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.
27 #warning "FIXME remove gettext hack"
33 g_shell_error_quark (void)
35 static GQuark quark = 0;
37 quark = g_quark_from_static_string ("g-shell-error-quark");
41 /* Single quotes preserve the literal string exactly. escape
42 * sequences are not allowed; not even \' - if you want a '
43 * in the quoted text, you have to do something like 'foo'\''bar'
45 * Double quotes allow $ ` " \ and newline to be escaped with backslash.
46 * Otherwise double quotes preserve things literally.
50 unquote_string_inplace (gchar* str, gchar** end, GError** err)
56 g_return_val_if_fail(end != NULL, FALSE);
57 g_return_val_if_fail(err == NULL || *err == NULL, FALSE);
58 g_return_val_if_fail(str != NULL, FALSE);
64 if (!(*s == '"' || *s == '\''))
67 *err = g_error_new(G_SHELL_ERROR,
68 G_SHELL_ERROR_BAD_QUOTING,
69 _("Quoted text doesn't begin with a quotation mark"));
74 /* Skip the initial quote mark */
77 if (quote_char == '"')
81 g_assert(s > dest); /* loop invariant */
86 /* End of the string, return now */
94 /* Possible escaped quote or \ */
109 /* not an escaped char */
112 /* ++s already done. */
124 g_assert(s > dest); /* loop invariant */
131 g_assert(s > dest); /* loop invariant */
135 /* End of the string, return now */
148 g_assert(s > dest); /* loop invariant */
152 /* If we reach here this means the close quote was never encountered */
157 *err = g_error_new(G_SHELL_ERROR,
158 G_SHELL_ERROR_BAD_QUOTING,
159 _("Unmatched quotation mark in command line or other shell-quoted text"));
166 * @unquoted_string: a literal string
168 * Quotes a string so that the shell (/bin/sh) will interpret the
169 * quoted string to mean @unquoted_string. If you pass a filename to
170 * the shell, for example, you should first quote it with this
171 * function. The return value must be freed with g_free(). The
172 * quoting style used is undefined (single or double quotes may be
175 * Return value: quoted string
178 g_shell_quote (const gchar *unquoted_string)
180 /* We always use single quotes, because the algorithm is cheesier.
181 * We could use double if we felt like it, that might be more
188 g_return_val_if_fail (unquoted_string != NULL, NULL);
190 dest = g_string_new ("'");
194 /* could speed this up a lot by appending chunks of text at a
199 /* Replace literal ' with a close ', a \', and a open ' */
201 g_string_append (dest, "'\''");
203 g_string_append_c (dest, *p);
208 /* close the quote */
209 g_string_append_c (dest, '\'');
211 return g_string_free (dest, FALSE);
216 * @quoted_string: shell-quoted string
217 * @error: error return location or NULL
219 * Unquotes a string as the shell (/bin/sh) would. Only handles
220 * quotes; if a string contains file globs, arithmetic operators,
221 * variables, backticks, redirections, or other special-to-the-shell
222 * features, the result will be different from the result a real shell
223 * would produce (the variables, backticks, etc. will be passed
224 * through literally instead of being expanded). This function is
225 * guaranteed to succeed if applied to the result of
226 * g_shell_quote(). If it fails, it returns NULL and sets the
227 * error. The @quoted_string need not actually contain quoted or
228 * escaped text; g_shell_unquote() simply goes through the string and
229 * unquotes/unescapes anything that the shell would. Both single and
230 * double quotes are handled, as are escapes including escaped
231 * newlines. The return value must be freed with g_free(). Possible
232 * errors are in the #G_SHELL_ERROR domain.
