+/* vi: set sw=4 ts=4: */
/*
* sed.c - very minimalist version of sed
*
- * Copyright (C) 1999,2000,2001 by Lineo, inc.
- * Written by Mark Whitley <markw@lineo.com>, <markw@codepoet.org>
+ * Copyright (C) 1999,2000,2001 by Lineo, inc. and Mark Whitley
+ * Copyright (C) 1999,2000,2001 by Mark Whitley <markw@codepoet.org>
+ * Copyright (C) 2002 Matt Kraai
+ * Copyright (C) 2003 by Glenn McGrath <bug1@iinet.net.au>
+ * Copyright (C) 2003,2004 by Rob Landley <rob@landley.net>
*
* 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
*
*/
+/* Code overview.
+
+ Files are laid out to avoid unnecessary function declarations. So for
+ example, every function add_cmd calls occurs before add_cmd in this file.
+
+ add_cmd() is called on each line of sed command text (from a file or from
+ the command line). It calls get_address() and parse_cmd_args(). The
+ resulting sed_cmd_t structures are appended to a linked list
+ (sed_cmd_head/sed_cmd_tail).
+
+ process_file() does actual sedding, reading data lines from an input FILE *
+ (which could be stdin) and applying the sed command list (sed_cmd_head) to
+ each of the resulting lines.
+
+ sed_main() is where external code calls into this, with a command line.
+*/
+
+
/*
Supported features and commands in this version of sed:
- edit commands: (a)ppend, (i)nsert, (c)hange
- file commands: (r)ead
- backreferences in substitution expressions (\1, \2...\9)
-
+ - grouped commands: {cmd1;cmd2}
+ - transliteration (y/source-chars/dest-chars/)
+ - pattern space hold space storing / swapping (g, h, x)
+ - labels / branching (: label, b, t)
+
(Note: Specifying an address (range) to match is *optional*; commands
default to the whole pattern space if no specific address match was
requested.)
Unsupported features:
- - transliteration (y/source-chars/dest-chars/) (use 'tr')
- - no pattern space hold space storing / swapping (x, etc.)
- - no labels / branching (: label, b, t, and friends)
- - and lots, lots more.
+ - GNU extensions
+ - and more.
+
+ Todo:
+
+ - Create a wrapper around regex to make libc's regex conform with sed
+ - Fix bugs
+
+
+ Reference http://www.opengroup.org/onlinepubs/007904975/utilities/sed.html
*/
#include <stdio.h>
-#include <unistd.h> /* for getopt() */
+#include <unistd.h> /* for getopt() */
#include <regex.h>
-#include <string.h> /* for strdup() */
+#include <string.h> /* for strdup() */
#include <errno.h>
-#include <ctype.h> /* for isspace() */
+#include <ctype.h> /* for isspace() */
#include <stdlib.h>
#include "busybox.h"
-/* externs */
-extern void xregcomp(regex_t *preg, const char *regex, int cflags);
-extern int optind; /* in unistd.h */
-extern char *optarg; /* ditto */
+typedef struct sed_cmd_s {
+ /* Ordered by alignment requirements: currently 36 bytes on x86 */
-/* options */
-static int be_quiet = 0;
+ /* address storage */
+ regex_t *beg_match; /* sed -e '/match/cmd' */
+ regex_t *end_match; /* sed -e '/match/,/end_match/cmd' */
+ regex_t *sub_match; /* For 's/sub_match/string/' */
+ int beg_line; /* 'sed 1p' 0 == apply commands to all lines */
+ int end_line; /* 'sed 1,3p' 0 == one line only. -1 = last line ($) */
+ FILE *file; /* File (sr) command writes to, -1 for none. */
+ char *string; /* Data string for (saicytb) commands. */
-struct sed_cmd {
+ unsigned short which_match; /* (s) Which match to replace (0 for all) */
+ /* Bitfields (gcc won't group them if we don't) */
+ unsigned int invert:1; /* the '!' after the address */
+ unsigned int in_match:1; /* Next line also included in match? */
+ unsigned int no_newline:1; /* Last line written by (sr) had no '\n' */
+ unsigned int sub_p:1; /* (s) print option */
- /* 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 */
- int end_line; /* 'sed 1,3p' 0 == no end line, use only beginning. -1 == $ */
- regex_t *beg_match; /* sed -e '/match/cmd' */
- regex_t *end_match; /* sed -e '/match/,/end_match/cmd' */
+ /* GENERAL FIELDS */
+ char cmd; /* The command char: abcdDgGhHilnNpPqrstwxy:={} */
+ struct sed_cmd_s *next; /* Next command (linked list, NULL terminated) */
+} sed_cmd_t;
- /* the command */
- char cmd; /* p,d,s (add more at your leisure :-) */
+/* globals */
+/* options */
+static int be_quiet = 0, in_place=0;
+FILE *nonstdout;
+char *outname;
- /* SUBSTITUTION COMMAND SPECIFIC FIELDS */
+static const char bad_format_in_subst[] =
+ "bad format in substitution expression";
+const char *const semicolon_whitespace = "; \n\r\t\v";
- /* sed -e 's/sub_match/replace/' */
- regex_t *sub_match;
- char *replace;
- unsigned int num_backrefs:4; /* how many back references (\1..\9) */
- /* Note: GNU/POSIX sed does not save more than nine backrefs, so
- * we only use 4 bits to hold the number */
- unsigned int sub_g:1; /* sed -e 's/foo/bar/g' (global) */
- unsigned int sub_p:2; /* sed -e 's/foo/bar/p' (print substitution) */
+regmatch_t regmatch[10];
+static regex_t *previous_regex_ptr = NULL;
+/* linked list of sed commands */
+static sed_cmd_t sed_cmd_head;
+static sed_cmd_t *sed_cmd_tail = &sed_cmd_head;
- /* EDIT COMMAND (a,i,c) SPEICIFIC FIELDS */
+/* Linked list of append lines */
+struct append_list {
+ char *string;
+ struct append_list *next;
+};
+struct append_list *append_head=NULL, *append_tail=NULL;
