| 1 |
/* |
| 2 |
* lexical analyzer |
| 3 |
* This file is #included by regcomp.c. |
| 4 |
* |
| 5 |
* Copyright (c) 1998, 1999 Henry Spencer. All rights reserved. |
| 6 |
* |
| 7 |
* Development of this software was funded, in part, by Cray Research Inc., |
| 8 |
* UUNET Communications Services Inc., Sun Microsystems Inc., and Scriptics |
| 9 |
* Corporation, none of whom are responsible for the results. The author |
| 10 |
* thanks all of them. |
| 11 |
* |
| 12 |
* Redistribution and use in source and binary forms -- with or without |
| 13 |
* modification -- are permitted for any purpose, provided that |
| 14 |
* redistributions in source form retain this entire copyright notice and |
| 15 |
* indicate the origin and nature of any modifications. |
| 16 |
* |
| 17 |
* I'd appreciate being given credit for this package in the documentation |
| 18 |
* of software which uses it, but that is not a requirement. |
| 19 |
* |
| 20 |
* THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, |
| 21 |
* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY |
| 22 |
* AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL |
| 23 |
* HENRY SPENCER BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, |
| 24 |
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
| 25 |
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; |
| 26 |
* OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, |
| 27 |
* WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR |
| 28 |
* OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF |
| 29 |
* ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 30 |
* |
| 31 |
* src/backend/regex/regc_lex.c |
| 32 |
* |
| 33 |
*/ |
| 34 |
|
| 35 |
/* scanning macros (know about v) */ |
| 36 |
#define ATEOS() (v->now >= v->stop) |
| 37 |
#define HAVE(n) (v->stop - v->now >= (n)) |
| 38 |
#define NEXT1(c) (!ATEOS() && *v->now == CHR(c)) |
| 39 |
#define NEXT2(a,b) (HAVE(2) && *v->now == CHR(a) && *(v->now+1) == CHR(b)) |
| 40 |
#define NEXT3(a,b,c) (HAVE(3) && *v->now == CHR(a) && \ |
| 41 |
*(v->now+1) == CHR(b) && \ |
| 42 |
*(v->now+2) == CHR(c)) |
| 43 |
#define SET(c) (v->nexttype = (c)) |
| 44 |
#define SETV(c, n) (v->nexttype = (c), v->nextvalue = (n)) |
| 45 |
#define RET(c) return (SET(c), 1) |
| 46 |
#define RETV(c, n) return (SETV(c, n), 1) |
| 47 |
#define FAILW(e) return (ERR(e), 0) /* ERR does SET(EOS) */ |
| 48 |
#define LASTTYPE(t) (v->lasttype == (t)) |
| 49 |
|
| 50 |
/* lexical contexts */ |
| 51 |
#define L_ERE 1 /* mainline ERE/ARE */ |
| 52 |
#define L_BRE 2 /* mainline BRE */ |
| 53 |
#define L_Q 3 /* REG_QUOTE */ |
| 54 |
#define L_EBND 4 /* ERE/ARE bound */ |
| 55 |
#define L_BBND 5 /* BRE bound */ |
| 56 |
#define L_BRACK 6 /* brackets */ |
| 57 |
#define L_CEL 7 /* collating element */ |
| 58 |
#define L_ECL 8 /* equivalence class */ |
| 59 |
#define L_CCL 9 /* character class */ |
| 60 |
#define INTOCON(c) (v->lexcon = (c)) |
| 61 |
#define INCON(con) (v->lexcon == (con)) |
| 62 |
|
| 63 |
/* construct pointer past end of chr array */ |
| 64 |
#define ENDOF(array) ((array) + sizeof(array)/sizeof(chr)) |
| 65 |
|
| 66 |
/* |
| 67 |
* lexstart - set up lexical stuff, scan leading options |
| 68 |
*/ |
| 69 |
static void |
| 70 |
lexstart(struct vars *v) |
| 71 |
{ |
| 72 |
prefixes(v); /* may turn on new type bits etc. */ |
| 73 |
NOERR(); |
| 74 |
|
| 75 |
if (v->cflags & REG_QUOTE) |
| 76 |
{ |
| 77 |
assert(!(v->cflags & (REG_ADVANCED | REG_EXPANDED | REG_NEWLINE))); |
| 78 |
INTOCON(L_Q); |
| 79 |
} |
| 80 |
else if (v->cflags & REG_EXTENDED) |
| 81 |
{ |
| 82 |
assert(!(v->cflags & REG_QUOTE)); |
| 83 |
INTOCON(L_ERE); |
| 84 |
} |
| 85 |
else |
| 86 |
{ |
| 87 |
assert(!(v->cflags & (REG_QUOTE | REG_ADVF))); |