234 * Return value: an unquoted string
237 g_shell_unquote (const gchar *quoted_string,
245 g_return_val_if_fail (quoted_string != NULL, NULL);
247 unquoted = g_strdup (quoted_string);
251 retval = g_string_new ("");
253 /* The loop allows cases such as
254 * "foo"blah blah'bar'woo foo"baz"la la la\'\''foo'
258 /* Append all non-quoted chars, honoring backslash escape
261 while (*start && !(*start == '"' || *start == '\''))
265 /* all characters can get escaped by backslash,
266 * except newline, which is removed if it follows
267 * a backslash outside of quotes
274 g_string_append_c (retval, *start);
280 g_string_append_c (retval, *start);
287 if (!unquote_string_inplace (start, &end, error))
293 g_string_append (retval, start);
299 return g_string_free (retval, FALSE);
302 g_assert (error == NULL || *error != NULL);
305 g_string_free (retval, TRUE);
309 /* g_parse_argv() does a semi-arbitrary weird subset of the way
310 * the shell parses a command line. We don't do variable expansion,
311 * don't understand that operators are tokens, don't do tilde expansion,
312 * don't do command substitution, no arithmetic expansion, IFS gets ignored,
313 * don't do filename globs, don't remove redirection stuff, etc.
315 * READ THE UNIX98 SPEC on "Shell Command Language" before changing
316 * the behavior of this code.
318 * Steps to parsing the argv string:
320 * - tokenize the string (but since we ignore operators,
321 * our tokenization may diverge from what the shell would do)
322 * note that tokenization ignores the internals of a quoted
323 * word and it always splits on spaces, not on IFS even
324 * if we used IFS. We also ignore "end of input indicator"
325 * (I guess this is control-D?)
327 * Tokenization steps, from UNIX98 with operator stuff removed,
330 * 1) "If the current character is backslash, single-quote or
331 * double-quote (\, ' or ") and it is not quoted, it will affect
332 * quoting for subsequent characters up to the end of the quoted
333 * text. The rules for quoting are as described in Quoting
334 * . During token recognition no substitutions will be actually
335 * performed, and the result token will contain exactly the
336 * characters that appear in the input (except for newline
337 * character joining), unmodified, including any embedded or
338 * enclosing quotes or substitution operators, between the quote
339 * mark and the end of the quoted text. The token will not be
340 * delimited by the end of the quoted field."
342 * 2) "If the current character is an unquoted newline character,
343 * the current token will be delimited."
345 * 3) "If the current character is an unquoted blank character, any
346 * token containing the previous character is delimited and the
347 * current character will be discarded."
349 * 4) "If the previous character was part of a word, the current
350 * character will be appended to that word."
352 * 5) "If the current character is a "#", it and all subsequent
353 * characters up to, but excluding, the next newline character
354 * will be discarded as a comment. The newline character that
355 * ends the line is not considered part of the comment. The
356 * "#" starts a comment only when it is at the beginning of a
357 * token. Since the search for the end-of-comment does not
358 * consider an escaped newline character specially, a comment
359 * cannot be continued to the next line."
361 * 6) "The current character will be used as the start of a new word."
364 * - for each token (word), perform portions of word expansion, namely
365 * field splitting (using default whitespace IFS) and quote
366 * removal. Field splitting may increase the number of words.
367 * Quote removal does not increase the number of words.
369 * "If the complete expansion appropriate for a word results in an
370 * empty field, that empty field will be deleted from the list of
371 * fields that form the completely expanded command, unless the
372 * original word contained single-quote or double-quote characters."
379 ensure_token (GString **token)
382 *token = g_string_new ("");
386 delimit_token (GString **token,
392 *retval = g_slist_prepend (*retval, g_string_free (*token, FALSE));
398 tokenize_command_line (const gchar *command_line,
403 GString *current_token = NULL;
404 GSList *retval = NULL;
406 current_quote = '\0';
411 if (current_quote == '\\')
415 /* we append nothing; backslash-newline become nothing */
419 /* we append the backslash and the current char,
420 * to be interpreted later after tokenization
422 ensure_token (¤t_token);
423 g_string_append_c (current_token, '\\');
424 g_string_append_c (current_token, *p);
427 current_quote = '\0';
429 else if (current_quote == '#')
431 /* Discard up to and including next newline */
432 while (*p && *p != '\n')
435 current_quote = '\0';
440 else if (current_quote)
442 if (*p == current_quote &&
443 /* check that it isn't an escaped double quote */
444 !(current_quote == '"' && p != command_line && *(p - 1) == '\\'))
446 /* close the quote */
447 current_quote = '\0';
450 /* Everything inside quotes, and the close quote,
451 * gets appended literally.
