[821] | 1 | /* vi: set sw=4 ts=4: */
|
---|
| 2 | /*
|
---|
| 3 | * sed.c - very minimalist version of sed
|
---|
| 4 | *
|
---|
| 5 | * Copyright (C) 1999,2000,2001 by Lineo, inc. and Mark Whitley
|
---|
| 6 | * Copyright (C) 1999,2000,2001 by Mark Whitley <markw@codepoet.org>
|
---|
| 7 | * Copyright (C) 2002 Matt Kraai
|
---|
[2725] | 8 | * Copyright (C) 2003 by Glenn McGrath
|
---|
[821] | 9 | * Copyright (C) 2003,2004 by Rob Landley <rob@landley.net>
|
---|
| 10 | *
|
---|
| 11 | * MAINTAINER: Rob Landley <rob@landley.net>
|
---|
| 12 | *
|
---|
[2725] | 13 | * Licensed under GPLv2, see file LICENSE in this source tree.
|
---|
[821] | 14 | */
|
---|
| 15 |
|
---|
| 16 | /* Code overview.
|
---|
[3232] | 17 | *
|
---|
| 18 | * Files are laid out to avoid unnecessary function declarations. So for
|
---|
| 19 | * example, every function add_cmd calls occurs before add_cmd in this file.
|
---|
| 20 | *
|
---|
| 21 | * add_cmd() is called on each line of sed command text (from a file or from
|
---|
| 22 | * the command line). It calls get_address() and parse_cmd_args(). The
|
---|
| 23 | * resulting sed_cmd_t structures are appended to a linked list
|
---|
| 24 | * (G.sed_cmd_head/G.sed_cmd_tail).
|
---|
| 25 | *
|
---|
| 26 | * add_input_file() adds a FILE* to the list of input files. We need to
|
---|
| 27 | * know all input sources ahead of time to find the last line for the $ match.
|
---|
| 28 | *
|
---|
| 29 | * process_files() does actual sedding, reading data lines from each input FILE*
|
---|
| 30 | * (which could be stdin) and applying the sed command list (sed_cmd_head) to
|
---|
| 31 | * each of the resulting lines.
|
---|
| 32 | *
|
---|
| 33 | * sed_main() is where external code calls into this, with a command line.
|
---|
| 34 | */
|
---|
[821] | 35 |
|
---|
[3232] | 36 | /* Supported features and commands in this version of sed:
|
---|
| 37 | *
|
---|
| 38 | * - comments ('#')
|
---|
| 39 | * - address matching: num|/matchstr/[,num|/matchstr/|$]command
|
---|
| 40 | * - commands: (p)rint, (d)elete, (s)ubstitue (with g & I flags)
|
---|
| 41 | * - edit commands: (a)ppend, (i)nsert, (c)hange
|
---|
| 42 | * - file commands: (r)ead
|
---|
| 43 | * - backreferences in substitution expressions (\0, \1, \2...\9)
|
---|
| 44 | * - grouped commands: {cmd1;cmd2}
|
---|
| 45 | * - transliteration (y/source-chars/dest-chars/)
|
---|
| 46 | * - pattern space hold space storing / swapping (g, h, x)
|
---|
| 47 | * - labels / branching (: label, b, t, T)
|
---|
| 48 | *
|
---|
| 49 | * (Note: Specifying an address (range) to match is *optional*; commands
|
---|
| 50 | * default to the whole pattern space if no specific address match was
|
---|
| 51 | * requested.)
|
---|
| 52 | *
|
---|
| 53 | * Todo:
|
---|
| 54 | * - Create a wrapper around regex to make libc's regex conform with sed
|
---|
| 55 | *
|
---|
| 56 | * Reference
|
---|
| 57 | * http://www.opengroup.org/onlinepubs/007904975/utilities/sed.html
|
---|
| 58 | * http://pubs.opengroup.org/onlinepubs/9699919799/utilities/sed.html
|
---|
| 59 | */
|
---|
[821] | 60 |
|
---|
[3232] | 61 | //usage:#define sed_trivial_usage
|
---|
| 62 | //usage: "[-inr] [-f FILE]... [-e CMD]... [FILE]...\n"
|
---|
| 63 | //usage: "or: sed [-inr] CMD [FILE]..."
|
---|
| 64 | //usage:#define sed_full_usage "\n\n"
|
---|
| 65 | //usage: " -e CMD Add CMD to sed commands to be executed"
|
---|
| 66 | //usage: "\n -f FILE Add FILE contents to sed commands to be executed"
|
---|
| 67 | //usage: "\n -i[SFX] Edit files in-place (otherwise sends to stdout)"
|
---|
| 68 | //usage: "\n Optionally back files up, appending SFX"
|
---|
| 69 | //usage: "\n -n Suppress automatic printing of pattern space"
|
---|
| 70 | //usage: "\n -r Use extended regex syntax"
|
---|
| 71 | //usage: "\n"
|
---|
| 72 | //usage: "\nIf no -e or -f, the first non-option argument is the sed command string."
|
---|
| 73 | //usage: "\nRemaining arguments are input files (stdin if none)."
|
---|
| 74 | //usage:
|
---|
| 75 | //usage:#define sed_example_usage
|
---|
| 76 | //usage: "$ echo \"foo\" | sed -e 's/f[a-zA-Z]o/bar/g'\n"
|
---|
| 77 | //usage: "bar\n"
|
---|
[821] | 78 |
|
---|
[3232] | 79 | #include "libbb.h"
|
---|
| 80 | #include "xregex.h"
|
---|
[821] | 81 |
|
---|
[3232] | 82 | #if 0
|
---|
| 83 | # define dbg(...) bb_error_msg(__VA_ARGS__)
|
---|
| 84 | #else
|
---|
| 85 | # define dbg(...) ((void)0)
|
---|
| 86 | #endif
|
---|
[821] | 87 |
|
---|
| 88 |
|
---|
[2725] | 89 | enum {
|
---|
| 90 | OPT_in_place = 1 << 0,
|
---|
| 91 | };
|
---|
| 92 |
|
---|
[821] | 93 | /* Each sed command turns into one of these structures. */
|
---|
| 94 | typedef struct sed_cmd_s {
|
---|
[1765] | 95 | /* Ordered by alignment requirements: currently 36 bytes on x86 */
|
---|
| 96 | struct sed_cmd_s *next; /* Next command (linked list, NULL terminated) */
|
---|
[821] | 97 |
|
---|
[1765] | 98 | /* address storage */
|
---|
| 99 | regex_t *beg_match; /* sed -e '/match/cmd' */
|
---|
| 100 | regex_t *end_match; /* sed -e '/match/,/end_match/cmd' */
|
---|
| 101 | regex_t *sub_match; /* For 's/sub_match/string/' */
|
---|
| 102 | int beg_line; /* 'sed 1p' 0 == apply commands to all lines */
|
---|
[3232] | 103 | int beg_line_orig; /* copy of the above, needed for -i */
|
---|
[1765] | 104 | int end_line; /* 'sed 1,3p' 0 == one line only. -1 = last line ($) */
|
---|
[821] | 105 |
|
---|
[1765] | 106 | FILE *sw_file; /* File (sw) command writes to, -1 for none. */
|
---|
| 107 | char *string; /* Data string for (saicytb) commands. */
|
---|
[821] | 108 |
|
---|
[2725] | 109 | unsigned which_match; /* (s) Which match to replace (0 for all) */
|
---|
[821] | 110 |
|
---|
[1765] | 111 | /* Bitfields (gcc won't group them if we don't) */
|
---|
| 112 | unsigned invert:1; /* the '!' after the address */
|
---|
| 113 | unsigned in_match:1; /* Next line also included in match? */
|
---|
| 114 | unsigned sub_p:1; /* (s) print option */
|
---|
[821] | 115 |
|
---|
[1765] | 116 | char sw_last_char; /* Last line written by (sw) had no '\n' */
|
---|
| 117 |
|
---|
| 118 | /* GENERAL FIELDS */
|
---|
| 119 | char cmd; /* The command char: abcdDgGhHilnNpPqrstwxy:={} */
|
---|
[821] | 120 | } sed_cmd_t;
|
---|
| 121 |
|
---|
[1765] | 122 | static const char semicolon_whitespace[] ALIGN1 = "; \n\r\t\v";
|
---|
[821] | 123 |
|
---|
[1765] | 124 | struct globals {
|
---|
[821] | 125 | /* options */
|
---|
[1765] | 126 | int be_quiet, regex_type;
|
---|
[821] | 127 | FILE *nonstdout;
|
---|
| 128 | char *outname, *hold_space;
|
---|
| 129 |
|
---|
| 130 | /* List of input files */
|
---|
[1765] | 131 | int input_file_count, current_input_file;
|
---|
[821] | 132 | FILE **input_file_list;
|
---|
| 133 |
|
---|
| 134 | regmatch_t regmatch[10];
|
---|
| 135 | regex_t *previous_regex_ptr;
|
---|
[1765] | 136 |
|
---|
[821] | 137 | /* linked list of sed commands */
|
---|
[3232] | 138 | sed_cmd_t *sed_cmd_head, **sed_cmd_tail;
|
---|
[821] | 139 |
|
---|
| 140 | /* Linked list of append lines */
|
---|
| 141 | llist_t *append_head;
|
---|
| 142 |
|
---|
| 143 | char *add_cmd_line;
|
---|
| 144 |
|
---|
| 145 | struct pipeline {
|
---|
[2725] | 146 | char *buf; /* Space to hold string */
|
---|
| 147 | int idx; /* Space used */
|
---|
| 148 | int len; /* Space allocated */
|
---|
[821] | 149 | } pipeline;
|
---|
[2725] | 150 | } FIX_ALIASING;
|
---|
| 151 | #define G (*(struct globals*)&bb_common_bufsiz1)
|
---|
| 152 | struct BUG_G_too_big {
|
---|
[3232] | 153 | char BUG_G_too_big[sizeof(G) <= COMMON_BUFSIZE ? 1 : -1];
|
---|
[1765] | 154 | };
|
---|
| 155 | #define INIT_G() do { \
|
---|
| 156 | G.sed_cmd_tail = &G.sed_cmd_head; \
|
---|
| 157 | } while (0)
|
---|
[821] | 158 |
|
---|
| 159 |
|
---|
| 160 | #if ENABLE_FEATURE_CLEAN_UP
|
---|
[1765] | 161 | static void sed_free_and_close_stuff(void)
|
---|
[821] | 162 | {
|
---|
[3232] | 163 | sed_cmd_t *sed_cmd = G.sed_cmd_head;
|
---|
[821] | 164 |
|
---|
