Removed trailing \n from error_msg{,_and_die} messages.
[oweals/busybox.git] / sed.c
diff --git a/sed.c b/sed.c
index a3e635d3a36990b1a53e2d5b691c5c5be4450341..e0351c3b80f6ae3029b68a4450467728a26dbc4d 100644 (file)
--- a/sed.c
+++ b/sed.c
@@ -1,9 +1,8 @@
 /*
- * Mini sed implementation for busybox
+ * sed.c - very minimalist version of sed
  *
- *
- * Copyright (C) 1999 by Lineo, inc.
- * Written by Erik Andersen <andersen@lineo.com>, <andersee@debian.org>
+ * 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
  *
  */
 
-#include "internal.h"
-#include "regexp.h"
+/*
+       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
+        - 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 <dirent.h>
+#include <unistd.h> /* for getopt() */
+#include <regex.h>
+#include <string.h> /* for strdup() */
 #include <errno.h>
-#include <fcntl.h>
-#include <signal.h>
-#include <time.h>
-#include <ctype.h>
-
-static const char sed_usage[] = 
-"sed [-n] [-e script] [file...]\n"
-"Allowed scripts come in the following form:\n\n"
-"'s/regexp/replacement/[gp]'\n"
-"\tattempt to match regexp against the pattern space\n"
-"\tand if successful replaces the matched portion with replacement.\n\n"
-"Options:\n"
-"-e\tadd the script to the commands to be executed\n"
-"-n\tsuppress automatic printing of pattern space\n\n"
-#if defined BB_REGEXP
-"This version of sed matches full regexps.\n";
-#else
-"This version of sed matches strings (not full regexps).\n";
+#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;
+};
+
+/* 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? */
+
+#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)
+{
+       for ( ; str[idx]; idx++) {
+               if (str[idx] == sed_cmd->delimiter && str[idx-1] != '\\')
+                       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 *line, regex_t **regex)
+{
+       char *my_str = strdup(str);
+       int idx = 0;
+
+       if (isdigit(my_str[idx])) {
+               do {
+                       idx++;
+               } while (isdigit(my_str[idx]));
+               my_str[idx] = 0;
+               *line = atoi(my_str);
+       }
+       else if (my_str[idx] == '$') {
+               *line = -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, 0);
+               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);
+       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;
+       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 = strdup_substr(substr, oldidx, idx);
+
+       /* 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 = strdup_substr(substr, oldidx, idx);
+
+       /* 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;
+}
+
+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:
+       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) */
+
+       idx += 3;
+
+       return idx;
+}
+
+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
+        */
+
+
+       /* 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 += 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");
+       if (!strchr("pdsaic", cmdstr[idx])) /* <-- XXX add new commands here */
+               error_msg_and_die("invalid command");
+       sed_cmd->cmd = cmdstr[idx];
 
