/*
* sed.c - very minimalist version of sed
*
- * Copyright (C) 1999,2000 by Lineo, inc.
- * Written by Mark Whitley <markw@lineo.com>, <markw@enol.com>
+ * Copyright (C) 1999,2000,2001 by Lineo, inc.
+ * Written by Mark Whitley <markw@lineo.com>, <markw@codepoet.org>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
- address matching: num|/matchstr/[,num|/matchstr/|$]command
- commands: (p)rint, (d)elete, (s)ubstitue (with g & I flags)
- edit commands: (a)ppend, (i)nsert, (c)hange
+ - file commands: (r)ead
- backreferences in substitution expressions (\1, \2...\9)
(Note: Specifying an address (range) to match is *optional*; commands
Unsupported features:
- transliteration (y/source-chars/dest-chars/) (use 'tr')
- - no support for characters other than the '/' character for regex matches
- no pattern space hold space storing / swapping (x, etc.)
- no labels / branching (: label, b, t, and friends)
- and lots, lots more.
-
*/
#include <stdio.h>
#include <string.h> /* for strdup() */
#include <errno.h>
#include <ctype.h> /* for isspace() */
+#include <stdlib.h>
#include "busybox.h"
-#define bb_need_full_version
-#define BB_DECLARE_EXTERN
-#include "messages.c"
-
/* externs */
+extern void xregcomp(regex_t *preg, const char *regex, int cflags);
extern int optind; /* in unistd.h */
extern char *optarg; /* ditto */
/* GENERAL FIELDS */
+ char delimiter; /* The delimiter used to separate regexps */
/* address storage */
int beg_line; /* 'sed 1p' 0 == no begining line, apply commands to all lines */
/* EDIT COMMAND (a,i,c) SPEICIFIC FIELDS */
char *editline;
+
+
+ /* FILE COMMAND (r) SPEICIFIC FIELDS */
+
+ char *filename;
};
/* globals */
}
#endif
-#if 0
-/*
- * trim_str - trims leading and trailing space from a string
- *
- * Note: This returns a malloc'ed string so you must store and free it
- * XXX: This should be in the utility.c file.
- * XXX: This is now obsolete. Maybe it belongs nowhere.
- */
-static char *trim_str(const char *str)
-{
- int i;
- char *retstr = strdup(str);
-
- /* trim leading whitespace */
- memmove(retstr, &retstr[strspn(retstr, " \n\t\v")], strlen(retstr));
-
- /* trim trailing whitespace */
- i = strlen(retstr) - 1;
- while (isspace(retstr[i]))
- i--;
- retstr[++i] = 0;
-
- /* Aside:
- *
- * you know, a strrspn() would really be nice cuz then we could say:
- *
- * retstr[strrspn(retstr, " \n\t\v") + 1] = 0;
- */
-
- return retstr;
-}
-#endif
-
-#if 0
-/*
- * strrspn - works just like strspn() but goes from right to left instead of
- * left to right
- */
-static size_t strrspn(const char *s, const char *accept)
-{
- size_t i = strlen(s);
-
- while (strchr(accept, s[--i]))
- ;
-
- return i;
-}
-#endif
/*
- * index_of_next_unescaped_slash - walks left to right through a string
- * beginning at a specified index and returns the index of the next forward
- * slash ('/') not preceeded by a backslash ('\').
+ * index_of_next_unescaped_regexp_delim - walks left to right through a string
+ * beginning at a specified index and returns the index of the next regular
+ * expression delimiter (typically a forward * slash ('/')) not preceeded by
+ * a backslash ('\').
*/
-static int index_of_next_unescaped_slash(const char *str, int idx)
+static int index_of_next_unescaped_regexp_delim(struct sed_cmd *sed_cmd, const char *str, int idx)
{
for ( ; str[idx]; idx++) {
- if (str[idx] == '/' && str[idx-1] != '\\')
+ if (str[idx] == sed_cmd->delimiter && str[idx-1] != '\\')
return idx;
}
/*
* returns the index in the string just past where the address ends.
