_name == Py_None) { |
| 689 | Py_DECREF(module_name); |
| 690 | goto search; |
| 691 | } |
| 692 | |
| 693 | if (module_name) { |
| 694 | return module_name; |
| 695 | } |
| 696 | if (PyErr_ExceptionMatches(PyExc_AttributeError)) |
| 697 | PyErr_Clear(); |
| 698 | else |
| 699 | return NULL; |
| 700 | |
| 701 | search: |
| 702 | modules_dict = PySys_GetObject("modules"); |
| 703 | if (modules_dict == NULL) |
| 704 | return NULL; |
| 705 | |
| 706 | i = 0; |
| 707 | module_name = NULL; |
| 708 | while ((j = PyDict_Next(modules_dict, &i, &module_name, &module))) { |
| 709 | if (PyObject_Compare(module_name, main_str) == 0) |
| 710 | continue; |
| 711 | |
| 712 | obj = PyObject_GetAttr(module, global_name); |
| 713 | if (obj == NULL) { |
| 714 | if (PyErr_ExceptionMatches(PyExc_AttributeError)) |
| 715 | PyErr_Clear(); |
| 716 | else |
| 717 | return NULL; |
| 718 | continue; |
| 719 | } |
| 720 | |
| 721 | if (obj != global) { |
| 722 | Py_DECREF(obj); |
| 723 | continue; |
| 724 | } |
| 725 | |
| 726 | Py_DECREF(obj); |
| 727 | break; |
| 728 | } |
| 729 | |
| 730 | /* If no module is found, use __main__. */ |
| 731 | if (!j) { |
| 732 | module_name = main_str; |
| 733 | } |
| 734 | |
| 735 | Py_INCREF(module_name); |
| 736 | return module_name; |
| 737 | } |
| 738 | |
| 739 | /* fast_save_enter() and fast_save_leave() are guards against recursive |
| 740 | objects when Pickler is used with the "fast mode" (i.e., with object |
| 741 | memoization disabled). If the nesting of a list or dict object exceed |
| 742 | FAST_NESTING_LIMIT, these guards will start keeping an internal |
| 743 | reference to the seen list or dict objects and check whether these objects |
| 744 | are recursive. These are not strictly necessary, since save() has a |
| 745 | hard-coded recursion limit, but they give a nicer error message than the |
| 746 | typical RuntimeError. */ |
| 747 | static int |
| 748 | fast_save_enter(PicklerObject *self, PyObject *obj) |
| 749 | { |
| 750 | /* if fast_nesting < 0, we're doing an error exit. */ |
| 751 | if (++self->fast_nesting >= FAST_NESTING_LIMIT) { |
| 752 | PyObject *key = NULL; |
| 753 | if (self->fast_memo == NULL) { |
| 754 | self->fast_memo = PyDict_New(); |
| 755 | if (self->fast_memo == NULL) { |
| 756 | self->fast_nesting = -1; |
| 757 | return 0; |
| 758 | } |
| 759 | } |
| 760 | key = PyLong_FromVoidPtr(obj); |
| 761 | if (key == NULL) |
| 762 | return 0; |
| 763 | if (PyDict_GetItem(self->fast_memo, key)) { |
| 764 | Py_DECREF(key); |
| 765 | PyErr_Format(PyExc_ValueError, |
| 766 | "fast mode: can't pickle cyclic objects " |
| 767 | "including object type %.200s at %p", |
| 768 | obj->ob_type->tp_name, obj); |
| 769 | self->fast_nesting = -1; |
| 770 | return 0; |
| 771 | } |
| 772 | if (PyDict_SetItem(self->fast_memo, key, Py_None) < 0) { |
| 773 | Py_DECREF(key); |
| 774 | self->fast_nesting = -1; |
| 775 | return 0; |
| 776 | } |
| 777 | Py_DECREF(key); |
| 778 | } |
| 779 | return 1; |
| 780 | } |
| 781 | |
| 782 | static int |
| 783 | fast_save_leave(PicklerObject *self, PyObject *obj) |
| 784 | { |
| 785 | if (self->fast_nesting-- >= FAST_NESTING_LIMIT) { |
| 786 | PyObject *key = PyLong_FromVoidPtr(obj); |
| 787 | if (key == NULL) |
| 788 | return 0; |
| 789 | if (PyDict_DelItem(self->fast_memo, key) < 0) { |
| 790 | Py_DECREF(key); |
| 791 | return 0; |
| 792 | } |
| 793 | Py_DECREF(key); |
| 794 | } |
| 795 | return 1; |
| 796 | } |
| 797 | |
| 798 | static int |
| 799 | save_none(PicklerObject *self, PyObject *obj) |
| 800 | { |
| 801 | const char none_op = NONE; |
| 802 | if (pickler_write(self, &none_op, 1) < 0) |
| 803 | return -1; |
| 804 | |
| 805 | return 0; |
| 806 | } |
| 807 | |
| 808 | static int |
| 809 | save_bool(PicklerObject *self, PyObject *obj) |
| 810 | { |
| 811 | static const char *buf[2] = { FALSE, TRUE }; |
| 812 | const char len[2] = {sizeof(FALSE) - 1, sizeof(TRUE) - 1}; |
| 813 | int p = (obj == Py_True); |
| 814 | |
| 815 | if (self->proto >= 2) { |
| 816 | const char bool_op = p ? NEWTRUE : NEWFALSE; |
| 817 | if (pickler_write(self, &bool_op, 1) < 0) |
| 818 | return -1; |
| 819 | } |
| 820 | else if (pickler_write(self, buf[p], len[p]) < 0) |
| 821 | return -1; |
| 822 | |
| 823 | return 0; |
| 824 | } |
| 825 | |
| 826 | static int |
| 827 | save_int(PicklerObject *self, long x) |
| 828 | { |
| 829 | char pdata[32]; |
| 830 | int len = 0; |
| 831 | |
| 832 | if (!self->bin |
| 833 | #if SIZEOF_LONG > 4 |
| 834 | || x > 0x7fffffffL || x < -0x80000000L |
| 835 | #endif |
| 836 | ) { |
| 837 | /* Text-mode pickle, or long too big to fit in the 4-byte |
| 838 | * signed BININT format: store as a string. |
| 839 | */ |
| 840 | pdata[0] = LONG; /* use LONG for consistence with pickle.py */ |
| 841 | PyOS_snprintf(pdata + 1, sizeof(pdata) - 1, "%ld\n", x); |
| 842 | if (pickler_write(self, pdata, strlen(pdata)) < 0) |
| 843 | return -1; |
| 844 | } |
| 845 | else { |
| 846 | /* Binary pickle and x fits in a signed 4-byte int. */ |
| 847 | pdata[1] = (unsigned char)(x & 0xff); |
| 848 | pdata[2] = (unsigned char)((x >> 8) & 0xff); |
| 849 | pdata[3] = (unsigned char)((x >> 16) & 0xff); |
| 850 | pdata[4] = (unsigned char)((x >> 24) & 0xff); |
| 851 | |
| 852 | if ((pdata[4] == 0) && (pdata[3] == 0)) { |
| 853 | if (pdata[2] == 0) { |
| 854 | pdata[0] = BININT1; |
| 855 | len = 2; |
| 856 | } |
| 857 | else { |
| 858 | pdata[0] = BININT2; |
| 859 | len = 3; |
| 860 | } |
| 861 | } |
| 862 | else { |
| 863 | pdata[0] = BININT; |
| 864 | len = 5; |
| 865 | } |
| 866 | |
| 867 | if (pickler_write(self, pdata, len) < 0) |
| 868 | return -1; |
| 869 | } |
| 870 | |
| 871 | return 0; |
| 872 | } |
| 873 | |
| 874 | static int |
| 875 | save_long(PicklerObject *self, PyObject *obj) |
| 876 | { |
| 877 | PyObject *repr = NULL; |
| 878 | Py_ssize_t size; |
| 879 | long val = PyLong_AsLong(obj); |
| 880 | int status = 0; |
| 881 | |
| 882 | const char long_op = LONG; |
| 883 | |
| 884 | if (val == -1 && PyErr_Occurred()) { |
| 885 | /* out of range for int pickling */ |
| 886 | PyErr_Clear(); |
| 887 | } |
| 888 | else |
| 889 | return save_int(self, val); |
| 890 | |
| 891 | if (self->proto >= 2) { |
| 892 | /* Linear-time pickling. */ |
| 893 | size_t nbits; |
| 894 | size_t nbytes; |
| 895 | unsigned char *pdata; |
| 896 | char header[5]; |
| 897 | int i; |
| 898 | int sign = _PyLong_Sign(obj); |
| 899 | |
| 900 | if (sign == 0) { |
| 901 | header[0] = LONG1; |
| 902 | header[1] = 0; /* It's 0 -- an empty bytestring. */ |
| 903 | if (pickler_write(self, header, 2) < 0) |
| 904 | goto error; |
| 905 | return 0; |
| 906 | } |
| 907 | nbits = _PyLong_NumBits(obj); |
| 908 | if (nbits == (size_t)-1 && PyErr_Occurred()) |
| 909 | goto error; |
| 910 | /* How many bytes do we need? There are nbits >> 3 full |
| 911 | * bytes of data, and nbits & 7 leftover bits. If there |
| 912 | * are any leftover bits, then we clearly need another |
| 913 | * byte. Wnat's not so obvious is that we *probably* |
| 914 | * need another byte even if there aren't any leftovers: |
| 915 | * the most-significant bit of the most-significant byte |
| 916 | * acts like a sign bit, and it's usually got a sense |
| 917 | * opposite of the one we need. The exception is longs |
| 918 | * of the form -(2**(8*j-1)) for j > 0. Such a long is |
| 919 | * its own 256's-complement, so has the right sign bit |
| 920 | * even without the extra byte. That's a pain to check |
| 921 | * for in advance, though, so we always grab an extra |
| 922 | * byte at the start, and cut it back later if possible. |
| 923 | */ |
| 924 | nbytes = (nbits >> 3) + 1; |
| 925 | if (nbytes > INT_MAX) { |
| 926 | PyErr_SetString(PyExc_OverflowError, |
| 927 | "long too large to pickle"); |
| 928 | goto error; |
| 929 | } |
| 930 | repr = PyUnicode_FromStringAndSize(NULL, (int)nbytes); |
| 931 | if (repr == NULL) |
| 932 | goto error; |
| 933 | pdata = (unsigned char *)PyUnicode_AsString(repr); |
| 934 | i = _PyLong_AsByteArray((PyLongObject *)obj, |
| 935 | pdata, nbytes, |
| 936 | 1 /* little endian */ , 1 /* signed */ ); |
| 937 | if (i < 0) |
| 938 | goto error; |
| 939 | /* If the long is negative, this may be a byte more than |
| 940 | * needed. This is so iff the MSB is all redundant sign |
| 941 | * bits. |
| 942 | */ |
| 943 | if (sign < 0 && |
| 944 | nbytes > 1 && |
| 945 | pdata[nbytes - 1] == 0xff && |
| 946 | (pdata[nbytes - 2] & 0x80) != 0) { |
| 947 | nbytes--; |
| 948 | } |
| 949 | |
| 950 | if (nbytes < 256) { |
| 951 | header[0] = LONG1; |
| 952 | header[1] = (unsigned char)nbytes; |
| 953 | size = 2; |
| 954 | } |
| 955 | else { |
| 956 | header[0] = LONG4; |
| 957 | size = (int)nbytes; |
| 958 | for (i = 1; i < 5; i++) { |
| 959 | header[i] = (unsigned char)(size & 0xff); |
| 960 | size >>= 8; |
| 961 | } |
| 962 | size = 5; |
| 963 | } |
| 964 | if (pickler_write(self, header, size) < 0 || |
| 965 | pickler_write(self, (char *)pdata, (int)nbytes) < 0) |
| 966 | goto error; |
| 967 | } |
| 968 | else { |
| 969 | char *string; |
| 970 | |
| 971 | /* proto < 2: write the repr and newline. This is quadratic-time |
| 972 | (in the number of digits), in both directions. */ |
| 973 | |
| 974 | repr = PyObject_Repr(obj); |
| 975 | if (repr == NULL) |
| 976 | goto error; |
| 977 | |
| 978 | string = PyUnicode_AsStringAndSize(repr, &size); |
| 979 | if (string == NULL) |
| 980 | goto error; |
| 981 | |
| 982 | if (pickler_write(self, &long_op, 1) < 0 || |
| 983 | pickler_write(self, string, size) < 0 || |
| 984 | pickler_write(self, "\n", 1) < 0) |
| 985 | goto error; |
| 986 | } |
| 987 | |
| 988 | if (0) { |
| 989 | error: |
| 990 | status = -1; |
| 991 | } |
| 992 | Py_XDECREF(repr); |
| 993 | |
| 994 | return status; |
| 995 | } |
| 996 | |
| 997 | static int |
| 998 | save_float(PicklerObject *self, PyObject *obj) |
| 999 | { |
| 1000 | double x = PyFloat_AS_DOUBLE((PyFloatObject *)obj); |
| 1001 | |
| 1002 | if (self->bin) { |
| 1003 | char pdata[9]; |
| 1004 | pdata[0] = BINFLOAT; |
| 1005 | if (_PyFloat_Pack8(x, (unsigned char *)&pdata[1], 0) < 0) |
| 1006 | return -1; |
| 1007 | if (pickler_write(self, pdata, 9) < 0) |
| 1008 | return -1; |
| 1009 | } |
| 1010 | else { |
| 1011 | char pdata[250]; |
| 1012 | pdata[0] = FLOAT; |
| 1013 | PyOS_ascii_formatd(pdata + 1, sizeof(pdata) - 2, "%.17g", x); |
| 1014 | /* Extend the formatted string with a newline character */ |
| 1015 | strcat(pdata, "\n"); |
| 1016 | |
| 1017 | if (pickler_write(self, pdata, strlen(pdata)) < 0) |
| 1018 | return -1; |
| 1019 | } |
| 1020 | |
| 1021 | return 0; |
| 1022 | } |
| 1023 | |
| 1024 | static int |
| 1025 | save_bytes(PicklerObject *self, PyObject *obj) |
| 1026 | { |
| 1027 | if (self->proto < 3) { |
| 1028 | /* Older pickle protocols do not have an opcode for pickling bytes |
| 1029 | objects. Therefore, we need to fake the copy protocol (i.e., |
| 1030 | the __reduce__ method) to permit bytes object unpickling. */ |
| 1031 | PyObject *reduce_value = NULL; |
| 1032 | PyObject *bytelist = NULL; |
| 1033 | int status; |
| 1034 | |
| 1035 | bytelist = PySequence_List(obj); |
| 1036 | if (bytelist == NULL) |
| 1037 | return -1; |
| 1038 | |
| 1039 | reduce_value = Py_BuildValue("(O(O))", (PyObject *)&PyBytes_Type, |
| 1040 | bytelist); |
| 1041 | if (reduce_value == NULL) { |
| 1042 | Py_DECREF(bytelist); |
| 1043 | return -1; |
| 1044 | } |
| 1045 | |
| 1046 | /* save_reduce() will memoize the object automatically. */ |
| 1047 | status = save_reduce(self, reduce_value, obj); |
| 1048 | Py_DECREF(reduce_value); |
| 1049 | Py_DECREF(bytelist); |
| 1050 | return status; |
| 1051 | } |
| 1052 | else { |
| 1053 | Py_ssize_t size; |
| 1054 | char header[5]; |
| 1055 | int len; |
| 1056 | |
| 1057 | size = PyBytes_Size(obj); |
| 1058 | if (size < 0) |
| 1059 | return -1; |
| 1060 | |
| 1061 | if (size < 256) { |
| 1062 | header[0] = SHORT_BINBYTES; |
| 1063 | header[1] = (unsigned char)size; |
| 1064 | len = 2; |
| 1065 | } |
| 1066 | else if (size <= 0xffffffffL) { |
| 1067 | header[0] = BINBYTES; |
| 1068 | header[1] = (unsigned char)(size & 0xff); |
| 1069 | header[2] = (unsigned char)((size >> 8) & 0xff); |
| 1070 | header[3] = (unsigned char)((size >> 16) & 0xff); |
| 1071 | header[4] = (unsigned char)((size >> 24) & 0xff); |
| 1072 | len = 5; |
| 1073 | } |
| 1074 | else { |
| 1075 | return -1; /* string too large */ |
| 1076 | } |
| 1077 | |
| 1078 | if (pickler_write(self, header, len) < 0) |
| 1079 | return -1; |
| 1080 | |
| 1081 | if (pickler_write(self, PyBytes_AS_STRING(obj), size) < 0) |
| 1082 | return -1; |
| 1083 | |
| 1084 | if (memo_put(self, obj) < 0) |
| 1085 | return -1; |
| 1086 | |
| 1087 | return 0; |
| 1088 | } |
| 1089 | } |
| 1090 | |
| 1091 | /* A copy of PyUnicode_EncodeRawUnicodeEscape() that also translates |
| 1092 | backslash and newline characters to \uXXXX escapes. */ |
| 1093 | static PyObject * |
| 1094 | raw_unicode_escape(const Py_UNICODE *s, Py_ssize_t size) |
| 1095 | { |
| 1096 | PyObject *repr, *result; |
| 1097 | char *p; |
| 1098 | char *q; |
| 1099 | |
| 1100 | static const char *hexdigits = "0123456789abcdef"; |
| 1101 | |
| 1102 | #ifdef Py_UNICODE_WIDE |
| 1103 | repr = PyBytes_FromStringAndSize(NULL, 10 * size); |
| 1104 | #else |
| 1105 | repr = PyBytes_FromStringAndSize(NULL, 6 * size); |
| 1106 | #endif |
| 1107 | if (repr == NULL) |
| 1108 | return NULL; |
| 1109 | if (size == 0) |
| 1110 | goto done; |
| 1111 | |
| 1112 | p = q = PyBytes_AS_STRING(repr); |
| 1113 | while (size-- > 0) { |
| 1114 | Py_UNICODE ch = *s++; |
| 1115 | #ifdef Py_UNICODE_WIDE |
| 1116 | /* Map 32-bit characters to '\Uxxxxxxxx' */ |
| 1117 | if (ch >= 0x10000) { |
| 1118 | *p++ = '\\'; |
| 1119 | *p++ = 'U'; |
| 1120 | *p++ = hexdigits[(ch >> 28) & 0xf]; |
| 1121 | *p++ = hexdigits[(ch >> 24) & 0xf]; |
| 1122 | *p++ = hexdigits[(ch >> 20) & 0xf]; |
| 1123 | *p++ = hexdigits[(ch >> 16) & 0xf]; |
| 1124 | *p++ = hexdigits[(ch >> 12) & 0xf]; |
| 1125 | *p++ = hexdigits[(ch >> 8) & 0xf]; |
| 1126 | *p++ = hexdigits[(ch >> 4) & 0xf]; |
| 1127 | *p++ = hexdigits[ch & 15]; |
| 1128 | } |
| 1129 | else |
| 1130 | #endif |
| 1131 | /* Map 16-bit characters to '\uxxxx' */ |
| 1132 | if (ch >= 256 || ch == '\\' || ch == '\n') { |
| 1133 | *p++ = '\\'; |
| 1134 | *p++ = 'u'; |
| 1135 | *p++ = hexdigits[(ch >> 12) & 0xf]; |
| 1136 | *p++ = hexdigits[(ch >> 8) & 0xf]; |
| 1137 | *p++ = hexdigits[(ch >> 4) & 0xf]; |
| 1138 | *p++ = hexdigits[ch & 15]; |
| 1139 | } |
| 1140 | /* Copy everything else as-is */ |
| 1141 | else |
| 1142 | *p++ = (char) ch; |
| 1143 | } |
| 1144 | size = p - q; |
| 1145 | |
| 1146 | done: |
| 1147 | result = PyBytes_FromStringAndSize(PyBytes_AS_STRING(repr), size); |
| 1148 | Py_DECREF(repr); |
| 1149 | return result; |
| 1150 | } |
| 1151 | |
| 1152 | static int |
| 1153 | save_unicode(PicklerObject *self, PyObject *obj) |
| 1154 | { |
| 1155 | Py_ssize_t size; |
| 1156 | PyObject *encoded = NULL; |
| 1157 | |
| 1158 | if (self->bin) { |
| 1159 | char pdata[5]; |
| 1160 | |
| 1161 | encoded = PyUnicode_AsUTF8String(obj); |
| 1162 | if (encoded == NULL) |
| 1163 | goto error; |
| 1164 | |
| 1165 | size = PyBytes_GET_SIZE(encoded); |
| 1166 | if (size < 0 || size > 0xffffffffL) |
| 1167 | goto error; /* string too large */ |
| 1168 | |
| 1169 | pdata[0] = BINUNICODE; |
| 1170 | pdata[1] = (unsigned char)(size & 0xff); |
| 1171 | pdata[2] = (unsigned char)((size >> 8) & 0xff); |
| 1172 | pdata[3] = (unsigned char)((size >> 16) & 0xff); |
