1 /* struct module -- pack values into and (out of) strings */ 2 3 /* New version supporting byte order, alignment and size options, 4 character strings, and unsigned numbers */ 5 6 #define PY_SSIZE_T_CLEAN 7 8 #include "Python.h" 9 #include "structseq.h" 10 #include "structmember.h" 11 #include <ctype.h> 12 13 static PyTypeObject PyStructType; 14 15 /* compatibility macros */ 16 #if (PY_VERSION_HEX < 0x02050000) 17 typedef int Py_ssize_t; 18 #endif 19 20 /* warning messages */ 21 #define FLOAT_COERCE_WARN "integer argument expected, got float" 22 #define NON_INTEGER_WARN "integer argument expected, got non-integer " \ 23 "(implicit conversion using __int__ is deprecated)" 24 25 26 /* The translation function for each format character is table driven */ 27 typedef struct _formatdef { 28 char format; 29 Py_ssize_t size; 30 Py_ssize_t alignment; 31 PyObject* (*unpack)(const char *, 32 const struct _formatdef *); 33 int (*pack)(char *, PyObject *, 34 const struct _formatdef *); 35 } formatdef; 36 37 typedef struct _formatcode { 38 const struct _formatdef *fmtdef; 39 Py_ssize_t offset; 40 Py_ssize_t size; 41 } formatcode; 42 43 /* Struct object interface */ 44 45 typedef struct { 46 PyObject_HEAD 47 Py_ssize_t s_size; 48 Py_ssize_t s_len; 49 formatcode *s_codes; 50 PyObject *s_format; 51 PyObject *weakreflist; /* List of weak references */ 52 } PyStructObject; 53 54 55 #define PyStruct_Check(op) PyObject_TypeCheck(op, &PyStructType) 56 #define PyStruct_CheckExact(op) (Py_TYPE(op) == &PyStructType) 57 58 59 /* Exception */ 60 61 static PyObject *StructError; 62 63 64 /* Define various structs to figure out the alignments of types */ 65 66 67 typedef struct { char c; short x; } st_short; 68 typedef struct { char c; int x; } st_int; 69 typedef struct { char c; long x; } st_long; 70 typedef struct { char c; float x; } st_float; 71 typedef struct { char c; double x; } st_double; 72 typedef struct { char c; void *x; } st_void_p; 73 74 #define SHORT_ALIGN (sizeof(st_short) - sizeof(short)) 75 #define INT_ALIGN (sizeof(st_int) - sizeof(int)) 76 #define LONG_ALIGN (sizeof(st_long) - sizeof(long)) 77 #define FLOAT_ALIGN (sizeof(st_float) - sizeof(float)) 78 #define DOUBLE_ALIGN (sizeof(st_double) - sizeof(double)) 79 #define VOID_P_ALIGN (sizeof(st_void_p) - sizeof(void *)) 80 81 /* We can't support q and Q in native mode unless the compiler does; 82 in std mode, they're 8 bytes on all platforms. */ 83 #ifdef HAVE_LONG_LONG 84 typedef struct { char c; PY_LONG_LONG x; } s_long_long; 85 #define LONG_LONG_ALIGN (sizeof(s_long_long) - sizeof(PY_LONG_LONG)) 86 #endif 87 88 #ifdef HAVE_C99_BOOL 89 #define BOOL_TYPE _Bool 90 typedef struct { char c; _Bool x; } s_bool; 91 #define BOOL_ALIGN (sizeof(s_bool) - sizeof(BOOL_TYPE)) 92 #else 93 #define BOOL_TYPE char 94 #define BOOL_ALIGN 0 95 #endif 96 97 #define STRINGIFY(x) #x 98 99 #ifdef __powerc 100 #pragma options align=reset 101 #endif 102 103 static char *integer_codes = "bBhHiIlLqQ"; 104 105 /* Helper to get a PyLongObject by hook or by crook. Caller should decref. */ 106 107 static PyObject * 108 get_pylong(PyObject *v) 109 { 110 PyObject *r, *w; 111 int converted = 0; 112 assert(v != NULL); 113 if (!PyInt_Check(v) && !PyLong_Check(v)) { 114 PyNumberMethods *m; 115 /* Not an integer; first try to use __index__ to 116 convert to an integer. If the __index__ method 117 doesn't exist, or raises a TypeError, try __int__. 118 Use of the latter is deprecated, and will fail in 119 Python 3.x. */ 120 121 m = Py_TYPE(v)->tp_as_number; 122 if (PyIndex_Check(v)) { 123 w = PyNumber_Index(v); 124 if (w != NULL) { 125 v = w; 126 /* successfully converted to an integer */ 127 converted = 1; 128 } 129 else if (PyErr_ExceptionMatches(PyExc_TypeError)) { 130 PyErr_Clear(); 131 } 132 else 133 return NULL; 134 } 135 if (!converted && m != NULL && m->nb_int != NULL) { 136 /* Special case warning message for floats, for 137 backwards compatibility. */ 138 if (PyFloat_Check(v)) { 139 if (PyErr_WarnEx( 140 PyExc_DeprecationWarning, 141 FLOAT_COERCE_WARN, 1)) 142 return NULL; 143 } 144 else { 145 if (PyErr_WarnEx( 146 PyExc_DeprecationWarning, 147 NON_INTEGER_WARN, 1)) 148 return NULL; 149 } 150 v = m->nb_int(v); 151 if (v == NULL) 152 return NULL; 153 if (!PyInt_Check(v) && !PyLong_Check(v)) { 154 PyErr_SetString(PyExc_TypeError, 155 "__int__ method returned " 156 "non-integer"); 157 return NULL; 158 } 159 converted = 1; 160 } 161 if (!converted) { 162 PyErr_SetString(StructError, 163 "cannot convert argument " 164 "to integer"); 165 return NULL; 166 } 167 } 168 else 169 /* Ensure we own a reference to v. */ 170 Py_INCREF(v); 171 172 assert(PyInt_Check(v) || PyLong_Check(v)); 173 if (PyInt_Check(v)) { 174 r = PyLong_FromLong(PyInt_AS_LONG(v)); 175 Py_DECREF(v); 176 } 177 else if (PyLong_Check(v)) { 178 assert(PyLong_Check(v)); 179 r = v; 180 } 181 else { 182 r = NULL; /* silence compiler warning about 183 possibly uninitialized variable */ 184 assert(0); /* shouldn't ever get here */ 185 } 186 187 return r; 188 } 189 190 /* Helper to convert a Python object to a C long. Sets an exception 191 (struct.error for an inconvertible type, OverflowError for 192 out-of-range values) and returns -1 on error. */ 193 194 static int 195 get_long(PyObject *v, long *p) 196 { 197 long x; 198 199 v = get_pylong(v); 200 if (v == NULL) 201 return -1; 202 assert(PyLong_Check(v)); 203 x = PyLong_AsLong(v); 204 Py_DECREF(v); 205 if (x == (long)-1 && PyErr_Occurred()) 206 return -1; 207 *p = x; 208 return 0; 209 } 210 211 /* Same, but handling unsigned long */ 212 213 static int 214 get_ulong(PyObject *v, unsigned long *p) 215 { 216 unsigned long x; 217 218 v = get_pylong(v); 219 if (v == NULL) 220 return -1; 221 assert(PyLong_Check(v)); 222 x = PyLong_AsUnsignedLong(v); 223 Py_DECREF(v); 224 if (x == (unsigned long)-1 && PyErr_Occurred()) 225 return -1; 226 *p = x; 227 return 0; 228 } 229 230 #ifdef HAVE_LONG_LONG 231 232 /* Same, but handling native long long. */ 233 234 static int 235 get_longlong(PyObject *v, PY_LONG_LONG *p) 236 { 237 PY_LONG_LONG x; 238 239 v = get_pylong(v); 240 if (v == NULL) 241 return -1; 242 assert(PyLong_Check(v)); 243 x = PyLong_AsLongLong(v); 244 Py_DECREF(v); 245 if (x == (PY_LONG_LONG)-1 && PyErr_Occurred()) 246 return -1; 247 *p = x; 248 return 0; 249 } 250 251 /* Same, but handling native unsigned long long. */ 252 253 static int 254 get_ulonglong(PyObject *v, unsigned PY_LONG_LONG *p) 255 { 256 unsigned PY_LONG_LONG x; 257 258 v = get_pylong(v); 259 if (v == NULL) 260 return -1; 261 assert(PyLong_Check(v)); 262 x = PyLong_AsUnsignedLongLong(v); 263 Py_DECREF(v); 264 if (x == (unsigned PY_LONG_LONG)-1 && PyErr_Occurred()) 265 return -1; 266 *p = x; 267 return 0; 268 } 269 270 #endif 271 272 /* Floating point helpers */ 273 274 static PyObject * 275 unpack_float(const char *p, /* start of 4-byte string */ 276 int le) /* true for little-endian, false for big-endian */ 277 { 278 double x; 279 280 x = _PyFloat_Unpack4((unsigned char *)p, le); 281 if (x == -1.0 && PyErr_Occurred()) 282 return NULL; 283 return PyFloat_FromDouble(x); 284 } 285 286 static PyObject * 287 unpack_double(const char *p, /* start of 8-byte string */ 288 int le) /* true for little-endian, false for big-endian */ 289 { 290 double x; 291 292 x = _PyFloat_Unpack8((unsigned char *)p, le); 293 if (x == -1.0 && PyErr_Occurred()) 294 return NULL; 295 return PyFloat_FromDouble(x); 296 } 297 298 /* Helper to format the range error exceptions */ 299 static int 300 _range_error(const formatdef *f, int is_unsigned) 301 { 302 /* ulargest is the largest unsigned value with f->size bytes. 