[1765] | 165 | llist_free(G.append_head, free);
|
---|
[821] | 166 |
|
---|
| 167 | while (sed_cmd) {
|
---|
| 168 | sed_cmd_t *sed_cmd_next = sed_cmd->next;
|
---|
| 169 |
|
---|
[1765] | 170 | if (sed_cmd->sw_file)
|
---|
| 171 | xprint_and_close_file(sed_cmd->sw_file);
|
---|
[821] | 172 |
|
---|
| 173 | if (sed_cmd->beg_match) {
|
---|
| 174 | regfree(sed_cmd->beg_match);
|
---|
| 175 | free(sed_cmd->beg_match);
|
---|
| 176 | }
|
---|
| 177 | if (sed_cmd->end_match) {
|
---|
| 178 | regfree(sed_cmd->end_match);
|
---|
| 179 | free(sed_cmd->end_match);
|
---|
| 180 | }
|
---|
| 181 | if (sed_cmd->sub_match) {
|
---|
| 182 | regfree(sed_cmd->sub_match);
|
---|
| 183 | free(sed_cmd->sub_match);
|
---|
| 184 | }
|
---|
| 185 | free(sed_cmd->string);
|
---|
| 186 | free(sed_cmd);
|
---|
| 187 | sed_cmd = sed_cmd_next;
|
---|
| 188 | }
|
---|
| 189 |
|
---|
[2725] | 190 | free(G.hold_space);
|
---|
[821] | 191 |
|
---|
[1765] | 192 | while (G.current_input_file < G.input_file_count)
|
---|
| 193 | fclose(G.input_file_list[G.current_input_file++]);
|
---|
[821] | 194 | }
|
---|
[1765] | 195 | #else
|
---|
| 196 | void sed_free_and_close_stuff(void);
|
---|
[821] | 197 | #endif
|
---|
| 198 |
|
---|
| 199 | /* If something bad happens during -i operation, delete temp file */
|
---|
| 200 |
|
---|
| 201 | static void cleanup_outname(void)
|
---|
| 202 | {
|
---|
[1765] | 203 | if (G.outname) unlink(G.outname);
|
---|
[821] | 204 | }
|
---|
| 205 |
|
---|
[2725] | 206 | /* strcpy, replacing "\from" with 'to'. If to is NUL, replacing "\any" with 'any' */
|
---|
[821] | 207 |
|
---|
[1765] | 208 | static void parse_escapes(char *dest, const char *string, int len, char from, char to)
|
---|
[821] | 209 | {
|
---|
[1765] | 210 | int i = 0;
|
---|
[821] | 211 |
|
---|
[1765] | 212 | while (i < len) {
|
---|
| 213 | if (string[i] == '\\') {
|
---|
| 214 | if (!to || string[i+1] == from) {
|
---|
| 215 | *dest++ = to ? to : string[i+1];
|
---|
| 216 | i += 2;
|
---|
[821] | 217 | continue;
|
---|
[1765] | 218 | }
|
---|
| 219 | *dest++ = string[i++];
|
---|
[821] | 220 | }
|
---|
[2725] | 221 | /* TODO: is it safe wrt a string with trailing '\\' ? */
|
---|
[1765] | 222 | *dest++ = string[i++];
|
---|
[821] | 223 | }
|
---|
[2725] | 224 | *dest = '\0';
|
---|
[821] | 225 | }
|
---|
| 226 |
|
---|
[1765] | 227 | static char *copy_parsing_escapes(const char *string, int len)
|
---|
[821] | 228 | {
|
---|
[3232] | 229 | const char *s;
|
---|
[1765] | 230 | char *dest = xmalloc(len + 1);
|
---|
[821] | 231 |
|
---|
[3232] | 232 | /* sed recognizes \n */
|
---|
| 233 | /* GNU sed also recognizes \t and \r */
|
---|
| 234 | for (s = "\nn\tt\rr"; *s; s += 2) {
|
---|
| 235 | parse_escapes(dest, string, len, s[1], s[0]);
|
---|
| 236 | string = dest;
|
---|
| 237 | len = strlen(dest);
|
---|
| 238 | }
|
---|
[821] | 239 | return dest;
|
---|
| 240 | }
|
---|
| 241 |
|
---|
| 242 |
|
---|
| 243 | /*
|
---|
| 244 | * index_of_next_unescaped_regexp_delim - walks left to right through a string
|
---|
| 245 | * beginning at a specified index and returns the index of the next regular
|
---|
[2725] | 246 | * expression delimiter (typically a forward slash ('/')) not preceded by
|
---|
[902] | 247 | * a backslash ('\'). A negative delimiter disables square bracket checking.
|
---|
[821] | 248 | */
|
---|
[1765] | 249 | static int index_of_next_unescaped_regexp_delim(int delimiter, const char *str)
|
---|
[821] | 250 | {
|
---|
| 251 | int bracket = -1;
|
---|
| 252 | int escaped = 0;
|
---|
| 253 | int idx = 0;
|
---|
| 254 | char ch;
|
---|
| 255 |
|
---|
[902] | 256 | if (delimiter < 0) {
|
---|
| 257 | bracket--;
|
---|
[1765] | 258 | delimiter = -delimiter;
|
---|
[902] | 259 | }
|
---|
| 260 |
|
---|
[3232] | 261 | for (; (ch = str[idx]) != '\0'; idx++) {
|
---|
[902] | 262 | if (bracket >= 0) {
|
---|
[3232] | 263 | if (ch == ']'
|
---|
| 264 | && !(bracket == idx - 1 || (bracket == idx - 2 && str[idx - 1] == '^'))
|
---|
| 265 | ) {
|
---|
[821] | 266 | bracket = -1;
|
---|
[3232] | 267 | }
|
---|
[821] | 268 | } else if (escaped)
|
---|
| 269 | escaped = 0;
|
---|
| 270 | else if (ch == '\\')
|
---|
| 271 | escaped = 1;
|
---|
[902] | 272 | else if (bracket == -1 && ch == '[')
|
---|
[821] | 273 | bracket = idx;
|
---|
| 274 | else if (ch == delimiter)
|
---|
| 275 | return idx;
|
---|
| 276 | }
|
---|
| 277 |
|
---|
| 278 | /* if we make it to here, we've hit the end of the string */
|
---|
[1765] | 279 | bb_error_msg_and_die("unmatched '%c'", delimiter);
|
---|
[821] | 280 | }
|
---|
| 281 |
|
---|
| 282 | /*
|
---|
| 283 | * Returns the index of the third delimiter
|
---|
| 284 | */
|
---|
[1765] | 285 | static int parse_regex_delim(const char *cmdstr, char **match, char **replace)
|
---|
[821] | 286 | {
|
---|
[1765] | 287 | const char *cmdstr_ptr = cmdstr;
|
---|
[3232] | 288 | unsigned char delimiter;
|
---|
[821] | 289 | int idx = 0;
|
---|
| 290 |
|
---|
| 291 | /* verify that the 's' or 'y' is followed by something. That something
|
---|
| 292 | * (typically a 'slash') is now our regexp delimiter... */
|
---|
[902] | 293 | if (*cmdstr == '\0')
|
---|
| 294 | bb_error_msg_and_die("bad format in substitution expression");
|
---|
[1765] | 295 | delimiter = *cmdstr_ptr++;
|
---|
[821] | 296 |
|
---|
| 297 | /* save the match string */
|
---|
| 298 | idx = index_of_next_unescaped_regexp_delim(delimiter, cmdstr_ptr);
|
---|
| 299 | *match = copy_parsing_escapes(cmdstr_ptr, idx);
|
---|
| 300 |
|
---|
| 301 | /* save the replacement string */
|
---|
| 302 | cmdstr_ptr += idx + 1;
|
---|
[3232] | 303 | idx = index_of_next_unescaped_regexp_delim(- (int)delimiter, cmdstr_ptr);
|
---|
[821] | 304 | *replace = copy_parsing_escapes(cmdstr_ptr, idx);
|
---|
| 305 |
|
---|
| 306 | return ((cmdstr_ptr - cmdstr) + idx);
|
---|
| 307 | }
|
---|
| 308 |
|
---|
| 309 | /*
|
---|
| 310 | * returns the index in the string just past where the address ends.
|
---|
| 311 | */
|
---|
[1765] | 312 | static int get_address(const char *my_str, int *linenum, regex_t ** regex)
|
---|
[821] | 313 | {
|
---|
[1765] | 314 | const char *pos = my_str;
|
---|
[821] | 315 |
|
---|
| 316 | if (isdigit(*my_str)) {
|
---|
[1765] | 317 | *linenum = strtol(my_str, (char**)&pos, 10);
|
---|
[821] | 318 | /* endstr shouldnt ever equal NULL */
|
---|
| 319 | } else if (*my_str == '$') {
|
---|
| 320 | *linenum = -1;
|
---|
| 321 | pos++;
|
---|
| 322 | } else if (*my_str == '/' || *my_str == '\\') {
|
---|
| 323 | int next;
|
---|
| 324 | char delimiter;
|
---|
| 325 | char *temp;
|
---|
| 326 |
|
---|
[1765] | 327 | delimiter = '/';
|
---|
[3232] | 328 | if (*my_str == '\\')
|
---|
| 329 | delimiter = *++pos;
|
---|
[821] | 330 | next = index_of_next_unescaped_regexp_delim(delimiter, ++pos);
|
---|
[1765] | 331 | temp = copy_parsing_escapes(pos, next);
|
---|
[3232] | 332 | *regex = xzalloc(sizeof(regex_t));
|
---|
[1765] | 333 | xregcomp(*regex, temp, G.regex_type|REG_NEWLINE);
|
---|
[821] | 334 | free(temp);
|
---|
| 335 | /* Move position to next character after last delimiter */
|
---|
[902] | 336 | pos += (next+1);
|
---|
[821] | 337 | }
|
---|
| 338 | return pos - my_str;
|
---|
| 339 | }
|
---|
| 340 |
|
---|
| 341 | /* Grab a filename. Whitespace at start is skipped, then goes to EOL. */
|
---|
[2725] | 342 | static int parse_file_cmd(/*sed_cmd_t *sed_cmd,*/ const char *filecmdstr, char **retval)
|
---|
[821] | 343 | {
|
---|
[1765] | 344 | int start = 0, idx, hack = 0;
|
---|
[821] | 345 |
|
---|
| 346 | /* Skip whitespace, then grab filename to end of line */
|
---|
[1765] | 347 | while (isspace(filecmdstr[start]))
|
---|
| 348 | start++;
|
---|
| 349 | idx = start;
|
---|
| 350 | while (filecmdstr[idx] && filecmdstr[idx] != '\n')
|
---|
| 351 | idx++;
|
---|
| 352 |
|
---|
[821] | 353 | /* If lines glued together, put backslash back. */
|
---|
[1765] | 354 | if (filecmdstr[idx] == '\n')
|
---|
| 355 | hack = 1;
|
---|
| 356 | if (idx == start)
|
---|