| 1173 | pdata[4] = (unsigned char)((size >> 24) & 0xff); |
| 1174 | |
| 1175 | if (pickler_write(self, pdata, 5) < 0) |
| 1176 | goto error; |
| 1177 | |
| 1178 | if (pickler_write(self, PyBytes_AS_STRING(encoded), size) < 0) |
| 1179 | goto error; |
| 1180 | } |
| 1181 | else { |
| 1182 | const char unicode_op = UNICODE; |
| 1183 | |
| 1184 | encoded = raw_unicode_escape(PyUnicode_AS_UNICODE(obj), |
| 1185 | PyUnicode_GET_SIZE(obj)); |
| 1186 | if (encoded == NULL) |
| 1187 | goto error; |
| 1188 | |
| 1189 | if (pickler_write(self, &unicode_op, 1) < 0) |
| 1190 | goto error; |
| 1191 | |
| 1192 | size = PyBytes_GET_SIZE(encoded); |
| 1193 | if (pickler_write(self, PyBytes_AS_STRING(encoded), size) < 0) |
| 1194 | goto error; |
| 1195 | |
| 1196 | if (pickler_write(self, "\n", 1) < 0) |
| 1197 | goto error; |
| 1198 | } |
| 1199 | if (memo_put(self, obj) < 0) |
| 1200 | goto error; |
| 1201 | |
| 1202 | Py_DECREF(encoded); |
| 1203 | return 0; |
| 1204 | |
| 1205 | error: |
| 1206 | Py_XDECREF(encoded); |
| 1207 | return -1; |
| 1208 | } |
| 1209 | |
| 1210 | /* A helper for save_tuple. Push the len elements in tuple t on the stack. */ |
| 1211 | static int |
| 1212 | store_tuple_elements(PicklerObject *self, PyObject *t, int len) |
| 1213 | { |
| 1214 | int i; |
| 1215 | |
| 1216 | assert(PyTuple_Size(t) == len); |
| 1217 | |
| 1218 | for (i = 0; i < len; i++) { |
| 1219 | PyObject *element = PyTuple_GET_ITEM(t, i); |
| 1220 | |
| 1221 | if (element == NULL) |
| 1222 | return -1; |
| 1223 | if (save(self, element, 0) < 0) |
| 1224 | return -1; |
| 1225 | } |
| 1226 | |
| 1227 | return 0; |
| 1228 | } |
| 1229 | |
| 1230 | /* Tuples are ubiquitous in the pickle protocols, so many techniques are |
| 1231 | * used across protocols to minimize the space needed to pickle them. |
| 1232 | * Tuples are also the only builtin immutable type that can be recursive |
| 1233 | * (a tuple can be reached from itself), and that requires some subtle |
| 1234 | * magic so that it works in all cases. IOW, this is a long routine. |
| 1235 | */ |
| 1236 | static int |
| 1237 | save_tuple(PicklerObject *self, PyObject *obj) |
| 1238 | { |
| 1239 | PyObject *memo_key = NULL; |
| 1240 | int len, i; |
| 1241 | int status = 0; |
| 1242 | |
| 1243 | const char mark_op = MARK; |
| 1244 | const char tuple_op = TUPLE; |
| 1245 | const char pop_op = POP; |
| 1246 | const char pop_mark_op = POP_MARK; |
| 1247 | const char len2opcode[] = {EMPTY_TUPLE, TUPLE1, TUPLE2, TUPLE3}; |
| 1248 | |
| 1249 | if ((len = PyTuple_Size(obj)) < 0) |
| 1250 | return -1; |
| 1251 | |
| 1252 | if (len == 0) { |
| 1253 | char pdata[2]; |
| 1254 | |
| 1255 | if (self->proto) { |
| 1256 | pdata[0] = EMPTY_TUPLE; |
| 1257 | len = 1; |
| 1258 | } |
| 1259 | else { |
| 1260 | pdata[0] = MARK; |
| 1261 | pdata[1] = TUPLE; |
| 1262 | len = 2; |
| 1263 | } |
| 1264 | if (pickler_write(self, pdata, len) < 0) |
| 1265 | return -1; |
| 1266 | return 0; |
| 1267 | } |
| 1268 | |
| 1269 | /* id(tuple) isn't in the memo now. If it shows up there after |
| 1270 | * saving the tuple elements, the tuple must be recursive, in |
| 1271 | * which case we'll pop everything we put on the stack, and fetch |
| 1272 | * its value from the memo. |
| 1273 | */ |
| 1274 | memo_key = PyLong_FromVoidPtr(obj); |
| 1275 | if (memo_key == NULL) |
| 1276 | return -1; |
| 1277 | |
| 1278 | if (len <= 3 && self->proto >= 2) { |
| 1279 | /* Use TUPLE{1,2,3} opcodes. */ |
| 1280 | if (store_tuple_elements(self, obj, len) < 0) |
| 1281 | goto error; |
| 1282 | |
| 1283 | if (PyDict_GetItem(self->memo, memo_key)) { |
| 1284 | /* pop the len elements */ |
| 1285 | for (i = 0; i < len; i++) |
| 1286 | if (pickler_write(self, &pop_op, 1) < 0) |
| 1287 | goto error; |
| 1288 | /* fetch from memo */ |
| 1289 | if (memo_get(self, memo_key) < 0) |
| 1290 | goto error; |
| 1291 | |
| 1292 | Py_DECREF(memo_key); |
| 1293 | return 0; |
| 1294 | } |
| 1295 | else { /* Not recursive. */ |
| 1296 | if (pickler_write(self, len2opcode + len, 1) < 0) |
| 1297 | goto error; |
| 1298 | } |
| 1299 | goto memoize; |
| 1300 | } |
| 1301 | |
| 1302 | /* proto < 2 and len > 0, or proto >= 2 and len > 3. |
| 1303 | * Generate MARK e1 e2 ... TUPLE |
| 1304 | */ |
| 1305 | if (pickler_write(self, &mark_op, 1) < 0) |
| 1306 | goto error; |
| 1307 | |
| 1308 | if (store_tuple_elements(self, obj, len) < 0) |
| 1309 | goto error; |
| 1310 | |
| 1311 | if (PyDict_GetItem(self->memo, memo_key)) { |
| 1312 | /* pop the stack stuff we pushed */ |
| 1313 | if (self->bin) { |
| 1314 | if (pickler_write(self, &pop_mark_op, 1) < 0) |
| 1315 | goto error; |
| 1316 | } |
| 1317 | else { |
| 1318 | /* Note that we pop one more than len, to remove |
| 1319 | * the MARK too. |
| 1320 | */ |
| 1321 | for (i = 0; i <= len; i++) |
| 1322 | if (pickler_write(self, &pop_op, 1) < 0) |
| 1323 | goto error; |
| 1324 | } |
| 1325 | /* fetch from memo */ |
| 1326 | if (memo_get(self, memo_key) < 0) |
| 1327 | goto error; |
| 1328 | |
| 1329 | Py_DECREF(memo_key); |
| 1330 | return 0; |
| 1331 | } |
| 1332 | else { /* Not recursive. */ |
| 1333 | if (pickler_write(self, &tuple_op, 1) < 0) |
| 1334 | goto error; |
| 1335 | } |
| 1336 | |
| 1337 | memoize: |
| 1338 | if (memo_put(self, obj) < 0) |
| 1339 | goto error; |
| 1340 | |
| 1341 | if (0) { |
| 1342 | error: |
| 1343 | status = -1; |
| 1344 | } |
| 1345 | |
| 1346 | Py_DECREF(memo_key); |
| 1347 | return status; |
| 1348 | } |
| 1349 | |
| 1350 | /* iter is an iterator giving items, and we batch up chunks of |
| 1351 | * MARK item item ... item APPENDS |
| 1352 | * opcode sequences. Calling code should have arranged to first create an |
| 1353 | * empty list, or list-like object, for the APPENDS to operate on. |
| 1354 | * Returns 0 on success, <0 on error. |
| 1355 | */ |
| 1356 | static int |
| 1357 | batch_list(PicklerObject *self, PyObject *iter) |
| 1358 | { |
| 1359 | PyObject *obj; |
| 1360 | PyObject *slice[BATCHSIZE]; |
| 1361 | int i, n; |
| 1362 | |
| 1363 | const char mark_op = MARK; |
| 1364 | const char append_op = APPEND; |
| 1365 | const char appends_op = APPENDS; |
| 1366 | |
| 1367 | assert(iter != NULL); |
| 1368 | |
| 1369 | /* XXX: I think this function could be made faster by avoiding the |
| 1370 | iterator interface and fetching objects directly from list using |
| 1371 | PyList_GET_ITEM. |
| 1372 | */ |
| 1373 | |
| 1374 | if (self->proto == 0) { |
| 1375 | /* APPENDS isn't available; do one at a time. */ |
| 1376 | for (;;) { |
| 1377 | obj = PyIter_Next(iter); |
| 1378 | if (obj == NULL) { |
| 1379 | if (PyErr_Occurred()) |
| 1380 | return -1; |
| 1381 | break; |
| 1382 | } |
| 1383 | i = save(self, obj, 0); |
| 1384 | Py_DECREF(obj); |
| 1385 | if (i < 0) |
| 1386 | return -1; |
| 1387 | if (pickler_write(self, &append_op, 1) < 0) |
| 1388 | return -1; |
| 1389 | } |
| 1390 | return 0; |
| 1391 | } |
| 1392 | |
| 1393 | /* proto > 0: write in batches of BATCHSIZE. */ |
| 1394 | do { |
| 1395 | /* Get next group of (no more than) BATCHSIZE elements. */ |
| 1396 | for (n = 0; n < BATCHSIZE; n++) { |
| 1397 | obj = PyIter_Next(iter); |
| 1398 | if (obj == NULL) { |
| 1399 | if (PyErr_Occurred()) |
| 1400 | goto error; |
| 1401 | break; |
| 1402 | } |
| 1403 | slice[n] = obj; |
| 1404 | } |
| 1405 | |
| 1406 | if (n > 1) { |
| 1407 | /* Pump out MARK, slice[0:n], APPENDS. */ |
| 1408 | if (pickler_write(self, &mark_op, 1) < 0) |
| 1409 | goto error; |
| 1410 | for (i = 0; i < n; i++) { |
| 1411 | if (save(self, slice[i], 0) < 0) |
| 1412 | goto error; |
| 1413 | } |
| 1414 | if (pickler_write(self, &appends_op, 1) < 0) |
| 1415 | goto error; |
| 1416 | } |
| 1417 | else if (n == 1) { |
| 1418 | if (save(self, slice[0], 0) < 0 || |
| 1419 | pickler_write(self, &append_op, 1) < 0) |
| 1420 | goto error; |
| 1421 | } |
| 1422 | |
| 1423 | for (i = 0; i < n; i++) { |
| 1424 | Py_DECREF(slice[i]); |
| 1425 | } |
| 1426 | } while (n == BATCHSIZE); |
| 1427 | return 0; |
| 1428 | |
| 1429 | error: |
| 1430 | while (--n >= 0) { |
| 1431 | Py_DECREF(slice[n]); |
| 1432 | } |
| 1433 | return -1; |
| 1434 | } |
| 1435 | |
| 1436 | static int |
| 1437 | save_list(PicklerObject *self, PyObject *obj) |
| 1438 | { |
| 1439 | PyObject *iter; |
| 1440 | char header[3]; |
| 1441 | int len; |
| 1442 | int status = 0; |
| 1443 | |
| 1444 | if (self->fast && !fast_save_enter(self, obj)) |
| 1445 | goto error; |
| 1446 | |
| 1447 | /* Create an empty list. */ |
| 1448 | if (self->bin) { |
| 1449 | header[0] = EMPTY_LIST; |
| 1450 | len = 1; |
| 1451 | } |
| 1452 | else { |
| 1453 | header[0] = MARK; |
| 1454 | header[1] = LIST; |
| 1455 | len = 2; |
| 1456 | } |
| 1457 | |
| 1458 | if (pickler_write(self, header, len) < 0) |
| 1459 | goto error; |
| 1460 | |
| 1461 | /* Get list length, and bow out early if empty. */ |
| 1462 | if ((len = PyList_Size(obj)) < 0) |
| 1463 | goto error; |
| 1464 | |
| 1465 | if (memo_put(self, obj) < 0) |
| 1466 | goto error; |
| 1467 | |
| 1468 | if (len != 0) { |
| 1469 | /* Save the list elements. */ |
| 1470 | iter = PyObject_GetIter(obj); |
| 1471 | if (iter == NULL) |
| 1472 | goto error; |
| 1473 | status = batch_list(self, iter); |
| 1474 | Py_DECREF(iter); |
| 1475 | } |
| 1476 | |
| 1477 | if (0) { |
| 1478 | error: |
| 1479 | status = -1; |
| 1480 | } |
| 1481 | |
| 1482 | if (self->fast && !fast_save_leave(self, obj)) |
| 1483 | status = -1; |
| 1484 | |
| 1485 | return status; |
| 1486 | } |
| 1487 | |
| 1488 | /* iter is an iterator giving (key, value) pairs, and we batch up chunks of |
| 1489 | * MARK key value ... key value SETITEMS |
| 1490 | * opcode sequences. Calling code should have arranged to first create an |
| 1491 | * empty dict, or dict-like object, for the SETITEMS to operate on. |
| 1492 | * Returns 0 on success, <0 on error. |
| 1493 | * |
| 1494 | * This is very much like batch_list(). The difference between saving |
| 1495 | * elements directly, and picking apart two-tuples, is so long-winded at |
| 1496 | * the C level, though, that attempts to combine these routines were too |
| 1497 | * ugly to bear. |
| 1498 | */ |
| 1499 | static int |
| 1500 | batch_dict(PicklerObject *self, PyObject *iter) |
| 1501 | { |
| 1502 | PyObject *obj; |
| 1503 | PyObject *slice[BATCHSIZE]; |
| 1504 | int i, n; |
| 1505 | |
| 1506 | const char mark_op = MARK; |
| 1507 | const char setitem_op = SETITEM; |
| 1508 | const char setitems_op = SETITEMS; |
| 1509 | |
| 1510 | assert(iter != NULL); |
| 1511 | |
| 1512 | if (self->proto == 0) { |
| 1513 | /* SETITEMS isn't available; do one at a time. */ |
| 1514 | for (;;) { |
| 1515 | obj = PyIter_Next(iter); |
| 1516 | if (obj == NULL) { |
| 1517 | if (PyErr_Occurred()) |
| 1518 | return -1; |
| 1519 | break; |
| 1520 | } |
| 1521 | if (!PyTuple_Check(obj) || PyTuple_Size(obj) != 2) { |
| 1522 | PyErr_SetString(PyExc_TypeError, "dict items " |
| 1523 | "iterator must return 2-tuples"); |
| 1524 | return -1; |
| 1525 | } |
| 1526 | i = save(self, PyTuple_GET_ITEM(obj, 0), 0); |
| 1527 | if (i >= 0) |
| 1528 | i = save(self, PyTuple_GET_ITEM(obj, 1), 0); |
| 1529 | Py_DECREF(obj); |
| 1530 | if (i < 0) |
| 1531 | return -1; |
| 1532 | if (pickler_write(self, &setitem_op, 1) < 0) |
| 1533 | return -1; |
| 1534 | } |
| 1535 | return 0; |
| 1536 | } |
| 1537 | |
| 1538 | /* proto > 0: write in batches of BATCHSIZE. */ |
| 1539 | do { |
| 1540 | /* Get next group of (no more than) BATCHSIZE elements. */ |
| 1541 | for (n = 0; n < BATCHSIZE; n++) { |
| 1542 | obj = PyIter_Next(iter); |
| 1543 | if (obj == NULL) { |
| 1544 | if (PyErr_Occurred()) |
| 1545 | goto error; |
| 1546 | break; |
| 1547 | } |
| 1548 | if (!PyTuple_Check(obj) || PyTuple_Size(obj) != 2) { |
| 1549 | PyErr_SetString(PyExc_TypeError, "dict items " |
| 1550 | "iterator must return 2-tuples"); |
| 1551 | goto error; |
| 1552 | } |
| 1553 | slice[n] = obj; |
| 1554 | } |
| 1555 | |
| 1556 | if (n > 1) { |
| 1557 | /* Pump out MARK, slice[0:n], SETITEMS. */ |
| 1558 | if (pickler_write(self, &mark_op, 1) < 0) |
| 1559 | goto error; |
| 1560 | for (i = 0; i < n; i++) { |
| 1561 | obj = slice[i]; |
| 1562 | if (save(self, PyTuple_GET_ITEM(obj, 0), 0) < 0 || |
| 1563 | save(self, PyTuple_GET_ITEM(obj, 1), 0) < 0) |
| 1564 | goto error; |
| 1565 | } |
| 1566 | if (pickler_write(self, &setitems_op, 1) < 0) |
| 1567 | goto error; |
| 1568 | } |
| 1569 | else if (n == 1) { |
| 1570 | obj = slice[0]; |
| 1571 | if (save(self, PyTuple_GET_ITEM(obj, 0), 0) < 0 || |
| 1572 | save(self, PyTuple_GET_ITEM(obj, 1), 0) < 0 || |
| 1573 | pickler_write(self, &setitem_op, 1) < 0) |
| 1574 | goto error; |
| 1575 | } |
| 1576 | |
| 1577 | for (i = 0; i < n; i++) { |
| 1578 | Py_DECREF(slice[i]); |
| 1579 | } |
| 1580 | } while (n == BATCHSIZE); |
| 1581 | return 0; |
| 1582 | |
| 1583 | error: |
| 1584 | while (--n >= 0) { |
| 1585 | Py_DECREF(slice[n]); |
| 1586 | } |
| 1587 | return -1; |
| 1588 | } |
| 1589 | |
| 1590 | static int |
| 1591 | save_dict(PicklerObject *self, PyObject *obj) |
| 1592 | { |
| 1593 | PyObject *items, *iter; |
| 1594 | char header[3]; |
| 1595 | int len; |
| 1596 | int status = 0; |
| 1597 | |
| 1598 | if (self->fast && !fast_save_enter(self, obj)) |
| 1599 | goto error; |
| 1600 | |
| 1601 | /* Create an empty dict. */ |
| 1602 | if (self->bin) { |
| 1603 | header[0] = EMPTY_DICT; |
| 1604 | len = 1; |
| 1605 | } |
| 1606 | else { |
| 1607 | header[0] = MARK; |
| 1608 | header[1] = DICT; |
| 1609 | len = 2; |
| 1610 | } |
| 1611 | |
| 1612 | if (pickler_write(self, header, len) < 0) |
| 1613 | goto error; |
| 1614 | |
| 1615 | /* Get dict size, and bow out early if empty. */ |
| 1616 | if ((len = PyDict_Size(obj)) < 0) |
| 1617 | goto error; |
| 1618 | |
| 1619 | if (memo_put(self, obj) < 0) |
| 1620 | goto error; |
| 1621 | |
| 1622 | if (len != 0) { |
| 1623 | /* Save the dict items. */ |
| 1624 | items = PyObject_CallMethod(obj, "items", "()"); |
| 1625 | if (items == NULL) |
| 1626 | goto error; |
| 1627 | iter = PyObject_GetIter(items); |
| 1628 | Py_DECREF(items); |
| 1629 | if (iter == NULL) |
| 1630 | goto error; |
| 1631 | status = batch_dict(self, iter); |
| 1632 | Py_DECREF(iter); |
| 1633 | } |
| 1634 | |
| 1635 | if (0) { |
| 1636 | error: |
| 1637 | status = -1; |
| 1638 | } |
| 1639 | |
| 1640 | if (self->fast && !fast_save_leave(self, obj)) |
| 1641 | status = -1; |
| 1642 | |
| 1643 | return status; |
| 1644 | } |
| 1645 | |
| 1646 | static int |
| 1647 | save_global(PicklerObject *self, PyObject *obj, PyObject *name) |
| 1648 | { |
| 1649 | static PyObject *name_str = NULL; |
| 1650 | PyObject *global_name = NULL; |
| 1651 | PyObject *module_name = NULL; |
| 1652 | PyObject *module = NULL; |
| 1653 | PyObject *cls; |
| 1654 | int status = 0; |
| 1655 | |
| 1656 | const char global_op = GLOBAL; |
| 1657 | |
| 1658 | if (name_str == NULL) { |
| 1659 | name_str = PyUnicode_InternFromString("__name__"); |
| 1660 | if (name_str == NULL) |
| 1661 | goto error; |
| 1662 | } |
| 1663 | |
| 1664 | if (name) { |
| 1665 | global_name = name; |
| 1666 | Py_INCREF(global_name); |
| 1667 | } |
| 1668 | else { |
| 1669 | global_name = PyObject_GetAttr(obj, name_str); |
| 1670 | if (global_name == NULL) |
| 1671 | goto error; |
| 1672 | } |
| 1673 | |
| 1674 | module_name = whichmodule(obj, global_name); |
| 1675 | if (module_name == NULL) |
| 1676 | goto error; |
| 1677 | |
| 1678 | /* XXX: Change to use the import C API directly with level=0 to disallow |
| 1679 | relative imports. |
| 1680 | |