303 * Note that the simpler: 304 * ((size_t)1 << (f->size * 8)) - 1 305 * doesn't work when f->size == sizeof(size_t) because C doesn't 306 * define what happens when a left shift count is >= the number of 307 * bits in the integer being shifted; e.g., on some boxes it doesn't 308 * shift at all when they're equal. 309 */ 310 const size_t ulargest = (size_t)-1 >> ((SIZEOF_SIZE_T - f->size)*8); 311 assert(f->size >= 1 && f->size <= SIZEOF_SIZE_T); 312 if (is_unsigned) 313 PyErr_Format(StructError, 314 "'%c' format requires 0 <= number <= %zu", 315 f->format, 316 ulargest); 317 else { 318 const Py_ssize_t largest = (Py_ssize_t)(ulargest >> 1); 319 PyErr_Format(StructError, 320 "'%c' format requires %zd <= number <= %zd", 321 f->format, 322 ~ largest, 323 largest); 324 } 325 return -1; 326 } 327 328 329 330 /* A large number of small routines follow, with names of the form 331 332 [bln][up]_TYPE 333 334 [bln] distiguishes among big-endian, little-endian and native. 335 [pu] distiguishes between pack (to struct) and unpack (from struct). 336 TYPE is one of char, byte, ubyte, etc. 337 */ 338 339 /* Native mode routines. ****************************************************/ 340 /* NOTE: 341 In all n[up]_<type> routines handling types larger than 1 byte, there is 342 *no* guarantee that the p pointer is properly aligned for each type, 343 therefore memcpy is called. An intermediate variable is used to 344 compensate for big-endian architectures. 345 Normally both the intermediate variable and the memcpy call will be 346 skipped by C optimisation in little-endian architectures (gcc >= 2.91 347 does this). */ 348 349 static PyObject * 350 nu_char(const char *p, const formatdef *f) 351 { 352 return PyString_FromStringAndSize(p, 1); 353 } 354 355 static PyObject * 356 nu_byte(const char *p, const formatdef *f) 357 { 358 return PyInt_FromLong((long) *(signed char *)p); 359 } 360 361 static PyObject * 362 nu_ubyte(const char *p, const formatdef *f) 363 { 364 return PyInt_FromLong((long) *(unsigned char *)p); 365 } 366 367 static PyObject * 368 nu_short(const char *p, const formatdef *f) 369 { 370 short x; 371 memcpy((char *)&x, p, sizeof x); 372 return PyInt_FromLong((long)x); 373 } 374 375 static PyObject * 376 nu_ushort(const char *p, const formatdef *f) 377 { 378 unsigned short x; 379 memcpy((char *)&x, p, sizeof x); 380 return PyInt_FromLong((long)x); 381 } 382 383 static PyObject * 384 nu_int(const char *p, const formatdef *f) 385 { 386 int x; 387 memcpy((char *)&x, p, sizeof x); 388 return PyInt_FromLong((long)x); 389 } 390 391 static PyObject * 392 nu_uint(const char *p, const formatdef *f) 393 { 394 unsigned int x; 395 memcpy((char *)&x, p, sizeof x); 396 #if (SIZEOF_LONG > SIZEOF_INT) 397 return PyInt_FromLong((long)x); 398 #else 399 if (x <= ((unsigned int)LONG_MAX)) 400 return PyInt_FromLong((long)x); 401 return PyLong_FromUnsignedLong((unsigned long)x); 402 #endif 403 } 404 405 static PyObject * 406 nu_long(const char *p, const formatdef *f) 407 { 408 long x; 409 memcpy((char *)&x, p, sizeof x); 410 return PyInt_FromLong(x); 411 } 412 413 static PyObject * 414 nu_ulong(const char *p, const formatdef *f) 415 { 416 unsigned long x; 417 memcpy((char *)&x, p, sizeof x); 418 if (x <= LONG_MAX) 419 return PyInt_FromLong((long)x); 420 return PyLong_FromUnsignedLong(x); 421 } 422 423 /* Native mode doesn't support q or Q unless the platform C supports 424 long long (or, on Windows, __int64). */ 425 426 #ifdef HAVE_LONG_LONG 427 428 static PyObject * 429 nu_longlong(const char *p, const formatdef *f) 430 { 431 PY_LONG_LONG x; 432 memcpy((char *)&x, p, sizeof x); 433 if (x >= LONG_MIN && x <= LONG_MAX) 434 return PyInt_FromLong(Py_SAFE_DOWNCAST(x, PY_LONG_LONG, long)); 435 return PyLong_FromLongLong(x); 436 } 437 438 static PyObject * 439 nu_ulonglong(const char *p, const formatdef *f) 440 { 441 unsigned PY_LONG_LONG x; 442 memcpy((char *)&x, p, sizeof x); 443 if (x <= LONG_MAX) 444 return PyInt_FromLong(Py_SAFE_DOWNCAST(x, unsigned PY_LONG_LONG, long)); 445 return PyLong_FromUnsignedLongLong(x); 446 } 447 448 #endif 449 450 static PyObject * 451 nu_bool(const char *p, const formatdef *f) 452 { 453 BOOL_TYPE x; 454 memcpy((char *)&x, p, sizeof x); 455 return PyBool_FromLong(x != 0); 456 } 457 458 459 static PyObject * 460 nu_float(const char *p, const formatdef *f) 461 { 462 float x; 463 memcpy((char *)&x, p, sizeof x); 464 return PyFloat_FromDouble((double)x); 465 } 466 467 static PyObject * 468 nu_double(const char *p, const formatdef *f) 469 { 470 double x; 471 memcpy((char *)&x, p, sizeof x); 472 return PyFloat_FromDouble(x); 473 } 474 475 static PyObject * 476 nu_void_p(const char *p, const formatdef *f) 477 { 478 void *x; 479 memcpy((char *)&x, p, sizeof x); 480 return PyLong_FromVoidPtr(x); 481 } 482 483 static int 484 np_byte(char *p, PyObject *v, const formatdef *f) 485 { 486 long x; 487 if (get_long(v, &x) < 0) 488 return -1; 489 if (x < -128 || x > 127){ 490 PyErr_SetString(StructError, 491 "byte format requires -128 <= number <= 127"); 492 return -1; 493 } 494 *p = (char)x; 495 return 0; 496 } 497 498 static int 499 np_ubyte(char *p, PyObject *v, const formatdef *f) 500 { 501 long x; 502 if (get_long(v, &x) < 0) 503 return -1; 504 if (x < 0 || x > 255){ 505 PyErr_SetString(StructError, 506 "ubyte format requires 0 <= number <= 255"); 507 return -1; 508 } 509 *p = (char)x; 510 return 0; 511 } 512 513 static int 514 np_char(char *p, PyObject *v, const formatdef *f) 515 { 516 if (!PyString_Check(v) || PyString_Size(v) != 1) { 517 PyErr_SetString(StructError, 518 "char format require string of length 1"); 519 return -1; 520 } 521 *p = *PyString_AsString(v); 522 return 0; 523 } 524 525 static int 526 np_short(char *p, PyObject *v, const formatdef *f) 527 { 528 long x; 529 short y; 530 if (get_long(v, &x) < 0) 531 return -1; 532 if (x < SHRT_MIN || x > SHRT_MAX){ 533 PyErr_SetString(StructError, 534 "short format requires " STRINGIFY(SHRT_MIN) 535 " <= number <= " STRINGIFY(SHRT_MAX)); 536 return -1; 537 } 538 y = (short)x; 539 memcpy(p, (char *)&y, sizeof y); 540 return 0; 541 } 542 543 static int 544 np_ushort(char *p, PyObject *v, const formatdef *f) 545 { 546 long x; 547 unsigned short y; 548 if (get_long(v, &x) < 