| 88 |
INTOCON(L_BRE); |
| 89 |
} |
| 90 |
|
| 91 |
v->nexttype = EMPTY; /* remember we were at the start */ |
| 92 |
next(v); /* set up the first token */ |
| 93 |
} |
| 94 |
|
| 95 |
/* |
| 96 |
* prefixes - implement various special prefixes |
| 97 |
*/ |
| 98 |
static void |
| 99 |
prefixes(struct vars *v) |
| 100 |
{ |
| 101 |
/* literal string doesn't get any of this stuff */ |
| 102 |
if (v->cflags & REG_QUOTE) |
| 103 |
return; |
| 104 |
|
| 105 |
/* initial "***" gets special things */ |
| 106 |
if (HAVE(4) && NEXT3('*', '*', '*')) |
| 107 |
switch (*(v->now + 3)) |
| 108 |
{ |
| 109 |
case CHR('?'): /* "***?" error, msg shows version */ |
| 110 |
ERR(REG_BADPAT); |
| 111 |
return; /* proceed no further */ |
| 112 |
break; |
| 113 |
case CHR('='): /* "***=" shifts to literal string */ |
| 114 |
NOTE(REG_UNONPOSIX); |
| 115 |
v->cflags |= REG_QUOTE; |
| 116 |
v->cflags &= ~(REG_ADVANCED | REG_EXPANDED | REG_NEWLINE); |
| 117 |
v->now += 4; |
| 118 |
return; /* and there can be no more prefixes */ |
| 119 |
break; |
| 120 |
case CHR(':'): /* "***:" shifts to AREs */ |
| 121 |
NOTE(REG_UNONPOSIX); |
| 122 |
v->cflags |= REG_ADVANCED; |
| 123 |
v->now += 4; |
| 124 |
break; |
| 125 |
default: /* otherwise *** is just an error */ |
| 126 |
ERR(REG_BADRPT); |
| 127 |
return; |
| 128 |
break; |
| 129 |
} |
| 130 |
|
| 131 |
/* BREs and EREs don't get embedded options */ |
| 132 |
if ((v->cflags & REG_ADVANCED) != REG_ADVANCED) |
| 133 |
return; |
| 134 |
|
| 135 |
/* embedded options (AREs only) */ |
| 136 |
if (HAVE(3) && NEXT2('(', '?') && iscalpha(*(v->now + 2))) |
| 137 |
{ |
| 138 |
NOTE(REG_UNONPOSIX); |
| 139 |
v->now += 2; |
| 140 |
for (; !ATEOS() && iscalpha(*v->now); v->now++) |
| 141 |
switch (*v->now) |
| 142 |
{ |
| 143 |
case CHR('b'): /* BREs (but why???) */ |
| 144 |
v->cflags &= ~(REG_ADVANCED | REG_QUOTE); |
| 145 |
break; |
| 146 |
case CHR('c'): /* case sensitive */ |
| 147 |
v->cflags &= ~REG_ICASE; |
| 148 |
break; |
| 149 |
case CHR('e'): /* plain EREs */ |
| 150 |
v->cflags |= REG_EXTENDED; |
| 151 |
v->cflags &= ~(REG_ADVF | REG_QUOTE); |
| 152 |
break; |
| 153 |
case CHR('i'): /* case insensitive */ |
| 154 |
v->cflags |= REG_ICASE; |
| 155 |
break; |
| 156 |
case CHR('m'): /* Perloid synonym for n */ |
| 157 |
case CHR('n'): /* \n affects ^ $ . [^ */ |
| 158 |
v->cflags |= REG_NEWLINE; |
| 159 |
break; |
| 160 |
case CHR('p'): /* ~Perl, \n affects . [^ */ |
| 161 |
v->cflags |= REG_NLSTOP; |
| 162 |
v->cflags &= ~REG_NLANCH; |
| 163 |
break; |
| 164 |
case CHR('q'): /* literal string */ |
| 165 |
v->cflags |= REG_QUOTE; |
| 166 |
v->cflags &= ~REG_ADVANCED; |
| 167 |
break; |
| 168 |
case CHR('s'): /* single line, \n ordinary */ |
| 169 |
v->cflags &= ~REG_NEWLINE; |
| 170 |
break; |
| 171 |
case CHR('t'): /* tight syntax */ |
| 172 |
v->cflags &= ~REG_EXPANDED; |
| 173 |
break; |
| 174 |
case CHR('w'): /* weird, \n affects ^ $ only */ |
| 175 |
v->cflags &= ~REG_NLSTOP; |
| 176 |
v->cflags |= REG_NLANCH; |
| 177 |
break; |
| 178 |
case CHR('x'): /* expanded syntax */ |
| 179 |
v->cflags |= REG_EXPANDED; |
| 180 |
break; |
| 181 |
default: |
| 182 |
ERR(REG_BADOPT); |
| 183 |
return; |
| 184 |
} |
| 185 |
if (!NEXT1(')')) |
| 186 |
{ |
| 187 |
ERR(REG_BADOPT); |
| 188 |
return; |
| 189 |
} |
| 190 |
v->now++; |
| 191 |
if (v->cflags & REG_QUOTE) |
| 192 |
v->cflags &= ~(REG_EXPANDED | REG_NEWLINE); |
| 193 |
} |
| 194 |
} |
| 195 |
|
| 196 |
/* |
| 197 |
* next - get next token |
| 198 |
*/ |
| 199 |
static int /* 1 normal, 0 failure */ |
| 200 |
next(struct vars *v) |
| 201 |
{ |
| 202 |
chr c; |
| 203 |
|
| 204 |
next_restart: /* loop here after eating a comment */ |
| 205 |
|
| 206 |
