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
|
---|
8 | * Copyright (C) 2003 by Glenn McGrath
|
---|
9 | * Copyright (C) 2003,2004 by Rob Landley <rob@landley.net>
|
---|
10 | *
|
---|
11 | * MAINTAINER: Rob Landley <rob@landley.net>
|
---|
12 | *
|
---|
13 | * Licensed under GPLv2, see file LICENSE in this source tree.
|
---|
14 | */
|
---|
15 |
|
---|
16 | /* Code overview.
|
---|
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 | */
|
---|
35 |
|
---|
36 |
|
---|
37 | /*
|
---|
38 | Supported features and commands in this version of sed:
|
---|
39 |
|
---|
40 | - comments ('#')
|
---|
41 | - address matching: num|/matchstr/[,num|/matchstr/|$]command
|
---|
42 | - commands: (p)rint, (d)elete, (s)ubstitue (with g & I flags)
|
---|
43 | - edit commands: (a)ppend, (i)nsert, (c)hange
|
---|
44 | - file commands: (r)ead
|
---|
45 | - backreferences in substitution expressions (\0, \1, \2...\9)
|
---|
46 | - grouped commands: {cmd1;cmd2}
|
---|
47 | - transliteration (y/source-chars/dest-chars/)
|
---|
48 | - pattern space hold space storing / swapping (g, h, x)
|
---|
49 | - labels / branching (: label, b, t, T)
|
---|
50 |
|
---|
51 | (Note: Specifying an address (range) to match is *optional*; commands
|
---|
52 | default to the whole pattern space if no specific address match was
|
---|
53 | requested.)
|
---|
54 |
|
---|
55 | Todo:
|
---|
56 | - Create a wrapper around regex to make libc's regex conform with sed
|
---|
57 |
|
---|
58 | Reference http://www.opengroup.org/onlinepubs/007904975/utilities/sed.html
|
---|
59 | */
|
---|
60 |
|
---|
61 | #include "libbb.h"
|
---|
62 | #include "xregex.h"
|
---|
63 |
|
---|
64 | enum {
|
---|
65 | OPT_in_place = 1 << 0,
|
---|
66 | };
|
---|
67 |
|
---|
68 | /* Each sed command turns into one of these structures. */
|
---|
69 | typedef struct sed_cmd_s {
|
---|
70 | /* Ordered by alignment requirements: currently 36 bytes on x86 */
|
---|
71 | struct sed_cmd_s *next; /* Next command (linked list, NULL terminated) */
|
---|
72 |
|
---|
73 | /* address storage */
|
---|
74 | regex_t *beg_match; /* sed -e '/match/cmd' */
|
---|
75 | regex_t *end_match; /* sed -e '/match/,/end_match/cmd' */
|
---|
76 | regex_t *sub_match; /* For 's/sub_match/string/' */
|
---|
77 | int beg_line; /* 'sed 1p' 0 == apply commands to all lines */
|
---|
78 | int end_line; /* 'sed 1,3p' 0 == one line only. -1 = last line ($) */
|
---|
79 |
|
---|
80 | FILE *sw_file; /* File (sw) command writes to, -1 for none. */
|
---|
81 | char *string; /* Data string for (saicytb) commands. */
|
---|
82 |
|
---|
83 | unsigned which_match; /* (s) Which match to replace (0 for all) */
|
---|
84 |
|
---|
85 | /* Bitfields (gcc won't group them if we don't) */
|
---|
86 | unsigned invert:1; /* the '!' after the address */
|
---|
87 | unsigned in_match:1; /* Next line also included in match? */
|
---|
88 | unsigned sub_p:1; /* (s) print option */
|
---|
89 |
|
---|
90 | char sw_last_char; /* Last line written by (sw) had no '\n' */
|
---|
91 |
|
---|
92 | /* GENERAL FIELDS */
|
---|
93 | char cmd; /* The command char: abcdDgGhHilnNpPqrstwxy:={} */
|
---|
94 | } sed_cmd_t;
|
---|
95 |
|
---|
96 | static const char semicolon_whitespace[] ALIGN1 = "; \n\r\t\v";
|
---|
97 |
|
---|
98 | struct globals {
|
---|
99 | /* options */
|
---|
100 | int be_quiet, regex_type;
|
---|
101 | FILE *nonstdout;
|
---|
102 | char *outname, *hold_space;
|
---|
103 |
|
---|
104 | /* List of input files */
|
---|
105 | int input_file_count, current_input_file;
|
---|
106 | FILE **input_file_list;
|
---|
107 |
|
---|
108 | regmatch_t regmatch[10];
|
---|
109 | regex_t *previous_regex_ptr;
|
---|
110 |
|
---|
111 | /* linked list of sed commands */
|
---|
112 | sed_cmd_t sed_cmd_head, *sed_cmd_tail;
|
---|
113 |
|
---|
114 | /* Linked list of append lines */
|
---|
115 | llist_t *append_head;
|
---|
116 |
|
---|
117 | char *add_cmd_line;
|
---|
118 |
|
---|
119 | struct pipeline {
|
---|
120 | char *buf; /* Space to hold string */
|
---|
121 | int idx; /* Space used */
|
---|
122 | int len; /* Space allocated */
|
---|
123 | } pipeline;
|
---|
124 | } FIX_ALIASING;
|
---|
125 | #define G (*(struct globals*)&bb_common_bufsiz1)
|
---|
126 | struct BUG_G_too_big {
|
---|
127 | char BUG_G_too_big[sizeof(G) <= COMMON_BUFSIZE ? 1 : -1];
|
---|
128 | };
|
---|
129 | #define INIT_G() do { \
|
---|
130 | G.sed_cmd_tail = &G.sed_cmd_head; \
|
---|
131 | } while (0)
|
---|
132 |
|
---|
133 |
|
---|
134 | #if ENABLE_FEATURE_CLEAN_UP
|
---|
135 | static void sed_free_and_close_stuff(void)
|
---|
136 | {
|
---|
137 | sed_cmd_t *sed_cmd = G.sed_cmd_head.next;
|
---|
138 |
|
---|
139 | llist_free(G.append_head, free);
|
---|
140 |
|
---|
141 | while (sed_cmd) {
|
---|
142 | sed_cmd_t *sed_cmd_next = sed_cmd->next;
|
---|
143 |
|
---|
144 | if (sed_cmd->sw_file)
|
---|
145 | xprint_and_close_file(sed_cmd->sw_file);
|
---|
146 |
|
---|
147 | if (sed_cmd->beg_match) {
|
---|
148 | regfree(sed_cmd->beg_match);
|
---|
149 | free(sed_cmd->beg_match);
|
---|
150 | }
|
---|
151 | if (sed_cmd->end_match) {
|
---|
152 | regfree(sed_cmd->end_match);
|
---|
153 | free(sed_cmd->end_match);
|
---|
154 | }
|
---|
155 | if (sed_cmd->sub_match) {
|
---|
156 | regfree(sed_cmd->sub_match);
|
---|
157 | free(sed_cmd->sub_match);
|
---|
158 | }
|
---|
159 | free(sed_cmd->string);
|
---|
160 | free(sed_cmd);
|
---|
161 | sed_cmd = sed_cmd_next;
|
---|
162 | }
|
---|
163 |
|
---|
164 | free(G.hold_space);
|
---|
165 |
|
---|
166 | while (G.current_input_file < G.input_file_count)
|
---|
167 | fclose(G.input_file_list[G.current_input_file++]);
|
---|
168 | }
|
---|
169 | #else
|
---|
170 | void sed_free_and_close_stuff(void);
|
---|
171 | #endif
|
---|
172 |
|
---|
173 | /* If something bad happens during -i operation, delete temp file */
|
---|
174 |
|
---|
175 | static void cleanup_outname(void)
|
---|
176 | {
|
---|
177 | if (G.outname) unlink(G.outname);
|
---|
178 | }
|
---|
179 |
|
---|
180 | /* strcpy, replacing "\from" with 'to'. If to is NUL, replacing "\any" with 'any' */
|
---|
181 |
|
---|
182 | static void parse_escapes(char *dest, const char *string, int len, char from, char to)
|
---|
183 | {
|
---|
184 | int i = 0;
|
---|
185 |
|
---|
186 | while (i < len) {
|
---|
187 | if (string[i] == '\\') {
|
---|
188 | if (!to || string[i+1] == from) {
|
---|
189 | *dest++ = to ? to : string[i+1];
|
---|
190 | i += 2;
|
---|
191 | continue;
|
---|
192 | }
|
---|
193 | *dest++ = string[i++];
|
---|
194 | }
|
---|
195 | /* TODO: is it safe wrt a string with trailing '\\' ? */
|
---|
196 | *dest++ = string[i++];
|
---|
197 | }
|
---|
198 | *dest = '\0';
|
---|
199 | }
|
---|
200 |
|
---|
201 | static char *copy_parsing_escapes(const char *string, int len)
|
---|
202 | {
|
---|
203 | char *dest = xmalloc(len + 1);
|
---|
204 |
|
---|
205 | parse_escapes(dest, string, len, 'n', '\n');
|
---|
206 | /* GNU sed also recognizes \t */
|
---|
207 | parse_escapes(dest, dest, strlen(dest), 't', '\t');
|
---|
208 | return dest;
|
---|
209 | }
|
---|
210 |
|
---|
211 |
|
---|
212 | /*
|
---|
213 | * index_of_next_unescaped_regexp_delim - walks left to right through a string
|
---|
214 | * beginning at a specified index and returns the index of the next regular
|
---|
215 | * expression delimiter (typically a forward slash ('/')) not preceded by
|
---|
216 | * a backslash ('\'). A negative delimiter disables square bracket checking.