0) 549 return -1; 550 if (x < 0 || x > USHRT_MAX){ 551 PyErr_SetString(StructError, 552 "ushort format requires 0 <= number <= " STRINGIFY(USHRT_MAX)); 553 return -1; 554 } 555 y = (unsigned short)x; 556 memcpy(p, (char *)&y, sizeof y); 557 return 0; 558 } 559 560 static int 561 np_int(char *p, PyObject *v, const formatdef *f) 562 { 563 long x; 564 int y; 565 if (get_long(v, &x) < 0) 566 return -1; 567 #if (SIZEOF_LONG > SIZEOF_INT) 568 if ((x < ((long)INT_MIN)) || (x > ((long)INT_MAX))) 569 return _range_error(f, 0); 570 #endif 571 y = (int)x; 572 memcpy(p, (char *)&y, sizeof y); 573 return 0; 574 } 575 576 static int 577 np_uint(char *p, PyObject *v, const formatdef *f) 578 { 579 unsigned long x; 580 unsigned int y; 581 if (get_ulong(v, &x) < 0) 582 return -1; 583 y = (unsigned int)x; 584 #if (SIZEOF_LONG > SIZEOF_INT) 585 if (x > ((unsigned long)UINT_MAX)) 586 return _range_error(f, 1); 587 #endif 588 memcpy(p, (char *)&y, sizeof y); 589 return 0; 590 } 591 592 static int 593 np_long(char *p, PyObject *v, const formatdef *f) 594 { 595 long x; 596 if (get_long(v, &x) < 0) 597 return -1; 598 memcpy(p, (char *)&x, sizeof x); 599 return 0; 600 } 601 602 static int 603 np_ulong(char *p, PyObject *v, const formatdef *f) 604 { 605 unsigned long x; 606 if (get_ulong(v, &x) < 0) 607 return -1; 608 memcpy(p, (char *)&x, sizeof x); 609 return 0; 610 } 611 612 #ifdef HAVE_LONG_LONG 613 614 static int 615 np_longlong(char *p, PyObject *v, const formatdef *f) 616 { 617 PY_LONG_LONG x; 618 if (get_longlong(v, &x) < 0) 619 return -1; 620 memcpy(p, (char *)&x, sizeof x); 621 return 0; 622 } 623 624 static int 625 np_ulonglong(char *p, PyObject *v, const formatdef *f) 626 { 627 unsigned PY_LONG_LONG x; 628 if (get_ulonglong(v, &x) < 0) 629 return -1; 630 memcpy(p, (char *)&x, sizeof x); 631 return 0; 632 } 633 #endif 634 635 636 static int 637 np_bool(char *p, PyObject *v, const formatdef *f) 638 { 639 int y; 640 BOOL_TYPE x; 641 y = PyObject_IsTrue(v); 642 if (y < 0) 643 return -1; 644 x = y; 645 memcpy(p, (char *)&x, sizeof x); 646 return 0; 647 } 648 649 static int 650 np_float(char *p, PyObject *v, const formatdef *f) 651 { 652 float x = (float)PyFloat_AsDouble(v); 653 if (x == -1 && PyErr_Occurred()) { 654 PyErr_SetString(StructError, 655 "required argument is not a float"); 656 return -1; 657 } 658 memcpy(p, (char *)&x, sizeof x); 659 return 0; 660 } 661 662 static int 663 np_double(char *p, PyObject *v, const formatdef *f) 664 { 665 double x = PyFloat_AsDouble(v); 666 if (x == -1 && PyErr_Occurred()) { 667 PyErr_SetString(StructError, 668 "required argument is not a float"); 669 return -1; 670 } 671 memcpy(p, (char *)&x, sizeof(double)); 672 return 0; 673 } 674 675 static int 676 np_void_p(char *p, PyObject *v, const formatdef *f) 677 { 678 void *x; 679 680 v = get_pylong(v); 681 if (v == NULL) 682 return -1; 683 assert(PyLong_Check(v)); 684 x = PyLong_AsVoidPtr(v); 685 Py_DECREF(v); 686 if (x == NULL && PyErr_Occurred()) 687 return -1; 688 memcpy(p, (char *)&x, sizeof x); 689 return 0; 690 } 691 692 static formatdef native_table[] = { 693 {'x', sizeof(char), 0, NULL}, 694 {'b', sizeof(char), 0, nu_byte, np_byte}, 695 {'B', sizeof(char), 0, nu_ubyte, np_ubyte}, 696 {'c', sizeof(char), 0, nu_char, np_char}, 697 {'s', sizeof(char), 0, NULL}, 698 {'p', sizeof(char), 0, NULL}, 699 {'h', sizeof(short), SHORT_ALIGN, nu_short, np_short}, 700 {'H', sizeof(short), SHORT_ALIGN, nu_ushort, np_ushort}, 701 {'i', sizeof(int), INT_ALIGN, nu_int, np_int}, 702 {'I', sizeof(int), INT_ALIGN, nu_uint, np_uint}, 703 {'l', sizeof(long), LONG_ALIGN, nu_long, np_long}, 704 {'L', sizeof(long), LONG_ALIGN, nu_ulong, np_ulong}, 705 #ifdef HAVE_LONG_LONG 706 {'q', sizeof(PY_LONG_LONG), LONG_LONG_ALIGN, nu_longlong, np_longlong}, 707 {'Q', sizeof(PY_LONG_LONG), LONG_LONG_ALIGN, nu_ulonglong,np_ulonglong}, 708 #endif 709 {'?', sizeof(BOOL_TYPE), BOOL_ALIGN, nu_bool, np_bool}, 710 {'f', sizeof(float), FLOAT_ALIGN, nu_float, np_float}, 711 {'d', sizeof(double), DOUBLE_ALIGN, nu_double, np_double}, 712 {'P', sizeof(void *), VOID_P_ALIGN, nu_void_p, np_void_p}, 713 {0} 714 }; 715 716 /* Big-endian routines. *****************************************************/ 717 718 static PyObject * 719 bu_int(const char *p, const formatdef *f) 720 { 721 long x = 0; 722 Py_ssize_t i = f->size; 723 const unsigned char *bytes = (const unsigned char *)p; 724 do { 725 x = (x<<8) | *bytes++; 726 } while (--i > 0); 727 /* Extend the sign bit. */ 728 if (SIZEOF_LONG > f->size) 729 x |= -(x & (1L << ((8 * f->size) - 1))); 730 return PyInt_FromLong(x); 731 } 732 733 static PyObject * 734 bu_uint(const char *p, const formatdef *f) 735 { 736 unsigned long x = 0; 737 Py_ssize_t i = f->size; 738 const unsigned char *bytes = (const unsigned char *)p; 739 do { 740 x = (x<<8) | *bytes++; 741 } while (--i > 0); 742 if (x <= LONG_MAX) 743 return PyInt_FromLong((long)x); 744 return PyLong_FromUnsignedLong(x); 745 } 746 747 static PyObject * 748 bu_longlong(const char *p, const formatdef *f) 749 { 750 #ifdef HAVE_LONG_LONG 751 PY_LONG_LONG x = 0; 752 Py_ssize_t i = f->size; 753 const unsigned char *bytes = (const unsigned char *)p; 754 do { 755 x = (x<<8) | *bytes++; 756 } while (--i > 0); 757 /* Extend the sign bit. */ 758 if (SIZEOF_LONG_LONG > f->size) 759 x |= -(x & ((PY_LONG_LONG)1 << ((8 * f->size) - 1))); 760 if (x >= LONG_MIN && x <= LONG_MAX) 761 return PyInt_FromLong(Py_SAFE_DOWNCAST(x, PY_LONG_LONG, long)); 762 return PyLong_FromLongLong(x); 763 #else 764 return _PyLong_FromByteArray((const unsigned char *)p, 765 8, 766 0, /* little-endian */ 767 1 /* signed */); 768 #endif 769 } 770 771 static PyObject * 772 bu_ulonglong(const char *p, const formatdef *f) 773 { 774 #ifdef HAVE_LONG_LONG 775 unsigned PY_LONG_LONG x = 0; 776 Py_ssize_t i = f->size; 777 const unsigned char *bytes = (const unsigned char *)p; 778 do { 779 x = (x<<8) | *bytes++; 780 } while (--i > 0); 781 if (x <= LONG_MAX) 782 return PyInt_FromLong(Py_SAFE_DOWNCAST(x, unsigned PY_LONG_LONG, long)); 783 return PyLong_FromUnsignedLongLong(x); 784 #else 785 return _PyLong_FromByteArray((const unsigned char *)p, 786 8, 787 0, /* little-endian */ 788 0 /* signed */); 789 #endif 790 } 791 792 static PyObject * 793 bu_float(const char *p, const formatdef *f) 794 { 795 return unpack_float(p, 0); 796 } 797 798 static PyObject * 799 bu_double(const char *p, const formatdef *f) 800 { 801 return unpack_double(p, 0); 802 } 803 804 static PyObject * 805 bu_bool(const char *p, const formatdef *f) 806 { 807 char x; 808 memcpy((char *)&x, p, sizeof x); 809 return PyBool_FromLong(x != 0); 810 } 811 812 static int 813 bp_int(char *p, PyObject *v, const formatdef *f) 814 { 815 long x; 816 Py_ssize_t i; 817 if (get_long(v, &x) < 0) 818 return -1; 819 i = f->size; 820 if (i != SIZEOF_LONG) { 821 if ((i == 2) && (x < -32768 || x > 32767)) 822 return _range_error(f, 0); 823 #if (SIZEOF_LONG != 4) 824 else