1 /* $NetBSD: process.c,v 1.30 2001/10/02 03:03:02 atatat Exp $ */ 2 3 /*- 4 * Copyright (c) 1992 Diomidis Spinellis. 5 * Copyright (c) 1992, 1993, 1994 6 * The Regents of the University of California. All rights reserved. 7 * 8 * This code is derived from software contributed to Berkeley by 9 * Diomidis Spinellis of Imperial College, University of London. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. All advertising materials mentioning features or use of this software 20 * must display the following acknowledgement: 21 * This product includes software developed by the University of 22 * California, Berkeley and its contributors. 23 * 4. Neither the name of the University nor the names of its contributors 24 * may be used to endorse or promote products derived from this software 25 * without specific prior written permission. 26 * 27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 37 * SUCH DAMAGE. 38 */ 39 40 #include <sys/cdefs.h> 41 #ifndef lint 42 #if 0 43 static char sccsid[] = "@(#)process.c 8.6 (Berkeley) 4/20/94"; 44 #else 45 __RCSID("$NetBSD: process.c,v 1.30 2001/10/02 03:03:02 atatat Exp $"); 46 #endif 47 #endif /* not lint */ 48 49 #include <sys/types.h> 50 #include <sys/stat.h> 51 #include <sys/ioctl.h> 52 #include <sys/uio.h> 53 54 #include <ctype.h> 55 #include <errno.h> 56 #include <fcntl.h> 57 #include <limits.h> 58 #include <regex.h> 59 #include <stdio.h> 60 #include <stdlib.h> 61 #include <string.h> 62 #include <unistd.h> 63 64 #include "defs.h" 65 #include "extern.h" 66 67 static SPACE HS, PS, SS; 68 #define pd PS.deleted 69 #define ps PS.space 70 #define psl PS.len 71 #define hs HS.space 72 #define hsl HS.len 73 74 static inline int applies __P((struct s_command *)); 75 static void flush_appends __P((void)); 76 static void lputs __P((char *)); 77 static inline int regexec_e __P((regex_t *, const char *, int, int, size_t)); 78 static void regsub __P((SPACE *, char *, char *)); 79 static int substitute __P((struct s_command *)); 80 81 struct s_appends *appends; /* Array of pointers to strings to append. */ 82 static int appendx; /* Index into appends array. */ 83 int appendnum; /* Size of appends array. */ 84 85 static int lastaddr; /* Set by applies if last address of a range. */ 86 static int sdone; /* If any substitutes since last line input. */ 87 /* Iov structure for 'w' commands. */ 88 static regex_t *defpreg; 89 size_t maxnsub; 90 regmatch_t *match; 91 92 #define OUT(s) { fwrite(s, sizeof(u_char), psl, stdout); } 93 94 void 95 process() 96 { 97 struct s_command *cp; 98 SPACE tspace; 99 size_t len, oldpsl; 100 char *p; 101 102 oldpsl = 0; 103 for (linenum = 0; mf_fgets(&PS, REPLACE);) { 104 pd = 0; 105 top: 106 cp = prog; 107 redirect: 108 while (cp != NULL) { 109 if (!applies(cp)) { 110 cp = cp->next; 111 continue; 112 } 113 switch (cp->code) { 114 case '{': 115 cp = cp->u.c; 116 goto redirect; 117 case 'a': 118 if (appendx >= appendnum) 119 appends = xrealloc(appends, 120 sizeof(struct s_appends) * 121 (appendnum *= 2)); 122 appends[appendx].type = AP_STRING; 123 appends[appendx].s = cp->t; 124 appends[appendx].len = strlen(cp->t); 125 appendx++; 126 break; 127 case 'b': 128 cp = cp->u.c; 129 goto redirect; 130 case 'c': 131 pd = 1; 132 psl = 0; 133 if (cp->a2 == NULL || lastaddr) 134 (void)printf("%s", cp->t); 135 break; 136 case 'd': 137 pd = 1; 138 goto new; 139 case 'D': 140 if (psl == 0) 141 pd = 1; 142 if (pd) 143 goto new; 144 if ((p = memchr(ps, '\n', psl - 1)) == NULL) { 145 pd = 1; 146 goto new; 147 } else { 148 psl -= (p + 1) - ps; 149 memmove(ps, p + 1, psl); 150 goto top; 151 } 152 case 'g': 153 cspace(&PS, hs, hsl, REPLACE); 154 break; 155 case 'G': 156 if (hs == NULL) 157 cspace(&HS, "\n", 1, REPLACE); 158 cspace(&PS, hs, hsl, 0); 159 break; 160 case 'h': 161 cspace(&HS, ps, psl, REPLACE); 162 break; 163 case 'H': 164 cspace(&HS, ps, psl, 0); 165 break; 166 case 'i': 167 (void)printf("%s", cp->t); 168 break; 169 case 'l': 170 lputs(ps); 171 break; 172 case 'n': 173 if (!nflag && !pd) 174 OUT(ps) 175 flush_appends(); 176 if (!mf_fgets(&PS, REPLACE)) 177 exit(0); 178 pd = 0; 179 break; 180 case 'N': 181 flush_appends(); 182 if (!mf_fgets(&PS, 0)) { 183 if (!nflag && !pd) 184 OUT(ps) 185 exit(0); 186 } 187 break; 188 case 'p': 189 if (pd) 190 break; 191 OUT(ps) 192 break; 193 case 'P': 194 if (pd) 195 break; 196 if ((p = memchr(ps, '\n', psl - 1)) != NULL) { 197 oldpsl = psl; 198 psl = (p + 1) - ps; 199 } 200 OUT(ps) 201 if (p != NULL) 202 psl = oldpsl; 203 break; 204 case 'q': 205 if (!nflag && !pd) 206 OUT(ps) 207 flush_appends(); 208 exit(0); 209 case 'r': 210 if (appendx >= appendnum) 211 appends = xrealloc(appends, 212 sizeof(struct s_appends) * 213 (appendnum *= 2)); 214 appends[appendx].type = AP_FILE; 215 appends[appendx].s = cp->t; 216 appends[appendx].len = strlen(cp->t); 217 appendx++; 218 break; 219 case 's': 220 sdone |= substitute(cp); 221 break; 222 case 't': 223 if (sdone) { 224 sdone = 0; 225 cp = cp->u.c; 226 goto redirect; 227 } 228 break; 229 case 'w': 230 if (pd) 231 break; 232 if (cp->u.fd == -1 && (cp->u.fd = open(cp->t, 233 O_WRONLY|O_APPEND|O_CREAT|O_TRUNC, 234 DEFFILEMODE)) == -1) 235 err(FATAL, "%s: %s\n", 236 cp->t, strerror(errno)); 237 if (write(cp->u.fd, ps, psl) != psl) 238 err(FATAL, "%s: %s\n", 239 cp->t, strerror(errno)); 240 break; 241 case 'x': 242 if (hs == NULL) 243 cspace(&HS, "\n", 1, REPLACE); 244 tspace = PS; 245 PS = HS; 246 HS = tspace; 247 break; 248 case 'y': 249 if (pd) 250 break; 251 for (p = ps, len = psl; --len; ++p) 252 *p = cp->u.y[(int)*p]; 253 break; 254 case ':': 255 case '}': 256 break; 257 case '=': 258 (void)printf("%lu\n", linenum); 259 } 260 cp = cp->next; 261 } /* for all cp */ 262 263 new: if (!nflag && !pd) 264 OUT(ps) 265 flush_appends(); 266 } /* for all lines */ 267 } 268 269 /* 270 * TRUE if the address passed matches the current program state 271 * (lastline, linenumber, ps). 