/* errors yield an infinite sequence of failures */ |
| 207 |
if (ISERR()) |
| 208 |
return 0; /* the error has set nexttype to EOS */ |
| 209 |
|
| 210 |
/* remember flavor of last token */ |
| 211 |
v->lasttype = v->nexttype; |
| 212 |
|
| 213 |
/* REG_BOSONLY */ |
| 214 |
if (v->nexttype == EMPTY && (v->cflags & REG_BOSONLY)) |
| 215 |
{ |
| 216 |
/* at start of a REG_BOSONLY RE */ |
| 217 |
RETV(SBEGIN, 0); /* same as \A */ |
| 218 |
} |
| 219 |
|
| 220 |
/* skip white space etc. if appropriate (not in literal or []) */ |
| 221 |
if (v->cflags & REG_EXPANDED) |
| 222 |
switch (v->lexcon) |
| 223 |
{ |
| 224 |
case L_ERE: |
| 225 |
case L_BRE: |
| 226 |
case L_EBND: |
| 227 |
case L_BBND: |
| 228 |
skip(v); |
| 229 |
break; |
| 230 |
} |
| 231 |
|
| 232 |
/* handle EOS, depending on context */ |
| 233 |
if (ATEOS()) |
| 234 |
{ |
| 235 |
switch (v->lexcon) |
| 236 |
{ |
| 237 |
case L_ERE: |
| 238 |
case L_BRE: |
| 239 |
case L_Q: |
| 240 |
RET(EOS); |
| 241 |
break; |
| 242 |
case L_EBND: |
| 243 |
case L_BBND: |
| 244 |
FAILW(REG_EBRACE); |
| 245 |
break; |
| 246 |
case L_BRACK: |
| 247 |
case L_CEL: |
| 248 |
case L_ECL: |
| 249 |
case L_CCL: |
| 250 |
FAILW(REG_EBRACK); |
| 251 |
break; |
| 252 |
} |
| 253 |
assert(NOTREACHED); |
| 254 |
} |
| 255 |
|
| 256 |
/* okay, time to actually get a character */ |
| 257 |
c = *v->now++; |
| 258 |
|
| 259 |
/* deal with the easy contexts, punt EREs to code below */ |
| 260 |
switch (v->lexcon) |
| 261 |
{ |
| 262 |
case L_BRE: /* punt BREs to separate function */ |
| 263 |
return brenext(v, c); |
| 264 |
break; |
| 265 |
case L_ERE: /* see below */ |
| 266 |
break; |
| 267 |
case L_Q: /* literal strings are easy */ |
| 268 |
RETV(PLAIN, c); |
| 269 |
break; |
| 270 |
case L_BBND: /* bounds are fairly simple */ |
| 271 |
case L_EBND: |
| 272 |
switch (c) |
| 273 |
{ |
| 274 |
case CHR('0'): |
| 275 |
case CHR('1'): |
| 276 |
case CHR('2'): |
| 277 |
case CHR('3'): |
| 278 |
case CHR('4'): |
| 279 |
case CHR('5'): |
| 280 |
case CHR('6'): |
| 281 |
case CHR('7'): |
| 282 |
case CHR('8'): |
| 283 |
case CHR('9'): |
| 284 |
RETV(DIGIT, (chr) DIGITVAL(c)); |
| 285 |
break; |
| 286 |
case CHR(','): |
| 287 |
RET(','); |
| 288 |
break; |
| 289 |
case CHR('}'): /* ERE bound ends with } */ |
| 290 |
if (INCON(L_EBND)) |
| 291 |
{ |
| 292 |
INTOCON(L_ERE); |
| 293 |
if ((v->cflags & REG_ADVF) && NEXT1('?')) |
| 294 |
{ |
| 295 |
v->now++; |
| 296 |
NOTE(REG_UNONPOSIX); |
| 297 |
RETV('}', 0); |
| 298 |
} |
| 299 |
RETV('}', 1); |
| 300 |
} |
| 301 |
else |
| 302 |
FAILW(REG_BADBR); |
| 303 |
break; |
| 304 |
case CHR('\\'): /* BRE bound ends with \} */ |
| 305 |
if (INCON(L_BBND) && NEXT1('}')) |
| 306 |
{ |
| 307 |
v->now++; |
| 308 |
INTOCON(L_BRE); |
| 309 |
RETV('}', 1); |
| 310 |
} |
| 311 |
else |
| 312 |
FAILW(REG_BADBR); |
| 313 |
break; |
| 314 |
default: |
| 315 |
FAILW(REG_BADBR); |
| 316 |
break; |
| 317 |
} |
| 318 |
assert(NOTREACHED); |
| 319 |
break; |
| 320 |
case L_BRACK: /* brackets are not too hard */ |
| 321 |
switch (c) |
| 322 |
{ |
| 323 |
case CHR(']'): |
| 324 |
if (LASTTYPE('[')) |
| 325 |
RETV(PLAIN, c); |
| 326 |
else |
| 327 |
{ |
| 328 |
INTOCON((v->cflags & REG_EXTENDED) ? |
| 329 |
L_ERE : L_BRE); |
| 330 |
RET(']'); |
| 331 |
} |
| 332 |
break; |
| 333 |
case CHR('\\'): |
| 334 |
NOTE(REG_UBBS); |
| 335 |
if (!(v->cflags & REG_ADVF)) |
| 336 |
RETV(PLAIN, c); |
| 337 |
NOTE(REG_UNONPOSIX); |
| 338 |
if (ATEOS()) |
| 339 |
FAILW(REG_EESCAPE); |
| 340 |
if (!lexescape(v)) |
| 341 |
return 0; |
| 342 |
switch (v->nexttype) |
| 343 |
{ /* not all escapes okay here */ |
| 344 |
case PLAIN: |
| 345 |
case CCLASSS: |
| 346 |
case CCLASSC: |
| 347 |
return 1; |
| 348 |
break; |
| 349 |
} |
| 350 |
