2 This file is part of GNUnet.
3 (C) 2009, 2011 Christian Grothoff (and other contributing authors)
5 GNUnet is free software; you can redistribute it and/or modify
6 it under the terms of the GNU General Public License as published
7 by the Free Software Foundation; either version 3, or (at your
8 option) any later version.
10 GNUnet is distributed in the hope that it will be useful, but
11 WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 General Public License for more details.
15 You should have received a copy of the GNU General Public License
16 along with GNUnet; see the file COPYING. If not, write to the
17 Free Software Foundation, Inc., 59 Temple Place - Suite 330,
18 Boston, MA 02111-1307, USA.
22 * @file fs/fs_file_information.c
23 * @brief Manage information for publishing directory hierarchies
24 * @author Christian Grothoff
27 * - metadata filename clean up code
28 * - metadata/ksk generation for directories from contained files
31 #include <extractor.h>
32 #include "gnunet_fs_service.h"
38 * Add meta data that libextractor finds to our meta data
41 * @param cls closure, our meta data container
42 * @param plugin_name name of the plugin that produced this value;
43 * special values can be used (i.e. '<zlib>' for zlib being
44 * used in the main libextractor library and yielding
46 * @param type libextractor-type describing the meta data
47 * @param format basic format information about data
48 * @param data_mime_type mime-type of data (not of the original file);
49 * can be NULL (if mime-type is not known)
50 * @param data actual meta-data found
51 * @param data_len number of bytes in data
52 * @return always 0 to continue extracting
56 const char *plugin_name,
57 enum EXTRACTOR_MetaType type,
58 enum EXTRACTOR_MetaFormat format,
59 const char *data_mime_type,
63 struct GNUNET_CONTAINER_MetaData *md = cls;
64 (void) GNUNET_CONTAINER_meta_data_insert (md,
76 * Extract meta-data from a file.
78 * @return GNUNET_SYSERR on error, otherwise the number
79 * of meta-data items obtained
82 GNUNET_FS_meta_data_extract_from_file (struct GNUNET_CONTAINER_MetaData
83 *md, const char *filename,
84 struct EXTRACTOR_PluginList *
91 if (extractors == NULL)
93 old = GNUNET_CONTAINER_meta_data_iterate (md, NULL, NULL);
94 GNUNET_assert (old >= 0);
95 EXTRACTOR_extract (extractors,
100 return (GNUNET_CONTAINER_meta_data_iterate (md, NULL, NULL) - old);
106 * Obtain the name under which this file information
107 * structure is stored on disk. Only works for top-level
108 * file information structures.
110 * @param s structure to get the filename for
111 * @return NULL on error, otherwise filename that
112 * can be passed to "GNUNET_FS_file_information_recover"
113 * to read this fi-struct from disk.
116 GNUNET_FS_file_information_get_id (struct GNUNET_FS_FileInformation *s)
120 return s->serialization;
125 * Create an entry for a file in a publish-structure.
127 * @param h handle to the file sharing subsystem
128 * @param client_info initial value for the client-info value for this entry
129 * @param filename name of the file or directory to publish
130 * @param keywords under which keywords should this file be available
131 * directly; can be NULL
132 * @param meta metadata for the file
133 * @param do_index GNUNET_YES for index, GNUNET_NO for insertion,
134 * GNUNET_SYSERR for simulation
135 * @param bo block options
136 * @return publish structure entry for the file
138 struct GNUNET_FS_FileInformation *
139 GNUNET_FS_file_information_create_from_file (struct GNUNET_FS_Handle *h,
141 const char *filename,
142 const struct GNUNET_FS_Uri *keywords,
143 const struct GNUNET_CONTAINER_MetaData *meta,
145 const struct GNUNET_FS_BlockOptions *bo)
149 struct GNUNET_FS_FileInformation *ret;
153 char fn_conv[MAX_PATH];
156 if (0 != STAT (filename, &sbuf))
158 GNUNET_log_strerror_file (GNUNET_ERROR_TYPE_WARNING,
163 fi = GNUNET_FS_make_file_reader_context_ (filename);
169 ret = GNUNET_FS_file_information_create_from_reader (h,
172 &GNUNET_FS_data_reader_file_,
181 ret->filename = GNUNET_strdup (filename);
185 plibc_conv_to_win_path (filename, fn_conv);
188 while (NULL != (ss = strstr (fn,
191 GNUNET_CONTAINER_meta_data_insert (ret->meta,
193 EXTRACTOR_METATYPE_GNUNET_ORIGINAL_FILENAME,
194 EXTRACTOR_METAFORMAT_C_STRING,
203 * Create an entry for a file in a publish-structure.