if ((i == 4) && (x < -2147483648L || x > 2147483647L)) 825 return _range_error(f, 0); 826 #endif 827 } 828 do { 829 p[--i] = (char)x; 830 x >>= 8; 831 } while (i > 0); 832 return 0; 833 } 834 835 static int 836 bp_uint(char *p, PyObject *v, const formatdef *f) 837 { 838 unsigned long x; 839 Py_ssize_t i; 840 if (get_ulong(v, &x) < 0) 841 return -1; 842 i = f->size; 843 if (i != SIZEOF_LONG) { 844 unsigned long maxint = 1; 845 maxint <<= (unsigned long)(i * 8); 846 if (x >= maxint) 847 return _range_error(f, 1); 848 } 849 do { 850 p[--i] = (char)x; 851 x >>= 8; 852 } while (i > 0); 853 return 0; 854 } 855 856 static int 857 bp_longlong(char *p, PyObject *v, const formatdef *f) 858 { 859 int res; 860 v = get_pylong(v); 861 if (v == NULL) 862 return -1; 863 res = _PyLong_AsByteArray((PyLongObject *)v, 864 (unsigned char *)p, 865 8, 866 0, /* little_endian */ 867 1 /* signed */); 868 Py_DECREF(v); 869 return res; 870 } 871 872 static int 873 bp_ulonglong(char *p, PyObject *v, const formatdef *f) 874 { 875 int res; 876 v = get_pylong(v); 877 if (v == NULL) 878 return -1; 879 res = _PyLong_AsByteArray((PyLongObject *)v, 880 (unsigned char *)p, 881 8, 882 0, /* little_endian */ 883 0 /* signed */); 884 Py_DECREF(v); 885 return res; 886 } 887 888 static int 889 bp_float(char *p, PyObject *v, const formatdef *f) 890 { 891 double x = PyFloat_AsDouble(v); 892 if (x == -1 && PyErr_Occurred()) { 893 PyErr_SetString(StructError, 894 "required argument is not a float"); 895 return -1; 896 } 897 return _PyFloat_Pack4(x, (unsigned char *)p, 0); 898 } 899 900 static int 901 bp_double(char *p, PyObject *v, const formatdef *f) 902 { 903 double x = PyFloat_AsDouble(v); 904 if (x == -1 && PyErr_Occurred()) { 905 PyErr_SetString(StructError, 906 "required argument is not a float"); 907 return -1; 908 } 909 return _PyFloat_Pack8(x, (unsigned char *)p, 0); 910 } 911 912 static int 913 bp_bool(char *p, PyObject *v, const formatdef *f) 914 { 915 int y; 916 y = PyObject_IsTrue(v); 917 if (y < 0) 918 return -1; 919 *p = (char)y; 920 return 0; 921 } 922 923 static formatdef bigendian_table[] = { 924 {'x', 1, 0, NULL}, 925 {'b', 1, 0, nu_byte, np_byte}, 926 {'B', 1, 0, nu_ubyte, np_ubyte}, 927 {'c', 1, 0, nu_char, np_char}, 928 {'s', 1, 0, NULL}, 929 {'p', 1, 0, NULL}, 930 {'h', 2, 0, bu_int, bp_int}, 931 {'H', 2, 0, bu_uint, bp_uint}, 932 {'i', 4, 0, bu_int, bp_int}, 933 {'I', 4, 0, bu_uint, bp_uint}, 934 {'l', 4, 0, bu_int, bp_int}, 935 {'L', 4, 0, bu_uint, bp_uint}, 936 {'q', 8, 0, bu_longlong, bp_longlong}, 937 {'Q', 8, 0, bu_ulonglong, bp_ulonglong}, 938 {'?', 1, 0, bu_bool, bp_bool}, 939 {'f', 4, 0, bu_float, bp_float}, 940 {'d', 8, 0, bu_double, bp_double}, 941 {0} 942 }; 943 944 /* Little-endian routines. *****************************************************/ 945 946 static PyObject * 947 lu_int(const char *p, const formatdef *f) 948 { 949 long x = 0; 950 Py_ssize_t i = f->size; 951 const unsigned char *bytes = (const unsigned char *)p; 952 do { 953 x = (x<<8) | bytes[--i]; 954 } while (i > 0); 955 /* Extend the sign bit. */ 956 if (SIZEOF_LONG > f->size) 957 x |= -(x & (1L << ((8 * f->size) - 1))); 958 return PyInt_FromLong(x); 959 } 960 961 static PyObject * 962 lu_uint(const char *p, const formatdef *f) 963 { 964 unsigned long x = 0; 965 Py_ssize_t i = f->size; 966 const unsigned char *bytes = (const unsigned char *)p; 967 do { 968 x = (x<<8) | bytes[--i]; 969 } while (i > 0); 970 if (x <= LONG_MAX) 971 return PyInt_FromLong((long)x); 972 return PyLong_FromUnsignedLong((long)x); 973 } 974 975 static PyObject * 976 lu_longlong(const char *p, const formatdef *f) 977 { 978 #ifdef HAVE_LONG_LONG 979 PY_LONG_LONG x = 0; 980 Py_ssize_t i = f->size; 981 const unsigned char *bytes = (const unsigned char *)p; 982 do { 983 x = (x<<8) | bytes[--i]; 984 } while (i > 0); 985 /* Extend the sign bit. */ 986 if (SIZEOF_LONG_LONG > f->size) 987 x |= -(x & ((PY_LONG_LONG)1 << ((8 * f->size) - 1))); 988 if (x >= LONG_MIN && x <= LONG_MAX) 989 return PyInt_FromLong(Py_SAFE_DOWNCAST(x, PY_LONG_LONG, long)); 990 return PyLong_FromLongLong(x); 991 #else 992 return _PyLong_FromByteArray((const unsigned char *)p, 993 8, 994 1, /* little-endian */ 995 1 /* signed */); 996 #endif 997 } 998 999 static PyObject * 1000 lu_ulonglong(const char *p, const formatdef *f) 1001 { 1002 #ifdef HAVE_LONG_LONG 1003 unsigned PY_LONG_LONG x = 0; 1004 Py_ssize_t i = f->size; 1005 const unsigned char *bytes = (const unsigned char *)p; 1006 do { 1007 x = (x<<8) | bytes[--i]; 1008 } while (i > 0); 1009 if (x <= LONG_MAX) 1010 return PyInt_FromLong(Py_SAFE_DOWNCAST(x, unsigned PY_LONG_LONG, long)); 1011 return PyLong_FromUnsignedLongLong(x); 1012 #else 1013 return _PyLong_FromByteArray((const unsigned char *)p, 1014 8, 1015 1, /* little-endian */ 1016 0 /* signed */); 1017 #endif 1018 } 1019 1020 static PyObject * 1021 lu_float(const char *p, const formatdef *f) 1022 { 1023 return unpack_float(p, 1); 1024 } 1025 1026 static PyObject * 1027 lu_double(const char *p, const formatdef *f) 1028 { 1029 return unpack_double(p, 1); 1030 } 1031 1032 static int 1033 lp_int(char *p, PyObject *v, const formatdef *f) 1034 { 1035 long x; 1036 Py_ssize_t i; 1037 if (get_long(v, &x) < 0) 1038 return -1; 1039 i = f->size; 1040 if (i != SIZEOF_LONG) { 1041 if ((i == 2) && (x < -32768 || x > 32767)) 1042 return _range_error(f, 0); 1043 #if (SIZEOF_LONG != 4) 1044 else if ((i == 4) && (x < -2147483648L || x > 2147483647L)) 1045 return _range_error(f, 0); 1046 #endif 1047 } 1048 do { 1049 *p++ = (char)x; 1050 x >>= 8; 1051 } while (--i > 0); 1052 return 0; 1053 } 1054 1055 static int 1056 lp_uint(char *p, PyObject *v, const formatdef *f) 1057 { 1058 unsigned long x; 1059 Py_ssize_t i; 1060 if (get_ulong(v, &x) < 0) 1061 return -1; 1062 i = f->size; 1063 if (i != SIZEOF_LONG) { 1064 unsigned long maxint = 1; 1065 maxint <<= (unsigned long)(i * 8); 1066 if (x >= maxint) 1067 return _range_error(f, 1); 1068 } 1069 do { 1070 *p++ = (char)x; 1071 x >>= 8; 1072 } while (--i > 0); 1073 return 0; 1074 } 1075 1076 static int 1077 lp_longlong(char *p, PyObject *v, const formatdef *f) 1078 { 1079 int res; 1080 v = get_pylong(v); 1081 if (v == NULL) 1082 return -1; 1083 res = _PyLong_AsByteArray((PyLongObject*)v, 1084 (unsigned char *)p, 1085 8, 1086 1, /* little_endian */ 1087 1 /* signed */); 1088 Py_DECREF(v); 1089 return res; 1090 } 1091 1092 static int 1093 lp_ulonglong(char *p, PyObject *v, const formatdef *f) 1094 { 1095 int res; 1096 v = get_pylong(v); 1097 if (v == NULL) 1098 return -1; 1099 res = _PyLong_AsByteArray((PyLongObject*)v, 1100 (unsigned char *)p, 1101 8, 1102 1, /* little_endian */ 1103 0 /* signed */); 1104 Py_DECREF(v); 1105 return res; 1106 } 1107 1108 static int 1109 lp_float(char *p, PyObject *v, const formatdef *f) 1110 { 1111 double x = PyFloat_AsDouble(v); 1112 if (x == -1 && PyErr_Occurred()) { 1113 PyErr_SetString(StructError, 1114 "required argument is