272 */ 273 #define MATCH(a) \ 274 (a)->type == AT_RE ? regexec_e((a)->u.r, ps, 0, 1, psl) : \ 275 (a)->type == AT_LINE ? linenum == (a)->u.l : lastline 276 277 /* 278 * Return TRUE if the command applies to the current line. Sets the inrange 279 * flag to process ranges. Interprets the non-select (``!'') flag. 280 */ 281 static inline int 282 applies(cp) 283 struct s_command *cp; 284 { 285 int r; 286 287 lastaddr = 0; 288 if (cp->a1 == NULL && cp->a2 == NULL) 289 r = 1; 290 else if (cp->a2) { 291 if (cp->inrange) { 292 if (MATCH(cp->a2)) { 293 cp->inrange = 0; 294 lastaddr = 1; 295 } 296 r = 1; 297 } else if (MATCH(cp->a1)) { 298 /* 299 * If the second address is a number less than or 300 * equal to the line number first selected, only 301 * one line shall be selected. 302 * -- POSIX 1003.2 303 */ 304 if (cp->a2->type == AT_LINE && 305 linenum >= cp->a2->u.l) 306 lastaddr = 1; 307 else 308 cp->inrange = 1; 309 r = 1; 310 } else 311 r = 0; 312 } else 313 r = MATCH(cp->a1); 314 return (cp->nonsel ? ! r : r); 315 } 316 317 /* 318 * substitute -- 319 * Do substitutions in the pattern space. Currently, we build a 320 * copy of the new pattern space in the substitute space structure 321 * and then swap them. 322 */ 323 static int 324 substitute(cp) 325 struct s_command *cp; 326 { 327 SPACE tspace; 328 regex_t *re; 329 size_t re_off, slen; 330 int lastempty, n; 331 char *s; 332 333 s = ps; 334 re = cp->u.s->re; 335 if (re == NULL) { 336 if (defpreg != NULL && cp->u.s->maxbref > defpreg->re_nsub) { 337 linenum = cp->u.s->linenum; 338 err(COMPILE, "\\%d not defined in the RE", 339 cp->u.s->maxbref); 340 } 341 } 342 if (!regexec_e(re, s, 0, 0, psl)) 343 return (0); 344 345 SS.len = 0; /* Clean substitute space. */ 346 slen = psl; 347 n = cp->u.s->n; 348 lastempty = 1; 349 350 switch (n) { 351 case 0: /* Global */ 352 do { 353 if (lastempty || match[0].rm_so != match[0].rm_eo) { 354 /* Locate start of replaced string. */ 355 re_off = match[0].rm_so; 356 /* Copy leading retained string. */ 357 cspace(&SS, s, re_off, APPEND); 358 /* Add in regular expression. */ 359 regsub(&SS, s, cp->u.s->new); 360 } 361 362 /* Move past this match. */ 363 if (match[0].rm_so != match[0].rm_eo) { 364 s += match[0].rm_eo; 365 slen -= match[0].rm_eo; 366 lastempty = 0; 367 } else { 368 if (match[0].rm_so == 0) 369 cspace(&SS, 370 s, match[0].rm_so + 1, APPEND); 371 else 372 cspace(&SS, 373 s + match[0].rm_so, 1, APPEND); 374 s += match[0].rm_so + 1; 375 slen -= match[0].rm_so + 1; 376 lastempty = 1; 377 } 378 } while (slen > 0 && regexec_e(re, s, REG_NOTBOL, 0, slen)); 379 /* Copy trailing retained string. */ 380 if (slen > 0) 381 cspace(&SS, s, slen, APPEND); 382 break; 383 default: /* Nth occurrence */ 384 while (--n) { 385 s += match[0].rm_eo; 386 slen -= match[0].rm_eo; 387 if (!regexec_e(re, s, REG_NOTBOL, 0, slen)) 388 return (0); 389 } 390 /* FALLTHROUGH */ 391 case 1: /* 1st occurrence */ 392 /* Locate start of replaced string. */ 393 re_off = match[0].rm_so + (s - ps); 394 /* Copy leading retained string. */ 395 cspace(&SS, ps, re_off, APPEND); 396 /* Add in regular expression. */ 397 regsub(&SS, s, cp->u.s->new); 398 /* Copy trailing retained string. */ 399 s += match[0].rm_eo; 400 slen -= match[0].rm_eo; 401 cspace(&SS, s, slen, APPEND); 402 break; 403 } 404 405 /* 406 * Swap