|
---|
217 | */
|
---|
218 | static int index_of_next_unescaped_regexp_delim(int delimiter, const char *str)
|
---|
219 | {
|
---|
220 | int bracket = -1;
|
---|
221 | int escaped = 0;
|
---|
222 | int idx = 0;
|
---|
223 | char ch;
|
---|
224 |
|
---|
225 | if (delimiter < 0) {
|
---|
226 | bracket--;
|
---|
227 | delimiter = -delimiter;
|
---|
228 | }
|
---|
229 |
|
---|
230 | for (; (ch = str[idx]); idx++) {
|
---|
231 | if (bracket >= 0) {
|
---|
232 | if (ch == ']' && !(bracket == idx - 1 || (bracket == idx - 2
|
---|
233 | && str[idx - 1] == '^')))
|
---|
234 | bracket = -1;
|
---|
235 | } else if (escaped)
|
---|
236 | escaped = 0;
|
---|
237 | else if (ch == '\\')
|
---|
238 | escaped = 1;
|
---|
239 | else if (bracket == -1 && ch == '[')
|
---|
240 | bracket = idx;
|
---|
241 | else if (ch == delimiter)
|
---|
242 | return idx;
|
---|
243 | }
|
---|
244 |
|
---|
245 | /* if we make it to here, we've hit the end of the string */
|
---|
246 | bb_error_msg_and_die("unmatched '%c'", delimiter);
|
---|
247 | }
|
---|
248 |
|
---|
249 | /*
|
---|
250 | * Returns the index of the third delimiter
|
---|
251 | */
|
---|
252 | static int parse_regex_delim(const char *cmdstr, char **match, char **replace)
|
---|
253 | {
|
---|
254 | const char *cmdstr_ptr = cmdstr;
|
---|
255 | char delimiter;
|
---|
256 | int idx = 0;
|
---|
257 |
|
---|
258 | /* verify that the 's' or 'y' is followed by something. That something
|
---|
259 | * (typically a 'slash') is now our regexp delimiter... */
|
---|
260 | if (*cmdstr == '\0')
|
---|
261 | bb_error_msg_and_die("bad format in substitution expression");
|
---|
262 | delimiter = *cmdstr_ptr++;
|
---|
263 |
|
---|
264 | /* save the match string */
|
---|
265 | idx = index_of_next_unescaped_regexp_delim(delimiter, cmdstr_ptr);
|
---|
266 | *match = copy_parsing_escapes(cmdstr_ptr, idx);
|
---|
267 |
|
---|
268 | /* save the replacement string */
|
---|
269 | cmdstr_ptr += idx + 1;
|
---|
270 | idx = index_of_next_unescaped_regexp_delim(-delimiter, cmdstr_ptr);
|
---|
271 | *replace = copy_parsing_escapes(cmdstr_ptr, idx);
|
---|
272 |
|
---|
273 | return ((cmdstr_ptr - cmdstr) + idx);
|
---|
274 | }
|
---|
275 |
|
---|
276 | /*
|
---|
277 | * returns the index in the string just past where the address ends.
|
---|
278 | */
|
---|
279 | static int get_address(const char *my_str, int *linenum, regex_t ** regex)
|
---|
280 | {
|
---|
281 | const char *pos = my_str;
|
---|
282 |
|
---|
283 | if (isdigit(*my_str)) {
|
---|
284 | *linenum = strtol(my_str, (char**)&pos, 10);
|
---|
285 | /* endstr shouldnt ever equal NULL */
|
---|
286 | } else if (*my_str == '$') {
|
---|
287 | *linenum = -1;
|
---|
288 | pos++;
|
---|
289 | } else if (*my_str == '/' || *my_str == '\\') {
|
---|
290 | int next;
|
---|
291 | char delimiter;
|
---|
292 | char *temp;
|
---|
293 |
|
---|
294 | delimiter = '/';
|
---|
295 | if (*my_str == '\\') delimiter = *++pos;
|
---|
296 | next = index_of_next_unescaped_regexp_delim(delimiter, ++pos);
|
---|
297 | temp = copy_parsing_escapes(pos, next);
|
---|
298 | *regex = xmalloc(sizeof(regex_t));
|
---|
299 | xregcomp(*regex, temp, G.regex_type|REG_NEWLINE);
|
---|
300 | free(temp);
|
---|
301 | /* Move position to next character after last delimiter */
|
---|
302 | pos += (next+1);
|
---|
303 | }
|
---|
304 | return pos - my_str;
|
---|
305 | }
|
---|
306 |
|
---|
307 | /* Grab a filename. Whitespace at start is skipped, then goes to EOL. */
|
---|
308 | static int parse_file_cmd(/*sed_cmd_t *sed_cmd,*/ const char *filecmdstr, char **retval)
|
---|
309 | {
|
---|
310 | int start = 0, idx, hack = 0;
|
---|
311 |
|
---|
312 | /* Skip whitespace, then grab filename to end of line */
|
---|
313 | while (isspace(filecmdstr[start]))
|
---|
314 | start++;
|
---|
315 | idx = start;
|
---|
316 | while (filecmdstr[idx] && filecmdstr[idx] != '\n')
|
---|
317 | idx++;
|
---|
318 |
|
---|
319 | /* If lines glued together, put backslash back. */
|
---|
320 | if (filecmdstr[idx] == '\n')
|
---|
321 | hack = 1;
|
---|
322 | if (idx == start)
|
---|
323 | bb_error_msg_and_die("empty filename");
|
---|
324 | *retval = xstrndup(filecmdstr+start, idx-start+hack+1);
|
---|
325 | if (hack)
|
---|
326 | (*retval)[idx] = '\\';
|
---|
327 |
|
---|
328 | return idx;
|
---|
329 | }
|
---|
330 |
|
---|
331 | static int parse_subst_cmd(sed_cmd_t *sed_cmd, const char *substr)
|
---|
332 | {
|
---|
333 | int cflags = G.regex_type;
|
---|
334 | char *match;
|
---|
335 | int idx;
|
---|
336 |
|
---|
337 | /*
|
---|
338 | * A substitution command should look something like this:
|
---|
339 | * s/match/replace/ #gIpw
|
---|
340 | * || | |||
|
---|
341 | * mandatory optional
|
---|
342 | */
|
---|
343 | idx = parse_regex_delim(substr, &match, &sed_cmd->string);
|
---|
344 |
|
---|
345 | /* determine the number of back references in the match string */
|
---|
346 | /* Note: we compute this here rather than in the do_subst_command()
|
---|
347 | * function to save processor time, at the expense of a little more memory
|
---|
348 | * (4 bits) per sed_cmd */
|
---|
349 |
|
---|
350 | /* process the flags */
|
---|
351 |
|
---|
352 | sed_cmd->which_match = 1;
|
---|
353 | while (substr[++idx]) {
|
---|
354 | /* Parse match number */
|
---|
355 | if (isdigit(substr[idx])) {
|
---|
356 | if (match[0] != '^') {
|
---|
357 | /* Match 0 treated as all, multiple matches we take the last one. */
|
---|
358 | const char *pos = substr + idx;
|
---|
359 | /* FIXME: error check? */
|
---|
360 | sed_cmd->which_match = (unsigned)strtol(substr+idx, (char**) &pos, 10);
|
---|
361 | idx = pos - substr;
|
---|
362 | }
|
---|
363 | continue;
|
---|
364 | }
|
---|
365 | /* Skip spaces */
|
---|
366 | if (isspace(substr[idx]))
|
---|
367 | continue;
|
---|
368 |
|
---|
369 | switch (substr[idx]) {
|
---|
370 | /* Replace all occurrences */
|
---|
371 | case 'g':
|
---|
372 | if (match[0] != '^')
|
---|
373 | sed_cmd->which_match = 0;
|
---|
374 | break;
|
---|
375 | /* Print pattern space */
|
---|
376 | case 'p':
|
---|
377 | sed_cmd->sub_p = 1;
|
---|
378 | break;
|
---|
