| 1 |
/************************************************************************** |
| 2 |
* This program is Copyright (C) 1986-2002 by Jonathan Payne. JOVE is * |
| 3 |
* provided by Jonathan and Jovehacks without charge and without * |
| 4 |
* warranty. You may copy, modify, and/or distribute JOVE, provided that * |
| 5 |
* this notice is included in all the source files and documentation. * |
| 6 |
**************************************************************************/ |
| 7 |
|
| 8 |
/* search package */ |
| 9 |
|
| 10 |
#include "jove.h" |
| 11 |
#include "re.h" |
| 12 |
#include "jctype.h" |
| 13 |
#include "ask.h" |
| 14 |
#include "disp.h" |
| 15 |
#include "fmt.h" |
| 16 |
#include "marks.h" |
| 17 |
|
| 18 |
private jbool |
| 19 |
do_comp(struct RE_block *,int); |
| 20 |
|
| 21 |
char |
| 22 |
rep_search[128], /* replace search string */ |
| 23 |
rep_str[128]; /* contains replacement string */ |
| 24 |
|
| 25 |
jbool |
| 26 |
CaseIgnore = NO, /* VAR: ignore case in search */ |
| 27 |
WrapScan = NO; /* VAR: make searches wrap */ |
| 28 |
|
| 29 |
private ZXchar REpeekc; |
| 30 |
private const char *REptr; |
| 31 |
|
| 32 |
private ZXchar |
| 33 |
REgetc(void) |
| 34 |
{ |
| 35 |
ZXchar c; |
| 36 |
|
| 37 |
if ((c = REpeekc) != EOF) |
| 38 |
REpeekc = EOF; |
| 39 |
else if (*REptr != '\0') |
| 40 |
c = ZXC(*REptr++); |
| 41 |
else |
| 42 |
c = EOF; |
| 43 |
|
| 44 |
return c; |
| 45 |
} |
| 46 |
|
| 47 |
#define STAR 01 /* Match any number of last RE (ORed into other ops). */ |
| 48 |
|
| 49 |
#define AT_BOL 2 /* ^ */ |
| 50 |
#define AT_EOL 4 /* $ */ |
| 51 |
#define AT_BOW 6 /* \< */ |
| 52 |
#define AT_EOW 8 /* \> */ |
| 53 |
#define OPENP 10 /* \( {chunk number} */ |
| 54 |
#define CLOSEP 12 /* \) {chunk number} */ |
| 55 |
#define CURLYB 14 /* \{ {number of alt, alts } */ |
| 56 |
|
| 57 |
#define NOSTR 14 /* Codes <= NOSTR can't be *'d. */ |
| 58 |
|
| 59 |
#define ANYC (NOSTR+2) /* . */ |
| 60 |
#define NORMC (ANYC+2) /* normal chars {len, char...} */ |
| 61 |
#define CINDC (NORMC+2) /* case independent chars {len, char...} */ |
| 62 |
#define ONE_OF (CINDC+2) /* [xxx] {bitmask} */ |
| 63 |
#define NONE_OF (ONE_OF+2) /* [^xxx] {bitmask} */ |
| 64 |
#define BACKREF (NONE_OF+2) /* \# {chunk number} */ |
| 65 |
#define EOP (BACKREF+2) /* end of pattern */ |
| 66 |
|
| 67 |
#define CHAR_MASK ((1 << CHAR_BIT) - 1) /* byte mask, really */ |
| 68 |
#define ALT_LEN_LEN 2 /* an alt starts with a two-byte length */ |
| 69 |
#define ALT_LEN(p) (((p)[0] & CHAR_MASK) + (((p)[1] & CHAR_MASK) << CHAR_BIT)) |
| 70 |
|
| 71 |
/* ONE_OF/NONE_OF is represented as a bit vector. |
| 72 |
* These symbols parameterize the representation. |
| 73 |
*/ |
| 74 |
|
| 75 |
#define SETSIZE (NCHARS / CHAR_BIT) |
| 76 |
#define SETBYTE(c) ((c) / CHAR_BIT) |
| 77 |
#define SETBIT(c) (1 << ((c) % CHAR_BIT)) |
| 78 |
|
| 79 |
#define NPAR 10 /* [0-9] - 0th is the entire matched string, i.e. & */ |
| 80 |
private char *comp_ptr, |
| 81 |
**alt_p, |
| 82 |
**alt_endp; |
| 83 |
|
| 84 |
void |
| 85 |
REcompile(const char *pattern, jbool re, struct RE_block *re_blk) |
| 86 |
{ |
| 87 |
REptr = pattern; |
| 88 |
REpeekc = EOF; |
| 89 |
comp_ptr = re_blk->r_compbuf; |
| 90 |
alt_p = re_blk->r_alternates; |
| 91 |
alt_endp = alt_p + NALTS - 1; |
| 92 |
*alt_p++ = comp_ptr; |
| 93 |
re_blk->r_nparens = 0; |
| 94 |
(void) do_comp(re_blk, re ? OKAY_RE : NORM); |
| 95 |
*alt_p = NULL; |
| 96 |
|
| 97 |
re_blk->r_anchored = NO; |
| 98 |
re_blk->r_firstc = EOF; |
| 99 |
