2 * sed.c - very minimalist version of sed
4 * Copyright (C) 1999,2000 by Lineo, inc.
5 * Written by Mark Whitley <markw@lineo.com>, <markw@enol.com>
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
24 Supported features and commands in this version of sed:
27 - address matching: num|/matchstr/[,num|/matchstr/|$]command
28 - commands: (p)rint, (d)elete, (s)ubstitue (with g & I flags)
29 - edit commands: (a)ppend, (i)nsert, (c)hange
30 - backreferences in substitution expressions (\1, \2...\9)
32 (Note: Specifying an address (range) to match is *optional*; commands
33 default to the whole pattern space if no specific address match was
38 - transliteration (y/source-chars/dest-chars/) (use 'tr')
39 - no support for characters other than the '/' character for regex matches
40 - no pattern space hold space storing / swapping (x, etc.)
41 - no labels / branching (: label, b, t, and friends)
42 - and lots, lots more.
47 #include <unistd.h> /* for getopt() */
49 #include <string.h> /* for strdup() */
51 #include <ctype.h> /* for isspace() */
55 extern int optind; /* in unistd.h */
56 extern char *optarg; /* ditto */
59 static int be_quiet = 0;
68 int beg_line; /* 'sed 1p' 0 == no begining line, apply commands to all lines */
69 int end_line; /* 'sed 1,3p' 0 == no end line, use only beginning. -1 == $ */
70 regex_t *beg_match; /* sed -e '/match/cmd' */
71 regex_t *end_match; /* sed -e '/match/,/end_match/cmd' */
74 char cmd; /* p,d,s (add more at your leisure :-) */
77 /* SUBSTITUTION COMMAND SPECIFIC FIELDS */
79 /* sed -e 's/sub_match/replace/' */
82 unsigned int num_backrefs:4; /* how many back references (\1..\9) */
83 /* Note: GNU/POSIX sed does not save more than nine backrefs, so
84 * we only use 4 bits to hold the number */
85 unsigned int sub_g:1; /* sed -e 's/foo/bar/g' (global) */
86 unsigned int sub_p:2; /* sed -e 's/foo/bar/p' (print substitution) */
89 /* EDIT COMMAND (a,i,c) SPEICIFIC FIELDS */
95 static struct sed_cmd *sed_cmds = NULL; /* growable arrary holding a sequence of sed cmds */
96 static int ncmds = 0; /* number of sed commands */
98 /*static char *cur_file = NULL;*/ /* file currently being processed XXX: do I need this? */
100 #ifdef BB_FEATURE_CLEAN_UP
101 static void destroy_cmd_strs()
103 if (sed_cmds == NULL)
106 /* destroy all the elements in the array */
107 while (--ncmds >= 0) {
109 if (sed_cmds[ncmds].beg_match) {
110 regfree(sed_cmds[ncmds].beg_match);
111 free(sed_cmds[ncmds].beg_match);
113 if (sed_cmds[ncmds].end_match) {
114 regfree(sed_cmds[ncmds].end_match);
115 free(sed_cmds[ncmds].end_match);
117 if (sed_cmds[ncmds].sub_match) {
118 regfree(sed_cmds[ncmds].sub_match);
119 free(sed_cmds[ncmds].sub_match);
121 if (sed_cmds[ncmds].replace)
122 free(sed_cmds[ncmds].replace);
125 /* destroy the array */
133 * trim_str - trims leading and trailing space from a string
135 * Note: This returns a malloc'ed string so you must store and free it
136 * XXX: This should be in the utility.c file.
137 * XXX: This is now obsolete. Maybe it belongs nowhere.