379 | /* Write to file */
|
---|
380 | case 'w':
|
---|
381 | {
|
---|
382 | char *temp;
|
---|
383 | idx += parse_file_cmd(/*sed_cmd,*/ substr+idx, &temp);
|
---|
384 | break;
|
---|
385 | }
|
---|
386 | /* Ignore case (gnu exension) */
|
---|
387 | case 'I':
|
---|
388 | cflags |= REG_ICASE;
|
---|
389 | break;
|
---|
390 | /* Comment */
|
---|
391 | case '#':
|
---|
392 | // while (substr[++idx]) continue;
|
---|
393 | idx += strlen(substr + idx); // same
|
---|
394 | /* Fall through */
|
---|
395 | /* End of command */
|
---|
396 | case ';':
|
---|
397 | case '}':
|
---|
398 | goto out;
|
---|
399 | default:
|
---|
400 | bb_error_msg_and_die("bad option in substitution expression");
|
---|
401 | }
|
---|
402 | }
|
---|
403 | out:
|
---|
404 | /* compile the match string into a regex */
|
---|
405 | if (*match != '\0') {
|
---|
406 | /* If match is empty, we use last regex used at runtime */
|
---|
407 | sed_cmd->sub_match = xmalloc(sizeof(regex_t));
|
---|
408 | xregcomp(sed_cmd->sub_match, match, cflags);
|
---|
409 | }
|
---|
410 | free(match);
|
---|
411 |
|
---|
412 | return idx;
|
---|
413 | }
|
---|
414 |
|
---|
415 | /*
|
---|
416 | * Process the commands arguments
|
---|
417 | */
|
---|
418 | static const char *parse_cmd_args(sed_cmd_t *sed_cmd, const char *cmdstr)
|
---|
419 | {
|
---|
420 | /* handle (s)ubstitution command */
|
---|
421 | if (sed_cmd->cmd == 's')
|
---|
422 | cmdstr += parse_subst_cmd(sed_cmd, cmdstr);
|
---|
423 | /* handle edit cmds: (a)ppend, (i)nsert, and (c)hange */
|
---|
424 | else if (strchr("aic", sed_cmd->cmd)) {
|
---|
425 | if ((sed_cmd->end_line || sed_cmd->end_match) && sed_cmd->cmd != 'c')
|
---|
426 | bb_error_msg_and_die("only a beginning address can be specified for edit commands");
|
---|
427 | for (;;) {
|
---|
428 | if (*cmdstr == '\n' || *cmdstr == '\\') {
|
---|
429 | cmdstr++;
|
---|
430 | break;
|
---|
431 | }
|
---|
432 | if (!isspace(*cmdstr))
|
---|
433 | break;
|
---|
434 | cmdstr++;
|
---|
435 | }
|
---|
436 | sed_cmd->string = xstrdup(cmdstr);
|
---|
437 | /* "\anychar" -> "anychar" */
|
---|
438 | parse_escapes(sed_cmd->string, sed_cmd->string, strlen(cmdstr), '\0', '\0');
|
---|
439 | cmdstr += strlen(cmdstr);
|
---|
440 | /* handle file cmds: (r)ead */
|
---|
441 | } else if (strchr("rw", sed_cmd->cmd)) {
|
---|
442 | if (sed_cmd->end_line || sed_cmd->end_match)
|
---|
443 | bb_error_msg_and_die("command only uses one address");
|
---|
444 | cmdstr += parse_file_cmd(/*sed_cmd,*/ cmdstr, &sed_cmd->string);
|
---|
445 | if (sed_cmd->cmd == 'w') {
|
---|
446 | sed_cmd->sw_file = xfopen_for_write(sed_cmd->string);
|
---|
447 | sed_cmd->sw_last_char = '\n';
|
---|
448 | }
|
---|
449 | /* handle branch commands */
|
---|
450 | } else if (strchr(":btT", sed_cmd->cmd)) {
|
---|
451 | int length;
|
---|
452 |
|
---|
453 | cmdstr = skip_whitespace(cmdstr);
|
---|
454 | length = strcspn(cmdstr, semicolon_whitespace);
|
---|
455 | if (length) {
|
---|
456 | sed_cmd->string = xstrndup(cmdstr, length);
|
---|
457 | cmdstr += length;
|
---|
458 | }
|
---|
459 | }
|
---|
460 | /* translation command */
|
---|
461 | else if (sed_cmd->cmd == 'y') {
|
---|
462 | char *match, *replace;
|
---|
463 | int i = cmdstr[0];
|
---|
464 |
|
---|
465 | cmdstr += parse_regex_delim(cmdstr, &match, &replace)+1;
|
---|
466 | /* \n already parsed, but \delimiter needs unescaping. */
|
---|
467 | parse_escapes(match, match, strlen(match), i, i);
|
---|
468 | parse_escapes(replace, replace, strlen(replace), i, i);
|
---|
469 |
|
---|
470 | sed_cmd->string = xzalloc((strlen(match) + 1) * 2);
|
---|
471 | for (i = 0; match[i] && replace[i]; i++) {
|
---|
472 | sed_cmd->string[i*2] = match[i];
|
---|
473 | sed_cmd->string[i*2+1] = replace[i];
|
---|
474 | }
|
---|
475 | free(match);
|
---|
476 | free(replace);
|
---|
477 | }
|
---|
478 | /* if it wasnt a single-letter command that takes no arguments
|
---|
479 | * then it must be an invalid command.
|
---|
480 | */
|
---|
481 | else if (strchr("dDgGhHlnNpPqx={}", sed_cmd->cmd) == 0) {
|
---|
482 | bb_error_msg_and_die("unsupported command %c", sed_cmd->cmd);
|
---|
483 | }
|
---|
484 |
|
---|
485 | /* give back whatever's left over */
|
---|
486 | return cmdstr;
|
---|
487 | }
|
---|
488 |
|
---|
489 |
|
---|
490 | /* Parse address+command sets, skipping comment lines. */
|
---|
491 |
|
---|
492 | static void add_cmd(const char *cmdstr)
|
---|
493 | {
|
---|
494 | sed_cmd_t *sed_cmd;
|
---|
495 | unsigned len, n;
|
---|
496 |
|
---|
497 | /* Append this line to any unfinished line from last time. */
|
---|
498 | if (G.add_cmd_line) {
|
---|
499 | char *tp = xasprintf("%s\n%s", G.add_cmd_line, cmdstr);
|
---|
500 | free(G.add_cmd_line);
|
---|
501 | cmdstr = G.add_cmd_line = tp;
|
---|
502 | }
|
---|
503 |
|
---|
504 | /* If this line ends with unescaped backslash, request next line. */
|
---|
505 | n = len = strlen(cmdstr);
|
---|
506 | while (n && cmdstr[n-1] == '\\')
|
---|
507 | n--;
|
---|
508 | if ((len - n) & 1) { /* if odd number of trailing backslashes */
|
---|
509 | if (!G.add_cmd_line)
|
---|
510 | G.add_cmd_line = xstrdup(cmdstr);
|
---|
511 | G.add_cmd_line[len-1] = '\0';
|
---|
512 | return;
|
---|
513 | }
|
---|
514 |
|
---|
515 | /* Loop parsing all commands in this line. */
|
---|
516 | while (*cmdstr) {
|
---|
517 | /* Skip leading whitespace and semicolons */
|
---|
518 | cmdstr += strspn(cmdstr, semicolon_whitespace);
|
---|
519 |
|
---|
520 | /* If no more commands, exit. */
|
---|
521 | if (!*cmdstr) break;
|
---|
522 |
|
---|
523 | /* if this is a comment, jump past it and keep going */
|
---|
524 | if (*cmdstr == '#') {
|
---|
525 | /* "#n" is the same as using -n on the command line */
|
---|
526 | if (cmdstr[1] == 'n')
|
---|
527 | G.be_quiet++;
|
---|
528 | cmdstr = strpbrk(cmdstr, "\n\r");
|
---|
529 | if (!cmdstr) break;
|
---|
530 | continue;
|
---|
531 | }
|
---|
532 |
|
---|
533 | /* parse the command
|
---|
534 | * format is: [addr][,addr][!]cmd
|
---|
535 | * |----||-----||-|
|
---|
536 | * part1 part2 part3
|
---|
537 | */
|
---|
538 |
|
---|
539 | sed_cmd = xzalloc(sizeof(sed_cmd_t));
