1 /*****************************************************************************\
4 * Author : Chris Koeritz *
6 *******************************************************************************
7 * Copyright (c) 1993-$now By Author. This program is free software; you can *
8 * redistribute it and/or modify it under the terms of the GNU General Public *
9 * License as published by the Free Software Foundation; either version 2 of *
10 * the License or (at your option) any later version. This is online at: *
11 * http://www.fsf.org/copyleft/gpl.html *
12 * Please send any updates to: fred@gruntose.com *
13 \*****************************************************************************/
15 // implementation note: the filename is kept canonicalized. any constructor
16 // or assignment operator should ensure this (except the blank constructor).
20 #include <basis/byte_array.h>
21 #include <basis/functions.h>
22 #include <textual/parser_bits.h>
23 #include <system_helper.h>
27 #include <sys/types.h>
28 #if defined(__UNIX__) || defined(__GNU_WINDOWS__)
35 #define LOG(to_print) printf("%s::%s: %s\n", static_class_name(), func, astring(to_print).s())
37 using namespace basis;
38 using namespace structures;
40 class status_info : public stat
44 namespace filesystem {
46 //#if defined(__WIN32__) || defined(__VMS__)
47 // const char DEFAULT_SEPARATOR = '\\';
48 //#elif defined(__UNIX__)
49 const char DEFAULT_SEPARATOR = '/';
51 // #error "We have no idea what the default path separator is."
54 const char *NO_PARENT_DEFAULT = ".";
55 // used when no directory name can be popped off.
62 filename::filename(const astring &name)
67 filename::filename(const astring &directory, const astring &name_of_file)
71 // if the directory is empty, use the current directory.
73 *this = astring(NO_PARENT_DEFAULT);
74 _had_directory = false;
76 // check for a slash on the end of the directory. add one if there is none
78 bool add_slash = false;
79 if ( (directory[directory.end()] != '\\')
80 && (directory[directory.end()] != '/') ) add_slash = true;
81 if (add_slash) *this += DEFAULT_SEPARATOR;
82 *this += name_of_file;
86 filename::filename(const filename &to_copy)
88 _had_directory(to_copy._had_directory)
91 filename::~filename() {}
93 astring filename::default_separator() { return astring(DEFAULT_SEPARATOR, 1); }
95 astring &filename::raw() { return *this; }
97 const astring &filename::raw() const { return *this; }
99 bool filename::good() const { return exists(); }
101 bool filename::unlink() const { return ::unlink(observe()) == 0; }
103 void filename::reset(const astring &name) {
105 _had_directory = true; // until we know better.
109 astring filename::null_device()
118 bool filename::separator(char is_it)
119 { return (is_it == pc_separator) || (is_it == unix_separator); }
121 filename &filename::operator = (const filename &to_copy)
123 if (this == &to_copy) return *this;
124 (astring &)(*this) = to_copy;
125 _had_directory = to_copy._had_directory;
129 filename &filename::operator = (const astring &to_copy)
131 _had_directory = true;
132 if (this == &to_copy) return *this;
133 (astring &)(*this) = to_copy;
138 astring filename::pop()
140 astring to_return = basename();
141 filename parent_dir = parent();
142 if (parent_dir.raw().equal_to(NO_PARENT_DEFAULT)) {
143 // we haven't gone anywhere.
144 return ""; // signal that nothing was removed.
150 filename filename::parent() const { return dirname(); }
152 void filename::push(const astring &to_push)
154 *this = filename(*this, to_push);
157 void filename::canonicalize()
159 FUNCDEF("canonicalize");
160 // turn all the non-default separators into the default.
161 bool found_sep = false;
162 for (int j = 0; j < length(); j++) {
163 if (separator(get(j))) {
165 put(j, DEFAULT_SEPARATOR);
169 // if there wasn't a single directory separator, then they must not have
170 // specified any directory name for this filename (although it could itself
172 if (!found_sep) _had_directory = false;
174 // remove all occurrences of double separators except for the first
175 // double set, which could be a UNC filename. that's why the index below
176 // starts at one rather than zero.
177 bool saw_sep = false;
178 for (int i = 1; i < length(); i++) {
179 if (separator(get(i))) {
182 // two in a row is no good, except for the first two.
183 i--; // skip back one and try again.
187 } else saw_sep = false;
191 // on windows, we want to translate away from any cygwin or msys format into a more palatable
192 // version that the rest of windows understands.