/* not one of the acceptable escapes */ |
| 351 |
FAILW(REG_EESCAPE); |
| 352 |
break; |
| 353 |
case CHR('-'): |
| 354 |
if (LASTTYPE('[') || NEXT1(']')) |
| 355 |
RETV(PLAIN, c); |
| 356 |
else |
| 357 |
RETV(RANGE, c); |
| 358 |
break; |
| 359 |
case CHR('['): |
| 360 |
if (ATEOS()) |
| 361 |
FAILW(REG_EBRACK); |
| 362 |
switch (*v->now++) |
| 363 |
{ |
| 364 |
case CHR('.'): |
| 365 |
INTOCON(L_CEL); |
| 366 |
/* might or might not be locale-specific */ |
| 367 |
RET(COLLEL); |
| 368 |
break; |
| 369 |
case CHR('='): |
| 370 |
INTOCON(L_ECL); |
| 371 |
NOTE(REG_ULOCALE); |
| 372 |
RET(ECLASS); |
| 373 |
break; |
| 374 |
case CHR(':'): |
| 375 |
INTOCON(L_CCL); |
| 376 |
NOTE(REG_ULOCALE); |
| 377 |
RET(CCLASS); |
| 378 |
break; |
| 379 |
default: /* oops */ |
| 380 |
v->now--; |
| 381 |
RETV(PLAIN, c); |
| 382 |
break; |
| 383 |
} |
| 384 |
assert(NOTREACHED); |
| 385 |
break; |
| 386 |
default: |
| 387 |
RETV(PLAIN, c); |
| 388 |
break; |
| 389 |
} |
| 390 |
assert(NOTREACHED); |
| 391 |
break; |
| 392 |
case L_CEL: /* collating elements are easy */ |
| 393 |
if (c == CHR('.') && NEXT1(']')) |
| 394 |
{ |
| 395 |
v->now++; |
| 396 |
INTOCON(L_BRACK); |
| 397 |
RETV(END, '.'); |
| 398 |
} |
| 399 |
else |
| 400 |
RETV(PLAIN, c); |
| 401 |
break; |
| 402 |
case L_ECL: /* ditto equivalence classes */ |
| 403 |
if (c == CHR('=') && NEXT1(']')) |
| 404 |
{ |
| 405 |
v->now++; |
| 406 |
INTOCON(L_BRACK); |
| 407 |
RETV(END, '='); |
| 408 |
} |
| 409 |
else |
| 410 |
RETV(PLAIN, c); |
| 411 |
break; |
| 412 |
case L_CCL: /* ditto character classes */ |
| 413 |
if (c == CHR(':') && NEXT1(']')) |
| 414 |
{ |
| 415 |
v->now++; |
| 416 |
INTOCON(L_BRACK); |
| 417 |
RETV(END, ':'); |
| 418 |
} |
| 419 |
else |
| 420 |
RETV(PLAIN, c); |
| 421 |
break; |
| 422 |
default: |
| 423 |
assert(NOTREACHED); |
| 424 |
break; |
| 425 |
} |
| 426 |
|
| 427 |
/* that got rid of everything except EREs and AREs */ |
| 428 |
assert(INCON(L_ERE)); |
| 429 |
|
| 430 |
/* deal with EREs and AREs, except for backslashes */ |
| 431 |
switch (c) |
| 432 |
{ |
| 433 |
case CHR('|'): |
| 434 |
RET('|'); |
| 435 |
break; |
| 436 |
case CHR('*'): |
| 437 |
if ((v->cflags & REG_ADVF) && NEXT1('?')) |
| 438 |
{ |
| 439 |
v->now++; |
| 440 |
NOTE(REG_UNONPOSIX); |
| 441 |
RETV('*', 0); |
| 442 |
} |
| 443 |
RETV('*', 1); |
| 444 |
break; |
| 445 |
case CHR('+'): |
| 446 |
if ((v->cflags & REG_ADVF) && NEXT1('?')) |
| 447 |
{ |
| 448 |
v->now++; |
| 449 |
NOTE(REG_UNONPOSIX); |
| 450 |
RETV('+', 0); |
| 451 |
} |
| 452 |
RETV('+', 1); |
| 453 |
break; |
| 454 |
case CHR('?'): |
| 455 |
if ((v->cflags & REG_ADVF) && NEXT1('?')) |
| 456 |
{ |
| 457 |
v->now++; |
| 458 |
NOTE(REG_UNONPOSIX); |
| 459 |
RETV('?', 0); |
| 460 |
} |
| 461 |
RETV('?', 1); |
| 462 |
break; |
| 463 |
case CHR('{'): /* bounds start or plain character */ |
| 464 |
if (v->cflags & REG_EXPANDED) |
| 465 |
skip(v); |
| 466 |
if (ATEOS() || !iscdigit(*v->now)) |
| 467 |
{ |
| 468 |
NOTE(REG_UBRACES); |
| 469 |
NOTE(REG_UUNSPEC); |
| 470 |
RETV(PLAIN, c); |
| 471 |
} |
| 472 |
else |
| 473 |
{ |
| 474 |
NOTE(REG_UBOUNDS); |
| 475 |
INTOCON(L_EBND); |
| 476 |
RET('{'); |
| 477 |
} |
| 478 |
assert(NOTREACHED); |
| 479 |
break; |
| 480 |
case CHR('('): /* parenthesis, or advanced extension */ |
| 481 |
if ((v->cflags & REG_ADVF) && NEXT1('?')) |
| 482 |
{ |
| 483 |
NOTE(REG_UNONPOSIX); |
| 484 |
v->now++; |
| 485 |
if (ATEOS()) |
| 486 |
FAILW(REG_BADRPT); |
| 487 |
switch (*v->now++) |
| 488 |
{ |
| 489 |
case CHR(':'): /* non-capturing paren */ |
| 490 |
RETV('(', 0); |
| 491 |
break; |
| 492 |
case CHR('#'): /* comment */ |
| 493 |
while (!ATEOS() && *v->now != CHR(')')) |
| 494 |
v->now++; |
| 495 |
if (!ATEOS()) |
| 496 |
