00001 /** 00002 * @copyright 00003 * ==================================================================== 00004 * Copyright (c) 2000-2004 CollabNet. All rights reserved. 00005 * 00006 * This software is licensed as described in the file COPYING, which 00007 * you should have received as part of this distribution. The terms 00008 * are also available at http://subversion.tigris.org/license-1.html. 00009 * If newer versions of this license are posted there, you may use a 00010 * newer version instead, at your option. 00011 * 00012 * This software consists of voluntary contributions made by many 00013 * individuals. For exact contribution history, see the revision 00014 * history and logs, available at http://subversion.tigris.org/. 00015 * ==================================================================== 00016 * @endcopyright 00017 * 00018 * @file svn_path.h 00019 * @brief A path manipulation library 00020 * 00021 * All incoming and outgoing paths are non-null and in UTF-8, unless 00022 * otherwise documented. 00023 * 00024 * No result path ever ends with a separator, no matter whether the 00025 * path is a file or directory, because we always canonicalize() it. 00026 * 00027 * All paths passed to the @c svn_path_xxx functions, with the exceptions of 00028 * the svn_path_canonicalize() and svn_path_internal_style() functions, 00029 * must be in canonical form. 00030 */ 00031 00032 #ifndef SVN_PATH_H 00033 #define SVN_PATH_H 00034 00035 00036 #include <apr_pools.h> 00037 #include <apr_tables.h> 00038 00039 #include "svn_string.h" 00040 #include "svn_error.h" 00041 00042 00043 #ifdef __cplusplus 00044 extern "C" { 00045 #endif /* __cplusplus */ 00046 00047 00048 00049 /** Convert @a path from the local style to the canonical internal style. */ 00050 const char *svn_path_internal_style(const char *path, apr_pool_t *pool); 00051 00052 /** Convert @a path from the canonical internal style to the local style. */ 00053 const char *svn_path_local_style(const char *path, apr_pool_t *pool); 00054 00055 00056 /** Join a base path (@a base) with a component (@a component), allocated in 00057 * @a pool. 00058 * 00059 * If either @a base or @a component is the empty path, then the other 00060 * argument will be copied and returned. If both are the empty path the 00061 * empty path is returned. 00062 * 00063 * If the @a component is an absolute path, then it is copied and returned. 00064 * Exactly one slash character ('/') is used to joined the components, 00065 * accounting for any trailing slash in @a base. 00066 * 00067 * Note that the contents of @a base are not examined, so it is possible to 00068 * use this function for constructing URLs, or for relative URLs or 00069 * repository paths. 00070 * 00071 * This function is NOT appropriate for native (local) file 00072 * paths. Only for "internal" canonicalized paths, since it uses '/' 00073 * for the separator. Further, an absolute path (for @a component) is 00074 * based on a leading '/' character. Thus, an "absolute URI" for the 00075 * @a component won't be detected. An absolute URI can only be used 00076 * for the base. 00077 */ 00078 char *svn_path_join(const char *base, 00079 const char *component, 00080 apr_pool_t *pool); 00081 00082 /** Join multiple components onto a @a base path, allocated in @a pool. The 00083 * components are terminated by a @c NULL. 00084 * 00085 * If any component is the empty string, it will be ignored. 00086 * 00087 * If any component is an absolute path, then it resets the base and 00088 * further components will be appended to it. 00089 * 00090 * See svn_path_join() for further notes about joining paths. 