205 * @param h handle to the file sharing subsystem
206 * @param client_info initial value for the client-info value for this entry
207 * @param length length of the file
208 * @param data data for the file (should not be used afterwards by
209 * the caller; callee will "free")
210 * @param keywords under which keywords should this file be available
211 * directly; can be NULL
212 * @param meta metadata for the file
213 * @param do_index GNUNET_YES for index, GNUNET_NO for insertion,
214 * GNUNET_SYSERR for simulation
215 * @param bo block options
216 * @return publish structure entry for the file
218 struct GNUNET_FS_FileInformation *
219 GNUNET_FS_file_information_create_from_data (struct GNUNET_FS_Handle *h,
223 const struct GNUNET_FS_Uri *keywords,
224 const struct GNUNET_CONTAINER_MetaData *meta,
226 const struct GNUNET_FS_BlockOptions *bo)
228 if (GNUNET_YES == do_index)
233 return GNUNET_FS_file_information_create_from_reader (h,
236 &GNUNET_FS_data_reader_copy_,
246 * Create an entry for a file in a publish-structure.
248 * @param h handle to the file sharing subsystem
249 * @param client_info initial value for the client-info value for this entry
250 * @param length length of the file
251 * @param reader function that can be used to obtain the data for the file
252 * @param reader_cls closure for "reader"
253 * @param keywords under which keywords should this file be available
254 * directly; can be NULL
255 * @param meta metadata for the file
256 * @param do_index GNUNET_YES for index, GNUNET_NO for insertion,
257 * GNUNET_SYSERR for simulation
258 * @param bo block options
259 * @return publish structure entry for the file
261 struct GNUNET_FS_FileInformation *
262 GNUNET_FS_file_information_create_from_reader (struct GNUNET_FS_Handle *h,
265 GNUNET_FS_DataReader reader,
267 const struct GNUNET_FS_Uri *keywords,
268 const struct GNUNET_CONTAINER_MetaData *meta,
270 const struct GNUNET_FS_BlockOptions *bo)
272 struct GNUNET_FS_FileInformation *ret;
274 if ( (GNUNET_YES == do_index) &&
275 (reader != &GNUNET_FS_data_reader_file_) )
280 ret = GNUNET_malloc (sizeof (struct GNUNET_FS_FileInformation));
282 ret->client_info = client_info;
283 ret->meta = GNUNET_CONTAINER_meta_data_duplicate (meta);
284 if (ret->meta == NULL)
285 ret->meta = GNUNET_CONTAINER_meta_data_create ();
286 ret->keywords = (keywords == NULL) ? NULL : GNUNET_FS_uri_dup (keywords);
287 ret->data.file.reader = reader;
288 ret->data.file.reader_cls = reader_cls;
289 ret->data.file.do_index = do_index;
290 ret->data.file.file_size = length;
297 * Closure for "dir_scan_cb".
302 * Metadata extractors to use.
304 struct EXTRACTOR_PluginList *extractors;
309 struct GNUNET_FS_Handle *h;
312 * Function to call on each directory entry.
314 GNUNET_FS_FileProcessor proc;
322 * Scanner to use for subdirectories.
324 GNUNET_FS_DirectoryScanner scanner;
327 * Closure for scanner.
332 * Set to an error message (if any).