| 1681 | XXX: PyImport_ImportModuleLevel could be used. However, this bypasses |
| 1682 | builtins.__import__. Therefore, _pickle, unlike pickle.py, will ignore |
| 1683 | custom import functions (IMHO, this would be a nice security |
| 1684 | feature). The import C API would need to be extended to support the |
| 1685 | extra parameters of __import__ to fix that. */ |
| 1686 | module = PyImport_Import(module_name); |
| 1687 | if (module == NULL) { |
| 1688 | PyErr_Format(PicklingError, |
| 1689 | "Can't pickle %R: import of module %R failed", |
| 1690 | obj, module_name); |
| 1691 | goto error; |
| 1692 | } |
| 1693 | cls = PyObject_GetAttr(module, global_name); |
| 1694 | if (cls == NULL) { |
| 1695 | PyErr_Format(PicklingError, |
| 1696 | "Can't pickle %R: attribute lookup %S.%S failed", |
| 1697 | obj, module_name, global_name); |
| 1698 | goto error; |
| 1699 | } |
| 1700 | if (cls != obj) { |
| 1701 | Py_DECREF(cls); |
| 1702 | PyErr_Format(PicklingError, |
| 1703 | "Can't pickle %R: it's not the same object as %S.%S", |
| 1704 | obj, module_name, global_name); |
| 1705 | goto error; |
| 1706 | } |
| 1707 | Py_DECREF(cls); |
| 1708 | |
| 1709 | if (self->proto >= 2) { |
| 1710 | /* See whether this is in the extension registry, and if |
| 1711 | * so generate an EXT opcode. |
| 1712 | */ |
| 1713 | PyObject *code_obj; /* extension code as Python object */ |
| 1714 | long code; /* extension code as C value */ |
| 1715 | char pdata[5]; |
| 1716 | int n; |
| 1717 | |
| 1718 | PyTuple_SET_ITEM(two_tuple, 0, module_name); |
| 1719 | PyTuple_SET_ITEM(two_tuple, 1, global_name); |
| 1720 | code_obj = PyDict_GetItem(extension_registry, two_tuple); |
| 1721 | /* The object is not registered in the extension registry. |
| 1722 | This is the most likely code path. */ |
| 1723 | if (code_obj == NULL) |
| 1724 | goto gen_global; |
| 1725 | |
| 1726 | /* XXX: pickle.py doesn't check neither the type, nor the range |
| 1727 | of the value returned by the extension_registry. It should for |
| 1728 | consistency. */ |
| 1729 | |
| 1730 | /* Verify code_obj has the right type and value. */ |
| 1731 | if (!PyLong_Check(code_obj)) { |
| 1732 | PyErr_Format(PicklingError, |
| 1733 | "Can't pickle %R: extension code %R isn't an integer", |
| 1734 | obj, code_obj); |
| 1735 | goto error; |
| 1736 | } |
| 1737 | code = PyLong_AS_LONG(code_obj); |
| 1738 | if (code <= 0 || code > 0x7fffffffL) { |
| 1739 | PyErr_Format(PicklingError, |
| 1740 | "Can't pickle %R: extension code %ld is out of range", |
| 1741 | obj, code); |
| 1742 | goto error; |
| 1743 | } |
| 1744 | |
| 1745 | /* Generate an EXT opcode. */ |
| 1746 | if (code <= 0xff) { |
| 1747 | pdata[0] = EXT1; |
| 1748 | pdata[1] = (unsigned char)code; |
| 1749 | n = 2; |
| 1750 | } |
| 1751 | else if (code <= 0xffff) { |
| 1752 | pdata[0] = EXT2; |
| 1753 | pdata[1] = (unsigned char)(code & 0xff); |
| 1754 | pdata[2] = (unsigned char)((code >> 8) & 0xff); |
| 1755 | n = 3; |
| 1756 | } |
| 1757 | else { |
| 1758 | pdata[0] = EXT4; |
| 1759 | pdata[1] = (unsigned char)(code & 0xff); |
| 1760 | pdata[2] = (unsigned char)((code >> 8) & 0xff); |
| 1761 | pdata[3] = (unsigned char)((code >> 16) & 0xff); |
| 1762 | pdata[4] = (unsigned char)((code >> 24) & 0xff); |
| 1763 | n = 5; |
| 1764 | } |
| 1765 | |
| 1766 | if (pickler_write(self, pdata, n) < 0) |
| 1767 | goto error; |
| 1768 | } |
| 1769 | else { |
| 1770 | /* Generate a normal global opcode if we are using a pickle |
| 1771 | protocol <= 2, or if the object is not registered in the |
| 1772 | extension registry. */ |
| 1773 | PyObject *encoded; |
| 1774 | PyObject *(*unicode_encoder)(PyObject *); |
| 1775 | |
| 1776 | gen_global: |
| 1777 | if (pickler_write(self, &global_op, 1) < 0) |
| 1778 | goto error; |
| 1779 | |
| 1780 | /* Since Python 3.0 now supports non-ASCII identifiers, we encode both |
| 1781 | the module name and the global name using UTF-8. We do so only when |
| 1782 | we are using the pickle protocol newer than version 3. This is to |
| 1783 | ensure compatibility with older Unpickler running on Python 2.x. */ |
| 1784 | if (self->proto >= 3) { |
| 1785 | unicode_encoder = PyUnicode_AsUTF8String; |
| 1786 | } |
| 1787 | else { |
| 1788 | unicode_encoder = PyUnicode_AsASCIIString; |
| 1789 | } |
| 1790 | |
| 1791 | /* Save the name of the module. */ |
| 1792 | encoded = unicode_encoder(module_name); |
| 1793 | if (encoded == NULL) { |
| 1794 | if (PyErr_ExceptionMatches(PyExc_UnicodeEncodeError)) |
| 1795 | PyErr_Format(PicklingError, |
| 1796 | "can't pickle module identifier '%S' using " |
| 1797 | "pickle protocol %i", module_name, self->proto); |
| 1798 | goto error; |
| 1799 | } |
| 1800 | if (pickler_write(self, PyBytes_AS_STRING(encoded), |
| 1801 | PyBytes_GET_SIZE(encoded)) < 0) { |
| 1802 | Py_DECREF(encoded); |
| 1803 | goto error; |
| 1804 | } |
| 1805 | Py_DECREF(encoded); |
| 1806 | if(pickler_write(self, "\n", 1) < 0) |
| 1807 | goto error; |
| 1808 | |
| 1809 | /* Save the name of the module. */ |
| 1810 | encoded = unicode_encoder(global_name); |
| 1811 | if (encoded == NULL) { |
| 1812 | if (PyErr_ExceptionMatches(PyExc_UnicodeEncodeError)) |
| 1813 | PyErr_Format(PicklingError, |
| 1814 | "can't pickle global identifier '%S' using " |
| 1815 | "pickle protocol %i", global_name, self->proto); |
| 1816 | goto error; |
| 1817 | } |
| 1818 | if (pickler_write(self, PyBytes_AS_STRING(encoded), |
| 1819 | PyBytes_GET_SIZE(encoded)) < 0) { |
| 1820 | Py_DECREF(encoded); |
| 1821 | goto error; |
| 1822 | } |
| 1823 | Py_DECREF(encoded); |
| 1824 | if(pickler_write(self, "\n", 1) < 0) |
| 1825 | goto error; |
| 1826 | |
| 1827 | /* Memoize the object. */ |
| 1828 | if (memo_put(self, obj) < 0) |
| 1829 | goto error; |
| 1830 | } |
| 1831 | |
| 1832 | if (0) { |
| 1833 | error: |
| 1834 | status = -1; |
| 1835 | } |
| 1836 | Py_XDECREF(module_name); |
| 1837 | Py_XDECREF(global_name); |
| 1838 | Py_XDECREF(module); |
| 1839 | |
| 1840 | return status; |
| 1841 | } |
| 1842 | |
| 1843 | static int |
| 1844 | save_pers(PicklerObject *self, PyObject *obj, PyObject *func) |
| 1845 | { |
| 1846 | PyObject *pid = NULL; |
| 1847 | int status = 0; |
| 1848 | |
| 1849 | const char persid_op = PERSID; |
| 1850 | const char binpersid_op = BINPERSID; |
| 1851 | |
| 1852 | Py_INCREF(obj); |
| 1853 | pid = pickler_call(self, func, obj); |
| 1854 | if (pid == NULL) |
| 1855 | return -1; |
| 1856 | |
| 1857 | if (pid != Py_None) { |
| 1858 | if (self->bin) { |
| 1859 | if (save(self, pid, 1) < 0 || |
| 1860 | pickler_write(self, &binpersid_op, 1) < 0) |
| 1861 | goto error; |
| 1862 | } |
| 1863 | else { |
| 1864 | PyObject *pid_str = NULL; |
| 1865 | char *pid_ascii_bytes; |
| 1866 | Py_ssize_t size; |
| 1867 | |
| 1868 | pid_str = PyObject_Str(pid); |
| 1869 | if (pid_str == NULL) |
| 1870 | goto error; |
| 1871 | |
| 1872 | /* XXX: Should it check whether the persistent id only contains |
| 1873 | ASCII characters? And what if the pid contains embedded |
| 1874 | newlines? */ |
| 1875 | pid_ascii_bytes = PyUnicode_AsStringAndSize(pid_str, &size); |
| 1876 | Py_DECREF(pid_str); |
| 1877 | if (pid_ascii_bytes == NULL) |
| 1878 | goto error; |
| 1879 | |
| 1880 | if (pickler_write(self, &persid_op, 1) < 0 || |
| 1881 | pickler_write(self, pid_ascii_bytes, size) < 0 || |
| 1882 | pickler_write(self, "\n", 1) < 0) |
| 1883 | goto error; |
| 1884 | } |
| 1885 | status = 1; |
| 1886 | } |
| 1887 | |
| 1888 | if (0) { |
| 1889 | error: |
| 1890 | status = -1; |
| 1891 | } |
| 1892 | Py_XDECREF(pid); |
| 1893 | |
| 1894 | return status; |
| 1895 | } |
| 1896 | |
| 1897 | /* We're saving obj, and args is the 2-thru-5 tuple returned by the |
| 1898 | * appropriate __reduce__ method for obj. |
| 1899 | */ |
| 1900 | static int |
| 1901 | save_reduce(PicklerObject *self, PyObject *args, PyObject *obj) |
| 1902 | { |
| 1903 | PyObject *callable; |
| 1904 | PyObject *argtup; |
| 1905 | PyObject *state = NULL; |
| 1906 | PyObject *listitems = NULL; |
| 1907 | PyObject *dictitems = NULL; |
| 1908 | |
| 1909 | int use_newobj = self->proto >= 2; |
| 1910 | |
| 1911 | const char reduce_op = REDUCE; |
| 1912 | const char build_op = BUILD; |
| 1913 | const char newobj_op = NEWOBJ; |
| 1914 | |
| 1915 | if (!PyArg_UnpackTuple(args, "save_reduce", 2, 5, |
| 1916 | &callable, &argtup, &state, &listitems, &dictitems)) |
| 1917 | return -1; |
| 1918 | |
| 1919 | if (!PyCallable_Check(callable)) { |
| 1920 | PyErr_SetString(PicklingError, |
| 1921 | "first argument of save_reduce() must be callable"); |
| 1922 | return -1; |
| 1923 | } |
| 1924 | if (!PyTuple_Check(argtup)) { |
| 1925 | PyErr_SetString(PicklingError, |
| 1926 | "second argument of save_reduce() must be a tuple"); |
| 1927 | return -1; |
| 1928 | } |
| 1929 | |
| 1930 | if (state == Py_None) |
| 1931 | state = NULL; |
| 1932 | if (listitems == Py_None) |
| 1933 | listitems = NULL; |
| 1934 | if (dictitems == Py_None) |
| 1935 | dictitems = NULL; |
| 1936 | |
| 1937 | /* Protocol 2 special case: if callable's name is __newobj__, use |
| 1938 | NEWOBJ. */ |
| 1939 | if (use_newobj) { |
| 1940 | static PyObject *newobj_str = NULL; |
| 1941 | PyObject *name_str; |
| 1942 | |
| 1943 | if (newobj_str == NULL) { |
| 1944 | newobj_str = PyUnicode_InternFromString("__newobj__"); |
| 1945 | } |
| 1946 | |
| 1947 | name_str = PyObject_GetAttrString(callable, "__name__"); |
| 1948 | if (name_str == NULL) { |
| 1949 | if (PyErr_ExceptionMatches(PyExc_AttributeError)) |
| 1950 | PyErr_Clear(); |
| 1951 | else |
| 1952 | return -1; |
| 1953 | use_newobj = 0; |
| 1954 | } |
| 1955 | else { |
| 1956 | use_newobj = PyUnicode_Check(name_str) && |
| 1957 | PyUnicode_Compare(name_str, newobj_str) == 0; |
| 1958 | Py_DECREF(name_str); |
| 1959 | } |
| 1960 | } |
| 1961 | if (use_newobj) { |
| 1962 | PyObject *cls; |
| 1963 | PyObject *newargtup; |
| 1964 | PyObject *obj_class; |
| 1965 | int p; |
| 1966 | |
| 1967 | /* Sanity checks. */ |
| 1968 | if (Py_SIZE(argtup) < 1) { |
| 1969 | PyErr_SetString(PicklingError, "__newobj__ arglist is empty"); |
| 1970 | return -1; |
| 1971 | } |
| 1972 | |
| 1973 | cls = PyTuple_GET_ITEM(argtup, 0); |
| 1974 | if (!PyObject_HasAttrString(cls, "__new__")) { |
| 1975 | PyErr_SetString(PicklingError, "args[0] from " |
| 1976 | "__newobj__ args has no __new__"); |
| 1977 | return -1; |
| 1978 | } |
| 1979 | |
| 1980 | if (obj != NULL) { |
| 1981 | obj_class = PyObject_GetAttrString(obj, "__class__"); |
| 1982 | if (obj_class == NULL) { |
| 1983 | if (PyErr_ExceptionMatches(PyExc_AttributeError)) |
| 1984 | PyErr_Clear(); |
| 1985 | else |
| 1986 | return -1; |
| 1987 | } |
| 1988 | p = obj_class != cls; /* true iff a problem */ |
| 1989 | Py_DECREF(obj_class); |
| 1990 | if (p) { |
| 1991 | PyErr_SetString(PicklingError, "args[0] from " |
| 1992 | "__newobj__ args has the wrong class"); |
| 1993 | return -1; |
| 1994 | } |
| 1995 | } |
| 1996 | /* XXX: These calls save() are prone to infinite recursion. Imagine |
| 1997 | what happen if the value returned by the __reduce__() method of |
| 1998 | some extension type contains another object of the same type. Ouch! |
| 1999 | |
| 2000 | Here is a quick example, that I ran into, to illustrate what I |
| 2001 | mean: |
| 2002 | |
| 2003 | >>> import pickle, copyreg |
| 2004 | >>> copyreg.dispatch_table.pop(complex) |
| 2005 | >>> pickle.dumps(1+2j) |
| 2006 | Traceback (most recent call last): |
| 2007 | ... |
| 2008 | RuntimeError: maximum recursion depth exceeded |
| 2009 | |
| 2010 | Removing the complex class from copyreg.dispatch_table made the |
| 2011 | __reduce_ex__() method emit another complex object: |
| 2012 | |
| 2013 | >>> (1+1j).__reduce_ex__(2) |
| 2014 | (<function __newobj__ at 0xb7b71c3c>, |
| 2015 | (<class 'complex'>, (1+1j)), None, None, None) |
| 2016 | |
| 2017 | Thus when save() was called on newargstup (the 2nd item) recursion |
| 2018 | ensued. Of course, the bug was in the complex class which had a |
| 2019 | broken __getnewargs__() that emitted another complex object. But, |
| 2020 | the point, here, is it is quite easy to end up with a broken reduce |
| 2021 | function. */ |
| 2022 | |
| 2023 | /* Save the class and its __new__ arguments. */ |
| 2024 | if (save(self, cls, 0) < 0) |
| 2025 | return -1; |
| 2026 | |
| 2027 | newargtup = PyTuple_GetSlice(argtup, 1, Py_SIZE(argtup)); |
| 2028 | if (newargtup == NULL) |
| 2029 | return -1; |
| 2030 | |
| 2031 | p = save(self, newargtup, 0); |
| 2032 | Py_DECREF(newargtup); |
| 2033 | if (p < 0) |
| 2034 | return -1; |
| 2035 | |
| 2036 | /* Add NEWOBJ opcode. */ |
| 2037 | if (pickler_write(self, &newobj_op, 1) < 0) |
| 2038 | return -1; |
| 2039 | } |
| 2040 | else { /* Not using NEWOBJ. */ |
| 2041 | if (save(self, callable, 0) < 0 || |
| 2042 | save(self, argtup, 0) < 0 || |
| 2043 | pickler_write(self, &reduce_op, 1) < 0) |
| 2044 | return -1; |
| 2045 | } |
| 2046 | |
| 2047 | /* obj can be NULL when save_reduce() is used directly. A NULL obj means |
| 2048 | the caller do not want to memoize the object. Not particularly useful, |
| 2049 | but that is to mimic the behavior save_reduce() in pickle.py when |
| 2050 | obj is None. */ |
| 2051 | if (obj && memo_put(self, obj) < 0) |
| 2052 | return -1; |
| 2053 | |
| 2054 | if (listitems && batch_list(self, listitems) < 0) |
| 2055 | return -1; |
| 2056 | |
| 2057 | if (dictitems && batch_dict(self, dictitems) < 0) |
| 2058 | return -1; |
| 2059 | |
| 2060 | if (state) { |
| 2061 | if (save(self, state, 0) < 0 || |
| 2062 | pickler_write(self, &build_op, 1) < 0) |
| 2063 | return -1; |
| 2064 | } |
| 2065 | |
| 2066 | return 0; |
| 2067 | } |
| 2068 | |
| 2069 | static int |
| 2070 | save(PicklerObject *self, PyObject *obj, int pers_save) |
| 2071 | { |
| 2072 | PyTypeObject *type; |
| 2073 | PyObject *reduce_func = NULL; |
| 2074 | PyObject *reduce_value = NULL; |
| 2075 | PyObject *memo_key = NULL; |
| 2076 | int status = 0; |
| 2077 | |
| 2078 | /* XXX: Use Py_EnterRecursiveCall()? */ |
| 2079 | if (++self->nesting > Py_GetRecursionLimit()) { |
| 2080 | PyErr_SetString(PyExc_RuntimeError, |
| 2081 | "maximum recursion depth exceeded"); |
| 2082 | goto error; |
| 2083 | } |
| 2084 | |
| 2085 | /* The extra pers_save argument is necessary to avoid calling save_pers() |
| 2086 | on its returned object. */ |
| 2087 | if (!pers_save && self->pers_func) { |
| 2088 | /* save_pers() returns: |
| 2089 | -1 to signal an error; |
| 2090 | 0 if it did nothing successfully; |
| 2091 | 1 if a persistent id was saved. |
| 2092 | */ |
| 2093 | if ((status = save_pers(self, obj, self->pers_func)) != 0) |
| 2094 | goto done; |
| 2095 | } |
| 2096 | |
| 2097 | type = Py_TYPE(obj); |
| 2098 | |
| 2099 | /* XXX: The old cPickle had an optimization that used switch-case |
| 2100 | statement dispatching on the first letter of the type name. It was |
| 2101 | probably not a bad idea after all. If benchmarks shows that particular |
| 2102 | optimization had some real benefits, it would be nice to add it |
| 2103 | back. */ |
| 2104 | |
| 2105 | /* Atom types; these aren't memoized, so don't check the memo. */ |
| 2106 | |
| 2107 | if (obj == Py_None) { |
| 2108 | status = save_none(self, obj); |
| 2109 | goto done; |
| 2110 | } |
| 2111 | else if (obj == Py_False || obj == Py_True) { |
| 2112 | status = save_bool(self, obj); |
| 2113 | goto done; |
| 2114 | } |
| 2115 | else if (type == &PyLong_Type) { |
| 2116 | status = save_long(self, obj); |
| 2117 | goto done; |
| 2118 | } |
