1 /* doop.c 2 * 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 4 * 2001, 2002, 2004, 2005, 2006, 2007, 2008, 2009 by Larry Wall and others 5 * 6 * You may distribute under the terms of either the GNU General Public 7 * License or the Artistic License, as specified in the README file. 8 * 9 */ 10 11 /* 12 * 'So that was the job I felt I had to do when I started,' thought Sam. 13 * 14 * [p.934 of _The Lord of the Rings_, VI/iii: "Mount Doom"] 15 */ 16 17 /* This file contains some common functions needed to carry out certain 18 * ops. For example, both pp_sprintf() and pp_prtf() call the function 19 * do_sprintf() found in this file. 20 */ 21 22 #include "EXTERN.h" 23 #define PERL_IN_DOOP_C 24 #include "perl.h" 25 26 #ifndef PERL_MICRO 27 #include <signal.h> 28 #endif 29 30 31 /* Helper function for do_trans(). 32 * Handles non-utf8 cases(*) not involving the /c, /d, /s flags, 33 * and where search and replacement charlists aren't identical. 34 * (*) i.e. where the search and replacement charlists are non-utf8. sv may 35 * or may not be utf8. 36 */ 37 38 STATIC Size_t 39 S_do_trans_simple(pTHX_ SV * const sv) 40 { 41 Size_t matches = 0; 42 STRLEN len; 43 U8 *s = (U8*)SvPV_nomg(sv,len); 44 U8 * const send = s+len; 45 const OPtrans_map * const tbl = (OPtrans_map*)cPVOP->op_pv; 46 47 PERL_ARGS_ASSERT_DO_TRANS_SIMPLE; 48 49 if (!tbl) 50 Perl_croak(aTHX_ "panic: do_trans_simple line %d",__LINE__); 51 52 /* First, take care of non-UTF-8 input strings, because they're easy */ 53 if (!SvUTF8(sv)) { 54 while (s < send) { 55 const short ch = tbl->map[*s]; 56 if (ch >= 0) { 57 matches++; 58 *s = (U8)ch; 59 } 60 s++; 61 } 62 SvSETMAGIC(sv); 63 } 64 else { 65 const bool grows = cBOOL(PL_op->op_private & OPpTRANS_GROWS); 66 U8 *d; 67 U8 *dstart; 68 69 /* Allow for expansion: $_="a".chr(400); tr/a/\xFE/, FE needs encoding */ 70 if (grows) 71 Newx(d, len*2+1, U8); 72 else 73 d = s; 74 dstart = d; 75 while (s < send) { 76 STRLEN ulen; 77 short ch; 78 79 /* Need to check this, otherwise 128..255 won't match */ 80 const UV c = utf8n_to_uvchr(s, send - s, &ulen, UTF8_ALLOW_DEFAULT); 81 if (c < 0x100 && (ch = tbl->map[c]) >= 0) { 82 matches++; 83 d = uvchr_to_utf8(d, (UV)ch); 84 s += ulen; 85 } 86 else { /* No match -> copy */ 87 Move(s, d, ulen, U8); 88 d += ulen; 89 s += ulen; 90 } 91 } 92 if (grows) { 93 sv_setpvn(sv, (char*)dstart, d - dstart); 94 Safefree(dstart); 95 } 96 else { 97 *d = '\0'; 98 SvCUR_set(sv, d - dstart); 99 } 100 SvUTF8_on(sv); 101 SvSETMAGIC(sv); 102 } 103 return matches; 104 } 105 106 107 /* Helper function for do_trans(). 108 * Handles non-utf8 cases(*) where search and replacement charlists are 109 * identical: so the string isn't modified, and only a count of modifiable 110 * chars is needed. 111 * Note that it doesn't handle /d or /s, since these modify the string 112 * even if the replacement list is empty. 113 * (*) i.e. where the search and replacement charlists are non-utf8. sv may 114 * or may not be utf8. 115 */ 116 117 STATIC Size_t 118 S_do_trans_count(pTHX_ SV * const sv) 119 { 120 STRLEN len; 121 const U8 *s = (const U8*)SvPV_nomg_const(sv, len); 122 const U8 * const send = s + len; 123 Size_t matches = 0; 124 const OPtrans_map * const tbl = (OPtrans_map*)cPVOP->op_pv; 125 126 PERL_ARGS_ASSERT_DO_TRANS_COUNT; 127 128 if (!tbl) 129 Perl_croak(aTHX_ "panic: do_trans_count line %d",__LINE__); 130 131 if (!SvUTF8(sv)) { 132 while (s < send) { 133 if (tbl->map[*s++] >= 0) 134 matches++; 135 } 136 } 137 else { 138 const bool complement = cBOOL(PL_op->op_private & OPpTRANS_COMPLEMENT); 139 while (s < send) { 140 STRLEN ulen; 141 const UV c = utf8n_to_uvchr(s, send - s, &ulen, UTF8_ALLOW_DEFAULT); 142 if (c < 0x100) { 143 if (tbl->map[c] >= 0) 144 matches++; 145 } else if (complement) 146 matches++; 147 s += ulen; 148 } 149 } 150 151 return matches; 152 } 153 154 155 /* Helper function for do_trans(). 156 * Handles non-utf8 cases(*) involving the /c, /d, /s flags, 157 * and where search and replacement charlists aren't identical. 158 * (*) i.e. where the search and replacement charlists are non-utf8. sv may 159 * or may not be utf8. 