not a float"); 1115 return -1; 1116 } 1117 return _PyFloat_Pack4(x, (unsigned char *)p, 1); 1118 } 1119 1120 static int 1121 lp_double(char *p, PyObject *v, const formatdef *f) 1122 { 1123 double x = PyFloat_AsDouble(v); 1124 if (x == -1 && PyErr_Occurred()) { 1125 PyErr_SetString(StructError, 1126 "required argument is not a float"); 1127 return -1; 1128 } 1129 return _PyFloat_Pack8(x, (unsigned char *)p, 1); 1130 } 1131 1132 static formatdef lilendian_table[] = { 1133 {'x', 1, 0, NULL}, 1134 {'b', 1, 0, nu_byte, np_byte}, 1135 {'B', 1, 0, nu_ubyte, np_ubyte}, 1136 {'c', 1, 0, nu_char, np_char}, 1137 {'s', 1, 0, NULL}, 1138 {'p', 1, 0, NULL}, 1139 {'h', 2, 0, lu_int, lp_int}, 1140 {'H', 2, 0, lu_uint, lp_uint}, 1141 {'i', 4, 0, lu_int, lp_int}, 1142 {'I', 4, 0, lu_uint, lp_uint}, 1143 {'l', 4, 0, lu_int, lp_int}, 1144 {'L', 4, 0, lu_uint, lp_uint}, 1145 {'q', 8, 0, lu_longlong, lp_longlong}, 1146 {'Q', 8, 0, lu_ulonglong, lp_ulonglong}, 1147 {'?', 1, 0, bu_bool, bp_bool}, /* Std rep not endian dep, 1148 but potentially different from native rep -- reuse bx_bool funcs. */ 1149 {'f', 4, 0, lu_float, lp_float}, 1150 {'d', 8, 0, lu_double, lp_double}, 1151 {0} 1152 }; 1153 1154 1155 static const formatdef * 1156 whichtable(char **pfmt) 1157 { 1158 const char *fmt = (*pfmt)++; /* May be backed out of later */ 1159 switch (*fmt) { 1160 case '<': 1161 return lilendian_table; 1162 case '>': 1163 case '!': /* Network byte order is big-endian */ 1164 return bigendian_table; 1165 case '=': { /* Host byte order -- different from native in alignment! */ 1166 int n = 1; 1167 char *p = (char *) &n; 1168 if (*p == 1) 1169 return lilendian_table; 1170 else 1171 return bigendian_table; 1172 } 1173 default: 1174 --*pfmt; /* Back out of pointer increment */ 1175 /* Fall through */ 1176 case '@': 1177 return native_table; 1178 } 1179 } 1180 1181 1182 /* Get the table entry for a format code */ 1183 1184 static const formatdef * 1185 getentry(int c, const formatdef *f) 1186 { 1187 for (; f->format != '\0'; f++) { 1188 if (f->format == c) { 1189 return f; 1190 } 1191 } 1192 PyErr_SetString(StructError, "bad char in struct format"); 1193 return NULL; 1194 } 1195 1196 1197 /* Align a size according to a format code. Return -1 on overflow. */ 1198 1199 static Py_ssize_t 1200 align(Py_ssize_t size, char c, const formatdef *e) 1201 { 1202 Py_ssize_t extra; 1203 1204 if (e->format == c) { 1205 if (e->alignment && size > 0) { 1206 extra = (e->alignment - 1) - (size - 1) % (e->alignment); 1207 if (extra > PY_SSIZE_T_MAX - size) 1208 return -1; 1209 size += extra; 1210 } 1211 } 1212 return size; 1213 } 1214 1215 1216 /* calculate the size of a format string */ 1217 1218 static int 1219 prepare_s(PyStructObject *self) 1220 { 1221 const formatdef *f; 1222 const formatdef *e; 1223 formatcode *codes; 1224 1225 const char *s; 1226 const char *fmt; 1227 char c; 1228 Py_ssize_t size, len, num, itemsize; 1229 1230 fmt = PyString_AS_STRING(self->s_format); 1231 1232 f = whichtable((char **)&fmt); 1233 1234 s = fmt; 1235 size = 0; 1236 len = 0; 1237 while ((c = *s++) != '\0') { 1238 if (isspace(Py_CHARMASK(c))) 1239 continue; 1240 if ('0' <= c && c <= '9') { 1241 num = c - '0'; 1242 while ('0' <= (c = *s++) && c <= '9') { 1243 /* overflow-safe version of 1244 if (num*10 + (c - '0') > PY_SSIZE_T_MAX) { ... } */ 1245 if (num >= PY_SSIZE_T_MAX / 10 && ( 1246 num > PY_SSIZE_T_MAX / 10 || 1247 (c - '0') > PY_SSIZE_T_MAX % 10)) 1248 goto overflow; 1249 num = num*10 + (c - '0'); 1250 } 1251 if (c == '\0') 1252 break; 1253 } 1254 else 1255 num = 1; 1256 1257 e = getentry(c, f); 1258 if (e == NULL) 1259 return -1; 1260 1261 switch (c) { 1262 case 's': /* fall through */ 1263 case 'p': len++; break; 1264 case 'x': break; 1265 default: len += num; break; 1266 } 1267 1268 itemsize = e->size; 1269 size = align(size, c, e); 1270 if (size == -1) 1271 goto overflow; 1272 1273 /* if (size + num * itemsize > PY_SSIZE_T_MAX) { ... } */ 1274 if (num > (PY_SSIZE_T_MAX - size) / itemsize) 1275 goto overflow; 1276 size += num * itemsize; 1277 } 1278 1279 /* check for overflow */ 1280 if ((len + 1) > (PY_SSIZE_T_MAX / sizeof(formatcode))) { 1281 PyErr_NoMemory(); 1282 return -1; 1283 } 1284 1285 self->s_size = size; 1286 self->s_len = len; 1287 codes = PyMem_MALLOC((len + 1) * sizeof(formatcode)); 1288 if (codes == NULL) { 1289 PyErr_NoMemory(); 1290 return -1; 1291 } 1292 /* Free any s_codes value left over from a previous initialization. */ 1293 if (self->s_codes != NULL) 1294 PyMem_FREE(self->s_codes); 1295 self->s_codes = codes; 1296 1297 s = fmt; 1298 size = 0; 1299 while ((c = *s++) != '\0') { 1300 if (isspace(Py_CHARMASK(c))) 1301 continue; 1302 if ('0' <= c && c <= '9') { 1303 num = c - '0'; 1304 while ('0' <= (c = *s++) && c <= '9') 1305 num = num*10 + (c - '0'); 1306 if (c == '\0') 1307 break; 1308 } 1309 else 1310 num = 1; 1311 1312 e = getentry(c, f); 1313 1314 size = align(size, c, e); 1315 if (c == 's' || c == 'p') { 1316 codes->offset = size; 1317 codes->size = num; 1318 codes->fmtdef = e; 1319 codes++; 1320 size += num; 1321 } else if (c == 'x') { 1322 size += num; 1323 } else { 1324 while (--num >= 0) { 1325 codes->offset = size; 1326 codes->size = e->size; 1327 codes->fmtdef = e; 1328 codes++; 1329 size += e->size; 1330 } 1331 } 1332 } 1333 codes->fmtdef = NULL; 1334 codes->offset = size; 1335 codes->size = 0; 1336 1337 return 0; 1338 1339 overflow: 1340 PyErr_SetString(StructError, 1341 "total struct size too long"); 1342 return -1; 1343 } 1344 1345 static PyObject * 1346 s_new(PyTypeObject *type, PyObject *args, PyObject *kwds) 1347 { 1348 PyObject *self; 1349 1350 assert(type != NULL && type->tp_alloc != NULL); 1351 1352 self = type->tp_alloc(type, 0); 1353 if (self != NULL) { 1354 PyStructObject *s = (PyStructObject*)self; 1355 Py_INCREF(Py_None); 1356 s->s_format = Py_None; 1357 s->s_codes = NULL; 1358 s->s_size = -1; 1359 s->s_len = -1; 1360 } 1361 return self; 1362 } 1363 1364 static int 1365 s_init(PyObject *self, PyObject *args, PyObject *kwds) 1366 { 1367 PyStructObject *soself = (PyStructObject *)self; 1368 PyObject *o_format = NULL; 1369 int ret = 0; 1370 static char *kwlist[] = {"format", 0}; 1371 1372 assert(PyStruct_Check(self)); 1373 1374 if (!PyArg_ParseTupleAndKeywords(args, kwds, "O:Struct", kwlist, 1375 &o_format)) 1376 return -1; 1377 1378 if (PyString_Check(o_format)) { 1379 Py_INCREF(o_format); 1380 Py_CLEAR(soself->s_format); 1381 soself->s_format = o_format; 1382 } 1383 else if (PyUnicode_Check(o_format)) { 1384 PyObject *str = PyUnicode_AsEncodedString(o_format, "ascii", NULL); 1385 if (str == NULL) 1386 return -1; 1387 Py_CLEAR(soself->s_format); 1388 soself->s_format = str; 1389 } 1390 else { 1391 PyErr_Format(PyExc_TypeError, 1392 "Struct() argument 1 must be string, not %s", 1393 Py_TYPE(o_format)->tp_name); 1394 return -1; 1395 } 1396 1397 ret = prepare_s(soself); 1398 return ret; 1399 } 1400 1401 static