| 2119 | else if (type == &PyFloat_Type) { |
| 2120 | status = save_float(self, obj); |
| 2121 | goto done; |
| 2122 | } |
| 2123 | |
| 2124 | /* Check the memo to see if it has the object. If so, generate |
| 2125 | a GET (or BINGET) opcode, instead of pickling the object |
| 2126 | once again. */ |
| 2127 | memo_key = PyLong_FromVoidPtr(obj); |
| 2128 | if (memo_key == NULL) |
| 2129 | goto error; |
| 2130 | if (PyDict_GetItem(self->memo, memo_key)) { |
| 2131 | if (memo_get(self, memo_key) < 0) |
| 2132 | goto error; |
| 2133 | goto done; |
| 2134 | } |
| 2135 | |
| 2136 | if (type == &PyBytes_Type) { |
| 2137 | status = save_bytes(self, obj); |
| 2138 | goto done; |
| 2139 | } |
| 2140 | else if (type == &PyUnicode_Type) { |
| 2141 | status = save_unicode(self, obj); |
| 2142 | goto done; |
| 2143 | } |
| 2144 | else if (type == &PyDict_Type) { |
| 2145 | status = save_dict(self, obj); |
| 2146 | goto done; |
| 2147 | } |
| 2148 | else if (type == &PyList_Type) { |
| 2149 | status = save_list(self, obj); |
| 2150 | goto done; |
| 2151 | } |
| 2152 | else if (type == &PyTuple_Type) { |
| 2153 | status = save_tuple(self, obj); |
| 2154 | goto done; |
| 2155 | } |
| 2156 | else if (type == &PyType_Type) { |
| 2157 | status = save_global(self, obj, NULL); |
| 2158 | goto done; |
| 2159 | } |
| 2160 | else if (type == &PyFunction_Type) { |
| 2161 | status = save_global(self, obj, NULL); |
| 2162 | if (status < 0 && PyErr_ExceptionMatches(PickleError)) { |
| 2163 | /* fall back to reduce */ |
| 2164 | PyErr_Clear(); |
| 2165 | } |
| 2166 | else { |
| 2167 | goto done; |
| 2168 | } |
| 2169 | } |
| 2170 | else if (type == &PyCFunction_Type) { |
| 2171 | status = save_global(self, obj, NULL); |
| 2172 | goto done; |
| 2173 | } |
| 2174 | else if (PyType_IsSubtype(type, &PyType_Type)) { |
| 2175 | status = save_global(self, obj, NULL); |
| 2176 | goto done; |
| 2177 | } |
| 2178 | |
| 2179 | /* XXX: This part needs some unit tests. */ |
| 2180 | |
| 2181 | /* Get a reduction callable, and call it. This may come from |
| 2182 | * copyreg.dispatch_table, the object's __reduce_ex__ method, |
| 2183 | * or the object's __reduce__ method. |
| 2184 | */ |
| 2185 | reduce_func = PyDict_GetItem(dispatch_table, (PyObject *)type); |
| 2186 | if (reduce_func != NULL) { |
| 2187 | /* Here, the reference count of the reduce_func object returned by |
| 2188 | PyDict_GetItem needs to be increased to be consistent with the one |
| 2189 | returned by PyObject_GetAttr. This is allow us to blindly DECREF |
| 2190 | reduce_func at the end of the save() routine. |
| 2191 | */ |
| 2192 | Py_INCREF(reduce_func); |
| 2193 | Py_INCREF(obj); |
| 2194 | reduce_value = pickler_call(self, reduce_func, obj); |
| 2195 | } |
| 2196 | else { |
| 2197 | static PyObject *reduce_str = NULL; |
| 2198 | static PyObject *reduce_ex_str = NULL; |
| 2199 | |
| 2200 | /* Cache the name of the reduce methods. */ |
| 2201 | if (reduce_str == NULL) { |
| 2202 | reduce_str = PyUnicode_InternFromString("__reduce__"); |
| 2203 | if (reduce_str == NULL) |
| 2204 | goto error; |
| 2205 | reduce_ex_str = PyUnicode_InternFromString("__reduce_ex__"); |
| 2206 | if (reduce_ex_str == NULL) |
| 2207 | goto error; |
| 2208 | } |
| 2209 | |
| 2210 | /* XXX: If the __reduce__ method is defined, __reduce_ex__ is |
| 2211 | automatically defined as __reduce__. While this is convenient, this |
| 2212 | make it impossible to know which method was actually called. Of |
| 2213 | course, this is not a big deal. But still, it would be nice to let |
| 2214 | the user know which method was called when something go |
| 2215 | wrong. Incidentally, this means if __reduce_ex__ is not defined, we |
| 2216 | don't actually have to check for a __reduce__ method. */ |
| 2217 | |
| 2218 | /* Check for a __reduce_ex__ method. */ |
| 2219 | reduce_func = PyObject_GetAttr(obj, reduce_ex_str); |
| 2220 | if (reduce_func != NULL) { |
| 2221 | PyObject *proto; |
| 2222 | proto = PyLong_FromLong(self->proto); |
| 2223 | if (proto != NULL) { |
| 2224 | reduce_value = pickler_call(self, reduce_func, proto); |
| 2225 | } |
| 2226 | } |
| 2227 | else { |
| 2228 | if (PyErr_ExceptionMatches(PyExc_AttributeError)) |
| 2229 | PyErr_Clear(); |
| 2230 | else |
| 2231 | goto error; |
| 2232 | /* Check for a __reduce__ method. */ |
| 2233 | reduce_func = PyObject_GetAttr(obj, reduce_str); |
| 2234 | if (reduce_func != NULL) { |
| 2235 | reduce_value = PyObject_Call(reduce_func, empty_tuple, NULL); |
| 2236 | } |
| 2237 | else { |
| 2238 | PyErr_Format(PicklingError, "can't pickle '%.200s' object: %R", |
| 2239 | type->tp_name, obj); |
| 2240 | goto error; |
| 2241 | } |
| 2242 | } |
| 2243 | } |
| 2244 | |
| 2245 | if (reduce_value == NULL) |
| 2246 | goto error; |
| 2247 | |
| 2248 | if (PyUnicode_Check(reduce_value)) { |
| 2249 | status = save_global(self, obj, reduce_value); |
| 2250 | goto done; |
| 2251 | } |
| 2252 | |
| 2253 | if (!PyTuple_Check(reduce_value)) { |
| 2254 | PyErr_SetString(PicklingError, |
| 2255 | "__reduce__ must return a string or tuple"); |
| 2256 | goto error; |
| 2257 | } |
| 2258 | if (Py_SIZE(reduce_value) < 2 || Py_SIZE(reduce_value) > 5) { |
| 2259 | PyErr_SetString(PicklingError, "tuple returned by __reduce__ " |
| 2260 | "must contain 2 through 5 elements"); |
| 2261 | goto error; |
| 2262 | } |
| 2263 | if (!PyTuple_Check(PyTuple_GET_ITEM(reduce_value, 1))) { |
| 2264 | PyErr_SetString(PicklingError, "second item of the tuple " |
| 2265 | "returned by __reduce__ must be a tuple"); |
| 2266 | goto error; |
| 2267 | } |
| 2268 | |
| 2269 | status = save_reduce(self, reduce_value, obj); |
| 2270 | |
| 2271 | if (0) { |
| 2272 | error: |
| 2273 | status = -1; |
| 2274 | } |
| 2275 | done: |
| 2276 | self->nesting--; |
| 2277 | Py_XDECREF(memo_key); |
| 2278 | Py_XDECREF(reduce_func); |
| 2279 | Py_XDECREF(reduce_value); |
| 2280 | |
| 2281 | return status; |
| 2282 | } |
| 2283 | |
| 2284 | static int |
| 2285 | dump(PicklerObject *self, PyObject *obj) |
| 2286 | { |
| 2287 | const char stop_op = STOP; |
| 2288 | |
| 2289 | if (self->proto >= 2) { |
| 2290 | char header[2]; |
| 2291 | |
| 2292 | header[0] = PROTO; |
| 2293 | assert(self->proto >= 0 && self->proto < 256); |
| 2294 | header[1] = (unsigned char)self->proto; |
| 2295 | if (pickler_write(self, header, 2) < 0) |
| 2296 | return -1; |
| 2297 | } |
| 2298 | |
| 2299 | if (save(self, obj, 0) < 0 || |
| 2300 | pickler_write(self, &stop_op, 1) < 0 || |
| 2301 | pickler_write(self, NULL, 0) < 0) |
| 2302 | return -1; |
| 2303 | |
| 2304 | return 0; |
| 2305 | } |
| 2306 | |
| 2307 | PyDoc_STRVAR(Pickler_clear_memo_doc, |
| 2308 | "clear_memo() -> None. Clears the pickler's \"memo\"." |
| 2309 | "\n" |
| 2310 | "The memo is the data structure that remembers which objects the\n" |
| 2311 | "pickler has already seen, so that shared or recursive objects are\n" |
| 2312 | "pickled by reference and not by value. This method is useful when\n" |
| 2313 | "re-using picklers."); |
| 2314 | |
| 2315 | static PyObject * |
| 2316 | Pickler_clear_memo(PicklerObject *self) |
| 2317 | { |
| 2318 | if (self->memo) |
| 2319 | PyDict_Clear(self->memo); |
| 2320 | |
| 2321 | Py_RETURN_NONE; |
| 2322 | } |
| 2323 | |
| 2324 | PyDoc_STRVAR(Pickler_dump_doc, |
| 2325 | "dump(obj) -> None. Write a pickled representation of obj to the open file."); |
| 2326 | |
| 2327 | static PyObject * |
| 2328 | Pickler_dump(PicklerObject *self, PyObject *args) |
| 2329 | { |
| 2330 | PyObject *obj; |
| 2331 | |
| 2332 | if (!PyArg_ParseTuple(args, "O:dump", &obj)) |
| 2333 | return NULL; |
| 2334 | |
| 2335 | if (dump(self, obj) < 0) |
| 2336 | return NULL; |
| 2337 | |
| 2338 | Py_RETURN_NONE; |
| 2339 | } |
| 2340 | |
| 2341 | static struct PyMethodDef Pickler_methods[] = { |
| 2342 | {"dump", (PyCFunction)Pickler_dump, METH_VARARGS, |
| 2343 | Pickler_dump_doc}, |
| 2344 | {"clear_memo", (PyCFunction)Pickler_clear_memo, METH_NOARGS, |
| 2345 | Pickler_clear_memo_doc}, |
| 2346 | {NULL, NULL} /* sentinel */ |
| 2347 | }; |
| 2348 | |
| 2349 | static void |
| 2350 | Pickler_dealloc(PicklerObject *self) |
| 2351 | { |
| 2352 | PyObject_GC_UnTrack(self); |
| 2353 | |
| 2354 | Py_XDECREF(self->write); |
| 2355 | Py_XDECREF(self->memo); |
| 2356 | Py_XDECREF(self->pers_func); |
| 2357 | Py_XDECREF(self->arg); |
| 2358 | Py_XDECREF(self->fast_memo); |
| 2359 | |
| 2360 | PyMem_Free(self->write_buf); |
| 2361 | |
| 2362 | Py_TYPE(self)->tp_free((PyObject *)self); |
| 2363 | } |
| 2364 | |
| 2365 | static int |
| 2366 | Pickler_traverse(PicklerObject *self, visitproc visit, void *arg) |
| 2367 | { |
| 2368 | Py_VISIT(self->write); |
| 2369 | Py_VISIT(self->memo); |
| 2370 | Py_VISIT(self->pers_func); |
| 2371 | Py_VISIT(self->arg); |
| 2372 | Py_VISIT(self->fast_memo); |
| 2373 | return 0; |
| 2374 | } |
| 2375 | |
| 2376 | static int |
| 2377 | Pickler_clear(PicklerObject *self) |
| 2378 | { |
| 2379 | Py_CLEAR(self->write); |
| 2380 | Py_CLEAR(self->memo); |
| 2381 | Py_CLEAR(self->pers_func); |
| 2382 | Py_CLEAR(self->arg); |
| 2383 | Py_CLEAR(self->fast_memo); |
| 2384 | |
| 2385 | PyMem_Free(self->write_buf); |
| 2386 | self->write_buf = NULL; |
| 2387 | |
| 2388 | return 0; |
| 2389 | } |
| 2390 | |
| 2391 | PyDoc_STRVAR(Pickler_doc, |
| 2392 | "Pickler(file, protocol=None)" |
| 2393 | "\n" |
| 2394 | "This takes a binary file for writing a pickle data stream.\n" |
| 2395 | "\n" |
| 2396 | "The optional protocol argument tells the pickler to use the\n" |
| 2397 | "given protocol; supported protocols are 0, 1, 2, 3. The default\n" |
| 2398 | "protocol is 3; a backward-incompatible protocol designed for\n" |
| 2399 | "Python 3.0.\n" |
| 2400 | "\n" |
| 2401 | "Specifying a negative protocol version selects the highest\n" |
| 2402 | "protocol version supported. The higher the protocol used, the\n" |
| 2403 | "more recent the version of Python needed to read the pickle\n" |
| 2404 | "produced.\n" |
| 2405 | "\n" |
| 2406 | "The file argument must have a write() method that accepts a single\n" |
| 2407 | "bytes argument. It can thus be a file object opened for binary\n" |
| 2408 | "writing, a io.BytesIO instance, or any other custom object that\n" |
| 2409 | "meets this interface.\n"); |
| 2410 | |
| 2411 | static int |
| 2412 | Pickler_init(PicklerObject *self, PyObject *args, PyObject *kwds) |
| 2413 | { |
| 2414 | static char *kwlist[] = {"file", "protocol", 0}; |
| 2415 | PyObject *file; |
| 2416 | PyObject *proto_obj = NULL; |
| 2417 | long proto = 0; |
| 2418 | |
| 2419 | if (!PyArg_ParseTupleAndKeywords(args, kwds, "O|O:Pickler", |
| 2420 | kwlist, &file, &proto_obj)) |
| 2421 | return -1; |
| 2422 | |
| 2423 | /* In case of multiple __init__() calls, clear previous content. */ |
| 2424 | if (self->write != NULL) |
| 2425 | (void)Pickler_clear(self); |
| 2426 | |
| 2427 | if (proto_obj == NULL || proto_obj == Py_None) |
| 2428 | proto = DEFAULT_PROTOCOL; |
| 2429 | else |
| 2430 | proto = PyLong_AsLong(proto_obj); |
| 2431 | |
| 2432 | if (proto < 0) |
| 2433 | proto = HIGHEST_PROTOCOL; |
| 2434 | if (proto > HIGHEST_PROTOCOL) { |
| 2435 | PyErr_Format(PyExc_ValueError, "pickle protocol must be <= %d", |
| 2436 | HIGHEST_PROTOCOL); |
| 2437 | return -1; |
| 2438 | } |
| 2439 | |
| 2440 | self->proto = proto; |
| 2441 | self->bin = proto > 0; |
| 2442 | self->arg = NULL; |
| 2443 | self->nesting = 0; |
| 2444 | self->fast = 0; |
| 2445 | self->fast_nesting = 0; |
| 2446 | self->fast_memo = NULL; |
| 2447 | |
| 2448 | if (!PyObject_HasAttrString(file, "write")) { |
| 2449 | PyErr_SetString(PyExc_TypeError, |
| 2450 | "file must have a 'write' attribute"); |
| 2451 | return -1; |
| 2452 | } |
| 2453 | self->write = PyObject_GetAttrString(file, "write"); |
| 2454 | if (self->write == NULL) |
| 2455 | return -1; |
| 2456 | self->buf_size = 0; |
| 2457 | self->write_buf = (char *)PyMem_Malloc(WRITE_BUF_SIZE); |
| 2458 | if (self->write_buf == NULL) { |
| 2459 | PyErr_NoMemory(); |
| 2460 | return -1; |
| 2461 | } |
| 2462 | self->pers_func = NULL; |
| 2463 | if (PyObject_HasAttrString((PyObject *)self, "persistent_id")) { |
| 2464 | self->pers_func = PyObject_GetAttrString((PyObject *)self, |
| 2465 | "persistent_id"); |
| 2466 | if (self->pers_func == NULL) |
| 2467 | return -1; |
| 2468 | } |
| 2469 | self->memo = PyDict_New(); |
| 2470 | if (self->memo == NULL) |
| 2471 | return -1; |
| 2472 | |
| 2473 | return 0; |
| 2474 | } |
| 2475 | |
| 2476 | static PyObject * |
| 2477 | Pickler_get_memo(PicklerObject *self) |
| 2478 | { |
| 2479 | if (self->memo == NULL) |
| 2480 | PyErr_SetString(PyExc_AttributeError, "memo"); |
| 2481 | else |
| 2482 | Py_INCREF(self->memo); |
| 2483 | return self->memo; |
| 2484 | } |
| 2485 | |
| 2486 | static int |
| 2487 | Pickler_set_memo(PicklerObject *self, PyObject *value) |
| 2488 | { |
| 2489 | PyObject *tmp; |
| 2490 | |
| 2491 | if (value == NULL) { |
| 2492 | PyErr_SetString(PyExc_TypeError, |
| 2493 | "attribute deletion is not supported"); |
| 2494 | return -1; |
| 2495 | } |
| 2496 | if (!PyDict_Check(value)) { |
| 2497 | PyErr_SetString(PyExc_TypeError, "memo must be a dictionary"); |
| 2498 | return -1; |
| 2499 | } |
| 2500 | |
| 2501 | tmp = self->memo; |
| 2502 | Py_INCREF(value); |
| 2503 | self->memo = value; |
| 2504 | Py_XDECREF(tmp); |
| 2505 | |
| 2506 | return 0; |
| 2507 | } |
| 2508 | |
| 2509 | static PyObject * |
| 2510 | Pickler_get_persid(PicklerObject *self) |
| 2511 | { |
| 2512 | if (self->pers_func == NULL) |
| 2513 | PyErr_SetString(PyExc_AttributeError, "persistent_id"); |
| 2514 | else |
| 2515 | Py_INCREF(self->pers_func); |
| 2516 | return self->pers_func; |
| 2517 | } |
| 2518 | |
| 2519 | static int |
| 2520 | Pickler_set_persid(PicklerObject *self, PyObject *value) |
| 2521 | { |
| 2522 | PyObject *tmp; |
| 2523 | |
| 2524 | if (value == NULL) { |
| 2525 | PyErr_SetString(PyExc_TypeError, |
| 2526 | "attribute deletion is not supported"); |
| 2527 | return -1; |
| 2528 | } |
| 2529 | if (!PyCallable_Check(value)) { |
| 2530 | PyErr_SetString(PyExc_TypeError, |
| 2531 | "persistent_id must be a callable taking one argument"); |
| 2532 | return -1; |
| 2533 | } |
| 2534 | |
| 2535 | tmp = self->pers_func; |
| 2536 | Py_INCREF(value); |
| 2537 | self->pers_func = value; |
| 2538 | Py_XDECREF(tmp); /* self->pers_func can be NULL, so be careful. */ |
| 2539 | |
| 2540 | return 0; |
| 2541 | } |
| 2542 | |
| 2543 | static PyMemberDef Pickler_members[] = { |
| 2544 | {"bin", T_INT, offsetof(PicklerObject, bin)}, |
| 2545 | {"fast", T_INT, offsetof(PicklerObject, fast)}, |
| 2546 | {NULL} |
| 2547 | }; |
| 2548 | |
| 2549 | static PyGetSetDef Pickler_getsets[] = { |
| 2550 | {"memo", (getter)Pickler_get_memo, |
| 2551 | (setter)Pickler_set_memo}, |
| 2552 | {"persistent_id", (getter)Pickler_get_persid, |
| 2553 | (setter)Pickler_set_persid}, |
| 2554 | {NULL} |
| 2555 | }; |
| 2556 | |
| 2557 | static PyTypeObject Pickler_Type = { |
| 2558 | PyVarObject_HEAD_INIT(NULL, 0) |
| 2559 | "_pickle.Pickler" , /*tp_name*/ |
| 2560 | sizeof(PicklerObject), /*tp_basicsize*/ |
| 2561 | 0, /*tp_itemsize*/ |
| 2562 | (destructor)Pickler_dealloc, /*tp_dealloc*/ |
| 2563 | 0, /*tp_print*/ |
| 2564 | 0, /*tp_getattr*/ |
| 2565 | 0, /*tp_setattr*/ |
| 2566 | 0, /*tp_compare*/ |
| 2567 | 0, /*tp_repr*/ |
| 2568 | 0, /*tp_as_number*/ |
| 2569 | 0, /*tp_as_sequence*/ |
| 2570 | 0, /*tp_as_mapping*/ |
| 2571 | 0, /*tp_hash*/ |
| 2572 | 0, /*tp_call*/ |
| 2573 | 0, /*tp_str*/ |