160 */ 161 162 STATIC Size_t 163 S_do_trans_complex(pTHX_ SV * const sv) 164 { 165 STRLEN len; 166 U8 *s = (U8*)SvPV_nomg(sv, len); 167 U8 * const send = s+len; 168 Size_t matches = 0; 169 const OPtrans_map * const tbl = (OPtrans_map*)cPVOP->op_pv; 170 171 PERL_ARGS_ASSERT_DO_TRANS_COMPLEX; 172 173 if (!tbl) 174 Perl_croak(aTHX_ "panic: do_trans_complex line %d",__LINE__); 175 176 if (!SvUTF8(sv)) { 177 U8 *d = s; 178 U8 * const dstart = d; 179 180 if (PL_op->op_private & OPpTRANS_SQUASH) { 181 const U8* p = send; 182 while (s < send) { 183 const short ch = tbl->map[*s]; 184 if (ch >= 0) { 185 *d = (U8)ch; 186 matches++; 187 if (p != d - 1 || *p != *d) 188 p = d++; 189 } 190 else if (ch == -1) /* -1 is unmapped character */ 191 *d++ = *s; 192 else if (ch == -2) /* -2 is delete character */ 193 matches++; 194 s++; 195 } 196 } 197 else { 198 while (s < send) { 199 const short ch = tbl->map[*s]; 200 if (ch >= 0) { 201 matches++; 202 *d++ = (U8)ch; 203 } 204 else if (ch == -1) /* -1 is unmapped character */ 205 *d++ = *s; 206 else if (ch == -2) /* -2 is delete character */ 207 matches++; 208 s++; 209 } 210 } 211 *d = '\0'; 212 SvCUR_set(sv, d - dstart); 213 } 214 else { /* is utf8 */ 215 const bool squash = cBOOL(PL_op->op_private & OPpTRANS_SQUASH); 216 const bool grows = cBOOL(PL_op->op_private & OPpTRANS_GROWS); 217 U8 *d; 218 U8 *dstart; 219 Size_t size = tbl->size; 220 UV pch = 0xfeedface; 221 222 if (grows) 223 Newx(d, len*2+1, U8); 224 else 225 d = s; 226 dstart = d; 227 228 while (s < send) { 229 STRLEN len; 230 const UV comp = utf8n_to_uvchr(s, send - s, &len, 231 UTF8_ALLOW_DEFAULT); 232 UV ch; 233 short sch; 234 235 sch = tbl->map[comp >= size ? size : comp]; 236 237 if (sch >= 0) { 238 ch = (UV)sch; 239 replace: 240 matches++; 241 if (LIKELY(!squash || ch != pch)) { 242 d = uvchr_to_utf8(d, ch); 243 pch = ch; 244 } 245 s += len; 246 continue; 247 } 248 else if (sch == -1) { /* -1 is unmapped character */ 249 Move(s, d, len, U8); 250 d += len; 251 } 252 else if (sch == -2) /* -2 is delete character */ 253 matches++; 254 else { 255 assert(sch == -3); /* -3 is empty replacement */ 256 ch = comp; 257 goto replace; 258 } 259 260 s += len; 261 pch = 0xfeedface; 262 } 263 264 if (grows) { 265 sv_setpvn(sv, (char*)dstart, d - dstart); 266 Safefree(dstart); 267 } 268 else { 269 *d = '\0'; 270 SvCUR_set(sv, d - dstart); 271 } 272 SvUTF8_on(sv); 273 } 274 SvSETMAGIC(sv); 275 return matches; 276 } 277 278 279 /* Helper function for do_trans(). 280 * Handles utf8 cases(*) not involving the /c, /d, /s flags, 281 * and where search and replacement charlists aren't identical. 282 * (*) i.e. where the search or replacement charlists are utf8. sv may 283 * or may not be utf8. 284 */ 285 286 STATIC Size_t 287 S_do_trans_simple_utf8(pTHX_ SV * const sv) 288 { 289 U8 *s; 290 U8 *send; 291 U8 *d; 292 U8 *start; 293 U8 *dstart, *dend; 294 Size_t matches = 0; 295 const bool grows = cBOOL(PL_op->op_private & OPpTRANS_GROWS); 296 STRLEN len; 297 SV* const rv = 298 #ifdef USE_ITHREADS 299 PAD_SVl(cPADOP->op_padix); 300 #else 301 MUTABLE_SV(cSVOP->op_sv); 302 #endif 303 HV* const hv = MUTABLE_HV(SvRV(rv)); 304 SV* const * svp = hv_fetchs(hv, "NONE", FALSE); 305 const UV none = svp ? SvUV(*svp) : 0x7fffffff; 306 const UV extra = none + 1; 307 UV final = 0; 308 U8 hibit = 0; 309 310 PERL_ARGS_ASSERT_DO_TRANS_SIMPLE_UTF8; 311 312 s = (U8*)SvPV_nomg(sv, len); 313 if (!SvUTF8(sv)) { 314 hibit = ! is_utf8_invariant_string(s, len); 315 if (hibit) { 316 s = bytes_to_utf8(s, &len); 317 } 318 } 319 send = s + len; 320 start = s; 321 322 svp = hv_fetchs(hv, "FINAL", FALSE); 323 if (svp) 324 final = SvUV(*svp); 325 326 if (grows) { 327 /* d needs to be bigger than s, in case e.g. upgrading is required */ 328 Newx(d, len * 3 + UTF8_MAXBYTES, U8); 329 dend = d + len * 3; 330 dstart = d; 331 } 332 else { 333 dstart = d = s; 334 dend = d + len; 335 } 336 337 while (s < send) { 338 const UV uv = swash_fetch(rv, s, TRUE); 339 if (uv < none) { 340 s += UTF8SKIP(s); 341 matches++; 342 d = uvchr_to_utf8(d, uv); 343 } 344 else if (uv == none) { 345 const int i = UTF8SKIP(s); 346 Move(s, d, i, U8); 347 d += i; 348 s += i; 349 } 350 else if (uv == extra) { 351 s += UTF8SKIP(s); 352 matches++; 353 d = uvchr_to_utf8(d, final); 354 } 355 else 356 s += UTF8SKIP(s); 357 358 if (d > dend) { 359 const STRLEN clen = d - dstart; 360 const STRLEN nlen = dend - dstart + len + UTF8_MAXBYTES; 361 if (!grows) 362 Perl_croak(aTHX_ "panic: do_trans_simple_utf8 line %d",__LINE__); 363 Renew(dstart, nlen + UTF8_MAXBYTES, U8); 364 d = dstart + clen; 365 dend = dstart + nlen; 366 } 367 } 368 if (grows || hibit) { 369 sv_setpvn(sv, (char*)dstart, d - dstart); 370 Safefree(dstart); 371 if (grows && hibit) 372 Safefree(start); 373 } 374 else { 375 *d = '\0'; 376 SvCUR_set(sv, d - dstart); 377 } 378 SvSETMAGIC(sv); 379 SvUTF8_on(sv); 380 381 return matches; 382 } 383 384 385 /* Helper function for do_trans(). 