00091 */ 00092 char *svn_path_join_many(apr_pool_t *pool, const char *base, ...); 00093 00094 00095 /** Get the basename of the specified canonicalized @a path. The 00096 * basename is defined as the last component of the path (ignoring any 00097 * trailing slashes). If the @a path is root ("/"), then that is 00098 * returned. Otherwise, the returned value will have no slashes in 00099 * it. 00100 * 00101 * Example: svn_path_basename("/foo/bar") -> "bar" 00102 * 00103 * The returned basename will be allocated in @a pool. 00104 * 00105 * @note If an empty string is passed, then an empty string will be returned. 00106 */ 00107 char *svn_path_basename(const char *path, apr_pool_t *pool); 00108 00109 /** Get the dirname of the specified canonicalized @a path, defined as 00110 * the path with its basename removed. 00111 * 00112 * Get the dirname of the specified @a path, defined as the path with its 00113 * basename removed. If @a path is root ("/"), it is returned unchanged. 00114 * 00115 * The returned dirname will be allocated in @a pool. 00116 */ 00117 char *svn_path_dirname(const char *path, apr_pool_t *pool); 00118 00119 /** Return the number of components in the canonicalized @a path. 00120 * 00121 * @since New in 1.1. 00122 */ 00123 apr_size_t 00124 svn_path_component_count(const char *path); 00125 00126 /** Add a @a component (a null-terminated C-string) to the 00127 * canonicalized @a path. @a component is allowed to contain 00128 * directory separators. 00129 * 00130 * If @a path is non-empty, append the appropriate directory separator 00131 * character, and then @a component. If @a path is empty, simply set it to 00132 * @a component; don't add any separator character. 00133 * 00134 * If the result ends in a separator character, then remove the separator. 00135 */ 00136 void svn_path_add_component(svn_stringbuf_t *path, 00137 const char *component); 00138 00139 /** Remove one component off the end of the canonicalized @a path. */ 00140 void svn_path_remove_component(svn_stringbuf_t *path); 00141 00142 /** Remove @a n components off the end of the canonicalized @a path. 00143 * Equivalent to calling svn_path_remove_component() @a n times. 00144 * 00145 * @since New in 1.1. 00146 */ 00147 void svn_path_remove_components(svn_stringbuf_t *path, apr_size_t n); 00148 00149 /** Divide the canonicalized @a path into @a *dirpath and @a 00150 * *base_name, allocated in @a pool. 00151 * 00152 * If @a dirpath or @a base_name is null, then don't set that one. 00153 * 00154 * Either @a dirpath or @a base_name may be @a path's own address, but they 00155 * may not both be the same address, or the results are undefined. 00156 * 00157 * If @a path has two or more components, the separator between @a dirpath 00158 * and @a base_name is not included in either of the new names. 00159 * 00160 * examples: 00161 * - <pre>"/foo/bar/baz" ==> "/foo/bar" and "baz"</pre> 00162 * - <pre>"/bar" ==> "/" and "bar"</pre> 00163 * - <pre>"/" ==> "/" and "/"</pre> 00164 * - <pre>"bar" ==> "" and "bar"</pre> 00165 * - <pre>"" ==> "" and ""</pre> 00166 */ 00167 void svn_path_split(const char *path, 00168 const char **dirpath, 00169 const char **base_name, 00170 apr_pool_t *pool); 00171 00172 00173 /** Return non-zero iff @a path is empty ("") or represents the current 00174 * directory -- that is, if prepending it as a component to an existing 00175 * path would result in no meaningful change. 00176 */ 00177 int svn_path_is_empty(const char *path); 00178 00179 00180 /** Return a new path (or URL) like @a path, but transformed such that 00181 * some types of path specification redundancies are removed. 00182 * 00183 * This involves collapsing redundant "/./" elements, removing 00184 * multiple adjacent separator characters, removing trailing 00185 * separator characters, and possibly other semantically inoperative 00186 * transformations. 00187 * 00188 * The returned path may be statically allocated, equal to @a path, or 00189 * allocated from @a pool. 00190 */ 00191 const char *svn_path_canonicalize(const char *path, apr_pool_t *pool); 00192 00193 00194 /** Return an integer greater than, equal to, or less than 0, according 00195 * as @a path1 is greater than, equal to, or less than @a path2. 