339 const struct GNUNET_FS_BlockOptions *bo;
342 * Should files be indexed?
350 * Function called on each entry in a file to
351 * cause default-publishing.
352 * @param cls closure (struct DirScanCls)
353 * @param filename name of the file to be published
354 * @return GNUNET_OK on success, GNUNET_SYSERR to abort
357 dir_scan_cb (void *cls,
358 const char *filename)
360 struct DirScanCls *dsc = cls;
362 struct GNUNET_FS_FileInformation *fi;
363 struct GNUNET_FS_Uri *ksk_uri;
364 struct GNUNET_FS_Uri *keywords;
365 struct GNUNET_CONTAINER_MetaData *meta;
367 if (0 != STAT (filename, &sbuf))
369 GNUNET_asprintf (&dsc->emsg,
370 _("`%s' failed on file `%s': %s"),
374 return GNUNET_SYSERR;
376 if (S_ISDIR (sbuf.st_mode))
378 fi = GNUNET_FS_file_information_create_from_directory (dsc->h,
388 GNUNET_assert (NULL != dsc->emsg);
389 return GNUNET_SYSERR;
394 meta = GNUNET_CONTAINER_meta_data_create ();
395 GNUNET_FS_meta_data_extract_from_file (meta,
398 keywords = GNUNET_FS_uri_ksk_create_from_meta_data (meta);
399 ksk_uri = GNUNET_FS_uri_ksk_canonicalize (keywords);
400 fi = GNUNET_FS_file_information_create_from_file (dsc->h,
407 GNUNET_CONTAINER_meta_data_destroy (meta);
408 GNUNET_FS_uri_destroy (keywords);
409 GNUNET_FS_uri_destroy (ksk_uri);
411 dsc->proc (dsc->proc_cls,
419 * Simple, useful default implementation of a directory scanner
420 * (GNUNET_FS_DirectoryScanner). This implementation expects to get a
421 * UNIX filename, will publish all files in the directory except hidden
422 * files (those starting with a "."). Metadata will be extracted
423 * using GNU libextractor; the specific list of plugins should be
424 * specified in "cls", passing NULL will disable (!) metadata
425 * extraction. Keywords will be derived from the metadata and be
426 * subject to default canonicalization. This is strictly a
427 * convenience function.
429 * @param cls must be of type "struct EXTRACTOR_Extractor*"
430 * @param h handle to the file sharing subsystem
431 * @param dirname name of the directory to scan
432 * @param do_index should files be indexed or inserted
433 * @param bo block options
434 * @param proc function called on each entry
435 * @param proc_cls closure for proc
436 * @param emsg where to store an error message (on errors)
437 * @return GNUNET_OK on success
440 GNUNET_FS_directory_scanner_default (void *cls,
441 struct GNUNET_FS_Handle *h,
444 const struct GNUNET_FS_BlockOptions *bo,
445 GNUNET_FS_FileProcessor proc,
449 struct EXTRACTOR_PluginList *ex = cls;
450 struct DirScanCls dsc;
455 dsc.proc_cls = proc_cls;
456 dsc.scanner = &GNUNET_FS_directory_scanner_default;
457 dsc.scanner_cls = cls;
458 dsc.do_index = do_index;
460 if (-1 == GNUNET_DISK_directory_scan (dirname,
464 GNUNET_assert (NULL != dsc.emsg);
466 return GNUNET_SYSERR;
473 * Closure for dirproc function.
478 * Linked list of directory entries that is being
481 struct GNUNET_FS_FileInformation *entries;
487 * Function that processes a directory entry that
488 * was obtained from the scanner.
489 * @param cls our closure
490 * @param filename name of the file (unused, why there???)