139 static char *trim_str(const char *str)
142 char *retstr = strdup(str);
144 /* trim leading whitespace */
145 memmove(retstr, &retstr[strspn(retstr, " \n\t\v")], strlen(retstr));
147 /* trim trailing whitespace */
148 i = strlen(retstr) - 1;
149 while (isspace(retstr[i]))
155 * you know, a strrspn() would really be nice cuz then we could say:
157 * retstr[strrspn(retstr, " \n\t\v") + 1] = 0;
166 * strrspn - works just like strspn() but goes from right to left instead of
169 static size_t strrspn(const char *s, const char *accept)
171 size_t i = strlen(s);
173 while (strchr(accept, s[--i]))
181 * index_of_next_unescaped_slash - walks left to right through a string
182 * beginning at a specified index and returns the index of the next forward
183 * slash ('/') not preceeded by a backslash ('\').
185 static int index_of_next_unescaped_slash(const char *str, int idx)
187 for ( ; str[idx]; idx++) {
188 if (str[idx] == '/' && str[idx-1] != '\\')
192 /* if we make it to here, we've hit the end of the string */
197 * returns the index in the string just past where the address ends.
199 static int get_address(const char *str, int *line, regex_t **regex)
201 char *my_str = strdup(str);
204 if (isdigit(my_str[idx])) {
207 } while (isdigit(my_str[idx]));
209 *line = atoi(my_str);
211 else if (my_str[idx] == '$') {
215 else if (my_str[idx] == '/') {
216 idx = index_of_next_unescaped_slash(my_str, ++idx);
218 error_msg_and_die("unterminated match expression\n");
220 *regex = (regex_t *)xmalloc(sizeof(regex_t));
221 xregcomp(*regex, my_str+1, 0);
222 idx++; /* so it points to the next character after the last '/' */
225 error_msg("get_address: no address found in string\n"
226 "\t(you probably didn't check the string you passed me)\n");
234 static char *strdup_substr(const char *str, int start, int end)
236 int size = end - start + 1;
237 char *newstr = xmalloc(size);
238 memcpy(newstr, str+start, size-1);
239 newstr[size-1] = '\0';
243 static int parse_subst_cmd(struct sed_cmd *sed_cmd, const char *substr)
245 int oldidx, cflags = REG_NEWLINE;
251 * the string that gets passed to this function should look like this:
252 * s/match/replace/gIp
256 * (all three of the '/' slashes are mandatory)
259 /* verify that the 's' is followed by a 'slash' */
260 if (substr[++idx] != '/')
261 error_msg_and_die("bad format in substitution expression\n");
263 /* save the match string */
265 idx = index_of_next_unescaped_slash(substr, ++idx);
267 error_msg_and_die("bad format in substitution expression\n");
268 match = strdup_substr(substr, oldidx, idx);
270 /* determine the number of back references in the match string */
271 /* Note: we compute this here rather than in the do_subst_command()
272 * function to save processor time, at the expense of a little more memory
273 * (4 bits) per sed_cmd */
275 /* sed_cmd->num_backrefs = 0; */ /* XXX: not needed? --apparently not */
276 for (j = 0; match[j]; j++) {
277 /* GNU/POSIX sed does not save more than nine backrefs */
278 if (match[j] == '\\' && match[j+1] == '(' && sed_cmd->num_backrefs <= 9)
279 sed_cmd->num_backrefs++;
282 /* save the replacement string */
284 idx = index_of_next_unescaped_slash(substr, ++idx);
286 error_msg_and_die("bad format in substitution expression\n");
287 sed_cmd->replace = strdup_substr(substr, oldidx, idx);
289 /* process the flags */
290 while (substr[++idx]) {
291 switch (substr[idx]) {
302 /* any whitespace or semicolon trailing after a s/// is ok */
303 if (strchr("; \t\v\n\r", substr[idx]))
306 error_msg_and_die("bad option in substitution expression\n");
311 /* compile the match string into a regex */
312 sed_cmd->sub_match = (regex_t *)xmalloc(sizeof(regex_t));
313 xregcomp(sed_cmd->sub_match, match, cflags);
319 static int parse_edit_cmd(struct sed_cmd *sed_cmd, const char *editstr)
322 int slashes_eaten = 0;
323 char *ptr; /* shorthand */
326 * the string that gets passed to this function should look like this:
330 * | this backslash (immediately following the edit command) is mandatory
337 * as soon as we hit a TEXT line that has no trailing '\', we're done.