v->now++; |
| 497 |
assert(v->nexttype == v->lasttype); |
| 498 |
goto next_restart; |
| 499 |
case CHR('='): /* positive lookahead */ |
| 500 |
NOTE(REG_ULOOKAROUND); |
| 501 |
RETV(LACON, LATYPE_AHEAD_POS); |
| 502 |
break; |
| 503 |
case CHR('!'): /* negative lookahead */ |
| 504 |
NOTE(REG_ULOOKAROUND); |
| 505 |
RETV(LACON, LATYPE_AHEAD_NEG); |
| 506 |
break; |
| 507 |
case CHR('<'): |
| 508 |
if (ATEOS()) |
| 509 |
FAILW(REG_BADRPT); |
| 510 |
switch (*v->now++) |
| 511 |
{ |
| 512 |
case CHR('='): /* positive lookbehind */ |
| 513 |
NOTE(REG_ULOOKAROUND); |
| 514 |
RETV(LACON, LATYPE_BEHIND_POS); |
| 515 |
break; |
| 516 |
case CHR('!'): /* negative lookbehind */ |
| 517 |
NOTE(REG_ULOOKAROUND); |
| 518 |
RETV(LACON, LATYPE_BEHIND_NEG); |
| 519 |
break; |
| 520 |
default: |
| 521 |
FAILW(REG_BADRPT); |
| 522 |
break; |
| 523 |
} |
| 524 |
assert(NOTREACHED); |
| 525 |
break; |
| 526 |
default: |
| 527 |
FAILW(REG_BADRPT); |
| 528 |
break; |
| 529 |
} |
| 530 |
assert(NOTREACHED); |
| 531 |
} |
| 532 |
RETV('(', 1); |
| 533 |
break; |
| 534 |
case CHR(')'): |
| 535 |
if (LASTTYPE('(')) |
| 536 |
NOTE(REG_UUNSPEC); |
| 537 |
RETV(')', c); |
| 538 |
break; |
| 539 |
case CHR('['): /* easy except for [[:<:]] and [[:>:]] */ |
| 540 |
if (HAVE(6) && *(v->now + 0) == CHR('[') && |
| 541 |
*(v->now + 1) == CHR(':') && |
| 542 |
(*(v->now + 2) == CHR('<') || |
| 543 |
*(v->now + 2) == CHR('>')) && |
| 544 |
*(v->now + 3) == CHR(':') && |
| 545 |
*(v->now + 4) == CHR(']') && |
| 546 |
*(v->now + 5) == CHR(']')) |
| 547 |
{ |
| 548 |
c = *(v->now + 2); |
| 549 |
v->now += 6; |
| 550 |
NOTE(REG_UNONPOSIX); |
| 551 |
RET((c == CHR('<')) ? '<' : '>'); |
| 552 |
} |
| 553 |
INTOCON(L_BRACK); |
| 554 |
if (NEXT1('^')) |
| 555 |
{ |
| 556 |
v->now++; |
| 557 |
RETV('[', 0); |
| 558 |
} |
| 559 |
RETV('[', 1); |
| 560 |
break; |
| 561 |
case CHR('.'): |
| 562 |
RET('.'); |
| 563 |
break; |
| 564 |
case CHR('^'): |
| 565 |
RET('^'); |
| 566 |
break; |
| 567 |
case CHR('$'): |
| 568 |
RET('$'); |
| 569 |
break; |
| 570 |
case CHR('\\'): /* mostly punt backslashes to code below */ |
| 571 |
if (ATEOS()) |
| 572 |
FAILW(REG_EESCAPE); |
| 573 |
break; |
| 574 |
default: /* ordinary character */ |
| 575 |
RETV(PLAIN, c); |
| 576 |
break; |
| 577 |
} |
| 578 |
|
| 579 |
/* ERE/ARE backslash handling; backslash already eaten */ |
| 580 |
assert(!ATEOS()); |
| 581 |
if (!(v->cflags & REG_ADVF)) |
| 582 |
{ /* only AREs have non-trivial escapes */ |
| 583 |
if (iscalnum(*v->now)) |
| 584 |
{ |
| 585 |
NOTE(REG_UBSALNUM); |
| 586 |
NOTE(REG_UUNSPEC); |
| 587 |
} |
| 588 |
RETV(PLAIN, *v->now++); |
| 589 |
} |
| 590 |
return lexescape(v); |
| 591 |
} |
| 592 |
|
| 593 |
/* |
| 594 |
* lexescape - parse an ARE backslash escape (backslash already eaten) |
| 595 |
* |
| 596 |
* This is used for ARE backslashes both normally and inside bracket |
| 597 |
* expressions. In the latter case, not all escape types are allowed, |
| 598 |
* but the caller must reject unwanted ones after we return. |
| 599 |
*/ |
| 600 |
static int |
| 601 |
lexescape(struct vars *v) |
| 602 |
{ |
| 603 |
chr c; |
| 604 |
static const chr alert[] = { |
| 605 |
CHR('a'), CHR('l'), CHR('e'), CHR('r'), CHR('t') |
| 606 |
}; |
| 607 |
static const chr esc[] = { |
| 608 |
CHR('E'), CHR('S'), CHR('C') |
| 609 |
}; |
| 610 |
const chr *save; |
| 611 |
|
| 612 |
assert(v->cflags & REG_ADVF); |
| 613 |
|
| 614 |
assert(!ATEOS()); |
| 615 |
c = *v->now++; |
| 616 |
|
| 617 |
/* if it's not alphanumeric ASCII, treat it as a plain character */ |
| 618 |
if (!('a' <= c && c <= 'z') && |
| 619 |
!('A' <= c && c <= 'Z') && |
| 620 |
!('0' <= c && c <= '9')) |
| 621 |
RETV(PLAIN, c); |
| 622 |
|
| 623 |