|
---|
540 |
|
---|
541 | /* first part (if present) is an address: either a '$', a number or a /regex/ */
|
---|
542 | cmdstr += get_address(cmdstr, &sed_cmd->beg_line, &sed_cmd->beg_match);
|
---|
543 |
|
---|
544 | /* second part (if present) will begin with a comma */
|
---|
545 | if (*cmdstr == ',') {
|
---|
546 | int idx;
|
---|
547 |
|
---|
548 | cmdstr++;
|
---|
549 | idx = get_address(cmdstr, &sed_cmd->end_line, &sed_cmd->end_match);
|
---|
550 | if (!idx)
|
---|
551 | bb_error_msg_and_die("no address after comma");
|
---|
552 | cmdstr += idx;
|
---|
553 | }
|
---|
554 |
|
---|
555 | /* skip whitespace before the command */
|
---|
556 | cmdstr = skip_whitespace(cmdstr);
|
---|
557 |
|
---|
558 | /* Check for inversion flag */
|
---|
559 | if (*cmdstr == '!') {
|
---|
560 | sed_cmd->invert = 1;
|
---|
561 | cmdstr++;
|
---|
562 |
|
---|
563 | /* skip whitespace before the command */
|
---|
564 | cmdstr = skip_whitespace(cmdstr);
|
---|
565 | }
|
---|
566 |
|
---|
567 | /* last part (mandatory) will be a command */
|
---|
568 | if (!*cmdstr)
|
---|
569 | bb_error_msg_and_die("missing command");
|
---|
570 | sed_cmd->cmd = *cmdstr++;
|
---|
571 | cmdstr = parse_cmd_args(sed_cmd, cmdstr);
|
---|
572 |
|
---|
573 | /* Add the command to the command array */
|
---|
574 | G.sed_cmd_tail->next = sed_cmd;
|
---|
575 | G.sed_cmd_tail = G.sed_cmd_tail->next;
|
---|
576 | }
|
---|
577 |
|
---|
578 | /* If we glued multiple lines together, free the memory. */
|
---|
579 | free(G.add_cmd_line);
|
---|
580 | G.add_cmd_line = NULL;
|
---|
581 | }
|
---|
582 |
|
---|
583 | /* Append to a string, reallocating memory as necessary. */
|
---|
584 |
|
---|
585 | #define PIPE_GROW 64
|
---|
586 |
|
---|
587 | static void pipe_putc(char c)
|
---|
588 | {
|
---|
589 | if (G.pipeline.idx == G.pipeline.len) {
|
---|
590 | G.pipeline.buf = xrealloc(G.pipeline.buf,
|
---|
591 | G.pipeline.len + PIPE_GROW);
|
---|
592 | G.pipeline.len += PIPE_GROW;
|
---|
593 | }
|
---|
594 | G.pipeline.buf[G.pipeline.idx++] = c;
|
---|
595 | }
|
---|
596 |
|
---|
597 | static void do_subst_w_backrefs(char *line, char *replace)
|
---|
598 | {
|
---|
599 | int i, j;
|
---|
600 |
|
---|
601 | /* go through the replacement string */
|
---|
602 | for (i = 0; replace[i]; i++) {
|
---|
603 | /* if we find a backreference (\1, \2, etc.) print the backref'ed * text */
|
---|
604 | if (replace[i] == '\\') {
|
---|
605 | unsigned backref = replace[++i] - '0';
|
---|
606 | if (backref <= 9) {
|
---|
607 | /* print out the text held in G.regmatch[backref] */
|
---|
608 | if (G.regmatch[backref].rm_so != -1) {
|
---|
609 | j = G.regmatch[backref].rm_so;
|
---|
610 | while (j < G.regmatch[backref].rm_eo)
|
---|
611 | pipe_putc(line[j++]);
|
---|
612 | }
|
---|
613 | continue;
|
---|
614 | }
|
---|
615 | /* I _think_ it is impossible to get '\' to be
|
---|
616 | * the last char in replace string. Thus we dont check
|
---|
617 | * for replace[i] == NUL. (counterexample anyone?) */
|
---|
618 | /* if we find a backslash escaped character, print the character */
|
---|
619 | pipe_putc(replace[i]);
|
---|
620 | continue;
|
---|
621 | }
|
---|
622 | /* if we find an unescaped '&' print out the whole matched text. */
|
---|
623 | if (replace[i] == '&') {
|
---|
624 | j = G.regmatch[0].rm_so;
|
---|
625 | while (j < G.regmatch[0].rm_eo)
|
---|
626 | pipe_putc(line[j++]);
|
---|
627 | continue;
|
---|
628 | }
|
---|
629 | /* Otherwise just output the character. */
|
---|
630 | pipe_putc(replace[i]);
|
---|
631 | }
|
---|
632 | }
|
---|
633 |
|
---|
634 | static int do_subst_command(sed_cmd_t *sed_cmd, char **line_p)
|
---|
635 | {
|
---|
636 | char *line = *line_p;
|
---|
637 | int altered = 0;
|
---|
638 | unsigned match_count = 0;
|
---|
639 | regex_t *current_regex;
|
---|
640 |
|
---|
641 | current_regex = sed_cmd->sub_match;
|
---|
642 | /* Handle empty regex. */
|
---|
643 | if (!current_regex) {
|
---|
644 | current_regex = G.previous_regex_ptr;
|
---|
645 | if (!current_regex)
|
---|
646 | bb_error_msg_and_die("no previous regexp");
|
---|
647 | }
|
---|
648 | G.previous_regex_ptr = current_regex;
|
---|
649 |
|
---|
650 | /* Find the first match */
|
---|
651 | if (REG_NOMATCH == regexec(current_regex, line, 10, G.regmatch, 0))
|
---|
652 | return 0;
|
---|
653 |
|
---|
654 | /* Initialize temporary output buffer. */
|
---|
655 | G.pipeline.buf = xmalloc(PIPE_GROW);
|
---|
656 | G.pipeline.len = PIPE_GROW;
|
---|
657 | G.pipeline.idx = 0;
|
---|
658 |
|
---|
659 | /* Now loop through, substituting for matches */
|
---|
660 | do {
|
---|
661 | int i;
|
---|
662 |
|
---|
663 | /* Work around bug in glibc regexec, demonstrated by:
|
---|
664 | echo " a.b" | busybox sed 's [^ .]* x g'
|
---|
665 | The match_count check is so not to break
|
---|
666 | echo "hi" | busybox sed 's/^/!/g' */
|
---|
667 | if (!G.regmatch[0].rm_so && !G.regmatch[0].rm_eo && match_count) {
|
---|
668 | pipe_putc(*line++);
|
---|
669 | continue;
|
---|
670 | }
|
---|
671 |
|
---|
672 | match_count++;
|
---|
673 |
|
---|
674 | /* If we aren't interested in this match, output old line to
|
---|
675 | end of match and continue */
|
---|
676 | if (sed_cmd->which_match
|
---|
677 | && (sed_cmd->which_match != match_count)
|
---|
678 | ) {
|
---|
679 | for (i = 0; i < G.regmatch[0].rm_eo; i++)
|
---|
680 | pipe_putc(*line++);
|
---|
681 | continue;
|
---|
682 | }
|
---|
683 |
|
---|
684 | /* print everything before the match */
|
---|
685 | for (i = 0; i < G.regmatch[0].rm_so; i++)
|
---|
686 | pipe_putc(line[i]);
|
---|
687 |
|
---|
688 | /* then print the substitution string */
|
---|
689 | do_subst_w_backrefs(line, sed_cmd->string);
|
---|
690 |
|
---|
691 | /* advance past the match */
|
---|
692 | line += G.regmatch[0].rm_eo;
|
---|
693 | /* flag that something has changed */
|
---|
694 | altered++;
|
---|
695 |
|
---|
696 | /* if we're not doing this globally, get out now */
|
---|
697 | if (sed_cmd->which_match)
|
---|
698 | break;
|
---|
699 |
|
---|
700 | //maybe (G.regmatch[0].rm_eo ? REG_NOTBOL : 0) instead of unconditional REG_NOTBOL?