338 * this means a command like:
347 if (editstr[1] != '\\' && (editstr[2] != '\n' || editstr[2] != '\r'))
348 error_msg_and_die("bad format in edit expression\n");
350 /* store the edit line text */
351 /* make editline big enough to accomodate the extra '\n' we will tack on
353 sed_cmd->editline = xmalloc(strlen(&editstr[3]) + 2);
354 strcpy(sed_cmd->editline, &editstr[3]);
355 ptr = sed_cmd->editline;
357 /* now we need to go through * and: s/\\[\r\n]$/\n/g on the edit line */
359 while (ptr[idx] != '\\' && (ptr[idx+1] != '\n' || ptr[idx+1] != '\r')) {
365 /* move the newline over the '\' before it (effectively eats the '\') */
366 memmove(&ptr[idx], &ptr[idx+1], strlen(&ptr[idx+1]));
367 ptr[strlen(ptr)-1] = 0;
369 /* substitue \r for \n if needed */
370 if (ptr[idx] == '\r')
378 /* this accounts for discrepancies between the modified string and the
379 * original string passed in to this function */
380 idx += slashes_eaten;
382 /* this accounts for the fact that A) we started at index 3, not at index
383 * 0 and B) that we added an extra '\n' at the end (if you think the next
384 * line should read 'idx += 4' remember, arrays are zero-based) */
391 static char *parse_cmd_str(struct sed_cmd *sed_cmd, const char *cmdstr)
396 * format is: [addr][,addr]cmd
402 /* first part (if present) is an address: either a number or a /regex/ */
403 if (isdigit(cmdstr[idx]) || cmdstr[idx] == '/')
404 idx = get_address(cmdstr, &sed_cmd->beg_line, &sed_cmd->beg_match);
406 /* second part (if present) will begin with a comma */
407 if (cmdstr[idx] == ',')
408 idx += get_address(&cmdstr[++idx], &sed_cmd->end_line, &sed_cmd->end_match);
410 /* last part (mandatory) will be a command */
411 if (cmdstr[idx] == '\0')
412 error_msg_and_die("missing command\n");
413 if (!strchr("pdsaic", cmdstr[idx])) /* <-- XXX add new commands here */
414 error_msg_and_die("invalid command\n");
415 sed_cmd->cmd = cmdstr[idx];
417 /* special-case handling for (s)ubstitution */
418 if (sed_cmd->cmd == 's') {
419 idx += parse_subst_cmd(sed_cmd, &cmdstr[idx]);
421 /* special-case handling for (a)ppend, (i)nsert, and (c)hange */
422 else if (strchr("aic", cmdstr[idx])) {
423 if (sed_cmd->end_line || sed_cmd->end_match)
424 error_msg_and_die("only a beginning address can be specified for edit commands\n");
425 idx += parse_edit_cmd(sed_cmd, &cmdstr[idx]);
427 /* if it was a single-letter command (such as 'p' or 'd') we need to
428 * increment the index past that command */
432 /* give back whatever's left over */
433 return (char *)&cmdstr[idx];
436 static void add_cmd_str(const char *cmdstr)
438 char *mystr = (char *)cmdstr;
442 /* trim leading whitespace and semicolons */
443 memmove(mystr, &mystr[strspn(mystr, "; \n\r\t\v")], strlen(mystr));
444 /* if we ate the whole thing, that means there was just trailing
445 * whitespace or a final / no-op semicolon. either way, get out */
446 if (strlen(mystr) == 0)
448 /* if this is a comment, jump past it and keep going */
449 if (mystr[0] == '#') {
450 mystr = strpbrk(mystr, ";\n\r");
454 sed_cmds = xrealloc(sed_cmds, sizeof(struct sed_cmd) * (++ncmds));
455 /* zero new element */
456 memset(&sed_cmds[ncmds-1], 0, sizeof(struct sed_cmd));
457 /* load command string into new array element, get remainder */
458 mystr = parse_cmd_str(&sed_cmds[ncmds-1], mystr);
460 } while (mystr && strlen(mystr));
464 static void load_cmd_file(char *filename)
470 cmdfile = xfopen(filename, "r");