| 357 | bb_error_msg_and_die("empty filename");
|
---|
| 358 | *retval = xstrndup(filecmdstr+start, idx-start+hack+1);
|
---|
| 359 | if (hack)
|
---|
| 360 | (*retval)[idx] = '\\';
|
---|
[821] | 361 |
|
---|
| 362 | return idx;
|
---|
| 363 | }
|
---|
| 364 |
|
---|
[1765] | 365 | static int parse_subst_cmd(sed_cmd_t *sed_cmd, const char *substr)
|
---|
[821] | 366 | {
|
---|
[1765] | 367 | int cflags = G.regex_type;
|
---|
[821] | 368 | char *match;
|
---|
[1765] | 369 | int idx;
|
---|
[821] | 370 |
|
---|
| 371 | /*
|
---|
| 372 | * A substitution command should look something like this:
|
---|
| 373 | * s/match/replace/ #gIpw
|
---|
| 374 | * || | |||
|
---|
| 375 | * mandatory optional
|
---|
| 376 | */
|
---|
| 377 | idx = parse_regex_delim(substr, &match, &sed_cmd->string);
|
---|
| 378 |
|
---|
| 379 | /* determine the number of back references in the match string */
|
---|
| 380 | /* Note: we compute this here rather than in the do_subst_command()
|
---|
| 381 | * function to save processor time, at the expense of a little more memory
|
---|
| 382 | * (4 bits) per sed_cmd */
|
---|
| 383 |
|
---|
| 384 | /* process the flags */
|
---|
| 385 |
|
---|
[1765] | 386 | sed_cmd->which_match = 1;
|
---|
[821] | 387 | while (substr[++idx]) {
|
---|
| 388 | /* Parse match number */
|
---|
[1765] | 389 | if (isdigit(substr[idx])) {
|
---|
| 390 | if (match[0] != '^') {
|
---|
[821] | 391 | /* Match 0 treated as all, multiple matches we take the last one. */
|
---|
[1765] | 392 | const char *pos = substr + idx;
|
---|
| 393 | /* FIXME: error check? */
|
---|
[2725] | 394 | sed_cmd->which_match = (unsigned)strtol(substr+idx, (char**) &pos, 10);
|
---|
[1765] | 395 | idx = pos - substr;
|
---|
[821] | 396 | }
|
---|
| 397 | continue;
|
---|
| 398 | }
|
---|
| 399 | /* Skip spaces */
|
---|
[2725] | 400 | if (isspace(substr[idx]))
|
---|
| 401 | continue;
|
---|
[821] | 402 |
|
---|
| 403 | switch (substr[idx]) {
|
---|
[1765] | 404 | /* Replace all occurrences */
|
---|
| 405 | case 'g':
|
---|
[2725] | 406 | if (match[0] != '^')
|
---|
| 407 | sed_cmd->which_match = 0;
|
---|
[1765] | 408 | break;
|
---|
| 409 | /* Print pattern space */
|
---|
| 410 | case 'p':
|
---|
| 411 | sed_cmd->sub_p = 1;
|
---|
| 412 | break;
|
---|
| 413 | /* Write to file */
|
---|
| 414 | case 'w':
|
---|
| 415 | {
|
---|
| 416 | char *temp;
|
---|
[2725] | 417 | idx += parse_file_cmd(/*sed_cmd,*/ substr+idx, &temp);
|
---|
[1765] | 418 | break;
|
---|
[821] | 419 | }
|
---|
[1765] | 420 | /* Ignore case (gnu exension) */
|
---|
| 421 | case 'I':
|
---|
| 422 | cflags |= REG_ICASE;
|
---|
| 423 | break;
|
---|
| 424 | /* Comment */
|
---|
| 425 | case '#':
|
---|
[2725] | 426 | // while (substr[++idx]) continue;
|
---|
| 427 | idx += strlen(substr + idx); // same
|
---|
[1765] | 428 | /* Fall through */
|
---|
| 429 | /* End of command */
|
---|
| 430 | case ';':
|
---|
| 431 | case '}':
|
---|
| 432 | goto out;
|
---|
| 433 | default:
|
---|
| 434 | bb_error_msg_and_die("bad option in substitution expression");
|
---|
| 435 | }
|
---|
[821] | 436 | }
|
---|
[2725] | 437 | out:
|
---|
[821] | 438 | /* compile the match string into a regex */
|
---|
| 439 | if (*match != '\0') {
|
---|
| 440 | /* If match is empty, we use last regex used at runtime */
|
---|
[3232] | 441 | sed_cmd->sub_match = xzalloc(sizeof(regex_t));
|
---|
| 442 | dbg("xregcomp('%s',%x)", match, cflags);
|
---|
[821] | 443 | xregcomp(sed_cmd->sub_match, match, cflags);
|
---|
[3232] | 444 | dbg("regcomp ok");
|
---|
[821] | 445 | }
|
---|
| 446 | free(match);
|
---|
| 447 |
|
---|
| 448 | return idx;
|
---|
| 449 | }
|
---|
| 450 |
|
---|
| 451 | /*
|
---|
| 452 | * Process the commands arguments
|
---|
| 453 | */
|
---|
[1765] | 454 | static const char *parse_cmd_args(sed_cmd_t *sed_cmd, const char *cmdstr)
|
---|
[821] | 455 | {
|
---|
[3232] | 456 | static const char cmd_letters[] = "saicrw:btTydDgGhHlnNpPqx={}";
|
---|
| 457 | enum {
|
---|
| 458 | IDX_s = 0,
|
---|
| 459 | IDX_a,
|
---|
| 460 | IDX_i,
|
---|
| 461 | IDX_c,
|
---|
| 462 | IDX_r,
|
---|
| 463 | IDX_w,
|
---|
| 464 | IDX_colon,
|
---|
| 465 | IDX_b,
|
---|
| 466 | IDX_t,
|
---|
| 467 | IDX_T,
|
---|
| 468 | IDX_y,
|
---|
| 469 | IDX_d,
|
---|
| 470 | IDX_D,
|
---|
| 471 | IDX_g,
|
---|
| 472 | IDX_G,
|
---|
| 473 | IDX_h,
|
---|
| 474 | IDX_H,
|
---|
| 475 | IDX_l,
|
---|
| 476 | IDX_n,
|
---|
| 477 | IDX_N,
|
---|
| 478 | IDX_p,
|
---|
| 479 | IDX_P,
|
---|
| 480 | IDX_q,
|
---|
| 481 | IDX_x,
|
---|
| 482 | IDX_equal,
|
---|
| 483 | IDX_lbrace,
|
---|
| 484 | IDX_rbrace,
|
---|
| 485 | IDX_nul
|
---|
| 486 | };
|
---|
| 487 | struct chk { char chk[sizeof(cmd_letters)-1 == IDX_nul ? 1 : -1]; };
|
---|
| 488 |
|
---|
| 489 | unsigned idx = strchrnul(cmd_letters, sed_cmd->cmd) - cmd_letters;
|
---|
| 490 |
|
---|
[821] | 491 | /* handle (s)ubstitution command */
|
---|
[3232] | 492 | if (idx == IDX_s) {
|
---|
[1765] | 493 | cmdstr += parse_subst_cmd(sed_cmd, cmdstr);
|
---|
[3232] | 494 | }
|
---|
[821] | 495 | /* handle edit cmds: (a)ppend, (i)nsert, and (c)hange */
|
---|
[3232] | 496 | else if (idx <= IDX_c) { /* a,i,c */
|
---|
| 497 | if (idx < IDX_c) { /* a,i */
|
---|
| 498 | if (sed_cmd->end_line || sed_cmd->end_match)
|
---|
| 499 | bb_error_msg_and_die("command '%c' uses only one address", sed_cmd->cmd);
|
---|
| 500 | }
|
---|
[1765] | 501 | for (;;) {
|
---|
| 502 | if (*cmdstr == '\n' || *cmdstr == '\\') {
|
---|
[821] | 503 | cmdstr++;
|
---|
| 504 | break;
|
---|
[2725] | 505 | }
|
---|
| 506 | if (!isspace(*cmdstr))
|
---|
[1765] | 507 | break;
|
---|
[2725] | 508 | cmdstr++;
|
---|
[821] | 509 | }
|
---|
[1765] | 510 | sed_cmd->string = xstrdup(cmdstr);
|
---|
[2725] | 511 | /* "\anychar" -> "anychar" */
|
---|
| 512 | parse_escapes(sed_cmd->string, sed_cmd->string, strlen(cmdstr), '\0', '\0');
|
---|
[821] | 513 | cmdstr += strlen(cmdstr);
|
---|
[3232] | 514 | }
|
---|
[821] | 515 | /* handle file cmds: (r)ead */
|
---|
[3232] | 516 | else if (idx <= IDX_w) { /* r,w */
|
---|
| 517 | if (idx < IDX_w) { /* r */
|
---|
| 518 | if (sed_cmd->end_line || sed_cmd->end_match)
|
---|
| 519 | bb_error_msg_and_die("command '%c' uses only one address", sed_cmd->cmd);
|
---|
| 520 | }
|
---|
[2725] | 521 | cmdstr += parse_file_cmd(/*sed_cmd,*/ cmdstr, &sed_cmd->string);
|
---|
[1765] | 522 | if (sed_cmd->cmd == 'w') {
|
---|
[2725] | 523 | sed_cmd->sw_file = xfopen_for_write(sed_cmd->string);
|
---|
[1765] | 524 | sed_cmd->sw_last_char = '\n';
|
---|
| 525 | }
|
---|
[3232] | 526 | }
|
---|
[821] | 527 | /* handle branch commands */
|
---|
[3232] | 528 | else if (idx <= IDX_T) { /* :,b,t,T */
|
---|
[821] | 529 | int length;
|
---|
| 530 |
|
---|
[1765] | 531 | cmdstr = skip_whitespace(cmdstr);
|
---|
[821] | 532 | length = strcspn(cmdstr, semicolon_whitespace);
|
---|
| 533 | if (length) {
|
---|
[1765] | 534 | sed_cmd->string = xstrndup(cmdstr, length);
|
---|
[821] | 535 | cmdstr += length;
|
---|
| 536 | }
|
---|
| 537 | }
|
---|
| 538 | /* translation command */
|
---|
[3232] | 539 | else if (idx == IDX_y) {
|
---|
[821] | 540 | char *match, *replace;
|
---|
[1765] | 541 | int i = cmdstr[0];
|
---|
[821] | 542 |
|
---|
[1765] | 543 | cmdstr += parse_regex_delim(cmdstr, &match, &replace)+1;
|
---|
[821] | 544 | /* \n already parsed, but \delimiter needs unescaping. */
|
---|
[1765] | 545 | parse_escapes(match, match, strlen(match), i, i);
|
---|
| 546 | parse_escapes(replace, replace, strlen(replace), i, i);
|
---|
[821] | 547 |
|
---|
| 548 | sed_cmd->string = xzalloc((strlen(match) + 1) * 2);
|
---|
| 549 | for (i = 0; match[i] && replace[i]; i++) {
|
---|
[1765] | 550 | sed_cmd->string[i*2] = match[i];
|
---|
| 551 | sed_cmd->string[i*2+1] = replace[i];
|
---|
[821] | 552 | }
|
---|
| 553 | free(match);
|
---|
| 554 | free(replace);
|
---|
| 555 | }
|
---|
| 556 | /* if it wasnt a single-letter command that takes no arguments
|
---|
| 557 | * then it must be an invalid command.
