sed.c

手机嵌入式Linux下可用的busybox源码

/*
 * sed.c - very minimalist version of sed
 *
 * Copyright (C) 1999,2000 by Lineo, inc. and Mark Whitley
 * Copyright (C) 1999,2000,2001 by Mark Whitley <[e-mail]> 
 *
 * 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
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
 *
 */

/*
	Supported features and commands in this version of sed:

	 - comments ('#')
	 - 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
	 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.
*/

#include <stdio.h>
#include <unistd.h> /* for getopt() */
#include <regex.h>
#include <string.h> /* for strdup() */
#include <errno.h>
#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 */

/* options */
static int be_quiet = 0;


struct sed_cmd {


	/* 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' */

	/* the command */
	char cmd; /* p,d,s (add more at your leisure :-) */


	/* SUBSTITUTION COMMAND SPECIFIC FIELDS */

	/* 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) */


	/* EDIT COMMAND (a,i,c) SPEICIFIC FIELDS */

	char *editline;


	/* FILE COMMAND (r) SPEICIFIC FIELDS */

	char *filename;
};

/* globals */
static struct sed_cmd *sed_cmds = NULL; /* growable arrary holding a sequence of sed cmds */
static int ncmds = 0; /* number of sed commands */

/*static char *cur_file = NULL;*/ /* file currently being processed XXX: do I need this? */

const char * const semicolon_whitespace = "; \n\r\t\v\0";

#ifdef BB_FEATURE_CLEAN_UP
static void destroy_cmd_strs()
{
	if (sed_cmds == NULL)
		return;

	/* destroy all the elements in the array */
	while (--ncmds >= 0) {

		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);
		}
		if (sed_cmds[ncmds].replace)
			free(sed_cmds[ncmds].replace);
	}

	/* destroy the array */
	free(sed_cmds);
	sed_cmds = NULL;
}
#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 
 * a backslash ('\').
 */
static int index_of_next_unescaped_regexp_delim(struct sed_cmd *sed_cmd, const char *str, int idx)
{
	int bracket = -1;
	int escaped = 0;

	for ( ; str[idx]; idx++) {
		if (bracket != -1) {
			if (str[idx] == ']' && !(bracket == idx - 1 ||
									 (bracket == idx - 2 && str[idx-1] == '^')))
				bracket = -1;
		} else if (escaped)
			escaped = 0;
		else if (str[idx] == '\\')
			escaped = 1;
		else if (str[idx] == '[')
			bracket = idx;
		else if (str[idx] == sed_cmd->delimiter)
			return idx;
	}

	/* if we make it to here, we've hit the end of the string */
	return -1;
}

/*
 * returns the index in the string just past where the address ends.
 */
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;
		*linenum = atoi(my_str);
	}
	else if (my_str[idx] == '$') {
		*linenum = -1;
		idx++;
	}
	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 '/' */
	}
	else {
		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 int parse_subst_cmd(struct sed_cmd * const sed_cmd, const char *substr)
{
	int oldidx, cflags = REG_NEWLINE;
	char *match;
	int idx = 0;
	int j;

	/*
	 * the string that gets passed to this function should look like this:
	 *    s/match/replace/gIp
	 *    ||     |        |||
	 *    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);

	/* 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 */
	while (substr[++idx]) {
		switch (substr[idx]) {
			case 'g':
				sed_cmd->sub_g = 1;
				break;
			case 'I':
				cflags |= REG_ICASE;
				break;
			case 'p':
				sed_cmd->sub_p = 1;
				break;
			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");
		}
	}

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);
	free(match);

	return idx;
}

#if 0
static void move_back(char *str, int offset)
{
    memmove(str, str + offset, strlen(str + offset) + 1);
}
#endif

static int parse_edit_cmd(struct sed_cmd *sed_cmd, const char *editstr)
{
	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;
			}
		}
		/* 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++;
	}

	/* 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 * const sed_cmd, const char *const cmdstr)
{
	int idx = 0;

	/* parse the command
	 * format is: [addr][,addr]cmd
	 *            |----||-----||-|
	 *            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(sed_cmd, cmdstr, &sed_cmd->beg_line, &sed_cmd->beg_match);

	/* second part (if present) will begin with a comma */
	if (cmdstr[idx] == ',') {
		idx++;
		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')
		error_msg_and_die("missing command");
	sed_cmd->cmd = cmdstr[idx];