472 while ((line = get_line_from_file(cmdfile)) != NULL) {
473 /* if a line ends with '\' it needs the next line appended to it */
474 while (line[strlen(line)-2] == '\\' &&
475 (nextline = get_line_from_file(cmdfile)) != NULL) {
476 line = xrealloc(line, strlen(line) + strlen(nextline) + 1);
477 strcat(line, nextline);
480 /* eat trailing newline (if any) --if I don't do this, edit commands
481 * (aic) will print an extra newline */
482 if (line[strlen(line)-1] == '\n')
483 line[strlen(line)-1] = 0;
489 static void print_subst_w_backrefs(const char *line, const char *replace, regmatch_t *regmatch)
493 /* go through the replacement string */
494 for (i = 0; replace[i]; i++) {
495 /* if we find a backreference (\1, \2, etc.) print the backref'ed * text */
496 if (replace[i] == '\\' && isdigit(replace[i+1])) {
500 ++i; /* i now indexes the backref number, instead of the leading slash */
501 tmpstr[0] = replace[i];
503 backref = atoi(tmpstr);
504 /* print out the text held in regmatch[backref] */
505 for (j = regmatch[backref].rm_so; j < regmatch[backref].rm_eo; j++)
506 fputc(line[j], stdout);
509 /* if we find a backslash escaped character, print the character */
510 else if (replace[i] == '\\') {
512 fputc(replace[i], stdout);
515 /* if we find an unescaped '&' print out the whole matched text.
516 * fortunately, regmatch[0] contains the indicies to the whole matched
517 * expression (kinda seems like it was designed for just such a
519 else if (replace[i] == '&' && replace[i-1] != '\\') {
521 for (j = regmatch[0].rm_so; j < regmatch[0].rm_eo; j++)
522 fputc(line[j], stdout);
524 /* nothing special, just print this char of the replacement string to stdout */
526 fputc(replace[i], stdout);
530 static int do_subst_command(const struct sed_cmd *sed_cmd, const char *line)
532 char *hackline = (char *)line;
534 regmatch_t *regmatch = NULL;
536 /* we only proceed if the substitution 'search' expression matches */
537 if (regexec(sed_cmd->sub_match, line, 0, NULL, 0) == REG_NOMATCH)
540 /* whaddaya know, it matched. get the number of back references */
541 regmatch = xmalloc(sizeof(regmatch_t) * (sed_cmd->num_backrefs+1));
543 /* and now, as long as we've got a line to try matching and if we can match
544 * the search string, we make substitutions */
545 while (*hackline && (regexec(sed_cmd->sub_match, hackline,
546 sed_cmd->num_backrefs+1, regmatch, 0) == 0) ) {
549 /* print everything before the match */
550 for (i = 0; i < regmatch[0].rm_so; i++)
551 fputc(hackline[i], stdout);
553 /* then print the substitution string */
554 print_subst_w_backrefs(hackline, sed_cmd->replace, regmatch);
556 /* advance past the match */
557 hackline += regmatch[0].rm_eo;
558 /* flag that something has changed */
561 /* if we're not doing this globally, get out now */
566 /* if there's anything left of the line, print it */
568 fputs(hackline, stdout);
576 static int do_sed_command(const struct sed_cmd *sed_cmd, const char *line)
580 switch (sed_cmd->cmd) {
593 * Some special cases for 's' printing to make it compliant with
594 * GNU sed printing behavior (aka "The -n | s///p Matrix"):
596 * -n ONLY = never print anything regardless of any successful
599 * s///p ONLY = always print successful substitutions, even if
600 * the line is going to be printed anyway (line will be printed
603 * -n AND s///p = print ONLY a successful substitution ONE TIME;
604 * no other lines are printed - this is the reason why the 'p'
605 * flag exists in the first place.
