1 /* gshell.c - Shell-related utilities
3 * Copyright 2000 Red Hat, Inc.
4 * g_execvpe implementation based on GNU libc execvp:
5 * Copyright 1991, 92, 95, 96, 97, 98, 99 Free Software Foundation, Inc.
7 * GLib is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public License as
9 * published by the Free Software Foundation; either version 2 of the
10 * License, or (at your option) any later version.
12 * GLib is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with GLib; see the file COPYING.LIB. If not, write
19 * to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
20 * Boston, MA 02111-1307, USA.
30 #warning "FIXME remove gettext hack"
36 g_shell_error_quark (void)
38 static GQuark quark = 0;
40 quark = g_quark_from_static_string ("g-shell-error-quark");
44 /* Single quotes preserve the literal string exactly. escape
45 * sequences are not allowed; not even \' - if you want a '
46 * in the quoted text, you have to do something like 'foo'\''bar'
48 * Double quotes allow $ ` " \ and newline to be escaped with backslash.
49 * Otherwise double quotes preserve things literally.
53 unquote_string_inplace (gchar* str, gchar** end, GError** err)
59 g_return_val_if_fail(end != NULL, FALSE);
60 g_return_val_if_fail(err == NULL || *err == NULL, FALSE);
61 g_return_val_if_fail(str != NULL, FALSE);
67 if (!(*s == '"' || *s == '\''))
70 *err = g_error_new(G_SHELL_ERROR,
71 G_SHELL_ERROR_BAD_QUOTING,
72 _("Quoted text doesn't begin with a quotation mark"));
77 /* Skip the initial quote mark */
80 if (quote_char == '"')
84 g_assert(s > dest); /* loop invariant */
89 /* End of the string, return now */
97 /* Possible escaped quote or \ */
112 /* not an escaped char */
115 /* ++s already done. */
127 g_assert(s > dest); /* loop invariant */
134 g_assert(s > dest); /* loop invariant */
138 /* End of the string, return now */
151 g_assert(s > dest); /* loop invariant */
155 /* If we reach here this means the close quote was never encountered */
160 *err = g_error_new(G_SHELL_ERROR,
161 G_SHELL_ERROR_BAD_QUOTING,
162 _("Unmatched quotation mark in command line or other shell-quoted text"));
169 * @unquoted_string: a literal string
171 * Quotes a string so that the shell (/bin/sh) will interpret the
172 * quoted string to mean @unquoted_string. If you pass a filename to
173 * the shell, for example, you should first quote it with this
174 * function. The return value must be freed with g_free(). The
175 * quoting style used is undefined (single or double quotes may be
178 * Return value: quoted string
181 g_shell_quote (const gchar *unquoted_string)
183 /* We always use single quotes, because the algorithm is cheesier.
184 * We could use double if we felt like it, that might be more
191 g_return_val_if_fail (unquoted_string != NULL, NULL);
193 dest = g_string_new ("'");
197 /* could speed this up a lot by appending chunks of text at a
202 /* Replace literal ' with a close ', a \', and a open ' */
204 g_string_append (dest, "'\\''");
206 g_string_append_c (dest, *p);
211 /* close the quote */
212 g_string_append_c (dest, '\'');
214 return g_string_free (dest, FALSE);
219 * @quoted_string: shell-quoted string
220 * @error: error return location or NULL
222 * Unquotes a string as the shell (/bin/sh) would. Only handles
223 * quotes; if a string contains file globs, arithmetic operators,
224 * variables, backticks, redirections, or other special-to-the-shell
225 * features, the result will be different from the result a real shell
226 * would produce (the variables, backticks, etc. will be passed
227 * through literally instead of being expanded). This function is
228 * guaranteed to succeed if applied to the result of
229 * g_shell_quote(). If it fails, it returns %NULL and sets the
230 * error. The @quoted_string need not actually contain quoted or
231 * escaped text; g_shell_unquote() simply goes through the string and
232 * unquotes/unescapes anything that the shell would. Both single and
233 * double quotes are handled, as are escapes including escaped
234 * newlines. The return value must be freed with g_free(). Possible
235 * errors are in the #G_SHELL_ERROR domain.