the substitute space and the pattern space, and make sure 407 * that any leftover pointers into stdio memory get lost. 408 */ 409 tspace = PS; 410 PS = SS; 411 SS = tspace; 412 SS.space = SS.back; 413 414 /* Handle the 'p' flag. */ 415 if (cp->u.s->p) 416 OUT(ps) 417 418 /* Handle the 'w' flag. */ 419 if (cp->u.s->wfile && !pd) { 420 if (cp->u.s->wfd == -1 && (cp->u.s->wfd = open(cp->u.s->wfile, 421 O_WRONLY|O_APPEND|O_CREAT|O_TRUNC, DEFFILEMODE)) == -1) 422 err(FATAL, "%s: %s\n", cp->u.s->wfile, strerror(errno)); 423 if (write(cp->u.s->wfd, ps, psl) != psl) 424 err(FATAL, "%s: %s\n", cp->u.s->wfile, strerror(errno)); 425 } 426 return (1); 427 } 428 429 /* 430 * Flush append requests. Always called before reading a line, 431 * therefore it also resets the substitution done (sdone) flag. 432 */ 433 static void 434 flush_appends() 435 { 436 FILE *f; 437 int count, i; 438 char buf[8 * 1024]; 439 440 for (i = 0; i < appendx; i++) 441 switch (appends[i].type) { 442 case AP_STRING: 443 fwrite(appends[i].s, sizeof(char), appends[i].len, 444 stdout); 445 break; 446 case AP_FILE: 447 /* 448 * Read files probably shouldn't be cached. Since 449 * it's not an error to read a non-existent file, 450 * it's possible that another program is interacting 451 * with the sed script through the file system. It 452 * would be truly bizarre, but possible. It's probably 453 * not that big a performance win, anyhow. 454 */ 455 if ((f = fopen(appends[i].s, "r")) == NULL) 456 break; 457 while ((count = 458 fread(buf, sizeof(char), sizeof(buf), f)) > 0) 459 (void)fwrite(buf, sizeof(char), count, stdout); 460 (void)fclose(f); 461 break; 462 } 463 if (ferror(stdout)) 464 err(FATAL, "stdout: %s", strerror(errno ? errno : EIO)); 465 appendx = sdone = 0; 466 } 467 468 static void 469 lputs(s) 470 char *s; 471 { 472 int count; 473 char *escapes, *p; 474 struct winsize win; 475 static int termwidth = -1; 476 477 if (termwidth == -1) { 478 if ((p = getenv("COLUMNS")) != NULL) 479 termwidth = atoi(p); 480 else if (ioctl(STDOUT_FILENO, TIOCGWINSZ, &win) == 0 && 481 win.ws_col > 0) 482 termwidth = win.ws_col; 483 else 484 termwidth = 60; 485 } 486 for (count = 0; *s; ++s) { 487 if (count >= termwidth) { 488 (void)printf("\\\n"); 489 count = 0; 490 } 491 if (isascii((unsigned char)*s) && isprint((unsigned char)*s) && 492 *s != '\\') { 493 (void)putchar(*s); 494 count++; 495 } else { 496 escapes = "\\\a\b\f\n\r\t\v"; 497 (void)putchar('\\'); 498 if ((p = strchr(escapes, *s)) != NULL) { 499 (void)putchar("\\abfnrtv"[p - escapes]); 500 count += 2; 501 } else { 502 (void)printf("%03o", *(u_char *)s); 503 count += 4; 504 } 505 } 506 } 507 (void)putchar('$'); 508 (void)putchar('\n'); 509 if (ferror(stdout)) 510 err(FATAL, "stdout: %s", strerror(errno ? errno : EIO)); 511 } 512 513 static inline int 514 regexec_e(preg, string, eflags, nomatch, slen) 515 regex_t *preg; 516 const char *string; 517 int eflags, nomatch; 518 size_t slen; 519 { 520 int eval; 521 #ifndef REG_STARTEND 522 char *buf; 523 #endif 524 525 if (preg == NULL) { 526 if (defpreg == NULL) 527 err(FATAL, "first RE may not be empty"); 528 } else 529 defpreg = preg; 530 531 /* Set anchors, discounting trailing newline (if any). */ 532 if (slen > 0 && string[slen - 1] == '\n') 533 slen--; 534 535 #ifndef REG_STARTEND 536 if ((buf = malloc(slen + 1)) == NULL) 537 err(1, NULL); 538 (void)memcpy(buf, string, slen); 539 buf[slen] = '\0'; 540 eval = regexec(defpreg, buf, 541 nomatch ? 