|
---|
701 | } while (*line && regexec(current_regex, line, 10, G.regmatch, REG_NOTBOL) != REG_NOMATCH);
|
---|
702 |
|
---|
703 | /* Copy rest of string into output pipeline */
|
---|
704 | while (1) {
|
---|
705 | char c = *line++;
|
---|
706 | pipe_putc(c);
|
---|
707 | if (c == '\0')
|
---|
708 | break;
|
---|
709 | }
|
---|
710 |
|
---|
711 | free(*line_p);
|
---|
712 | *line_p = G.pipeline.buf;
|
---|
713 | return altered;
|
---|
714 | }
|
---|
715 |
|
---|
716 | /* Set command pointer to point to this label. (Does not handle null label.) */
|
---|
717 | static sed_cmd_t *branch_to(char *label)
|
---|
718 | {
|
---|
719 | sed_cmd_t *sed_cmd;
|
---|
720 |
|
---|
721 | for (sed_cmd = G.sed_cmd_head.next; sed_cmd; sed_cmd = sed_cmd->next) {
|
---|
722 | if (sed_cmd->cmd == ':' && sed_cmd->string && !strcmp(sed_cmd->string, label)) {
|
---|
723 | return sed_cmd;
|
---|
724 | }
|
---|
725 | }
|
---|
726 | bb_error_msg_and_die("can't find label for jump to '%s'", label);
|
---|
727 | }
|
---|
728 |
|
---|
729 | static void append(char *s)
|
---|
730 | {
|
---|
731 | llist_add_to_end(&G.append_head, xstrdup(s));
|
---|
732 | }
|
---|
733 |
|
---|
734 | static void flush_append(void)
|
---|
735 | {
|
---|
736 | char *data;
|
---|
737 |
|
---|
738 | /* Output appended lines. */
|
---|
739 | while ((data = (char *)llist_pop(&G.append_head))) {
|
---|
740 | fprintf(G.nonstdout, "%s\n", data);
|
---|
741 | free(data);
|
---|
742 | }
|
---|
743 | }
|
---|
744 |
|
---|
745 | static void add_input_file(FILE *file)
|
---|
746 | {
|
---|
747 | G.input_file_list = xrealloc_vector(G.input_file_list, 2, G.input_file_count);
|
---|
748 | G.input_file_list[G.input_file_count++] = file;
|
---|
749 | }
|
---|
750 |
|
---|
751 | /* Get next line of input from G.input_file_list, flushing append buffer and
|
---|
752 | * noting if we ran out of files without a newline on the last line we read.
|
---|
753 | */
|
---|
754 | enum {
|
---|
755 | NO_EOL_CHAR = 1,
|
---|
756 | LAST_IS_NUL = 2,
|
---|
757 | };
|
---|
758 | static char *get_next_line(char *gets_char)
|
---|
759 | {
|
---|
760 | char *temp = NULL;
|
---|
761 | int len;
|
---|
762 | char gc;
|
---|
763 |
|
---|
764 | flush_append();
|
---|
765 |
|
---|
766 | /* will be returned if last line in the file
|
---|
767 | * doesn't end with either '\n' or '\0' */
|
---|
768 | gc = NO_EOL_CHAR;
|
---|
769 | while (G.current_input_file < G.input_file_count) {
|
---|
770 | FILE *fp = G.input_file_list[G.current_input_file];
|
---|
771 | /* Read line up to a newline or NUL byte, inclusive,
|
---|
772 | * return malloc'ed char[]. length of the chunk read
|
---|
773 | * is stored in len. NULL if EOF/error */
|
---|
774 | temp = bb_get_chunk_from_file(fp, &len);
|
---|
775 | if (temp) {
|
---|
776 | /* len > 0 here, it's ok to do temp[len-1] */
|
---|
777 | char c = temp[len-1];
|
---|
778 | if (c == '\n' || c == '\0') {
|
---|
779 | temp[len-1] = '\0';
|
---|
780 | gc = c;
|
---|
781 | if (c == '\0') {
|
---|
782 | int ch = fgetc(fp);
|
---|
783 | if (ch != EOF)
|
---|
784 | ungetc(ch, fp);
|
---|
785 | else
|
---|
786 | gc = LAST_IS_NUL;
|
---|
787 | }
|
---|
788 | }
|
---|
789 | /* else we put NO_EOL_CHAR into *gets_char */
|
---|
790 | break;
|
---|
791 |
|
---|
792 | /* NB: I had the idea of peeking next file(s) and returning
|
---|
793 | * NO_EOL_CHAR only if it is the *last* non-empty
|
---|
794 | * input file. But there is a case where this won't work:
|
---|
795 | * file1: "a woo\nb woo"
|
---|
796 | * file2: "c no\nd no"
|
---|
797 | * sed -ne 's/woo/bang/p' input1 input2 => "a bang\nb bang"
|
---|
798 | * (note: *no* newline after "b bang"!) */
|
---|
799 | }
|
---|
800 | /* Close this file and advance to next one */
|
---|
801 | fclose(fp);
|
---|
802 | G.current_input_file++;
|
---|
803 | }
|
---|
804 | *gets_char = gc;
|
---|
805 | return temp;
|
---|
806 | }
|
---|
807 |
|
---|
808 | /* Output line of text. */
|
---|
809 | /* Note:
|
---|
810 | * The tricks with NO_EOL_CHAR and last_puts_char are there to emulate gnu sed.
|
---|
811 | * Without them, we had this:
|
---|
812 | * echo -n thingy >z1
|
---|
813 | * echo -n again >z2
|
---|
814 | * >znull
|
---|
815 | * sed "s/i/z/" z1 z2 znull | hexdump -vC
|
---|
816 | * output:
|
---|
817 | * gnu sed 4.1.5:
|
---|
818 | * 00000000 74 68 7a 6e 67 79 0a 61 67 61 7a 6e |thzngy.agazn|
|
---|
819 | * bbox:
|
---|
820 | * 00000000 74 68 7a 6e 67 79 61 67 61 7a 6e |thzngyagazn|
|
---|
821 | */
|
---|
822 | static void puts_maybe_newline(char *s, FILE *file, char *last_puts_char, char last_gets_char)
|
---|
823 | {
|
---|
824 | char lpc = *last_puts_char;
|
---|
825 |
|
---|
826 | /* Need to insert a '\n' between two files because first file's
|
---|
827 | * last line wasn't terminated? */
|
---|
828 | if (lpc != '\n' && lpc != '\0') {
|
---|
829 | fputc('\n', file);
|
---|
830 | lpc = '\n';
|
---|
831 | }
|
---|
832 | fputs(s, file);
|
---|
833 |
|
---|
834 | /* 'x' - just something which is not '\n', '\0' or NO_EOL_CHAR */
|
---|
835 | if (s[0])
|
---|
836 | lpc = 'x';
|
---|
837 |
|
---|
838 | /* had trailing '\0' and it was last char of file? */
|
---|
839 | if (last_gets_char == LAST_IS_NUL) {
|
---|
840 | fputc('\0', file);
|
---|
841 | lpc = 'x'; /* */
|
---|
842 | } else
|
---|
843 | /* had trailing '\n' or '\0'? */
|
---|
844 | if (last_gets_char != NO_EOL_CHAR) {
|
---|
845 | fputc(last_gets_char, file);
|
---|
846 | lpc = last_gets_char;
|
---|
847 | }
|
---|
848 |
|
---|
849 | if (ferror(file)) {
|
---|
850 | xfunc_error_retval = 4; /* It's what gnu sed exits with... */
|
---|
851 | bb_error_msg_and_die(bb_msg_write_error);
|
---|
852 | }
|
---|
853 | *last_puts_char = lpc;
|
---|
854 | }
|
---|
855 |
|
---|
856 | #define sed_puts(s, n) (puts_maybe_newline(s, G.nonstdout, &last_puts_char, n))
|
---|
857 |
|
---|
858 | static int beg_match(sed_cmd_t *sed_cmd, const char *pattern_space)
|
---|
859 | {
|
---|
860 | int retval = sed_cmd->beg_match && !regexec(sed_cmd->beg_match, pattern_space, 0, NULL, 0);
|
---|
861 | if (retval)
|
---|
862 | G.previous_regex_ptr = sed_cmd->beg_match;
|
---|
863 | return retval;
|
---|
864 | }
|
---|
865 |
|
---|
866 | /* Process all the lines in all the files */
|
---|
867 |
|
---|
868 | static void process_files(void)
|
---|
869 | {
|
---|
870 | char *pattern_space, *next_line;
|
---|
871 | int linenum = 0;
|
---|
872 | char last_puts_char = '\n';
|
---|
873 | char last_gets_char, next_gets_char;
|
---|
874 | sed_cmd_t *sed_cmd;
|
---|
875 | int substituted;
|
---|
876 |
|
---|
877 | /* Prime the pump */
|
---|
878 | next_line = get_next_line(&next_gets_char);
|
---|
879 |
|
---|
880 | /* Go through every line in each file */