void 1402 s_dealloc(PyStructObject *s) 1403 { 1404 if (s->weakreflist != NULL) 1405 PyObject_ClearWeakRefs((PyObject *)s); 1406 if (s->s_codes != NULL) { 1407 PyMem_FREE(s->s_codes); 1408 } 1409 Py_XDECREF(s->s_format); 1410 Py_TYPE(s)->tp_free((PyObject *)s); 1411 } 1412 1413 static PyObject * 1414 s_unpack_internal(PyStructObject *soself, char *startfrom) { 1415 formatcode *code; 1416 Py_ssize_t i = 0; 1417 PyObject *result = PyTuple_New(soself->s_len); 1418 if (result == NULL) 1419 return NULL; 1420 1421 for (code = soself->s_codes; code->fmtdef != NULL; code++) { 1422 PyObject *v; 1423 const formatdef *e = code->fmtdef; 1424 const char *res = startfrom + code->offset; 1425 if (e->format == 's') { 1426 v = PyString_FromStringAndSize(res, code->size); 1427 } else if (e->format == 'p') { 1428 Py_ssize_t n = *(unsigned char*)res; 1429 if (n >= code->size) 1430 n = code->size - 1; 1431 v = PyString_FromStringAndSize(res + 1, n); 1432 } else { 1433 v = e->unpack(res, e); 1434 } 1435 if (v == NULL) 1436 goto fail; 1437 PyTuple_SET_ITEM(result, i++, v); 1438 } 1439 1440 return result; 1441 fail: 1442 Py_DECREF(result); 1443 return NULL; 1444 } 1445 1446 1447 PyDoc_STRVAR(s_unpack__doc__, 1448 "S.unpack(str) -> (v1, v2, ...)\n\ 1449 \n\ 1450 Return tuple containing values unpacked according to this Struct's format.\n\ 1451 Requires len(str) == self.size. See struct.__doc__ for more on format\n\ 1452 strings."); 1453 1454 static PyObject * 1455 s_unpack(PyObject *self, PyObject *inputstr) 1456 { 1457 Py_buffer buf; 1458 char *start; 1459 Py_ssize_t len; 1460 PyObject *args=NULL, *result; 1461 PyStructObject *soself = (PyStructObject *)self; 1462 assert(PyStruct_Check(self)); 1463 assert(soself->s_codes != NULL); 1464 if (inputstr == NULL) 1465 goto fail; 1466 if (PyString_Check(inputstr) && 1467 PyString_GET_SIZE(inputstr) == soself->s_size) { 1468 return s_unpack_internal(soself, PyString_AS_STRING(inputstr)); 1469 } 1470 args = PyTuple_Pack(1, inputstr); 1471 if (args == NULL) 1472 return NULL; 1473 if (!PyArg_ParseTuple(args, "s*:unpack", &buf)) 1474 goto fail; 1475 start = buf.buf; 1476 len = buf.len; 1477 if (soself->s_size != len) { 1478 PyBuffer_Release(&buf); 1479 goto fail; 1480 } 1481 result = s_unpack_internal(soself, start); 1482 Py_DECREF(args); 1483 PyBuffer_Release(&buf); 1484 return result; 1485 1486 fail: 1487 Py_XDECREF(args); 1488 PyErr_Format(StructError, 1489 "unpack requires a string argument of length %zd", 1490 soself->s_size); 1491 return NULL; 1492 } 1493 1494 PyDoc_STRVAR(s_unpack_from__doc__, 1495 "S.unpack_from(buffer[, offset]) -> (v1, v2, ...)\n\ 1496 \n\ 1497 Return tuple containing values unpacked according to this Struct's format.\n\ 1498 Unlike unpack, unpack_from can unpack values from any object supporting\n\ 1499 the buffer API, not just str. Requires len(buffer[offset:]) >= self.size.\n\ 1500 See struct.__doc__ for more on format strings."); 1501 1502 static PyObject * 1503 s_unpack_from(PyObject *self, PyObject *args, PyObject *kwds) 1504 { 1505 static char *kwlist[] = {"buffer", "offset", 0}; 1506 static char *fmt = "z*|n:unpack_from"; 1507 Py_buffer buf; 1508 Py_ssize_t buffer_len = 0, offset = 0; 1509 char *buffer = NULL; 1510 PyStructObject *soself = (PyStructObject *)self; 1511 PyObject *result; 1512 assert(PyStruct_Check(self)); 1513 assert(soself->s_codes != NULL); 1514 1515 if (!PyArg_ParseTupleAndKeywords(args, kwds, fmt, kwlist, 1516 &buf, &offset)) 1517 return NULL; 1518 buffer = buf.buf; 1519 buffer_len = buf.len; 1520 if (buffer == NULL) { 1521 PyErr_Format(StructError, 1522 "unpack_from requires a buffer argument"); 1523 PyBuffer_Release(&buf); 1524 return NULL; 1525 } 1526 1527 if (offset < 0) 1528 offset += buffer_len; 1529 1530 if (offset < 0 || (buffer_len - offset) < soself->s_size) { 1531 PyErr_Format(StructError, 1532 "unpack_from requires a buffer of at least %zd bytes", 1533 soself->s_size); 1534 PyBuffer_Release(&buf); 1535 return NULL; 1536 } 1537 result = s_unpack_internal(soself, buffer + offset); 1538 PyBuffer_Release(&buf); 1539 return result; 1540 } 1541 1542 1543 /* 1544 * Guts of the pack function. 1545 * 1546 * Takes a struct object, a tuple of arguments, and offset in that tuple of 1547 * argument for where to start processing the arguments for packing, and a 1548 * character buffer for writing the packed string. The caller must insure 1549 * that the buffer may contain the required length for packing the arguments. 1550 * 0 is returned on success, 1 is returned if there is an error. 1551 * 1552 */ 1553 static int 1554 s_pack_internal(PyStructObject *soself, PyObject *args, int offset, char* buf) 1555 { 1556 formatcode *code; 1557 /* XXX(nnorwitz): why does i need to be a local? can we use 1558 the offset parameter or do we need the wider width? */ 1559 Py_ssize_t i; 1560 1561 memset(buf, '\0', soself->s_size); 1562 i = offset; 1563 for (code = soself->s_codes; code->fmtdef != NULL; code++) { 1564 Py_ssize_t n; 1565 PyObject *v = PyTuple_GET_ITEM(args, i++); 1566 const formatdef *e = code->fmtdef; 1567 char *res = buf + code->offset; 1568 if (e->format == 's') { 1569 if (!PyString_Check(v)) { 1570 PyErr_SetString(StructError, 1571 "argument for 's' must " 1572 "be a string"); 1573 return -1; 1574 } 1575 n = PyString_GET_SIZE(v); 1576 if (n > code->size) 1577 n = code->size; 1578 if (n > 0) 1579 memcpy(res, PyString_AS_STRING(v), n); 1580 } else if (e->format == 'p') { 1581 if (!PyString_Check(v)) { 1582 PyErr_SetString(StructError, 1583 "argument for 'p' must " 1584 "be a string"); 1585 return -1; 1586 } 1587 n = PyString_GET_SIZE(v); 1588 if (n > (code->size - 1)) 1589 n = code->size - 1; 1590 if (n > 0) 1591 memcpy(res + 1, PyString_AS_STRING(v), n); 1592 if (n > 255) 1593 n = 255; 1594 *res = Py_SAFE_DOWNCAST(n, Py_ssize_t, unsigned char); 1595 } else if (e->pack(res, v, e) < 0) { 1596 if (strchr(integer_codes, e->format) != NULL && 1597 PyErr_ExceptionMatches(PyExc_OverflowError)) 1598 PyErr_Format(StructError, 1599 "integer out of range for " 1600 "'%c' format code", 1601 e->format); 1602 return -1; 1603 } 1604 } 1605 1606 /* Success */ 1607 return 0; 1608 } 1609 1610 1611 PyDoc_STRVAR(s_pack__doc__, 1612 "S.pack(v1, v2, ...) -> string\n\ 1613 \n\ 1614 Return a string containing values v1, v2, ... packed according to this\n\ 1615 Struct's format. See struct.__doc__ for more on format strings."); 1616 1617 static PyObject * 1618 s_pack(PyObject *self, PyObject *args) 1619 { 1620 PyStructObject *soself; 1621 PyObject *result; 1622 1623 /* Validate arguments. */ 1624 soself = (PyStructObject *)self; 1625 assert(PyStruct_Check(self)); 1626 assert(soself->s_codes != NULL); 1627 if (PyTuple_GET_SIZE(args) != soself->s_len) 1628 { 1629 PyErr_Format(StructError, 1630 "pack expected %zd items for packing (got %zd)", soself->s_len, PyTuple_GET_SIZE(args)); 1631 return NULL; 1632 } 1633 1634 /* Allocate a new string */ 1635 result = PyString_FromStringAndSize((char *)NULL, soself->s_size); 1636 if (result == NULL) 1637 return NULL; 1638 1639 /* Call the guts */ 1640 if ( s_pack_internal(soself, args, 0, PyString_AS_STRING(result)) != 0 ) { 1641 Py_DECREF(result); 1642 return NULL; 1643 } 1644 1645 return result; 1646 } 1647 1648 PyDoc_STRVAR(s_pack_into__doc__, 1649 "S.pack_into(buffer, offset, v1, v2, ...)