|
---|
| 558 | */
|
---|
[3232] | 559 | else if (idx >= IDX_nul) { /* not d,D,g,G,h,H,l,n,N,p,P,q,x,=,{,} */
|
---|
[1765] | 560 | bb_error_msg_and_die("unsupported command %c", sed_cmd->cmd);
|
---|
[821] | 561 | }
|
---|
| 562 |
|
---|
| 563 | /* give back whatever's left over */
|
---|
[1765] | 564 | return cmdstr;
|
---|
[821] | 565 | }
|
---|
| 566 |
|
---|
| 567 |
|
---|
| 568 | /* Parse address+command sets, skipping comment lines. */
|
---|
| 569 |
|
---|
[1765] | 570 | static void add_cmd(const char *cmdstr)
|
---|
[821] | 571 | {
|
---|
| 572 | sed_cmd_t *sed_cmd;
|
---|
[2725] | 573 | unsigned len, n;
|
---|
[821] | 574 |
|
---|
| 575 | /* Append this line to any unfinished line from last time. */
|
---|
[1765] | 576 | if (G.add_cmd_line) {
|
---|
| 577 | char *tp = xasprintf("%s\n%s", G.add_cmd_line, cmdstr);
|
---|
| 578 | free(G.add_cmd_line);
|
---|
| 579 | cmdstr = G.add_cmd_line = tp;
|
---|
[821] | 580 | }
|
---|
| 581 |
|
---|
[2725] | 582 | /* If this line ends with unescaped backslash, request next line. */
|
---|
| 583 | n = len = strlen(cmdstr);
|
---|
| 584 | while (n && cmdstr[n-1] == '\\')
|
---|
| 585 | n--;
|
---|
| 586 | if ((len - n) & 1) { /* if odd number of trailing backslashes */
|
---|
[1765] | 587 | if (!G.add_cmd_line)
|
---|
| 588 | G.add_cmd_line = xstrdup(cmdstr);
|
---|
[2725] | 589 | G.add_cmd_line[len-1] = '\0';
|
---|
[821] | 590 | return;
|
---|
| 591 | }
|
---|
| 592 |
|
---|
| 593 | /* Loop parsing all commands in this line. */
|
---|
[1765] | 594 | while (*cmdstr) {
|
---|
[821] | 595 | /* Skip leading whitespace and semicolons */
|
---|
| 596 | cmdstr += strspn(cmdstr, semicolon_whitespace);
|
---|
| 597 |
|
---|
| 598 | /* If no more commands, exit. */
|
---|
[1765] | 599 | if (!*cmdstr) break;
|
---|
[821] | 600 |
|
---|
| 601 | /* if this is a comment, jump past it and keep going */
|
---|
| 602 | if (*cmdstr == '#') {
|
---|
| 603 | /* "#n" is the same as using -n on the command line */
|
---|
[1765] | 604 | if (cmdstr[1] == 'n')
|
---|
| 605 | G.be_quiet++;
|
---|
| 606 | cmdstr = strpbrk(cmdstr, "\n\r");
|
---|
| 607 | if (!cmdstr) break;
|
---|
[821] | 608 | continue;
|
---|
| 609 | }
|
---|
| 610 |
|
---|
| 611 | /* parse the command
|
---|
| 612 | * format is: [addr][,addr][!]cmd
|
---|
| 613 | * |----||-----||-|
|
---|
| 614 | * part1 part2 part3
|
---|
| 615 | */
|
---|
| 616 |
|
---|
| 617 | sed_cmd = xzalloc(sizeof(sed_cmd_t));
|
---|
| 618 |
|
---|
| 619 | /* first part (if present) is an address: either a '$', a number or a /regex/ */
|
---|
| 620 | cmdstr += get_address(cmdstr, &sed_cmd->beg_line, &sed_cmd->beg_match);
|
---|
[3232] | 621 | sed_cmd->beg_line_orig = sed_cmd->beg_line;
|
---|
[821] | 622 |
|
---|
| 623 | /* second part (if present) will begin with a comma */
|
---|
| 624 | if (*cmdstr == ',') {
|
---|
| 625 | int idx;
|
---|
| 626 |
|
---|
| 627 | cmdstr++;
|
---|
| 628 | idx = get_address(cmdstr, &sed_cmd->end_line, &sed_cmd->end_match);
|
---|
[1765] | 629 | if (!idx)
|
---|
| 630 | bb_error_msg_and_die("no address after comma");
|
---|
[821] | 631 | cmdstr += idx;
|
---|
| 632 | }
|
---|
| 633 |
|
---|
| 634 | /* skip whitespace before the command */
|
---|
[1765] | 635 | cmdstr = skip_whitespace(cmdstr);
|
---|
[821] | 636 |
|
---|
| 637 | /* Check for inversion flag */
|
---|
| 638 | if (*cmdstr == '!') {
|
---|
| 639 | sed_cmd->invert = 1;
|
---|
| 640 | cmdstr++;
|
---|
| 641 |
|
---|
| 642 | /* skip whitespace before the command */
|
---|
[1765] | 643 | cmdstr = skip_whitespace(cmdstr);
|
---|
[821] | 644 | }
|
---|
| 645 |
|
---|
| 646 | /* last part (mandatory) will be a command */
|
---|
[1765] | 647 | if (!*cmdstr)
|
---|
| 648 | bb_error_msg_and_die("missing command");
|
---|
[2725] | 649 | sed_cmd->cmd = *cmdstr++;
|
---|
[821] | 650 | cmdstr = parse_cmd_args(sed_cmd, cmdstr);
|
---|
| 651 |
|
---|
| 652 | /* Add the command to the command array */
|
---|
[3232] | 653 | *G.sed_cmd_tail = sed_cmd;
|
---|
| 654 | G.sed_cmd_tail = &sed_cmd->next;
|
---|
[821] | 655 | }
|
---|
| 656 |
|
---|
| 657 | /* If we glued multiple lines together, free the memory. */
|
---|
[1765] | 658 | free(G.add_cmd_line);
|
---|
| 659 | G.add_cmd_line = NULL;
|
---|
[821] | 660 | }
|
---|
| 661 |
|
---|
| 662 | /* Append to a string, reallocating memory as necessary. */
|
---|
| 663 |
|
---|
| 664 | #define PIPE_GROW 64
|
---|
| 665 |
|
---|
| 666 | static void pipe_putc(char c)
|
---|
| 667 | {
|
---|
[1765] | 668 | if (G.pipeline.idx == G.pipeline.len) {
|
---|
| 669 | G.pipeline.buf = xrealloc(G.pipeline.buf,
|
---|
| 670 | G.pipeline.len + PIPE_GROW);
|
---|
| 671 | G.pipeline.len += PIPE_GROW;
|
---|
[821] | 672 | }
|
---|
[1765] | 673 | G.pipeline.buf[G.pipeline.idx++] = c;
|
---|
[821] | 674 | }
|
---|
| 675 |
|
---|
[902] | 676 | static void do_subst_w_backrefs(char *line, char *replace)
|
---|
[821] | 677 | {
|
---|
[2725] | 678 | int i, j;
|
---|
[821] | 679 |
|
---|
| 680 | /* go through the replacement string */
|
---|
| 681 | for (i = 0; replace[i]; i++) {
|
---|
[3232] | 682 | /* if we find a backreference (\1, \2, etc.) print the backref'ed text */
|
---|
[1765] | 683 | if (replace[i] == '\\') {
|
---|
| 684 | unsigned backref = replace[++i] - '0';
|
---|
| 685 | if (backref <= 9) {
|
---|
| 686 | /* print out the text held in G.regmatch[backref] */
|
---|
| 687 | if (G.regmatch[backref].rm_so != -1) {
|
---|
| 688 | j = G.regmatch[backref].rm_so;
|
---|
| 689 | while (j < G.regmatch[backref].rm_eo)
|
---|
| 690 | pipe_putc(line[j++]);
|
---|
| 691 | }
|
---|
| 692 | continue;
|
---|
| 693 | }
|
---|
| 694 | /* I _think_ it is impossible to get '\' to be
|
---|
| 695 | * the last char in replace string. Thus we dont check
|
---|
| 696 | * for replace[i] == NUL. (counterexample anyone?) */
|
---|
| 697 | /* if we find a backslash escaped character, print the character */
|
---|
| 698 | pipe_putc(replace[i]);
|
---|
| 699 | continue;
|
---|
[821] | 700 | }
|
---|
| 701 | /* if we find an unescaped '&' print out the whole matched text. */
|
---|
[1765] | 702 | if (replace[i] == '&') {
|
---|
| 703 | j = G.regmatch[0].rm_so;
|
---|
| 704 | while (j < G.regmatch[0].rm_eo)
|
---|
| 705 | pipe_putc(line[j++]);
|
---|
| 706 | continue;
|
---|
| 707 | }
|
---|
[821] | 708 | /* Otherwise just output the character. */
|
---|
[1765] | 709 | pipe_putc(replace[i]);
|
---|
[821] | 710 | }
|
---|
| 711 | }
|
---|
| 712 |
|
---|
[2725] | 713 | static int do_subst_command(sed_cmd_t *sed_cmd, char **line_p)
|
---|
[821] | 714 | {
|
---|
[2725] | 715 | char *line = *line_p;
|
---|
| 716 | unsigned match_count = 0;
|
---|
[3232] | 717 | bool altered = 0;
|
---|
| 718 | bool prev_match_empty = 1;
|
---|
| 719 | bool tried_at_eol = 0;
|
---|
[821] | 720 | regex_t *current_regex;
|
---|
| 721 |
|
---|
[2725] | 722 | current_regex = sed_cmd->sub_match;
|
---|
[821] | 723 | /* Handle empty regex. */
|
---|
[2725] | 724 | if (!current_regex) {
|
---|
[1765] | 725 | current_regex = G.previous_regex_ptr;
|
---|
| 726 | if (!current_regex)
|
---|
| 727 | bb_error_msg_and_die("no previous regexp");
|
---|
[2725] | 728 | }
|
---|
| 729 | G.previous_regex_ptr = current_regex;
|
---|
[821] | 730 |
|
---|
| 731 | /* Find the first match */
|
---|
[3232] | 732 | dbg("matching '%s'", line);
|
---|
| 733 | if (REG_NOMATCH == regexec(current_regex, line, 10, G.regmatch, 0)) {
|
---|
| 734 | dbg("no match");
|
---|
[821] | 735 | return 0;
|
---|
[3232] | 736 | }
|
---|
| 737 | dbg("match");
|
---|
[821] | 738 |
|
---|
| 739 | /* Initialize temporary output buffer. */
|
---|
[1765] | 740 | G.pipeline.buf = xmalloc(PIPE_GROW);
|
---|
| 741 | G.pipeline.len = PIPE_GROW;
|
---|
| 742 | G.pipeline.idx = 0;
|
---|
[821] | 743 |
|
---|
| 744 | /* Now loop through, substituting for matches */
|
---|
| 745 | do {
|
---|
[3232] | 746 | int start = G.regmatch[0].rm_so;
|
---|
| 747 | int end = G.regmatch[0].rm_eo;
|
---|
[821] | 748 | int i;
|
---|
| 749 |
|
---|
| 750 | match_count++;
|
---|
| 751 |
|
---|
| 752 | /* If we aren't interested in this match, output old line to
|
---|
[3232] | 753 | * end of match and continue */
|
---|
[2725] | 754 | if (sed_cmd->which_match
|
---|
| 755 | && (sed_cmd->which_match != match_count)
|
---|
| 756 | ) {
|
---|
[3232] | 757 | for (i = 0; i < end; i++)
|
---|
[2725] | 758 | pipe_putc(*line++);
|
---|
[3232] | 759 | /* Null match? Print one more char */
|
---|
| 760 | if (start == end && *line)
|
---|
| 761 | pipe_putc(*line++);
|
---|
| 762 | goto next;
|
---|
[821] | 763 | }
|
---|
| 764 |
|
---|
[3232] | 765 | /* Print everything before the match */
|
---|
| 766 | for (i = 0; i < start; i++)
|
---|
[2725] | 767 | pipe_putc(line[i]);
|
---|
[821] | 768 |
|
---|
[3232] | 769 | /* Then print the substitution string,
|
---|
| 770 | * unless we just matched empty string after non-empty one.
|
---|
| 771 | * Example: string "cccd", pattern "c*", repl "R":
|
---|
| 772 | * result is "RdR", not "RRdR": first match "ccc",
|
---|
| 773 | * second is "" before "d", third is "" after "d".
