1 /*- 2 * Copyright (c) 1992 Diomidis Spinellis. 3 * Copyright (c) 1992, 1993, 1994 4 * The Regents of the University of California. All rights reserved. 5 * 6 * This code is derived from software contributed to Berkeley by 7 * Diomidis Spinellis of Imperial College, University of London. 8 * 9 * %sccs.include.redist.c% 10 */ 11 12 #ifndef lint 13 static char sccsid[] = "@(#)process.c 8.5 (Berkeley) 04/20/94"; 14 #endif /* not lint */ 15 16 #include <sys/types.h> 17 #include <sys/stat.h> 18 #include <sys/ioctl.h> 19 #include <sys/uio.h> 20 21 #include <ctype.h> 22 #include <errno.h> 23 #include <fcntl.h> 24 #include <limits.h> 25 #include <regex.h> 26 #include <stdio.h> 27 #include <stdlib.h> 28 #include <string.h> 29 #include <unistd.h> 30 31 #include "defs.h" 32 #include "extern.h" 33 34 static SPACE HS, PS, SS; 35 #define pd PS.deleted 36 #define ps PS.space 37 #define psl PS.len 38 #define hs HS.space 39 #define hsl HS.len 40 41 static inline int applies __P((struct s_command *)); 42 static void flush_appends __P((void)); 43 static void lputs __P((char *)); 44 static inline int regexec_e __P((regex_t *, const char *, int, int, size_t)); 45 static void regsub __P((SPACE *, char *, char *)); 46 static int substitute __P((struct s_command *)); 47 48 struct s_appends *appends; /* Array of pointers to strings to append. */ 49 static int appendx; /* Index into appends array. */ 50 int appendnum; /* Size of appends array. */ 51 52 static int lastaddr; /* Set by applies if last address of a range. */ 53 static int sdone; /* If any substitutes since last line input. */ 54 /* Iov structure for 'w' commands. */ 55 static regex_t *defpreg; 56 size_t maxnsub; 57 regmatch_t *match; 58 59 #define OUT(s) { fwrite(s, sizeof(u_char), psl, stdout); } 60 61 void 62 process() 63 { 64 struct s_command *cp; 65 SPACE tspace; 66 size_t len; 67 int r; 68 char oldc, *p; 69 70 for (linenum = 0; mf_fgets(&PS, REPLACE);) { 71 pd = 0; 72 cp = prog; 73 redirect: 74 while (cp != NULL) { 75 if (!applies(cp)) { 76 cp = cp->next; 77 continue; 78 } 79 switch (cp->code) { 80 case '{': 81 cp = cp->u.c; 82 goto redirect; 83 case 'a': 84 if (appendx >= appendnum) 85 appends = xrealloc(appends, 86 sizeof(struct s_appends) * 87 (appendnum *= 2)); 88 appends[appendx].type = AP_STRING; 89 appends[appendx].s = cp->t; 90 appends[appendx].len = strlen(cp->t); 91 appendx++; 92 break; 93 case 'b': 94 cp = cp->u.c; 95 goto redirect; 96 case 'c': 97 pd = 1; 98 psl = 0; 99 if (cp->a2 == NULL || lastaddr) 100 (void)printf("%s", cp->t); 101 break; 102 case 'd': 103 pd = 1; 104 goto new; 105 case 'D': 106 if (pd) 107 goto new; 108 if ((p = memchr(ps, '\n', psl)) == NULL) 109 pd = 1; 110 else { 111 psl -= (p - ps) + 1; 112 memmove(ps, p + 1, psl); 113 } 114 goto new; 115 case 'g': 116 cspace(&PS, hs, hsl, REPLACE); 117 break; 118 case 'G': 119 cspace(&PS, hs, hsl, 0); 120 break; 121 case 'h': 122 cspace(&HS, ps, psl, REPLACE); 123 break; 124 case 'H': 125 cspace(&HS, ps, psl, 0); 126 break; 127 case 'i': 128 (void)printf("%s", cp->t); 129 break; 130 case 'l': 131 lputs(ps); 132 break; 133 case 'n': 134 if (!nflag && !pd) 135 OUT(ps) 136 flush_appends(); 137 r = mf_fgets(&PS, REPLACE); 138 #ifdef HISTORIC_PRACTICE 139 if (!r) 140 exit(0); 141 #endif 142 pd = 0; 143 break; 144 case 'N': 145 flush_appends(); 146 if (!mf_fgets(&PS, 0)) { 147 if (!nflag && !pd) 148 OUT(ps) 149 exit(0); 150 } 151 break; 152 case 'p': 153 if (pd) 154 break; 155 OUT(ps) 156 break; 157 case 'P': 158 if (pd) 159 break; 160 if ((p = memchr(ps, '\n', psl)) != NULL) { 