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() */
54 #define bb_need_full_version
55 #define BB_DECLARE_EXTERN
59 extern int optind; /* in unistd.h */
60 extern char *optarg; /* ditto */
63 static int be_quiet = 0;
72 int beg_line; /* 'sed 1p' 0 == no begining line, apply commands to all lines */
73 int end_line; /* 'sed 1,3p' 0 == no end line, use only beginning. -1 == $ */
74 regex_t *beg_match; /* sed -e '/match/cmd' */
75 regex_t *end_match; /* sed -e '/match/,/end_match/cmd' */
78 char cmd; /* p,d,s (add more at your leisure :-) */
81 /* SUBSTITUTION COMMAND SPECIFIC FIELDS */
83 /* sed -e 's/sub_match/replace/' */
86 unsigned int num_backrefs:4; /* how many back references (\1..\9) */
87 /* Note: GNU/POSIX sed does not save more than nine backrefs, so
88 * we only use 4 bits to hold the number */
89 unsigned int sub_g:1; /* sed -e 's/foo/bar/g' (global) */
90 unsigned int sub_p:2; /* sed -e 's/foo/bar/p' (print substitution) */
93 /* EDIT COMMAND (a,i,c) SPEICIFIC FIELDS */
99 static struct sed_cmd *sed_cmds = NULL; /* growable arrary holding a sequence of sed cmds */
100 static int ncmds = 0; /* number of sed commands */
102 /*static char *cur_file = NULL;*/ /* file currently being processed XXX: do I need this? */
104 #ifdef BB_FEATURE_CLEAN_UP
105 static void destroy_cmd_strs()
107 if (sed_cmds == NULL)
110 /* destroy all the elements in the array */
111 while (--ncmds >= 0) {
113 if (sed_cmds[ncmds].beg_match) {
114 regfree(sed_cmds[ncmds].beg_match);
115 free(sed_cmds[ncmds].beg_match);
117 if (sed_cmds[ncmds].end_match) {
118 regfree(sed_cmds[ncmds].end_match);
119 free(sed_cmds[ncmds].end_match);
121 if (sed_cmds[ncmds].sub_match) {
122 regfree(sed_cmds[ncmds].sub_match);
123 free(sed_cmds[ncmds].sub_match);
125 if (sed_cmds[ncmds].replace)
126 free(sed_cmds[ncmds].replace);
129 /* destroy the array */
137 * trim_str - trims leading and trailing space from a string
139 * Note: This returns a malloc'ed string so you must store and free it
140 * XXX: This should be in the utility.c file.
141 * XXX: This is now obsolete. Maybe it belongs nowhere.
143 static char *trim_str(const char *str)
146 char *retstr = strdup(str);
148 /* trim leading whitespace */
149 memmove(retstr, &retstr[strspn(retstr, " \n\t\v")], strlen(retstr));
151 /* trim trailing whitespace */
152 i = strlen(retstr) - 1;
153 while (isspace(retstr[i]))
159 * you know, a strrspn() would really be nice cuz then we could say:
161 * retstr[strrspn(retstr, " \n\t\v") + 1] = 0;
170 * strrspn - works just like strspn() but goes from right to left instead of
173 static size_t strrspn(const char *s, const char *accept)
175 size_t i = strlen(s);
177 while (strchr(accept, s[--i]))
185 * index_of_next_unescaped_slash - walks left to right through a string
186 * beginning at a specified index and returns the index of the next forward
187 * slash ('/') not preceeded by a backslash ('\').
189 static int index_of_next_unescaped_slash(const char *str, int idx)
191 for ( ; str[idx]; idx++) {
192 if (str[idx] == '/' && str[idx-1] != '\\')
196 /* if we make it to here, we've hit the end of the string */
201 * returns the index in the string just past where the address ends.