NOTE(REG_UNONPOSIX); |
| 624 |
switch (c) |
| 625 |
{ |
| 626 |
case CHR('a'): |
| 627 |
RETV(PLAIN, chrnamed(v, alert, ENDOF(alert), CHR('\007'))); |
| 628 |
break; |
| 629 |
case CHR('A'): |
| 630 |
RETV(SBEGIN, 0); |
| 631 |
break; |
| 632 |
case CHR('b'): |
| 633 |
RETV(PLAIN, CHR('\b')); |
| 634 |
break; |
| 635 |
case CHR('B'): |
| 636 |
RETV(PLAIN, CHR('\\')); |
| 637 |
break; |
| 638 |
case CHR('c'): |
| 639 |
NOTE(REG_UUNPORT); |
| 640 |
if (ATEOS()) |
| 641 |
FAILW(REG_EESCAPE); |
| 642 |
RETV(PLAIN, (chr) (*v->now++ & 037)); |
| 643 |
break; |
| 644 |
case CHR('d'): |
| 645 |
NOTE(REG_ULOCALE); |
| 646 |
RETV(CCLASSS, CC_DIGIT); |
| 647 |
break; |
| 648 |
case CHR('D'): |
| 649 |
NOTE(REG_ULOCALE); |
| 650 |
RETV(CCLASSC, CC_DIGIT); |
| 651 |
break; |
| 652 |
case CHR('e'): |
| 653 |
NOTE(REG_UUNPORT); |
| 654 |
RETV(PLAIN, chrnamed(v, esc, ENDOF(esc), CHR('\033'))); |
| 655 |
break; |
| 656 |
case CHR('f'): |
| 657 |
RETV(PLAIN, CHR('\f')); |
| 658 |
break; |
| 659 |
case CHR('m'): |
| 660 |
RET('<'); |
| 661 |
break; |
| 662 |
case CHR('M'): |
| 663 |
RET('>'); |
| 664 |
break; |
| 665 |
case CHR('n'): |
| 666 |
RETV(PLAIN, CHR('\n')); |
| 667 |
break; |
| 668 |
case CHR('r'): |
| 669 |
RETV(PLAIN, CHR('\r')); |
| 670 |
break; |
| 671 |
case CHR('s'): |
| 672 |
NOTE(REG_ULOCALE); |
| 673 |
RETV(CCLASSS, CC_SPACE); |
| 674 |
break; |
| 675 |
case CHR('S'): |
| 676 |
NOTE(REG_ULOCALE); |
| 677 |
RETV(CCLASSC, CC_SPACE); |
| 678 |
break; |
| 679 |
case CHR('t'): |
| 680 |
RETV(PLAIN, CHR('\t')); |
| 681 |
break; |
| 682 |
case CHR('u'): |
| 683 |
c = lexdigits(v, 16, 4, 4); |
| 684 |
if (ISERR() || !CHR_IS_IN_RANGE(c)) |
| 685 |
FAILW(REG_EESCAPE); |
| 686 |
RETV(PLAIN, c); |
| 687 |
break; |
| 688 |
case CHR('U'): |
| 689 |
c = lexdigits(v, 16, 8, 8); |
| 690 |
if (ISERR() || !CHR_IS_IN_RANGE(c)) |
| 691 |
FAILW(REG_EESCAPE); |
| 692 |
RETV(PLAIN, c); |
| 693 |
break; |
| 694 |
case CHR('v'): |
| 695 |
RETV(PLAIN, CHR('\v')); |
| 696 |
break; |
| 697 |
case CHR('w'): |
| 698 |
NOTE(REG_ULOCALE); |
| 699 |
RETV(CCLASSS, CC_WORD); |
| 700 |
break; |
| 701 |
case CHR('W'): |
| 702 |
NOTE(REG_ULOCALE); |
| 703 |
RETV(CCLASSC, CC_WORD); |
| 704 |
break; |
| 705 |
case CHR('x'): |
| 706 |
NOTE(REG_UUNPORT); |
| 707 |
c = lexdigits(v, 16, 1, 255); /* REs >255 long outside spec */ |
| 708 |
if (ISERR() || !CHR_IS_IN_RANGE(c)) |
| 709 |
FAILW(REG_EESCAPE); |
| 710 |
RETV(PLAIN, c); |
| 711 |
break; |
| 712 |
case CHR('y'): |
| 713 |
NOTE(REG_ULOCALE); |
| 714 |
RETV(WBDRY, 0); |
| 715 |
break; |
| 716 |
case CHR('Y'): |
| 717 |
NOTE(REG_ULOCALE); |
| 718 |
RETV(NWBDRY, 0); |
| 719 |
break; |
| 720 |
case CHR('Z'): |
| 721 |
RETV(SEND, 0); |
| 722 |
break; |
| 723 |
case CHR('1'): |
| 724 |
case CHR('2'): |
| 725 |
case CHR('3'): |
| 726 |
case CHR('4'): |
| 727 |
case CHR('5'): |
| 728 |
case CHR('6'): |
| 729 |
case CHR('7'): |
| 730 |
case CHR('8'): |
| 731 |
case CHR('9'): |
| 732 |
save = v->now; |
| 733 |
v->now--; /* put first digit back */ |
| 734 |
c = lexdigits(v, 10, 1, 255); /* REs >255 long outside spec */ |
| 735 |
if (ISERR()) |
| 736 |
FAILW(REG_EESCAPE); |
| 737 |
/* ugly heuristic (first test is "exactly 1 digit?") */ |
| 738 |
if (v->now == save || ((int) c > 0 && (int) c <= v->nsubexp)) |
| 739 |
{ |
| 740 |
NOTE(REG_UBACKREF); |
| 741 |
RETV(BACKREF, c); |
| 742 |
} |
| 743 |
/* oops, doesn't look like it's a backref after all... */ |
| 744 |
v->now = save; |
| 745 |
/* and fall through into octal number */ |
| 746 |
/* FALLTHROUGH */ |
| 747 |
case CHR('0'): |
| 748 |
NOTE(REG_UUNPORT); |
| 749 |
v->now--; /* put first digit back */ |
| 750 |
c = lexdigits(v, 8, 1, 3); |
| 751 |
if (ISERR()) |
| 752 |
FAILW(REG_EESCAPE); |
| 753 |