*/
-static int get_address(const char *str, int *line, regex_t **regex)
+static int get_address(struct sed_cmd *sed_cmd, const char *str, int *linenum, regex_t **regex)
{
char *my_str = strdup(str);
int idx = 0;
+ char olddelimiter;
+ olddelimiter = sed_cmd->delimiter;
+ sed_cmd->delimiter = '/';
if (isdigit(my_str[idx])) {
do {
idx++;
} while (isdigit(my_str[idx]));
my_str[idx] = 0;
- *line = atoi(my_str);
+ *linenum = atoi(my_str);
}
else if (my_str[idx] == '$') {
- *line = -1;
+ *linenum = -1;
idx++;
}
else if (my_str[idx] == '/') {
- idx = index_of_next_unescaped_slash(my_str, ++idx);
+ idx = index_of_next_unescaped_regexp_delim(sed_cmd, my_str, ++idx);
if (idx == -1)
- fatalError("unterminated match expression\n");
+ error_msg_and_die("unterminated match expression");
my_str[idx] = '\0';
*regex = (regex_t *)xmalloc(sizeof(regex_t));
xregcomp(*regex, my_str+1, REG_NEWLINE);
idx++; /* so it points to the next character after the last '/' */
}
else {
- errorMsg("get_address: no address found in string\n"
- "\t(you probably didn't check the string you passed me)\n");
+ error_msg("get_address: no address found in string\n"
+ "\t(you probably didn't check the string you passed me)");
idx = -1;
}
free(my_str);
+ sed_cmd->delimiter = olddelimiter;
return idx;
}
-static char *strdup_substr(const char *str, int start, int end)
-{
- int size = end - start + 1;
- char *newstr = xmalloc(size);
- memcpy(newstr, str+start, size-1);
- newstr[size-1] = '\0';
- return newstr;
-}
-
static int parse_subst_cmd(struct sed_cmd *sed_cmd, const char *substr)
{
int oldidx, cflags = REG_NEWLINE;
* (all three of the '/' slashes are mandatory)
*/
- /* verify that the 's' is followed by a 'slash' */
- if (substr[++idx] != '/')
- fatalError("bad format in substitution expression\n");
+ /* verify that the 's' is followed by something. That something
+ * (typically a 'slash') is now our regexp delimiter... */
+ if (!substr[++idx])
+ error_msg_and_die("bad format in substitution expression");
+ else
+ sed_cmd->delimiter=substr[idx];
/* save the match string */
oldidx = idx+1;
- idx = index_of_next_unescaped_slash(substr, ++idx);
+ idx = index_of_next_unescaped_regexp_delim(sed_cmd, substr, ++idx);
if (idx == -1)
- fatalError("bad format in substitution expression\n");
- match = strdup_substr(substr, oldidx, idx);
+ error_msg_and_die("bad format in substitution expression");
+ match = xstrndup(substr + oldidx, idx - oldidx);
/* determine the number of back references in the match string */
/* Note: we compute this here rather than in the do_subst_command()
/* save the replacement string */
oldidx = idx+1;
- idx = index_of_next_unescaped_slash(substr, ++idx);
+ idx = index_of_next_unescaped_regexp_delim(sed_cmd, substr, ++idx);
if (idx == -1)
- fatalError("bad format in substitution expression\n");
- sed_cmd->replace = strdup_substr(substr, oldidx, idx);
+ error_msg_and_die("bad format in substitution expression");
+ sed_cmd->replace = xstrndup(substr + oldidx, idx - oldidx);
/* process the flags */
while (substr[++idx]) {
if (strchr("; \t\v\n\r", substr[idx]))
goto out;
/* else */