- char *editline;
+#ifdef CONFIG_FEATURE_CLEAN_UP
+static void free_and_close_stuff(void)
+{
+ sed_cmd_t *sed_cmd = sed_cmd_head.next;
+ while(append_head) {
+ append_tail=append_head->next;
+ free(append_head->string);
+ free(append_head);
+ append_head=append_tail;
+ }
- /* FILE COMMAND (r) SPEICIFIC FIELDS */
+ while (sed_cmd) {
+ sed_cmd_t *sed_cmd_next = sed_cmd->next;
- char *filename;
-};
+ if(sed_cmd->file)
+ bb_xprint_and_close_file(sed_cmd->file);
-/* globals */
-static struct sed_cmd *sed_cmds = NULL; /* growable arrary holding a sequence of sed cmds */
-static int ncmds = 0; /* number of sed commands */
+ if (sed_cmd->beg_match) {
+ regfree(sed_cmd->beg_match);
+ free(sed_cmd->beg_match);
+ }
+ if (sed_cmd->end_match) {
+ regfree(sed_cmd->end_match);
+ free(sed_cmd->end_match);
+ }
+ if (sed_cmd->sub_match) {
+ regfree(sed_cmd->sub_match);
+ free(sed_cmd->sub_match);
+ }
+ free(sed_cmd->string);
+ free(sed_cmd);
+ sed_cmd = sed_cmd_next;
+ }
+}
+#endif
-/*static char *cur_file = NULL;*/ /* file currently being processed XXX: do I need this? */
+/* If something bad happens during -i operation, delete temp file */
-#ifdef BB_FEATURE_CLEAN_UP
-static void destroy_cmd_strs()
+static void cleanup_outname(void)
{
- if (sed_cmds == NULL)
- return;
+ if(outname) unlink(outname);
+}
- /* destroy all the elements in the array */
- while (--ncmds >= 0) {
+/* strdup, replacing "\n" with '\n', and "\delimiter" with 'delimiter' */
- if (sed_cmds[ncmds].beg_match) {
- regfree(sed_cmds[ncmds].beg_match);
- free(sed_cmds[ncmds].beg_match);
- }
- if (sed_cmds[ncmds].end_match) {
- regfree(sed_cmds[ncmds].end_match);
- free(sed_cmds[ncmds].end_match);
- }
- if (sed_cmds[ncmds].sub_match) {
- regfree(sed_cmds[ncmds].sub_match);
- free(sed_cmds[ncmds].sub_match);
+static void parse_escapes(char *dest, const char *string, int len, char from, char to)
+{
+ int i=0;
+
+ while(i<len) {
+ if(string[i] == '\\') {
+ if(!to || string[i+1] == from) {
+ *(dest++) = to ? to : string[i+1];
+ i+=2;
+ continue;
+ } else *(dest++)=string[i++];
}
- if (sed_cmds[ncmds].replace)
- free(sed_cmds[ncmds].replace);
+ *(dest++) = string[i++];
}
+ *dest=0;
+}
+
+static char *copy_parsing_slashn(const char *string, int len)
+{
+ char *dest=xmalloc(len+1);
- /* destroy the array */
- free(sed_cmds);
- sed_cmds = NULL;
+ parse_escapes(dest,string,len,'n','\n');
+ return dest;
}
-#endif
/*
* 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
+ * expression delimiter (typically a forward * slash ('/')) not preceded by
* a backslash ('\').
*/
-static int index_of_next_unescaped_regexp_delim(struct sed_cmd *sed_cmd, const char *str, int idx)
+static int index_of_next_unescaped_regexp_delim(const char delimiter,
+ const char *str)
{
int bracket = -1;
int escaped = 0;
+ int idx = 0;
+ char ch;
- for ( ; str[idx]; idx++) {
+ for (; (ch = str[idx]); idx++) {
if (bracket != -1) {
- if (str[idx] == ']' && !(bracket == idx - 1 ||
- (bracket == idx - 2 && str[idx-1] == '^')))
+ if (ch == ']' && !(bracket == idx - 1 || (bracket == idx - 2
+ && str[idx - 1] == '^')))
bracket = -1;
} else if (escaped)
escaped = 0;
- else if (str[idx] == '\\')
+ else if (ch == '\\')
escaped = 1;
- else if (str[idx] == '[')
+ else if (ch == '[')
bracket = idx;
- else if (str[idx] == sed_cmd->delimiter)
+ else if (ch == delimiter)
return idx;
}
}
/*
- * returns the index in the string just past where the address ends.
+ * Returns the index of the third delimiter
*/
-static int get_address(struct sed_cmd *sed_cmd, const char *str, int *linenum, regex_t **regex)
+static int parse_regex_delim(const char *cmdstr, char **match, char **replace)
{
- char *my_str = strdup(str);
+ const char *cmdstr_ptr = cmdstr;
+ char delimiter;
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;
- *linenum = atoi(my_str);
- }
- else if (my_str[idx] == '$') {
- *linenum = -1;
- idx++;
+
+ /* verify that the 's' or 'y' is followed by something. That something
+ * (typically a 'slash') is now our regexp delimiter... */
+ if (*cmdstr == '\0') bb_error_msg_and_die(bad_format_in_subst);
+ delimiter = *(cmdstr_ptr++);
+
+ /* save the match string */
+ idx = index_of_next_unescaped_regexp_delim(delimiter, cmdstr_ptr);
+ if (idx == -1) {
+ bb_error_msg_and_die(bad_format_in_subst);
}
- else if (my_str[idx] == '/') {
- idx = index_of_next_unescaped_regexp_delim(sed_cmd, my_str, ++idx);
- if (idx == -1)
- error_msg_and_die("unterminated match expression");
- my_str[idx] = '\0';
- *regex = (regex_t *)xmalloc(sizeof(regex_t));
- xregcomp(*regex, my_str+1, REG_NEWLINE);
- idx++; /* so it points to the next character after the last '/' */
+ *match = copy_parsing_slashn(cmdstr_ptr, idx);
+
+ /* save the replacement string */
+ cmdstr_ptr += idx + 1;
+ idx = index_of_next_unescaped_regexp_delim(delimiter, cmdstr_ptr);
+ if (idx == -1) {
+ bb_error_msg_and_die(bad_format_in_subst);
}
- else {
- error_msg("get_address: no address found in string\n"
- "\t(you probably didn't check the string you passed me)");
- idx = -1;
+ *replace = copy_parsing_slashn(cmdstr_ptr, idx);
+
+ return ((cmdstr_ptr - cmdstr) + idx);
+}
+
+/*
+ * returns the index in the string just past where the address ends.