| 2574 | 0, /*tp_getattro*/ |
| 2575 | 0, /*tp_setattro*/ |
| 2576 | 0, /*tp_as_buffer*/ |
| 2577 | Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE | Py_TPFLAGS_HAVE_GC, |
| 2578 | Pickler_doc, /*tp_doc*/ |
| 2579 | (traverseproc)Pickler_traverse, /*tp_traverse*/ |
| 2580 | (inquiry)Pickler_clear, /*tp_clear*/ |
| 2581 | 0, /*tp_richcompare*/ |
| 2582 | 0, /*tp_weaklistoffset*/ |
| 2583 | 0, /*tp_iter*/ |
| 2584 | 0, /*tp_iternext*/ |
| 2585 | Pickler_methods, /*tp_methods*/ |
| 2586 | Pickler_members, /*tp_members*/ |
| 2587 | Pickler_getsets, /*tp_getset*/ |
| 2588 | 0, /*tp_base*/ |
| 2589 | 0, /*tp_dict*/ |
| 2590 | 0, /*tp_descr_get*/ |
| 2591 | 0, /*tp_descr_set*/ |
| 2592 | 0, /*tp_dictoffset*/ |
| 2593 | (initproc)Pickler_init, /*tp_init*/ |
| 2594 | PyType_GenericAlloc, /*tp_alloc*/ |
| 2595 | PyType_GenericNew, /*tp_new*/ |
| 2596 | PyObject_GC_Del, /*tp_free*/ |
| 2597 | 0, /*tp_is_gc*/ |
| 2598 | }; |
| 2599 | |
| 2600 | /* Temporary helper for calling self.find_class(). |
| 2601 | |
| 2602 | XXX: It would be nice to able to avoid Python function call overhead, by |
| 2603 | using directly the C version of find_class(), when find_class() is not |
| 2604 | overridden by a subclass. Although, this could become rather hackish. A |
| 2605 | simpler optimization would be to call the C function when self is not a |
| 2606 | subclass instance. */ |
| 2607 | static PyObject * |
| 2608 | find_class(UnpicklerObject *self, PyObject *module_name, PyObject *global_name) |
| 2609 | { |
| 2610 | return PyObject_CallMethod((PyObject *)self, "find_class", "OO", |
| 2611 | module_name, global_name); |
| 2612 | } |
| 2613 | |
| 2614 | static int |
| 2615 | marker(UnpicklerObject *self) |
| 2616 | { |
| 2617 | if (self->num_marks < 1) { |
| 2618 | PyErr_SetString(UnpicklingError, "could not find MARK"); |
| 2619 | return -1; |
| 2620 | } |
| 2621 | |
| 2622 | return self->marks[--self->num_marks]; |
| 2623 | } |
| 2624 | |
| 2625 | static int |
| 2626 | load_none(UnpicklerObject *self) |
| 2627 | { |
| 2628 | PDATA_APPEND(self->stack, Py_None, -1); |
| 2629 | return 0; |
| 2630 | } |
| 2631 | |
| 2632 | static int |
| 2633 | bad_readline(void) |
| 2634 | { |
| 2635 | PyErr_SetString(UnpicklingError, "pickle data was truncated"); |
| 2636 | return -1; |
| 2637 | } |
| 2638 | |
| 2639 | static int |
| 2640 | load_int(UnpicklerObject *self) |
| 2641 | { |
| 2642 | PyObject *value; |
| 2643 | char *endptr, *s; |
| 2644 | Py_ssize_t len; |
| 2645 | long x; |
| 2646 | |
| 2647 | if ((len = unpickler_readline(self, &s)) < 0) |
| 2648 | return -1; |
| 2649 | if (len < 2) |
| 2650 | return bad_readline(); |
| 2651 | |
| 2652 | errno = 0; |
| 2653 | /* XXX: Should the base argument of strtol() be explicitly set to 10? */ |
| 2654 | x = strtol(s, &endptr, 0); |
| 2655 | |
| 2656 | if (errno || (*endptr != '\n') || (endptr[1] != '\0')) { |
| 2657 | /* Hm, maybe we've got something long. Let's try reading |
| 2658 | * it as a Python long object. */ |
| 2659 | errno = 0; |
| 2660 | /* XXX: Same thing about the base here. */ |
| 2661 | value = PyLong_FromString(s, NULL, 0); |
| 2662 | if (value == NULL) { |
| 2663 | PyErr_SetString(PyExc_ValueError, |
| 2664 | "could not convert string to int"); |
| 2665 | return -1; |
| 2666 | } |
| 2667 | } |
| 2668 | else { |
| 2669 | if (len == 3 && (x == 0 || x == 1)) { |
| 2670 | if ((value = PyBool_FromLong(x)) == NULL) |
| 2671 | return -1; |
| 2672 | } |
| 2673 | else { |
| 2674 | if ((value = PyLong_FromLong(x)) == NULL) |
| 2675 | return -1; |
| 2676 | } |
| 2677 | } |
| 2678 | |
| 2679 | PDATA_PUSH(self->stack, value, -1); |
| 2680 | return 0; |
| 2681 | } |
| 2682 | |
| 2683 | static int |
| 2684 | load_bool(UnpicklerObject *self, PyObject *boolean) |
| 2685 | { |
| 2686 | assert(boolean == Py_True || boolean == Py_False); |
| 2687 | PDATA_APPEND(self->stack, boolean, -1); |
| 2688 | return 0; |
| 2689 | } |
| 2690 | |
| 2691 | /* s contains x bytes of a little-endian integer. Return its value as a |
| 2692 | * C int. Obscure: when x is 1 or 2, this is an unsigned little-endian |
| 2693 | * int, but when x is 4 it's a signed one. This is an historical source |
| 2694 | * of x-platform bugs. |
| 2695 | */ |
| 2696 | static long |
| 2697 | calc_binint(char *bytes, int size) |
| 2698 | { |
| 2699 | unsigned char *s = (unsigned char *)bytes; |
| 2700 | int i = size; |
| 2701 | long x = 0; |
| 2702 | |
| 2703 | for (i = 0; i < size; i++) { |
| 2704 | x |= (long)s[i] << (i * 8); |
| 2705 | } |
| 2706 | |
| 2707 | /* Unlike BININT1 and BININT2, BININT (more accurately BININT4) |
| 2708 | * is signed, so on a box with longs bigger than 4 bytes we need |
| 2709 | * to extend a BININT's sign bit to the full width. |
| 2710 | */ |
| 2711 | if (SIZEOF_LONG > 4 && size == 4) { |
| 2712 | x |= -(x & (1L << 31)); |
| 2713 | } |
| 2714 | |
| 2715 | return x; |
| 2716 | } |
| 2717 | |
| 2718 | static int |
| 2719 | load_binintx(UnpicklerObject *self, char *s, int size) |
| 2720 | { |
| 2721 | PyObject *value; |
| 2722 | long x; |
| 2723 | |
| 2724 | x = calc_binint(s, size); |
| 2725 | |
| 2726 | if ((value = PyLong_FromLong(x)) == NULL) |
| 2727 | return -1; |
| 2728 | |
| 2729 | PDATA_PUSH(self->stack, value, -1); |
| 2730 | return 0; |
| 2731 | } |
| 2732 | |
| 2733 | static int |
| 2734 | load_binint(UnpicklerObject *self) |
| 2735 | { |
| 2736 | char *s; |
| 2737 | |
| 2738 | if (unpickler_read(self, &s, 4) < 0) |
| 2739 | return -1; |
| 2740 | |
| 2741 | return load_binintx(self, s, 4); |
| 2742 | } |
| 2743 | |
| 2744 | static int |
| 2745 | load_binint1(UnpicklerObject *self) |
| 2746 | { |
| 2747 | char *s; |
| 2748 | |
| 2749 | if (unpickler_read(self, &s, 1) < 0) |
| 2750 | return -1; |
| 2751 | |
| 2752 | return load_binintx(self, s, 1); |
| 2753 | } |
| 2754 | |
| 2755 | static int |
| 2756 | load_binint2(UnpicklerObject *self) |
| 2757 | { |
| 2758 | char *s; |
| 2759 | |
| 2760 | if (unpickler_read(self, &s, 2) < 0) |
| 2761 | return -1; |
| 2762 | |
| 2763 | return load_binintx(self, s, 2); |
| 2764 | } |
| 2765 | |
| 2766 | static int |
| 2767 | load_long(UnpicklerObject *self) |
| 2768 | { |
| 2769 | PyObject *value; |
| 2770 | char *s; |
| 2771 | Py_ssize_t len; |
| 2772 | |
| 2773 | if ((len = unpickler_readline(self, &s)) < 0) |
| 2774 | return -1; |
| 2775 | if (len < 2) |
| 2776 | return bad_readline(); |
| 2777 | |
| 2778 | /* XXX: Should the base argument explicitly set to 10? */ |
| 2779 | if ((value = PyLong_FromString(s, NULL, 0)) == NULL) |
| 2780 | return -1; |
| 2781 | |
| 2782 | PDATA_PUSH(self->stack, value, -1); |
| 2783 | return 0; |
| 2784 | } |
| 2785 | |
| 2786 | /* 'size' bytes contain the # of bytes of little-endian 256's-complement |
| 2787 | * data following. |
| 2788 | */ |
| 2789 | static int |
| 2790 | load_counted_long(UnpicklerObject *self, int size) |
| 2791 | { |
| 2792 | PyObject *value; |
| 2793 | char *nbytes; |
| 2794 | char *pdata; |
| 2795 | |
| 2796 | assert(size == 1 || size == 4); |
| 2797 | if (unpickler_read(self, &nbytes, size) < 0) |
| 2798 | return -1; |
| 2799 | |
| 2800 | size = calc_binint(nbytes, size); |
| 2801 | if (size < 0) { |
| 2802 | /* Corrupt or hostile pickle -- we never write one like this */ |
| 2803 | PyErr_SetString(UnpicklingError, |
| 2804 | "LONG pickle has negative byte count"); |
| 2805 | return -1; |
| 2806 | } |
| 2807 | |
| 2808 | if (size == 0) |
| 2809 | value = PyLong_FromLong(0L); |
| 2810 | else { |
| 2811 | /* Read the raw little-endian bytes and convert. */ |
| 2812 | if (unpickler_read(self, &pdata, size) < 0) |
| 2813 | return -1; |
| 2814 | value = _PyLong_FromByteArray((unsigned char *)pdata, (size_t)size, |
| 2815 | 1 /* little endian */ , 1 /* signed */ ); |
| 2816 | } |
| 2817 | if (value == NULL) |
| 2818 | return -1; |
| 2819 | PDATA_PUSH(self->stack, value, -1); |
| 2820 | return 0; |
| 2821 | } |
| 2822 | |
| 2823 | static int |
| 2824 | load_float(UnpicklerObject *self) |
| 2825 | { |
| 2826 | PyObject *value; |
| 2827 | char *endptr, *s; |
| 2828 | Py_ssize_t len; |
| 2829 | double d; |
| 2830 | |
| 2831 | if ((len = unpickler_readline(self, &s)) < 0) |
| 2832 | return -1; |
| 2833 | if (len < 2) |
| 2834 | return bad_readline(); |
| 2835 | |
| 2836 | errno = 0; |
| 2837 | d = PyOS_ascii_strtod(s, &endptr); |
| 2838 | |
| 2839 | if (errno || (endptr[0] != '\n') || (endptr[1] != '\0')) { |
| 2840 | PyErr_SetString(PyExc_ValueError, "could not convert string to float"); |
| 2841 | return -1; |
| 2842 | } |
| 2843 | |
| 2844 | if ((value = PyFloat_FromDouble(d)) == NULL) |
| 2845 | return -1; |
| 2846 | |
| 2847 | PDATA_PUSH(self->stack, value, -1); |
| 2848 | return 0; |
| 2849 | } |
| 2850 | |
| 2851 | static int |
| 2852 | load_binfloat(UnpicklerObject *self) |
| 2853 | { |
| 2854 | PyObject *value; |
| 2855 | double x; |
| 2856 | char *s; |
| 2857 | |
| 2858 | if (unpickler_read(self, &s, 8) < 0) |
| 2859 | return -1; |
| 2860 | |
| 2861 | x = _PyFloat_Unpack8((unsigned char *)s, 0); |
| 2862 | if (x == -1.0 && PyErr_Occurred()) |
| 2863 | return -1; |
| 2864 | |
| 2865 | if ((value = PyFloat_FromDouble(x)) == NULL) |
| 2866 | return -1; |
| 2867 | |
| 2868 | PDATA_PUSH(self->stack, value, -1); |
| 2869 | return 0; |
| 2870 | } |
| 2871 | |
| 2872 | static int |
| 2873 | load_string(UnpicklerObject *self) |
| 2874 | { |
| 2875 | PyObject *bytes; |
| 2876 | PyObject *str = NULL; |
| 2877 | Py_ssize_t len; |
| 2878 | char *s, *p; |
| 2879 | |
| 2880 | if ((len = unpickler_readline(self, &s)) < 0) |
| 2881 | return -1; |
| 2882 | if (len < 3) |
| 2883 | return bad_readline(); |
| 2884 | if ((s = strdup(s)) == NULL) { |
| 2885 | PyErr_NoMemory(); |
| 2886 | return -1; |
| 2887 | } |
| 2888 | |
| 2889 | /* Strip outermost quotes */ |
| 2890 | while (s[len - 1] <= ' ') |
| 2891 | len--; |
| 2892 | if (s[0] == '"' && s[len - 1] == '"') { |
| 2893 | s[len - 1] = '\0'; |
| 2894 | p = s + 1; |
| 2895 | len -= 2; |
| 2896 | } |
| 2897 | else if (s[0] == '\'' && s[len - 1] == '\'') { |
| 2898 | s[len - 1] = '\0'; |
| 2899 | p = s + 1; |
| 2900 | len -= 2; |
| 2901 | } |
| 2902 | else { |
| 2903 | free(s); |
| 2904 | PyErr_SetString(PyExc_ValueError, "insecure string pickle"); |
| 2905 | return -1; |
| 2906 | } |
| 2907 | |
| 2908 | /* Use the PyBytes API to decode the string, since that is what is used |
| 2909 | to encode, and then coerce the result to Unicode. */ |
| 2910 | bytes = PyBytes_DecodeEscape(p, len, NULL, 0, NULL); |
| 2911 | free(s); |
| 2912 | if (bytes == NULL) |
| 2913 | return -1; |
| 2914 | str = PyUnicode_FromEncodedObject(bytes, self->encoding, self->errors); |
| 2915 | Py_DECREF(bytes); |
| 2916 | if (str == NULL) |
| 2917 | return -1; |
| 2918 | |
| 2919 | PDATA_PUSH(self->stack, str, -1); |
| 2920 | return 0; |
| 2921 | } |
| 2922 | |
| 2923 | static int |
| 2924 | load_binbytes(UnpicklerObject *self) |
| 2925 | { |
| 2926 | PyObject *bytes; |
| 2927 | long x; |
| 2928 | char *s; |
| 2929 | |
| 2930 | if (unpickler_read(self, &s, 4) < 0) |
| 2931 | return -1; |
| 2932 | |
| 2933 | x = calc_binint(s, 4); |
| 2934 | if (x < 0) { |
| 2935 | PyErr_SetString(UnpicklingError, |
| 2936 | "BINBYTES pickle has negative byte count"); |
| 2937 | return -1; |
| 2938 | } |
| 2939 | |
| 2940 | if (unpickler_read(self, &s, x) < 0) |
| 2941 | return -1; |
| 2942 | bytes = PyBytes_FromStringAndSize(s, x); |
| 2943 | if (bytes == NULL) |
| 2944 | return -1; |
| 2945 | |
| 2946 | PDATA_PUSH(self->stack, bytes, -1); |
| 2947 | return 0; |
| 2948 | } |
| 2949 | |
| 2950 | static int |
| 2951 | load_short_binbytes(UnpicklerObject *self) |
| 2952 | { |
| 2953 | PyObject *bytes; |
| 2954 | unsigned char x; |
| 2955 | char *s; |
| 2956 | |
| 2957 | if (unpickler_read(self, &s, 1) < 0) |
| 2958 | return -1; |
| 2959 | |
| 2960 | x = (unsigned char)s[0]; |
| 2961 | |
| 2962 | if (unpickler_read(self, &s, x) < 0) |
| 2963 | return -1; |
| 2964 | |
| 2965 | bytes = PyBytes_FromStringAndSize(s, x); |
| 2966 | if (bytes == NULL) |
| 2967 | return -1; |
| 2968 | |
| 2969 | PDATA_PUSH(self->stack, bytes, -1); |
| 2970 | return 0; |
| 2971 | } |
| 2972 | |
| 2973 | static int |
| 2974 | load_binstring(UnpicklerObject *self) |
| 2975 | { |
| 2976 | PyObject *str; |
| 2977 | long x; |
| 2978 | char *s; |
| 2979 | |
| 2980 | if (unpickler_read(self, &s, 4) < 0) |
| 2981 | return -1; |
| 2982 | |
| 2983 | x = calc_binint(s, 4); |
| 2984 | if (x < 0) { |
| 2985 | PyErr_SetString(UnpicklingError, |
| 2986 | "BINSTRING pickle has negative byte count"); |
| 2987 | return -1; |
| 2988 | } |
| 2989 | |
| 2990 | if (unpickler_read(self, &s, x) < 0) |
| 2991 | return -1; |
| 2992 | |
| 2993 | /* Convert Python 2.x strings to unicode. */ |
| 2994 | str = PyUnicode_Decode(s, x, self->encoding, self->errors); |
| 2995 | if (str == NULL) |
| 2996 | return -1; |
| 2997 | |
| 2998 | PDATA_PUSH(self->stack, str, -1); |
| 2999 | return 0; |
| 3000 | } |
| 3001 | |
| 3002 | static int |
| 3003 | load_short_binstring(UnpicklerObject *self) |
| 3004 | { |
| 3005 | PyObject *str; |
| 3006 | unsigned char x; |
| 3007 | char *s; |
| 3008 | |
| 3009 | if (unpickler_read(self, &s, 1) < 0) |
| 3010 | return -1; |
| 3011 | |
| 3012 | x = (unsigned char)s[0]; |
| 3013 | |
| 3014 | if (unpickler_read(self, &s, x) < 0) |
| 3015 | return -1; |
| 3016 | |
| 3017 | /* Convert Python 2.x strings to unicode. */ |
| 3018 | str = PyUnicode_Decode(s, x, self->encoding, self->errors); |
| 3019 | if (str == NULL) |
| 3020 | return -1; |
| 3021 | |
| 3022 | PDATA_PUSH(self->stack, str, -1); |
| 3023 | return 0; |
| 3024 | } |
| 3025 | |
| 3026 | static int |
| 3027 | load_unicode(UnpicklerObject *self) |
| 3028 | { |
| 3029 | PyObject *str; |
| 3030 | Py_ssize_t len; |
| 3031 | char *s; |
| 3032 | |
| 3033 | if ((len = unpickler_readline(self, &s)) < 0) |
| 3034 | return -1; |
| 3035 | if (len < 1) |
| 3036 | return bad_readline(); |
| 3037 | |
| 3038 | str = PyUnicode_DecodeRawUnicodeEscape(s, len - 1, NULL); |
| 3039 | if (str == NULL) |
| 3040 | return -1; |
| 3041 | |
| 3042 | PDATA_PUSH(self->stack, str, -1); |
| 3043 | return 0; |
| 3044 | } |
| 3045 | |
| 3046 | static int |
| 3047 | load_binunicode(UnpicklerObject *self) |
| 3048 | { |
| 3049 | PyObject *str; |
| 3050 | long size; |
| 3051 | char *s; |
| 3052 | |
| 3053 | if (unpickler_read(self, &s, 4) < 0) |
| 3054 | return -1; |
| 3055 | |
| 3056 | size = calc_binint(s, 4); |
| 3057 | if (size < 0) { |
| 3058 | PyErr_SetString(UnpicklingError, |
| 3059 | "BINUNICODE pickle has negative byte count"); |
| 3060 | return -1; |
| 3061 | } |
| 3062 | |
| 3063 | if (unpickler_read(self, &s, size) < 0) |
| 3064 | return -1; |
| 3065 | |
| 3066 | str = PyUnicode_DecodeUTF8(s, size, NULL); |
| 3067 | if (str == NULL) |
| 3068 | return -1; |
| 3069 | |
| 3070 | PDATA_PUSH(self->stack, str, -1); |
| 3071 | return 0; |
| 3072 | } |
| 3073 | |
| 3074 | static int |
| 3075 | load_tuple(UnpicklerObject *self) |
| 3076 | { |
| 3077 | PyObject *tuple; |
| 3078 | int i; |
| 3079 | |
| 3080 | if ((i = marker(self)) < 0) |
| 3081 | return -1; |
| 3082 | |
| 3083 | tuple = Pdata_poptuple(self->stack, i); |
| 3084 | if (tuple == NULL) |
| 3085 | return -1; |
| 3086 | PDATA_PUSH(self->stack, tuple, -1); |
| 3087 | return 0; |
| 3088 | } |
| 3089 | |
| 3090 | static int |
| 3091 | load_counted_tuple(UnpicklerObject *self, int len) |
| 3092 | { |
| 3093 | PyObject *tuple; |
| 3094 | |
| 3095 | tuple = PyTuple_New(len); |
| 3096 | if (tuple == NULL) |
| 3097 | return -1; |
| 3098 | |