161 oldc = *p; 162 *p = '\0'; 163 } 164 OUT(ps) 165 if (p != NULL) 166 *p = oldc; 167 break; 168 case 'q': 169 if (!nflag && !pd) 170 OUT(ps) 171 flush_appends(); 172 exit(0); 173 case 'r': 174 if (appendx >= appendnum) 175 appends = xrealloc(appends, 176 sizeof(struct s_appends) * 177 (appendnum *= 2)); 178 appends[appendx].type = AP_FILE; 179 appends[appendx].s = cp->t; 180 appends[appendx].len = strlen(cp->t); 181 appendx++; 182 break; 183 case 's': 184 sdone |= substitute(cp); 185 break; 186 case 't': 187 if (sdone) { 188 sdone = 0; 189 cp = cp->u.c; 190 goto redirect; 191 } 192 break; 193 case 'w': 194 if (pd) 195 break; 196 if (cp->u.fd == -1 && (cp->u.fd = open(cp->t, 197 O_WRONLY|O_APPEND|O_CREAT|O_TRUNC, 198 DEFFILEMODE)) == -1) 199 err(FATAL, "%s: %s\n", 200 cp->t, strerror(errno)); 201 if (write(cp->u.fd, ps, psl) != psl) 202 err(FATAL, "%s: %s\n", 203 cp->t, strerror(errno)); 204 break; 205 case 'x': 206 if (hs == NULL) 207 cspace(&HS, "", 0, REPLACE); 208 tspace = PS; 209 PS = HS; 210 HS = tspace; 211 break; 212 case 'y': 213 if (pd) 214 break; 215 for (p = ps, len = psl; --len; ++p) 216 *p = cp->u.y[*p]; 217 break; 218 case ':': 219 case '}': 220 break; 221 case '=': 222 (void)printf("%lu\n", linenum); 223 } 224 cp = cp->next; 225 } /* for all cp */ 226 227 new: if (!nflag && !pd) 228 OUT(ps) 229 flush_appends(); 230 } /* for all lines */ 231 } 232 233 /* 234 * TRUE if the address passed matches the current program state 235 * (lastline, linenumber, ps). 236 */ 237 #define MATCH(a) \ 238 (a)->type == AT_RE ? regexec_e((a)->u.r, ps, 0, 1, psl) : \ 239 (a)->type == AT_LINE ? linenum == (a)->u.l : lastline 240 241 /* 242 * Return TRUE if the command applies to the current line. Sets the inrange 243 * flag to process ranges. Interprets the non-select (``!'') flag. 244 */ 245 static inline int 246 applies(cp) 247 struct s_command *cp; 248 { 249 int r; 250 251 lastaddr = 0; 252 if (cp->a1 == NULL && cp->a2 == NULL) 253 r = 1; 254 else if (cp->a2) 255 if (cp->inrange) { 256 if (MATCH(cp->a2)) { 257 cp->inrange = 0; 258 lastaddr = 1; 259 } 260 r = 1; 261 } else if (MATCH(cp->a1)) { 262 /* 263 * If the second address is a number less than or 264 * equal to the line number first selected, only 265 * one line shall be selected. 266 * -- POSIX 1003.2 267 */ 268 if (cp->a2->type == AT_LINE && 269 linenum >= cp->a2->u.l) 270 lastaddr = 1; 271 else 272 cp->inrange = 1; 273 r = 1; 274 } else 275 r = 0; 276 else 277 r = MATCH(cp->a1); 278 return (cp->nonsel ? ! r : r); 279 } 280 281 /* 282 * substitute -- 283 * Do substitutions in the pattern space. Currently, we build a 284 * copy of the new pattern space in the substitute space structure 285 * and then swap them. 286 */ 287 static int 288 substitute(cp) 289 struct s_command *cp; 290 { 291 SPACE tspace; 292 regex_t *re; 293 size_t re_off, slen; 294 int lastempty, n; 295 char *s; 296 297 s = ps; 298 re = cp->u.s->re; 299 if (re == NULL) { 300 if (defpreg != NULL && cp->u.s->maxbref > defpreg->re_nsub) { 301 linenum = cp->u.s->linenum; 302 err(COMPILE, "\\%d not defined in the RE", 303 cp->u.s->maxbref); 304 } 305 } 306 if (!regexec_e(re, s, 0, 0, psl)) 307 return (0); 308 309 SS.len = 0; /* Clean substitute space. */ 310 slen = psl; 311 n = cp->u.s->n; 312 lastempty = 1; 313 314 switch (n) { 315 case 0: /* Global */ 316 do { 317 if (lastempty || match[0].rm_so != match[0].rm_eo) { 318 /* Locate start of replaced string. */ 319 re_off = match[0].rm_so; 320 /* Copy leading retained string. */ 321 cspace(&SS, s, re_off, APPEND); 