\n\ 1650 \n\ 1651 Pack the values v1, v2, ... according to this Struct's format, write \n\ 1652 the packed bytes into the writable buffer buf starting at offset. Note\n\ 1653 that the offset is not an optional argument. See struct.__doc__ for \n\ 1654 more on format strings."); 1655 1656 static PyObject * 1657 s_pack_into(PyObject *self, PyObject *args) 1658 { 1659 PyStructObject *soself; 1660 Py_buffer buf; 1661 Py_ssize_t offset; 1662 1663 /* Validate arguments. +1 is for the first arg as buffer. */ 1664 soself = (PyStructObject *)self; 1665 assert(PyStruct_Check(self)); 1666 assert(soself->s_codes != NULL); 1667 if (PyTuple_GET_SIZE(args) != (soself->s_len + 2)) 1668 { 1669 if (PyTuple_GET_SIZE(args) == 0) { 1670 PyErr_Format(StructError, 1671 "pack_into expected buffer argument"); 1672 } 1673 else if (PyTuple_GET_SIZE(args) == 1) { 1674 PyErr_Format(StructError, 1675 "pack_into expected offset argument"); 1676 } 1677 else { 1678 PyErr_Format(StructError, 1679 "pack_into expected %zd items for packing (got %zd)", 1680 soself->s_len, (PyTuple_GET_SIZE(args) - 2)); 1681 } 1682 return NULL; 1683 } 1684 1685 /* Extract a writable memory buffer from the first argument */ 1686 if (!PyArg_Parse(PyTuple_GET_ITEM(args, 0), "w*", &buf)) 1687 return NULL; 1688 1689 /* Extract the offset from the first argument */ 1690 offset = PyInt_AsSsize_t(PyTuple_GET_ITEM(args, 1)); 1691 if (offset == -1 && PyErr_Occurred()) { 1692 PyBuffer_Release(&buf); 1693 return NULL; 1694 } 1695 1696 /* Support negative offsets. */ 1697 if (offset < 0) 1698 offset += buf.len; 1699 1700 /* Check boundaries */ 1701 if (offset < 0 || (buf.len - offset) < soself->s_size) { 1702 PyErr_Format(StructError, 1703 "pack_into requires a buffer of at least %zd bytes", 1704 soself->s_size); 1705 PyBuffer_Release(&buf); 1706 return NULL; 1707 } 1708 1709 /* Call the guts */ 1710 if (s_pack_internal(soself, args, 2, (char *)buf.buf + offset) != 0) { 1711 PyBuffer_Release(&buf); 1712 return NULL; 1713 } 1714 PyBuffer_Release(&buf); 1715 1716 Py_RETURN_NONE; 1717 } 1718 1719 static PyObject * 1720 s_get_format(PyStructObject *self, void *unused) 1721 { 1722 Py_INCREF(self->s_format); 1723 return self->s_format; 1724 } 1725 1726 static PyObject * 1727 s_get_size(PyStructObject *self, void *unused) 1728 { 1729 return PyInt_FromSsize_t(self->s_size); 1730 } 1731 1732 PyDoc_STRVAR(s_sizeof__doc__, 1733 "S.__sizeof__() -> size of S in memory, in bytes"); 1734 1735 static PyObject * 1736 s_sizeof(PyStructObject *self, void *unused) 1737 { 1738 Py_ssize_t size; 1739 1740 size = sizeof(PyStructObject) + sizeof(formatcode) * (self->s_len + 1); 1741 return PyLong_FromSsize_t(size); 1742 } 1743 1744 /* List of functions */ 1745 1746 static struct PyMethodDef s_methods[] = { 1747 {"pack", s_pack, METH_VARARGS, s_pack__doc__}, 1748 {"pack_into", s_pack_into, METH_VARARGS, s_pack_into__doc__}, 1749 {"unpack", s_unpack, METH_O, s_unpack__doc__}, 1750 {"unpack_from", (PyCFunction)s_unpack_from, METH_VARARGS|METH_KEYWORDS, 1751 s_unpack_from__doc__}, 1752 {"__sizeof__", (PyCFunction)s_sizeof, METH_NOARGS, s_sizeof__doc__}, 1753 {NULL, NULL} /* sentinel */ 1754 }; 1755 1756 PyDoc_STRVAR(s__doc__, "Compiled struct object"); 1757 1758 #define OFF(x) offsetof(PyStructObject, x) 1759 1760 static PyGetSetDef s_getsetlist[] = { 1761 {"format", (getter)s_get_format, (setter)NULL, "struct format string", NULL}, 1762 {"size", (getter)s_get_size, (setter)NULL, "struct size in bytes", NULL}, 1763 {NULL} /* sentinel */ 1764 }; 1765 1766 static 1767 PyTypeObject PyStructType = { 1768 PyVarObject_HEAD_INIT(NULL, 0) 1769 "Struct", 1770 sizeof(PyStructObject), 1771 0, 1772 (destructor)s_dealloc, /* tp_dealloc */ 1773 0, /* tp_print */ 1774 0, /* tp_getattr */ 1775 0, /* tp_setattr */ 1776 0, /* tp_compare */ 1777 0, /* tp_repr */ 1778 0, /* tp_as_number */ 1779 0, /* tp_as_sequence */ 1780 0, /* tp_as_mapping */ 1781 0, /* tp_hash */ 1782 0, /* tp_call */ 1783 0, /* tp_str */ 1784 PyObject_GenericGetAttr, /* tp_getattro */ 1785 PyObject_GenericSetAttr, /* tp_setattro */ 1786 0, /* tp_as_buffer */ 1787 Py_TPFLAGS_DEFAULT | Py_TPFLAGS_BASETYPE | Py_TPFLAGS_HAVE_WEAKREFS,/* tp_flags */ 1788 s__doc__, /* tp_doc */ 1789 0, /* tp_traverse */ 1790 0, /* tp_clear */ 1791 0, /* tp_richcompare */ 1792 offsetof(PyStructObject, weakreflist), /* tp_weaklistoffset */ 1793 0, /* tp_iter */ 1794 0, /* tp_iternext */ 1795 s_methods, /* tp_methods */ 1796 NULL, /* tp_members */ 1797 s_getsetlist, /* tp_getset */ 1798 0, /* tp_base */ 1799 0, /* tp_dict */ 1800 0, /* tp_descr_get */ 1801 0, /* tp_descr_set */ 1802 0, /* tp_dictoffset */ 1803 s_init, /* tp_init */ 1804 PyType_GenericAlloc,/* tp_alloc */ 1805 s_new, /* tp_new */ 1806 PyObject_Del, /* tp_free */ 1807 }; 1808 1809 1810 /* ---- Standalone functions ---- */ 1811 1812 #define MAXCACHE 100 1813 static PyObject *cache = NULL; 1814 1815 static PyObject * 1816 cache_struct(PyObject *fmt) 1817 { 1818 PyObject * s_object; 1819 1820 if (cache == NULL) { 1821 cache = PyDict_New(); 1822 if (cache == NULL) 1823 return NULL; 1824 } 1825 1826 s_object = PyDict_GetItem(cache, fmt); 1827 if (s_object != NULL) { 1828 Py_INCREF(s_object); 1829 return s_object; 1830 } 1831 1832 s_object = PyObject_CallFunctionObjArgs((PyObject *)(&PyStructType), fmt, NULL); 1833 if (s_object != NULL) { 1834 if (PyDict_Size(cache) >= MAXCACHE) 1835 PyDict_Clear(cache); 1836 /* Attempt to cache the result */ 1837 if (PyDict_SetItem(cache, fmt, s_object) == -1) 1838 PyErr_Clear(); 1839 } 1840 return s_object; 1841 } 1842 1843 PyDoc_STRVAR(clearcache_doc, 1844 "Clear the internal cache."); 1845 1846 static PyObject * 1847 clearcache(PyObject *self) 1848 { 1849 Py_CLEAR(cache); 1850 Py_RETURN_NONE; 1851 } 1852 1853 PyDoc_STRVAR(calcsize_doc, 1854 "Return size of C struct described by format string fmt."); 1855 1856 static PyObject * 1857 calcsize(PyObject *self, PyObject *fmt) 1858 { 1859 Py_ssize_t n; 1860 PyObject *s_object = cache_struct(fmt); 1861 if (s_object == NULL) 1862 return NULL; 1863 n = ((PyStructObject *)s_object)->s_size; 1864 Py_DECREF(s_object); 1865 return PyInt_FromSsize_t(n); 1866 } 1867 1868 PyDoc_STRVAR(pack_doc, 1869 "Return string containing values v1, v2, ... packed according to fmt."); 1870 1871 static PyObject * 1872 pack(PyObject *self, PyObject *args) 1873 { 1874 PyObject *s_object, *fmt, *newargs, *result; 1875 Py_ssize_t n = PyTuple_GET_SIZE(args); 1876 1877 if (n == 0) { 1878 PyErr_SetString(PyExc_TypeError, "missing format argument"); 1879 return NULL; 1880 } 1881 fmt = PyTuple_GET_ITEM(args, 0); 1882 newargs = PyTuple_GetSlice(args, 1, n); 1883 if (newargs == NULL) 1884 return NULL; 1885 1886 s_object = cache_struct(fmt); 1887 if (s_object == NULL) { 1888 Py_DECREF(newargs); 1889 return NULL; 1890 } 1891 result = s_pack(s_object, newargs); 1892 Py_DECREF(newargs); 1893 Py_DECREF(s_object); 1894 return result; 1895 } 1896 1897 PyDoc_STRVAR(pack_into_doc, 1898 "Pack the values v1, v2, ... according to fmt.