|
---|
| 774 | * Second match is NOT replaced!
|
---|
| 775 | */
|
---|
| 776 | if (prev_match_empty || start != 0 || start != end) {
|
---|
| 777 | //dbg("%d %d %d", prev_match_empty, start, end);
|
---|
| 778 | dbg("inserting replacement at %d in '%s'", start, line);
|
---|
| 779 | do_subst_w_backrefs(line, sed_cmd->string);
|
---|
| 780 | /* Flag that something has changed */
|
---|
| 781 | altered = 1;
|
---|
| 782 | } else {
|
---|
| 783 | dbg("NOT inserting replacement at %d in '%s'", start, line);
|
---|
| 784 | }
|
---|
[821] | 785 |
|
---|
[3232] | 786 | /* If matched string is empty (f.e. "c*" pattern),
|
---|
| 787 | * copy verbatim one char after it before attempting more matches
|
---|
| 788 | */
|
---|
| 789 | prev_match_empty = (start == end);
|
---|
| 790 | if (prev_match_empty) {
|
---|
| 791 | if (!line[end]) {
|
---|
| 792 | tried_at_eol = 1;
|
---|
| 793 | } else {
|
---|
| 794 | pipe_putc(line[end]);
|
---|
| 795 | end++;
|
---|
| 796 | }
|
---|
| 797 | }
|
---|
[821] | 798 |
|
---|
[3232] | 799 | /* Advance past the match */
|
---|
| 800 | dbg("line += %d", end);
|
---|
| 801 | line += end;
|
---|
| 802 |
|
---|
[821] | 803 | /* if we're not doing this globally, get out now */
|
---|
[3232] | 804 | if (sed_cmd->which_match != 0)
|
---|
[2725] | 805 | break;
|
---|
[3232] | 806 | next:
|
---|
| 807 | /* Exit if we are at EOL and already tried matching at it */
|
---|
| 808 | if (*line == '\0') {
|
---|
| 809 | if (tried_at_eol)
|
---|
| 810 | break;
|
---|
| 811 | tried_at_eol = 1;
|
---|
| 812 | }
|
---|
[821] | 813 |
|
---|
[3232] | 814 | //maybe (end ? REG_NOTBOL : 0) instead of unconditional REG_NOTBOL?
|
---|
| 815 | } while (regexec(current_regex, line, 10, G.regmatch, REG_NOTBOL) != REG_NOMATCH);
|
---|
[2725] | 816 |
|
---|
[821] | 817 | /* Copy rest of string into output pipeline */
|
---|
[2725] | 818 | while (1) {
|
---|
| 819 | char c = *line++;
|
---|
| 820 | pipe_putc(c);
|
---|
| 821 | if (c == '\0')
|
---|
| 822 | break;
|
---|
| 823 | }
|
---|
[821] | 824 |
|
---|
[2725] | 825 | free(*line_p);
|
---|
| 826 | *line_p = G.pipeline.buf;
|
---|
[821] | 827 | return altered;
|
---|
| 828 | }
|
---|
| 829 |
|
---|
| 830 | /* Set command pointer to point to this label. (Does not handle null label.) */
|
---|
[902] | 831 | static sed_cmd_t *branch_to(char *label)
|
---|
[821] | 832 | {
|
---|
| 833 | sed_cmd_t *sed_cmd;
|
---|
| 834 |
|
---|
[3232] | 835 | for (sed_cmd = G.sed_cmd_head; sed_cmd; sed_cmd = sed_cmd->next) {
|
---|
[1765] | 836 | if (sed_cmd->cmd == ':' && sed_cmd->string && !strcmp(sed_cmd->string, label)) {
|
---|
| 837 | return sed_cmd;
|
---|
[821] | 838 | }
|
---|
| 839 | }
|
---|
[1765] | 840 | bb_error_msg_and_die("can't find label for jump to '%s'", label);
|
---|
[821] | 841 | }
|
---|
| 842 |
|
---|
| 843 | static void append(char *s)
|
---|
| 844 | {
|
---|
[1765] | 845 | llist_add_to_end(&G.append_head, xstrdup(s));
|
---|
[821] | 846 | }
|
---|
| 847 |
|
---|
| 848 | static void flush_append(void)
|
---|
| 849 | {
|
---|
| 850 | char *data;
|
---|
| 851 |
|
---|
| 852 | /* Output appended lines. */
|
---|
[1765] | 853 | while ((data = (char *)llist_pop(&G.append_head))) {
|
---|
| 854 | fprintf(G.nonstdout, "%s\n", data);
|
---|
[821] | 855 | free(data);
|
---|
| 856 | }
|
---|
| 857 | }
|
---|
| 858 |
|
---|
| 859 | static void add_input_file(FILE *file)
|
---|
| 860 | {
|
---|
[2725] | 861 | G.input_file_list = xrealloc_vector(G.input_file_list, 2, G.input_file_count);
|
---|
[1765] | 862 | G.input_file_list[G.input_file_count++] = file;
|
---|
[821] | 863 | }
|
---|
| 864 |
|
---|
[1765] | 865 | /* Get next line of input from G.input_file_list, flushing append buffer and
|
---|
[821] | 866 | * noting if we ran out of files without a newline on the last line we read.
|
---|
| 867 | */
|
---|
[1765] | 868 | enum {
|
---|
| 869 | NO_EOL_CHAR = 1,
|
---|
| 870 | LAST_IS_NUL = 2,
|
---|
| 871 | };
|
---|
| 872 | static char *get_next_line(char *gets_char)
|
---|
[821] | 873 | {
|
---|
[1765] | 874 | char *temp = NULL;
|
---|
[821] | 875 | int len;
|
---|
[1765] | 876 | char gc;
|
---|
[821] | 877 |
|
---|
| 878 | flush_append();
|
---|
[1765] | 879 |
|
---|
| 880 | /* will be returned if last line in the file
|
---|
| 881 | * doesn't end with either '\n' or '\0' */
|
---|
| 882 | gc = NO_EOL_CHAR;
|
---|
| 883 | while (G.current_input_file < G.input_file_count) {
|
---|
| 884 | FILE *fp = G.input_file_list[G.current_input_file];
|
---|
| 885 | /* Read line up to a newline or NUL byte, inclusive,
|
---|
| 886 | * return malloc'ed char[]. length of the chunk read
|
---|
| 887 | * is stored in len. NULL if EOF/error */
|
---|
| 888 | temp = bb_get_chunk_from_file(fp, &len);
|
---|
[821] | 889 | if (temp) {
|
---|
[1765] | 890 | /* len > 0 here, it's ok to do temp[len-1] */
|
---|
| 891 | char c = temp[len-1];
|
---|
| 892 | if (c == '\n' || c == '\0') {
|
---|
| 893 | temp[len-1] = '\0';
|
---|
| 894 | gc = c;
|
---|
| 895 | if (c == '\0') {
|
---|
| 896 | int ch = fgetc(fp);
|
---|
| 897 | if (ch != EOF)
|
---|
| 898 | ungetc(ch, fp);
|
---|
| 899 | else
|
---|
| 900 | gc = LAST_IS_NUL;
|
---|
| 901 | }
|
---|
| 902 | }
|
---|
| 903 | /* else we put NO_EOL_CHAR into *gets_char */
|
---|
[821] | 904 | break;
|
---|
[1765] | 905 |
|
---|
| 906 | /* NB: I had the idea of peeking next file(s) and returning
|
---|
| 907 | * NO_EOL_CHAR only if it is the *last* non-empty
|
---|
| 908 | * input file. But there is a case where this won't work:
|
---|
| 909 | * file1: "a woo\nb woo"
|
---|
| 910 | * file2: "c no\nd no"
|
---|
| 911 | * sed -ne 's/woo/bang/p' input1 input2 => "a bang\nb bang"
|
---|
| 912 | * (note: *no* newline after "b bang"!) */
|
---|
| 913 | }
|
---|
| 914 | /* Close this file and advance to next one */
|
---|
| 915 | fclose(fp);
|
---|
| 916 | G.current_input_file++;
|
---|
[821] | 917 | }
|
---|
[1765] | 918 | *gets_char = gc;
|
---|
[821] | 919 | return temp;
|
---|
| 920 | }
|
---|
| 921 |
|
---|
[1765] | 922 | /* Output line of text. */
|
---|
| 923 | /* Note:
|
---|
| 924 | * The tricks with NO_EOL_CHAR and last_puts_char are there to emulate gnu sed.