386 * Handles utf8 cases(*) where search and replacement charlists are 387 * identical: so the string isn't modified, and only a count of modifiable 388 * chars is needed. 389 * Note that it doesn't handle /d or /s, since these modify the string 390 * even if the replacement charlist is empty. 391 * (*) i.e. where the search or replacement charlists are utf8. sv may 392 * or may not be utf8. 393 */ 394 395 STATIC Size_t 396 S_do_trans_count_utf8(pTHX_ SV * const sv) 397 { 398 const U8 *s; 399 const U8 *start = NULL; 400 const U8 *send; 401 Size_t matches = 0; 402 STRLEN len; 403 SV* const rv = 404 #ifdef USE_ITHREADS 405 PAD_SVl(cPADOP->op_padix); 406 #else 407 MUTABLE_SV(cSVOP->op_sv); 408 #endif 409 HV* const hv = MUTABLE_HV(SvRV(rv)); 410 SV* const * const svp = hv_fetchs(hv, "NONE", FALSE); 411 const UV none = svp ? SvUV(*svp) : 0x7fffffff; 412 const UV extra = none + 1; 413 U8 hibit = 0; 414 415 PERL_ARGS_ASSERT_DO_TRANS_COUNT_UTF8; 416 417 s = (const U8*)SvPV_nomg_const(sv, len); 418 if (!SvUTF8(sv)) { 419 hibit = ! is_utf8_invariant_string(s, len); 420 if (hibit) { 421 start = s = bytes_to_utf8(s, &len); 422 } 423 } 424 send = s + len; 425 426 while (s < send) { 427 const UV uv = swash_fetch(rv, s, TRUE); 428 if (uv < none || uv == extra) 429 matches++; 430 s += UTF8SKIP(s); 431 } 432 if (hibit) 433 Safefree(start); 434 435 return matches; 436 } 437 438 439 /* Helper function for do_trans(). 440 * Handles utf8 cases(*) involving the /c, /d, /s flags, 441 * and where search and replacement charlists aren't identical. 442 * (*) i.e. where the search or replacement charlists are utf8. sv may 443 * or may not be utf8. 444 */ 445 446 STATIC Size_t 447 S_do_trans_complex_utf8(pTHX_ SV * const sv) 448 { 449 U8 *start, *send; 450 U8 *d; 451 Size_t matches = 0; 452 const bool squash = cBOOL(PL_op->op_private & OPpTRANS_SQUASH); 453 const bool del = cBOOL(PL_op->op_private & OPpTRANS_DELETE); 454 const bool grows = cBOOL(PL_op->op_private & OPpTRANS_GROWS); 455 SV* const rv = 456 #ifdef USE_ITHREADS 457 PAD_SVl(cPADOP->op_padix); 458 #else 459 MUTABLE_SV(cSVOP->op_sv); 460 #endif 461 HV * const hv = MUTABLE_HV(SvRV(rv)); 462 SV * const *svp = hv_fetchs(hv, "NONE", FALSE); 463 const UV none = svp ? SvUV(*svp) : 0x7fffffff; 464 const UV extra = none + 1; 465 UV final = 0; 466 bool havefinal = FALSE; 467 STRLEN len; 468 U8 *dstart, *dend; 469 U8 hibit = 0; 470 U8 *s = (U8*)SvPV_nomg(sv, len); 471 472 PERL_ARGS_ASSERT_DO_TRANS_COMPLEX_UTF8; 473 474 if (!SvUTF8(sv)) { 475 hibit = ! is_utf8_invariant_string(s, len); 476 if (hibit) { 477 s = bytes_to_utf8(s, &len); 478 } 479 } 480 send = s + len; 481 start = s; 482 483 svp = hv_fetchs(hv, "FINAL", FALSE); 484 if (svp) { 485 final = SvUV(*svp); 486 havefinal = TRUE; 487 } 488 489 if (grows) { 490 /* d needs to be bigger than s, in case e.g. upgrading is required */ 491 Newx(d, len * 3 + UTF8_MAXBYTES, U8); 492 dend = d + len * 3; 493 dstart = d; 494 } 495 else { 496 dstart = d = s; 497 dend = d + len; 498 } 499 500 if (squash) { 501 UV puv = 0xfeedface; 502 while (s < send) { 503 UV uv = swash_fetch(rv, s, TRUE); 504 505 if (d > dend) { 506 const STRLEN clen = d - dstart; 507 const STRLEN nlen = dend - dstart + len + UTF8_MAXBYTES; 508 if (!grows) 509 Perl_croak(aTHX_ "panic: do_trans_complex_utf8 line %d",__LINE__); 510 Renew(dstart, nlen + UTF8_MAXBYTES, U8); 511 d = dstart + clen; 512 dend = dstart + nlen; 513 } 514 if (uv < none) { 515 matches++; 516 s += UTF8SKIP(s); 517 if (uv != puv) { 518 d = uvchr_to_utf8(d, uv); 519 puv = uv; 520 } 521 continue; 522 } 523 else if (uv == none) { /* "none" is unmapped character */ 524 const int i = UTF8SKIP(s); 525 Move(s, d, i, U8); 526 d += i; 527 s += i; 528 puv = 0xfeedface; 529 continue; 530 } 531 else if (uv == extra && !del) { 532 matches++; 533 if (havefinal) { 534 s += UTF8SKIP(s); 535 if (puv != final) { 536 d = uvchr_to_utf8(d, final); 537 puv = final; 538 } 539 } 540 else { 541 STRLEN len; 542 uv = utf8n_to_uvchr(s, send - s, &len, UTF8_ALLOW_DEFAULT); 543 if (uv != puv) { 544 Move(s, d, len, U8); 545 d += len; 546 puv = uv; 547 } 548 s += len; 549 } 550 continue; 551 } 552 matches++; /* "none+1" is delete character */ 553 s += UTF8SKIP(s); 554 } 555 } 556 else { 557 while (s < send) { 558 const UV uv = swash_fetch(rv, s, TRUE); 559 if (d > dend) { 560 const STRLEN clen = d - dstart; 561 const STRLEN nlen = dend - dstart + len + UTF8_MAXBYTES; 562 if (!grows) 563 Perl_croak(aTHX_ "panic: do_trans_complex_utf8 line %d",__LINE__); 564 Renew(dstart, nlen + UTF8_MAXBYTES, U8); 565 d = dstart + clen; 566 dend = dstart + nlen; 567 } 568 if (uv < none) { 569 matches++; 570 s += UTF8SKIP(s); 571 d = uvchr_to_utf8(d, uv); 572 continue; 573 } 574 else if (uv == none) { /* "none" is unmapped character */ 575 const int i = UTF8SKIP(s); 576 Move(s, d, i, U8); 577 d += i; 578 s += i; 579 continue; 580 } 581 else if (uv == extra && !del) { 582 matches++; 583 s += UTF8SKIP(s); 584 d = uvchr_to_utf8(d, final); 585 continue; 586 } 587 matches++; /* "none+1" is delete character */ 588 s += UTF8SKIP(s); 589 } 590 } 591 if (grows || hibit) { 592 sv_setpvn(sv, (char*)dstart, d - dstart); 593 Safefree(dstart); 594 if (grows && hibit) 595 Safefree(start); 596 } 597 else { 598 *d = '\0'; 599 SvCUR_set(sv, d - dstart); 600 } 601 SvUTF8_on(sv); 602 SvSETMAGIC(sv); 603 604 return matches; 605 } 606 607 608 /* Execute a tr//. sv is the value to be translated, while PL_op 609 * should be an OP_TRANS or OP_TRANSR op, whose op_pv field contains a 610 * translation table or whose op_sv field contains a swash. 611 * Returns a count of number of characters translated 612 */ 613 614 Size_t 615 Perl_do_trans(pTHX_ SV *sv) 616 { 617 STRLEN len; 618 const U8 flags = PL_op->op_private; 619 const U8 hasutf = flags & (OPpTRANS_FROM_UTF | OPpTRANS_TO_UTF); 620 621 PERL_ARGS_ASSERT_DO_TRANS; 622 623 if (SvREADONLY(sv) && !(flags & OPpTRANS_IDENTICAL)) { 624 Perl_croak_no_modify(); 625 } 626 (void)SvPV_const(sv, len); 627 if (!len) 628 return 0; 629 if (!(flags & OPpTRANS_IDENTICAL)) { 630 if (!SvPOKp(sv) || SvTHINKFIRST(sv)) 631 (void)SvPV_force_nomg(sv, len); 632 (void)SvPOK_only_UTF8(sv); 633 } 634 635 /* If we use only OPpTRANS_IDENTICAL to bypass the READONLY check, 636 * we must also rely on it to choose the readonly strategy. 637 */ 638 if (flags & OPpTRANS_IDENTICAL) { 639 return hasutf ? do_trans_count_utf8(sv) : do_trans_count(sv); 640 } else if (flags & (OPpTRANS_SQUASH|OPpTRANS_DELETE|OPpTRANS_COMPLEMENT)) { 641 return hasutf ? do_trans_complex_utf8(sv) : do_trans_complex(sv); 642 } else { 643 return hasutf ? do_trans_simple_utf8(sv) : do_trans_simple(sv); 644 } 645 } 646 647 void 648 Perl_do_join(pTHX_ SV *sv, SV *delim, SV **mark, SV **sp) 649 { 650 SV ** const oldmark = mark; 651 I32 items = sp - mark; 652 STRLEN len; 653 STRLEN delimlen; 654 const char * const delims = SvPV_const(delim, delimlen); 655 656 PERL_ARGS_ASSERT_DO_JOIN; 657 658 mark++; 659 len = (items > 0 ? (delimlen * (items - 1) ) : 0); 660 SvUPGRADE(sv, SVt_PV); 661 if (SvLEN(sv) < len + items) { /* current length is way too short */ 662 while (items-- > 0) { 663 if (*mark && !SvGAMAGIC(*mark) && SvOK(*mark)) { 664 STRLEN tmplen; 665 SvPV_const(*mark, tmplen); 666 len += tmplen; 667 } 668 mark++; 669 } 670 SvGROW(sv, len + 1); /* so try to pre-extend */ 671 672 mark = oldmark; 673 items = sp - mark; 674 ++mark; 675 } 676 677 SvPVCLEAR(sv); 678 /* sv_setpv retains old UTF8ness [perl #24846] */ 679 SvUTF8_off(sv); 680 681 if (TAINTING_get && SvMAGICAL(sv)) 682 SvTAINTED_off(sv); 683 684 if (items-- > 0) { 685 if (*mark) 686 sv_catsv(sv, *mark); 687 mark++; 688 } 689 690 if (delimlen) { 691 const U32 delimflag = DO_UTF8(delim) ? SV_CATUTF8 : SV_CATBYTES; 692 for (; items > 0; items--,mark++) { 693 STRLEN len; 694 const char *s; 695 sv_catpvn_flags(sv,delims,delimlen,delimflag); 696 s = SvPV_const(*mark,len); 697 sv_catpvn_flags(sv,s,len, 698 DO_UTF8(*mark) ? SV_CATUTF8 : SV_CATBYTES); 699 } 700 } 701 else { 702 for (; items > 0; items--,mark++) 703 { 704 STRLEN len; 705 const char *s = SvPV_const(*mark,len); 706 sv_catpvn_flags(sv,s,len, 707 DO_UTF8(*mark) ? SV_CATUTF8 : SV_CATBYTES); 708 } 709 } 710 SvSETMAGIC(sv); 711 } 712 713 void 714 Perl_do_sprintf(pTHX_ SV *sv, SSize_t len, SV **sarg) 715 { 716 STRLEN patlen; 717 const char * const pat = SvPV_const(*sarg, patlen); 718 bool do_taint = FALSE; 719 720 PERL_ARGS_ASSERT_DO_SPRINTF; 721 assert(len >= 1); 722 723 if (SvTAINTED(*sarg)) 724 TAINT_PROPER( 725 (PL_op && PL_op->op_type < OP_max) 726 ? (PL_op->op_type == OP_PRTF) 727 ? "printf" 728 : PL_op_name[PL_op->op_type] 729 : "(unknown)" 730 ); 731 SvUTF8_off(sv); 732 if (DO_UTF8(*sarg)) 733 SvUTF8_on(sv); 734 sv_vsetpvfn(sv, pat, patlen, NULL, sarg + 1, (Size_t)(len - 1), &do_taint); 735 SvSETMAGIC(sv); 736 if (do_taint) 737 SvTAINTED_on(sv); 738 } 739 740 /* currently converts input to bytes if possible, but doesn't sweat failure */ 741 UV 742 Perl_do_vecget(pTHX_ SV *sv, STRLEN offset, int size) 743 { 744 STRLEN srclen, len, avail, uoffset, bitoffs = 0; 745 const I32 svpv_flags = ((PL_op->op_flags & OPf_MOD || LVRET) 746 ? SV_UNDEF_RETURNS_NULL : 0); 747 unsigned char *s = (unsigned char *) 748 SvPV_flags(sv, srclen, (svpv_flags|SV_GMAGIC)); 749 UV retnum = 0; 750 751 if (!s) { 752 s = (unsigned char *)""; 753 } 754 755 PERL_ARGS_ASSERT_DO_VECGET; 756 757 if (size < 1 || (size & (size-1))) /* size < 1 or not a power of two */ 758 Perl_croak(aTHX_ "Illegal number of bits in vec"); 759 760 if (SvUTF8(sv)) { 761 if (Perl_sv_utf8_downgrade(aTHX_ sv, TRUE)) { 762 /* PVX may have changed */ 763 s = (unsigned char *) SvPV_flags(sv, srclen, svpv_flags); 764 } 765 else { 766 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), 767 "Use of strings with code points over 0xFF as" 768 " arguments to vec is deprecated. This will" 769 " be a fatal error in Perl 5.32"); 770 } 771 } 772 773 if (size < 8) { 774 bitoffs = ((offset%8)*size)%8; 775 uoffset = offset/(8/size); 776 } 777 else if (size > 8) { 778 int n = size/8; 779 if (offset > Size_t_MAX / n - 1) /* would overflow */ 780 return 0; 781 uoffset = offset*n; 782 } 783 else 784 uoffset = offset; 785 786 if (uoffset >= srclen) 787 return 0; 788 789 len = (bitoffs + size + 7)/8; /* required number of bytes */ 790 avail = srclen - uoffset; /* available number of bytes */ 791 792 /* Does the byte range overlap the end of the string? If so, 793 * handle specially. */ 794 if (avail < len) { 795 if (size <= 8) 796 retnum = 0; 797 else { 798 if (size == 16) { 799 assert(avail == 1); 800 retnum = (UV) s[uoffset] << 8; 801 } 802 else if (size == 32) { 803 assert(avail >= 1 && avail <= 3); 804 if (avail == 1) 805 retnum = 806 ((UV) s[uoffset ] << 24); 807 else if (avail == 2) 808 retnum = 809 ((UV) s[uoffset ] << 24) + 810 ((UV) s[uoffset + 1] << 16); 811 else 812 retnum = 813 ((UV) s[uoffset ] << 24) + 814 ((UV) s[uoffset + 1] << 16) + 815 ( s[uoffset + 2] << 8); 816 } 817 #ifdef UV_IS_QUAD 818 else if (size == 64) { 819 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE), 820 "Bit vector size > 32 non-portable"); 821 assert(avail >= 1 && avail <= 7); 822 if (avail == 1) 823 retnum = 824 (UV) s[uoffset ] << 56; 825 else if (avail == 2) 826 retnum = 827 ((UV) s[uoffset ] << 56) + 828 ((UV) s[uoffset + 1] << 48); 829 else if (avail == 3) 830 retnum = 831 ((UV) s[uoffset ] << 56) + 832 ((UV) s[uoffset + 1] << 48) + 833 ((UV) s[uoffset + 2] << 40); 834 else if (avail == 4) 835 retnum = 836 ((UV) s[uoffset ] << 56) + 837 ((UV) s[uoffset + 1] << 48) + 838 ((UV) s[uoffset + 2] << 40) + 839 ((UV) s[uoffset + 3] << 32); 840 else if (avail == 5) 841 retnum = 842 ((UV) s[uoffset ] << 56) + 843 ((UV) s[uoffset + 1] << 48) + 844 ((UV) s[uoffset + 2] << 40) + 845 ((UV) s[uoffset + 3] << 32) + 846 ((UV) s[uoffset + 4] << 24); 847 else if (avail == 6) 848 retnum = 849 ((UV) s[uoffset ] << 56) + 850 ((UV) s[uoffset + 1] << 48) + 851 ((UV) s[uoffset + 2] << 40) + 852 ((UV) s[uoffset + 3] << 32) + 853 ((UV) s[uoffset + 4] << 24) + 854 ((UV) s[uoffset + 5] << 16); 855 else 856 retnum = 857 ((UV) s[uoffset ] << 56) + 858 ((UV) s[uoffset + 1] << 48) + 859 ((UV) s[uoffset + 2] << 40) + 860 ((UV) s[uoffset + 3] << 32) + 861 ((UV) s[uoffset + 4] << 24) + 862 ((UV) s[uoffset + 5] << 16) + 863 ((UV) s[uoffset + 6] << 8); 864 } 865 #endif 866 } 867 } 868 else if (size < 8) 869 retnum = (s[uoffset] >> bitoffs) & ((1 << size) - 1); 870 else { 871 if (size == 8) 872 retnum = s[uoffset]; 873 else if (size == 16) 874 retnum = 875 ((UV) s[uoffset] << 8) + 876 s[uoffset + 1]; 877 else if (size == 32) 878 retnum = 879 ((UV) s[uoffset ] << 24) + 880 ((UV) s[uoffset + 1] << 16) + 881 ( s[uoffset + 