/* do a little post processing */ |
| 100 |
if (re_blk->r_alternates[1] == NULL) { |
| 101 |
char *p; |
| 102 |
|
| 103 |
p = re_blk->r_alternates[0]; |
| 104 |
for (;;) { |
| 105 |
switch (*p) { |
| 106 |
case OPENP: |
| 107 |
case CLOSEP: |
| 108 |
p += 2; |
| 109 |
continue; |
| 110 |
|
| 111 |
case AT_BOW: |
| 112 |
case AT_EOW: |
| 113 |
p += 1; |
| 114 |
continue; |
| 115 |
|
| 116 |
case AT_BOL: |
| 117 |
re_blk->r_anchored = YES; |
| 118 |
/* don't set firstc -- won't work */ |
| 119 |
break; |
| 120 |
|
| 121 |
case NORMC: |
| 122 |
case CINDC: |
| 123 |
re_blk->r_firstc = CharUpcase(p[2]); |
| 124 |
break; |
| 125 |
|
| 126 |
default: |
| 127 |
break; |
| 128 |
} |
| 129 |
break; |
| 130 |
} |
| 131 |
} |
| 132 |
} |
| 133 |
|
| 134 |
/* compile the pattern into an internal code */ |
| 135 |
|
| 136 |
private jbool |
| 137 |
do_comp(struct RE_block *re_blk, int kind) |
| 138 |
{ |
| 139 |
char *this_verb, |
| 140 |
*prev_verb, |
| 141 |
*start_p, |
| 142 |
*comp_endp; |
| 143 |
int parens[NPAR], |
| 144 |
*parenp, |
| 145 |
outer_max_paren = -1; |
| 146 |
ZXchar c; |
| 147 |
jbool done_cb = NO; |
| 148 |
|
| 149 |
parenp = parens; |
| 150 |
this_verb = NULL; |
| 151 |
comp_endp = &re_blk->r_compbuf[COMPSIZE - 6]; |
| 152 |
|
| 153 |
/* wrap the whole expression around (implied) parens */ |
| 154 |
if (kind != IN_CB) { |
| 155 |
if (re_blk->r_nparens >= NPAR) { |
| 156 |
complain("Too many ('s; max is %d.", NPAR); |
| 157 |
/* NOTREACHED */ |
| 158 |
} |
| 159 |
*comp_ptr++ = OPENP; |
| 160 |
*parenp++ = *comp_ptr++ = re_blk->r_nparens++; |
| 161 |
} |
| 162 |
|
| 163 |
start_p = comp_ptr; |
| 164 |
|
| 165 |
while ((c = REgetc()) != EOF) { |
| 166 |
if (comp_ptr > comp_endp) { |
| 167 |
toolong: |
| 168 |
complain("Search string too long/complex."); |
| 169 |
/* NOTREACHED */ |
| 170 |
} |
| 171 |
prev_verb = this_verb; |
| 172 |
this_verb = comp_ptr; |
| 173 |
|
| 174 |
/* The following test ought to be |
| 175 |
* kind == NORM && c != '\\' |
| 176 |
* but Jon likes to put ^, $, and \ in i-searches. |
| 177 |
* Don't tell him, but $ only sort of works. -- DHR |
| 178 |
*/ |
| 179 |
if (kind == NORM && strchr("^$\\", c) == NULL) |
| 180 |
goto defchar; |
| 181 |
switch (c) { |
| 182 |
case '\\': |
| 183 |
switch (c = REgetc()) { |
| 184 |
case EOF: |
| 185 |
complain("[Premature end of pattern]"); |
| 186 |
/*NOTREACHED*/ |
| 187 |
|
| 188 |
case '{': |
| 189 |
{ |
| 190 |
char *altcntp; /* alternate count */ |
| 191 |
int |
| 192 |
init_paren = re_blk->r_nparens, |
| 193 |
max_paren = -1; |
| 194 |
|
| 195 |
*comp_ptr++ = CURLYB; |
| 196 |
altcntp = comp_ptr; |
| 197 |
*comp_ptr++ = 0; /* initialize alt-count */ |
| 198 |
for (;;) { |
| 199 |
char *comp_len = comp_ptr; |
| 200 |
jbool done; |
| 201 |
long len; |
| 202 |
|
| 203 |
comp_ptr += ALT_LEN_LEN; |
| 204 |
re_blk->r_nparens = init_paren; |
| 205 |
done = do_comp(re_blk, IN_CB); |
| 206 |
|
| 207 |
/* We demand that each alternate has the same number |
| 208 |
* of parens because we currently have no mechanism to |
| 209 |
* set the matching strings to a meaningful default. |
| 210 |
*/ |
| 211 |
if (max_paren == -1) |
| 212 |
max_paren = re_blk->r_nparens; |
| 213 |
if (max_paren != re_blk->r_nparens) { |
| 214 |
complain("[each alternate must have the same number of \\( \\)]"); |
| 215 |
/* NOTREACHED */ |
| 216 |
} |
| 217 |
len = comp_ptr - comp_len; |
| 218 |
comp_len[0] = (char) len; /* truncate */ |
| 219 |
comp_len[1] = (char) (len >> CHAR_BIT); /* truncate */ |
| 220 |