237 * Shell quoting rules are a bit strange. Single quotes preserve the
238 * literal string exactly. escape sequences are not allowed; not even
239 * \' - if you want a ' in the quoted text, you have to do something
240 * like 'foo'\''bar'. Double quotes allow $, `, ", \, and newline to
241 * be escaped with backslash. Otherwise double quotes preserve things
244 * Return value: an unquoted string
247 g_shell_unquote (const gchar *quoted_string,
255 g_return_val_if_fail (quoted_string != NULL, NULL);
257 unquoted = g_strdup (quoted_string);
261 retval = g_string_new (NULL);
263 /* The loop allows cases such as
264 * "foo"blah blah'bar'woo foo"baz"la la la\'\''foo'
268 /* Append all non-quoted chars, honoring backslash escape
271 while (*start && !(*start == '"' || *start == '\''))
275 /* all characters can get escaped by backslash,
276 * except newline, which is removed if it follows
277 * a backslash outside of quotes
284 g_string_append_c (retval, *start);
290 g_string_append_c (retval, *start);
297 if (!unquote_string_inplace (start, &end, error))
303 g_string_append (retval, start);
310 return g_string_free (retval, FALSE);
313 g_assert (error == NULL || *error != NULL);
316 g_string_free (retval, TRUE);
320 /* g_parse_argv() does a semi-arbitrary weird subset of the way
321 * the shell parses a command line. We don't do variable expansion,
322 * don't understand that operators are tokens, don't do tilde expansion,
323 * don't do command substitution, no arithmetic expansion, IFS gets ignored,
324 * don't do filename globs, don't remove redirection stuff, etc.
326 * READ THE UNIX98 SPEC on "Shell Command Language" before changing
327 * the behavior of this code.
329 * Steps to parsing the argv string:
331 * - tokenize the string (but since we ignore operators,
332 * our tokenization may diverge from what the shell would do)
333 * note that tokenization ignores the internals of a quoted
334 * word and it always splits on spaces, not on IFS even
335 * if we used IFS. We also ignore "end of input indicator"
336 * (I guess this is control-D?)
338 * Tokenization steps, from UNIX98 with operator stuff removed,
341 * 1) "If the current character is backslash, single-quote or
342 * double-quote (\, ' or ") and it is not quoted, it will affect
343 * quoting for subsequent characters up to the end of the quoted
344 * text. The rules for quoting are as described in Quoting
345 * . During token recognition no substitutions will be actually
346 * performed, and the result token will contain exactly the
347 * characters that appear in the input (except for newline
348 * character joining), unmodified, including any embedded or
349 * enclosing quotes or substitution operators, between the quote
350 * mark and the end of the quoted text. The token will not be
351 * delimited by the end of the quoted field."
353 * 2) "If the current character is an unquoted newline character,
354 * the current token will be delimited."
356 * 3) "If the current character is an unquoted blank character, any
357 * token containing the previous character is delimited and the
358 * current character will be discarded."
360 * 4) "If the previous character was part of a word, the current
361 * character will be appended to that word."
363 * 5) "If the current character is a "#", it and all subsequent
364 * characters up to, but excluding, the next newline character
365 * will be discarded as a comment. The newline character that
366 * ends the line is not considered part of the comment. The
367 * "#" starts a comment only when it is at the beginning of a
368 * token. Since the search for the end-of-comment does not
369 * consider an escaped newline character specially, a comment
370 * cannot be continued to the next line."
372 * 6) "The current character will be used as the start of a new word."
375 * - for each token (word), perform portions of word expansion, namely
376 * field splitting (using default whitespace IFS) and quote
377 * removal. Field splitting may increase the number of words.
378 * Quote removal does not increase the number of words.
380 * "If the complete expansion appropriate for a word results in an
381 * empty field, that empty field will be deleted from the list of
382 * fields that form the completely expanded command, unless the
383 * original word contained single-quote or double-quote characters."
390 ensure_token (GString **token)
393 *token = g_string_new (NULL);
397 delimit_token (GString **token,
403 *retval = g_slist_prepend (*retval, g_string_free (*token, FALSE));
409 tokenize_command_line (const gchar *command_line,
414 GString *current_token = NULL;
415 GSList *retval = NULL;
418 current_quote = '\0';
424 if (current_quote == '\\')
428 /* we append nothing; backslash-newline become nothing */
432 /* we append the backslash and the current char,
433 * to be interpreted later after tokenization
435 ensure_token (¤t_token);
436 g_string_append_c (current_token, '\\');
437 g_string_append_c (current_token, *p);
440 current_quote = '\0';
442 else if (current_quote == '#')
444 /* Discard up to and including next newline */
445 while (*p && *p != '\n')
448 current_quote = '\0';
453 else if (current_quote)
455 if (*p == current_quote &&
456 /* check that it isn't an escaped double quote */
457 !(current_quote == '"' && quoted))
459 /* close the quote */
460 current_quote = '\0';
463 /* Everything inside quotes, and the close quote,
464 * gets appended literally.