- fatalError("bad option in substitution expression\n");
+ error_msg_and_die("bad option in substitution expression");
}
}
*/
if (editstr[1] != '\\' && (editstr[2] != '\n' || editstr[2] != '\r'))
- fatalError("bad format in edit expression\n");
+ error_msg_and_die("bad format in edit expression");
/* store the edit line text */
/* make editline big enough to accomodate the extra '\n' we will tack on
/* now we need to go through * and: s/\\[\r\n]$/\n/g on the edit line */
while (ptr[idx]) {
- while (ptr[idx] != '\\' && (ptr[idx+1] != '\n' || ptr[idx+1] != '\r')) {
+ while (ptr[idx] != '\\' || (ptr[idx+1] != '\n' && ptr[idx+1] != '\r')) {
idx++;
if (!ptr[idx]) {
goto out;
}
out:
- ptr[idx] = '\n';
- ptr[idx+1] = 0;
-
/* this accounts for discrepancies between the modified string and the
* original string passed in to this function */
idx += slashes_eaten;
- /* this accounts for the fact that A) we started at index 3, not at index
- * 0 and B) that we added an extra '\n' at the end (if you think the next
- * line should read 'idx += 4' remember, arrays are zero-based) */
+ /* figure out if we need to add a newline */
+ if (ptr[idx-1] != '\n') {
+ ptr[idx] = '\n';
+ idx++;
+ }
- idx += 3;
+ /* terminate string */
+ ptr[idx]= 0;
+ /* adjust for opening 2 chars [aic]\ */
+ idx += 2;
return idx;
}
+
+static int parse_file_cmd(struct sed_cmd *sed_cmd, const char *filecmdstr)
+{
+ int idx = 0;
+ int filenamelen = 0;
+
+ /*
+ * the string that gets passed to this function should look like this:
+ * '[ ]filename'
+ * | |
+ * | a filename
+ * |
+ * optional whitespace
+
+ * re: the file to be read, the GNU manual says the following: "Note that
+ * if filename cannot be read, it is treated as if it were an empty file,
+ * without any error indication." Thus, all of the following commands are
+ * perfectly leagal:
+ *
+ * sed -e '1r noexist'
+ * sed -e '1r ;'
+ * sed -e '1r'
+ */
+
+ /* the file command may be followed by whitespace; move past it. */
+ while (isspace(filecmdstr[++idx]))
+ { ; }
+
+ /* the first non-whitespace we get is a filename. the filename ends when we
+ * hit a normal sed command terminator or end of string */
+ filenamelen = strcspn(&filecmdstr[idx], "; \n\r\t\v\0");
+ sed_cmd->filename = xmalloc(filenamelen + 1);
+ safe_strncpy(sed_cmd->filename, &filecmdstr[idx], filenamelen + 1);
+
+ return idx + filenamelen;
+}
+
+
static char *parse_cmd_str(struct sed_cmd *sed_cmd, const char *cmdstr)
{
int idx = 0;
* part1 part2 part3
*/
-
/* first part (if present) is an address: either a number or a /regex/ */
if (isdigit(cmdstr[idx]) || cmdstr[idx] == '/')
- idx = get_address(cmdstr, &sed_cmd->beg_line, &sed_cmd->beg_match);
+ idx = get_address(sed_cmd, cmdstr, &sed_cmd->beg_line, &sed_cmd->beg_match);
/* second part (if present) will begin with a comma */
if (cmdstr[idx] == ',')
- idx += get_address(&cmdstr[++idx], &sed_cmd->end_line, &sed_cmd->end_match);
+ idx += get_address(sed_cmd, &cmdstr[++idx], &sed_cmd->end_line, &sed_cmd->end_match);
/* last part (mandatory) will be a command */
if (cmdstr[idx] == '\0')
- fatalError("missing command\n");