-extern int sed_main (int argc, char **argv)
+       /* special-case handling for (s)ubstitution */
+       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)
+                       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++;
+
+       /* give back whatever's left over */
+       return (char *)&cmdstr[idx];
+}
+
+static void add_cmd_str(const char *cmdstr)
+{
+       char *mystr = (char *)cmdstr;
+
+       do {
+
+               /* 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));
+}
+
+
+static void load_cmd_file(char *filename)
 {
-    FILE *fp;
-    char *needle=NULL, *newNeedle=NULL;
-    char *name;
-    char *cp;
-    int ignoreCase=FALSE;
-    int foundOne=FALSE;
-    int noprintFlag=FALSE;
-    int stopNow;
-    char *haystack;
-
-    argc--;
-    argv++;
-    if (argc < 1) {
-       usage(sed_usage);
-    }
-
-    if (**argv == '-') {
-       argc--;
-       cp = *argv++;
-       stopNow=FALSE;
-
-       while (*++cp && stopNow==FALSE)
-           switch (*cp) {
-           case 'n':
-               noprintFlag = TRUE;
-               break;
-           case 'e':
-               if (*(cp+1)==0 && --argc < 0) {
-                   fprintf(stderr, "A\n");
-                   usage( sed_usage);
+       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 */
+               if (line[strlen(line)-1] == '\n')
+                       line[strlen(line)-1] = 0;
+               add_cmd_str(line);
+               free(line);
+       }
+}
+
+static void print_subst_w_backrefs(const char *line, const char *replace, regmatch_t *regmatch)
+{
+       int i;
+
+       /* 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);
+                       /* print out the text held in regmatch[backref] */
+                       for (j = regmatch[backref].rm_so; j < regmatch[backref].rm_eo; j++)
+                               fputc(line[j], stdout);
+               }
+
+               /* if we find a backslash escaped character, print the character */
+               else if (replace[i] == '\\') {
+                       ++i;
+                       fputc(replace[i], stdout);
+               }
+
+               /* 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;
+                       for (j = regmatch[0].rm_so; j < regmatch[0].rm_eo; j++)
+                               fputc(line[j], stdout);
                }
-               cp = *argv++;
-               while( *cp ) {
-                   if (*cp == 's' && strlen(cp) > 3 && *(cp+1) == '/') {
-                       char* pos=needle=cp+2;
-                       for(;;) {
-                           pos = strchr(pos, '/');
-                           if (pos==NULL) {
-                               fprintf(stderr, "B\n");
-                               usage( sed_usage);
-                           }
-                           if (*(pos-1) == '\\') {
-                               pos++;
-                               continue;
-                           }
-                           break;
+               /* nothing special, just print this char of the replacement string to stdout */
+               else
+                       fputc(replace[i], stdout);
+       }
+}
+
+static int do_subst_command(const struct sed_cmd *sed_cmd, const char *line)
+{
+       char *hackline = (char *)line;
+       int altered = 0;
+       regmatch_t *regmatch = NULL;
+
+       /* we only proceed if the substitution 'search' expression matches */
+       if (regexec(sed_cmd->sub_match, line, 0, NULL, 0) == REG_NOMATCH)
+               return 0;
+
+       /* whaddaya know, it matched. get the number of back references */
+       regmatch = xmalloc(sizeof(regmatch_t) * (sed_cmd->num_backrefs+1));
+
+       /* 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) ) {
+               int i;
+
+               /* print everything before the match */
+               for (i = 0; i < regmatch[0].rm_so; i++)
+                       fputc(hackline[i], stdout);
+
+               /* then print the substitution string */
+               print_subst_w_backrefs(hackline, sed_cmd->replace, regmatch);
+
+               /* advance past the match */
+               hackline += 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 there's anything left of the line, print it */
+       if (*hackline)
+               fputs(hackline, stdout);
+
+       /* 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.
+                        */
+
+                       /* 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)
+                               break;
+
+                       /* 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 i;
+
+       /* go through every line in the file */
+       while ((line = get_line_from_file(file)) != NULL) {
+
+               linenum++;
+               line_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;
+                               }
                        }
-                       *pos=0;
-                       newNeedle=++pos;
-                       for(;;) {
-                           pos = strchr(pos, '/');
-                           if (pos==NULL) {
-                               fprintf(stderr, "C\n");
-                               usage( sed_usage);
-                           }
-                           if (*(pos-1) == '\\') {
-                               pos++;
-                               continue;
-                           }
-                           break;
+
+                       /* 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);
                        }
-                       *pos=0;
-                   }
-                   cp++;
+
+                       /* 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);
+                       }
+
+                       /* 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);
+                       }
+
+                       /* not acting on matches or line numbers. act on every line */
+                       else 
+                               line_altered += do_sed_command(&sed_cmds[i], line);
+
                }
-               fprintf(stderr, "replace '%s' with '%s'\n", needle, newNeedle);
-               stopNow=TRUE;
-               break;
-
-           default:
-               fprintf(stderr, "D\n");
-               usage(sed_usage);
-           }
-    }
-
-    fprintf(stderr, "argc=%d\n", argc);
-    while (argc-- > 0) {
-       name = *argv++;
-
-       fp = fopen (name, "r");
-       if (fp == NULL) {
-           perror (name);
-           continue;
-       }
-       fprintf(stderr, "filename is '%s'\n", name);
-
-       haystack = (char*)malloc( 80);
-       while (fgets (haystack, sizeof (haystack), fp)) {
-
-           foundOne = replace_match(haystack, needle, newNeedle, ignoreCase);
-           if (noprintFlag==TRUE && foundOne==TRUE)
-               fputs (haystack, stdout);
-           else
-               fputs (haystack, stdout);
-           /* Avoid any mem leaks */
-           free(haystack);
-           haystack = (char*)malloc( BUF_SIZE);
-       }
-
-       if (ferror (fp))
-           perror (name);
-
-       fclose (fp);
-    }
-    exit( TRUE);
+
+               /* 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);
+
+               free(line);
+       }
 }
 
+extern int sed_main(int argc, char **argv)
+{
+       int opt;
 
-/* END CODE */
+#ifdef BB_FEATURE_CLEAN_UP
+       /* destroy command strings on exit */
+       if (atexit(destroy_cmd_strs) == -1)
+               perror_msg_and_die("atexit");
+#endif
+
+       /* do normal option parsing */
+       while ((opt = getopt(argc, argv, "hne:f:")) > 0) {
+               switch (opt) {
+                       case 'h':
+                               usage(sed_usage);
+                               break;
+                       case 'n':
+                               be_quiet++;
+                               break;
+                       case 'e':
+                               add_cmd_str(optarg);
+                               break;
+                       case 'f': 
+                               load_cmd_file(optarg);
+                               break;
+               }
+       }
+
+       /* 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 (argv[optind] == NULL)
+                       usage(sed_usage);
+               else {
+                       add_cmd_str(argv[optind]);
+                       optind++;
+               }
+       }
 
 
+       /* 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)) {
+               process_file(stdin);
+       }
+       else {
+               int i;
+               FILE *file;
+               for (i = optind; i < argc; i++) {
+                       file = fopen(argv[i], "r");
+                       if (file == NULL) {
+                               perror_msg("%s", argv[i]);
+                       } else {
+                               process_file(file);
+                               fclose(file);
+                       }
+               }
+       }
+       
+       return 0;
+}