+ */
+static int get_address(char *my_str, int *linenum, regex_t ** regex)
+{
+ char *pos = my_str;
+
+ if (isdigit(*my_str)) {
+ *linenum = strtol(my_str, &pos, 10);
+ /* endstr shouldnt ever equal NULL */
+ } else if (*my_str == '$') {
+ *linenum = -1;
+ pos++;
+ } else if (*my_str == '/' || *my_str == '\\') {
+ int next;
+ char delimiter;
+ char *temp;
+
+ if (*my_str == '\\') delimiter = *(++pos);
+ else delimiter = '/';
+ next = index_of_next_unescaped_regexp_delim(delimiter, ++pos);
+ if (next == -1)
+ bb_error_msg_and_die("unterminated match expression");
+
+ temp=copy_parsing_slashn(pos,next);
+ *regex = (regex_t *) xmalloc(sizeof(regex_t));
+ xregcomp(*regex, temp, REG_NEWLINE);
+ free(temp);
+ /* Move position to next character after last delimiter */
+ pos+=(next+1);
}
+ return pos - my_str;
+}
+
+/* Grab a filename. Whitespace at start is skipped, then goes to EOL. */
+static int parse_file_cmd(sed_cmd_t * sed_cmd, const char *filecmdstr, char **retval)
+{
+ int start = 0, idx, hack=0;
+
+ /* Skip whitespace, then grab filename to end of line */
+ while (isspace(filecmdstr[start])) start++;
+ idx=start;
+ while(filecmdstr[idx] && filecmdstr[idx]!='\n') idx++;
+ /* If lines glued together, put backslash back. */
+ if(filecmdstr[idx]=='\n') hack=1;
+ if(idx==start) bb_error_msg_and_die("Empty filename");
+ *retval = bb_xstrndup(filecmdstr+start, idx-start+hack+1);
+ if(hack) *(idx+*retval)='\\';
- free(my_str);
- sed_cmd->delimiter = olddelimiter;
return idx;
}
-static int parse_subst_cmd(struct sed_cmd *sed_cmd, const char *substr)
+static int parse_subst_cmd(sed_cmd_t * const sed_cmd, char *substr)
{
- int oldidx, cflags = REG_NEWLINE;
+ int cflags = 0;
char *match;
int idx = 0;
- int j;
/*
- * the string that gets passed to this function should look like this:
- * s/match/replace/gIp
+ * A substitution command should look something like this:
+ * s/match/replace/ #gIpw
* || | |||
* mandatory optional
- *
- * (all three of the '/' slashes are mandatory)
*/
-
- /* 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_regexp_delim(sed_cmd, substr, ++idx);
- if (idx == -1)
- error_msg_and_die("bad format in substitution expression");
- match = xstrndup(substr + oldidx, idx - oldidx);
+ idx = parse_regex_delim(substr, &match, &sed_cmd->string);
/* determine the number of back references in the match string */
/* Note: we compute this here rather than in the do_subst_command()
* function to save processor time, at the expense of a little more memory
* (4 bits) per sed_cmd */
-
- /* sed_cmd->num_backrefs = 0; */ /* XXX: not needed? --apparently not */
- for (j = 0; match[j]; j++) {
- /* GNU/POSIX sed does not save more than nine backrefs */
- if (match[j] == '\\' && match[j+1] == '(' && sed_cmd->num_backrefs <= 9)
- sed_cmd->num_backrefs++;
- }
-
- /* save the replacement string */
- oldidx = idx+1;
- idx = index_of_next_unescaped_regexp_delim(sed_cmd, substr, ++idx);
- if (idx == -1)
- error_msg_and_die("bad format in substitution expression");
- sed_cmd->replace = xstrndup(substr + oldidx, idx - oldidx);
/* process the flags */
+
+ sed_cmd->which_match=1;
while (substr[++idx]) {
+ /* Parse match number */
+ if(isdigit(substr[idx])) {
+ if(match[0]!='^') {
+ /* Match 0 treated as all, multiple matches we take the last one. */
+ char *pos=substr+idx;
+ sed_cmd->which_match=(unsigned short)strtol(substr+idx,&pos,10);
+ idx=pos-substr;
+ }
+ continue;
+ }
+ /* Skip spaces */
+ if(isspace(substr[idx])) continue;
+
switch (substr[idx]) {
+ /* Replace all occurrences */
case 'g':
- sed_cmd->sub_g = 1;
- break;
- case 'I':
- cflags |= REG_ICASE;
+ if (match[0] != '^') sed_cmd->which_match = 0;
break;
+ /* Print pattern space */
case 'p':
sed_cmd->sub_p = 1;
break;
+ case 'w':
+ {
+ char *temp;
+ idx+=parse_file_cmd(sed_cmd,substr+idx,&temp);
+
+ break;
+ }
+ /* Ignore case (gnu exension) */
+ case 'I':
+ cflags |= REG_ICASE;
+ break;
+ case ';':
+ case '}':
+ goto out;
default:
- /* any whitespace or semicolon trailing after a s/// is ok */
- if (strchr("; \t\v\n\r", substr[idx]))
- goto out;
- /* else */
- error_msg_and_die("bad option in substitution expression");
+ bb_error_msg_and_die("bad option in substitution expression");
}
}
-
-out:
+out:
/* compile the match string into a regex */
- sed_cmd->sub_match = (regex_t *)xmalloc(sizeof(regex_t));
- xregcomp(sed_cmd->sub_match, match, cflags);
+ if (*match != '\0') {
+ /* If match is empty, we use last regex used at runtime */
+ sed_cmd->sub_match = (regex_t *) xmalloc(sizeof(regex_t));
+ xregcomp(sed_cmd->sub_match, match, cflags);
+ }
free(match);
return idx;
}
-static int parse_edit_cmd(struct sed_cmd *sed_cmd, const char *editstr)
+/*
+ * Process the commands arguments
+ */
+static char *parse_cmd_args(sed_cmd_t *sed_cmd, char *cmdstr)
{
- int idx = 0;
- int slashes_eaten = 0;
- char *ptr; /* shorthand */
-
- /*
- * the string that gets passed to this function should look like this:
- *
- * need one of these
- * |
- * | this backslash (immediately following the edit command) is mandatory
- * | |
- * [aic]\
- * TEXT1\
- * TEXT2\
- * TEXTN
- *
- * as soon as we hit a TEXT line that has no trailing '\', we're done.
- * this means a command like:
- *
- * i\
- * INSERTME
- *
- * is a-ok.