- if (!strchr("pdsaic", cmdstr[idx])) /* <-- XXX add new commands here */
- fatalError("invalid command\n");
+ error_msg_and_die("missing command");
sed_cmd->cmd = cmdstr[idx];
- /* special-case handling for (s)ubstitution */
- if (sed_cmd->cmd == 's') {
+ /* if it was a single-letter command that takes no arguments (such as 'p'
+ * or 'd') all we need to do is increment the index past that command */
+ if (strchr("pd", cmdstr[idx])) {
+ idx++;
+ }
+ /* handle (s)ubstitution command */
+ else if (sed_cmd->cmd == 's') {
idx += parse_subst_cmd(sed_cmd, &cmdstr[idx]);
}
- /* special-case handling for (a)ppend, (i)nsert, and (c)hange */
- else if (strchr("aic", cmdstr[idx])) {
- if (sed_cmd->end_line || sed_cmd->end_match)
- fatalError("only a beginning address can be specified for edit commands\n");
+ /* handle edit cmds: (a)ppend, (i)nsert, and (c)hange */
+ else if (strchr("aic", sed_cmd->cmd)) {
+ if ((sed_cmd->end_line || sed_cmd->end_match) && sed_cmd->cmd != 'c')
+ error_msg_and_die("only a beginning address can be specified for edit commands");
idx += parse_edit_cmd(sed_cmd, &cmdstr[idx]);
}
- /* if it was a single-letter command (such as 'p' or 'd') we need to
- * increment the index past that command */
- else
- idx++;
+ /* handle file cmds: (r)ead */
+ else if (sed_cmd->cmd == 'r') {
+ if (sed_cmd->end_line || sed_cmd->end_match)
+ error_msg_and_die("Command only uses one address");
+ idx += parse_file_cmd(sed_cmd, &cmdstr[idx]);
+ }
+ else {
+ error_msg_and_die("invalid command");
+ }
/* give back whatever's left over */
return (char *)&cmdstr[idx];
}
/* eat trailing newline (if any) --if I don't do this, edit commands
* (aic) will print an extra newline */
- if (line[strlen(line)-1] == '\n')
- line[strlen(line)-1] = 0;
+ chomp(line);
add_cmd_str(line);
free(line);
}
int altered = 0;
regmatch_t *regmatch = NULL;
- /* if the user specified that they didn't want anything printed (i.e. a -n
- * flag and no 'p' flag after the s///), then there's really no point doing
- * anything here. */
- if (be_quiet && !sed_cmd->sub_p)
- return 0;
-
/* we only proceed if the substitution 'search' expression matches */
if (regexec(sed_cmd->sub_match, line, 0, NULL, 0) == REG_NOMATCH)
return 0;
break;
}
- /* if there's anything left of the line, print it */
- if (*hackline)
- fputs(hackline, stdout);
+ puts(hackline);
/* cleanup */
free(regmatch);
return altered;
}
-static int do_sed_command(const struct sed_cmd *sed_cmd, const char *line)
-{
- int altered = 0;
-
- switch (sed_cmd->cmd) {
-
- case 'p':
- fputs(line, stdout);
- break;
-
- case 'd':
- altered++;
- break;
-
- case 's':
-
- /*
- * Some special cases for 's' printing to make it compliant with
- * GNU sed printing behavior (aka "The -n | s///p Matrix"):
- *
- * -n ONLY = never print anything regardless of any successful
- * substitution
- *
- * s///p ONLY = always print successful substitutions, even if
- * the line is going to be printed anyway (line will be printed
- * twice).
- *
- * -n AND s///p = print ONLY a successful substitution ONE TIME;
- * no other lines are printed - this is the reason why the 'p'
- * flag exists in the first place.