2] << 8) + 882 s[uoffset + 3]; 883 #ifdef UV_IS_QUAD 884 else if (size == 64) { 885 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE), 886 "Bit vector size > 32 non-portable"); 887 retnum = 888 ((UV) s[uoffset ] << 56) + 889 ((UV) s[uoffset + 1] << 48) + 890 ((UV) s[uoffset + 2] << 40) + 891 ((UV) s[uoffset + 3] << 32) + 892 ((UV) s[uoffset + 4] << 24) + 893 ((UV) s[uoffset + 5] << 16) + 894 ( s[uoffset + 6] << 8) + 895 s[uoffset + 7]; 896 } 897 #endif 898 } 899 900 return retnum; 901 } 902 903 /* currently converts input to bytes if possible but doesn't sweat failures, 904 * although it does ensure that the string it clobbers is not marked as 905 * utf8-valid any more 906 */ 907 void 908 Perl_do_vecset(pTHX_ SV *sv) 909 { 910 STRLEN offset, bitoffs = 0; 911 int size; 912 unsigned char *s; 913 UV lval; 914 I32 mask; 915 STRLEN targlen; 916 STRLEN len; 917 SV * const targ = LvTARG(sv); 918 char errflags = LvFLAGS(sv); 919 920 PERL_ARGS_ASSERT_DO_VECSET; 921 922 /* some out-of-range errors have been deferred if/until the LV is 923 * actually written to: f(vec($s,-1,8)) is not always fatal */ 924 if (errflags) { 925 assert(!(errflags & ~(LVf_NEG_OFF|LVf_OUT_OF_RANGE))); 926 if (errflags & LVf_NEG_OFF) 927 Perl_croak_nocontext("Negative offset to vec in lvalue context"); 928 Perl_croak_nocontext("Out of memory!"); 929 } 930 931 if (!targ) 932 return; 933 s = (unsigned char*)SvPV_force_flags(targ, targlen, 934 SV_GMAGIC | SV_UNDEF_RETURNS_NULL); 935 if (SvUTF8(targ)) { 936 /* This is handled by the SvPOK_only below... 937 if (!Perl_sv_utf8_downgrade(aTHX_ targ, TRUE)) 938 SvUTF8_off(targ); 939 */ 940 (void) Perl_sv_utf8_downgrade(aTHX_ targ, TRUE); 941 } 942 943 (void)SvPOK_only(targ); 944 lval = SvUV(sv); 945 offset = LvTARGOFF(sv); 946 size = LvTARGLEN(sv); 947 948 if (size < 1 || (size & (size-1))) /* size < 1 or not a power of two */ 949 Perl_croak(aTHX_ "Illegal number of bits in vec"); 950 951 if (size < 8) { 952 bitoffs = ((offset%8)*size)%8; 953 offset /= 8/size; 954 } 955 else if (size > 8) { 956 int n = size/8; 957 if (offset > Size_t_MAX / n - 1) /* would overflow */ 958 Perl_croak_nocontext("Out of memory!"); 959 offset *= n; 960 } 961 962 len = (bitoffs + size + 7)/8; /* required number of bytes */ 963 if (targlen < offset || targlen - offset < len) { 964 STRLEN newlen = offset > Size_t_MAX - len - 1 ? /* avoid overflow */ 965 Size_t_MAX : offset + len + 1; 966 s = (unsigned char*)SvGROW(targ, newlen); 967 (void)memzero((char *)(s + targlen), newlen - targlen); 968 SvCUR_set(targ, newlen - 1); 969 } 970 971 if (size < 8) { 972 mask = (1 << size) - 1; 973 lval &= mask; 974 s[offset] &= ~(mask << bitoffs); 975 s[offset] |= lval << bitoffs; 976 } 977 else { 978 if (size == 8) 979 s[offset ] = (U8)( lval & 0xff); 980 else if (size == 16) { 981 s[offset ] = (U8)((lval >> 8) & 0xff); 982 s[offset+1] = (U8)( lval & 0xff); 983 } 984 else if (size == 32) { 985 s[offset ] = (U8)((lval >> 24) & 0xff); 986 s[offset+1] = (U8)((lval >> 16) & 0xff); 987 s[offset+2] = (U8)((lval >> 8) & 0xff); 988 s[offset+3] = (U8)( lval & 0xff); 989 } 990 #ifdef UV_IS_QUAD 991 else if (size == 64) { 992 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE), 993 "Bit vector size > 32 non-portable"); 994 s[offset ] = (U8)((lval >> 56) & 0xff); 995 s[offset+1] = (U8)((lval >> 48) & 0xff); 996 s[offset+2] = (U8)((lval >> 40) & 0xff); 997 s[offset+3] = (U8)((lval >> 32) & 0xff); 998 s[offset+4] = (U8)((lval >> 24) & 0xff); 999 s[offset+5] = (U8)((lval >> 16) & 0xff); 1000 s[offset+6] = (U8)((lval >> 8) & 0xff); 1001 s[offset+7] = (U8)( lval & 0xff); 1002 } 1003 #endif 1004 } 1005 SvSETMAGIC(targ); 1006 } 1007 1008 void 1009 Perl_do_vop(pTHX_ I32 optype, SV *sv, SV *left, SV *right) 1010 { 1011 #ifdef LIBERAL 1012 long *dl; 1013 long *ll; 1014 long *rl; 1015 #endif 1016 char *dc; 1017 STRLEN leftlen; 1018 STRLEN rightlen; 1019 const char *lc; 1020 const char *rc; 1021 STRLEN len = 0; 1022 STRLEN lensave; 1023 const char *lsave; 1024 const char *rsave; 1025 STRLEN needlen = 0; 1026 bool result_needs_to_be_utf8 = FALSE; 1027 bool left_utf8 = FALSE; 1028 bool right_utf8 = FALSE; 1029 U8 * left_non_downgraded = NULL; 1030 U8 * right_non_downgraded = NULL; 1031 Size_t left_non_downgraded_len = 0; 1032 Size_t right_non_downgraded_len = 0; 1033 char * non_downgraded = NULL; 1034 Size_t