00196 */ 00197 int svn_path_compare_paths(const char *path1, const char *path2); 00198 00199 00200 /** Return the longest common path shared by two canonicalized paths, 00201 * @a path1 and @a path2. If there's no common ancestor, return the 00202 * empty path. 00203 * 00204 * @a path1 and @a path2 may be URLs. In order for two URLs to have 00205 * a common ancestor, they must (a) have the same protocol (since two URLs 00206 * with the same path but different protocols may point at completely 00207 * different resources), and (b) share a common ancestor in their path 00208 * component, i.e. 'protocol://' is not a sufficient ancestor. 00209 */ 00210 char *svn_path_get_longest_ancestor(const char *path1, 00211 const char *path2, 00212 apr_pool_t *pool); 00213 00214 /** Convert @a relative canonicalized path to an absolute path and 00215 * return the results in @a *pabsolute, allocated in @a pool. 00216 * 00217 * @a relative may be a URL, in which case no attempt is made to convert it, 00218 * and a copy of the URL is returned. 00219 */ 00220 svn_error_t * 00221 svn_path_get_absolute(const char **pabsolute, 00222 const char *relative, 00223 apr_pool_t *pool); 00224 00225 /** Return the path part of the canonicalized @a path in @a 00226 * *pdirectory, and the file part in @a *pfile. If @a path is a 00227 * directory, set @a *pdirectory to @a path, and @a *pfile to the 00228 * empty string. If @a path does not exist it is treated as if it is 00229 * a file, since directories do not normally vanish. 00230 */ 00231 svn_error_t * 00232 svn_path_split_if_file(const char *path, 00233 const char **pdirectory, 00234 const char **pfile, 00235 apr_pool_t *pool); 00236 00237 /** Find the common prefix of the canonicalized paths in @a targets 00238 * (an array of @a const char *'s), and remove redundant paths if @a 00239 * remove_redundancies is true. 00240 * 00241 * - Set @a *pcommon to the absolute path of the path or URL common to 00242 * all of the targets. If the targets have no common prefix, or 00243 * are a mix of URLs and local paths, set @a *pcommon to the 00244 * empty string. 00245 * 00246 * - If @a pcondensed_targets is non-null, set @a *pcondensed_targets 00247 * to an array of targets relative to @a *pcommon, and if 00248 * @a remove_redundancies is true, omit any paths/URLs that are 00249 * descendants of another path/URL in @a targets. If *pcommon 00250 * is empty, @a *pcondensed_targets will contain full URLs and/or 00251 * absolute paths; redundancies can still be removed (from both URLs 00252 * and paths). If @a pcondensed_targets is null, leave it alone. 00253 * 00254 * Else if there is exactly one target, then 00255 * 00256 * - Set @a *pcommon to that target, and 00257 * 00258 * - If @a pcondensed_targets is non-null, set @a *pcondensed_targets 00259 * to an array containing zero elements. Else if 00260 * @a pcondensed_targets is null, leave it alone. 00261 * 00262 * If there are no items in @a targets, set @a *pcommon and (if 00263 * applicable) @a *pcondensed_targets to @c NULL. 00264 * 00265 * @note There is no guarantee that @a *pcommon is within a working 00266 * copy. */ 00267 svn_error_t * 00268 svn_path_condense_targets(const char **pcommon, 00269 apr_array_header_t **pcondensed_targets, 00270 const apr_array_header_t *targets, 00271 svn_boolean_t remove_redundancies, 00272 apr_pool_t *pool); 00273 00274 00275 /** Copy a list of canonicalized @a targets, one at a time, into @a 00276 * pcondensed_targets, omitting any targets that are found earlier in 00277 * the list, or whose ancestor is found earlier in the list. Ordering 00278 * of targets in the original list is preserved in the condensed list 00279 * of targets. Use @a pool for any allocations. 