203 static int get_address(const char *str, int *line, regex_t **regex)
205 char *my_str = strdup(str);
208 if (isdigit(my_str[idx])) {
211 } while (isdigit(my_str[idx]));
213 *line = atoi(my_str);
215 else if (my_str[idx] == '$') {
219 else if (my_str[idx] == '/') {
220 idx = index_of_next_unescaped_slash(my_str, ++idx);
222 fatalError("unterminated match expression\n");
224 *regex = (regex_t *)xmalloc(sizeof(regex_t));
225 xregcomp(*regex, my_str+1, REG_NEWLINE);
226 idx++; /* so it points to the next character after the last '/' */
229 errorMsg("get_address: no address found in string\n"
230 "\t(you probably didn't check the string you passed me)\n");
238 static char *strdup_substr(const char *str, int start, int end)
240 int size = end - start + 1;
241 char *newstr = xmalloc(size);
242 memcpy(newstr, str+start, size-1);
243 newstr[size-1] = '\0';
247 static int parse_subst_cmd(struct sed_cmd *sed_cmd, const char *substr)
249 int oldidx, cflags = REG_NEWLINE;
255 * the string that gets passed to this function should look like this:
256 * s/match/replace/gIp
260 * (all three of the '/' slashes are mandatory)
263 /* verify that the 's' is followed by a 'slash' */
264 if (substr[++idx] != '/')
265 fatalError("bad format in substitution expression\n");
267 /* save the match string */
269 idx = index_of_next_unescaped_slash(substr, ++idx);
271 fatalError("bad format in substitution expression\n");
272 match = strdup_substr(substr, oldidx, idx);
274 /* determine the number of back references in the match string */
275 /* Note: we compute this here rather than in the do_subst_command()
276 * function to save processor time, at the expense of a little more memory
277 * (4 bits) per sed_cmd */
279 /* sed_cmd->num_backrefs = 0; */ /* XXX: not needed? --apparently not */
280 for (j = 0; match[j]; j++) {
281 /* GNU/POSIX sed does not save more than nine backrefs */
282 if (match[j] == '\\' && match[j+1] == '(' && sed_cmd->num_backrefs <= 9)
283 sed_cmd->num_backrefs++;
286 /* save the replacement string */
288 idx = index_of_next_unescaped_slash(substr, ++idx);
290 fatalError("bad format in substitution expression\n");
291 sed_cmd->replace = strdup_substr(substr, oldidx, idx);
293 /* process the flags */
294 while (substr[++idx]) {
295 switch (substr[idx]) {
306 /* any whitespace or semicolon trailing after a s/// is ok */
307 if (strchr("; \t\v\n\r", substr[idx]))
310 fatalError("bad option in substitution expression\n");
315 /* compile the match string into a regex */
316 sed_cmd->sub_match = (regex_t *)xmalloc(sizeof(regex_t));
317 xregcomp(sed_cmd->sub_match, match, cflags);
323 static int parse_edit_cmd(struct sed_cmd *sed_cmd, const char *editstr)
326 int slashes_eaten = 0;
327 char *ptr; /* shorthand */
330 * the string that gets passed to this function should look like this:
334 * | this backslash (immediately following the edit command) is mandatory
341 * as soon as we hit a TEXT line that has no trailing '\', we're done.
342 * this means a command like:
351 if (editstr[1] != '\\' && (editstr[2] != '\n' || editstr[2] != '\r'))
352 fatalError("bad format in edit expression\n");
354 /* store the edit line text */
355 /* make editline big enough to accomodate the extra '\n' we will tack on
357 sed_cmd->editline = xmalloc(strlen(&editstr[3]) + 2);
358 strcpy(sed_cmd->editline, &editstr[3]);
359 ptr = sed_cmd->editline;
361 /* now we need to go through * and: s/\\[\r\n]$/\n/g on the edit line */
363 while (ptr[idx] != '\\' && (ptr[idx+1] != '\n' || ptr[idx+1] != '\r')) {
369 /* move the newline over the '\' before it (effectively eats the '\') */
370 memmove(&ptr[idx], &ptr[idx+1], strlen(&ptr[idx+1]));
371 ptr[strlen(ptr)-1] = 0;
373 /* substitue \r for \n if needed */
374 if (ptr[idx] == '\r')
382 /* this accounts for discrepancies between the modified string and the
383 * original string passed in to this function */
384 idx += slashes_eaten;
386 /* this accounts for the fact that A) we started at index 3, not at index
387 * 0 and B) that we added an extra '\n' at the end (if you think the next
388 * line should read 'idx += 4' remember, arrays are zero-based) */
395 static char *parse_cmd_str(struct sed_cmd *sed_cmd, const char *cmdstr)
400 * format is: [addr][,addr]cmd
406 /* first part (if present) is an address: either a number or a /regex/ */
407 if (isdigit(cmdstr[idx]) || cmdstr[idx] == '/')
408 idx = get_address(cmdstr, &sed_cmd->beg_line, &sed_cmd->beg_match);
410 /* second part (if present) will begin with a comma */
411 if (cmdstr[idx] == ',')
412 idx += get_address(&cmdstr[++idx], &sed_cmd->end_line, &sed_cmd->end_match);
414 /* last part (mandatory) will be a command */
415 if (cmdstr[idx] == '\0')
416 fatalError("missing command\n");
417 if (!strchr("pdsaic", cmdstr[idx])) /* <-- XXX add new commands here */
418 fatalError("invalid command\n");
419 sed_cmd->cmd = cmdstr[idx];
421 /* special-case handling for (s)ubstitution */