(*altcntp) += 1; |
| 221 |
if (done) |
| 222 |
break; |
| 223 |
} |
| 224 |
break; |
| 225 |
} |
| 226 |
|
| 227 |
case '}': |
| 228 |
if (kind != IN_CB) { |
| 229 |
complain("Unexpected \\}."); |
| 230 |
/* NOTREACHED */ |
| 231 |
} |
| 232 |
done_cb = YES; |
| 233 |
goto outahere; |
| 234 |
|
| 235 |
case '(': |
| 236 |
if (re_blk->r_nparens >= NPAR) { |
| 237 |
complain("Too many ('s; max is %d.", NPAR); |
| 238 |
/* NOTREACHED */ |
| 239 |
} |
| 240 |
*comp_ptr++ = OPENP; |
| 241 |
*parenp++ = *comp_ptr++ = re_blk->r_nparens++; |
| 242 |
break; |
| 243 |
|
| 244 |
case ')': |
| 245 |
if (parenp == parens) { |
| 246 |
complain("Too many )'s."); |
| 247 |
/* NOTREACHED */ |
| 248 |
} |
| 249 |
*comp_ptr++ = CLOSEP; |
| 250 |
*comp_ptr++ = *--parenp; |
| 251 |
break; |
| 252 |
|
| 253 |
case '|': |
| 254 |
if (kind == IN_CB) |
| 255 |
goto outahere; |
| 256 |
if (alt_p >= alt_endp) { |
| 257 |
complain("Too many alternates; max %d.", NALTS); |
| 258 |
/* NOTREACHED */ |
| 259 |
} |
| 260 |
/* close off previous alternate */ |
| 261 |
*comp_ptr++ = CLOSEP; |
| 262 |
*comp_ptr++ = *--parenp; |
| 263 |
if (parenp != parens) { |
| 264 |
complain("Unmatched \\(."); |
| 265 |
/* NOTREACHED */ |
| 266 |
} |
| 267 |
*comp_ptr++ = EOP; |
| 268 |
|
| 269 |
/* We demand that each alternate has the same number |
| 270 |
* of parens because we currently have no mechanism to |
| 271 |
* set the matching strings to a meaningful default. |
| 272 |
*/ |
| 273 |
if (outer_max_paren == -1) |
| 274 |
outer_max_paren = re_blk->r_nparens; |
| 275 |
if (outer_max_paren != re_blk->r_nparens) { |
| 276 |
complain("[each alternate must have the same number of \\( \\)]"); |
| 277 |
/* NOTREACHED */ |
| 278 |
} |
| 279 |
/* start a new alt */ |
| 280 |
*alt_p++ = comp_ptr; |
| 281 |
re_blk->r_nparens = 0; |
| 282 |
*comp_ptr++ = OPENP; |
| 283 |
*parenp++ = *comp_ptr++ = re_blk->r_nparens++; |
| 284 |
start_p = comp_ptr; |
| 285 |
break; |
| 286 |
|
| 287 |
case '1': |
| 288 |
case '2': |
| 289 |
case '3': |
| 290 |
case '4': |
| 291 |
case '5': |
| 292 |
case '6': |
| 293 |
case '7': |
| 294 |
case '8': |
| 295 |
case '9': |
| 296 |
*comp_ptr++ = BACKREF; |
| 297 |
*comp_ptr++ = c - '0'; |
| 298 |
break; |
| 299 |
|
| 300 |
case '<': |
| 301 |
*comp_ptr++ = AT_BOW; |
| 302 |
break; |
| 303 |
|
| 304 |
case '>': |
| 305 |
*comp_ptr++ = AT_EOW; |
| 306 |
break; |
| 307 |
|
| 308 |
default: |
| 309 |
goto defchar; |
| 310 |
} |
| 311 |
break; |
| 312 |
|
| 313 |
case '.': |
| 314 |
*comp_ptr++ = ANYC; |
| 315 |
break; |
| 316 |
|
| 317 |
case '^': |
| 318 |
if (comp_ptr == start_p) { |
| 319 |
*comp_ptr++ = AT_BOL; |
| 320 |
break; |
| 321 |
} |
| 322 |
goto defchar; |
| 323 |
|
| 324 |
case '$': |
| 325 |
if ((REpeekc = REgetc()) != EOF && REpeekc != '\\') |
| 326 |
goto defchar; |
| 327 |
*comp_ptr++ = AT_EOL; |
| 328 |
break; |
| 329 |
|
| 330 |
case '[': |
| 331 |
*comp_ptr++ = ONE_OF; |
| 332 |
if (comp_ptr + SETSIZE >= comp_endp) |
| 333 |
goto toolong; |
| 334 |
byte_zero(comp_ptr, (size_t) SETSIZE); |
| 335 |
c = REgetc(); |
| 336 |
if (c == '^') { |
| 337 |
*this_verb = NONE_OF; |
| 338 |
c = REgetc(); |
| 339 |
} |
| 340 |
do { |
| 341 |
if (c == EOF) |
| 342 |
break; |
| 343 |
|
| 344 |
if (c == '\\') { |
| 345 |
c = REgetc(); |
| 346 |
if (c == EOF) |
| 347 |
break; |
| 348 |
} |
| 349 |
if ((REpeekc = REgetc()) == '-') { |
| 350 |
/* possibly a range */ |
| 351 |
ZXchar i = c; |
| 352 |
|
| 353 |
REpeekc = EOF; /* discard '-' */ |
| 354 |
c = REgetc(); |