|
---|
881 | again:
|
---|
882 | substituted = 0;
|
---|
883 |
|
---|
884 | /* Advance to next line. Stop if out of lines. */
|
---|
885 | pattern_space = next_line;
|
---|
886 | if (!pattern_space)
|
---|
887 | return;
|
---|
888 | last_gets_char = next_gets_char;
|
---|
889 |
|
---|
890 | /* Read one line in advance so we can act on the last line,
|
---|
891 | * the '$' address */
|
---|
892 | next_line = get_next_line(&next_gets_char);
|
---|
893 | linenum++;
|
---|
894 |
|
---|
895 | /* For every line, go through all the commands */
|
---|
896 | restart:
|
---|
897 | for (sed_cmd = G.sed_cmd_head.next; sed_cmd; sed_cmd = sed_cmd->next) {
|
---|
898 | int old_matched, matched;
|
---|
899 |
|
---|
900 | old_matched = sed_cmd->in_match;
|
---|
901 |
|
---|
902 | /* Determine if this command matches this line: */
|
---|
903 |
|
---|
904 | //bb_error_msg("match1:%d", sed_cmd->in_match);
|
---|
905 | //bb_error_msg("match2:%d", (!sed_cmd->beg_line && !sed_cmd->end_line
|
---|
906 | // && !sed_cmd->beg_match && !sed_cmd->end_match));
|
---|
907 | //bb_error_msg("match3:%d", (sed_cmd->beg_line > 0
|
---|
908 | // && (sed_cmd->end_line || sed_cmd->end_match
|
---|
909 | // ? (sed_cmd->beg_line <= linenum)
|
---|
910 | // : (sed_cmd->beg_line == linenum)
|
---|
911 | // )
|
---|
912 | // )
|
---|
913 | //bb_error_msg("match4:%d", (beg_match(sed_cmd, pattern_space)));
|
---|
914 | //bb_error_msg("match5:%d", (sed_cmd->beg_line == -1 && next_line == NULL));
|
---|
915 |
|
---|
916 | /* Are we continuing a previous multi-line match? */
|
---|
917 | sed_cmd->in_match = sed_cmd->in_match
|
---|
918 | /* Or is no range necessary? */
|
---|
919 | || (!sed_cmd->beg_line && !sed_cmd->end_line
|
---|
920 | && !sed_cmd->beg_match && !sed_cmd->end_match)
|
---|
921 | /* Or did we match the start of a numerical range? */
|
---|
922 | || (sed_cmd->beg_line > 0
|
---|
923 | && (sed_cmd->end_line || sed_cmd->end_match
|
---|
924 | /* note: even if end is numeric and is < linenum too,
|
---|
925 | * GNU sed matches! We match too */
|
---|
926 | ? (sed_cmd->beg_line <= linenum) /* N,end */
|
---|
927 | : (sed_cmd->beg_line == linenum) /* N */
|
---|
928 | )
|
---|
929 | )
|
---|
930 | /* Or does this line match our begin address regex? */
|
---|
931 | || (beg_match(sed_cmd, pattern_space))
|
---|
932 | /* Or did we match last line of input? */
|
---|
933 | || (sed_cmd->beg_line == -1 && next_line == NULL);
|
---|
934 |
|
---|
935 | /* Snapshot the value */
|
---|
936 | matched = sed_cmd->in_match;
|
---|
937 |
|
---|
938 | //bb_error_msg("cmd:'%c' matched:%d beg_line:%d end_line:%d linenum:%d",
|
---|
939 | //sed_cmd->cmd, matched, sed_cmd->beg_line, sed_cmd->end_line, linenum);
|
---|
940 |
|
---|
941 | /* Is this line the end of the current match? */
|
---|
942 |
|
---|
943 | if (matched) {
|
---|
944 | /* once matched, "n,xxx" range is dead, disabling it */
|
---|
945 | if (sed_cmd->beg_line > 0
|
---|
946 | && !(option_mask32 & OPT_in_place) /* but not for -i */
|
---|
947 | ) {
|
---|
948 | sed_cmd->beg_line = -2;
|
---|
949 | }
|
---|
950 | sed_cmd->in_match = !(
|
---|
951 | /* has the ending line come, or is this a single address command? */
|
---|
952 | (sed_cmd->end_line ?
|
---|
953 | sed_cmd->end_line == -1 ?
|
---|
954 | !next_line
|
---|
955 | : (sed_cmd->end_line <= linenum)
|
---|
956 | : !sed_cmd->end_match
|
---|
957 | )
|
---|
958 | /* or does this line matches our last address regex */
|
---|
959 | || (sed_cmd->end_match && old_matched
|
---|
960 | && (regexec(sed_cmd->end_match,
|
---|
961 | pattern_space, 0, NULL, 0) == 0))
|
---|
962 | );
|
---|
963 | }
|
---|
964 |
|
---|
965 | /* Skip blocks of commands we didn't match */
|
---|
966 | if (sed_cmd->cmd == '{') {
|
---|
967 | if (sed_cmd->invert ? matched : !matched) {
|
---|
968 | unsigned nest_cnt = 0;
|
---|
969 | while (1) {
|
---|
970 | if (sed_cmd->cmd == '{')
|
---|
971 | nest_cnt++;
|
---|
972 | if (sed_cmd->cmd == '}') {
|
---|
973 | nest_cnt--;
|
---|
974 | if (nest_cnt == 0)
|
---|
975 | break;
|
---|
976 | }
|
---|
977 | sed_cmd = sed_cmd->next;
|
---|
978 | if (!sed_cmd)
|
---|
979 | bb_error_msg_and_die("unterminated {");
|
---|
980 | }
|
---|
981 | }
|
---|
982 | continue;
|
---|
983 | }
|
---|
984 |
|
---|
985 | /* Okay, so did this line match? */
|
---|
986 | if (sed_cmd->invert ? matched : !matched)
|
---|
987 | continue; /* no */
|
---|
988 |
|
---|
989 | /* Update last used regex in case a blank substitute BRE is found */
|
---|
990 | if (sed_cmd->beg_match) {
|
---|
991 | G.previous_regex_ptr = sed_cmd->beg_match;
|
---|
992 | }
|
---|
993 |
|
---|
994 | /* actual sedding */
|
---|
995 | //bb_error_msg("pattern_space:'%s' next_line:'%s' cmd:%c",
|
---|
996 | //pattern_space, next_line, sed_cmd->cmd);
|
---|
997 | switch (sed_cmd->cmd) {
|
---|
998 |
|
---|
999 | /* Print line number */
|
---|
1000 | case '=':
|
---|
1001 | fprintf(G.nonstdout, "%d\n", linenum);
|
---|
1002 | break;
|
---|
1003 |
|
---|
1004 | /* Write the current pattern space up to the first newline */
|
---|
1005 | case 'P':
|
---|
1006 | {
|
---|
1007 | char *tmp = strchr(pattern_space, '\n');
|
---|
1008 | if (tmp) {
|
---|
1009 | *tmp = '\0';
|
---|
1010 | /* TODO: explain why '\n' below */
|
---|
1011 | sed_puts(pattern_space, '\n');
|
---|
1012 | *tmp = '\n';
|
---|
1013 | break;
|
---|
1014 | }
|
---|
1015 | /* Fall Through */
|
---|
1016 | }
|
---|
1017 |
|
---|
1018 | /* Write the current pattern space to output */
|
---|
1019 | case 'p':
|
---|
1020 | /* NB: we print this _before_ the last line
|
---|
1021 | * (of current file) is printed. Even if
|
---|
1022 | * that line is nonterminated, we print
|
---|
1023 | * '\n' here (gnu sed does the same) */
|
---|
1024 | sed_puts(pattern_space, '\n');
|
---|
1025 | break;
|
---|
1026 | /* Delete up through first newline */
|
---|
1027 | case 'D':
|
---|
1028 | {
|
---|
1029 | char *tmp = strchr(pattern_space, '\n');
|
---|
1030 | if (tmp) {
|
---|
1031 | overlapping_strcpy(pattern_space, tmp + 1);
|
---|
1032 | goto restart;
|
---|
1033 | }
|
---|
1034 | }
|
---|
1035 | /* discard this line. */
|
---|
1036 | case 'd':
|
---|
1037 | goto discard_line;
|
---|
1038 |
|
---|
1039 | /* Substitute with regex */
|
---|
1040 | case 's':
|
---|
1041 | if (!do_subst_command(sed_cmd, &pattern_space))
|
---|
1042 | break;
|
---|
1043 | substituted |= 1;
|
---|
1044 |
|
---|
1045 | /* handle p option */
|
---|
1046 | if (sed_cmd->sub_p)
|
---|
1047 | sed_puts(pattern_space, last_gets_char);
|
---|
1048 | /* handle w option */
|
---|
1049 | if (sed_cmd->sw_file)
|
---|
1050 | puts_maybe_newline(
|
---|
1051 | pattern_space, sed_cmd->sw_file,