322 /* Add in regular expression. */ 323 regsub(&SS, s, cp->u.s->new); 324 } 325 326 /* Move past this match. */ 327 if (match[0].rm_so != match[0].rm_eo) { 328 s += match[0].rm_eo; 329 slen -= match[0].rm_eo; 330 lastempty = 0; 331 } else { 332 if (match[0].rm_so == 0) 333 cspace(&SS, 334 s, match[0].rm_so + 1, APPEND); 335 else 336 cspace(&SS, 337 s + match[0].rm_so, 1, APPEND); 338 s += match[0].rm_so + 1; 339 slen -= match[0].rm_so + 1; 340 lastempty = 1; 341 } 342 } while (slen > 0 && regexec_e(re, s, REG_NOTBOL, 0, slen)); 343 /* Copy trailing retained string. */ 344 if (slen > 0) 345 cspace(&SS, s, slen, APPEND); 346 break; 347 default: /* Nth occurrence */ 348 while (--n) { 349 s += match[0].rm_eo; 350 slen -= match[0].rm_eo; 351 if (!regexec_e(re, s, REG_NOTBOL, 0, slen)) 352 return (0); 353 } 354 /* FALLTHROUGH */ 355 case 1: /* 1st occurrence */ 356 /* Locate start of replaced string. */ 357 re_off = match[0].rm_so + (s - ps); 358 /* Copy leading retained string. */ 359 cspace(&SS, ps, re_off, APPEND); 360 /* Add in regular expression. */ 361 regsub(&SS, s, cp->u.s->new); 362 /* Copy trailing retained string. */ 363 s += match[0].rm_eo; 364 slen -= match[0].rm_eo; 365 cspace(&SS, s, slen, APPEND); 366 break; 367 } 368 369 /* 370 * Swap the substitute space and the pattern space, and make sure 371 * that any leftover pointers into stdio memory get lost. 372 */ 373 tspace = PS; 374 PS = SS; 375 SS = tspace; 376 SS.space = SS.back; 377 378 /* Handle the 'p' flag. */ 379 if (cp->u.s->p) 380 OUT(ps) 381 382 /* Handle the 'w' flag. */ 383 if (cp->u.s->wfile && !pd) { 384 if (cp->u.s->wfd == -1 && (cp->u.s->wfd = open(cp->u.s->wfile, 385 O_WRONLY|O_APPEND|O_CREAT|O_TRUNC, DEFFILEMODE)) == -1) 386 err(FATAL, "%s: %s\n", cp->u.s->wfile, strerror(errno)); 387 if (write(cp->u.s->wfd, ps, psl) != psl) 388 err(FATAL, "%s: %s\n", cp->u.s->wfile, strerror(errno)); 389 } 390 return (1); 391 } 392 393 /* 394 * Flush append requests. Always called before reading a line, 395 * therefore it also resets the substitution done (sdone) flag. 396 */ 397 static void 398 flush_appends() 399 { 400 FILE *f; 401 int count, i; 402 char buf[8 * 1024]; 403 404 for (i = 0; i < appendx; i++) 405 switch (appends[i].type) { 406 case AP_STRING: 407 fwrite(appends[i].s, sizeof(char), appends[i].len, 408 stdout); 409 break; 410 case AP_FILE: 411 /* 412 * Read files probably shouldn't be cached. Since 413 * it's not an error to read a non-existent file, 414 * it's possible that another program is interacting 415 * with the sed script through the file system. It 416 * would be truly bizarre, but possible. It's probably 417 * not that big a performance win, anyhow. 418 */ 419 if ((f = fopen(appends[i].s, "r")) == NULL) 420 break; 421 while (count = fread(buf, sizeof(char), sizeof(buf), f)) 422 (void)fwrite(buf, sizeof(char), count, stdout); 423 (void)fclose(f); 424 break; 425 } 426 if (ferror(stdout)) 427 err(FATAL, "stdout: %s", strerror(errno ? errno : EIO)); 428 appendx = sdone = 0; 429 } 430 431 static void 432 lputs(s) 433 register char *s; 434 { 435 register int count; 436 register char *escapes, *p; 437 struct winsize win; 438 static int termwidth = -1; 439 440 if (termwidth == -1) 441 if (p = getenv("COLUMNS")) 442 termwidth = atoi(p); 443 else if (ioctl(STDOUT_FILENO, TIOCGWINSZ, &win) == 0 && 444 win.ws_col > 0) 445 termwidth = win.ws_col; 446 else 447 termwidth = 60; 448 449 for (count = 0; *s; ++s) { 450 if (count >= termwidth) { 451 (void)printf("\\\n"); 452 count = 0; 453 } 454 