| 355 |
if (c == EOF) |
| 356 |
break; |
| 357 |
|
| 358 |
if (c == ']') { |
| 359 |
/* not a range after all */ |
| 360 |
REpeekc = c; /* push back ']' */ |
| 361 |
c = '-'; /* recycle '-' */ |
| 362 |
comp_ptr[SETBYTE(i)] |= SETBIT(i); /* handle initial char */ |
| 363 |
} else { |
| 364 |
/* really a range: add members up to c */ |
| 365 |
if (c == '\\') { |
| 366 |
c = REgetc(); |
| 367 |
if (c == EOF) |
| 368 |
break; |
| 369 |
} |
| 370 |
while (i < c) { |
| 371 |
comp_ptr[SETBYTE(i)] |= SETBIT(i); |
| 372 |
i += 1; |
| 373 |
} |
| 374 |
} |
| 375 |
} |
| 376 |
comp_ptr[SETBYTE(c)] |= SETBIT(c); |
| 377 |
c = REgetc(); |
| 378 |
} while (c != ']'); |
| 379 |
if (c == EOF) { |
| 380 |
complain("Missing ]."); |
| 381 |
/* NOTREACHED */ |
| 382 |
} |
| 383 |
comp_ptr += SETSIZE; |
| 384 |
break; |
| 385 |
|
| 386 |
case '*': |
| 387 |
if (prev_verb == NULL || *prev_verb <= NOSTR || (*prev_verb&STAR)!=0) |
| 388 |
goto defchar; |
| 389 |
|
| 390 |
if (*prev_verb == NORMC || *prev_verb == CINDC) { |
| 391 |
char lastc = comp_ptr[-1]; |
| 392 |
|
| 393 |
/* The * operator applies only to the |
| 394 |
* previous character. Since we were |
| 395 |
* building a string-matching command |
| 396 |
* (NORMC or CINDC), we must split it |
| 397 |
* up and work with the last character. |
| 398 |
* |
| 399 |
* Note that the STARed versions of these |
| 400 |
* commands do not operate on strings, and |
| 401 |
* so do not need or have character counts. |
| 402 |
*/ |
| 403 |
|
| 404 |
if (prev_verb[1] == 1) { |
| 405 |
/* Only one char in string: |
| 406 |
* delete old command. |
| 407 |
*/ |
| 408 |
this_verb = prev_verb; |
| 409 |
} else { |
| 410 |
/* Several chars in string: |
| 411 |
* strip off the last. |
| 412 |
* New verb is derived from old. |
| 413 |
*/ |
| 414 |
prev_verb[1] -= 1; |
| 415 |
this_verb -= 1; |
| 416 |
*this_verb = *prev_verb; |
| 417 |
} |
| 418 |
comp_ptr = this_verb + 1; |
| 419 |
*comp_ptr++ = lastc; |
| 420 |
} else { |
| 421 |
/* This command is just the previous one, |
| 422 |
* whose verb we will modify. |
| 423 |
*/ |
| 424 |
this_verb = prev_verb; |
| 425 |
} |
| 426 |
*this_verb |= STAR; |
| 427 |
break; |
| 428 |
default: |
| 429 |
defchar: |
| 430 |
if (prev_verb == NULL |
| 431 |
|| !(*prev_verb == NORMC || *prev_verb == CINDC)) |
| 432 |
{ |
| 433 |
/* create new string command */ |
| 434 |
*comp_ptr++ = (CaseIgnore) ? CINDC : NORMC; |
| 435 |
*comp_ptr++ = 0; |
| 436 |
} else { |
| 437 |
/* merge this into previous string command */ |
| 438 |
this_verb = prev_verb; |
| 439 |
} |
| 440 |
this_verb[1] += 1; |
| 441 |
*comp_ptr++ = c; |
| 442 |
break; |
| 443 |
} |
| 444 |
} |
| 445 |
outahere: |
| 446 |
|
| 447 |
/* End of pattern, let's do some error checking. */ |
| 448 |
if (kind != IN_CB) { |
| 449 |
*comp_ptr++ = CLOSEP; |
| 450 |
*comp_ptr++ = *--parenp; |
| 451 |
} |
| 452 |
if (parenp != parens) { |
| 453 |
complain("Unmatched \\(."); |
| 454 |
/* NOTREACHED */ |
| 455 |
} |
| 456 |
if (kind == IN_CB && c == EOF) { /* end of pattern with missing \}. */ |
| 457 |
complain("Missing \\}."); |
| 458 |
/* NOTREACHED */ |
| 459 |
} |
| 460 |
*comp_ptr++ = EOP; |
| 461 |
|
| 462 |
/* We demand that each alternate has the same number |
| 463 |
* of parens because we currently have no mechanism to |
| 464 |
* set the matching strings to a meaningful default. |
| 465 |
*/ |
| 466 |
if (outer_max_paren != -1 && outer_max_paren != re_blk->r_nparens) { |
| 467 |
complain("[each alternate must have the same number of \\( \\)]"); |