|
---|
| 925 | * Without them, we had this:
|
---|
| 926 | * echo -n thingy >z1
|
---|
| 927 | * echo -n again >z2
|
---|
| 928 | * >znull
|
---|
| 929 | * sed "s/i/z/" z1 z2 znull | hexdump -vC
|
---|
| 930 | * output:
|
---|
| 931 | * gnu sed 4.1.5:
|
---|
| 932 | * 00000000 74 68 7a 6e 67 79 0a 61 67 61 7a 6e |thzngy.agazn|
|
---|
| 933 | * bbox:
|
---|
| 934 | * 00000000 74 68 7a 6e 67 79 61 67 61 7a 6e |thzngyagazn|
|
---|
| 935 | */
|
---|
| 936 | static void puts_maybe_newline(char *s, FILE *file, char *last_puts_char, char last_gets_char)
|
---|
[821] | 937 | {
|
---|
[1765] | 938 | char lpc = *last_puts_char;
|
---|
[821] | 939 |
|
---|
[1765] | 940 | /* Need to insert a '\n' between two files because first file's
|
---|
| 941 | * last line wasn't terminated? */
|
---|
| 942 | if (lpc != '\n' && lpc != '\0') {
|
---|
| 943 | fputc('\n', file);
|
---|
| 944 | lpc = '\n';
|
---|
| 945 | }
|
---|
| 946 | fputs(s, file);
|
---|
| 947 |
|
---|
| 948 | /* 'x' - just something which is not '\n', '\0' or NO_EOL_CHAR */
|
---|
| 949 | if (s[0])
|
---|
| 950 | lpc = 'x';
|
---|
| 951 |
|
---|
| 952 | /* had trailing '\0' and it was last char of file? */
|
---|
| 953 | if (last_gets_char == LAST_IS_NUL) {
|
---|
| 954 | fputc('\0', file);
|
---|
| 955 | lpc = 'x'; /* */
|
---|
| 956 | } else
|
---|
| 957 | /* had trailing '\n' or '\0'? */
|
---|
| 958 | if (last_gets_char != NO_EOL_CHAR) {
|
---|
| 959 | fputc(last_gets_char, file);
|
---|
| 960 | lpc = last_gets_char;
|
---|
| 961 | }
|
---|
| 962 |
|
---|
| 963 | if (ferror(file)) {
|
---|
| 964 | xfunc_error_retval = 4; /* It's what gnu sed exits with... */
|
---|
[821] | 965 | bb_error_msg_and_die(bb_msg_write_error);
|
---|
| 966 | }
|
---|
[1765] | 967 | *last_puts_char = lpc;
|
---|
| 968 | }
|
---|
[821] | 969 |
|
---|
[1765] | 970 | #define sed_puts(s, n) (puts_maybe_newline(s, G.nonstdout, &last_puts_char, n))
|
---|
| 971 |
|
---|
| 972 | static int beg_match(sed_cmd_t *sed_cmd, const char *pattern_space)
|
---|
| 973 | {
|
---|
| 974 | int retval = sed_cmd->beg_match && !regexec(sed_cmd->beg_match, pattern_space, 0, NULL, 0);
|
---|
| 975 | if (retval)
|
---|
| 976 | G.previous_regex_ptr = sed_cmd->beg_match;
|
---|
| 977 | return retval;
|
---|
[821] | 978 | }
|
---|
| 979 |
|
---|
| 980 | /* Process all the lines in all the files */
|
---|
| 981 |
|
---|
| 982 | static void process_files(void)
|
---|
| 983 | {
|
---|
| 984 | char *pattern_space, *next_line;
|
---|
[1765] | 985 | int linenum = 0;
|
---|
| 986 | char last_puts_char = '\n';
|
---|
| 987 | char last_gets_char, next_gets_char;
|
---|
| 988 | sed_cmd_t *sed_cmd;
|
---|
| 989 | int substituted;
|
---|
[821] | 990 |
|
---|
| 991 | /* Prime the pump */
|
---|
[1765] | 992 | next_line = get_next_line(&next_gets_char);
|
---|
[821] | 993 |
|
---|
[2725] | 994 | /* Go through every line in each file */
|
---|
| 995 | again:
|
---|
[1765] | 996 | substituted = 0;
|
---|
[821] | 997 |
|
---|
[1765] | 998 | /* Advance to next line. Stop if out of lines. */
|
---|
| 999 | pattern_space = next_line;
|
---|
[2725] | 1000 | if (!pattern_space)
|
---|
| 1001 | return;
|
---|
[1765] | 1002 | last_gets_char = next_gets_char;
|
---|
[821] | 1003 |
|
---|
[1765] | 1004 | /* Read one line in advance so we can act on the last line,
|
---|
| 1005 | * the '$' address */
|
---|
| 1006 | next_line = get_next_line(&next_gets_char);
|
---|
| 1007 | linenum++;
|
---|
[2725] | 1008 |
|
---|
| 1009 | /* For every line, go through all the commands */
|
---|
| 1010 | restart:
|
---|
[3232] | 1011 | for (sed_cmd = G.sed_cmd_head; sed_cmd; sed_cmd = sed_cmd->next) {
|
---|
[1765] | 1012 | int old_matched, matched;
|
---|
[821] | 1013 |
|
---|
[1765] | 1014 | old_matched = sed_cmd->in_match;
|
---|
[821] | 1015 |
|
---|
[1765] | 1016 | /* Determine if this command matches this line: */
|
---|
[821] | 1017 |
|
---|
[3232] | 1018 | dbg("match1:%d", sed_cmd->in_match);
|
---|
| 1019 | dbg("match2:%d", (!sed_cmd->beg_line && !sed_cmd->end_line
|
---|
| 1020 | && !sed_cmd->beg_match && !sed_cmd->end_match));
|
---|
| 1021 | dbg("match3:%d", (sed_cmd->beg_line > 0
|
---|
| 1022 | && (sed_cmd->end_line || sed_cmd->end_match
|
---|
| 1023 | ? (sed_cmd->beg_line <= linenum)
|
---|
| 1024 | : (sed_cmd->beg_line == linenum)
|
---|
| 1025 | )
|
---|
| 1026 | ));
|
---|
| 1027 | dbg("match4:%d", (beg_match(sed_cmd, pattern_space)));
|
---|
| 1028 | dbg("match5:%d", (sed_cmd->beg_line == -1 && next_line == NULL));
|
---|
[2725] | 1029 |
|
---|
[1765] | 1030 | /* Are we continuing a previous multi-line match? */
|
---|
| 1031 | sed_cmd->in_match = sed_cmd->in_match
|
---|
[821] | 1032 | /* Or is no range necessary? */
|
---|
[1765] | 1033 | || (!sed_cmd->beg_line && !sed_cmd->end_line
|
---|
| 1034 | && !sed_cmd->beg_match && !sed_cmd->end_match)
|
---|
[821] | 1035 | /* Or did we match the start of a numerical range? */
|
---|
[2725] | 1036 | || (sed_cmd->beg_line > 0
|
---|
| 1037 | && (sed_cmd->end_line || sed_cmd->end_match
|
---|
| 1038 | /* note: even if end is numeric and is < linenum too,
|
---|
[3232] | 1039 | * GNU sed matches! We match too, therefore we don't
|
---|
| 1040 | * check here that linenum <= end.
|
---|
| 1041 | * Example:
|
---|
| 1042 | * printf '1\n2\n3\n4\n' | sed -n '1{N;N;d};1p;2,3p;3p;4p'
|
---|
| 1043 | * first three input lines are deleted;
|
---|
| 1044 | * 4th line is matched and printed
|
---|
| 1045 | * by "2,3" (!) and by "4" ranges
|
---|
| 1046 | */
|
---|
[2725] | 1047 | ? (sed_cmd->beg_line <= linenum) /* N,end */
|
---|
| 1048 | : (sed_cmd->beg_line == linenum) /* N */
|
---|
| 1049 | )
|
---|
| 1050 | )
|
---|
[821] | 1051 | /* Or does this line match our begin address regex? */
|
---|
[1765] | 1052 | || (beg_match(sed_cmd, pattern_space))
|
---|
[821] | 1053 | /* Or did we match last line of input? */
|
---|
[1765] | 1054 | || (sed_cmd->beg_line == -1 && next_line == NULL);
|
---|
[821] | 1055 |
|
---|
[1765] | 1056 | /* Snapshot the value */
|
---|
| 1057 | matched = sed_cmd->in_match;
|
---|
[821] | 1058 |
|
---|
[3232] | 1059 | dbg("cmd:'%c' matched:%d beg_line:%d end_line:%d linenum:%d",
|
---|
| 1060 | sed_cmd->cmd, matched, sed_cmd->beg_line, sed_cmd->end_line, linenum);
|
---|
[2725] | 1061 |
|
---|
[1765] | 1062 | /* Is this line the end of the current match? */
|
---|
[821] | 1063 |
|
---|
[1765] | 1064 | if (matched) {
|
---|
[2725] | 1065 | /* once matched, "n,xxx" range is dead, disabling it */
|
---|
[3232] | 1066 | if (sed_cmd->beg_line > 0) {
|
---|
[2725] | 1067 | sed_cmd->beg_line = -2;
|
---|
| 1068 | }
|
---|
[1765] | 1069 | sed_cmd->in_match = !(
|
---|
| 1070 | /* has the ending line come, or is this a single address command? */
|
---|
[3232] | 1071 | (sed_cmd->end_line
|
---|
| 1072 | ? sed_cmd->end_line == -1
|
---|
| 1073 | ? !next_line
|
---|
[1765] | 1074 | : (sed_cmd->end_line <= linenum)
|
---|
| 1075 | : !sed_cmd->end_match
|
---|
| 1076 | )
|
---|
| 1077 | /* or does this line matches our last address regex */
|
---|
| 1078 | || (sed_cmd->end_match && old_matched
|
---|
| 1079 | && (regexec(sed_cmd->end_match,
|
---|
[3232] | 1080 | pattern_space, 0, NULL, 0) == 0)
|
---|
| 1081 | )
|
---|
[1765] | 1082 | );
|
---|
| 1083 | }
|
---|
| 1084 |
|
---|
[2725] | 1085 | /* Skip blocks of commands we didn't match */
|
---|
[1765] | 1086 | if (sed_cmd->cmd == '{') {
|
---|
| 1087 | if (sed_cmd->invert ? matched : !matched) {
|
---|
[2725] | 1088 | unsigned nest_cnt = 0;
|
---|
| 1089 | while (1) {
|
---|
| 1090 | if (sed_cmd->cmd == '{')
|
---|
| 1091 | nest_cnt++;
|
---|
| 1092 | if (sed_cmd->cmd == '}') {
|
---|
| 1093 | nest_cnt--;
|
---|
| 1094 | if (nest_cnt == 0)
|
---|
| 1095 | break;
|
---|
| 1096 | }
|
---|
[1765] | 1097 | sed_cmd = sed_cmd->next;
|
---|
| 1098 | if (!sed_cmd)
|
---|
| 1099 | bb_error_msg_and_die("unterminated {");
|
---|
| 1100 | }
|
---|
[821] | 1101 | }
|
---|
[1765] | 1102 | continue;
|
---|
| 1103 | }
|
---|
[821] | 1104 |
|
---|
[1765] | 1105 | /* Okay, so did this line match? */
|
---|
[2725] | 1106 | if (sed_cmd->invert ? matched : !matched)
|
---|
| 1107 | continue; /* no */
|
---|
[821] | 1108 |
|
---|
[2725] | 1109 | /* Update last used regex in case a blank substitute BRE is found */
|
---|
| 1110 | if (sed_cmd->beg_match) {
|
---|
| 1111 | G.previous_regex_ptr = sed_cmd->beg_match;
|
---|
| 1112 | }
|
---|
[821] | 1113 |
|
---|
[2725] | 1114 | /* actual sedding */
|
---|
[3232] | 1115 | dbg("pattern_space:'%s' next_line:'%s' cmd:%c",
|
---|
| 1116 | pattern_space, next_line, sed_cmd->cmd);
|
---|
[2725] | 1117 | switch (sed_cmd->cmd) {
|
---|
[821] | 1118 |
|
---|
[2725] | 1119 | /* Print line number */
|
---|
| 1120 | case '=':
|
---|
| 1121 | fprintf(G.nonstdout, "%d\n", linenum);
|
---|
| 1122 | break;
|
---|
[821] | 1123 |
|
---|
[2725] | 1124 | /* Write the current pattern space up to the first newline */
|
---|
| 1125 | case 'P':
|
---|
| 1126 | {
|
---|
| 1127 | char *tmp = strchr(pattern_space, '\n');
|
---|
| 1128 | if (tmp) {
|
---|
| 1129 | *tmp = '\0';
|
---|
| 1130 | /* TODO: explain why '\n' below */
|
---|
[1765] | 1131 | sed_puts(pattern_space, '\n');
|
---|
[2725] | 1132 | *tmp = '\n';
|
---|
[1765] | 1133 | break;
|
---|
[2725] | 1134 | }
|
---|
| 1135 | /* Fall Through */
|
---|
| 1136 | }
|
---|
[821] | 1137 |
|
---|
[2725] | 1138 | /* Write the current pattern space to output */
|
---|