- *
- */
-
- if (editstr[1] != '\\' && (editstr[2] != '\n' || editstr[2] != '\r'))
- 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
- * to the end */
- sed_cmd->editline = xmalloc(strlen(&editstr[3]) + 2);
- strcpy(sed_cmd->editline, &editstr[3]);
- ptr = sed_cmd->editline;
-
- /* 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')) {
- idx++;
- if (!ptr[idx]) {
- goto out;
- }
+ /* handle (s)ubstitution command */
+ if (sed_cmd->cmd == 's') cmdstr += parse_subst_cmd(sed_cmd, cmdstr);
+ /* 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')
+ bb_error_msg_and_die
+ ("only a beginning address can be specified for edit commands");
+ for(;;) {
+ if(*cmdstr=='\n' || *cmdstr=='\\') {
+ cmdstr++;
+ break;
+ } else if(isspace(*cmdstr)) cmdstr++;
+ else break;
+ }
+ sed_cmd->string = bb_xstrdup(cmdstr);
+ parse_escapes(sed_cmd->string,sed_cmd->string,strlen(cmdstr),0,0);
+ cmdstr += strlen(cmdstr);
+ /* handle file cmds: (r)ead */
+ } else if(strchr("rw", sed_cmd->cmd)) {
+ if (sed_cmd->end_line || sed_cmd->end_match)
+ bb_error_msg_and_die("Command only uses one address");
+ cmdstr += parse_file_cmd(sed_cmd, cmdstr, &sed_cmd->string);
+ if(sed_cmd->cmd=='w')
+ sed_cmd->file=bb_xfopen(sed_cmd->string,"w");
+ /* handle branch commands */
+ } else if (strchr(":bt", sed_cmd->cmd)) {
+ int length;
+
+ while(isspace(*cmdstr)) cmdstr++;
+ length = strcspn(cmdstr, semicolon_whitespace);
+ if (length) {
+ sed_cmd->string = strndup(cmdstr, length);
+ cmdstr += length;
}
- /* move the newline over the '\' before it (effectively eats the '\') */
- memmove(&ptr[idx], &ptr[idx+1], strlen(&ptr[idx+1]));
- ptr[strlen(ptr)-1] = 0;
- slashes_eaten++;
- /* substitue \r for \n if needed */
- if (ptr[idx] == '\r')
- ptr[idx] = '\n';
}
-
-out:
- /* this accounts for discrepancies between the modified string and the
- * original string passed in to this function */
- idx += slashes_eaten;
-
- /* figure out if we need to add a newline */
- if (ptr[idx-1] != '\n') {
- ptr[idx] = '\n';
- idx++;
+ /* translation command */
+ else if (sed_cmd->cmd == 'y') {
+ char *match, *replace;
+ int i=cmdstr[0];
+
+ cmdstr+=parse_regex_delim(cmdstr, &match, &replace)+1;
+ /* \n already parsed, but \delimiter needs unescaping. */
+ parse_escapes(match,match,strlen(match),i,i);
+ parse_escapes(replace,replace,strlen(replace),i,i);
+
+ sed_cmd->string = xcalloc(1, (strlen(match) + 1) * 2);
+ for (i = 0; match[i] && replace[i]; i++) {
+ sed_cmd->string[i * 2] = match[i];
+ sed_cmd->string[(i * 2) + 1] = replace[i];
+ }
+ free(match);
+ free(replace);
+ }
+ /* if it wasnt a single-letter command that takes no arguments
+ * then it must be an invalid command.
+ */
+ else if (strchr("dDgGhHlnNpPqx={}", sed_cmd->cmd) == 0) {
+ bb_error_msg_and_die("Unsupported command %c", sed_cmd->cmd);
}
- /* terminate string */
- ptr[idx]= 0;
- /* adjust for opening 2 chars [aic]\ */
- idx += 2;
-
- return idx;
+ /* give back whatever's left over */
+ return (cmdstr);
}
-static int parse_file_cmd(struct sed_cmd *sed_cmd, const char *filecmdstr)
+/* Parse address+command sets, skipping comment lines. */
+
+void add_cmd(char *cmdstr)
{
- int idx = 0;
- int filenamelen = 0;
+ static char *add_cmd_line=NULL;
+ sed_cmd_t *sed_cmd;
+ int temp;
+
+ /* Append this line to any unfinished line from last time. */
+ if(add_cmd_line) {
+ int lastlen=strlen(add_cmd_line);
+ char *tmp=xmalloc(lastlen+strlen(cmdstr)+2);
+
+ memcpy(tmp,add_cmd_line,lastlen);
+ tmp[lastlen]='\n';
+ strcpy(tmp+lastlen+1,cmdstr);
+ free(add_cmd_line);
+ cmdstr=add_cmd_line=tmp;
+ } else add_cmd_line=NULL;
+
+ /* If this line ends with backslash, request next line. */
+ temp=strlen(cmdstr);
+ if(temp && cmdstr[temp-1]=='\\') {
+ if(!add_cmd_line) add_cmd_line=strdup(cmdstr);
+ add_cmd_line[temp-1]=0;
+ return;
+ }
- /*
- * 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'
- */
+ /* Loop parsing all commands in this line. */
+ while(*cmdstr) {
+ /* Skip leading whitespace and semicolons */
+ cmdstr += strspn(cmdstr, semicolon_whitespace);
- /* 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;
-}
+ /* If no more commands, exit. */
+ if(!*cmdstr) break;
+ /* if this is a comment, jump past it and keep going */
+ if (*cmdstr == '#') {
+ /* "#n" is the same as using -n on the command line */
+ if (cmdstr[1] == 'n') be_quiet++;
+ if(!(cmdstr=strpbrk(cmdstr, "\n\r"))) break;
+ continue;
+ }
-static char *parse_cmd_str(struct sed_cmd *sed_cmd, const char *cmdstr)
-{
- int idx = 0;
+ /* parse the command
+ * format is: [addr][,addr][!]cmd
+ * |----||-----||-|
+ * part1 part2 part3
+ */
- /* parse the command
- * format is: [addr][,addr]cmd
- * |----||-----||-|
- * part1 part2 part3
- */
+ sed_cmd = xcalloc(1, sizeof(sed_cmd_t));
- /* first part (if present) is an address: either a number or a /regex/ */
- if (isdigit(cmdstr[idx]) || cmdstr[idx] == '/')
- idx = get_address(sed_cmd, cmdstr, &sed_cmd->beg_line, &sed_cmd->beg_match);
+ /* first part (if present) is an address: either a '$', a number or a /regex/ */
+ cmdstr += get_address(cmdstr, &sed_cmd->beg_line, &sed_cmd->beg_match);
- /* second part (if present) will begin with a comma */
- if (cmdstr[idx] == ',')
- idx += get_address(sed_cmd, &cmdstr[++idx], &sed_cmd->end_line, &sed_cmd->end_match);
+ /* second part (if present) will begin with a comma */
+ if (*cmdstr == ',') {
+ int idx;
- /* last part (mandatory) will be a command */
- if (cmdstr[idx] == '\0')
- error_msg_and_die("missing command");
- sed_cmd->cmd = cmdstr[idx];
+ cmdstr++;
+ idx = get_address(cmdstr, &sed_cmd->end_line, &sed_cmd->end_match);
+ if (!idx) bb_error_msg_and_die("get_address: no address found in string\n");
+ cmdstr += idx;
+ }
- /* 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]);
- }
- /* 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]);
- }
- /* 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");
- }
+ /* skip whitespace before the command */
+ while (isspace(*cmdstr)) cmdstr++;
- /* give back whatever's left over */
- return (char *)&cmdstr[idx];
-}