- */
-
- /* we print the line once, unless we were told to be quiet */
- if (!be_quiet)
- altered = do_subst_command(sed_cmd, line);
-
- /* we also print the line if we were given the 'p' flag
- * (this is quite possibly the second printing) */
- if (sed_cmd->sub_p)
- altered = do_subst_command(sed_cmd, line);
-
- break;
-
- case 'a':
- fputs(line, stdout);
- fputs(sed_cmd->editline, stdout);
- altered++;
- break;
-
- case 'i':
- fputs(sed_cmd->editline, stdout);
- break;
-
- case 'c':
- fputs(sed_cmd->editline, stdout);
- altered++;
- break;
- }
-
- return altered;
-}
static void process_file(FILE *file)
{
char *line = NULL;
static int linenum = 0; /* GNU sed does not restart counting lines at EOF */
unsigned int still_in_range = 0;
- int line_altered;
+ int altered;
int i;
/* go through every line in the file */
while ((line = get_line_from_file(file)) != NULL) {
+ chomp(line);
linenum++;
- line_altered = 0;
+ altered = 0;
/* for every line, go through all the commands */
for (i = 0; i < ncmds; i++) {
- /* are we acting on a range of matched lines? */
- if (sed_cmds[i].beg_match && sed_cmds[i].end_match) {
- if (still_in_range || regexec(sed_cmds[i].beg_match, line, 0, NULL, 0) == 0) {
- line_altered += do_sed_command(&sed_cmds[i], line);
- still_in_range = 1;
- if (regexec(sed_cmds[i].end_match, line, 0, NULL, 0) == 0)
- still_in_range = 0;
- }
- }
- /* are we trying to match a single line? */
- else if (sed_cmds[i].beg_match) {
- if (regexec(sed_cmds[i].beg_match, line, 0, NULL, 0) == 0)
- line_altered += do_sed_command(&sed_cmds[i], line);
- }
+ /*
+ * entry point into sedding...
+ */
+ if (
+ /* no range necessary */
+ (sed_cmds[i].beg_line == 0 && sed_cmds[i].end_line == 0 &&
+ sed_cmds[i].beg_match == NULL &&
+ sed_cmds[i].end_match == NULL) ||
+ /* this line number is the first address we're looking for */
+ (sed_cmds[i].beg_line && (sed_cmds[i].beg_line == linenum)) ||
+ /* this line matches our first address regex */
+ (sed_cmds[i].beg_match && (regexec(sed_cmds[i].beg_match, line, 0, NULL, 0) == 0)) ||
+ /* we are currently within the beginning & ending address range */
+ still_in_range
+ ) {
+
+ /*
+ * actual sedding
+ */
+ switch (sed_cmds[i].cmd) {
+
+ case 'p':
+ puts(line);
+ break;
+
+ case 'd':
+ altered++;
+ break;
+
+ case 's':
+
+ /*
+ * Some special cases for 's' printing to make it compliant with
+ * GNU sed printing behavior (aka "The -n | s///p Matrix"):
+ *
+ * -n ONLY = never print anything regardless of any successful
+ * substitution
+ *
+ * s///p ONLY = always print successful substitutions, even if
+ * the line is going to be printed anyway (line will be printed
+ * twice).
+ *
+ * -n AND s///p = print ONLY a successful substitution ONE TIME;
+ * no other lines are printed - this is the reason why the 'p'
+ * flag exists in the first place.