| 468 |
/* NOTREACHED */ |
| 469 |
} |
| 470 |
return done_cb; |
| 471 |
} |
| 472 |
|
| 473 |
private char *pstrtlst[NPAR], /* index into re_blk->r_lbuf */ |
| 474 |
*pendlst[NPAR], |
| 475 |
*REbolp, /* begining-of-line pointer */ |
| 476 |
*locrater, /* roof of last substitution */ |
| 477 |
*loc1, /* start of matched text */ |
| 478 |
*loc2; /* roof of matched text */ |
| 479 |
|
| 480 |
int REbom, /* beginning and end columns of match */ |
| 481 |
REeom, |
| 482 |
REdelta; /* increase in line length due to last re_dosub */ |
| 483 |
|
| 484 |
private jbool |
| 485 |
backref(int n, char *linep) |
| 486 |
{ |
| 487 |
char *backsp = pstrtlst[n], |
| 488 |
*backep = pendlst[n]; |
| 489 |
|
| 490 |
while (*backsp++ == *linep++) |
| 491 |
if (backsp >= backep) |
| 492 |
return YES; |
| 493 |
return NO; |
| 494 |
} |
| 495 |
|
| 496 |
private jbool |
| 497 |
member(char *comp_ptr, ZXchar c, jbool af) |
| 498 |
{ |
| 499 |
return c != '\0' && ((comp_ptr[SETBYTE(c)] & SETBIT(c))? af : !af); |
| 500 |
} |
| 501 |
|
| 502 |
private jbool |
| 503 |
REmatch(char *linep, char *comp_ptr) |
| 504 |
{ |
| 505 |
char *first_p; |
| 506 |
int n; |
| 507 |
|
| 508 |
for (;;) switch (*comp_ptr++) { |
| 509 |
case NORMC: |
| 510 |
n = *comp_ptr++; |
| 511 |
while (--n >= 0) |
| 512 |
if (*linep++ != *comp_ptr++) |
| 513 |
return NO; |
| 514 |
continue; |
| 515 |
|
| 516 |
case CINDC: /* case independent comparison */ |
| 517 |
n = *comp_ptr++; |
| 518 |
while (--n >= 0) |
| 519 |
if (!cind_eq(*linep++, *comp_ptr++)) |
| 520 |
return NO; |
| 521 |
continue; |
| 522 |
|
| 523 |
case EOP: |
| 524 |
loc2 = linep; |
| 525 |
REeom = (loc2 - REbolp); |
| 526 |
return YES; /* Success! */ |
| 527 |
|
| 528 |
case AT_BOL: |
| 529 |
if (linep == REbolp && linep != locrater) |
| 530 |
continue; |
| 531 |
|
| 532 |
return NO; |
| 533 |
|
| 534 |
case AT_EOL: |
| 535 |
if (*linep == '\0') |
| 536 |
continue; |
| 537 |
|
| 538 |
return NO; |
| 539 |
|
| 540 |
case ANYC: |
| 541 |
if (*linep++ != '\0') |
| 542 |
continue; |
| 543 |
|
| 544 |
return NO; |
| 545 |
|
| 546 |
case AT_BOW: |
| 547 |
if (linep != locrater && jisident(*linep) |
| 548 |
&& (linep == REbolp || !jisident(linep[-1]))) |
| 549 |
continue; |
| 550 |
|
| 551 |
return NO; |
| 552 |
|
| 553 |
case AT_EOW: |
| 554 |
if (linep != locrater && (*linep == '\0' || !jisident(*linep)) |
| 555 |
&& (linep != REbolp && jisident(linep[-1]))) |
| 556 |
continue; |
| 557 |
|
| 558 |
return NO; |
| 559 |
|
| 560 |
case ONE_OF: |
| 561 |
case NONE_OF: |
| 562 |
if (member(comp_ptr, ZXC(*linep++), comp_ptr[-1] == ONE_OF)) { |
| 563 |
comp_ptr += SETSIZE; |
| 564 |
continue; |
| 565 |
} |
| 566 |
return NO; |
| 567 |
|
| 568 |
case OPENP: |
| 569 |
pstrtlst[(int) *comp_ptr++] = linep; |
| 570 |
continue; |
| 571 |
|
| 572 |
case CLOSEP: |
| 573 |
pendlst[(int) *comp_ptr++] = linep; |
| 574 |
continue; |
| 575 |
|
| 576 |
case BACKREF: |
| 577 |
if (pstrtlst[n = *comp_ptr++] == NULL) { |
| 578 |
s_mess("\\%d was not specified.", n + 1); |
| 579 |
} else if (backref(n, linep)) { |
| 580 |
linep += pendlst[n] - pstrtlst[n]; |
| 581 |
continue; |
| 582 |
} |
| 583 |
return NO; |
| 584 |
|
| 585 |
case CURLYB: |
| 586 |
{ |
| 587 |
int altcnt = *comp_ptr++; |
| 588 |
jbool any = NO; |
| 589 |
|
| 590 |
while (--altcnt >= 0) { |
| 591 |
if (!any) |
| 592 |
any = REmatch(linep, comp_ptr + ALT_LEN_LEN); |
| 593 |
comp_ptr += ALT_LEN(comp_ptr); |
| 594 |
} |
| 595 |
if (!any) |
| 596 |
return NO; |
| 597 |
|
| 598 |
linep = loc2; |