|
---|
1052 | &sed_cmd->sw_last_char, last_gets_char);
|
---|
1053 | break;
|
---|
1054 |
|
---|
1055 | /* Append line to linked list to be printed later */
|
---|
1056 | case 'a':
|
---|
1057 | append(sed_cmd->string);
|
---|
1058 | break;
|
---|
1059 |
|
---|
1060 | /* Insert text before this line */
|
---|
1061 | case 'i':
|
---|
1062 | sed_puts(sed_cmd->string, '\n');
|
---|
1063 | break;
|
---|
1064 |
|
---|
1065 | /* Cut and paste text (replace) */
|
---|
1066 | case 'c':
|
---|
1067 | /* Only triggers on last line of a matching range. */
|
---|
1068 | if (!sed_cmd->in_match)
|
---|
1069 | sed_puts(sed_cmd->string, '\n');
|
---|
1070 | goto discard_line;
|
---|
1071 |
|
---|
1072 | /* Read file, append contents to output */
|
---|
1073 | case 'r':
|
---|
1074 | {
|
---|
1075 | FILE *rfile;
|
---|
1076 | rfile = fopen_for_read(sed_cmd->string);
|
---|
1077 | if (rfile) {
|
---|
1078 | char *line;
|
---|
1079 |
|
---|
1080 | while ((line = xmalloc_fgetline(rfile))
|
---|
1081 | != NULL)
|
---|
1082 | append(line);
|
---|
1083 | xprint_and_close_file(rfile);
|
---|
1084 | }
|
---|
1085 |
|
---|
1086 | break;
|
---|
1087 | }
|
---|
1088 |
|
---|
1089 | /* Write pattern space to file. */
|
---|
1090 | case 'w':
|
---|
1091 | puts_maybe_newline(
|
---|
1092 | pattern_space, sed_cmd->sw_file,
|
---|
1093 | &sed_cmd->sw_last_char, last_gets_char);
|
---|
1094 | break;
|
---|
1095 |
|
---|
1096 | /* Read next line from input */
|
---|
1097 | case 'n':
|
---|
1098 | if (!G.be_quiet)
|
---|
1099 | sed_puts(pattern_space, last_gets_char);
|
---|
1100 | if (next_line) {
|
---|
1101 | free(pattern_space);
|
---|
1102 | pattern_space = next_line;
|
---|
1103 | last_gets_char = next_gets_char;
|
---|
1104 | next_line = get_next_line(&next_gets_char);
|
---|
1105 | substituted = 0;
|
---|
1106 | linenum++;
|
---|
1107 | break;
|
---|
1108 | }
|
---|
1109 | /* fall through */
|
---|
1110 |
|
---|
1111 | /* Quit. End of script, end of input. */
|
---|
1112 | case 'q':
|
---|
1113 | /* Exit the outer while loop */
|
---|
1114 | free(next_line);
|
---|
1115 | next_line = NULL;
|
---|
1116 | goto discard_commands;
|
---|
1117 |
|
---|
1118 | /* Append the next line to the current line */
|
---|
1119 | case 'N':
|
---|
1120 | {
|
---|
1121 | int len;
|
---|
1122 | /* If no next line, jump to end of script and exit. */
|
---|
1123 | /* http://www.gnu.org/software/sed/manual/sed.html:
|
---|
1124 | * "Most versions of sed exit without printing anything
|
---|
1125 | * when the N command is issued on the last line of
|
---|
1126 | * a file. GNU sed prints pattern space before exiting
|
---|
1127 | * unless of course the -n command switch has been
|
---|
1128 | * specified. This choice is by design."
|
---|
1129 | */
|
---|
1130 | if (next_line == NULL) {
|
---|
1131 | //goto discard_line;
|
---|
1132 | goto discard_commands; /* GNU behavior */
|
---|
1133 | }
|
---|
1134 | /* Append next_line, read new next_line. */
|
---|
1135 | len = strlen(pattern_space);
|
---|
1136 | pattern_space = xrealloc(pattern_space, len + strlen(next_line) + 2);
|
---|
1137 | pattern_space[len] = '\n';
|
---|
1138 | strcpy(pattern_space + len+1, next_line);
|
---|
1139 | last_gets_char = next_gets_char;
|
---|
1140 | next_line = get_next_line(&next_gets_char);
|
---|
1141 | linenum++;
|
---|
1142 | break;
|
---|
1143 | }
|
---|
1144 |
|
---|
1145 | /* Test/branch if substitution occurred */
|
---|
1146 | case 't':
|
---|
1147 | if (!substituted) break;
|
---|
1148 | substituted = 0;
|
---|
1149 | /* Fall through */
|
---|
1150 | /* Test/branch if substitution didn't occur */
|
---|
1151 | case 'T':
|
---|
1152 | if (substituted) break;
|
---|
1153 | /* Fall through */
|
---|
1154 | /* Branch to label */
|
---|
1155 | case 'b':
|
---|
1156 | if (!sed_cmd->string) goto discard_commands;
|
---|
1157 | else sed_cmd = branch_to(sed_cmd->string);
|
---|
1158 | break;
|
---|
1159 | /* Transliterate characters */
|
---|
1160 | case 'y':
|
---|
1161 | {
|
---|
1162 | int i, j;
|
---|
1163 | for (i = 0; pattern_space[i]; i++) {
|
---|
1164 | for (j = 0; sed_cmd->string[j]; j += 2) {
|
---|
1165 | if (pattern_space[i] == sed_cmd->string[j]) {
|
---|
1166 | pattern_space[i] = sed_cmd->string[j + 1];
|
---|
1167 | break;
|
---|
1168 | }
|
---|
1169 | }
|
---|
1170 | }
|
---|
1171 |
|
---|
1172 | break;
|
---|
1173 | }
|
---|
1174 | case 'g': /* Replace pattern space with hold space */
|
---|
1175 | free(pattern_space);
|
---|
1176 | pattern_space = xstrdup(G.hold_space ? G.hold_space : "");
|
---|
1177 | break;
|
---|
1178 | case 'G': /* Append newline and hold space to pattern space */
|
---|
1179 | {
|
---|
1180 | int pattern_space_size = 2;
|
---|
1181 | int hold_space_size = 0;
|
---|
1182 |
|
---|
1183 | if (pattern_space)
|
---|
1184 | pattern_space_size += strlen(pattern_space);
|
---|
1185 | if (G.hold_space)
|
---|
1186 | hold_space_size = strlen(G.hold_space);
|
---|
1187 | pattern_space = xrealloc(pattern_space,
|
---|
1188 | pattern_space_size + hold_space_size);
|
---|
1189 | if (pattern_space_size == 2)
|
---|
1190 | pattern_space[0] = 0;
|
---|
1191 | strcat(pattern_space, "\n");
|
---|
1192 | if (G.hold_space)
|
---|
1193 | strcat(pattern_space, G.hold_space);
|
---|
1194 | last_gets_char = '\n';
|
---|
1195 |
|
---|
1196 | break;
|
---|
1197 | }
|
---|
1198 | case 'h': /* Replace hold space with pattern space */
|
---|
1199 | free(G.hold_space);
|
---|
1200 | G.hold_space = xstrdup(pattern_space);
|
---|
1201 | break;
|
---|
1202 | case 'H': /* Append newline and pattern space to hold space */
|
---|
1203 | {
|
---|
1204 | int hold_space_size = 2;
|
---|
1205 | int pattern_space_size = 0;
|
---|
1206 |
|
---|
1207 | if (G.hold_space)
|
---|
1208 | hold_space_size += strlen(G.hold_space);
|
---|
1209 | if (pattern_space)
|
---|
1210 | pattern_space_size = strlen(pattern_space);
|
---|
1211 | G.hold_space = xrealloc(G.hold_space,
|
---|
1212 | hold_space_size + pattern_space_size);
|
---|
1213 |
|
---|
1214 | if (hold_space_size == 2)
|
---|
1215 | *G.hold_space = 0;
|
---|
1216 | strcat(G.hold_space, "\n");
|
---|
1217 | if (pattern_space)
|
---|
1218 | strcat(G.hold_space, pattern_space);
|
---|
1219 |
|
---|
1220 | break;
|
---|
1221 | }
|
---|
1222 | case 'x': /* Exchange hold and pattern space */
|
---|
1223 | {
|
---|
1224 | char *tmp = pattern_space;
|
---|
1225 | pattern_space = G.hold_space ? G.hold_space : xzalloc(1);
|
---|
1226 | last_gets_char = '\n';
|
---|
1227 | G.hold_space = tmp;
|
---|
1228 | break;
|
---|
1229 | }
|
---|
1230 | } /* switch */
|
---|
1231 | } /* for each cmd */
|
---|
1232 |
|
---|
1233 | /*
|
---|
1234 | * Exit point from sedding...