+ */
+
+ /* if the user specified that they didn't want anything printed (i.e., a -n
+ * flag and no 'p' flag after the s///), then there's really no point doing
+ * anything here. */
+ if (be_quiet && !sed_cmds[i].sub_p)
+ break;
+
+ /* we print the line once, unless we were told to be quiet */
+ if (!be_quiet)
+ altered |= do_subst_command(&sed_cmds[i], line);
+
+ /* we also print the line if we were given the 'p' flag
+ * (this is quite possibly the second printing) */
+ if (sed_cmds[i].sub_p)
+ altered |= do_subst_command(&sed_cmds[i], line);
+
+ break;
+
+ case 'a':
+ puts(line);
+ fputs(sed_cmds[i].editline, stdout);
+ altered++;
+ break;
+
+ case 'i':
+ fputs(sed_cmds[i].editline, stdout);
+ break;
+
+ case 'c':
+ /* single-address case */
+ if (sed_cmds[i].end_match == NULL && sed_cmds[i].end_line == 0) {
+ fputs(sed_cmds[i].editline, stdout);
+ }
+ /* multi-address case */
+ else {
+ /* matching text */
+ if (sed_cmds[i].end_match && (regexec(sed_cmds[i].end_match, line, 0, NULL, 0) == 0))
+ fputs(sed_cmds[i].editline, stdout);
+ /* matching line numbers */
+ if (sed_cmds[i].end_line > 0 && sed_cmds[i].end_line == linenum)
+ fputs(sed_cmds[i].editline, stdout);
+ }
+ altered++;
+
+ break;
+
+ case 'r': {
+ FILE *outfile;
+ puts(line);
+ outfile = fopen(sed_cmds[i].filename, "r");
+ if (outfile)
+ print_file(outfile);
+ /* else if we couldn't open the output file,
+ * no biggie, just don't print anything */
+ altered++;
+ }
+ break;
+ }
- /* are we acting on a range of line numbers? */
- else if (sed_cmds[i].beg_line > 0 && sed_cmds[i].end_line != 0) {
- if (linenum >= sed_cmds[i].beg_line &&
- (sed_cmds[i].end_line == -1 || linenum <= sed_cmds[i].end_line))
- line_altered += do_sed_command(&sed_cmds[i], line);
- }
+ /*
+ * exit point from sedding...
+ */
+ if (
+ /* this is a single-address command or... */
+ (sed_cmds[i].end_line == 0 && sed_cmds[i].end_match == NULL) || (
+ /* we were in the middle of our address range (this
+ * isn't the first time through) and.. */
+ (still_in_range == 1) && (
+ /* this line number is the last address we're looking for or... */
+ (sed_cmds[i].end_line && (sed_cmds[i].end_line == linenum)) ||
+ /* this line matches our last address regex */
+ (sed_cmds[i].end_match && (regexec(sed_cmds[i].end_match, line, 0, NULL, 0) == 0))
+ )
+ )
+ ) {
+ /* we're out of our address range */
+ still_in_range = 0;
+ }
- /* are we acting on a specified line number */
- else if (sed_cmds[i].beg_line > 0) {
- if (linenum == sed_cmds[i].beg_line)
- line_altered += do_sed_command(&sed_cmds[i], line);
+ /* didn't hit the exit? then we're still in the middle of an address range */
+ else {
+ still_in_range = 1;
+ }
}
-
- /* not acting on matches or line numbers. act on every line */
- else
- line_altered += do_sed_command(&sed_cmds[i], line);
-
}
/* we will print the line unless we were told to be quiet or if the
* line was altered (via a 'd'elete or 's'ubstitution), in which case
* the altered line was already printed */
- if (!be_quiet && !line_altered)
- fputs(line, stdout);
+ if (!be_quiet && !altered)
+ puts(line);
free(line);
}
#ifdef BB_FEATURE_CLEAN_UP
/* destroy command strings on exit */
- if (atexit(destroy_cmd_strs) == -1) {
- perror("sed");
- exit(1);
- }
+ if (atexit(destroy_cmd_strs) == -1)
+ perror_msg_and_die("atexit");
#endif
/* do normal option parsing */
- while ((opt = getopt(argc, argv, "hne:f:")) > 0) {
+ while ((opt = getopt(argc, argv, "ne:f:")) > 0) {
switch (opt) {
- case 'h':
- usage(sed_usage);
- break;
case 'n':
be_quiet++;
break;
case 'f':
load_cmd_file(optarg);
break;
+ default:
+ show_usage();
}
}
* argv[optind] should be the pattern. no pattern, no worky */
if (ncmds == 0) {
if (argv[optind] == NULL)
- usage(sed_usage);
+ show_usage();
else {
add_cmd_str(argv[optind]);
optind++;
for (i = optind; i < argc; i++) {
file = fopen(argv[i], "r");
if (file == NULL) {
- errorMsg("%s: %s\n", argv[i], strerror(errno));
+ perror_msg("%s", argv[i]);
} else {
process_file(file);
fclose(file);