491 * @param fi information for publishing the file
495 const char *filename,
496 struct GNUNET_FS_FileInformation *fi)
498 struct EntryProcCls *dc = cls;
500 GNUNET_assert (fi->next == NULL);
501 GNUNET_assert (fi->dir == NULL);
502 fi->next = dc->entries;
508 * Create a publish-structure from an existing file hierarchy, inferring
509 * and organizing keywords and metadata as much as possible. This
510 * function primarily performs the recursive build and re-organizes
511 * keywords and metadata; for automatically getting metadata
512 * extraction, scanning of directories and creation of the respective
513 * GNUNET_FS_FileInformation entries the default scanner should be
514 * passed (GNUNET_FS_directory_scanner_default). This is strictly a
515 * convenience function.
517 * @param h handle to the file sharing subsystem
518 * @param client_info initial value for the client-info value for this entry
519 * @param filename name of the top-level file or directory
520 * @param scanner function used to get a list of files in a directory
521 * @param scanner_cls closure for scanner
522 * @param do_index should files in the hierarchy be indexed?
523 * @param bo block options
524 * @param emsg where to store an error message
525 * @return publish structure entry for the directory, NULL on error
527 struct GNUNET_FS_FileInformation *
528 GNUNET_FS_file_information_create_from_directory (struct GNUNET_FS_Handle *h,
530 const char *filename,
531 GNUNET_FS_DirectoryScanner scanner,
534 const struct GNUNET_FS_BlockOptions *bo,
537 struct GNUNET_FS_FileInformation *ret;
538 struct EntryProcCls dc;
539 struct GNUNET_FS_Uri *ksk;
540 struct GNUNET_CONTAINER_MetaData *meta;
546 meta = GNUNET_CONTAINER_meta_data_create ();
547 GNUNET_FS_meta_data_make_directory (meta);
548 scanner (scanner_cls,
556 ksk = NULL; // FIXME...
557 // FIXME: create meta!
558 ret = GNUNET_FS_file_information_create_empty_directory (h,
563 GNUNET_CONTAINER_meta_data_destroy (meta);
564 ret->data.dir.entries = dc.entries;
565 while (dc.entries != NULL)
567 dc.entries->dir = ret;
568 dc.entries = dc.entries->next;
571 while ( (NULL != (ss = strstr (fn,
572 DIR_SEPARATOR_STR))) &&
575 GNUNET_asprintf (&dn,
578 GNUNET_CONTAINER_meta_data_insert (ret->meta,
580 EXTRACTOR_METATYPE_GNUNET_ORIGINAL_FILENAME,
581 EXTRACTOR_METAFORMAT_C_STRING,
586 ret->filename = GNUNET_strdup (filename);
592 * Test if a given entry represents a directory.
594 * @param ent check if this FI represents a directory
595 * @return GNUNET_YES if so, GNUNET_NO if not
598 GNUNET_FS_file_information_is_directory (const struct GNUNET_FS_FileInformation *ent)
600 return ent->is_directory;
605 * Create an entry for an empty directory in a publish-structure.
606 * This function should be used by applications for which the
607 * use of "GNUNET_FS_file_information_create_from_directory"
608 * is not appropriate.
610 * @param h handle to the file sharing subsystem
611 * @param client_info initial value for the client-info value for this entry
612 * @param meta metadata for the directory
613 * @param keywords under which keywords should this directory be available
614 * directly; can be NULL
615 * @param bo block options
616 * @return publish structure entry for the directory , NULL on error
618 struct GNUNET_FS_FileInformation *
619 GNUNET_FS_file_information_create_empty_directory (struct GNUNET_FS_Handle *h,
621 const struct GNUNET_FS_Uri *keywords,
622 const struct GNUNET_CONTAINER_MetaData *meta,
623 const struct GNUNET_FS_BlockOptions *bo)
625 struct GNUNET_FS_FileInformation *ret;
627 ret = GNUNET_malloc (sizeof (struct GNUNET_FS_FileInformation));
629 ret->client_info = client_info;
630 ret->meta = GNUNET_CONTAINER_meta_data_duplicate (meta);
631 ret->keywords = GNUNET_FS_uri_dup (keywords);
633 ret->is_directory = GNUNET_YES;
639 * Add an entry to a directory in a publish-structure. Clients
640 * should never modify publish structures that were passed to
641 * "GNUNET_FS_publish_start" already.