+ /* Check for inversion flag */
+ if (*cmdstr == '!') {
+ sed_cmd->invert = 1;
+ cmdstr++;
-static void add_cmd_str(const char *cmdstr)
-{
- char *mystr = (char *)cmdstr;
+ /* skip whitespace before the command */
+ while (isspace(*cmdstr)) cmdstr++;
+ }
- do {
+ /* last part (mandatory) will be a command */
+ if (!*cmdstr) bb_error_msg_and_die("missing command");
+ sed_cmd->cmd = *(cmdstr++);
+ cmdstr = parse_cmd_args(sed_cmd, cmdstr);
- /* trim leading whitespace and semicolons */
- memmove(mystr, &mystr[strspn(mystr, "; \n\r\t\v")], strlen(mystr));
- /* if we ate the whole thing, that means there was just trailing
- * whitespace or a final / no-op semicolon. either way, get out */
- if (strlen(mystr) == 0)
- return;
- /* if this is a comment, jump past it and keep going */
- if (mystr[0] == '#') {
- mystr = strpbrk(mystr, ";\n\r");
- continue;
- }
- /* grow the array */
- sed_cmds = xrealloc(sed_cmds, sizeof(struct sed_cmd) * (++ncmds));
- /* zero new element */
- memset(&sed_cmds[ncmds-1], 0, sizeof(struct sed_cmd));
- /* load command string into new array element, get remainder */
- mystr = parse_cmd_str(&sed_cmds[ncmds-1], mystr);
-
- } while (mystr && strlen(mystr));
+ /* Add the command to the command array */
+ sed_cmd_tail->next = sed_cmd;
+ sed_cmd_tail = sed_cmd_tail->next;
+ }
+
+ /* If we glued multiple lines together, free the memory. */
+ if(add_cmd_line) {
+ free(add_cmd_line);
+ add_cmd_line=NULL;
+ }
}
+struct pipeline {
+ char *buf; /* Space to hold string */
+ int idx; /* Space used */
+ int len; /* Space allocated */
+} pipeline;
+
+#define PIPE_GROW 64
-static void load_cmd_file(char *filename)
+void pipe_putc(char c)
{
- FILE *cmdfile;
- char *line;
- char *nextline;
-
- cmdfile = xfopen(filename, "r");
-
- while ((line = get_line_from_file(cmdfile)) != NULL) {
- /* if a line ends with '\' it needs the next line appended to it */
- while (line[strlen(line)-2] == '\\' &&
- (nextline = get_line_from_file(cmdfile)) != NULL) {
- line = xrealloc(line, strlen(line) + strlen(nextline) + 1);
- strcat(line, nextline);
- free(nextline);
- }
- /* eat trailing newline (if any) --if I don't do this, edit commands
- * (aic) will print an extra newline */
- chomp(line);
- add_cmd_str(line);
- free(line);
+ if(pipeline.idx==pipeline.len) {
+ pipeline.buf = xrealloc(pipeline.buf, pipeline.len + PIPE_GROW);
+ pipeline.len+=PIPE_GROW;
}
+ pipeline.buf[pipeline.idx++] = (c);
}
-#define PIPE_MAGIC 0x7f
-#define PIPE_GROW 64
-#define pipeputc(c) \
-{ if (pipeline[pipeline_idx] == PIPE_MAGIC) { \
- pipeline = xrealloc(pipeline, pipeline_len+PIPE_GROW); \
- memset(pipeline+pipeline_len, 0, PIPE_GROW); \
- pipeline_len += PIPE_GROW; \
- pipeline[pipeline_len-1] = PIPE_MAGIC; } \
- pipeline[pipeline_idx++] = (c); }
-
-static void print_subst_w_backrefs(const char *line, const char *replace,
- regmatch_t *regmatch, char **pipeline_p, int *pipeline_idx_p,
- int *pipeline_len_p, int matches)
+static void do_subst_w_backrefs(const char *line, const char *replace)
{
- char *pipeline = *pipeline_p;
- int pipeline_idx = *pipeline_idx_p;
- int pipeline_len = *pipeline_len_p;
- int i;
+ int i,j;
/* go through the replacement string */
for (i = 0; replace[i]; i++) {
/* if we find a backreference (\1, \2, etc.) print the backref'ed * text */
- if (replace[i] == '\\' && isdigit(replace[i+1])) {
- int j;
- char tmpstr[2];
- int backref;
- ++i; /* i now indexes the backref number, instead of the leading slash */
- tmpstr[0] = replace[i];
- tmpstr[1] = 0;
- backref = atoi(tmpstr);
+ if (replace[i] == '\\' && replace[i+1]>'0' && replace[i+1]<='9') {
+ int backref=replace[++i]-'0';
+
/* print out the text held in regmatch[backref] */
- if (backref <= matches && regmatch[backref].rm_so != -1)
+ if(regmatch[backref].rm_so != -1)
for (j = regmatch[backref].rm_so; j < regmatch[backref].rm_eo; j++)
- pipeputc(line[j]);
+ pipe_putc(line[j]);
}
/* if we find a backslash escaped character, print the character */
- else if (replace[i] == '\\') {
- ++i;
- pipeputc(replace[i]);
- }
+ else if (replace[i] == '\\') pipe_putc(replace[++i]);
- /* if we find an unescaped '&' print out the whole matched text.
- * fortunately, regmatch[0] contains the indicies to the whole matched
- * expression (kinda seems like it was designed for just such a
- * purpose...) */
- else if (replace[i] == '&' && replace[i-1] != '\\') {
- int j;
+ /* if we find an unescaped '&' print out the whole matched text. */
+ else if (replace[i] == '&')
for (j = regmatch[0].rm_so; j < regmatch[0].rm_eo; j++)
- pipeputc(line[j]);
- }
- /* nothing special, just print this char of the replacement string to stdout */
- else
- pipeputc(replace[i]);
+ pipe_putc(line[j]);
+ /* Otherwise just output the character. */
+ else pipe_putc(replace[i]);
}
- *pipeline_p = pipeline;
- *pipeline_idx_p = pipeline_idx;
- *pipeline_len_p = pipeline_len;
}
-static int do_subst_command(const struct sed_cmd *sed_cmd, char **line)
+static int do_subst_command(sed_cmd_t * sed_cmd, char **line)
{
- char *hackline = *line;
- char *pipeline = 0;
- int pipeline_idx = 0;
- int pipeline_len = 0;
+ char *oldline = *line;
int altered = 0;
- regmatch_t *regmatch = NULL;
-
- /* we only proceed if the substitution 'search' expression matches */
- if (regexec(sed_cmd->sub_match, hackline, 0, NULL, 0) == REG_NOMATCH)
+ int match_count=0;
+ regex_t *current_regex;
+
+ /* Handle empty regex. */
+ if (sed_cmd->sub_match == NULL) {
+ current_regex = previous_regex_ptr;
+ if(!current_regex)
+ bb_error_msg_and_die("No previous regexp.");
+ } else previous_regex_ptr = current_regex = sed_cmd->sub_match;
+
+ /* Find the first match */
+ if(REG_NOMATCH==regexec(current_regex, oldline, 10, regmatch, 0))
return 0;
- /* whaddaya know, it matched. get the number of back references */
- regmatch = xmalloc(sizeof(regmatch_t) * (sed_cmd->num_backrefs+1));
-
- /* allocate more PIPE_GROW bytes
- if replaced string is larger than original */
- pipeline_len = strlen(hackline)+PIPE_GROW;
- pipeline = xmalloc(pipeline_len);
- memset(pipeline, 0, pipeline_len);
- /* buffer magic */