| 599 |
continue; |
| 600 |
} |
| 601 |
|
| 602 |
case ANYC | STAR: |
| 603 |
first_p = linep; |
| 604 |
do {} while (*linep++ != '\0'); |
| 605 |
goto star; |
| 606 |
|
| 607 |
case NORMC | STAR: |
| 608 |
first_p = linep; |
| 609 |
do {} while (*comp_ptr == *linep++); |
| 610 |
comp_ptr += 1; |
| 611 |
goto star; |
| 612 |
|
| 613 |
case CINDC | STAR: |
| 614 |
first_p = linep; |
| 615 |
do {} while (cind_eq(*comp_ptr, *linep++)); |
| 616 |
comp_ptr += 1; |
| 617 |
goto star; |
| 618 |
|
| 619 |
case ONE_OF | STAR: |
| 620 |
case NONE_OF | STAR: |
| 621 |
first_p = linep; |
| 622 |
do {} while (member(comp_ptr, ZXC(*linep++), comp_ptr[-1] == (ONE_OF | STAR))); |
| 623 |
comp_ptr += SETSIZE; |
| 624 |
/* fall through */ |
| 625 |
star: |
| 626 |
/* linep points *after* first unmatched char. |
| 627 |
* first_p points at where starred element started matching. |
| 628 |
*/ |
| 629 |
while (--linep > first_p) { |
| 630 |
if ((*comp_ptr != NORMC || *linep == comp_ptr[2]) |
| 631 |
&& REmatch(linep, comp_ptr)) |
| 632 |
return YES; |
| 633 |
} |
| 634 |
continue; |
| 635 |
|
| 636 |
case BACKREF | STAR: |
| 637 |
first_p = linep; |
| 638 |
n = *comp_ptr++; |
| 639 |
while (backref(n, linep)) |
| 640 |
linep += pendlst[n] - pstrtlst[n]; |
| 641 |
while (linep > first_p) { |
| 642 |
if (REmatch(linep, comp_ptr)) |
| 643 |
return YES; |
| 644 |
|
| 645 |
linep -= pendlst[n] - pstrtlst[n]; |
| 646 |
} |
| 647 |
continue; |
| 648 |
|
| 649 |
default: |
| 650 |
complain("RE error match (%d).", comp_ptr[-1]); |
| 651 |
/* NOTREACHED */ |
| 652 |
} |
| 653 |
/* NOTREACHED */ |
| 654 |
} |
| 655 |
|
| 656 |
private void |
| 657 |
REreset(void) |
| 658 |
{ |
| 659 |
int i; |
| 660 |
|
| 661 |
for (i = 0; i < NPAR; i++) |
| 662 |
pstrtlst[i] = pendlst[i] = NULL; |
| 663 |
} |
| 664 |
|
| 665 |
/* Index LINE at OFFSET. If lbuf_okay is YES it's okay to use linebuf |
| 666 |
* if LINE is the current line. This should save lots of time in things |
| 667 |
* like paren matching in LISP mode. Saves all that copying from linebuf |
| 668 |
* to a local buffer. substitute() is the guy who calls re_lindex with |
| 669 |
* lbuf_okay as NO, since the substitution gets placed in linebuf ... |
| 670 |
* doesn't work too well when the source and destination strings are the |
| 671 |
* same. I hate all these arguments! |
| 672 |
* |
| 673 |
* This code is cumbersome, repetetive for reasons of efficiency. Fast |
| 674 |
* search is a must as far as I am concerned. |
| 675 |
*/ |
| 676 |
|
| 677 |
jbool |
| 678 |
re_lindex(LinePtr line, int offset, int dir, |
| 679 |
struct RE_block *re_blk, jbool lbuf_okay, |
| 680 |
int crater) /* offset of previous substitute (or -1). */ |
| 681 |
{ |
| 682 |
char *p; |
| 683 |
ZXchar firstc = re_blk->r_firstc; |
| 684 |
jbool anchored = re_blk->r_anchored; |
| 685 |
char **alts = re_blk->r_alternates; |
| 686 |
|
| 687 |
REreset(); |
| 688 |
if (lbuf_okay) { |
| 689 |
REbolp = lbptr(line); |
| 690 |
if (offset == -1) |
| 691 |
offset = strlen(REbolp); /* arg! */ |
| 692 |
} else { |
| 693 |
REbolp = ltobuf(line, re_blk->r_lbuf); |
| 694 |
if (offset == -1) { /* Reverse search, find end of line. */ |
| 695 |
offset = Jr_Len; /* Just Read Len. */ |
| 696 |
} |
| 697 |
} |
| 698 |
|
| 699 |
if (anchored) { |
| 700 |
if (dir == FORWARD) { |
| 701 |
if (offset != 0 || crater != -1) |
| 702 |
return NO; |
| 703 |
} else { |
| 704 |
offset = 0; |
| 705 |
} |
| 706 |
} |
| 707 |
|
| 708 |
p = REbolp + offset; |
| 709 |
locrater = REbolp + crater; |