454 ensure_token (¤t_token);
455 g_string_append_c (current_token, *p);
462 delimit_token (¤t_token, &retval);
467 /* If the current token contains the previous char, delimit
468 * the current token. A nonzero length
469 * token should always contain the previous char.
472 current_token->len > 0)
474 delimit_token (¤t_token, &retval);
477 /* discard all unquoted blanks (don't add them to a token) */
481 /* single/double quotes are appended to the token,
482 * escapes are maybe appended next time through the loop,
483 * comment chars are never appended.
488 ensure_token (¤t_token);
489 g_string_append_c (current_token, *p);
499 /* Combines rules 4) and 6) - if we have a token, append to it,
500 * otherwise create a new token.
502 ensure_token (¤t_token);
503 g_string_append_c (current_token, *p);
511 delimit_token (¤t_token, &retval);
515 if (current_quote == '\\')
518 G_SHELL_ERROR_BAD_QUOTING,
519 _("Text ended just after a '\' character."
520 " (The text was '%s')"),
525 G_SHELL_ERROR_BAD_QUOTING,
526 _("Text ended before matching quote was found for %c."
527 " (The text was '%s')"),
528 current_quote, command_line);
537 G_SHELL_ERROR_EMPTY_STRING,
538 _("Text was empty (or contained only whitespace)"));
543 /* we appended backward */
544 retval = g_slist_reverse (retval);
549 g_assert (error == NULL || *error != NULL);
553 g_slist_foreach (retval, (GFunc)g_free, NULL);
554 g_slist_free (retval);
561 * g_shell_parse_argv:
562 * @command_line: command line to parse
563 * @argcp: return location for number of args
564 * @argvp: return location for array of args
565 * @error: return location for error
567 * Parses a command line into an argument vector, in much the same way
568 * the shell would, but without many of the expansions the shell would
569 * perform (variable expansion, globs, operators, filename expansion,
570 * etc. are not supported). The results are defined to be the same as
571 * those you would get from a UNIX98 /bin/sh, as long as the input
572 * contains none of the unsupported shell expansions. If the input
573 * does contain such expansions, they are passed through
574 * literally. Possible errors are those from the #G_SHELL_ERROR
577 * Return value: TRUE on success, FALSE if error set
580 g_shell_parse_argv (const gchar *command_line,
585 /* Code based on poptParseArgvString() from libpopt */
588 GSList *tokens = NULL;
592 g_return_val_if_fail (command_line != NULL, FALSE);
594 tokens = tokenize_command_line (command_line, error);
598 /* Because we can't have introduced any new blank space into the
599 * tokens (we didn't do any new expansions), we don't need to
600 * perform field splitting. If we were going to honor IFS or do any
601 * expansions, we would have to do field splitting on each word
602 * here. Also, if we were going to do any expansion we would need to
603 * remove any zero-length words that didn't contain quotes
604 * originally; but since there's no expansion we know all words have
605 * nonzero length, unless they contain quotes.
607 * So, we simply remove quotes, and don't do any field splitting or
608 * empty word removal, since we know there was no way to introduce
612 argc = g_slist_length (tokens);
613 argv = g_new0 (gchar*, argc + 1);
618 argv[i] = g_shell_unquote (tmp_list->data, error);
620 /* Since we already checked that quotes matched up in the
621 * tokenizer, this shouldn't be possible to reach I guess.
626 tmp_list = g_slist_next (tmp_list);
630 g_slist_foreach (tokens, (GFunc)g_free, NULL);
631 g_slist_free (tokens);
645 g_assert (error == NULL || *error != NULL);
647 g_slist_foreach (tokens, (GFunc) g_free, NULL);
648 g_slist_free (tokens);