- pipeline[pipeline_len-1] = PIPE_MAGIC;
-
- /* and now, as long as we've got a line to try matching and if we can match
- * the search string, we make substitutions */
- while (*hackline && (regexec(sed_cmd->sub_match, hackline,
- sed_cmd->num_backrefs+1, regmatch, 0) == 0) ) {
+ /* Initialize temporary output buffer. */
+ pipeline.buf=xmalloc(PIPE_GROW);
+ pipeline.len=PIPE_GROW;
+ pipeline.idx=0;
+
+ /* Now loop through, substituting for matches */
+ do {
int i;
+ /* Work around bug in glibc regexec, demonstrated by:
+ echo " a.b" | busybox sed 's [^ .]* x g'
+ The match_count check is so not to break
+ echo "hi" | busybox sed 's/^/!/g' */
+ if(!regmatch[0].rm_so && !regmatch[0].rm_eo && match_count) {
+ pipe_putc(*(oldline++));
+ continue;
+ }
+
+ match_count++;
+
+ /* If we aren't interested in this match, output old line to
+ end of match and continue */
+ if(sed_cmd->which_match && sed_cmd->which_match!=match_count) {
+ for(i=0;i<regmatch[0].rm_eo;i++)
+ pipe_putc(oldline[i]);
+ continue;
+ }
+
/* print everything before the match */
- for (i = 0; i < regmatch[0].rm_so; i++)
- pipeputc(hackline[i]);
+ for (i = 0; i < regmatch[0].rm_so; i++) pipe_putc(oldline[i]);
/* then print the substitution string */
- print_subst_w_backrefs(hackline, sed_cmd->replace, regmatch,
- &pipeline, &pipeline_idx, &pipeline_len,
- sed_cmd->num_backrefs);
+ do_subst_w_backrefs(oldline, sed_cmd->string);
/* advance past the match */
- hackline += regmatch[0].rm_eo;
+ oldline += regmatch[0].rm_eo;
/* flag that something has changed */
altered++;
/* if we're not doing this globally, get out now */
- if (!sed_cmd->sub_g)
- break;
- }
+ if (sed_cmd->which_match) break;
+ } while (*oldline && (regexec(current_regex, oldline, 10, regmatch, 0) != REG_NOMATCH));
- for (; *hackline; hackline++) pipeputc(*hackline);
- if (pipeline[pipeline_idx] == PIPE_MAGIC) pipeline[pipeline_idx] = 0;
+ /* Copy rest of string into output pipeline */
- /* cleanup */
- free(regmatch);
+ while(*oldline) pipe_putc(*(oldline++));
+ pipe_putc(0);
free(*line);
- *line = pipeline;
+ *line = pipeline.buf;
return altered;
}
+/* Set command pointer to point to this label. (Does not handle null label.) */
+static sed_cmd_t *branch_to(const char *label)
+{
+ sed_cmd_t *sed_cmd;
+
+ for (sed_cmd = sed_cmd_head.next; sed_cmd; sed_cmd = sed_cmd->next) {
+ if ((sed_cmd->cmd == ':') && (sed_cmd->string) && (strcmp(sed_cmd->string, label) == 0)) {
+ return (sed_cmd);
+ }
+ }
+ bb_error_msg_and_die("Can't find label for jump to `%s'", label);
+}
+
+/* Append copy of string to append buffer */
+static void append(char *s)
+{
+ struct append_list *temp=calloc(1,sizeof(struct append_list));
+
+ if(append_head)
+ append_tail=(append_tail->next=temp);
+ else append_head=append_tail=temp;
+ temp->string=strdup(s);
+}
+
+static void flush_append(void)
+{
+ /* Output appended lines. */
+ while(append_head) {
+ fprintf(nonstdout,"%s\n",append_head->string);
+ append_tail=append_head->next;
+ free(append_head->string);
+ free(append_head);
+ append_head=append_tail;
+ }
+ append_head=append_tail=NULL;
+}
+
+/* Get next line of input, flushing append buffer and noting if we hit EOF
+ * without a newline on the last line.
+ */
+static char *get_next_line(FILE * file, int *no_newline)
+{
+ char *temp;
+ int len;
+
+ flush_append();
+ temp=bb_get_line_from_file(file);
+ if(temp) {
+ len=strlen(temp);
+ if(len && temp[len-1]=='\n') temp[len-1]=0;
+ else *no_newline=1;
+ }
+
+ return temp;
+}
+
+/* Output line of text. missing_newline means the last line output did not
+ end with a newline. no_newline means this line does not end with a
+ newline. */
+
+static int puts_maybe_newline(char *s, FILE *file, int missing_newline, int no_newline)
+{
+ if(missing_newline) fputc('\n',file);
+ fputs(s,file);
+ if(!no_newline) fputc('\n',file);
+
+ if(ferror(file)) {
+ fprintf(stderr,"Write failed.\n");
+ exit(4); /* It's what gnu sed exits with... */
+ }
+
+ return no_newline;
+}
+
+#define sed_puts(s,n) missing_newline=puts_maybe_newline(s,nonstdout,missing_newline,n)
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 altered;
- int i;
+ char *pattern_space, *next_line, *hold_space=NULL;
+ static int linenum = 0, missing_newline=0;
+ int no_newline,next_no_newline=0;
+
+ next_line = get_next_line(file,&next_no_newline);
/* go through every line in the file */
- while ((line = get_line_from_file(file)) != NULL) {
+ for(;;) {
+ sed_cmd_t *sed_cmd;
+ int substituted=0;
- chomp(line);
- linenum++;
- altered = 0;
+ /* Advance to next line. Stop if out of lines. */
+ if(!(pattern_space=next_line)) break;
+ no_newline=next_no_newline;
+ /* Read one line in advance so we can act on the last line, the '$' address */
+ next_line = get_next_line(file,&next_no_newline);
+ linenum++;
+restart:
/* for every line, go through all the commands */
- for (i = 0; i < ncmds; i++) {
-
-
- /*
- * 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) {
+ for (sed_cmd = sed_cmd_head.next; sed_cmd; sed_cmd = sed_cmd->next) {
+ int old_matched, matched;
- case 'p':
- puts(line);
- break;
+ old_matched = sed_cmd->in_match;
- case 'd':
- altered++;
+ /* Determine if this command matches this line: */
+
+ /* Are we continuing a previous multi-line match? */
+
+ sed_cmd->in_match = sed_cmd->in_match
+
+ /* Or is no range necessary? */
+ || (!sed_cmd->beg_line && !sed_cmd->end_line
+ && !sed_cmd->beg_match && !sed_cmd->end_match)
+
+ /* Or did we match the start of a numerical range? */
+ || (sed_cmd->beg_line > 0 && (sed_cmd->beg_line == linenum))
+
+ /* Or does this line match our begin address regex? */
+ || (sed_cmd->beg_match &&
+ !regexec(sed_cmd->beg_match, pattern_space, 0, NULL, 0))
+
+ /* Or did we match last line of input? */
+ || (sed_cmd->beg_line == -1 && next_line == NULL);
+
+ /* Snapshot the value */
+
+ matched = sed_cmd->in_match;
+
+ /* Is this line the end of the current match? */
+
+ if(matched) {
+ sed_cmd->in_match = !(
+ /* has the ending line come, or is this a single address command? */
+ (sed_cmd->end_line ?