non_downgraded_len = 0; 1035 1036 PERL_ARGS_ASSERT_DO_VOP; 1037 1038 if (sv != left || (optype != OP_BIT_AND && !SvOK(sv))) 1039 SvPVCLEAR(sv); /* avoid undef warning on |= and ^= */ 1040 if (sv == left) { 1041 lc = SvPV_force_nomg(left, leftlen); 1042 } 1043 else { 1044 lc = SvPV_nomg_const(left, leftlen); 1045 SvPV_force_nomg_nolen(sv); 1046 } 1047 rc = SvPV_nomg_const(right, rightlen); 1048 1049 /* This needs to come after SvPV to ensure that string overloading has 1050 fired off. */ 1051 1052 /* Create downgraded temporaries of any UTF-8 encoded operands */ 1053 if (DO_UTF8(left)) { 1054 const U8 * save_lc = (U8 *) lc; 1055 1056 left_utf8 = TRUE; 1057 result_needs_to_be_utf8 = TRUE; 1058 1059 left_non_downgraded_len = leftlen; 1060 lc = (char *) bytes_from_utf8_loc((const U8 *) lc, &leftlen, 1061 &left_utf8, 1062 (const U8 **) &left_non_downgraded); 1063 /* Calculate the number of trailing unconvertible bytes. This quantity 1064 * is the original length minus the length of the converted portion. */ 1065 left_non_downgraded_len -= left_non_downgraded - save_lc; 1066 SAVEFREEPV(lc); 1067 } 1068 if (DO_UTF8(right)) { 1069 const U8 * save_rc = (U8 *) rc; 1070 1071 right_utf8 = TRUE; 1072 result_needs_to_be_utf8 = TRUE; 1073 1074 right_non_downgraded_len = rightlen; 1075 rc = (char *) bytes_from_utf8_loc((const U8 *) rc, &rightlen, 1076 &right_utf8, 1077 (const U8 **) &right_non_downgraded); 1078 right_non_downgraded_len -= right_non_downgraded - save_rc; 1079 SAVEFREEPV(rc); 1080 } 1081 1082 /* We set 'len' to the length that the operation actually operates on. The 1083 * dangling part of the longer operand doesn't actually participate in the 1084 * operation. What happens is that we pretend that the shorter operand has 1085 * been extended to the right by enough imaginary zeros to match the length 1086 * of the longer one. But we know in advance the result of the operation 1087 * on zeros without having to do it. In the case of '&', the result is 1088 * zero, and the dangling portion is simply discarded. For '|' and '^', the 1089 * result is the same as the other operand, so the dangling part is just 1090 * appended to the final result, unchanged. We currently accept above-FF 1091 * code points in the dangling portion, as that's how it has long worked, 1092 * and code depends on it staying that way. But it is now fatal for 1093 * above-FF to appear in the portion that does get operated on. Hence, any 1094 * above-FF must come only in the longer operand, and only in its dangling 1095 * portion. That means that at least one of the operands has to be 1096 * entirely non-UTF-8, and the length of that operand has to be before the 1097 * first above-FF in the other */ 1098 if (left_utf8 || right_utf8) { 1099 if (left_utf8) { 1100 if (right_utf8 || rightlen > leftlen) { 1101 Perl_croak(aTHX_ FATAL_ABOVE_FF_MSG, PL_op_desc[optype]); 1102 } 1103 len = rightlen; 1104 } 1105 else if (right_utf8) { 1106 if (leftlen > rightlen) { 1107 Perl_croak(aTHX_ FATAL_ABOVE_FF_MSG, PL_op_desc[optype]); 1108 } 1109 len = leftlen; 1110 } 1111 1112 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), 1113 DEPRECATED_ABOVE_FF_MSG, PL_op_desc[optype]); 1114 } 1115 else { /* Neither is UTF-8 */ 1116 len = MIN(leftlen, rightlen); 1117 } 1118 1119 lensave = len; 1120 lsave = lc; 1121 rsave = rc; 1122 1123 SvCUR_set(sv, len); 1124 (void)SvPOK_only(sv); 1125 if (SvOK(sv) || SvTYPE(sv) > SVt_PVMG) { 1126 dc = SvPV_force_nomg_nolen(sv); 1127 if (SvLEN(sv) < len + 1) { 1128 dc = SvGROW(sv, len + 1); 1129 (void)memzero(dc + SvCUR(sv), len - SvCUR(sv) + 1); 1130 } 1131 } 1132 else { 1133 needlen = optype == OP_BIT_AND 1134 ? len : (leftlen > rightlen ? leftlen : rightlen); 1135 Newxz(dc, needlen + 1, char); 1136 sv_usepvn_flags(sv, dc, needlen, SV_HAS_TRAILING_NUL); 1137 dc = SvPVX(sv); /* sv_usepvn() calls Renew() */ 1138 } 1139 1140 #ifdef LIBERAL 1141 if (len >= sizeof(long)*4 && 1142 !((unsigned long)dc % sizeof(long)) && 1143 !((unsigned long)lc % sizeof(long)) && 1144 !