0 : maxnsub + 1, match, eflags); 542 free(buf); 543 #else 544 match[0].rm_so = 0; 545 match[0].rm_eo = slen; 546 eval = regexec(defpreg, string, 547 nomatch ? 0 : maxnsub + 1, match, eflags | REG_STARTEND); 548 #endif 549 switch(eval) { 550 case 0: 551 return (1); 552 case REG_NOMATCH: 553 return (0); 554 } 555 err(FATAL, "RE error: %s", strregerror(eval, defpreg)); 556 /* NOTREACHED */ 557 return (0); 558 } 559 560 /* 561 * regsub - perform substitutions after a regexp match 562 * Based on a routine by Henry Spencer 563 */ 564 static void 565 regsub(sp, string, src) 566 SPACE *sp; 567 char *string, *src; 568 { 569 int len, no; 570 char c, *dst; 571 572 #define NEEDSP(reqlen) \ 573 if (sp->len >= sp->blen - (reqlen) - 1) { \ 574 sp->blen += (reqlen) + 1024; \ 575 sp->space = sp->back = xrealloc(sp->back, sp->blen); \ 576 dst = sp->space + sp->len; \ 577 } 578 579 dst = sp->space + sp->len; 580 while ((c = *src++) != '\0') { 581 if (c == '&') 582 no = 0; 583 else if (c == '\\' && isdigit((unsigned char)*src)) 584 no = *src++ - '0'; 585 else 586 no = -1; 587 if (no < 0) { /* Ordinary character. */ 588 if (c == '\\' && (*src == '\\' || *src == '&')) 589 c = *src++; 590 NEEDSP(1); 591 *dst++ = c; 592 ++sp->len; 593 } else if (match[no].rm_so != -1 && match[no].rm_eo != -1) { 594 len = match[no].rm_eo - match[no].rm_so; 595 NEEDSP(len); 596 memmove(dst, string + match[no].rm_so, len); 597 dst += len; 598 sp->len += len; 599 } 600 } 601 NEEDSP(1); 602 *dst = '\0'; 603 } 604 605 /* 606 * aspace -- 607 * Append the source space to the destination space, allocating new 608 * space as necessary. 609 */ 610 void 611 cspace(sp, p, len, spflag) 612 SPACE *sp; 613 char *p; 614 size_t len; 615 enum e_spflag spflag; 616 { 617 size_t tlen; 618 619 /* Make sure SPACE has enough memory and ramp up quickly. */ 620 tlen = sp->len + len + 1; 621 if (tlen > sp->blen) { 622 sp->blen = tlen + 1024; 623 sp->space = sp->back = xrealloc(sp->back, sp->blen); 624 } 625 626 if (spflag == REPLACE) 627 sp->len = 0; 628 629 memmove(sp->space + sp->len, p, len); 630 631 sp->space[sp->len += len] = '\0'; 632 } 633 634 /* 635 * Close all cached opened files and report any errors 636 */ 637 void 638 cfclose(cp, end) 639 struct s_command *cp, *end; 640 { 641 642 for (; cp != end; cp = cp->next) 643 switch(cp->code) { 644 case 's': 645 if (cp->u.s->wfd != -1 && close(cp->u.s->wfd)) 646 err(FATAL, 647 "%s: %s", cp->u.s->wfile, strerror(errno)); 648 cp->u.s->wfd = -1; 649 break; 650 case 'w': 651 if (cp->u.fd != -1 && close(cp->u.fd)) 652 err(FATAL, "%s: %s", cp->t, strerror(errno)); 653 cp->u.fd = -1; 654 break; 655 case '{': 656 cfclose(cp->u.c, cp->next); 657 break; 658 } 659 } 660