+ sed_cmd->end_line==-1 ?
+ !next_line
+ : sed_cmd->end_line<=linenum
+ : !sed_cmd->end_match)
+ /* or does this line matches our last address regex */
+ || (sed_cmd->end_match && old_matched && (regexec(sed_cmd->end_match, pattern_space, 0, NULL, 0) == 0))
+ );
+ }
+
+ /* Skip blocks of commands we didn't match. */
+ if (sed_cmd->cmd == '{') {
+ if(sed_cmd->invert ? matched : !matched)
+ while(sed_cmd && sed_cmd->cmd!='}') sed_cmd=sed_cmd->next;
+ if(!sed_cmd) bb_error_msg_and_die("Unterminated {");
+ continue;
+ }
+
+ /* Okay, so did this line match? */
+ if (sed_cmd->invert ? !matched : matched) {
+ /* Update last used regex in case a blank substitute BRE is found */
+ if (sed_cmd->beg_match) {
+ previous_regex_ptr = sed_cmd->beg_match;
+ }
+
+ /* actual sedding */
+ switch (sed_cmd->cmd) {
+
+ /* Print line number */
+ case '=':
+ fprintf(nonstdout,"%d\n", linenum);
break;
- case 's':
+ /* Write the current pattern space up to the first newline */
+ case 'P':
+ {
+ char *tmp = strchr(pattern_space, '\n');
- /*
- * 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)
+ if (tmp) {
+ *tmp = '\0';
+ sed_puts(pattern_space,1);
+ *tmp = '\n';
break;
+ }
+ /* Fall Through */
+ }
- /* we print the line once, unless we were told to be quiet */
- if (!be_quiet)
- altered |= do_subst_command(&sed_cmds[i], &line);
+ /* Write the current pattern space to output */
+ case 'p':
+ sed_puts(pattern_space,no_newline);
+ break;
+ /* Delete up through first newline */
+ case 'D':
+ {
+ char *tmp = strchr(pattern_space,'\n');
+
+ if(tmp) {
+ tmp=bb_xstrdup(tmp+1);
+ free(pattern_space);
+ pattern_space=tmp;
+ goto restart;
+ }
+ }
+ /* discard this line. */
+ case 'd':
+ goto discard_line;
+
+ /* Substitute with regex */
+ case 's':
+ if(do_subst_command(sed_cmd, &pattern_space)) {
+ substituted|=1;
- /* 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);
- if (altered && (i+1 >= ncmds || sed_cmds[i+1].cmd != 's'))
- puts(line);
+ /* handle p option */
+ if(sed_cmd->sub_p)
+ sed_puts(pattern_space,no_newline);
+ /* handle w option */
+ if(sed_cmd->file)
+ sed_cmd->no_newline=puts_maybe_newline(pattern_space, sed_cmd->file, sed_cmd->no_newline, no_newline);
+ }
break;
+ /* Append line to linked list to be printed later */
case 'a':
- puts(line);
- fputs(sed_cmds[i].editline, stdout);
- altered++;
+ {
+ append(sed_cmd->string);
break;
+ }
+ /* Insert text before this line */
case 'i':
- fputs(sed_cmds[i].editline, stdout);
+ sed_puts(sed_cmd->string,1);
break;
+ /* Cut and paste text (replace) */
case 'c':
- /* single-address case */
- if (sed_cmds[i].end_match == NULL && sed_cmds[i].end_line == 0) {
- fputs(sed_cmds[i].editline, stdout);
+ /* Only triggers on last line of a matching range. */
+ if (!sed_cmd->in_match) sed_puts(sed_cmd->string,1);
+ goto discard_line;
+
+ /* Read file, append contents to output */
+ case 'r':
+ {
+ FILE *outfile;
+
+ outfile = fopen(sed_cmd->string, "r");
+ if (outfile) {
+ char *line;
+
+ while ((line = bb_get_chomped_line_from_file(outfile))
+ != NULL)
+ append(line);
+ bb_xprint_and_close_file(outfile);
+ }
+
+ break;
+ }
+
+ /* Write pattern space to file. */
+ case 'w':
+ sed_cmd->no_newline=puts_maybe_newline(pattern_space,sed_cmd->file, sed_cmd->no_newline,no_newline);
+ break;
+
+ /* Read next line from input */
+ case 'n':
+ if (!be_quiet)
+ sed_puts(pattern_space,no_newline);
+ if (next_line) {
+ free(pattern_space);
+ pattern_space = next_line;
+ no_newline=next_no_newline;
+ next_line = get_next_line(file,&next_no_newline);
+ linenum++;
+ break;
+ }
+ /* fall through */
+
+ /* Quit. End of script, end of input. */
+ case 'q':
+ /* Exit the outer while loop */
+ free(next_line);
+ next_line = NULL;
+ goto discard_commands;
+
+ /* Append the next line to the current line */
+ case 'N':
+ {
+ /* If no next line, jump to end of script and exit. */
+ if (next_line == NULL) {
+ /* Jump to end of script and exit */
+ free(next_line);
+ next_line = NULL;
+ goto discard_line;
+ /* append next_line, read new next_line. */
+ } else {
+ int len=strlen(pattern_space);
+
+ pattern_space = realloc(pattern_space, len + strlen(next_line) + 2);
+ pattern_space[len]='\n';
+ strcpy(pattern_space+len+1, next_line);
+ no_newline=next_no_newline;
+ next_line = get_next_line(file,&next_no_newline);
+ linenum++;
+ }
+ break;
+ }
+
+ /* Test if substition worked, branch if so. */
+ case 't':
+ if (!substituted) break;
+ substituted=0;
+ /* Fall through */
+ /* Branch to label */
+ case 'b':
+ if (!sed_cmd->string) goto discard_commands;
+ else sed_cmd = branch_to(sed_cmd->string);
+ break;
+ /* Transliterate characters */
+ case 'y':
+ {
+ int i;
+
+ for (i = 0; pattern_space[i]; i++) {
+ int j;
+
+ for (j = 0; sed_cmd->string[j]; j += 2) {
+ if (pattern_space[i] == sed_cmd->string[j]) {
+ pattern_space[i] = sed_cmd->string[j + 1];
+ }
+ }
}
- /* 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);
+
+ break;
+ }
+ case 'g': /* Replace pattern space with hold space */
+ free(pattern_space);
+ if (hold_space) {
+ pattern_space = strdup(hold_space);
+ no_newline=0;
}
- altered++;
+ break;
+ case 'G': /* Append newline and hold space to pattern space */
+ {
+ int pattern_space_size = 2;
+ int hold_space_size = 0;
+
+ if (pattern_space)
+ pattern_space_size += strlen(pattern_space);
+ if (hold_space) hold_space_size = strlen(hold_space);
+ pattern_space = xrealloc(pattern_space, pattern_space_size + hold_space_size);
+ if (pattern_space_size == 2) pattern_space[0]=0;
+ strcat(pattern_space, "\n");
+ if (hold_space) strcat(pattern_space, hold_space);
+ no_newline=0;
break;
+ }
+ case 'h': /* Replace hold space with pattern space */
+ free(hold_space);
+ hold_space = strdup(pattern_space);
+ break;
+ case 'H': /* Append newline and pattern space to hold space */
+ {
+ int hold_space_size = 2;
+ int pattern_space_size = 0;
- 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;
- }
+ if (hold_space) hold_space_size += strlen(hold_space);
+ if (pattern_space)
+ pattern_space_size = strlen(pattern_space);
+ hold_space = xrealloc(hold_space,
+ hold_space_size + pattern_space_size);
- /*
- * 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;
- }
+ if (hold_space_size == 2) hold_space[0]=0;
+ strcat(hold_space, "\n");
+ if (pattern_space) strcat(hold_space, pattern_space);
- /* didn't hit the exit? then we're still in the middle of an address range */
- else {
- still_in_range = 1;
+ break;
+ }
+ case 'x': /* Exchange hold and pattern space */
+ {
+ char *tmp = pattern_space;
+ pattern_space = hold_space;
+ no_newline=0;
+ hold_space = tmp;
+ break;
+ }
}
}
}
- /* 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 && !altered)
- puts(line);
+ /*
+ * exit point from sedding...