((unsigned long)rc % sizeof(long))) /* It's almost always aligned... */ 1145 { 1146 const STRLEN remainder = len % (sizeof(long)*4); 1147 len /= (sizeof(long)*4); 1148 1149 dl = (long*)dc; 1150 ll = (long*)lc; 1151 rl = (long*)rc; 1152 1153 switch (optype) { 1154 case OP_BIT_AND: 1155 while (len--) { 1156 *dl++ = *ll++ & *rl++; 1157 *dl++ = *ll++ & *rl++; 1158 *dl++ = *ll++ & *rl++; 1159 *dl++ = *ll++ & *rl++; 1160 } 1161 break; 1162 case OP_BIT_XOR: 1163 while (len--) { 1164 *dl++ = *ll++ ^ *rl++; 1165 *dl++ = *ll++ ^ *rl++; 1166 *dl++ = *ll++ ^ *rl++; 1167 *dl++ = *ll++ ^ *rl++; 1168 } 1169 break; 1170 case OP_BIT_OR: 1171 while (len--) { 1172 *dl++ = *ll++ | *rl++; 1173 *dl++ = *ll++ | *rl++; 1174 *dl++ = *ll++ | *rl++; 1175 *dl++ = *ll++ | *rl++; 1176 } 1177 } 1178 1179 dc = (char*)dl; 1180 lc = (char*)ll; 1181 rc = (char*)rl; 1182 1183 len = remainder; 1184 } 1185 #endif 1186 switch (optype) { 1187 case OP_BIT_AND: 1188 while (len--) 1189 *dc++ = *lc++ & *rc++; 1190 *dc = '\0'; 1191 break; 1192 case OP_BIT_XOR: 1193 while (len--) 1194 *dc++ = *lc++ ^ *rc++; 1195 goto mop_up; 1196 case OP_BIT_OR: 1197 while (len--) 1198 *dc++ = *lc++ | *rc++; 1199 mop_up: 1200 len = lensave; 1201 if (rightlen > len) { 1202 if (dc == rc) 1203 SvCUR(sv) = rightlen; 1204 else 1205 sv_catpvn_nomg(sv, rsave + len, rightlen - len); 1206 } 1207 else if (leftlen > len) { 1208 if (dc == lc) 1209 SvCUR(sv) = leftlen; 1210 else 1211 sv_catpvn_nomg(sv, lsave + len, leftlen - len); 1212 } 1213 *SvEND(sv) = '\0'; 1214 1215 /* If there is trailing stuff that couldn't be converted from UTF-8, it 1216 * is appended as-is for the ^ and | operators. This preserves 1217 * backwards compatibility */ 1218 if (right_non_downgraded) { 1219 non_downgraded = (char *) right_non_downgraded; 1220 non_downgraded_len = right_non_downgraded_len; 1221 } 1222 else if (left_non_downgraded) { 1223 non_downgraded = (char *) left_non_downgraded; 1224 non_downgraded_len = left_non_downgraded_len; 1225 } 1226 1227 break; 1228 } 1229 1230 if (result_needs_to_be_utf8) { 1231 sv_utf8_upgrade_nomg(sv); 1232 1233 /* Append any trailing UTF-8 as-is. */ 1234 if (non_downgraded) { 1235 sv_catpvn_nomg(sv, non_downgraded, non_downgraded_len); 1236 } 1237 } 1238 1239 SvTAINT(sv); 1240 } 1241 1242 1243 /* Perl_do_kv() may be: 1244 * * called directly as the pp function for pp_keys() and pp_values(); 1245 * * It may also be called directly when the op is OP_AVHVSWITCH, to 1246 * implement CORE::keys(), CORE::values(). 1247 * 1248 * In all cases it expects an HV on the stack and returns a list of keys, 1249 * values, or key-value pairs, depending on PL_op. 1250 */ 1251 1252 OP * 1253 Perl_do_kv(pTHX) 1254 { 1255 dSP; 1256 HV * const keys = MUTABLE_HV(POPs); 1257 const U8 gimme = GIMME_V; 1258 1259 const I32 dokeys = (PL_op->op_type == OP_KEYS) 1260 || ( PL_op->op_type == OP_AVHVSWITCH 1261 && (PL_op->op_private & OPpAVHVSWITCH_MASK) 1262 + OP_EACH == OP_KEYS); 1263 1264 const I32 dovalues = (PL_op->op_type == OP_VALUES) 1265 || ( PL_op->op_type == OP_AVHVSWITCH 1266 && (PL_op->op_private & OPpAVHVSWITCH_MASK) 1267 + OP_EACH == OP_VALUES); 1268 1269 assert( PL_op->op_type == OP_KEYS 1270 || PL_op->op_type == OP_VALUES 1271 || PL_op->op_type == OP_AVHVSWITCH); 1272 1273 assert(!( PL_op->op_type == OP_VALUES 1274 && (PL_op->op_private & OPpMAYBE_LVSUB))); 1275 1276 (void)hv_iterinit(keys); /* always reset iterator regardless */ 1277 1278 if (gimme == G_VOID) 1279 RETURN; 1280 1281 if (gimme == G_SCALAR) { 1282 if (PL_op->op_flags & OPf_MOD || LVRET) { /* lvalue */ 1283 SV * const ret = sv_2mortal(newSV_type(SVt_PVLV)); /* Not TARG RT#67838 */ 1284 sv_magic(ret, NULL, PERL_MAGIC_nkeys, NULL, 0); 1285 LvTYPE(ret) = 'k'; 1286 LvTARG(ret) = SvREFCNT_inc_simple(keys); 1287 PUSHs(ret); 1288 } 1289 else { 1290 IV i; 1291 dTARGET; 1292 1293 /* note that in 'scalar(keys %h)' the OP_KEYS is usually 1294 * optimised away and the action is performed directly by the 1295 * padhv or rv2hv op. We now only get here via OP_AVHVSWITCH 1296 * and \&CORE::keys 1297 */ 1298 if (! SvTIED_mg((const SV *)keys, PERL_MAGIC_tied) ) { 1299 i = HvUSEDKEYS(keys); 1300 } 1301 else { 1302 i = 0; 1303 while (hv_iternext(keys)) i++; 1304 } 1305 PUSHi( i ); 1306 } 1307 RETURN; 1308 } 1309 1310 if (UNLIKELY(PL_op->op_private & OPpMAYBE_LVSUB)) { 1311 const I32 flags = is_lvalue_sub(); 1312 if (flags && !(flags & OPpENTERSUB_INARGS)) 1313 /* diag_listed_as: Can't modify %s in %s */ 1314 Perl_croak(aTHX_ "Can't modify keys in list assignment"); 1315 } 1316 1317 PUTBACK; 1318 hv_pushkv(keys, (dokeys | (dovalues << 1))); 1319 return NORMAL; 1320 } 1321 1322 /* 1323 * ex: set ts=8 sts=4 sw=4 et: 1324 */ 1325