| 3099 | while (--len >= 0) { |
| 3100 | PyObject *item; |
| 3101 | |
| 3102 | PDATA_POP(self->stack, item); |
| 3103 | if (item == NULL) |
| 3104 | return -1; |
| 3105 | PyTuple_SET_ITEM(tuple, len, item); |
| 3106 | } |
| 3107 | PDATA_PUSH(self->stack, tuple, -1); |
| 3108 | return 0; |
| 3109 | } |
| 3110 | |
| 3111 | static int |
| 3112 | load_empty_list(UnpicklerObject *self) |
| 3113 | { |
| 3114 | PyObject *list; |
| 3115 | |
| 3116 | if ((list = PyList_New(0)) == NULL) |
| 3117 | return -1; |
| 3118 | PDATA_PUSH(self->stack, list, -1); |
| 3119 | return 0; |
| 3120 | } |
| 3121 | |
| 3122 | static int |
| 3123 | load_empty_dict(UnpicklerObject *self) |
| 3124 | { |
| 3125 | PyObject *dict; |
| 3126 | |
| 3127 | if ((dict = PyDict_New()) == NULL) |
| 3128 | return -1; |
| 3129 | PDATA_PUSH(self->stack, dict, -1); |
| 3130 | return 0; |
| 3131 | } |
| 3132 | |
| 3133 | static int |
| 3134 | load_list(UnpicklerObject *self) |
| 3135 | { |
| 3136 | PyObject *list; |
| 3137 | int i; |
| 3138 | |
| 3139 | if ((i = marker(self)) < 0) |
| 3140 | return -1; |
| 3141 | |
| 3142 | list = Pdata_poplist(self->stack, i); |
| 3143 | if (list == NULL) |
| 3144 | return -1; |
| 3145 | PDATA_PUSH(self->stack, list, -1); |
| 3146 | return 0; |
| 3147 | } |
| 3148 | |
| 3149 | static int |
| 3150 | load_dict(UnpicklerObject *self) |
| 3151 | { |
| 3152 | PyObject *dict, *key, *value; |
| 3153 | int i, j, k; |
| 3154 | |
| 3155 | if ((i = marker(self)) < 0) |
| 3156 | return -1; |
| 3157 | j = self->stack->length; |
| 3158 | |
| 3159 | if ((dict = PyDict_New()) == NULL) |
| 3160 | return -1; |
| 3161 | |
| 3162 | for (k = i + 1; k < j; k += 2) { |
| 3163 | key = self->stack->data[k - 1]; |
| 3164 | value = self->stack->data[k]; |
| 3165 | if (PyDict_SetItem(dict, key, value) < 0) { |
| 3166 | Py_DECREF(dict); |
| 3167 | return -1; |
| 3168 | } |
| 3169 | } |
| 3170 | Pdata_clear(self->stack, i); |
| 3171 | PDATA_PUSH(self->stack, dict, -1); |
| 3172 | return 0; |
| 3173 | } |
| 3174 | |
| 3175 | static PyObject * |
| 3176 | instantiate(PyObject *cls, PyObject *args) |
| 3177 | { |
| 3178 | PyObject *r = NULL; |
| 3179 | |
| 3180 | /* XXX: The pickle.py module does not create instances this way when the |
| 3181 | args tuple is empty. See Unpickler._instantiate(). */ |
| 3182 | if ((r = PyObject_CallObject(cls, args))) |
| 3183 | return r; |
| 3184 | |
| 3185 | /* XXX: Is this still nescessary? */ |
| 3186 | { |
| 3187 | PyObject *tp, *v, *tb, *tmp_value; |
| 3188 | |
| 3189 | PyErr_Fetch(&tp, &v, &tb); |
| 3190 | tmp_value = v; |
| 3191 | /* NULL occurs when there was a KeyboardInterrupt */ |
| 3192 | if (tmp_value == NULL) |
| 3193 | tmp_value = Py_None; |
| 3194 | if ((r = PyTuple_Pack(3, tmp_value, cls, args))) { |
| 3195 | Py_XDECREF(v); |
| 3196 | v = r; |
| 3197 | } |
| 3198 | PyErr_Restore(tp, v, tb); |
| 3199 | } |
| 3200 | return NULL; |
| 3201 | } |
| 3202 | |
| 3203 | static int |
| 3204 | load_obj(UnpicklerObject *self) |
| 3205 | { |
| 3206 | PyObject *cls, *args, *obj = NULL; |
| 3207 | int i; |
| 3208 | |
| 3209 | if ((i = marker(self)) < 0) |
| 3210 | return -1; |
| 3211 | |
| 3212 | args = Pdata_poptuple(self->stack, i + 1); |
| 3213 | if (args == NULL) |
| 3214 | return -1; |
| 3215 | |
| 3216 | PDATA_POP(self->stack, cls); |
| 3217 | if (cls) { |
| 3218 | obj = instantiate(cls, args); |
| 3219 | Py_DECREF(cls); |
| 3220 | } |
| 3221 | Py_DECREF(args); |
| 3222 | if (obj == NULL) |
| 3223 | return -1; |
| 3224 | |
| 3225 | PDATA_PUSH(self->stack, obj, -1); |
| 3226 | return 0; |
| 3227 | } |
| 3228 | |
| 3229 | static int |
| 3230 | load_inst(UnpicklerObject *self) |
| 3231 | { |
| 3232 | PyObject *cls = NULL; |
| 3233 | PyObject *args = NULL; |
| 3234 | PyObject *obj = NULL; |
| 3235 | PyObject *module_name; |
| 3236 | PyObject *class_name; |
| 3237 | Py_ssize_t len; |
| 3238 | int i; |
| 3239 | char *s; |
| 3240 | |
| 3241 | if ((i = marker(self)) < 0) |
| 3242 | return -1; |
| 3243 | if ((len = unpickler_readline(self, &s)) < 0) |
| 3244 | return -1; |
| 3245 | if (len < 2) |
| 3246 | return bad_readline(); |
| 3247 | |
| 3248 | /* Here it is safe to use PyUnicode_DecodeASCII(), even though non-ASCII |
| 3249 | identifiers are permitted in Python 3.0, since the INST opcode is only |
| 3250 | supported by older protocols on Python 2.x. */ |
| 3251 | module_name = PyUnicode_DecodeASCII(s, len - 1, "strict"); |
| 3252 | if (module_name == NULL) |
| 3253 | return -1; |
| 3254 | |
| 3255 | if ((len = unpickler_readline(self, &s)) >= 0) { |
| 3256 | if (len < 2) |
| 3257 | return bad_readline(); |
| 3258 | class_name = PyUnicode_DecodeASCII(s, len - 1, "strict"); |
| 3259 | if (class_name == NULL) { |
| 3260 | cls = find_class(self, module_name, class_name); |
| 3261 | Py_DECREF(class_name); |
| 3262 | } |
| 3263 | } |
| 3264 | Py_DECREF(module_name); |
| 3265 | |
| 3266 | if (cls == NULL) |
| 3267 | return -1; |
| 3268 | |
| 3269 | if ((args = Pdata_poptuple(self->stack, i)) != NULL) { |
| 3270 | obj = instantiate(cls, args); |
| 3271 | Py_DECREF(args); |
| 3272 | } |
| 3273 | Py_DECREF(cls); |
| 3274 | |
| 3275 | if (obj == NULL) |
| 3276 | return -1; |
| 3277 | |
| 3278 | PDATA_PUSH(self->stack, obj, -1); |
| 3279 | return 0; |
| 3280 | } |
| 3281 | |
| 3282 | static int |
| 3283 | load_newobj(UnpicklerObject *self) |
| 3284 | { |
| 3285 | PyObject *args = NULL; |
| 3286 | PyObject *clsraw = NULL; |
| 3287 | PyTypeObject *cls; /* clsraw cast to its true type */ |
| 3288 | PyObject *obj; |
| 3289 | |
| 3290 | /* Stack is ... cls argtuple, and we want to call |
| 3291 | * cls.__new__(cls, *argtuple). |
| 3292 | */ |
| 3293 | PDATA_POP(self->stack, args); |
| 3294 | if (args == NULL) |
| 3295 | goto error; |
| 3296 | if (!PyTuple_Check(args)) { |
| 3297 | PyErr_SetString(UnpicklingError, "NEWOBJ expected an arg " "tuple."); |
| 3298 | goto error; |
| 3299 | } |
| 3300 | |
| 3301 | PDATA_POP(self->stack, clsraw); |
| 3302 | cls = (PyTypeObject *)clsraw; |
| 3303 | if (cls == NULL) |
| 3304 | goto error; |
| 3305 | if (!PyType_Check(cls)) { |
| 3306 | PyErr_SetString(UnpicklingError, "NEWOBJ class argument " |
| 3307 | "isn't a type object"); |
| 3308 | goto error; |
| 3309 | } |
| 3310 | if (cls->tp_new == NULL) { |
| 3311 | PyErr_SetString(UnpicklingError, "NEWOBJ class argument " |
| 3312 | "has NULL tp_new"); |
| 3313 | goto error; |
| 3314 | } |
| 3315 | |
| 3316 | /* Call __new__. */ |
| 3317 | obj = cls->tp_new(cls, args, NULL); |
| 3318 | if (obj == NULL) |
| 3319 | goto error; |
| 3320 | |
| 3321 | Py_DECREF(args); |
| 3322 | Py_DECREF(clsraw); |
| 3323 | PDATA_PUSH(self->stack, obj, -1); |
| 3324 | return 0; |
| 3325 | |
| 3326 | error: |
| 3327 | Py_XDECREF(args); |
| 3328 | Py_XDECREF(clsraw); |
| 3329 | return -1; |
| 3330 | } |
| 3331 | |
| 3332 | static int |
| 3333 | load_global(UnpicklerObject *self) |
| 3334 | { |
| 3335 | PyObject *global = NULL; |
| 3336 | PyObject *module_name; |
| 3337 | PyObject *global_name; |
| 3338 | Py_ssize_t len; |
| 3339 | char *s; |
| 3340 | |
| 3341 | if ((len = unpickler_readline(self, &s)) < 0) |
| 3342 | return -1; |
| 3343 | if (len < 2) |
| 3344 | return bad_readline(); |
| 3345 | module_name = PyUnicode_DecodeUTF8(s, len - 1, "strict"); |
| 3346 | if (!module_name) |
| 3347 | return -1; |
| 3348 | |
| 3349 | if ((len = unpickler_readline(self, &s)) >= 0) { |
| 3350 | if (len < 2) { |
| 3351 | Py_DECREF(module_name); |
| 3352 | return bad_readline(); |
| 3353 | } |
| 3354 | global_name = PyUnicode_DecodeUTF8(s, len - 1, "strict"); |
| 3355 | if (global_name) { |
| 3356 | global = find_class(self, module_name, global_name); |
| 3357 | Py_DECREF(global_name); |
| 3358 | } |
| 3359 | } |
| 3360 | Py_DECREF(module_name); |
| 3361 | |
| 3362 | if (global == NULL) |
| 3363 | return -1; |
| 3364 | PDATA_PUSH(self->stack, global, -1); |
| 3365 | return 0; |
| 3366 | } |
| 3367 | |
| 3368 | static int |
| 3369 | load_persid(UnpicklerObject *self) |
| 3370 | { |
| 3371 | PyObject *pid; |
| 3372 | Py_ssize_t len; |
| 3373 | char *s; |
| 3374 | |
| 3375 | if (self->pers_func) { |
| 3376 | if ((len = unpickler_readline(self, &s)) < 0) |
| 3377 | return -1; |
| 3378 | if (len < 2) |
| 3379 | return bad_readline(); |
| 3380 | |
| 3381 | pid = PyBytes_FromStringAndSize(s, len - 1); |
| 3382 | if (pid == NULL) |
| 3383 | return -1; |
| 3384 | |
| 3385 | /* Ugh... this does not leak since unpickler_call() steals the |
| 3386 | reference to pid first. */ |
| 3387 | pid = unpickler_call(self, self->pers_func, pid); |
| 3388 | if (pid == NULL) |
| 3389 | return -1; |
| 3390 | |
| 3391 | PDATA_PUSH(self->stack, pid, -1); |
| 3392 | return 0; |
| 3393 | } |
| 3394 | else { |
| 3395 | PyErr_SetString(UnpicklingError, |
| 3396 | "A load persistent id instruction was encountered,\n" |
| 3397 | "but no persistent_load function was specified."); |
| 3398 | return -1; |
| 3399 | } |
| 3400 | } |
| 3401 | |
| 3402 | static int |
| 3403 | load_binpersid(UnpicklerObject *self) |
| 3404 | { |
| 3405 | PyObject *pid; |
| 3406 | |
| 3407 | if (self->pers_func) { |
| 3408 | PDATA_POP(self->stack, pid); |
| 3409 | if (pid == NULL) |
| 3410 | return -1; |
| 3411 | |
| 3412 | /* Ugh... this does not leak since unpickler_call() steals the |
| 3413 | reference to pid first. */ |
| 3414 | pid = unpickler_call(self, self->pers_func, pid); |
| 3415 | if (pid == NULL) |
| 3416 | return -1; |
| 3417 | |
| 3418 | PDATA_PUSH(self->stack, pid, -1); |
| 3419 | return 0; |
| 3420 | } |
| 3421 | else { |
| 3422 | PyErr_SetString(UnpicklingError, |
| 3423 | "A load persistent id instruction was encountered,\n" |
| 3424 | "but no persistent_load function was specified."); |
| 3425 | return -1; |
| 3426 | } |
| 3427 | } |
| 3428 | |
| 3429 | static int |
| 3430 | load_pop(UnpicklerObject *self) |
| 3431 | { |
| 3432 | int len; |
| 3433 | |
| 3434 | if ((len = self->stack->length) <= 0) |
| 3435 | return stack_underflow(); |
| 3436 | |
| 3437 | /* Note that we split the (pickle.py) stack into two stacks, |
| 3438 | * an object stack and a mark stack. We have to be clever and |
| 3439 | * pop the right one. We do this by looking at the top of the |
| 3440 | * mark stack. |
| 3441 | */ |
| 3442 | |
| 3443 | if ((self->num_marks > 0) && (self->marks[self->num_marks - 1] == len)) |
| 3444 | self->num_marks--; |
| 3445 | else { |
| 3446 | len--; |
| 3447 | Py_DECREF(self->stack->data[len]); |
| 3448 | self->stack->length = len; |
| 3449 | } |
| 3450 | |
| 3451 | return 0; |
| 3452 | } |
| 3453 | |
| 3454 | static int |
| 3455 | load_pop_mark(UnpicklerObject *self) |
| 3456 | { |
| 3457 | int i; |
| 3458 | |
| 3459 | if ((i = marker(self)) < 0) |
| 3460 | return -1; |
| 3461 | |
| 3462 | Pdata_clear(self->stack, i); |
| 3463 | |
| 3464 | return 0; |
| 3465 | } |
| 3466 | |
| 3467 | static int |
| 3468 | load_dup(UnpicklerObject *self) |
| 3469 | { |
| 3470 | PyObject *last; |
| 3471 | int len; |
| 3472 | |
| 3473 | if ((len = self->stack->length) <= 0) |
| 3474 | return stack_underflow(); |
| 3475 | last = self->stack->data[len - 1]; |
| 3476 | PDATA_APPEND(self->stack, last, -1); |
| 3477 | return 0; |
| 3478 | } |
| 3479 | |
| 3480 | static int |
| 3481 | load_get(UnpicklerObject *self) |
| 3482 | { |
| 3483 | PyObject *key, *value; |
| 3484 | Py_ssize_t len; |
| 3485 | char *s; |
| 3486 | |
| 3487 | if ((len = unpickler_readline(self, &s)) < 0) |
| 3488 | return -1; |
| 3489 | if (len < 2) |
| 3490 | return bad_readline(); |
| 3491 | |
| 3492 | key = PyLong_FromString(s, NULL, 10); |
| 3493 | if (key == NULL) |
| 3494 | return -1; |
| 3495 | |
| 3496 | value = PyDict_GetItemWithError(self->memo, key); |
| 3497 | if (value == NULL) { |
| 3498 | if (!PyErr_Occurred()) |
| 3499 | PyErr_SetObject(PyExc_KeyError, key); |
| 3500 | Py_DECREF(key); |
| 3501 | return -1; |
| 3502 | } |
| 3503 | Py_DECREF(key); |
| 3504 | |
| 3505 | PDATA_APPEND(self->stack, value, -1); |
| 3506 | return 0; |
| 3507 | } |
| 3508 | |
| 3509 | static int |
| 3510 | load_binget(UnpicklerObject *self) |
| 3511 | { |
| 3512 | PyObject *key, *value; |
| 3513 | char *s; |
| 3514 | |
| 3515 | if (unpickler_read(self, &s, 1) < 0) |
| 3516 | return -1; |
| 3517 | |
| 3518 | /* Here, the unsigned cast is necessary to avoid negative values. */ |
| 3519 | key = PyLong_FromLong((long)(unsigned char)s[0]); |
| 3520 | if (key == NULL) |
| 3521 | return -1; |
| 3522 | |
| 3523 | value = PyDict_GetItemWithError(self->memo, key); |
| 3524 | if (value == NULL) { |
| 3525 | if (!PyErr_Occurred()) |
| 3526 | PyErr_SetObject(PyExc_KeyError, key); |
| 3527 | Py_DECREF(key); |
| 3528 | return -1; |
| 3529 | } |
| 3530 | Py_DECREF(key); |
| 3531 | |
| 3532 | PDATA_APPEND(self->stack, value, -1); |
| 3533 | return 0; |
| 3534 | } |
| 3535 | |
| 3536 | static int |
| 3537 | load_long_binget(UnpicklerObject *self) |
| 3538 | { |
| 3539 | PyObject *key, *value; |
| 3540 | char *s; |
| 3541 | long k; |
| 3542 | |
| 3543 | if (unpickler_read(self, &s, 4) < 0) |
| 3544 | return -1; |
| 3545 | |
| 3546 | k = (long)(unsigned char)s[0]; |
| 3547 | k |= (long)(unsigned char)s[1] << 8; |
| 3548 | k |= (long)(unsigned char)s[2] << 16; |
| 3549 | k |= (long)(unsigned char)s[3] << 24; |
| 3550 | |
| 3551 | key = PyLong_FromLong(k); |
| 3552 | if (key == NULL) |
| 3553 | return -1; |
| 3554 | |
| 3555 | value = PyDict_GetItemWithError(self->memo, key); |
| 3556 | if (value == NULL) { |
| 3557 | if (!PyErr_Occurred()) |
| 3558 | PyErr_SetObject(PyExc_KeyError, key); |
| 3559 | Py_DECREF(key); |
| 3560 | return -1; |
| 3561 | } |
| 3562 | Py_DECREF(key); |
| 3563 | |
| 3564 | PDATA_APPEND(self->stack, value, -1); |
| 3565 | return 0; |
| 3566 | } |
| 3567 | |
| 3568 | /* Push an object from the extension registry (EXT[124]). nbytes is |
| 3569 | * the number of bytes following the opcode, holding the index (code) value. |
| 3570 | */ |
| 3571 | static int |
| 3572 | load_extension(UnpicklerObject *self, int nbytes) |
| 3573 | { |
| 3574 | char *codebytes; /* the nbytes bytes after the opcode */ |
| 3575 | long code; /* calc_binint returns long */ |
| 3576 | PyObject *py_code; /* code as a Python int */ |
| 3577 | PyObject *obj; /* the object to push */ |
| 3578 | PyObject *pair; /* (module_name, class_name) */ |
| 3579 | PyObject *module_name, *class_name; |
| 3580 | |
| 3581 | assert(nbytes == 1 || nbytes == 2 || nbytes == 4); |
| 3582 | if (unpickler_read(self, &codebytes, nbytes) < 0) |
| 3583 | return -1; |
| 3584 | code = calc_binint(codebytes, nbytes); |
| 3585 | if (code <= 0) { /* note that 0 is forbidden */ |
| 3586 | /* Corrupt or hostile pickle. */ |
| 3587 | PyErr_SetString(UnpicklingError, "EXT specifies code <= 0"); |
| 3588 | return -1; |
| 3589 | } |
| 3590 | |
| 3591 | /* Look for the code in the cache. */ |
| 3592 | py_code = PyLong_FromLong(code); |
| 3593 | if (py_code == NULL) |
| 3594 | return -1; |
| 3595 | obj = PyDict_GetItem(extension_cache, py_code); |
| 3596 | if (obj != NULL) { |
| 3597 | /* Bingo. */ |
| 3598 | Py_DECREF(py_code); |
| 3599 | PDATA_APPEND(self->stack, obj, -1); |
| 3600 | return 0; |
| 3601 | } |
| 3602 | |
| 3603 | /* Look up the (module_name, class_name) pair. */ |
| 3604 | pair = PyDict_GetItem(inverted_registry, py_code); |
| 3605 | if (pair == NULL) { |
| 3606 | Py_DECREF(py_code); |
| 3607 | PyErr_Format(PyExc_ValueError, "unregistered extension " |
| 3608 | "code %ld", code); |
| 3609 | return -1; |
| 3610 | } |
| 3611 | /* Since the extension registry is manipulable via Python code, |