+ */
+discard_commands:
+ /* we will print the line unless we were told to be quiet ('-n')
+ or if the line was suppressed (ala 'd'elete) */
+ if (!be_quiet) sed_puts(pattern_space,no_newline);
+
+ /* Delete and such jump here. */
+discard_line:
+ flush_append();
+ free(pattern_space);
+ }
+}
+
+/* It is possible to have a command line argument with embedded
+ newlines. This counts as multiple command lines. */
- free(line);
+static void add_cmd_block(char *cmdstr)
+{
+ int go=1;
+ char *temp=bb_xstrdup(cmdstr),*temp2=temp;
+
+ while(go) {
+ int len=strcspn(temp2,"\n");
+ if(!temp2[len]) go=0;
+ else temp2[len]=0;
+ add_cmd(temp2);
+ temp2+=len+1;
}
+ free(temp);
}
extern int sed_main(int argc, char **argv)
{
+ int status = EXIT_SUCCESS;
int opt;
+ uint8_t getpat = 1;
-#ifdef BB_FEATURE_CLEAN_UP
+#ifdef CONFIG_FEATURE_CLEAN_UP
/* destroy command strings on exit */
- if (atexit(destroy_cmd_strs) == -1)
- perror_msg_and_die("atexit");
+ if (atexit(free_and_close_stuff) == -1)
+ bb_perror_msg_and_die("atexit");
+#endif
+
+#define LIE_TO_AUTOCONF
+#ifdef LIE_TO_AUTOCONF
+ if(argc==2 && !strcmp(argv[1],"--version")) {
+ printf("This is not GNU sed version 4.0\n");
+ exit(0);
+ }
#endif
/* do normal option parsing */
- while ((opt = getopt(argc, argv, "ne:f:")) > 0) {
+ while ((opt = getopt(argc, argv, "ine:f:")) > 0) {
switch (opt) {
- case 'n':
- be_quiet++;
- break;
- case 'e':
- add_cmd_str(optarg);
- break;
- case 'f':
- load_cmd_file(optarg);
- break;
- default:
- show_usage();
+ case 'i':
+ in_place++;
+ atexit(cleanup_outname);
+ break;
+ case 'n':
+ be_quiet++;
+ break;
+ case 'e':
+ add_cmd_block(optarg);
+ getpat=0;
+ break;
+ case 'f':
+ {
+ FILE *cmdfile;
+ char *line;
+
+ cmdfile = bb_xfopen(optarg, "r");
+
+ while ((line = bb_get_chomped_line_from_file(cmdfile))
+ != NULL) {
+ add_cmd(line);
+ getpat=0;
+ free(line);
+ }
+ bb_xprint_and_close_file(cmdfile);
+
+ break;
+ }
+ default:
+ bb_show_usage();
}
}
/* if we didn't get a pattern from a -e and no command file was specified,
* argv[optind] should be the pattern. no pattern, no worky */
- if (ncmds == 0) {
+ if(getpat) {
if (argv[optind] == NULL)
- show_usage();
- else {
- add_cmd_str(argv[optind]);
- optind++;
- }
+ bb_show_usage();
+ else
+ add_cmd_block(argv[optind++]);
}
+ /* Flush any unfinished commands. */
+ add_cmd("");
+ /* By default, we write to stdout */
+ nonstdout=stdout;
/* argv[(optind)..(argc-1)] should be names of file to process. If no
* files were specified or '-' was specified, take input from stdin.
* Otherwise, we process all the files specified. */
- if (argv[optind] == NULL || (strcmp(argv[optind], "-") == 0)) {
+ if (argv[optind] == NULL) {
+ if(in_place) {
+ fprintf(stderr,"sed: Filename required for -i\n");
+ exit(1);
+ }
process_file(stdin);
- }
- else {
+ } else {
int i;
FILE *file;
+
for (i = optind; i < argc; i++) {
- file = fopen(argv[i], "r");
- if (file == NULL) {
- perror_msg("%s", argv[i]);
+ if(!strcmp(argv[i], "-") && !in_place) {
+ process_file(stdin);
} else {
- process_file(file);
- fclose(file);
+ file = bb_wfopen(argv[i], "r");
+ if (file) {
+ if(in_place) {
+ struct stat statbuf;
+ outname=bb_xstrndup(argv[i],strlen(argv[i])+6);
+ strcat(outname,"XXXXXX");
+ /* Set permissions of output file */
+ fstat(fileno(file),&statbuf);
+ mkstemp(outname);
+ nonstdout=bb_wfopen(outname,"w");
+ /* Set permissions of output file */
+ fstat(fileno(file),&statbuf);
+ fchmod(fileno(nonstdout),statbuf.st_mode);
+ atexit(cleanup_outname);
+ }
+ process_file(file);
+ fclose(file);
+ if(in_place) {
+ fclose(nonstdout);
+ nonstdout=stdout;
+ unlink(argv[i]);
+ rename(outname,argv[i]);
+ free(outname);
+ outname=0;
+ }
+ } else {
+ status = EXIT_FAILURE;
+ }
}
}
}
-
- return 0;
+
+ return status;
}
projects / oweals / busybox.git / blobdiff