|
---|
1235 | */
|
---|
1236 | discard_commands:
|
---|
1237 | /* we will print the line unless we were told to be quiet ('-n')
|
---|
1238 | or if the line was suppressed (ala 'd'elete) */
|
---|
1239 | if (!G.be_quiet)
|
---|
1240 | sed_puts(pattern_space, last_gets_char);
|
---|
1241 |
|
---|
1242 | /* Delete and such jump here. */
|
---|
1243 | discard_line:
|
---|
1244 | flush_append();
|
---|
1245 | free(pattern_space);
|
---|
1246 |
|
---|
1247 | goto again;
|
---|
1248 | }
|
---|
1249 |
|
---|
1250 | /* It is possible to have a command line argument with embedded
|
---|
1251 | * newlines. This counts as multiple command lines.
|
---|
1252 | * However, newline can be escaped: 's/e/z\<newline>z/'
|
---|
1253 | * We check for this.
|
---|
1254 | */
|
---|
1255 |
|
---|
1256 | static void add_cmd_block(char *cmdstr)
|
---|
1257 | {
|
---|
1258 | char *sv, *eol;
|
---|
1259 |
|
---|
1260 | cmdstr = sv = xstrdup(cmdstr);
|
---|
1261 | do {
|
---|
1262 | eol = strchr(cmdstr, '\n');
|
---|
1263 | next:
|
---|
1264 | if (eol) {
|
---|
1265 | /* Count preceding slashes */
|
---|
1266 | int slashes = 0;
|
---|
1267 | char *sl = eol;
|
---|
1268 |
|
---|
1269 | while (sl != cmdstr && *--sl == '\\')
|
---|
1270 | slashes++;
|
---|
1271 | /* Odd number of preceding slashes - newline is escaped */
|
---|
1272 | if (slashes & 1) {
|
---|
1273 | overlapping_strcpy(eol - 1, eol);
|
---|
1274 | eol = strchr(eol, '\n');
|
---|
1275 | goto next;
|
---|
1276 | }
|
---|
1277 | *eol = '\0';
|
---|
1278 | }
|
---|
1279 | add_cmd(cmdstr);
|
---|
1280 | cmdstr = eol + 1;
|
---|
1281 | } while (eol);
|
---|
1282 | free(sv);
|
---|
1283 | }
|
---|
1284 |
|
---|
1285 | int sed_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
|
---|
1286 | int sed_main(int argc UNUSED_PARAM, char **argv)
|
---|
1287 | {
|
---|
1288 | unsigned opt;
|
---|
1289 | llist_t *opt_e, *opt_f;
|
---|
1290 | int status = EXIT_SUCCESS;
|
---|
1291 |
|
---|
1292 | INIT_G();
|
---|
1293 |
|
---|
1294 | /* destroy command strings on exit */
|
---|
1295 | if (ENABLE_FEATURE_CLEAN_UP) atexit(sed_free_and_close_stuff);
|
---|
1296 |
|
---|
1297 | /* Lie to autoconf when it starts asking stupid questions. */
|
---|
1298 | if (argv[1] && !strcmp(argv[1], "--version")) {
|
---|
1299 | puts("This is not GNU sed version 4.0");
|
---|
1300 | return 0;
|
---|
1301 | }
|
---|
1302 |
|
---|
1303 | /* do normal option parsing */
|
---|
1304 | opt_e = opt_f = NULL;
|
---|
1305 | opt_complementary = "e::f::" /* can occur multiple times */
|
---|
1306 | "nn"; /* count -n */
|
---|
1307 | /* -i must be first, to match OPT_in_place definition */
|
---|
1308 | opt = getopt32(argv, "irne:f:", &opt_e, &opt_f,
|
---|
1309 | &G.be_quiet); /* counter for -n */
|
---|
1310 | //argc -= optind;
|
---|
1311 | argv += optind;
|
---|
1312 | if (opt & OPT_in_place) { // -i
|
---|
1313 | atexit(cleanup_outname);
|
---|
1314 | }
|
---|
1315 | if (opt & 0x2) G.regex_type |= REG_EXTENDED; // -r
|
---|
1316 | //if (opt & 0x4) G.be_quiet++; // -n
|
---|
1317 | while (opt_e) { // -e
|
---|
1318 | add_cmd_block(llist_pop(&opt_e));
|
---|
1319 | }
|
---|
1320 | while (opt_f) { // -f
|
---|
1321 | char *line;
|
---|
1322 | FILE *cmdfile;
|
---|
1323 | cmdfile = xfopen_for_read(llist_pop(&opt_f));
|
---|
1324 | while ((line = xmalloc_fgetline(cmdfile)) != NULL) {
|
---|
1325 | add_cmd(line);
|
---|
1326 | free(line);
|
---|
1327 | }
|
---|
1328 | fclose(cmdfile);
|
---|
1329 | }
|
---|
1330 | /* if we didn't get a pattern from -e or -f, use argv[0] */
|
---|
1331 | if (!(opt & 0x18)) {
|
---|
1332 | if (!*argv)
|
---|
1333 | bb_show_usage();
|
---|
1334 | add_cmd_block(*argv++);
|
---|
1335 | }
|
---|
1336 | /* Flush any unfinished commands. */
|
---|
1337 | add_cmd("");
|
---|
1338 |
|
---|
1339 | /* By default, we write to stdout */
|
---|
1340 | G.nonstdout = stdout;
|
---|
1341 |
|
---|
1342 | /* argv[0..(argc-1)] should be names of file to process. If no
|
---|
1343 | * files were specified or '-' was specified, take input from stdin.
|
---|
1344 | * Otherwise, we process all the files specified. */
|
---|
1345 | if (argv[0] == NULL) {
|
---|
1346 | if (opt & OPT_in_place)
|
---|
1347 | bb_error_msg_and_die(bb_msg_requires_arg, "-i");
|
---|
1348 | add_input_file(stdin);
|
---|
1349 | } else {
|
---|
1350 | int i;
|
---|
1351 | FILE *file;
|
---|
1352 |
|
---|
1353 | for (i = 0; argv[i]; i++) {
|
---|
1354 | struct stat statbuf;
|
---|
1355 | int nonstdoutfd;
|
---|
1356 |
|
---|
1357 | if (LONE_DASH(argv[i]) && !(opt & OPT_in_place)) {
|
---|
1358 | add_input_file(stdin);
|
---|
1359 | process_files();
|
---|
1360 | continue;
|
---|
1361 | }
|
---|
1362 | file = fopen_or_warn(argv[i], "r");
|
---|
1363 | if (!file) {
|
---|
1364 | status = EXIT_FAILURE;
|
---|
1365 | continue;
|
---|
1366 | }
|
---|
1367 | if (!(opt & OPT_in_place)) {
|
---|
1368 | add_input_file(file);
|
---|
1369 | continue;
|
---|
1370 | }
|
---|
1371 |
|
---|
1372 | G.outname = xasprintf("%sXXXXXX", argv[i]);
|
---|
1373 | nonstdoutfd = xmkstemp(G.outname);
|
---|
1374 | G.nonstdout = xfdopen_for_write(nonstdoutfd);
|
---|
1375 |
|
---|
1376 | /* Set permissions/owner of output file */
|
---|
1377 | fstat(fileno(file), &statbuf);
|
---|
1378 | /* chmod'ing AFTER chown would preserve suid/sgid bits,
|
---|
1379 | * but GNU sed 4.2.1 does not preserve them either */
|
---|
1380 | fchmod(nonstdoutfd, statbuf.st_mode);
|
---|
1381 | fchown(nonstdoutfd, statbuf.st_uid, statbuf.st_gid);
|
---|
1382 | add_input_file(file);
|
---|
1383 | process_files();
|
---|
1384 | fclose(G.nonstdout);
|
---|
1385 |
|
---|
1386 | G.nonstdout = stdout;
|
---|
1387 | /* unlink(argv[i]); */
|
---|
1388 | xrename(G.outname, argv[i]);
|
---|
1389 | free(G.outname);
|
---|
1390 | G.outname = NULL;
|
---|
1391 | }
|
---|
1392 | /* Here, to handle "sed 'cmds' nonexistent_file" case we did:
|
---|
1393 | * if (G.current_input_file >= G.input_file_count)
|
---|
1394 | * return status;
|
---|
1395 | * but it's not needed since process_files() works correctly
|
---|
1396 | * in this case too. */
|
---|
1397 | }
|
---|
1398 | process_files();
|
---|
1399 |
|
---|
1400 | return status;
|
---|
1401 | }
|
---|