| 1139 | case 'p':
|
---|
| 1140 | /* NB: we print this _before_ the last line
|
---|
| 1141 | * (of current file) is printed. Even if
|
---|
| 1142 | * that line is nonterminated, we print
|
---|
| 1143 | * '\n' here (gnu sed does the same) */
|
---|
| 1144 | sed_puts(pattern_space, '\n');
|
---|
| 1145 | break;
|
---|
| 1146 | /* Delete up through first newline */
|
---|
| 1147 | case 'D':
|
---|
| 1148 | {
|
---|
| 1149 | char *tmp = strchr(pattern_space, '\n');
|
---|
| 1150 | if (tmp) {
|
---|
| 1151 | overlapping_strcpy(pattern_space, tmp + 1);
|
---|
| 1152 | goto restart;
|
---|
[1765] | 1153 | }
|
---|
[2725] | 1154 | }
|
---|
| 1155 | /* discard this line. */
|
---|
| 1156 | case 'd':
|
---|
| 1157 | goto discard_line;
|
---|
[821] | 1158 |
|
---|
[2725] | 1159 | /* Substitute with regex */
|
---|
| 1160 | case 's':
|
---|
| 1161 | if (!do_subst_command(sed_cmd, &pattern_space))
|
---|
[1765] | 1162 | break;
|
---|
[3232] | 1163 | dbg("do_subst_command succeeded:'%s'", pattern_space);
|
---|
[2725] | 1164 | substituted |= 1;
|
---|
[821] | 1165 |
|
---|
[2725] | 1166 | /* handle p option */
|
---|
| 1167 | if (sed_cmd->sub_p)
|
---|
| 1168 | sed_puts(pattern_space, last_gets_char);
|
---|
| 1169 | /* handle w option */
|
---|
| 1170 | if (sed_cmd->sw_file)
|
---|
| 1171 | puts_maybe_newline(
|
---|
| 1172 | pattern_space, sed_cmd->sw_file,
|
---|
| 1173 | &sed_cmd->sw_last_char, last_gets_char);
|
---|
| 1174 | break;
|
---|
[821] | 1175 |
|
---|
[2725] | 1176 | /* Append line to linked list to be printed later */
|
---|
| 1177 | case 'a':
|
---|
| 1178 | append(sed_cmd->string);
|
---|
| 1179 | break;
|
---|
[821] | 1180 |
|
---|
[2725] | 1181 | /* Insert text before this line */
|
---|
| 1182 | case 'i':
|
---|
| 1183 | sed_puts(sed_cmd->string, '\n');
|
---|
| 1184 | break;
|
---|
[821] | 1185 |
|
---|
[2725] | 1186 | /* Cut and paste text (replace) */
|
---|
| 1187 | case 'c':
|
---|
| 1188 | /* Only triggers on last line of a matching range. */
|
---|
| 1189 | if (!sed_cmd->in_match)
|
---|
| 1190 | sed_puts(sed_cmd->string, '\n');
|
---|
| 1191 | goto discard_line;
|
---|
[821] | 1192 |
|
---|
[2725] | 1193 | /* Read file, append contents to output */
|
---|
| 1194 | case 'r':
|
---|
| 1195 | {
|
---|
| 1196 | FILE *rfile;
|
---|
| 1197 | rfile = fopen_for_read(sed_cmd->string);
|
---|
| 1198 | if (rfile) {
|
---|
| 1199 | char *line;
|
---|
[821] | 1200 |
|
---|
[2725] | 1201 | while ((line = xmalloc_fgetline(rfile))
|
---|
| 1202 | != NULL)
|
---|
| 1203 | append(line);
|
---|
| 1204 | xprint_and_close_file(rfile);
|
---|
[1765] | 1205 | }
|
---|
[821] | 1206 |
|
---|
[2725] | 1207 | break;
|
---|
| 1208 | }
|
---|
[821] | 1209 |
|
---|
[2725] | 1210 | /* Write pattern space to file. */
|
---|
| 1211 | case 'w':
|
---|
| 1212 | puts_maybe_newline(
|
---|
| 1213 | pattern_space, sed_cmd->sw_file,
|
---|
| 1214 | &sed_cmd->sw_last_char, last_gets_char);
|
---|
| 1215 | break;
|
---|
[821] | 1216 |
|
---|
[2725] | 1217 | /* Read next line from input */
|
---|
| 1218 | case 'n':
|
---|
| 1219 | if (!G.be_quiet)
|
---|
| 1220 | sed_puts(pattern_space, last_gets_char);
|
---|
| 1221 | if (next_line) {
|
---|
| 1222 | free(pattern_space);
|
---|
| 1223 | pattern_space = next_line;
|
---|
[1765] | 1224 | last_gets_char = next_gets_char;
|
---|
| 1225 | next_line = get_next_line(&next_gets_char);
|
---|
[2725] | 1226 | substituted = 0;
|
---|
[1765] | 1227 | linenum++;
|
---|
| 1228 | break;
|
---|
| 1229 | }
|
---|
[2725] | 1230 | /* fall through */
|
---|
[821] | 1231 |
|
---|
[2725] | 1232 | /* Quit. End of script, end of input. */
|
---|
| 1233 | case 'q':
|
---|
| 1234 | /* Exit the outer while loop */
|
---|
| 1235 | free(next_line);
|
---|
| 1236 | next_line = NULL;
|
---|
| 1237 | goto discard_commands;
|
---|
[821] | 1238 |
|
---|
[2725] | 1239 | /* Append the next line to the current line */
|
---|
| 1240 | case 'N':
|
---|
| 1241 | {
|
---|
| 1242 | int len;
|
---|
| 1243 | /* If no next line, jump to end of script and exit. */
|
---|
| 1244 | /* http://www.gnu.org/software/sed/manual/sed.html:
|
---|
| 1245 | * "Most versions of sed exit without printing anything
|
---|
| 1246 | * when the N command is issued on the last line of
|
---|
| 1247 | * a file. GNU sed prints pattern space before exiting
|
---|
| 1248 | * unless of course the -n command switch has been
|
---|
| 1249 | * specified. This choice is by design."
|
---|
| 1250 | */
|
---|
| 1251 | if (next_line == NULL) {
|
---|
| 1252 | //goto discard_line;
|
---|
| 1253 | goto discard_commands; /* GNU behavior */
|
---|
| 1254 | }
|
---|
| 1255 | /* Append next_line, read new next_line. */
|
---|
| 1256 | len = strlen(pattern_space);
|
---|
| 1257 | pattern_space = xrealloc(pattern_space, len + strlen(next_line) + 2);
|
---|
| 1258 | pattern_space[len] = '\n';
|
---|
| 1259 | strcpy(pattern_space + len+1, next_line);
|
---|
| 1260 | last_gets_char = next_gets_char;
|
---|
| 1261 | next_line = get_next_line(&next_gets_char);
|
---|
| 1262 | linenum++;
|
---|
| 1263 | break;
|
---|
| 1264 | }
|
---|
| 1265 |
|
---|
| 1266 | /* Test/branch if substitution occurred */
|
---|
| 1267 | case 't':
|
---|
| 1268 | if (!substituted) break;
|
---|
| 1269 | substituted = 0;
|
---|
| 1270 | /* Fall through */
|
---|
| 1271 | /* Test/branch if substitution didn't occur */
|
---|
| 1272 | case 'T':
|
---|
| 1273 | if (substituted) break;
|
---|
| 1274 | /* Fall through */
|
---|
| 1275 | /* Branch to label */
|
---|
| 1276 | case 'b':
|
---|
| 1277 | if (!sed_cmd->string) goto discard_commands;
|
---|
| 1278 | else sed_cmd = branch_to(sed_cmd->string);
|
---|
| 1279 | break;
|
---|
| 1280 | /* Transliterate characters */
|
---|
| 1281 | case 'y':
|
---|
| 1282 | {
|
---|
| 1283 | int i, j;
|
---|
| 1284 | for (i = 0; pattern_space[i]; i++) {
|
---|
| 1285 | for (j = 0; sed_cmd->string[j]; j += 2) {
|
---|
| 1286 | if (pattern_space[i] == sed_cmd->string[j]) {
|
---|
| 1287 | pattern_space[i] = sed_cmd->string[j + 1];
|
---|
| 1288 | break;
|
---|
[821] | 1289 | }
|
---|
[1765] | 1290 | }
|
---|
| 1291 | }
|
---|
[821] | 1292 |
|
---|
[2725] | 1293 | break;
|
---|
| 1294 | }
|
---|
| 1295 | case 'g': /* Replace pattern space with hold space */
|
---|
| 1296 | free(pattern_space);
|
---|
| 1297 | pattern_space = xstrdup(G.hold_space ? G.hold_space : "");
|
---|
| 1298 | break;
|
---|
| 1299 | case 'G': /* Append newline and hold space to pattern space */
|
---|
| 1300 | {
|
---|
| 1301 | int pattern_space_size = 2;
|
---|
| 1302 | int hold_space_size = 0;
|
---|
[821] | 1303 |
|
---|
[2725] | 1304 | if (pattern_space)
|
---|
| 1305 | pattern_space_size += strlen(pattern_space);
|
---|
| 1306 | if (G.hold_space)
|
---|
| 1307 | hold_space_size = strlen(G.hold_space);
|
---|
| 1308 | pattern_space = xrealloc(pattern_space,
|
---|
| 1309 | pattern_space_size + hold_space_size);
|
---|
| 1310 | if (pattern_space_size == 2)
|
---|
| 1311 | pattern_space[0] = 0;
|
---|
| 1312 | strcat(pattern_space, "\n");
|
---|
| 1313 | if (G.hold_space)
|
---|
| 1314 | strcat(pattern_space, G.hold_space);
|
---|
| 1315 | last_gets_char = '\n';
|
---|
[821] | 1316 |
|
---|
[2725] | 1317 | break;
|
---|
| 1318 | }
|
---|
| 1319 | case 'h': /* Replace hold space with pattern space */
|
---|
| 1320 | free(G.hold_space);
|
---|
| 1321 | G.hold_space = xstrdup(pattern_space);
|
---|
| 1322 | break;
|
---|
| 1323 | case 'H': /* Append newline and pattern space to hold space */
|
---|
| 1324 | {
|
---|
| 1325 | int hold_space_size = 2;
|
---|
| 1326 | int pattern_space_size = 0;
|
---|
[821] | 1327 |
|
---|
[2725] | 1328 | if (G.hold_space)
|
---|
| 1329 | hold_space_size += strlen(G.hold_space);
|
---|
| 1330 | if (pattern_space)
|
---|
| 1331 | pattern_space_size = strlen(pattern_space);
|
---|
| 1332 | G.hold_space = xrealloc(G.hold_space,
|
---|
| 1333 | hold_space_size + pattern_space_size);
|
---|
[821] | 1334 |
|
---|
[2725] | 1335 | if (hold_space_size == 2)
|
---|
| 1336 | *G.hold_space = 0;
|
---|
| 1337 | strcat(G.hold_space, "\n");
|
---|
| 1338 | if (pattern_space)
|
---|
| 1339 | strcat(G.hold_space, pattern_space);
|
---|
| 1340 |
|
---|
| 1341 | break;
|
---|
[821] | 1342 | }
|
---|
[2725] | 1343 | case 'x': /* Exchange hold and pattern space */
|
---|
| 1344 | {
|
---|
| 1345 | char *tmp = pattern_space;
|
---|
| 1346 | pattern_space = G.hold_space ? G.hold_space : xzalloc(1);
|
---|
| 1347 | last_gets_char = '\n';
|
---|
| 1348 | G.hold_space = tmp;
|
---|
| 1349 | break;
|
---|
| 1350 | }
|
---|
| 1351 | } /* switch */
|
---|
| 1352 | } /* for each cmd */
|
---|
[821] | 1353 |
|
---|
[1765] | 1354 | /*
|
---|
[2725] | 1355 | * Exit point from sedding...
|
---|
[1765] | 1356 | */
|
---|
| 1357 | discard_commands:
|
---|
| 1358 | /* we will print the line unless we were told to be quiet ('-n')
|
---|
| 1359 | or if the line was suppressed (ala 'd'elete) */
|
---|
| 1360 | if (!G.be_quiet)
|
---|
| 1361 | sed_puts(pattern_space, last_gets_char);
|
---|
[821] | 1362 |
|
---|
[1765] | 1363 | /* Delete and such jump here. */
|
---|
| 1364 | discard_line:
|
---|
| 1365 | flush_append();
|
---|
| 1366 | free(pattern_space);
|
---|
| 1367 |
|
---|
| 1368 | goto again;
|
---|
[821] | 1369 | }
|
---|
| 1370 |
|
---|
| 1371 | /* It is possible to have a command line argument with embedded
|
---|
[1765] | 1372 | * newlines. This counts as multiple command lines.