422 if (sed_cmd->cmd == 's') {
423 idx += parse_subst_cmd(sed_cmd, &cmdstr[idx]);
425 /* special-case handling for (a)ppend, (i)nsert, and (c)hange */
426 else if (strchr("aic", cmdstr[idx])) {
427 if (sed_cmd->end_line || sed_cmd->end_match)
428 fatalError("only a beginning address can be specified for edit commands\n");
429 idx += parse_edit_cmd(sed_cmd, &cmdstr[idx]);
431 /* if it was a single-letter command (such as 'p' or 'd') we need to
432 * increment the index past that command */
436 /* give back whatever's left over */
437 return (char *)&cmdstr[idx];
440 static void add_cmd_str(const char *cmdstr)
442 char *mystr = (char *)cmdstr;
446 /* trim leading whitespace and semicolons */
447 memmove(mystr, &mystr[strspn(mystr, "; \n\r\t\v")], strlen(mystr));
448 /* if we ate the whole thing, that means there was just trailing
449 * whitespace or a final / no-op semicolon. either way, get out */
450 if (strlen(mystr) == 0)
452 /* if this is a comment, jump past it and keep going */
453 if (mystr[0] == '#') {
454 mystr = strpbrk(mystr, ";\n\r");
458 sed_cmds = xrealloc(sed_cmds, sizeof(struct sed_cmd) * (++ncmds));
459 /* zero new element */
460 memset(&sed_cmds[ncmds-1], 0, sizeof(struct sed_cmd));
461 /* load command string into new array element, get remainder */
462 mystr = parse_cmd_str(&sed_cmds[ncmds-1], mystr);
464 } while (mystr && strlen(mystr));
468 static void load_cmd_file(char *filename)
474 cmdfile = xfopen(filename, "r");
476 while ((line = get_line_from_file(cmdfile)) != NULL) {
477 /* if a line ends with '\' it needs the next line appended to it */
478 while (line[strlen(line)-2] == '\\' &&
479 (nextline = get_line_from_file(cmdfile)) != NULL) {
480 line = xrealloc(line, strlen(line) + strlen(nextline) + 1);
481 strcat(line, nextline);
484 /* eat trailing newline (if any) --if I don't do this, edit commands
485 * (aic) will print an extra newline */
486 if (line[strlen(line)-1] == '\n')
487 line[strlen(line)-1] = 0;
493 static void print_subst_w_backrefs(const char *line, const char *replace, regmatch_t *regmatch)
497 /* go through the replacement string */
498 for (i = 0; replace[i]; i++) {
499 /* if we find a backreference (\1, \2, etc.) print the backref'ed * text */
500 if (replace[i] == '\\' && isdigit(replace[i+1])) {
504 ++i; /* i now indexes the backref number, instead of the leading slash */
505 tmpstr[0] = replace[i];
507 backref = atoi(tmpstr);
508 /* print out the text held in regmatch[backref] */
509 for (j = regmatch[backref].rm_so; j < regmatch[backref].rm_eo; j++)
510 fputc(line[j], stdout);
513 /* if we find a backslash escaped character, print the character */
514 else if (replace[i] == '\\') {
516 fputc(replace[i], stdout);
519 /* if we find an unescaped '&' print out the whole matched text.
520 * fortunately, regmatch[0] contains the indicies to the whole matched
521 * expression (kinda seems like it was designed for just such a
523 else if (replace[i] == '&' && replace[i-1] != '\\') {
525 for (j = regmatch[0].rm_so; j < regmatch[0].rm_eo; j++)
526 fputc(line[j], stdout);
528 /* nothing special, just print this char of the replacement string to stdout */
530 fputc(replace[i], stdout);
534 static int do_subst_command(const struct sed_cmd *sed_cmd, const char *line)
536 char *hackline = (char *)line;
538 regmatch_t *regmatch = NULL;
540 /* we only proceed if the substitution 'search' expression matches */
541 if (regexec(sed_cmd->sub_match, line, 0, NULL, 0) == REG_NOMATCH)
544 /* whaddaya know, it matched. get the number of back references */
545 regmatch = xmalloc(sizeof(regmatch_t) * (sed_cmd->num_backrefs+1));
547 /* and now, as long as we've got a line to try matching and if we can match
548 * the search string, we make substitutions */
549 while (*hackline && (regexec(sed_cmd->sub_match, hackline,
550 sed_cmd->num_backrefs+1, regmatch, 0) == 0) ) {
553 /* print everything before the match */
554 for (i = 0; i < regmatch[0].rm_so; i++)
555 fputc(hackline[i], stdout);
557 /* then print the substitution string */
558 print_subst_w_backrefs(hackline, sed_cmd->replace, regmatch);
560 /* advance past the match */
561 hackline += regmatch[0].rm_eo;
562 /* flag that something has changed */
565 /* if we're not doing this globally, get out now */
570 /* if there's anything left of the line, print it */
572 fputs(hackline, stdout);
580 static int do_sed_command(const struct sed_cmd *sed_cmd, const char *line)
584 switch (sed_cmd->cmd) {
597 * Some special cases for 's' printing to make it compliant with
598 * GNU sed printing behavior (aka "The -n | s///p Matrix"):
600 * -n ONLY = never print anything regardless of any successful
603 * s///p ONLY = always print successful substitutions, even if
604 * the line is going to be printed anyway (line will be printed
607 * -n AND s///p = print ONLY a successful substitution ONE TIME;
608 * no other lines are printed - this is the reason why the 'p'
609 * flag exists in the first place.
