2 * sed.c - very minimalist version of sed
4 * Copyright (C) 1999,2000,2001 by Lineo, inc.
5 * Written by Mark Whitley <markw@lineo.com>, <markw@codepoet.org>
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 pattern space hold space storing / swapping (x, etc.)
40 - no labels / branching (: label, b, t, and friends)
41 - and lots, lots more.
45 #include <unistd.h> /* for getopt() */
47 #include <string.h> /* for strdup() */
49 #include <ctype.h> /* for isspace() */
54 extern void xregcomp(regex_t *preg, const char *regex, int cflags);
55 extern int optind; /* in unistd.h */
56 extern char *optarg; /* ditto */
59 static int be_quiet = 0;
66 char delimiter; /* The delimiter used to separate regexps */
69 int beg_line; /* 'sed 1p' 0 == no begining line, apply commands to all lines */
70 int end_line; /* 'sed 1,3p' 0 == no end line, use only beginning. -1 == $ */
71 regex_t *beg_match; /* sed -e '/match/cmd' */
72 regex_t *end_match; /* sed -e '/match/,/end_match/cmd' */
75 char cmd; /* p,d,s (add more at your leisure :-) */
78 /* SUBSTITUTION COMMAND SPECIFIC FIELDS */
80 /* sed -e 's/sub_match/replace/' */
83 unsigned int num_backrefs:4; /* how many back references (\1..\9) */
84 /* Note: GNU/POSIX sed does not save more than nine backrefs, so
85 * we only use 4 bits to hold the number */
86 unsigned int sub_g:1; /* sed -e 's/foo/bar/g' (global) */
87 unsigned int sub_p:2; /* sed -e 's/foo/bar/p' (print substitution) */
90 /* EDIT COMMAND (a,i,c) SPEICIFIC FIELDS */
96 static struct sed_cmd *sed_cmds = NULL; /* growable arrary holding a sequence of sed cmds */
97 static int ncmds = 0; /* number of sed commands */
99 /*static char *cur_file = NULL;*/ /* file currently being processed XXX: do I need this? */
101 #ifdef BB_FEATURE_CLEAN_UP
102 static void destroy_cmd_strs()
104 if (sed_cmds == NULL)
107 /* destroy all the elements in the array */
108 while (--ncmds >= 0) {
110 if (sed_cmds[ncmds].beg_match) {
111 regfree(sed_cmds[ncmds].beg_match);
112 free(sed_cmds[ncmds].beg_match);
114 if (sed_cmds[ncmds].end_match) {
115 regfree(sed_cmds[ncmds].end_match);
116 free(sed_cmds[ncmds].end_match);
118 if (sed_cmds[ncmds].sub_match) {
119 regfree(sed_cmds[ncmds].sub_match);
120 free(sed_cmds[ncmds].sub_match);
122 if (sed_cmds[ncmds].replace)
123 free(sed_cmds[ncmds].replace);
126 /* destroy the array */
134 * index_of_next_unescaped_regexp_delim - walks left to right through a string
135 * beginning at a specified index and returns the index of the next regular
136 * expression delimiter (typically a forward * slash ('/')) not preceeded by
139 static int index_of_next_unescaped_regexp_delim(struct sed_cmd *sed_cmd, const char *str, int idx)
141 for ( ; str[idx]; idx++) {
142 if (str[idx] == sed_cmd->delimiter && str[idx-1] != '\\')
146 /* if we make it to here, we've hit the end of the string */
151 * returns the index in the string just past where the address ends.