643 * @param dir the directory
644 * @param ent the entry to add; the entry must not have been
645 * added to any other directory at this point and
646 * must not include "dir" in its structure
647 * @return GNUNET_OK on success, GNUNET_SYSERR on error
650 GNUNET_FS_file_information_add (struct GNUNET_FS_FileInformation *dir,
651 struct GNUNET_FS_FileInformation *ent)
653 if ( (ent->dir != NULL) ||
654 (ent->next != NULL) ||
655 (! dir->is_directory) )
658 return GNUNET_SYSERR;
661 ent->next = dir->data.dir.entries;
662 dir->data.dir.entries = ent;
663 dir->data.dir.dir_size = 0;
669 * Inspect a file or directory in a publish-structure. Clients
670 * should never modify publish structures that were passed to
671 * "GNUNET_FS_publish_start" already. When called on a directory,
672 * this function will FIRST call "proc" with information about
673 * the directory itself and then for each of the files in the
674 * directory (but not for files in subdirectories). When called
675 * on a file, "proc" will be called exactly once (with information
676 * about the specific file).
678 * @param dir the directory
679 * @param proc function to call on each entry
680 * @param proc_cls closure for proc
683 GNUNET_FS_file_information_inspect (struct GNUNET_FS_FileInformation *dir,
684 GNUNET_FS_FileInformationProcessor proc,
687 struct GNUNET_FS_FileInformation *pos;
694 (dir->is_directory) ? dir->data.dir.dir_size : dir->data.file.file_size,
698 (dir->is_directory) ? &no : &dir->data.file.do_index,
701 if (! dir->is_directory)
703 pos = dir->data.dir.entries;
710 (pos->is_directory) ? pos->data.dir.dir_size : pos->data.file.file_size,
714 (dir->is_directory) ? &no : &dir->data.file.do_index,
723 * Destroy publish-structure. Clients should never destroy publish
724 * structures that were passed to "GNUNET_FS_publish_start" already.
726 * @param fi structure to destroy
727 * @param cleaner function to call on each entry in the structure
728 * (useful to clean up client_info); can be NULL; return
730 * @param cleaner_cls closure for cleaner
733 GNUNET_FS_file_information_destroy (struct GNUNET_FS_FileInformation *fi,
734 GNUNET_FS_FileInformationProcessor cleaner,
737 struct GNUNET_FS_FileInformation *pos;
741 if (fi->is_directory)
743 /* clean up directory */
744 while (NULL != (pos = fi->data.dir.entries))
746 fi->data.dir.entries = pos->next;
747 GNUNET_FS_file_information_destroy (pos, cleaner, cleaner_cls);
749 /* clean up client-info */
751 cleaner (cleaner_cls,
753 fi->data.dir.dir_size,
759 GNUNET_free_non_null (fi->data.dir.dir_data);
763 /* call clean-up function of the reader */
764 if (fi->data.file.reader != NULL)
765 fi->data.file.reader (fi->data.file.reader_cls, 0, 0,
767 /* clean up client-info */
769 cleaner (cleaner_cls,
771 fi->data.file.file_size,
775 &fi->data.file.do_index,
778 GNUNET_free_non_null (fi->filename);
779 GNUNET_free_non_null (fi->emsg);
780 GNUNET_free_non_null (fi->chk_uri);
781 /* clean up serialization */
782 if ( (NULL != fi->serialization) &&
783 (0 != UNLINK (fi->serialization)) )
784 GNUNET_log_strerror_file (GNUNET_ERROR_TYPE_WARNING,
787 if (NULL != fi->keywords)
788 GNUNET_FS_uri_destroy (fi->keywords);
789 if (NULL != fi->meta)
790 GNUNET_CONTAINER_meta_data_destroy (fi->meta);
791 GNUNET_free_non_null (fi->serialization);
794 GNUNET_FS_tree_encoder_finish (fi->te,
802 /* end of fs_file_information.c */