if (isascii(*s) && isprint(*s) && *s != '\\') { 455 (void)putchar(*s); 456 count++; 457 } else { 458 escapes = "\\\a\b\f\n\r\t\v"; 459 (void)putchar('\\'); 460 if (p = strchr(escapes, *s)) { 461 (void)putchar("\\abfnrtv"[p - escapes]); 462 count += 2; 463 } else { 464 (void)printf("%03o", *(u_char *)s); 465 count += 4; 466 } 467 } 468 } 469 (void)putchar('$'); 470 (void)putchar('\n'); 471 if (ferror(stdout)) 472 err(FATAL, "stdout: %s", strerror(errno ? errno : EIO)); 473 } 474 475 static inline int 476 regexec_e(preg, string, eflags, nomatch, slen) 477 regex_t *preg; 478 const char *string; 479 int eflags, nomatch; 480 size_t slen; 481 { 482 int eval; 483 484 if (preg == NULL) { 485 if (defpreg == NULL) 486 err(FATAL, "first RE may not be empty"); 487 } else 488 defpreg = preg; 489 490 /* Set anchors, discounting trailing newline (if any). */ 491 if (slen > 0 && string[slen - 1] == '\n') 492 slen--; 493 match[0].rm_so = 0; 494 match[0].rm_eo = slen; 495 496 eval = regexec(defpreg, string, 497 nomatch ? 0 : maxnsub + 1, match, eflags | REG_STARTEND); 498 switch(eval) { 499 case 0: 500 return (1); 501 case REG_NOMATCH: 502 return (0); 503 } 504 err(FATAL, "RE error: %s", strregerror(eval, defpreg)); 505 /* NOTREACHED */ 506 } 507 508 /* 509 * regsub - perform substitutions after a regexp match 510 * Based on a routine by Henry Spencer 511 */ 512 static void 513 regsub(sp, string, src) 514 SPACE *sp; 515 char *string, *src; 516 { 517 register int len, no; 518 register char c, *dst; 519 520 #define NEEDSP(reqlen) \ 521 if (sp->len >= sp->blen - (reqlen) - 1) { \ 522 sp->blen += (reqlen) + 1024; \ 523 sp->space = sp->back = xrealloc(sp->back, sp->blen); \ 524 dst = sp->space + sp->len; \ 525 } 526 527 dst = sp->space + sp->len; 528 while ((c = *src++) != '\0') { 529 if (c == '&') 530 no = 0; 531 else if (c == '\\' && isdigit(*src)) 532 no = *src++ - '0'; 533 else 534 no = -1; 535 if (no < 0) { /* Ordinary character. */ 536 if (c == '\\' && (*src == '\\' || *src == '&')) 537 c = *src++; 538 NEEDSP(1); 539 *dst++ = c; 540 ++sp->len; 541 } else if (match[no].rm_so != -1 && match[no].rm_eo != -1) { 542 len = match[no].rm_eo - match[no].rm_so; 543 NEEDSP(len); 544 memmove(dst, string + match[no].rm_so, len); 545 dst += len; 546 sp->len += len; 547 } 548 } 549 NEEDSP(1); 550 *dst = '\0'; 551 } 552 553 /* 554 * aspace -- 555 * Append the source space to the destination space, allocating new 556 * space as necessary. 557 */ 558 void 559 cspace(sp, p, len, spflag) 560 SPACE *sp; 561 char *p; 562 size_t len; 563 enum e_spflag spflag; 564 { 565 size_t tlen; 566 567 /* Make sure SPACE has enough memory and ramp up quickly. */ 568 tlen = sp->len + len + 1; 569 if (tlen > sp->blen) { 570 sp->blen = tlen + 1024; 571 sp->space = sp->back = xrealloc(sp->back, sp->blen); 572 } 573 574 if (spflag == REPLACE) 575 sp->len = 0; 576 577 memmove(sp->space + sp->len, p, len); 578 579 sp->space[sp->len += len] = '\0'; 580 } 581 582 /* 583 * Close all cached opened files and report any errors 584 */ 585 void 586 cfclose(cp, end) 587 register struct s_command *cp, *end; 588 { 589 590 for (; cp != end; cp = cp->next) 591 switch(cp->code) { 592 case 's': 593 if (cp->u.s->wfd != -1 && close(cp->u.s->wfd)) 594 err(FATAL, 595 "%s: %s", cp->u.s->wfile, strerror(errno)); 596 cp->u.s->wfd = -1; 597 break; 598 case 'w': 599 if (cp->u.fd != -1 && close(cp->u.fd)) 600 err(FATAL, "%s: %s", cp->t, strerror(errno)); 601 cp->u.fd = -1; 602 break; 603 case '{': 604 cfclose(cp->u.c, cp->next); 605 break; 606 } 607 } 608