if (c > 0xff) |
| 754 |
{ |
| 755 |
/* out of range, so we handled one digit too much */ |
| 756 |
v->now--; |
| 757 |
c >>= 3; |
| 758 |
} |
| 759 |
RETV(PLAIN, c); |
| 760 |
break; |
| 761 |
default: |
| 762 |
|
| 763 |
/* |
| 764 |
* Throw an error for unrecognized ASCII alpha escape sequences, |
| 765 |
* which reserves them for future use if needed. |
| 766 |
*/ |
| 767 |
FAILW(REG_EESCAPE); |
| 768 |
break; |
| 769 |
} |
| 770 |
assert(NOTREACHED); |
| 771 |
} |
| 772 |
|
| 773 |
/* |
| 774 |
* lexdigits - slurp up digits and return chr value |
| 775 |
* |
| 776 |
* This does not account for overflow; callers should range-check the result |
| 777 |
* if maxlen is large enough to make that possible. |
| 778 |
*/ |
| 779 |
static chr /* chr value; errors signalled via ERR */ |
| 780 |
lexdigits(struct vars *v, |
| 781 |
int base, |
| 782 |
int minlen, |
| 783 |
int maxlen) |
| 784 |
{ |
| 785 |
uchr n; /* unsigned to avoid overflow misbehavior */ |
| 786 |
int len; |
| 787 |
chr c; |
| 788 |
int d; |
| 789 |
const uchr ub = (uchr) base; |
| 790 |
|
| 791 |
n = 0; |
| 792 |
for (len = 0; len < maxlen && !ATEOS(); len++) |
| 793 |
{ |
| 794 |
c = *v->now++; |
| 795 |
switch (c) |
| 796 |
{ |
| 797 |
case CHR('0'): |
| 798 |
case CHR('1'): |
| 799 |
case CHR('2'): |
| 800 |
case CHR('3'): |
| 801 |
case CHR('4'): |
| 802 |
case CHR('5'): |
| 803 |
case CHR('6'): |
| 804 |
case CHR('7'): |
| 805 |
case CHR('8'): |
| 806 |
case CHR('9'): |
| 807 |
d = DIGITVAL(c); |
| 808 |
break; |
| 809 |
case CHR('a'): |
| 810 |
case CHR('A'): |
| 811 |
d = 10; |
| 812 |
break; |
| 813 |
case CHR('b'): |
| 814 |
case CHR('B'): |
| 815 |
d = 11; |
| 816 |
break; |
| 817 |
case CHR('c'): |
| 818 |
case CHR('C'): |
| 819 |
d = 12; |
| 820 |
break; |
| 821 |
case CHR('d'): |
| 822 |
case CHR('D'): |
| 823 |
d = 13; |
| 824 |
break; |
| 825 |
case CHR('e'): |
| 826 |
case CHR('E'): |
| 827 |
d = 14; |
| 828 |
break; |
| 829 |
case CHR('f'): |
| 830 |
case CHR('F'): |
| 831 |
d = 15; |
| 832 |
break; |
| 833 |
default: |
| 834 |
v->now--; /* oops, not a digit at all */ |
| 835 |
d = -1; |
| 836 |
break; |
| 837 |
} |
| 838 |
|
| 839 |
if (d >= base) |
| 840 |
{ /* not a plausible digit */ |
| 841 |
v->now--; |
| 842 |
d = -1; |
| 843 |
} |
| 844 |
if (d < 0) |
| 845 |
break; /* NOTE BREAK OUT */ |
| 846 |
n = n * ub + (uchr) d; |
| 847 |
} |
| 848 |
if (len < minlen) |
| 849 |
ERR(REG_EESCAPE); |
| 850 |
|
| 851 |
return (chr) n; |
| 852 |
} |
| 853 |
|
| 854 |
/* |
| 855 |
* brenext - get next BRE token |
| 856 |
* |
| 857 |
* This is much like EREs except for all the stupid backslashes and the |
| 858 |
* context-dependency of some things. |
| 859 |
*/ |
| 860 |
static int /* 1 normal, 0 failure */ |
| 861 |
brenext(struct vars *v, |
| 862 |
chr c) |
| 863 |
{ |
| 864 |
switch (c) |
| 865 |
{ |
| 866 |
case CHR('*'): |
| 867 |
if (LASTTYPE(EMPTY) || LASTTYPE('(') || LASTTYPE('^')) |
| 868 |
RETV(PLAIN, c); |
| 869 |
RETV('*', 1); |
| 870 |
break; |
| 871 |
case CHR('['): |
| 872 |
if (HAVE(6) && *(v->now + 0) == CHR('[') && |
| 873 |
*(v->now + 1) == CHR(':') && |
| 874 |
(*(v->now + 2) == CHR('<') || |
| 875 |
*(v->now + 2) == CHR('>')) && |
| 876 |
*(v->now + 3) == CHR(':') && |
| 877 |
*(v->now + 4) == CHR(']') && |
| 878 |
*(v->now + 5) == CHR(']')) |
| 879 |
{ |
| 880 |
c = *(v->now + 2); |
| 881 |
v->now += 6; |
| 882 |
NOTE(REG_UNONPOSIX); |
| 883 |
RET((c == CHR('<')) ? '<' : '>'); |
| 884 |
} |
| 885 |
INTOCON(L_BRACK); |
| 886 |
if (NEXT1('^')) |
| 887 |
{ |
| 888 |
v->now++; |
| 889 |
RETV('[', 0); |
| 890 |
} |
| 891 |
RETV('[', 1); |
| 892 |
break; |
| 893 |
case CHR('.'): |
| 894 |
RET('.'); |
| 895 |
break; |
| 896 |
case CHR('^'): |
| 897 |