194 //hmmm: make these into statics!
195 const astring CYGDRIVE_SENTINEL = "cygdrive";
196 const astring CYGDRIVE_PATH = astring(astring(DEFAULT_SEPARATOR, 1)
197 + CYGDRIVE_SENTINEL + astring(DEFAULT_SEPARATOR, 1));
199 // must be at least as long as the string we're looking for, plus a drive letter afterwards.
200 if ( (length() >= CYGDRIVE_PATH.length() + 1)
202 && separator(get(CYGDRIVE_PATH.length() - 1))
203 && compare(CYGDRIVE_SENTINEL, 1,
204 0, CYGDRIVE_SENTINEL.length(), true) ) {
205 zap(0, CYGDRIVE_PATH.length() - 1); // whack the cygdrive portion plus two slashes.
206 insert(1, ":"); // add a colon after the imputed drive letter.
207 //LOG(astring("turned cygdrive path string into: ") + *this);
209 //LOG(astring("path didn't match so left as: ") + *this);
212 // now we convert msys...
213 if ( (length() >= 2) && (get(0) == DEFAULT_SEPARATOR)
214 && textual::parser_bits::is_alpha(get(1)) ) {
215 // we seem reasonably sure now that this is a windows path hiding in msys format, but
216 // the next character needs to be a slash (if there is a next character) for it to be
217 // the windows drive form. otherwise it could be /tmp, which would obviously not be
218 // intended as a windows path.
219 if ( (length() == 2) || (get(2) == DEFAULT_SEPARATOR) ) {
220 // cool, this should be interpretable as an msys path, except for those wacky types
221 // of folks that might use a top-level single character directory name. we cannot
222 // help them, because we have made a design decision to support msys-style paths.
223 // note that this would only affect someone if they were referring to their directory on
224 // the current windows partition (c:, d:, etc.) without providing the drive letter,
225 // if they had that single character directory name (e.g., c:\x, d:\q, etc.) and even
226 // then only on the near defunct windows platform.
227 zap(0, 0); // take off initial slash.
228 insert(1, ":"); // add the obligatory colon.
229 //LOG(astring("turned msys string into: ") + *this);
233 // if we still have a unix style path here on windows, then there will be
234 // trouble when we pass that to the OS. we are not using any cygwin or
235 // other virtualization libraries directly, so we can't rely on those to
236 // fix the path. but if we built under something like cygwin, we should
237 // have stored the real dos-style location of the virtual unix root. we
238 // will use that to replace the root '/' and this should fix most of that
240 bool inject_root = false; // assume we don't need to do anything.
242 //LOG(astring("before root injection: ") + raw());
244 // condition here just checks if the path is only the root.
245 if ( (length() == 1) && separator(get(0)) ) { inject_root = true; }
247 //if (inject_root) LOG("decided to inject root since path is '/'.");
249 // condition is looking for first character being a slash, and second char as alphanumeric or dash or underscore.
250 // we will currently fail detecting freaky paths that don't start off with alphanumeric or one of that small set of special chars.
253 && ( textual::parser_bits::is_alphanumeric(get(1)) || ('-' == get(1)) || ('_' == get(1)) ) ) {
255 //if (inject_root) LOG(astring("decided to inject root since path is compatible: ") + *this);
258 //LOG(astring("after second phase root injection: ") + raw());
261 // inject the actual path to the unix root in front, if we know it.
262 // if we don't know it, then a default path that's unlikely to work is idiotically plugged in.
263 insert(0, FEISTY_MEOW_VIRTUAL_UNIX_ROOT);
264 //LOG(astring("turned cygdrive path string into: ") + raw());
268 // we don't crop the last separator if the name's too small. for msdos
269 // names, that would be chopping a slash off the c:\ style name.
271 // zap any separators that are hiding on the end.
272 const int last = end();
273 if (separator(get(last))) zap(last, last);
274 } else if ( (length() == 2) && (get(1) == ':') ) {
275 // special case for dos drive names. we turn it back into a valid
276 // directory rather than leaving it as just "X:". that form of the name
277 // means something else under dos/windows.
278 *this += astring(DEFAULT_SEPARATOR, 1);
282 char filename::drive(bool interact_with_fs) const
284 // first guess: if second letter's a colon, first letter's the drive.