153 static int get_address(struct sed_cmd *sed_cmd, const char *str, int *line, regex_t **regex)
155 char *my_str = strdup(str);
158 olddelimiter = sed_cmd->delimiter;
159 sed_cmd->delimiter = '/';
161 if (isdigit(my_str[idx])) {
164 } while (isdigit(my_str[idx]));
166 *line = atoi(my_str);
168 else if (my_str[idx] == '$') {
172 else if (my_str[idx] == '/') {
173 idx = index_of_next_unescaped_regexp_delim(sed_cmd, my_str, ++idx);
175 error_msg_and_die("unterminated match expression");
177 *regex = (regex_t *)xmalloc(sizeof(regex_t));
178 xregcomp(*regex, my_str+1, 0);
179 idx++; /* so it points to the next character after the last '/' */
182 error_msg("get_address: no address found in string\n"
183 "\t(you probably didn't check the string you passed me)");
188 sed_cmd->delimiter = olddelimiter;
192 static char *strdup_substr(const char *str, int start, int end)
194 int size = end - start + 1;
195 char *newstr = xmalloc(size);
196 memcpy(newstr, str+start, size-1);
197 newstr[size-1] = '\0';
201 static int parse_subst_cmd(struct sed_cmd *sed_cmd, const char *substr)
203 int oldidx, cflags = REG_NEWLINE;
209 * the string that gets passed to this function should look like this:
210 * s/match/replace/gIp
214 * (all three of the '/' slashes are mandatory)
217 /* verify that the 's' is followed by something. That something
218 * (typically a 'slash') is now our regexp delimiter... */
220 error_msg_and_die("bad format in substitution expression");
222 sed_cmd->delimiter=substr[idx];
224 /* save the match string */
226 idx = index_of_next_unescaped_regexp_delim(sed_cmd, substr, ++idx);
228 error_msg_and_die("bad format in substitution expression");
229 match = strdup_substr(substr, oldidx, idx);
231 /* determine the number of back references in the match string */
232 /* Note: we compute this here rather than in the do_subst_command()
233 * function to save processor time, at the expense of a little more memory
234 * (4 bits) per sed_cmd */
236 /* sed_cmd->num_backrefs = 0; */ /* XXX: not needed? --apparently not */
237 for (j = 0; match[j]; j++) {
238 /* GNU/POSIX sed does not save more than nine backrefs */
239 if (match[j] == '\\' && match[j+1] == '(' && sed_cmd->num_backrefs <= 9)
240 sed_cmd->num_backrefs++;
243 /* save the replacement string */
245 idx = index_of_next_unescaped_regexp_delim(sed_cmd, substr, ++idx);
247 error_msg_and_die("bad format in substitution expression");
248 sed_cmd->replace = strdup_substr(substr, oldidx, idx);
250 /* process the flags */
251 while (substr[++idx]) {
252 switch (substr[idx]) {
263 /* any whitespace or semicolon trailing after a s/// is ok */
264 if (strchr("; \t\v\n\r", substr[idx]))
267 error_msg_and_die("bad option in substitution expression");
272 /* compile the match string into a regex */
273 sed_cmd->sub_match = (regex_t *)xmalloc(sizeof(regex_t));
274 xregcomp(sed_cmd->sub_match, match, cflags);
280 static int parse_edit_cmd(struct sed_cmd *sed_cmd, const char *editstr)
283 int slashes_eaten = 0;
284 char *ptr; /* shorthand */
287 * the string that gets passed to this function should look like this:
291 * | this backslash (immediately following the edit command) is mandatory
298 * as soon as we hit a TEXT line that has no trailing '\', we're done.
299 * this means a command like:
308 if (editstr[1] != '\\' && (editstr[2] != '\n' || editstr[2] != '\r'))
309 error_msg_and_die("bad format in edit expression");
311 /* store the edit line text */
312 /* make editline big enough to accomodate the extra '\n' we will tack on
314 sed_cmd->editline = xmalloc(strlen(&editstr[3]) + 2);
315 strcpy(sed_cmd->editline, &editstr[3]);
316 ptr = sed_cmd->editline;
318 /* now we need to go through * and: s/\\[\r\n]$/\n/g on the edit line */
320 while (ptr[idx] != '\\' || (ptr[idx+1] != '\n' && ptr[idx+1] != '\r')) {
326 /* move the newline over the '\' before it (effectively eats the '\') */
327 memmove(&ptr[idx], &ptr[idx+1], strlen(&ptr[idx+1]));
328 ptr[strlen(ptr)-1] = 0;
330 /* substitue \r for \n if needed */
331 if (ptr[idx] == '\r')
336 /* this accounts for discrepancies between the modified string and the
337 * original string passed in to this function */
338 idx += slashes_eaten;
340 /* figure out if we need to add a newline */
341 if (ptr[idx-1] != '\n') {
346 /* terminate string */
348 /* adjust for opening 2 chars [aic]\ */
354 static char *parse_cmd_str(struct sed_cmd *sed_cmd, const char *cmdstr)
359 * format is: [addr][,addr]cmd
365 /* first part (if present) is an address: either a number or a /regex/ */
366 if (isdigit(cmdstr[idx]) || cmdstr[idx] == '/')
367 idx = get_address(sed_cmd, cmdstr, &sed_cmd->beg_line, &sed_cmd->beg_match);
369 /* second part (if present) will begin with a comma */
370 if (cmdstr[idx] == ',')
371 idx += get_address(sed_cmd, &cmdstr[++idx], &sed_cmd->end_line, &sed_cmd->end_match);
373 /* last part (mandatory) will be a command */
374 if (cmdstr[idx] == '\0')
375 error_msg_and_die("missing command");
376 if (!strchr("pdsaic", cmdstr[idx])) /* <-- XXX add new commands here */