| 710 |
|
| 711 |
if (firstc != EOF) { |
| 712 |
char *first_alt = *alts; |
| 713 |
|
| 714 |
if (dir == FORWARD) { |
| 715 |
while (CharUpcase(*p) != firstc || !REmatch(p, first_alt)) |
| 716 |
if (*p++ == '\0') |
| 717 |
return NO; |
| 718 |
} else { |
| 719 |
while (CharUpcase(*p) != firstc || !REmatch(p, first_alt)) |
| 720 |
if (--p < REbolp) |
| 721 |
return NO; |
| 722 |
} |
| 723 |
} else { |
| 724 |
for (;;) { |
| 725 |
char **altp = alts; |
| 726 |
|
| 727 |
while (*altp != NULL) |
| 728 |
if (REmatch(p, *altp++)) |
| 729 |
goto success; |
| 730 |
if (anchored |
| 731 |
|| (dir == FORWARD ? *p++ == '\0' : --p < REbolp)) |
| 732 |
return NO; |
| 733 |
} |
| 734 |
success:; |
| 735 |
} |
| 736 |
loc1 = p; |
| 737 |
REbom = loc1 - REbolp; |
| 738 |
|
| 739 |
return YES; |
| 740 |
} |
| 741 |
|
| 742 |
jbool okay_wrap = NO; /* Do a wrap search ... not when we're |
| 743 |
parsing errors ... */ |
| 744 |
|
| 745 |
Bufpos * |
| 746 |
dosearch(const char *pattern, int dir, jbool re) |
| 747 |
{ |
| 748 |
Bufpos *pos; |
| 749 |
struct RE_block re_blk; /* global re-compiled buffer */ |
| 750 |
|
| 751 |
if (bobp() && eobp()) /* Can't match! There's no buffer. */ |
| 752 |
return NULL; |
| 753 |
|
| 754 |
REcompile(pattern, re, &re_blk); |
| 755 |
|
| 756 |
pos = docompiled(dir, &re_blk); |
| 757 |
return pos; |
| 758 |
} |
| 759 |
|
| 760 |
Bufpos * |
| 761 |
docompiled(int dir, struct RE_block *re_blk) |
| 762 |
{ |
| 763 |
static Bufpos |
| 764 |
ret; |
| 765 |
LinePtr lp; |
| 766 |
int offset; |
| 767 |
jbool we_wrapped = NO; |
| 768 |
|
| 769 |
lsave(); |
| 770 |
/* Search now lsave()'s so it doesn't make any assumptions on |
| 771 |
* whether the the contents of curline/curchar are in linebuf. |
| 772 |
* Nowhere does search write all over linebuf. However, we have to |
| 773 |
* be careful about what calls we make here, because many of them |
| 774 |
* assume (and rightly so) that curline is in linebuf. |
| 775 |
*/ |
| 776 |
lp = curline; |
| 777 |
offset = curchar; |
| 778 |
if (dir == BACKWARD) { |
| 779 |
if (bobp()) { |
| 780 |
if (okay_wrap && WrapScan) |
| 781 |
goto doit; |
| 782 |
|
| 783 |
return NULL; |
| 784 |
} |
| 785 |
/* here we simulate BackChar() */ |
| 786 |
if (bolp()) { |
| 787 |
lp = lp->l_prev; |
| 788 |
offset = length(lp); |
| 789 |
} else { |
| 790 |
offset -= 1; |
| 791 |
} |
| 792 |
} else if (dir==FORWARD && lbptr(lp)[offset]=='\0' && !lastp(lp)) { |
| 793 |
lp = lp->l_next; |
| 794 |
offset = 0; |
| 795 |
} |
| 796 |
|
| 797 |
do { |
| 798 |
if (re_lindex(lp, offset, dir, re_blk, YES, -1)) |
| 799 |
break; |
| 800 |
doit: |
| 801 |
lp = (dir == FORWARD) ? lp->l_next : lp->l_prev; |
| 802 |
if (lp == NULL) { |
| 803 |
if (okay_wrap && WrapScan) { |
| 804 |
lp = (dir == FORWARD) |
| 805 |
? curbuf->b_first : curbuf->b_last; |
| 806 |
we_wrapped = YES; |
| 807 |
} else |
| 808 |
break; |
| 809 |
} |
| 810 |
if (dir == FORWARD) |
| 811 |
offset = 0; |
| 812 |
else |
| 813 |
offset = -1; /* signals re_lindex ... */ |
| 814 |
} while (lp != curline); |
| 815 |
|
| 816 |
if (lp == curline && we_wrapped) |
| 817 |
lp = NULL; |
| 818 |
if (lp == NULL) |
| 819 |
return NULL; |
| 820 |
|
| 821 |
ret.p_line = lp; |
| 822 |
ret.p_char = (dir == FORWARD) ? REeom : REbom; |
| 823 |
return &ret; |
| 824 |
} |
| 825 |
|
| 826 |
private char * |
| 827 |
insert(char *off, char *endp, int which) |
| 828 |
{ |
| 829 |
char *pp = pstrtlst[which]; |
| 830 |
int n; |
| 831 |
|
| 832 |
if (pp == NULL) { |
| 833 |