608 /* if the user specified that they didn't want anything printed (i.e. a -n
609 * flag and no 'p' flag after the s///), then there's really no point doing
611 if (be_quiet && !sed_cmd->sub_p)
614 /* we print the line once, unless we were told to be quiet */
616 altered = do_subst_command(sed_cmd, line);
618 /* we also print the line if we were given the 'p' flag
619 * (this is quite possibly the second printing) */
621 altered = do_subst_command(sed_cmd, line);
627 fputs(sed_cmd->editline, stdout);
632 fputs(sed_cmd->editline, stdout);
636 fputs(sed_cmd->editline, stdout);
644 static void process_file(FILE *file)
647 static int linenum = 0; /* GNU sed does not restart counting lines at EOF */
648 unsigned int still_in_range = 0;
652 /* go through every line in the file */
653 while ((line = get_line_from_file(file)) != NULL) {
658 /* for every line, go through all the commands */
659 for (i = 0; i < ncmds; i++) {
661 /* are we acting on a range of matched lines? */
662 if (sed_cmds[i].beg_match && sed_cmds[i].end_match) {
663 if (still_in_range || regexec(sed_cmds[i].beg_match, line, 0, NULL, 0) == 0) {
664 line_altered += do_sed_command(&sed_cmds[i], line);
666 if (regexec(sed_cmds[i].end_match, line, 0, NULL, 0) == 0)
671 /* are we trying to match a single line? */
672 else if (sed_cmds[i].beg_match) {
673 if (regexec(sed_cmds[i].beg_match, line, 0, NULL, 0) == 0)
674 line_altered += do_sed_command(&sed_cmds[i], line);
677 /* are we acting on a range of line numbers? */
678 else if (sed_cmds[i].beg_line > 0 && sed_cmds[i].end_line != 0) {
679 if (linenum >= sed_cmds[i].beg_line &&
680 (sed_cmds[i].end_line == -1 || linenum <= sed_cmds[i].end_line))
681 line_altered += do_sed_command(&sed_cmds[i], line);
684 /* are we acting on a specified line number */
685 else if (sed_cmds[i].beg_line > 0) {
686 if (linenum == sed_cmds[i].beg_line)
687 line_altered += do_sed_command(&sed_cmds[i], line);
690 /* not acting on matches or line numbers. act on every line */
692 line_altered += do_sed_command(&sed_cmds[i], line);
696 /* we will print the line unless we were told to be quiet or if the
697 * line was altered (via a 'd'elete or 's'ubstitution), in which case
698 * the altered line was already printed */
699 if (!be_quiet && !line_altered)
706 extern int sed_main(int argc, char **argv)
710 #ifdef BB_FEATURE_CLEAN_UP
711 /* destroy command strings on exit */
712 if (atexit(destroy_cmd_strs) == -1) {
718 /* do normal option parsing */
719 while ((opt = getopt(argc, argv, "hne:f:")) > 0) {
731 load_cmd_file(optarg);
736 /* if we didn't get a pattern from a -e and no command file was specified,
737 * argv[optind] should be the pattern. no pattern, no worky */
739 if (argv[optind] == NULL)
742 add_cmd_str(argv[optind]);
748 /* argv[(optind)..(argc-1)] should be names of file to process. If no
749 * files were specified or '-' was specified, take input from stdin.
750 * Otherwise, we process all the files specified. */
751 if (argv[optind] == NULL || (strcmp(argv[optind], "-") == 0)) {
757 for (i = optind; i < argc; i++) {
758 file = fopen(argv[i], "r");
760 perror_msg("%s", argv[i]);