377 error_msg_and_die("invalid command");
378 sed_cmd->cmd = cmdstr[idx];
380 /* special-case handling for (s)ubstitution */
381 if (sed_cmd->cmd == 's') {
382 idx += parse_subst_cmd(sed_cmd, &cmdstr[idx]);
384 /* special-case handling for (a)ppend, (i)nsert, and (c)hange */
385 else if (strchr("aic", cmdstr[idx])) {
386 if (sed_cmd->end_line || sed_cmd->end_match)
387 error_msg_and_die("only a beginning address can be specified for edit commands");
388 idx += parse_edit_cmd(sed_cmd, &cmdstr[idx]);
390 /* if it was a single-letter command (such as 'p' or 'd') we need to
391 * increment the index past that command */
395 /* give back whatever's left over */
396 return (char *)&cmdstr[idx];
399 static void add_cmd_str(const char *cmdstr)
401 char *mystr = (char *)cmdstr;
405 /* trim leading whitespace and semicolons */
406 memmove(mystr, &mystr[strspn(mystr, "; \n\r\t\v")], strlen(mystr));
407 /* if we ate the whole thing, that means there was just trailing
408 * whitespace or a final / no-op semicolon. either way, get out */
409 if (strlen(mystr) == 0)
411 /* if this is a comment, jump past it and keep going */
412 if (mystr[0] == '#') {
413 mystr = strpbrk(mystr, ";\n\r");
417 sed_cmds = xrealloc(sed_cmds, sizeof(struct sed_cmd) * (++ncmds));
418 /* zero new element */
419 memset(&sed_cmds[ncmds-1], 0, sizeof(struct sed_cmd));
420 /* load command string into new array element, get remainder */
421 mystr = parse_cmd_str(&sed_cmds[ncmds-1], mystr);
423 } while (mystr && strlen(mystr));
427 static void load_cmd_file(char *filename)
433 cmdfile = xfopen(filename, "r");
435 while ((line = get_line_from_file(cmdfile)) != NULL) {
436 /* if a line ends with '\' it needs the next line appended to it */
437 while (line[strlen(line)-2] == '\\' &&
438 (nextline = get_line_from_file(cmdfile)) != NULL) {
439 line = xrealloc(line, strlen(line) + strlen(nextline) + 1);
440 strcat(line, nextline);
443 /* eat trailing newline (if any) --if I don't do this, edit commands
444 * (aic) will print an extra newline */
451 static void print_subst_w_backrefs(const char *line, const char *replace, regmatch_t *regmatch)
455 /* go through the replacement string */
456 for (i = 0; replace[i]; i++) {
457 /* if we find a backreference (\1, \2, etc.) print the backref'ed * text */
458 if (replace[i] == '\\' && isdigit(replace[i+1])) {
462 ++i; /* i now indexes the backref number, instead of the leading slash */
463 tmpstr[0] = replace[i];
465 backref = atoi(tmpstr);
466 /* print out the text held in regmatch[backref] */
467 for (j = regmatch[backref].rm_so; j < regmatch[backref].rm_eo; j++)
468 fputc(line[j], stdout);
471 /* if we find a backslash escaped character, print the character */
472 else if (replace[i] == '\\') {
474 fputc(replace[i], stdout);
477 /* if we find an unescaped '&' print out the whole matched text.
478 * fortunately, regmatch[0] contains the indicies to the whole matched
479 * expression (kinda seems like it was designed for just such a
481 else if (replace[i] == '&' && replace[i-1] != '\\') {
483 for (j = regmatch[0].rm_so; j < regmatch[0].rm_eo; j++)
484 fputc(line[j], stdout);
486 /* nothing special, just print this char of the replacement string to stdout */
488 fputc(replace[i], stdout);
492 static int do_subst_command(const struct sed_cmd *sed_cmd, const char *line)
494 char *hackline = (char *)line;
496 regmatch_t *regmatch = NULL;
498 /* we only proceed if the substitution 'search' expression matches */
499 if (regexec(sed_cmd->sub_match, line, 0, NULL, 0) == REG_NOMATCH)
502 /* whaddaya know, it matched. get the number of back references */
503 regmatch = xmalloc(sizeof(regmatch_t) * (sed_cmd->num_backrefs+1));
505 /* and now, as long as we've got a line to try matching and if we can match
506 * the search string, we make substitutions */
507 while (*hackline && (regexec(sed_cmd->sub_match, hackline,
508 sed_cmd->num_backrefs+1, regmatch, 0) == 0) ) {
511 /* print everything before the match */
512 for (i = 0; i < regmatch[0].rm_so; i++)
513 fputc(hackline[i], stdout);
515 /* then print the substitution string */
516 print_subst_w_backrefs(hackline, sed_cmd->replace, regmatch);
518 /* advance past the match */
519 hackline += regmatch[0].rm_eo;
520 /* flag that something has changed */
523 /* if we're not doing this globally, get out now */
528 /* if there's anything left of the line, print it */
530 fputs(hackline, stdout);
538 static int do_sed_command(const struct sed_cmd *sed_cmd, const char *line)
542 switch (sed_cmd->cmd) {
555 * Some special cases for 's' printing to make it compliant with
556 * GNU sed printing behavior (aka "The -n | s///p Matrix"):
558 * -n ONLY = never print anything regardless of any successful
561 * s///p ONLY = always print successful substitutions, even if
562 * the line is going to be printed anyway (line will be printed
565 * -n AND s///p = print ONLY a successful substitution ONE TIME;
566 * no other lines are printed - this is the reason why the 'p'
567 * flag exists in the first place.