\n\ 1899 Write the packed bytes into the writable buffer buf starting at offset."); 1900 1901 static PyObject * 1902 pack_into(PyObject *self, PyObject *args) 1903 { 1904 PyObject *s_object, *fmt, *newargs, *result; 1905 Py_ssize_t n = PyTuple_GET_SIZE(args); 1906 1907 if (n == 0) { 1908 PyErr_SetString(PyExc_TypeError, "missing format argument"); 1909 return NULL; 1910 } 1911 fmt = PyTuple_GET_ITEM(args, 0); 1912 newargs = PyTuple_GetSlice(args, 1, n); 1913 if (newargs == NULL) 1914 return NULL; 1915 1916 s_object = cache_struct(fmt); 1917 if (s_object == NULL) { 1918 Py_DECREF(newargs); 1919 return NULL; 1920 } 1921 result = s_pack_into(s_object, newargs); 1922 Py_DECREF(newargs); 1923 Py_DECREF(s_object); 1924 return result; 1925 } 1926 1927 PyDoc_STRVAR(unpack_doc, 1928 "Unpack the string containing packed C structure data, according to fmt.\n\ 1929 Requires len(string) == calcsize(fmt)."); 1930 1931 static PyObject * 1932 unpack(PyObject *self, PyObject *args) 1933 { 1934 PyObject *s_object, *fmt, *inputstr, *result; 1935 1936 if (!PyArg_UnpackTuple(args, "unpack", 2, 2, &fmt, &inputstr)) 1937 return NULL; 1938 1939 s_object = cache_struct(fmt); 1940 if (s_object == NULL) 1941 return NULL; 1942 result = s_unpack(s_object, inputstr); 1943 Py_DECREF(s_object); 1944 return result; 1945 } 1946 1947 PyDoc_STRVAR(unpack_from_doc, 1948 "Unpack the buffer, containing packed C structure data, according to\n\ 1949 fmt, starting at offset. Requires len(buffer[offset:]) >= calcsize(fmt)."); 1950 1951 static PyObject * 1952 unpack_from(PyObject *self, PyObject *args, PyObject *kwds) 1953 { 1954 PyObject *s_object, *fmt, *newargs, *result; 1955 Py_ssize_t n = PyTuple_GET_SIZE(args); 1956 1957 if (n == 0) { 1958 PyErr_SetString(PyExc_TypeError, "missing format argument"); 1959 return NULL; 1960 } 1961 fmt = PyTuple_GET_ITEM(args, 0); 1962 newargs = PyTuple_GetSlice(args, 1, n); 1963 if (newargs == NULL) 1964 return NULL; 1965 1966 s_object = cache_struct(fmt); 1967 if (s_object == NULL) { 1968 Py_DECREF(newargs); 1969 return NULL; 1970 } 1971 result = s_unpack_from(s_object, newargs, kwds); 1972 Py_DECREF(newargs); 1973 Py_DECREF(s_object); 1974 return result; 1975 } 1976 1977 static struct PyMethodDef module_functions[] = { 1978 {"_clearcache", (PyCFunction)clearcache, METH_NOARGS, clearcache_doc}, 1979 {"calcsize", calcsize, METH_O, calcsize_doc}, 1980 {"pack", pack, METH_VARARGS, pack_doc}, 1981 {"pack_into", pack_into, METH_VARARGS, pack_into_doc}, 1982 {"unpack", unpack, METH_VARARGS, unpack_doc}, 1983 {"unpack_from", (PyCFunction)unpack_from, 1984 METH_VARARGS|METH_KEYWORDS, unpack_from_doc}, 1985 {NULL, NULL} /* sentinel */ 1986 }; 1987 1988 1989 /* Module initialization */ 1990 1991 PyDoc_STRVAR(module_doc, 1992 "Functions to convert between Python values and C structs represented\n\ 1993 as Python strings. It uses format strings (explained below) as compact\n\ 1994 descriptions of the lay-out of the C structs and the intended conversion\n\ 1995 to/from Python values.\n\ 1996 \n\ 1997 The optional first format char indicates byte order, size and alignment:\n\ 1998 @: native order, size & alignment (default)\n\ 1999 =: native order, std. size & alignment\n\ 2000 <: little-endian, std. size & alignment\n\ 2001 >: big-endian, std. size & alignment\n\ 2002 !: same as >\n\ 2003 \n\ 2004 The remaining chars indicate types of args and must match exactly;\n\ 2005 these can be preceded by a decimal repeat count:\n\ 2006 x: pad byte (no data); c:char; b:signed byte; B:unsigned byte;\n\ 2007 ?: _Bool (requires C99; if not available, char is used instead)\n\ 2008 h:short; H:unsigned short; i:int; I:unsigned int;\n\ 2009 l:long; L:unsigned long; f:float; d:double.\n\ 2010 Special cases (preceding decimal count indicates length):\n\ 2011 s:string (array of char); p: pascal string (with count byte).\n\ 2012 Special case (only available in native format):\n\ 2013 P:an integer type that is wide enough to hold a pointer.\n\ 2014 Special case (not in native mode unless 'long long' in platform C):\n\ 2015 q:long long; Q:unsigned long long\n\ 2016 Whitespace between formats is ignored.\n\ 2017 \n\ 2018 The variable struct.error is an exception raised on errors.\n"); 2019 2020 PyMODINIT_FUNC 2021 init_struct(void) 2022 { 2023 PyObject *ver, *m; 2024 2025 ver = PyString_FromString("0.2"); 2026 if (ver == NULL) 2027 return; 2028 2029 m = Py_InitModule3("_struct", module_functions, module_doc); 2030 if (m == NULL) 2031 return; 2032 2033 Py_TYPE(&PyStructType) = &PyType_Type; 2034 if (PyType_Ready(&PyStructType) < 0) 2035 return; 2036 2037 /* This speed trick can't be used until overflow masking goes 2038 away, because native endian always raises exceptions 2039 instead of overflow masking. */ 2040 2041 /* Check endian and swap in faster functions */ 2042 { 2043 int one = 1; 2044 formatdef *native = native_table; 2045 formatdef *other, *ptr; 2046 if ((int)*(unsigned char*)&one) 2047 other = lilendian_table; 2048 else 2049 other = bigendian_table; 2050 /* Scan through the native table, find a matching 2051 entry in the endian table and swap in the 2052 native implementations whenever possible 2053 (64-bit platforms may not have "standard" sizes) */ 2054 while (native->format != '\0' && other->format != '\0') { 2055 ptr = other; 2056 while (ptr->format != '\0') { 2057 if (ptr->format == native->format) { 2058 /* Match faster when formats are 2059 listed in the same order */ 2060 if (ptr == other) 2061 other++; 2062 /* Only use the trick if the 2063 size matches */ 2064 if (ptr->size != native->size) 2065 break; 2066 /* Skip float and double, could be 2067 "unknown" float format */ 2068 if (ptr->format == 'd' || ptr->format == 'f') 2069 break; 2070 ptr->pack = native->pack; 2071 ptr->unpack = native->unpack; 2072 break; 2073 } 2074 ptr++; 2075 } 2076 native++; 2077 } 2078 } 2079 2080 /* Add some symbolic constants to the module */ 2081 if (StructError == NULL) { 2082 StructError = PyErr_NewException("struct.error", NULL, NULL); 2083 if (StructError == NULL) 2084 return; 2085 } 2086 2087 Py_INCREF(StructError); 2088 PyModule_AddObject(m, "error", StructError); 2089 2090 Py_INCREF((PyObject*)&PyStructType); 2091 PyModule_AddObject(m, "Struct", (PyObject*)&PyStructType); 2092 2093 PyModule_AddObject(m, "__version__", ver); 2094 2095 PyModule_AddIntConstant(m, "_PY_STRUCT_RANGE_CHECKING", 1); 2096 PyModule_AddIntConstant(m, "_PY_STRUCT_FLOAT_COERCE", 1); 2097 } 2098