289 if (!interact_with_fs)
292 // otherwise, retrieve the file system's record for the file.
294 if (!get_info(&fill))
296 return char('A' + fill.st_dev);
299 astring filename::extension() const
301 astring base(basename().raw());
302 int posn = base.find('.', base.end(), true);
305 return base.substring(posn + 1, base.length() - 1);
308 astring filename::rootname() const
310 astring base(basename().raw());
311 int posn = base.find('.', base.end(), true);
314 return base.substring(0, posn - 1);
317 bool filename::get_info(status_info *to_fill) const
319 int ret = stat(observe(), to_fill);
325 bool filename::is_directory() const
328 if (!get_info(&fill))
330 return !!(fill.st_mode & S_IFDIR);
333 bool filename::is_writable() const
336 if (!get_info(&fill))
338 return !!(fill.st_mode & S_IWRITE);
341 bool filename::is_readable() const
344 if (!get_info(&fill))
346 return !!(fill.st_mode & S_IREAD);
349 bool filename::is_executable() const
352 if (!get_info(&fill))
354 return !!(fill.st_mode & S_IEXEC);
357 bool filename::is_normal() const
360 if (!get_info(&fill))
362 #if defined(__WIN32__) || defined(__VMS__)
363 //hmmm: is there a corresponding set of functions for windows, where applicable?
366 bool weird = S_ISCHR(fill.st_mode)
367 || S_ISBLK(fill.st_mode)
368 || S_ISFIFO(fill.st_mode)
369 || S_ISSOCK(fill.st_mode);
374 int filename::find_last_separator(const astring &look_at) const
379 sep = look_at.find(DEFAULT_SEPARATOR, last_sep + 1);
380 if (sep >= 0) last_sep = sep;
385 filename filename::basename() const
387 astring basename = *this;
388 int last_sep = find_last_separator(basename);
389 if (last_sep >= 0) basename.zap(0, last_sep);
393 filename filename::dirname() const
395 astring dirname = *this;
396 int last_sep = find_last_separator(dirname);
397 // we don't accept ripping off the first slash.
399 // we can rip the slash and suffix off to get the directory name. however,
400 // this might be in the form X: on windows. if they want the slash to
401 // remain, they can use the dirname that appends it.
402 dirname.zap(last_sep, dirname.end());
404 if (get(0) == DEFAULT_SEPARATOR) {
405 // handle when we're up at the top of the filesystem. on unix, once
406 // you hit the root, you can keep going up but you still remain at
407 // the root. similarly on windoze, if there's no drive name in there.
408 dirname = astring(DEFAULT_SEPARATOR, 1);
410 // there's no slash at all in the filename any more. we assume that
411 // the directory is the current one, if no other information is
412 // available. this default is already used by some code.
413 dirname = NO_PARENT_DEFAULT;
419 astring filename::dirname(bool add_slash) const
421 astring tempname = dirname().raw();
422 if (add_slash) tempname += DEFAULT_SEPARATOR;
426 bool filename::exists() const
430 // if the file name is empty, that cannot exist.
433 return is_readable();
436 bool filename::legal_character(char to_check)
441 case '*': case '?': case '$': case '&': case '|':
442 case '\'': case '"': case '`':
448 default: return true;
452 void filename::detooth_filename(astring &to_clean, char replacement)
454 for (int i = 0; i < to_clean.length(); i++) {
455 if (!legal_character(to_clean[i]))
456 to_clean[i] = replacement;
460 int filename::packed_size() const
462 return PACKED_SIZE_INT32 + astring::packed_size();
465 void filename::pack(byte_array &packed_form) const
467 attach(packed_form, int(_had_directory));
468 astring::pack(packed_form);
471 bool filename::unpack(byte_array &packed_form)
474 if (!detach(packed_form, temp))
476 _had_directory = temp;
477 if (!astring::unpack(packed_form))
482 void filename::separate(bool &rooted, string_array &pieces) const
485 const astring &raw_form = raw();
486 astring accumulator; // holds the names we find.
487 rooted = raw_form.length() && separator(raw_form[0]);
488 for (int i = 0; i < raw_form.length(); i++) {
489 if (separator(raw_form[i])) {
490 // this is a separator character, so eat it and add the accumulated
491 // string to the list.
492 if (i && accumulator.length()) pieces += accumulator;
493 // now reset our accumulated text.