complain("\\%d not defined", which); |
| 834 |
/* NOTREACHED */ |
| 835 |
} |
| 836 |
n = pendlst[which] - pp; |
| 837 |
|
| 838 |
/* note: ensure space will be left for a NUL */ |
| 839 |
if (off + n >= endp) { |
| 840 |
len_error(JMP_ERROR); |
| 841 |
/* NOTREACHED */ |
| 842 |
} |
| 843 |
while (--n >= 0) |
| 844 |
*off++ = *pp++; |
| 845 |
return off; |
| 846 |
} |
| 847 |
|
| 848 |
/* Perform the substitution. If DELP is YES the matched string is |
| 849 |
* deleted, i.e., the substitution string is not inserted. |
| 850 |
*/ |
| 851 |
void |
| 852 |
re_dosub(struct RE_block *re_blk, char *tobuf, jbool delp) |
| 853 |
{ |
| 854 |
char *tp, |
| 855 |
*rp, |
| 856 |
*endp; |
| 857 |
|
| 858 |
tp = tobuf; |
| 859 |
endp = tp + LBSIZE; |
| 860 |
rp = re_blk->r_lbuf; |
| 861 |
|
| 862 |
while (rp < loc1) |
| 863 |
*tp++ = *rp++; |
| 864 |
|
| 865 |
if (!delp) { |
| 866 |
char c; |
| 867 |
|
| 868 |
rp = rep_str; |
| 869 |
while ((c = *rp++) != '\0') { |
| 870 |
if (c == '\\') { |
| 871 |
c = *rp++; |
| 872 |
if (c >= '0' && c < re_blk->r_nparens + '0') { |
| 873 |
tp = insert(tp, endp, c - '0'); |
| 874 |
continue; |
| 875 |
} |
| 876 |
if (c == '\0') { |
| 877 |
/* treat \ at the end as if it were \\ */ |
| 878 |
c = '\\'; |
| 879 |
rp--; /* be sure to hit again */ |
| 880 |
} |
| 881 |
} |
| 882 |
*tp++ = c; |
| 883 |
if (tp >= endp) { |
| 884 |
len_error(JMP_ERROR); |
| 885 |
/* NOTREACHED */ |
| 886 |
} |
| 887 |
} |
| 888 |
} |
| 889 |
rp = loc2; |
| 890 |
REdelta = -REeom; |
| 891 |
REeom = tp - tobuf; |
| 892 |
REdelta += REeom; |
| 893 |
if (loc1==rp && *rp!='\0') { |
| 894 |
/* Skip an extra character if the matched text was a null |
| 895 |
* string, but don't skip over the end of line. This is to |
| 896 |
* prevent an infinite number of replacements in the same |
| 897 |
* position, e.g., replace "^" with "". |
| 898 |
*/ |
| 899 |
REeom += 1; |
| 900 |
} |
| 901 |
loc2 = re_blk->r_lbuf + REeom; |
| 902 |
while ((*tp++ = *rp++) != '\0') { |
| 903 |
if (tp >= endp) { |
| 904 |
len_error(JMP_ERROR); |
| 905 |
/* NOTREACHED */ |
| 906 |
} |
| 907 |
} |
| 908 |
} |
| 909 |
|
| 910 |
void |
| 911 |
putmatch(int which, char *buf, size_t size) |
| 912 |
{ |
| 913 |
*(insert(buf, buf + size, which)) = '\0'; |
| 914 |
} |
| 915 |
|
| 916 |
void |
| 917 |
RErecur(void) |
| 918 |
{ |
| 919 |
char repbuf[sizeof rep_str]; |
| 920 |
Mark *m = MakeMark(curline, REbom); |
| 921 |
|
| 922 |
message("Type ^X ^C to continue with query replace."); |
| 923 |
|
| 924 |
byte_copy(rep_str, repbuf, sizeof rep_str); |
| 925 |
Recur(); |
| 926 |
byte_copy(repbuf, rep_str, sizeof rep_str); |
| 927 |
if (!is_an_arg()) |
| 928 |
ToMark(m); |
| 929 |
DelMark(m); |
| 930 |
} |
| 931 |
|
| 932 |
/* Do we match PATTERN at OFFSET in BUF? */ |
| 933 |
|
| 934 |
jbool |
| 935 |
LookingAt(const char *pattern, char *buf, int offset) |
| 936 |
{ |
| 937 |
struct RE_block re_blk; |
| 938 |
char **alt = re_blk.r_alternates; |
| 939 |
|
| 940 |
REcompile(pattern, YES, &re_blk); |
| 941 |
REreset(); |
| 942 |
locrater = NULL; |
| 943 |
REbolp = buf; |
| 944 |
|
| 945 |
while (*alt) |
| 946 |
if (REmatch(buf + offset, *alt++)) |
| 947 |
return YES; |
| 948 |
return NO; |
| 949 |
} |
| 950 |
|
| 951 |
jbool |
| 952 |
look_at(char *expr) |
| 953 |
{ |
| 954 |
struct RE_block re_blk; |
| 955 |
|
| 956 |
REcompile(expr, NO, &re_blk); |
| 957 |
REreset(); |
| 958 |
locrater = NULL; |
| 959 |
REbolp = linebuf; |
| 960 |
return REmatch(linebuf + curchar, re_blk.r_alternates[0]); |
| 961 |
} |