612 /* if the user specified that they didn't want anything printed (i.e. a -n
613 * flag and no 'p' flag after the s///), then there's really no point doing
615 if (be_quiet && !sed_cmd->sub_p)
618 /* we print the line once, unless we were told to be quiet */
620 altered = do_subst_command(sed_cmd, line);
622 /* we also print the line if we were given the 'p' flag
623 * (this is quite possibly the second printing) */
625 altered = do_subst_command(sed_cmd, line);
631 fputs(sed_cmd->editline, stdout);
636 fputs(sed_cmd->editline, stdout);
640 fputs(sed_cmd->editline, stdout);
648 static void process_file(FILE *file)
651 static int linenum = 0; /* GNU sed does not restart counting lines at EOF */
652 unsigned int still_in_range = 0;
656 /* go through every line in the file */
657 while ((line = get_line_from_file(file)) != NULL) {
662 /* for every line, go through all the commands */
663 for (i = 0; i < ncmds; i++) {
665 /* are we acting on a range of matched lines? */
666 if (sed_cmds[i].beg_match && sed_cmds[i].end_match) {
667 if (still_in_range || regexec(sed_cmds[i].beg_match, line, 0, NULL, 0) == 0) {
668 line_altered += do_sed_command(&sed_cmds[i], line);
670 if (regexec(sed_cmds[i].end_match, line, 0, NULL, 0) == 0)
675 /* are we trying to match a single line? */
676 else if (sed_cmds[i].beg_match) {
677 if (regexec(sed_cmds[i].beg_match, line, 0, NULL, 0) == 0)
678 line_altered += do_sed_command(&sed_cmds[i], line);
681 /* are we acting on a range of line numbers? */
682 else if (sed_cmds[i].beg_line > 0 && sed_cmds[i].end_line != 0) {
683 if (linenum >= sed_cmds[i].beg_line &&
684 (sed_cmds[i].end_line == -1 || linenum <= sed_cmds[i].end_line))
685 line_altered += do_sed_command(&sed_cmds[i], line);
688 /* are we acting on a specified line number */
689 else if (sed_cmds[i].beg_line > 0) {
690 if (linenum == sed_cmds[i].beg_line)
691 line_altered += do_sed_command(&sed_cmds[i], line);
694 /* not acting on matches or line numbers. act on every line */
696 line_altered += do_sed_command(&sed_cmds[i], line);
700 /* we will print the line unless we were told to be quiet or if the
701 * line was altered (via a 'd'elete or 's'ubstitution), in which case
702 * the altered line was already printed */
703 if (!be_quiet && !line_altered)
710 extern int sed_main(int argc, char **argv)
714 #ifdef BB_FEATURE_CLEAN_UP
715 /* destroy command strings on exit */
716 if (atexit(destroy_cmd_strs) == -1) {
722 /* do normal option parsing */
723 while ((opt = getopt(argc, argv, "hne:f:")) > 0) {
735 load_cmd_file(optarg);
740 /* if we didn't get a pattern from a -e and no command file was specified,
741 * argv[optind] should be the pattern. no pattern, no worky */
743 if (argv[optind] == NULL)
746 add_cmd_str(argv[optind]);
752 /* argv[(optind)..(argc-1)] should be names of file to process. If no
753 * files were specified or '-' was specified, take input from stdin.
754 * Otherwise, we process all the files specified. */
755 if (argv[optind] == NULL || (strcmp(argv[optind], "-") == 0)) {
761 for (i = optind; i < argc; i++) {
762 file = fopen(argv[i], "r");
764 errorMsg("%s: %s\n", argv[i], strerror(errno));