00280 * 00281 * How does this differ in functionality from svn_path_condense_targets()? 00282 * 00283 * Here's the short version: 00284 * 00285 * 1. Disclaimer: if you wish to debate the following, talk to Karl. :-) 00286 * Order matters for updates because a multi-arg update is not 00287 * atomic, and CVS users are used to, when doing 'cvs up targetA 00288 * targetB' seeing targetA get updated, then targetB. I think the 00289 * idea is that if you're in a time-sensitive or flaky-network 00290 * situation, a user can say, "I really *need* to update 00291 * wc/A/D/G/tau, but I might as well update my whole working copy if 00292 * I can." So that user will do 'svn up wc/A/D/G/tau wc', and if 00293 * something dies in the middles of the 'wc' update, at least the 00294 * user has 'tau' up-to-date. 00295 * 00296 * 2. Also, we have this notion of an anchor and a target for updates 00297 * (the anchor is where the update editor is rooted, the target is 00298 * the actual thing we want to update). I needed a function that 00299 * would NOT screw with my input paths so that I could tell the 00300 * difference between someone being in A/D and saying 'svn up G' and 00301 * being in A/D/G and saying 'svn up .' -- believe it or not, these 00302 * two things don't mean the same thing. svn_path_condense_targets() 00303 * plays with absolute paths (which is fine, so does 00304 * svn_path_remove_redundancies()), but the difference is that it 00305 * actually tweaks those targets to be relative to the "grandfather 00306 * path" common to all the targets. Updates don't require a 00307 * "grandfather path" at all, and even if it did, the whole 00308 * conversion to an absolute path drops the crucial difference 00309 * between saying "i'm in foo, update bar" and "i'm in foo/bar, 00310 * update '.'" 00311 */ 00312 svn_error_t * 00313 svn_path_remove_redundancies(apr_array_header_t **pcondensed_targets, 00314 const apr_array_header_t *targets, 00315 apr_pool_t *pool); 00316 00317 00318 /** Decompose the canonicalized @a path into an array of <tt>const 00319 * char *</tt> components, allocated in @a pool. If @a path is 00320 * absolute, the first component will be a lone dir separator (the 00321 * root directory). 00322 */ 00323 apr_array_header_t *svn_path_decompose(const char *path, 00324 apr_pool_t *pool); 00325 00326 00327 /** Test that @a name is a single path component, that is: 00328 * - not @c NULL or empty. 00329 * - not a `/'-separated directory path 00330 * - not empty or `..' 00331 */ 00332 svn_boolean_t svn_path_is_single_path_component(const char *name); 00333 00334 00335 /** 00336 * Test to see if a backpath, i.e. '..', is present in @a path. 00337 * If not, return @c FALSE. 00338 * If so, return @c TRUE. 00339 * 00340 * @since New in 1.1. 00341 */ 00342 svn_boolean_t svn_path_is_backpath_present(const char *path); 00343 00344 00345 /** Test if @a path2 is a child of @a path1. 00346 * If not, return @c NULL. 00347 * If so, return a copy of the remainder path, allocated in @a pool. 00348 * (The remainder is the component which, added to @a path1, yields 00349 * @a path2. The remainder does not begin with a dir separator.) 00350 * 00351 * Both paths must be in canonical form, and must either be absolute, 00352 * or contain no ".." components. 00353 * 00354 * If @a path2 is the same as @a path1, it is not considered a child, so the 00355 * result is @c NULL; an empty string is never returned. 00356 * 00357 * ### todo: the ".." restriction is unfortunate, and would ideally 00358 * be lifted by making the implementation smarter. But this is not 00359 * trivial: if the path is "../foo", how do you know whether or not 00360 * the current directory is named "foo" in its parent? 