570 /* if the user specified that they didn't want anything printed (i.e. a -n
571 * flag and no 'p' flag after the s///), then there's really no point doing
573 if (be_quiet && !sed_cmd->sub_p)
576 /* we print the line once, unless we were told to be quiet */
578 altered = do_subst_command(sed_cmd, line);
580 /* we also print the line if we were given the 'p' flag
581 * (this is quite possibly the second printing) */
583 altered = do_subst_command(sed_cmd, line);
589 fputs(sed_cmd->editline, stdout);
594 fputs(sed_cmd->editline, stdout);
598 fputs(sed_cmd->editline, stdout);
606 static void process_file(FILE *file)
609 static int linenum = 0; /* GNU sed does not restart counting lines at EOF */
610 unsigned int still_in_range = 0;
614 /* go through every line in the file */
615 while ((line = get_line_from_file(file)) != NULL) {
620 /* for every line, go through all the commands */
621 for (i = 0; i < ncmds; i++) {
623 /* are we acting on a range of matched lines? */
624 if (sed_cmds[i].beg_match && sed_cmds[i].end_match) {
625 if (still_in_range || regexec(sed_cmds[i].beg_match, line, 0, NULL, 0) == 0) {
626 line_altered += do_sed_command(&sed_cmds[i], line);
627 if (still_in_range && regexec(sed_cmds[i].end_match, line, 0, NULL, 0) == 0)
634 /* are we trying to match a single line? */
635 else if (sed_cmds[i].beg_match) {
636 if (regexec(sed_cmds[i].beg_match, line, 0, NULL, 0) == 0)
637 line_altered += do_sed_command(&sed_cmds[i], line);
640 /* are we acting on a range of line numbers? */
641 else if (sed_cmds[i].beg_line > 0 && sed_cmds[i].end_line != 0) {
642 if (linenum >= sed_cmds[i].beg_line &&
643 (sed_cmds[i].end_line == -1 || linenum <= sed_cmds[i].end_line))
644 line_altered += do_sed_command(&sed_cmds[i], line);
647 /* are we acting on a specified line number */
648 else if (sed_cmds[i].beg_line > 0) {
649 if (linenum == sed_cmds[i].beg_line)
650 line_altered += do_sed_command(&sed_cmds[i], line);
653 /* not acting on matches or line numbers. act on every line */
655 line_altered += do_sed_command(&sed_cmds[i], line);
659 /* we will print the line unless we were told to be quiet or if the
660 * line was altered (via a 'd'elete or 's'ubstitution), in which case
661 * the altered line was already printed */
662 if (!be_quiet && !line_altered)
669 extern int sed_main(int argc, char **argv)
673 #ifdef BB_FEATURE_CLEAN_UP
674 /* destroy command strings on exit */
675 if (atexit(destroy_cmd_strs) == -1)
676 perror_msg_and_die("atexit");
679 /* do normal option parsing */
680 while ((opt = getopt(argc, argv, "ne:f:")) > 0) {
689 load_cmd_file(optarg);
696 /* if we didn't get a pattern from a -e and no command file was specified,
697 * argv[optind] should be the pattern. no pattern, no worky */
699 if (argv[optind] == NULL)
702 add_cmd_str(argv[optind]);
708 /* argv[(optind)..(argc-1)] should be names of file to process. If no
709 * files were specified or '-' was specified, take input from stdin.
710 * Otherwise, we process all the files specified. */
711 if (argv[optind] == NULL || (strcmp(argv[optind], "-") == 0)) {
717 for (i = optind; i < argc; i++) {
718 file = fopen(argv[i], "r");
720 perror_msg("%s", argv[i]);