if (LASTTYPE(EMPTY)) |
| 898 |
RET('^'); |
| 899 |
if (LASTTYPE('(')) |
| 900 |
{ |
| 901 |
NOTE(REG_UUNSPEC); |
| 902 |
RET('^'); |
| 903 |
} |
| 904 |
RETV(PLAIN, c); |
| 905 |
break; |
| 906 |
case CHR('$'): |
| 907 |
if (v->cflags & REG_EXPANDED) |
| 908 |
skip(v); |
| 909 |
if (ATEOS()) |
| 910 |
RET('$'); |
| 911 |
if (NEXT2('\\', ')')) |
| 912 |
{ |
| 913 |
NOTE(REG_UUNSPEC); |
| 914 |
RET('$'); |
| 915 |
} |
| 916 |
RETV(PLAIN, c); |
| 917 |
break; |
| 918 |
case CHR('\\'): |
| 919 |
break; /* see below */ |
| 920 |
default: |
| 921 |
RETV(PLAIN, c); |
| 922 |
break; |
| 923 |
} |
| 924 |
|
| 925 |
assert(c == CHR('\\')); |
| 926 |
|
| 927 |
if (ATEOS()) |
| 928 |
FAILW(REG_EESCAPE); |
| 929 |
|
| 930 |
c = *v->now++; |
| 931 |
switch (c) |
| 932 |
{ |
| 933 |
case CHR('{'): |
| 934 |
INTOCON(L_BBND); |
| 935 |
NOTE(REG_UBOUNDS); |
| 936 |
RET('{'); |
| 937 |
break; |
| 938 |
case CHR('('): |
| 939 |
RETV('(', 1); |
| 940 |
break; |
| 941 |
case CHR(')'): |
| 942 |
RETV(')', c); |
| 943 |
break; |
| 944 |
case CHR('<'): |
| 945 |
NOTE(REG_UNONPOSIX); |
| 946 |
RET('<'); |
| 947 |
break; |
| 948 |
case CHR('>'): |
| 949 |
NOTE(REG_UNONPOSIX); |
| 950 |
RET('>'); |
| 951 |
break; |
| 952 |
case CHR('1'): |
| 953 |
case CHR('2'): |
| 954 |
case CHR('3'): |
| 955 |
case CHR('4'): |
| 956 |
case CHR('5'): |
| 957 |
case CHR('6'): |
| 958 |
case CHR('7'): |
| 959 |
case CHR('8'): |
| 960 |
case CHR('9'): |
| 961 |
NOTE(REG_UBACKREF); |
| 962 |
RETV(BACKREF, (chr) DIGITVAL(c)); |
| 963 |
break; |
| 964 |
default: |
| 965 |
if (iscalnum(c)) |
| 966 |
{ |
| 967 |
NOTE(REG_UBSALNUM); |
| 968 |
NOTE(REG_UUNSPEC); |
| 969 |
} |
| 970 |
RETV(PLAIN, c); |
| 971 |
break; |
| 972 |
} |
| 973 |
|
| 974 |
assert(NOTREACHED); |
| 975 |
return 0; |
| 976 |
} |
| 977 |
|
| 978 |
/* |
| 979 |
* skip - skip white space and comments in expanded form |
| 980 |
*/ |
| 981 |
static void |
| 982 |
skip(struct vars *v) |
| 983 |
{ |
| 984 |
const chr *start = v->now; |
| 985 |
|
| 986 |
assert(v->cflags & REG_EXPANDED); |
| 987 |
|
| 988 |
for (;;) |
| 989 |
{ |
| 990 |
while (!ATEOS() && iscspace(*v->now)) |
| 991 |
v->now++; |
| 992 |
if (ATEOS() || *v->now != CHR('#')) |
| 993 |
break; /* NOTE BREAK OUT */ |
| 994 |
assert(NEXT1('#')); |
| 995 |
while (!ATEOS() && *v->now != CHR('\n')) |
| 996 |
v->now++; |
| 997 |
/* leave the newline to be picked up by the iscspace loop */ |
| 998 |
} |
| 999 |
|
| 1000 |
if (v->now != start) |
| 1001 |
NOTE(REG_UNONPOSIX); |
| 1002 |
} |
| 1003 |
|
| 1004 |
/* |
| 1005 |
* newline - return the chr for a newline |
| 1006 |
* |
| 1007 |
* This helps confine use of CHR to this source file. |
| 1008 |
*/ |
| 1009 |
static chr |
| 1010 |
newline(void) |
| 1011 |
{ |
| 1012 |
return CHR('\n'); |
| 1013 |
} |
| 1014 |
|
| 1015 |
/* |
| 1016 |
* chrnamed - return the chr known by a given (chr string) name |
| 1017 |
* |
| 1018 |
* The code is a bit clumsy, but this routine gets only such specialized |
| 1019 |
* use that it hardly matters. |
| 1020 |
*/ |
| 1021 |
static chr |
| 1022 |
chrnamed(struct vars *v, |
| 1023 |
const chr *startp, /* start of name */ |
| 1024 |
const chr *endp, /* just past end of name */ |
| 1025 |
chr lastresort) /* what to return if name lookup fails */ |
| 1026 |
{ |
| 1027 |
chr c; |
| 1028 |
int errsave; |
| 1029 |
int e; |
| 1030 |
struct cvec *cv; |
| 1031 |
|
| 1032 |
errsave = v->err; |
| 1033 |
v->err = 0; |
| 1034 |
c = element(v, startp, endp); |
| 1035 |
e = v->err; |
| 1036 |
v->err = errsave; |
| 1037 |
|
| 1038 |
if (e != 0) |
| 1039 |
return lastresort; |
| 1040 |
|
| 1041 |
cv = range(v, c, c, 0); |
| 1042 |
if (cv->nchrs == 0) |
| 1043 |
return lastresort; |
| 1044 |
return cv->chrs[0]; |
| 1045 |
} |