| 3612 | * confirm that pair is really a 2-tuple of strings. |
| 3613 | */ |
| 3614 | if (!PyTuple_Check(pair) || PyTuple_Size(pair) != 2 || |
| 3615 | !PyUnicode_Check(module_name = PyTuple_GET_ITEM(pair, 0)) || |
| 3616 | !PyUnicode_Check(class_name = PyTuple_GET_ITEM(pair, 1))) { |
| 3617 | Py_DECREF(py_code); |
| 3618 | PyErr_Format(PyExc_ValueError, "_inverted_registry[%ld] " |
| 3619 | "isn't a 2-tuple of strings", code); |
| 3620 | return -1; |
| 3621 | } |
| 3622 | /* Load the object. */ |
| 3623 | obj = find_class(self, module_name, class_name); |
| 3624 | if (obj == NULL) { |
| 3625 | Py_DECREF(py_code); |
| 3626 | return -1; |
| 3627 | } |
| 3628 | /* Cache code -> obj. */ |
| 3629 | code = PyDict_SetItem(extension_cache, py_code, obj); |
| 3630 | Py_DECREF(py_code); |
| 3631 | if (code < 0) { |
| 3632 | Py_DECREF(obj); |
| 3633 | return -1; |
| 3634 | } |
| 3635 | PDATA_PUSH(self->stack, obj, -1); |
| 3636 | return 0; |
| 3637 | } |
| 3638 | |
| 3639 | static int |
| 3640 | load_put(UnpicklerObject *self) |
| 3641 | { |
| 3642 | PyObject *key, *value; |
| 3643 | Py_ssize_t len; |
| 3644 | char *s; |
| 3645 | int x; |
| 3646 | |
| 3647 | if ((len = unpickler_readline(self, &s)) < 0) |
| 3648 | return -1; |
| 3649 | if (len < 2) |
| 3650 | return bad_readline(); |
| 3651 | if ((x = self->stack->length) <= 0) |
| 3652 | return stack_underflow(); |
| 3653 | |
| 3654 | key = PyLong_FromString(s, NULL, 10); |
| 3655 | if (key == NULL) |
| 3656 | return -1; |
| 3657 | value = self->stack->data[x - 1]; |
| 3658 | |
| 3659 | x = PyDict_SetItem(self->memo, key, value); |
| 3660 | Py_DECREF(key); |
| 3661 | return x; |
| 3662 | } |
| 3663 | |
| 3664 | static int |
| 3665 | load_binput(UnpicklerObject *self) |
| 3666 | { |
| 3667 | PyObject *key, *value; |
| 3668 | char *s; |
| 3669 | int x; |
| 3670 | |
| 3671 | if (unpickler_read(self, &s, 1) < 0) |
| 3672 | return -1; |
| 3673 | if ((x = self->stack->length) <= 0) |
| 3674 | return stack_underflow(); |
| 3675 | |
| 3676 | key = PyLong_FromLong((long)(unsigned char)s[0]); |
| 3677 | if (key == NULL) |
| 3678 | return -1; |
| 3679 | value = self->stack->data[x - 1]; |
| 3680 | |
| 3681 | x = PyDict_SetItem(self->memo, key, value); |
| 3682 | Py_DECREF(key); |
| 3683 | return x; |
| 3684 | } |
| 3685 | |
| 3686 | static int |
| 3687 | load_long_binput(UnpicklerObject *self) |
| 3688 | { |
| 3689 | PyObject *key, *value; |
| 3690 | long k; |
| 3691 | char *s; |
| 3692 | int x; |
| 3693 | |
| 3694 | if (unpickler_read(self, &s, 4) < 0) |
| 3695 | return -1; |
| 3696 | if ((x = self->stack->length) <= 0) |
| 3697 | return stack_underflow(); |
| 3698 | |
| 3699 | k = (long)(unsigned char)s[0]; |
| 3700 | k |= (long)(unsigned char)s[1] << 8; |
| 3701 | k |= (long)(unsigned char)s[2] << 16; |
| 3702 | k |= (long)(unsigned char)s[3] << 24; |
| 3703 | |
| 3704 | key = PyLong_FromLong(k); |
| 3705 | if (key == NULL) |
| 3706 | return -1; |
| 3707 | value = self->stack->data[x - 1]; |
| 3708 | |
| 3709 | x = PyDict_SetItem(self->memo, key, value); |
| 3710 | Py_DECREF(key); |
| 3711 | return x; |
| 3712 | } |
| 3713 | |
| 3714 | static int |
| 3715 | do_append(UnpicklerObject *self, int x) |
| 3716 | { |
| 3717 | PyObject *value; |
| 3718 | PyObject *list; |
| 3719 | int len, i; |
| 3720 | |
| 3721 | len = self->stack->length; |
| 3722 | if (x > len || x <= 0) |
| 3723 | return stack_underflow(); |
| 3724 | if (len == x) /* nothing to do */ |
| 3725 | return 0; |
| 3726 | |
| 3727 | list = self->stack->data[x - 1]; |
| 3728 | |
| 3729 | if (PyList_Check(list)) { |
| 3730 | PyObject *slice; |
| 3731 | Py_ssize_t list_len; |
| 3732 | |
| 3733 | slice = Pdata_poplist(self->stack, x); |
| 3734 | if (!slice) |
| 3735 | return -1; |
| 3736 | list_len = PyList_GET_SIZE(list); |
| 3737 | i = PyList_SetSlice(list, list_len, list_len, slice); |
| 3738 | Py_DECREF(slice); |
| 3739 | return i; |
| 3740 | } |
| 3741 | else { |
| 3742 | PyObject *append_func; |
| 3743 | |
| 3744 | append_func = PyObject_GetAttrString(list, "append"); |
| 3745 | if (append_func == NULL) |
| 3746 | return -1; |
| 3747 | for (i = x; i < len; i++) { |
| 3748 | PyObject *result; |
| 3749 | |
| 3750 | value = self->stack->data[i]; |
| 3751 | result = unpickler_call(self, append_func, value); |
| 3752 | if (result == NULL) { |
| 3753 | Pdata_clear(self->stack, i + 1); |
| 3754 | self->stack->length = x; |
| 3755 | return -1; |
| 3756 | } |
| 3757 | Py_DECREF(result); |
| 3758 | } |
| 3759 | self->stack->length = x; |
| 3760 | } |
| 3761 | |
| 3762 | return 0; |
| 3763 | } |
| 3764 | |
| 3765 | static int |
| 3766 | load_append(UnpicklerObject *self) |
| 3767 | { |
| 3768 | return do_append(self, self->stack->length - 1); |
| 3769 | } |
| 3770 | |
| 3771 | static int |
| 3772 | load_appends(UnpicklerObject *self) |
| 3773 | { |
| 3774 | return do_append(self, marker(self)); |
| 3775 | } |
| 3776 | |
| 3777 | static int |
| 3778 | do_setitems(UnpicklerObject *self, int x) |
| 3779 | { |
| 3780 | PyObject *value, *key; |
| 3781 | PyObject *dict; |
| 3782 | int len, i; |
| 3783 | int status = 0; |
| 3784 | |
| 3785 | len = self->stack->length; |
| 3786 | if (x > len || x <= 0) |
| 3787 | return stack_underflow(); |
| 3788 | if (len == x) /* nothing to do */ |
| 3789 | return 0; |
| 3790 | if ((len - x) % 2 != 0) { |
| 3791 | /* Currupt or hostile pickle -- we never write one like this. */ |
| 3792 | PyErr_SetString(UnpicklingError, "odd number of items for SETITEMS"); |
| 3793 | return -1; |
| 3794 | } |
| 3795 | |
| 3796 | /* Here, dict does not actually need to be a PyDict; it could be anything |
| 3797 | that supports the __setitem__ attribute. */ |
| 3798 | dict = self->stack->data[x - 1]; |
| 3799 | |
| 3800 | for (i = x + 1; i < len; i += 2) { |
| 3801 | key = self->stack->data[i - 1]; |
| 3802 | value = self->stack->data[i]; |
| 3803 | if (PyObject_SetItem(dict, key, value) < 0) { |
| 3804 | status = -1; |
| 3805 | break; |
| 3806 | } |
| 3807 | } |
| 3808 | |
| 3809 | Pdata_clear(self->stack, x); |
| 3810 | return status; |
| 3811 | } |
| 3812 | |
| 3813 | static int |
| 3814 | load_setitem(UnpicklerObject *self) |
| 3815 | { |
| 3816 | return do_setitems(self, self->stack->length - 2); |
| 3817 | } |
| 3818 | |
| 3819 | static int |
| 3820 | load_setitems(UnpicklerObject *self) |
| 3821 | { |
| 3822 | return do_setitems(self, marker(self)); |
| 3823 | } |
| 3824 | |
| 3825 | static int |
| 3826 | load_build(UnpicklerObject *self) |
| 3827 | { |
| 3828 | PyObject *state, *inst, *slotstate; |
| 3829 | PyObject *setstate; |
| 3830 | int status = 0; |
| 3831 | |
| 3832 | /* Stack is ... instance, state. We want to leave instance at |
| 3833 | * the stack top, possibly mutated via instance.__setstate__(state). |
| 3834 | */ |
| 3835 | if (self->stack->length < 2) |
| 3836 | return stack_underflow(); |
| 3837 | |
| 3838 | PDATA_POP(self->stack, state); |
| 3839 | if (state == NULL) |
| 3840 | return -1; |
| 3841 | |
| 3842 | inst = self->stack->data[self->stack->length - 1]; |
| 3843 | |
| 3844 | setstate = PyObject_GetAttrString(inst, "__setstate__"); |
| 3845 | if (setstate == NULL && PyErr_ExceptionMatches(PyExc_AttributeError)) { |
| 3846 | PyErr_Clear(); |
| 3847 | } |
| 3848 | else { |
| 3849 | PyObject *result; |
| 3850 | |
| 3851 | /* The explicit __setstate__ is responsible for everything. */ |
| 3852 | result = unpickler_call(self, setstate, state); |
| 3853 | Py_DECREF(setstate); |
| 3854 | if (result == NULL) |
| 3855 | return -1; |
| 3856 | Py_DECREF(result); |
| 3857 | return 0; |
| 3858 | } |
| 3859 | |
| 3860 | /* A default __setstate__. First see whether state embeds a |
| 3861 | * slot state dict too (a proto 2 addition). |
| 3862 | */ |
| 3863 | if (PyTuple_Check(state) && Py_SIZE(state) == 2) { |
| 3864 | PyObject *tmp = state; |
| 3865 | |
| 3866 | state = PyTuple_GET_ITEM(tmp, 0); |
| 3867 | slotstate = PyTuple_GET_ITEM(tmp, 1); |
| 3868 | Py_INCREF(state); |
| 3869 | Py_INCREF(slotstate); |
| 3870 | Py_DECREF(tmp); |
| 3871 | } |
| 3872 | else |
| 3873 | slotstate = NULL; |
| 3874 | |
| 3875 | /* Set inst.__dict__ from the state dict (if any). */ |
| 3876 | if (state != Py_None) { |
| 3877 | PyObject *dict; |
| 3878 | |
| 3879 | if (!PyDict_Check(state)) { |
| 3880 | PyErr_SetString(UnpicklingError, "state is not a dictionary"); |
| 3881 | goto error; |
| 3882 | } |
| 3883 | dict = PyObject_GetAttrString(inst, "__dict__"); |
| 3884 | if (dict == NULL) |
| 3885 | goto error; |
| 3886 | |
| 3887 | PyDict_Update(dict, state); |
| 3888 | Py_DECREF(dict); |
| 3889 | } |
| 3890 | |
| 3891 | /* Also set instance attributes from the slotstate dict (if any). */ |
| 3892 | if (slotstate != NULL) { |
| 3893 | PyObject *d_key, *d_value; |
| 3894 | Py_ssize_t i; |
| 3895 | |
| 3896 | if (!PyDict_Check(slotstate)) { |
| 3897 | PyErr_SetString(UnpicklingError, |
| 3898 | "slot state is not a dictionary"); |
| 3899 | goto error; |
| 3900 | } |
| 3901 | i = 0; |
| 3902 | while (PyDict_Next(slotstate, &i, &d_key, &d_value)) { |
| 3903 | if (PyObject_SetAttr(inst, d_key, d_value) < 0) |
| 3904 | goto error; |
| 3905 | } |
| 3906 | } |
| 3907 | |
| 3908 | if (0) { |
| 3909 | error: |
| 3910 | status = -1; |
| 3911 | } |
| 3912 | |
| 3913 | Py_DECREF(state); |
| 3914 | Py_XDECREF(slotstate); |
| 3915 | return status; |
| 3916 | } |
| 3917 | |
| 3918 | static int |
| 3919 | load_mark(UnpicklerObject *self) |
| 3920 | { |
| 3921 | |
| 3922 | /* Note that we split the (pickle.py) stack into two stacks, an |
| 3923 | * object stack and a mark stack. Here we push a mark onto the |
| 3924 | * mark stack. |
| 3925 | */ |
| 3926 | |
| 3927 | if ((self->num_marks + 1) >= self->marks_size) { |
| 3928 | size_t alloc; |
| 3929 | int *marks; |
| 3930 | |
| 3931 | /* Use the size_t type to check for overflow. */ |
| 3932 | alloc = ((size_t)self->num_marks << 1) + 20; |
Alexandre Vassalotti | 7634ff5 | 2008-06-13 02:16:06 +0000 | [diff] [blame] | 3933 | if (alloc > PY_SSIZE_T_MAX || |
| 3934 | alloc <= ((size_t)self->num_marks + 1)) { |
Alexandre Vassalotti | ca2d610 | 2008-06-12 18:26:05 +0000 | [diff] [blame] | 3935 | PyErr_NoMemory(); |
| 3936 | return -1; |
| 3937 | } |
| 3938 | |
| 3939 | if (self->marks == NULL) |
| 3940 | marks = (int *)PyMem_Malloc(alloc * sizeof(int)); |
| 3941 | else |
| 3942 | marks = (int *)PyMem_Realloc(self->marks, alloc * sizeof(int)); |
| 3943 | if (marks == NULL) { |
| 3944 | PyErr_NoMemory(); |
| 3945 | return -1; |
| 3946 | } |
| 3947 | self->marks = marks; |
| 3948 | self->marks_size = (Py_ssize_t)alloc; |
| 3949 | } |
| 3950 | |
| 3951 | self->marks[self->num_marks++] = self->stack->length; |
| 3952 | |
| 3953 | return 0; |
| 3954 | } |
| 3955 | |
| 3956 | static int |
| 3957 | load_reduce(UnpicklerObject *self) |
| 3958 | { |
| 3959 | PyObject *callable = NULL; |
| 3960 | PyObject *argtup = NULL; |
| 3961 | PyObject *obj = NULL; |
| 3962 | |
| 3963 | PDATA_POP(self->stack, argtup); |
| 3964 | if (argtup == NULL) |
| 3965 | return -1; |
| 3966 | PDATA_POP(self->stack, callable); |
| 3967 | if (callable) { |
| 3968 | obj = instantiate(callable, argtup); |
| 3969 | Py_DECREF(callable); |
| 3970 | } |
| 3971 | Py_DECREF(argtup); |
| 3972 | |
| 3973 | if (obj == NULL) |
| 3974 | return -1; |
| 3975 | |
| 3976 | PDATA_PUSH(self->stack, obj, -1); |
| 3977 | return 0; |
| 3978 | } |
| 3979 | |
| 3980 | /* Just raises an error if we don't know the protocol specified. PROTO |
| 3981 | * is the first opcode for protocols >= 2. |
| 3982 | */ |
| 3983 | static int |
| 3984 | load_proto(UnpicklerObject *self) |
| 3985 | { |
| 3986 | char *s; |
| 3987 | int i; |
| 3988 | |
| 3989 | if (unpickler_read(self, &s, 1) < 0) |
| 3990 | return -1; |
| 3991 | |
| 3992 | i = (unsigned char)s[0]; |
| 3993 | if (i <= HIGHEST_PROTOCOL) |
| 3994 | return 0; |
| 3995 | |
| 3996 | PyErr_Format(PyExc_ValueError, "unsupported pickle protocol: %d", i); |
| 3997 | return -1; |
| 3998 | } |
| 3999 | |
| 4000 | static PyObject * |
| 4001 | load(UnpicklerObject *self) |
| 4002 | { |
| 4003 | PyObject *err; |
| 4004 | PyObject *value = NULL; |
| 4005 | char *s; |
| 4006 | |
| 4007 | self->num_marks = 0; |
| 4008 | if (self->stack->length) |
| 4009 | Pdata_clear(self->stack, 0); |
| 4010 | |
| 4011 | /* Convenient macros for the dispatch while-switch loop just below. */ |
| 4012 | #define OP(opcode, load_func) \ |
| 4013 | case opcode: if (load_func(self) < 0) break; continue; |
| 4014 | |
| 4015 | #define OP_ARG(opcode, load_func, arg) \ |
| 4016 | case opcode: if (load_func(self, (arg)) < 0) break; continue; |
| 4017 | |
| 4018 | while (1) { |
| 4019 | if (unpickler_read(self, &s, 1) < 0) |
| 4020 | break; |
| 4021 | |
| 4022 | switch ((enum opcode)s[0]) { |
| 4023 | OP(NONE, load_none) |
| 4024 | OP(BININT, load_binint) |
| 4025 | OP(BININT1, load_binint1) |
| 4026 | OP(BININT2, load_binint2) |
| 4027 | OP(INT, load_int) |
| 4028 | OP(LONG, load_long) |
| 4029 | OP_ARG(LONG1, load_counted_long, 1) |
| 4030 | OP_ARG(LONG4, load_counted_long, 4) |
| 4031 | OP(FLOAT, load_float) |
| 4032 | OP(BINFLOAT, load_binfloat) |
| 4033 | OP(BINBYTES, load_binbytes) |
| 4034 | OP(SHORT_BINBYTES, load_short_binbytes) |
| 4035 | OP(BINSTRING, load_binstring) |
| 4036 | OP(SHORT_BINSTRING, load_short_binstring) |
| 4037 | OP(STRING, load_string) |
| 4038 | OP(UNICODE, load_unicode) |
| 4039 | OP(BINUNICODE, load_binunicode) |
| 4040 | OP_ARG(EMPTY_TUPLE, load_counted_tuple, 0) |
| 4041 | OP_ARG(TUPLE1, load_counted_tuple, 1) |
| 4042 | OP_ARG(TUPLE2, load_counted_tuple, 2) |
| 4043 | OP_ARG(TUPLE3, load_counted_tuple, 3) |
| 4044 | OP(TUPLE, load_tuple) |
| 4045 | OP(EMPTY_LIST, load_empty_list) |
| 4046 | OP(LIST, load_list) |
| 4047 | OP(EMPTY_DICT, load_empty_dict) |
| 4048 | OP(DICT, load_dict) |
| 4049 | OP(OBJ, load_obj) |
| 4050 | OP(INST, load_inst) |
| 4051 | OP(NEWOBJ, load_newobj) |
| 4052 | OP(GLOBAL, load_global) |
| 4053 | OP(APPEND, load_append) |
| 4054 | OP(APPENDS, load_appends) |
| 4055 | OP(BUILD, load_build) |
| 4056 | OP(DUP, load_dup) |
| 4057 | OP(BINGET, load_binget) |
| 4058 | OP(LONG_BINGET, load_long_binget) |
| 4059 | OP(GET, load_get) |
| 4060 | OP(MARK, load_mark) |
| 4061 | OP(BINPUT, load_binput) |
| 4062 | OP(LONG_BINPUT, load_long_binput) |
| 4063 | OP(PUT, load_put) |
| 4064 | OP(POP, load_pop) |
| 4065 | OP(POP_MARK, load_pop_mark) |
| 4066 | OP(SETITEM, load_setitem) |
| 4067 | OP(SETITEMS, load_setitems) |
| 4068 | OP(PERSID, load_persid) |
| 4069 | OP(BINPERSID, load_binpersid) |
| 4070 | OP(REDUCE, load_reduce) |
| 4071 | OP(PROTO, load_proto) |
| 4072 | OP_ARG(EXT1, load_extension, 1) |
| 4073 | OP_ARG(EXT2, load_extension, 2) |
| 4074 | OP_ARG(EXT4, load_extension, 4) |
| 4075 | OP_ARG(NEWTRUE, load_bool, Py_True) |
| 4076 | OP_ARG(NEWFALSE, load_bool, Py_False) |
| 4077 | |
| 4078 | case STOP: |
| 4079 | break; |
| 4080 | |
| 4081 | case '\0': |
| 4082 | PyErr_SetNone(PyExc_EOFError); |
| 4083 | return NULL; |
| 4084 | |
| 4085 | default: |
| 4086 | PyErr_Format(UnpicklingError, |
| 4087 | "invalid load key, '%c'.", s[0]); |
| 4088 | return NULL; |
| 4089 | } |
| 4090 | |
| 4091 | break; /* and we are done! */ |
| 4092 | } |
| 4093 | |
| 4094 | /* XXX: It is not clear what this is actually for. */ |
| 4095 | if ((err = PyErr_Occurred())) { |
| 4096 | if (err == PyExc_EOFError) { |
| 4097 | PyErr_SetNone(PyExc_EOFError); |
| 4098 | } |
| 4099 | return NULL; |
| 4100 | } |
| 4101 | |
| 4102 | PDATA_POP(self->stack, value); |
| 4103 | return value; |
| 4104 | } |