00361 */ 00362 const char *svn_path_is_child(const char *path1, 00363 const char *path2, 00364 apr_pool_t *pool); 00365 00366 /** Return true if @a path1 is an ancestor of @a path2 or the paths are equal 00367 * and false otherwise. 00368 * 00369 * @since New in 1.3. 00370 */ 00371 svn_boolean_t 00372 svn_path_is_ancestor(const char *path1, const char *path2); 00373 00374 /** 00375 * Check whether @a path is a valid Subversion path. 00376 * 00377 * A valid Subversion pathname is a UTF-8 string without control 00378 * characters. "Valid" means Subversion can store the pathname in 00379 * a repository. There may be other, OS-specific, limitations on 00380 * what paths can be represented in a working copy. 00381 * 00382 * ASSUMPTION: @a path is a valid UTF-8 string. This function does 00383 * not check UTF-8 validity. 00384 * 00385 * Return @c SVN_NO_ERROR if valid and @c SVN_ERR_FS_PATH_SYNTAX if 00386 * invalid. 00387 * 00388 * @since New in 1.2. 00389 */ 00390 svn_error_t *svn_path_check_valid(const char *path, apr_pool_t *pool); 00391 00392 00393 /** URI/URL stuff 00394 * 00395 * @defgroup svn_path_uri_stuff URI/URL stuff 00396 * @{ 00397 */ 00398 00399 /** Return @c TRUE iff @a path looks like a valid URL, @c FALSE otherwise. */ 00400 svn_boolean_t svn_path_is_url(const char *path); 00401 00402 /** Return @c TRUE iff @a path is URI-safe, @c FALSE otherwise. */ 00403 svn_boolean_t svn_path_is_uri_safe(const char *path); 00404 00405 /** Return a URI-encoded copy of @a path, allocated in @a pool. */ 00406 const char *svn_path_uri_encode(const char *path, apr_pool_t *pool); 00407 00408 /** Return a URI-decoded copy of @a path, allocated in @a pool. */ 00409 const char *svn_path_uri_decode(const char *path, apr_pool_t *pool); 00410 00411 /** Extend @a url by a single @a component, URI-encoding that @a component 00412 * before adding it to the @a url. Return the new @a url, allocated in 00413 * @a pool. Notes: if @a component is @c NULL, just return a copy or @a url 00414 * allocated in @a pool; if @a component is already URI-encoded, calling 00415 * code should just use <tt>svn_path_join (url, component, pool)</tt>. @a url 00416 * does not need to be a canonical path, it may have trailing '/'. 00417 */ 00418 const char *svn_path_url_add_component(const char *url, 00419 const char *component, 00420 apr_pool_t *pool); 00421 00422 /** 00423 * Convert @a iri (Internationalized URI) to an URI. 00424 * The return value may be the same as @a iri if it was already 00425 * a URI. Else, allocate the return value in @a pool. 00426 * 00427 * @since New in 1.1. 00428 */ 00429 const char *svn_path_uri_from_iri(const char *iri, 00430 apr_pool_t *pool); 00431 00432 /** 00433 * URI-encode certain characters in @a uri that are not valid in an URI, but 00434 * doesn't have any special meaning in @a uri at their positions. If no 00435 * characters need escaping, just return @a uri. 00436 * 00437 * @note Currently, this function escapes <, >, ", space, {, }, |, \, ^, and `. 00438 * This may be extended in the future to do context-dependent escaping. 00439 * 00440 * @since New in 1.1. 00441 */ 00442 const char *svn_path_uri_autoescape(const char *uri, 00443 apr_pool_t *pool); 00444 00445 /** @} */ 00446 00447 /** Charset conversion stuff 00448 * 00449 * @defgroup svn_path_charset_stuff Charset conversion stuff 00450 * @{ 00451 */ 00452 00453 /** Convert @a path_utf8 from UTF-8 to the internal encoding used by APR. */ 00454 svn_error_t *svn_path_cstring_from_utf8(const char **path_apr, 00455 const char *path_utf8, 00456 apr_pool_t *pool); 00457 00458 /** Convert @a path_apr from the internal encoding used by APR to UTF-8. */ 00459 svn_error_t *svn_path_cstring_to_utf8(const char **path_utf8, 00460 const char *path_apr, 00461 apr_pool_t *pool); 00462 00463 00464 /** @} */ 00465 00466 #ifdef __cplusplus 00467 } 00468 #endif /* __cplusplus */ 00469 00470 00471 #endif /* SVN_PATH_H */