|
---|
| 1373 | * However, newline can be escaped: 's/e/z\<newline>z/'
|
---|
| 1374 | * We check for this.
|
---|
| 1375 | */
|
---|
[821] | 1376 |
|
---|
| 1377 | static void add_cmd_block(char *cmdstr)
|
---|
| 1378 | {
|
---|
[1765] | 1379 | char *sv, *eol;
|
---|
[821] | 1380 |
|
---|
[1765] | 1381 | cmdstr = sv = xstrdup(cmdstr);
|
---|
| 1382 | do {
|
---|
| 1383 | eol = strchr(cmdstr, '\n');
|
---|
| 1384 | next:
|
---|
| 1385 | if (eol) {
|
---|
| 1386 | /* Count preceding slashes */
|
---|
| 1387 | int slashes = 0;
|
---|
| 1388 | char *sl = eol;
|
---|
| 1389 |
|
---|
| 1390 | while (sl != cmdstr && *--sl == '\\')
|
---|
| 1391 | slashes++;
|
---|
| 1392 | /* Odd number of preceding slashes - newline is escaped */
|
---|
| 1393 | if (slashes & 1) {
|
---|
[2725] | 1394 | overlapping_strcpy(eol - 1, eol);
|
---|
[1765] | 1395 | eol = strchr(eol, '\n');
|
---|
| 1396 | goto next;
|
---|
| 1397 | }
|
---|
| 1398 | *eol = '\0';
|
---|
| 1399 | }
|
---|
| 1400 | add_cmd(cmdstr);
|
---|
| 1401 | cmdstr = eol + 1;
|
---|
| 1402 | } while (eol);
|
---|
| 1403 | free(sv);
|
---|
[821] | 1404 | }
|
---|
| 1405 |
|
---|
[2725] | 1406 | int sed_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
|
---|
| 1407 | int sed_main(int argc UNUSED_PARAM, char **argv)
|
---|
[821] | 1408 | {
|
---|
[1765] | 1409 | unsigned opt;
|
---|
| 1410 | llist_t *opt_e, *opt_f;
|
---|
[3232] | 1411 | char *opt_i;
|
---|
| 1412 |
|
---|
| 1413 | #if ENABLE_LONG_OPTS
|
---|
| 1414 | static const char sed_longopts[] ALIGN1 =
|
---|
| 1415 | /* name has_arg short */
|
---|
| 1416 | "in-place\0" Optional_argument "i"
|
---|
| 1417 | "regexp-extended\0" No_argument "r"
|
---|
| 1418 | "quiet\0" No_argument "n"
|
---|
| 1419 | "silent\0" No_argument "n"
|
---|
| 1420 | "expression\0" Required_argument "e"
|
---|
| 1421 | "file\0" Required_argument "f";
|
---|
| 1422 | #endif
|
---|
| 1423 |
|
---|
[1765] | 1424 | int status = EXIT_SUCCESS;
|
---|
[821] | 1425 |
|
---|
[1765] | 1426 | INIT_G();
|
---|
[821] | 1427 |
|
---|
| 1428 | /* destroy command strings on exit */
|
---|
| 1429 | if (ENABLE_FEATURE_CLEAN_UP) atexit(sed_free_and_close_stuff);
|
---|
| 1430 |
|
---|
| 1431 | /* Lie to autoconf when it starts asking stupid questions. */
|
---|
[3232] | 1432 | if (argv[1] && strcmp(argv[1], "--version") == 0) {
|
---|
[1765] | 1433 | puts("This is not GNU sed version 4.0");
|
---|
| 1434 | return 0;
|
---|
[821] | 1435 | }
|
---|
| 1436 |
|
---|
| 1437 | /* do normal option parsing */
|
---|
[1765] | 1438 | opt_e = opt_f = NULL;
|
---|
[3232] | 1439 | opt_i = NULL;
|
---|
[1765] | 1440 | opt_complementary = "e::f::" /* can occur multiple times */
|
---|
| 1441 | "nn"; /* count -n */
|
---|
[3232] | 1442 |
|
---|
| 1443 | IF_LONG_OPTS(applet_long_options = sed_longopts);
|
---|
| 1444 |
|
---|
[2725] | 1445 | /* -i must be first, to match OPT_in_place definition */
|
---|
[3232] | 1446 | opt = getopt32(argv, "i::rne:f:", &opt_i, &opt_e, &opt_f,
|
---|
[1765] | 1447 | &G.be_quiet); /* counter for -n */
|
---|
[2725] | 1448 | //argc -= optind;
|
---|
[1765] | 1449 | argv += optind;
|
---|
| 1450 | if (opt & OPT_in_place) { // -i
|
---|
| 1451 | atexit(cleanup_outname);
|
---|
| 1452 | }
|
---|
| 1453 | if (opt & 0x2) G.regex_type |= REG_EXTENDED; // -r
|
---|
| 1454 | //if (opt & 0x4) G.be_quiet++; // -n
|
---|
| 1455 | while (opt_e) { // -e
|
---|
[2725] | 1456 | add_cmd_block(llist_pop(&opt_e));
|
---|
[1765] | 1457 | }
|
---|
| 1458 | while (opt_f) { // -f
|
---|
| 1459 | char *line;
|
---|
| 1460 | FILE *cmdfile;
|
---|
[2725] | 1461 | cmdfile = xfopen_for_read(llist_pop(&opt_f));
|
---|
| 1462 | while ((line = xmalloc_fgetline(cmdfile)) != NULL) {
|
---|
[1765] | 1463 | add_cmd(line);
|
---|
| 1464 | free(line);
|
---|
[821] | 1465 | }
|
---|
[1765] | 1466 | fclose(cmdfile);
|
---|
[821] | 1467 | }
|
---|
[1765] | 1468 | /* if we didn't get a pattern from -e or -f, use argv[0] */
|
---|
| 1469 | if (!(opt & 0x18)) {
|
---|
[2725] | 1470 | if (!*argv)
|
---|
[821] | 1471 | bb_show_usage();
|
---|
[1765] | 1472 | add_cmd_block(*argv++);
|
---|
[821] | 1473 | }
|
---|
| 1474 | /* Flush any unfinished commands. */
|
---|
| 1475 | add_cmd("");
|
---|
| 1476 |
|
---|
| 1477 | /* By default, we write to stdout */
|
---|
[1765] | 1478 | G.nonstdout = stdout;
|
---|
[821] | 1479 |
|
---|
[1765] | 1480 | /* argv[0..(argc-1)] should be names of file to process. If no
|
---|
[821] | 1481 | * files were specified or '-' was specified, take input from stdin.
|
---|
| 1482 | * Otherwise, we process all the files specified. */
|
---|
[1765] | 1483 | if (argv[0] == NULL) {
|
---|
| 1484 | if (opt & OPT_in_place)
|
---|
| 1485 | bb_error_msg_and_die(bb_msg_requires_arg, "-i");
|
---|
[821] | 1486 | add_input_file(stdin);
|
---|
| 1487 | } else {
|
---|
| 1488 | int i;
|
---|
| 1489 |
|
---|
[2725] | 1490 | for (i = 0; argv[i]; i++) {
|
---|
[1765] | 1491 | struct stat statbuf;
|
---|
| 1492 | int nonstdoutfd;
|
---|
[3232] | 1493 | FILE *file;
|
---|
| 1494 | sed_cmd_t *sed_cmd;
|
---|
[1765] | 1495 |
|
---|
| 1496 | if (LONE_DASH(argv[i]) && !(opt & OPT_in_place)) {
|
---|
[821] | 1497 | add_input_file(stdin);
|
---|
| 1498 | process_files();
|
---|
[1765] | 1499 | continue;
|
---|
| 1500 | }
|
---|
| 1501 | file = fopen_or_warn(argv[i], "r");
|
---|
| 1502 | if (!file) {
|
---|
| 1503 | status = EXIT_FAILURE;
|
---|
| 1504 | continue;
|
---|
| 1505 | }
|
---|
[3232] | 1506 | add_input_file(file);
|
---|
[1765] | 1507 | if (!(opt & OPT_in_place)) {
|
---|
| 1508 | continue;
|
---|
| 1509 | }
|
---|
[821] | 1510 |
|
---|
[3232] | 1511 | /* -i: process each FILE separately: */
|
---|
| 1512 |
|
---|
[1765] | 1513 | G.outname = xasprintf("%sXXXXXX", argv[i]);
|
---|
[2725] | 1514 | nonstdoutfd = xmkstemp(G.outname);
|
---|
| 1515 | G.nonstdout = xfdopen_for_write(nonstdoutfd);
|
---|
[821] | 1516 |
|
---|
[2725] | 1517 | /* Set permissions/owner of output file */
|
---|
[1765] | 1518 | fstat(fileno(file), &statbuf);
|
---|
[2725] | 1519 | /* chmod'ing AFTER chown would preserve suid/sgid bits,
|
---|
| 1520 | * but GNU sed 4.2.1 does not preserve them either */
|
---|
[1765] | 1521 | fchmod(nonstdoutfd, statbuf.st_mode);
|
---|
[2725] | 1522 | fchown(nonstdoutfd, statbuf.st_uid, statbuf.st_gid);
|
---|
[3232] | 1523 |
|
---|
[1765] | 1524 | process_files();
|
---|
| 1525 | fclose(G.nonstdout);
|
---|
[3232] | 1526 | G.nonstdout = stdout;
|
---|
[821] | 1527 |
|
---|
[3232] | 1528 | if (opt_i) {
|
---|
| 1529 | char *backupname = xasprintf("%s%s", argv[i], opt_i);
|
---|
| 1530 | xrename(argv[i], backupname);
|
---|
| 1531 | free(backupname);
|
---|
| 1532 | }
|
---|
| 1533 | /* else unlink(argv[i]); - rename below does this */
|
---|
| 1534 | xrename(G.outname, argv[i]); //TODO: rollback backup on error?
|
---|
[1765] | 1535 | free(G.outname);
|
---|
[2725] | 1536 | G.outname = NULL;
|
---|
[3232] | 1537 |
|
---|
| 1538 | /* Re-enable disabled range matches */
|
---|
| 1539 | for (sed_cmd = G.sed_cmd_head; sed_cmd; sed_cmd = sed_cmd->next) {
|
---|
| 1540 | sed_cmd->beg_line = sed_cmd->beg_line_orig;
|
---|
| 1541 | }
|
---|
[821] | 1542 | }
|
---|
[2725] | 1543 | /* Here, to handle "sed 'cmds' nonexistent_file" case we did:
|
---|
| 1544 | * if (G.current_input_file >= G.input_file_count)
|
---|
| 1545 | * return status;
|
---|
| 1546 | * but it's not needed since process_files() works correctly
|
---|
| 1547 | * in this case too. */
|
---|
[821] | 1548 | }
|
---|
[2725] | 1549 | process_files();
|
---|
[821] | 1550 |
|
---|
| 1551 | return status;
|
---|
| 1552 | }
|
---|