467 ensure_token (¤t_token);
468 g_string_append_c (current_token, *p);
475 delimit_token (¤t_token, &retval);
480 /* If the current token contains the previous char, delimit
481 * the current token. A nonzero length
482 * token should always contain the previous char.
485 current_token->len > 0)
487 delimit_token (¤t_token, &retval);
490 /* discard all unquoted blanks (don't add them to a token) */
494 /* single/double quotes are appended to the token,
495 * escapes are maybe appended next time through the loop,
496 * comment chars are never appended.
501 ensure_token (¤t_token);
502 g_string_append_c (current_token, *p);
512 /* Combines rules 4) and 6) - if we have a token, append to it,
513 * otherwise create a new token.
515 ensure_token (¤t_token);
516 g_string_append_c (current_token, *p);
521 /* We need to count consecutive backslashes mod 2,
522 * to detect escaped doublequotes.
532 delimit_token (¤t_token, &retval);
536 if (current_quote == '\\')
539 G_SHELL_ERROR_BAD_QUOTING,
540 _("Text ended just after a '\\' character."
541 " (The text was '%s')"),
546 G_SHELL_ERROR_BAD_QUOTING,
547 _("Text ended before matching quote was found for %c."
548 " (The text was '%s')"),
549 current_quote, command_line);
558 G_SHELL_ERROR_EMPTY_STRING,
559 _("Text was empty (or contained only whitespace)"));
564 /* we appended backward */
565 retval = g_slist_reverse (retval);
570 g_assert (error == NULL || *error != NULL);
574 g_slist_foreach (retval, (GFunc)g_free, NULL);
575 g_slist_free (retval);
582 * g_shell_parse_argv:
583 * @command_line: command line to parse
584 * @argcp: return location for number of args
585 * @argvp: return location for array of args
586 * @error: return location for error
588 * Parses a command line into an argument vector, in much the same way
589 * the shell would, but without many of the expansions the shell would
590 * perform (variable expansion, globs, operators, filename expansion,
591 * etc. are not supported). The results are defined to be the same as
592 * those you would get from a UNIX98 /bin/sh, as long as the input
593 * contains none of the unsupported shell expansions. If the input
594 * does contain such expansions, they are passed through
595 * literally. Possible errors are those from the #G_SHELL_ERROR
596 * domain. Free the returned vector with g_strfreev().
598 * Return value: %TRUE on success, %FALSE if error set
601 g_shell_parse_argv (const gchar *command_line,
606 /* Code based on poptParseArgvString() from libpopt */
609 GSList *tokens = NULL;
613 g_return_val_if_fail (command_line != NULL, FALSE);
615 tokens = tokenize_command_line (command_line, error);
619 /* Because we can't have introduced any new blank space into the
620 * tokens (we didn't do any new expansions), we don't need to
621 * perform field splitting. If we were going to honor IFS or do any
622 * expansions, we would have to do field splitting on each word
623 * here. Also, if we were going to do any expansion we would need to
624 * remove any zero-length words that didn't contain quotes
625 * originally; but since there's no expansion we know all words have
626 * nonzero length, unless they contain quotes.
628 * So, we simply remove quotes, and don't do any field splitting or
629 * empty word removal, since we know there was no way to introduce
633 argc = g_slist_length (tokens);
634 argv = g_new0 (gchar*, argc + 1);
639 argv[i] = g_shell_unquote (tmp_list->data, error);
641 /* Since we already checked that quotes matched up in the
642 * tokenizer, this shouldn't be possible to reach I guess.
647 tmp_list = g_slist_next (tmp_list);
651 g_slist_foreach (tokens, (GFunc)g_free, NULL);
652 g_slist_free (tokens);
666 g_assert (error == NULL || *error != NULL);
668 g_slist_foreach (tokens, (GFunc) g_free, NULL);
669 g_slist_free (tokens);