| 4105 | |
| 4106 | PyDoc_STRVAR(Unpickler_load_doc, |
| 4107 | "load() -> object. Load a pickle." |
| 4108 | "\n" |
| 4109 | "Read a pickled object representation from the open file object given in\n" |
| 4110 | "the constructor, and return the reconstituted object hierarchy specified\n" |
| 4111 | "therein.\n"); |
| 4112 | |
| 4113 | static PyObject * |
| 4114 | Unpickler_load(UnpicklerObject *self) |
| 4115 | { |
| 4116 | /* Check whether the Unpickler was initialized correctly. This prevents |
| 4117 | segfaulting if a subclass overridden __init__ with a function that does |
| 4118 | not call Unpickler.__init__(). Here, we simply ensure that self->read |
| 4119 | is not NULL. */ |
| 4120 | if (self->read == NULL) { |
| 4121 | PyErr_Format(UnpicklingError, |
| 4122 | "Unpickler.__init__() was not called by %s.__init__()", |
| 4123 | Py_TYPE(self)->tp_name); |
| 4124 | return NULL; |
| 4125 | } |
| 4126 | |
| 4127 | return load(self); |
| 4128 | } |
| 4129 | |
| 4130 | /* The name of find_class() is misleading. In newer pickle protocols, this |
| 4131 | function is used for loading any global (i.e., functions), not just |
| 4132 | classes. The name is kept only for backward compatibility. */ |
| 4133 | |
| 4134 | PyDoc_STRVAR(Unpickler_find_class_doc, |
| 4135 | "find_class(module_name, global_name) -> object.\n" |
| 4136 | "\n" |
| 4137 | "Return an object from a specified module, importing the module if\n" |
| 4138 | "necessary. Subclasses may override this method (e.g. to restrict\n" |
| 4139 | "unpickling of arbitrary classes and functions).\n" |
| 4140 | "\n" |
| 4141 | "This method is called whenever a class or a function object is\n" |
| 4142 | "needed. Both arguments passed are str objects.\n"); |
| 4143 | |
| 4144 | static PyObject * |
| 4145 | Unpickler_find_class(UnpicklerObject *self, PyObject *args) |
| 4146 | { |
| 4147 | PyObject *global; |
| 4148 | PyObject *modules_dict; |
| 4149 | PyObject *module; |
| 4150 | PyObject *module_name, *global_name; |
| 4151 | |
| 4152 | if (!PyArg_UnpackTuple(args, "find_class", 2, 2, |
| 4153 | &module_name, &global_name)) |
| 4154 | return NULL; |
| 4155 | |
| 4156 | modules_dict = PySys_GetObject("modules"); |
| 4157 | if (modules_dict == NULL) |
| 4158 | return NULL; |
| 4159 | |
| 4160 | module = PyDict_GetItem(modules_dict, module_name); |
| 4161 | if (module == NULL) { |
| 4162 | module = PyImport_Import(module_name); |
| 4163 | if (module == NULL) |
| 4164 | return NULL; |
| 4165 | global = PyObject_GetAttr(module, global_name); |
| 4166 | Py_DECREF(module); |
| 4167 | } |
| 4168 | else { |
| 4169 | global = PyObject_GetAttr(module, global_name); |
| 4170 | } |
| 4171 | return global; |
| 4172 | } |
| 4173 | |
| 4174 | static struct PyMethodDef Unpickler_methods[] = { |
| 4175 | {"load", (PyCFunction)Unpickler_load, METH_NOARGS, |
| 4176 | Unpickler_load_doc}, |
| 4177 | {"find_class", (PyCFunction)Unpickler_find_class, METH_VARARGS, |
| 4178 | Unpickler_find_class_doc}, |
| 4179 | {NULL, NULL} /* sentinel */ |
| 4180 | }; |
| 4181 | |
| 4182 | static void |
| 4183 | Unpickler_dealloc(UnpicklerObject *self) |
| 4184 | { |
| 4185 | PyObject_GC_UnTrack((PyObject *)self); |
| 4186 | Py_XDECREF(self->readline); |
| 4187 | Py_XDECREF(self->read); |
| 4188 | Py_XDECREF(self->memo); |
| 4189 | Py_XDECREF(self->stack); |
| 4190 | Py_XDECREF(self->pers_func); |
| 4191 | Py_XDECREF(self->arg); |
| 4192 | Py_XDECREF(self->last_string); |
| 4193 | |
| 4194 | PyMem_Free(self->marks); |
| 4195 | free(self->encoding); |
| 4196 | free(self->errors); |
| 4197 | |
| 4198 | Py_TYPE(self)->tp_free((PyObject *)self); |
| 4199 | } |
| 4200 | |
| 4201 | static int |
| 4202 | Unpickler_traverse(UnpicklerObject *self, visitproc visit, void *arg) |
| 4203 | { |
| 4204 | Py_VISIT(self->readline); |
| 4205 | Py_VISIT(self->read); |
| 4206 | Py_VISIT(self->memo); |
| 4207 | Py_VISIT(self->stack); |
| 4208 | Py_VISIT(self->pers_func); |
| 4209 | Py_VISIT(self->arg); |
| 4210 | Py_VISIT(self->last_string); |
| 4211 | return 0; |
| 4212 | } |
| 4213 | |
| 4214 | static int |
| 4215 | Unpickler_clear(UnpicklerObject *self) |
| 4216 | { |
| 4217 | Py_CLEAR(self->readline); |
| 4218 | Py_CLEAR(self->read); |
| 4219 | Py_CLEAR(self->memo); |
| 4220 | Py_CLEAR(self->stack); |
| 4221 | Py_CLEAR(self->pers_func); |
| 4222 | Py_CLEAR(self->arg); |
| 4223 | Py_CLEAR(self->last_string); |
| 4224 | |
| 4225 | PyMem_Free(self->marks); |
| 4226 | self->marks = NULL; |
| 4227 | free(self->encoding); |
| 4228 | self->encoding = NULL; |
| 4229 | free(self->errors); |
| 4230 | self->errors = NULL; |
| 4231 | |
| 4232 | return 0; |
| 4233 | } |
| 4234 | |
| 4235 | PyDoc_STRVAR(Unpickler_doc, |
| 4236 | "Unpickler(file, *, encoding='ASCII', errors='strict')" |
| 4237 | "\n" |
| 4238 | "This takes a binary file for reading a pickle data stream.\n" |
| 4239 | "\n" |
| 4240 | "The protocol version of the pickle is detected automatically, so no\n" |
| 4241 | "proto argument is needed.\n" |
| 4242 | "\n" |
| 4243 | "The file-like object must have two methods, a read() method\n" |
| 4244 | "that takes an integer argument, and a readline() method that\n" |
| 4245 | "requires no arguments. Both methods should return bytes.\n" |
| 4246 | "Thus file-like object can be a binary file object opened for\n" |
| 4247 | "reading, a BytesIO object, or any other custom object that\n" |
| 4248 | "meets this interface.\n" |
| 4249 | "\n" |
| 4250 | "Optional keyword arguments are encoding and errors, which are\n" |
| 4251 | "used to decode 8-bit string instances pickled by Python 2.x.\n" |
| 4252 | "These default to 'ASCII' and 'strict', respectively.\n"); |
| 4253 | |
| 4254 | static int |
| 4255 | Unpickler_init(UnpicklerObject *self, PyObject *args, PyObject *kwds) |
| 4256 | { |
| 4257 | static char *kwlist[] = {"file", "encoding", "errors", 0}; |
| 4258 | PyObject *file; |
| 4259 | char *encoding = NULL; |
| 4260 | char *errors = NULL; |
| 4261 | |
| 4262 | /* XXX: That is an horrible error message. But, I don't know how to do |
| 4263 | better... */ |
| 4264 | if (Py_SIZE(args) != 1) { |
| 4265 | PyErr_Format(PyExc_TypeError, |
| 4266 | "%s takes exactly one positional argument (%zd given)", |
| 4267 | Py_TYPE(self)->tp_name, Py_SIZE(args)); |
| 4268 | return -1; |
| 4269 | } |
| 4270 | |
| 4271 | /* Arguments parsing needs to be done in the __init__() method to allow |
| 4272 | subclasses to define their own __init__() method, which may (or may |
| 4273 | not) support Unpickler arguments. However, this means we need to be |
| 4274 | extra careful in the other Unpickler methods, since a subclass could |
| 4275 | forget to call Unpickler.__init__() thus breaking our internal |
| 4276 | invariants. */ |
| 4277 | if (!PyArg_ParseTupleAndKeywords(args, kwds, "O|ss:Unpickler", kwlist, |
| 4278 | &file, &encoding, &errors)) |
| 4279 | return -1; |
| 4280 | |
| 4281 | /* In case of multiple __init__() calls, clear previous content. */ |
| 4282 | if (self->read != NULL) |
| 4283 | (void)Unpickler_clear(self); |
| 4284 | |
| 4285 | self->read = PyObject_GetAttrString(file, "read"); |
| 4286 | self->readline = PyObject_GetAttrString(file, "readline"); |
| 4287 | if (self->readline == NULL || self->read == NULL) |
| 4288 | return -1; |
| 4289 | |
| 4290 | if (encoding == NULL) |
| 4291 | encoding = "ASCII"; |
| 4292 | if (errors == NULL) |
| 4293 | errors = "strict"; |
| 4294 | |
| 4295 | self->encoding = strdup(encoding); |
| 4296 | self->errors = strdup(errors); |
| 4297 | if (self->encoding == NULL || self->errors == NULL) { |
| 4298 | PyErr_NoMemory(); |
| 4299 | return -1; |
| 4300 | } |
| 4301 | |
| 4302 | if (PyObject_HasAttrString((PyObject *)self, "persistent_load")) { |
| 4303 | self->pers_func = PyObject_GetAttrString((PyObject *)self, |
| 4304 | "persistent_load"); |
| 4305 | if (self->pers_func == NULL) |
| 4306 | return -1; |
| 4307 | } |
| 4308 | else { |
| 4309 | self->pers_func = NULL; |
| 4310 | } |
| 4311 | |
| 4312 | self->stack = (Pdata *)Pdata_New(); |
| 4313 | if (self->stack == NULL) |
| 4314 | return -1; |
| 4315 | |
| 4316 | self->memo = PyDict_New(); |
| 4317 | if (self->memo == NULL) |
| 4318 | return -1; |
| 4319 | |
| 4320 | return 0; |
| 4321 | } |
| 4322 | |
| 4323 | static PyObject * |
| 4324 | Unpickler_get_memo(UnpicklerObject *self) |
| 4325 | { |
| 4326 | if (self->memo == NULL) |
| 4327 | PyErr_SetString(PyExc_AttributeError, "memo"); |
| 4328 | else |
| 4329 | Py_INCREF(self->memo); |
| 4330 | return self->memo; |
| 4331 | } |
| 4332 | |
| 4333 | static int |
| 4334 | Unpickler_set_memo(UnpicklerObject *self, PyObject *value) |
| 4335 | { |
| 4336 | PyObject *tmp; |
| 4337 | |
| 4338 | if (value == NULL) { |
| 4339 | PyErr_SetString(PyExc_TypeError, |
| 4340 | "attribute deletion is not supported"); |
| 4341 | return -1; |
| 4342 | } |
| 4343 | if (!PyDict_Check(value)) { |
| 4344 | PyErr_SetString(PyExc_TypeError, "memo must be a dictionary"); |
| 4345 | return -1; |
| 4346 | } |
| 4347 | |
| 4348 | tmp = self->memo; |
| 4349 | Py_INCREF(value); |
| 4350 | self->memo = value; |
| 4351 | Py_XDECREF(tmp); |
| 4352 | |
| 4353 | return 0; |
| 4354 | } |
| 4355 | |
| 4356 | static PyObject * |
| 4357 | Unpickler_get_persload(UnpicklerObject *self) |
| 4358 | { |
| 4359 | if (self->pers_func == NULL) |
| 4360 | PyErr_SetString(PyExc_AttributeError, "persistent_load"); |
| 4361 | else |
| 4362 | Py_INCREF(self->pers_func); |
| 4363 | return self->pers_func; |
| 4364 | } |
| 4365 | |
| 4366 | static int |
| 4367 | Unpickler_set_persload(UnpicklerObject *self, PyObject *value) |
| 4368 | { |
| 4369 | PyObject *tmp; |
| 4370 | |
| 4371 | if (value == NULL) { |
| 4372 | PyErr_SetString(PyExc_TypeError, |
| 4373 | "attribute deletion is not supported"); |
| 4374 | return -1; |
| 4375 | } |
| 4376 | if (!PyCallable_Check(value)) { |
| 4377 | PyErr_SetString(PyExc_TypeError, |
| 4378 | "persistent_load must be a callable taking " |
| 4379 | "one argument"); |
| 4380 | return -1; |
| 4381 | } |
| 4382 | |
| 4383 | tmp = self->pers_func; |
| 4384 | Py_INCREF(value); |
| 4385 | self->pers_func = value; |
| 4386 | Py_XDECREF(tmp); /* self->pers_func can be NULL, so be careful. */ |
| 4387 | |
| 4388 | return 0; |
| 4389 | } |
| 4390 | |
| 4391 | static PyGetSetDef Unpickler_getsets[] = { |
| 4392 | {"memo", (getter)Unpickler_get_memo, (setter)Unpickler_set_memo}, |
| 4393 | {"persistent_load", (getter)Unpickler_get_persload, |
| 4394 | (setter)Unpickler_set_persload}, |
| 4395 | {NULL} |
| 4396 | }; |
| 4397 | |
| 4398 | static PyTypeObject Unpickler_Type = { |
| 4399 | PyVarObject_HEAD_INIT(NULL, 0) |
| 4400 | "_pickle.Unpickler", /*tp_name*/ |
| 4401 | sizeof(UnpicklerObject), /*tp_basicsize*/ |
| 4402 | 0, /*tp_itemsize*/ |
| 4403 | (destructor)Unpickler_dealloc, /*tp_dealloc*/ |
| 4404 | 0, /*tp_print*/ |
| 4405 | 0, /*tp_getattr*/ |
| 4406 | 0, /*tp_setattr*/ |
| 4407 | 0, /*tp_compare*/ |
| 4408 | 0, /*tp_repr*/ |
| 4409 | 0, /*tp_as_number*/ |
| 4410 | 0, /*tp_as_sequence*/ |
| 4411 | 0, /*tp_as_mapping*/ |
| 4412 | 0, /*tp_hash*/ |
| 4413 | 0, /*tp_call*/ |
| 4414 | 0, /*tp_str*/ |
| 4415 | 0, /*tp_getattro*/ |
| 4416 | 0, /*tp_setattro*/ |
| 4417 | 0, /*tp_as_buffer*/ |
| 4418 | Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE | Py_TPFLAGS_HAVE_GC, |
| 4419 | Unpickler_doc, /*tp_doc*/ |
| 4420 | (traverseproc)Unpickler_traverse, /*tp_traverse*/ |
| 4421 | (inquiry)Unpickler_clear, /*tp_clear*/ |
| 4422 | 0, /*tp_richcompare*/ |
| 4423 | 0, /*tp_weaklistoffset*/ |
| 4424 | 0, /*tp_iter*/ |
| 4425 | 0, /*tp_iternext*/ |
| 4426 | Unpickler_methods, /*tp_methods*/ |
| 4427 | 0, /*tp_members*/ |
| 4428 | Unpickler_getsets, /*tp_getset*/ |
| 4429 | 0, /*tp_base*/ |
| 4430 | 0, /*tp_dict*/ |
| 4431 | 0, /*tp_descr_get*/ |
| 4432 | 0, /*tp_descr_set*/ |
| 4433 | 0, /*tp_dictoffset*/ |
| 4434 | (initproc)Unpickler_init, /*tp_init*/ |
| 4435 | PyType_GenericAlloc, /*tp_alloc*/ |
| 4436 | PyType_GenericNew, /*tp_new*/ |
| 4437 | PyObject_GC_Del, /*tp_free*/ |
| 4438 | 0, /*tp_is_gc*/ |
| 4439 | }; |
| 4440 | |
| 4441 | static int |
| 4442 | init_stuff(void) |
| 4443 | { |
| 4444 | PyObject *copyreg; |
| 4445 | |
| 4446 | copyreg = PyImport_ImportModule("copyreg"); |
| 4447 | if (!copyreg) |
| 4448 | return -1; |
| 4449 | |
| 4450 | dispatch_table = PyObject_GetAttrString(copyreg, "dispatch_table"); |
| 4451 | if (!dispatch_table) |
| 4452 | goto error; |
| 4453 | |
| 4454 | extension_registry = \ |
| 4455 | PyObject_GetAttrString(copyreg, "_extension_registry"); |
| 4456 | if (!extension_registry) |
| 4457 | goto error; |
| 4458 | |
| 4459 | inverted_registry = PyObject_GetAttrString(copyreg, "_inverted_registry"); |
| 4460 | if (!inverted_registry) |
| 4461 | goto error; |
| 4462 | |
| 4463 | extension_cache = PyObject_GetAttrString(copyreg, "_extension_cache"); |
| 4464 | if (!extension_cache) |
| 4465 | goto error; |
| 4466 | |
| 4467 | Py_DECREF(copyreg); |
| 4468 | |
| 4469 | empty_tuple = PyTuple_New(0); |
| 4470 | if (empty_tuple == NULL) |
| 4471 | return -1; |
| 4472 | |
| 4473 | two_tuple = PyTuple_New(2); |
| 4474 | if (two_tuple == NULL) |
| 4475 | return -1; |
| 4476 | /* We use this temp container with no regard to refcounts, or to |
| 4477 | * keeping containees alive. Exempt from GC, because we don't |
| 4478 | * want anything looking at two_tuple() by magic. |
| 4479 | */ |
| 4480 | PyObject_GC_UnTrack(two_tuple); |
| 4481 | |
| 4482 | return 0; |
| 4483 | |
| 4484 | error: |
| 4485 | Py_DECREF(copyreg); |
| 4486 | return -1; |
| 4487 | } |
| 4488 | |
| 4489 | static struct PyModuleDef _picklemodule = { |
| 4490 | PyModuleDef_HEAD_INIT, |
| 4491 | "_pickle", |
| 4492 | pickle_module_doc, |
| 4493 | -1, |
| 4494 | NULL, |
| 4495 | NULL, |
| 4496 | NULL, |
| 4497 | NULL, |
| 4498 | NULL |
| 4499 | }; |
| 4500 | |
| 4501 | PyMODINIT_FUNC |
| 4502 | PyInit__pickle(void) |
| 4503 | { |
| 4504 | PyObject *m; |
| 4505 | |
| 4506 | if (PyType_Ready(&Unpickler_Type) < 0) |
| 4507 | return NULL; |
| 4508 | if (PyType_Ready(&Pickler_Type) < 0) |
| 4509 | return NULL; |
| 4510 | if (PyType_Ready(&Pdata_Type) < 0) |
| 4511 | return NULL; |
| 4512 | |
| 4513 | /* Create the module and add the functions. */ |
| 4514 | m = PyModule_Create(&_picklemodule); |
| 4515 | if (m == NULL) |
| 4516 | return NULL; |
| 4517 | |
| 4518 | if (PyModule_AddObject(m, "Pickler", (PyObject *)&Pickler_Type) < 0) |
| 4519 | return NULL; |
| 4520 | if (PyModule_AddObject(m, "Unpickler", (PyObject *)&Unpickler_Type) < 0) |
| 4521 | return NULL; |
| 4522 | |
| 4523 | /* Initialize the exceptions. */ |
| 4524 | PickleError = PyErr_NewException("_pickle.PickleError", NULL, NULL); |
| 4525 | if (PickleError == NULL) |
| 4526 | return NULL; |
| 4527 | PicklingError = \ |
| 4528 | PyErr_NewException("_pickle.PicklingError", PickleError, NULL); |
| 4529 | if (PicklingError == NULL) |
| 4530 | return NULL; |
| 4531 | UnpicklingError = \ |
| 4532 | PyErr_NewException("_pickle.UnpicklingError", PickleError, NULL); |
| 4533 | if (UnpicklingError == NULL) |
| 4534 | return NULL; |
| 4535 | |
| 4536 | if (PyModule_AddObject(m, "PickleError", PickleError) < 0) |
| 4537 | return NULL; |
| 4538 | if (PyModule_AddObject(m, "PicklingError", PicklingError) < 0) |
| 4539 | return NULL; |
| 4540 | if (PyModule_AddObject(m, "UnpicklingError", UnpicklingError) < 0) |
| 4541 | return NULL; |
| 4542 | |
| 4543 | if (init_stuff() < 0) |
| 4544 | return NULL; |
| 4545 | |
| 4546 | return m; |
| 4547 | } |