494 accumulator = astring::empty_string();
496 // not a separator, so just accumulate it.
497 accumulator += raw_form[i];
500 if (accumulator.length()) pieces += accumulator;
503 void filename::join(bool rooted, const string_array &pieces)
505 astring constructed_name; // we'll make a filename here.
506 if (rooted) constructed_name += DEFAULT_SEPARATOR;
507 for (int i = 0; i < pieces.length(); i++) {
508 constructed_name += pieces[i];
509 if (!i || (i != pieces.length() - 1))
510 constructed_name += DEFAULT_SEPARATOR;
512 *this = constructed_name;
515 bool filename::base_compare_prefix(const filename &to_compare,
516 string_array &first, string_array &second)
519 separate(first_rooted, first);
521 to_compare.separate(second_rooted, second);
522 if (first_rooted != second_rooted) {
525 // that case should never be allowed, since there are some bits missing
526 // in the name to be compared.
527 if (first.length() > second.length())
530 // compare each of the pieces.
531 for (int i = 0; i < first.length(); i++) {
532 #if defined(__WIN32__) || defined(__VMS__)
533 // case-insensitive compare.
534 if (!first[i].iequals(second[i]))
537 // case-sensitive compare.
538 if (first[i] != second[i])
545 bool filename::compare_prefix(const filename &to_compare, astring &sequel)
547 sequel = astring::empty_string(); // clean our output parameter.
550 if (!base_compare_prefix(to_compare, first, second))
553 // create the sequel string.
554 int extra_strings = second.length() - first.length();
555 for (int i = second.length() - extra_strings; i < second.length(); i++) {
557 if (i != second.length() - 1) sequel += DEFAULT_SEPARATOR;
563 bool filename::compare_prefix(const filename &to_compare)
567 return base_compare_prefix(to_compare, first, second);
570 bool filename::base_compare_suffix(const filename &to_compare,
571 string_array &first, string_array &second)
574 separate(first_rooted, first);
576 to_compare.separate(second_rooted, second);
577 // that case should never be allowed, since there are some bits missing
578 // in the name to be compared.
579 if (first.length() > second.length())
582 // compare each of the pieces.
583 for (int i = first.length() - 1; i >= 0; i--) {
584 //clean up this computation; the difference in lengths is constant--use that.
585 int distance_from_end = first.length() - 1 - i;
586 int j = second.length() - 1 - distance_from_end;
587 #if defined(__WIN32__) || defined(__VMS__)
588 // case-insensitive compare.
589 if (!first[i].iequals(second[j]))
592 // case-sensitive compare.
593 if (first[i] != second[j])
600 bool filename::compare_suffix(const filename &to_compare, astring &prequel)
602 prequel = astring::empty_string(); // clean our output parameter.
605 if (!base_compare_suffix(to_compare, first, second))
608 // create the prequel string.
609 int extra_strings = second.length() - first.length();
610 for (int i = 0; i < extra_strings; i++) {
611 prequel += second[i];
612 if (i != second.length() - 1) prequel += DEFAULT_SEPARATOR;
617 bool filename::compare_suffix(const filename &to_compare)
621 return base_compare_suffix(to_compare, first, second);
624 bool filename::chmod(int write_mode, int owner_mode) const
628 if (write_mode & ALLOW_READ) {
629 if (owner_mode & USER_RIGHTS) chmod_value |= S_IRUSR;
630 if (owner_mode & GROUP_RIGHTS) chmod_value |= S_IRGRP;
631 if (owner_mode & OTHER_RIGHTS) chmod_value |= S_IROTH;
633 if (write_mode & ALLOW_WRITE) {
634 if (owner_mode & USER_RIGHTS) chmod_value |= S_IWUSR;
635 if (owner_mode & GROUP_RIGHTS) chmod_value |= S_IWGRP;
636 if (owner_mode & OTHER_RIGHTS) chmod_value |= S_IWOTH;
638 //// chmod_value = S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH;
639 #elif defined(__WIN32__)
640 if (write_mode & ALLOW_READ) {
641 chmod_value |= _S_IREAD;
643 if (write_mode & ALLOW_WRITE) {
644 chmod_value |= _S_IWRITE;
647 #error unsupported OS type currently.
649 int chmod_result = ::chmod(raw().s(), chmod_value);
651 // LOG(astring("there was a problem changing permissions on ") + raw());