1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006, 2009, 2010, 2011, 2012, 2013 Free Software Foundation, Inc.
4 This program is free software: you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation, either version 3 of the License, or
7 (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program. If not, see <http://www.gnu.org/licenses/>. */
19 #include "data/file-handle-def.h"
26 #include "data/dataset.h"
27 #include "data/variable.h"
28 #include "libpspp/assertion.h"
29 #include "libpspp/cast.h"
30 #include "libpspp/compiler.h"
31 #include "libpspp/hash-functions.h"
32 #include "libpspp/hmap.h"
33 #include "libpspp/i18n.h"
34 #include "libpspp/message.h"
35 #include "libpspp/str.h"
39 #include "gl/dirname.h"
40 #include "gl/xalloc.h"
43 #define _(msgid) gettext (msgid)
45 #if defined _WIN32 || defined __WIN32__
46 #define WIN32_LEAN_AND_MEAN /* avoid including junk */
53 struct hmap_node name_node; /* Element in named_handles hmap. */
54 size_t ref_cnt; /* Number of references. */
55 char *id; /* Identifier token, NULL if none. */
56 char *name; /* User-friendly identifying name. */
57 enum fh_referent referent; /* What the file handle refers to. */
59 /* FH_REF_FILE only. */
60 char *file_name; /* File name as provided by user. */
61 char *file_name_encoding; /* The character encoding of file_name,
62 This is NOT the encoding of the file contents! */
63 enum fh_mode mode; /* File mode. */
64 enum fh_line_ends line_ends; /* Line ends for text files. */
66 /* FH_REF_FILE and FH_REF_INLINE only. */
67 size_t record_width; /* Length of fixed-format records. */
68 size_t tab_width; /* Tab width, 0=do not expand tabs. */
69 char *encoding; /* Charset for contents. */
71 /* FH_REF_DATASET only. */
72 struct dataset *ds; /* Dataset. */
75 /* All "struct file_handle"s with nonnull 'id' member. */
76 static struct hmap named_handles = HMAP_INITIALIZER (named_handles);
78 /* Default file handle for DATA LIST, REREAD, REPEATING DATA
80 static struct file_handle *default_handle;
82 /* The "file" that reads from BEGIN DATA...END DATA. */
83 static struct file_handle *inline_file;
85 static struct file_handle *create_handle (const char *id,
86 char *name, enum fh_referent,
87 const char *encoding);
88 static void free_handle (struct file_handle *);
89 static void unname_handle (struct file_handle *);
91 /* Hash table of all active locks. */
92 static struct hmap locks = HMAP_INITIALIZER (locks);
94 static struct file_identity *fh_get_identity (const struct file_handle *);
95 static void fh_free_identity (struct file_identity *);
96 static int fh_compare_file_identities (const struct file_identity *,
97 const struct file_identity *);
98 static unsigned int fh_hash_identity (const struct file_identity *);
100 /* File handle initialization routine. */
104 inline_file = create_handle ("INLINE", xstrdup ("INLINE"), FH_REF_INLINE,
106 inline_file->record_width = 80;
107 inline_file->tab_width = 8;
110 /* Removes all named file handles from the global list. */
114 struct file_handle *handle, *next;
116 HMAP_FOR_EACH_SAFE (handle, next,
117 struct file_handle, name_node, &named_handles)
118 unname_handle (handle);
120 free_handle (inline_file);
123 /* Free HANDLE and remove it from the global list. */
125 free_handle (struct file_handle *handle)
130 /* Remove handle from global list. */
131 if (handle->id != NULL)
132 hmap_delete (&named_handles, &handle->name_node);
137 free (handle->file_name);
138 free (handle->file_name_encoding);
139 free (handle->encoding);
143 /* Make HANDLE unnamed, so that it can no longer be referenced by
144 name. The caller must hold a reference to HANDLE, which is
145 not affected by this function. */
147 unname_handle (struct file_handle *handle)
149 assert (handle->id != NULL);
152 hmap_delete (&named_handles, &handle->name_node);
154 /* Drop the reference held by the named_handles table. */
158 /* Increments HANDLE's reference count and returns HANDLE. */
160 fh_ref (struct file_handle *handle)
162 if (handle == fh_inline_file ())
164 assert (handle->ref_cnt > 0);
169 /* Decrements HANDLE's reference count.
170 If the reference count drops to 0, HANDLE is destroyed. */
172 fh_unref (struct file_handle *handle)
176 if (handle == fh_inline_file ())
178 assert (handle->ref_cnt > 0);
179 if (--handle->ref_cnt == 0)
180 free_handle (handle);
184 /* Make HANDLE unnamed, so that it can no longer be referenced by
185 name. The caller must hold a reference to HANDLE, which is
186 not affected by this function.
188 This function ignores a null pointer as input. It has no
189 effect on the inline handle, which is always named INLINE.*/
191 fh_unname (struct file_handle *handle)
193 assert (handle->ref_cnt > 1);
194 if (handle != fh_inline_file () && handle->id != NULL)
195 unname_handle (handle);
198 /* Returns the handle with the given ID, or a null pointer if
201 fh_from_id (const char *id)
203 struct file_handle *handle;
205 HMAP_FOR_EACH_WITH_HASH (handle, struct file_handle, name_node,
206 utf8_hash_case_string (id, 0), &named_handles)
207 if (!utf8_strcasecmp (id, handle->id))
208 return fh_ref (handle);
213 /* Creates a new handle with identifier ID (which may be null)
214 and name HANDLE_NAME that refers to REFERENT. Links the new
215 handle into the global list. Returns the new handle.
217 The new handle is not fully initialized. The caller is
218 responsible for completing its initialization. */
219 static struct file_handle *
220 create_handle (const char *id, char *handle_name, enum fh_referent referent,
221 const char *encoding)
223 struct file_handle *handle = xmalloc (sizeof *handle);
224 *handle = (struct file_handle) {
226 .id = xstrdup_if_nonnull (id),
228 .referent = referent,
229 .encoding = xstrdup (encoding),
233 hmap_insert (&named_handles, &handle->name_node,
234 utf8_hash_case_string (handle->id, 0));
239 /* Returns the unique handle of referent type FH_REF_INLINE,
240 which refers to the "inline file" that represents character
241 data in the command file between BEGIN DATA and END DATA. */
243 fh_inline_file (void)
248 /* Creates and returns a new file handle with the given ID, which may be null.
249 If it is non-null, it must be a UTF-8 encoded string that is unique among
250 existing file identifiers. The new handle is associated with file FILE_NAME
251 and the given PROPERTIES. */
253 fh_create_file (const char *id, const char *file_name, const char *file_name_encoding,
254 const struct fh_properties *properties)
256 char *handle_name = id ? xstrdup (id) : xasprintf ("`%s'", file_name);
257 struct file_handle *handle = create_handle (id, handle_name, FH_REF_FILE,
258 properties->encoding);
259 handle->file_name = xstrdup (file_name);
260 handle->file_name_encoding = xstrdup_if_nonnull (file_name_encoding);
261 handle->mode = properties->mode;
262 handle->line_ends = properties->line_ends;
263 handle->record_width = properties->record_width;
264 handle->tab_width = properties->tab_width;
268 /* Creates a new file handle with the given ID, which must be
269 unique among existing file identifiers. The new handle is
270 associated with a dataset file (initially empty). */
272 fh_create_dataset (struct dataset *ds)
275 struct file_handle *handle;
277 name = dataset_name (ds);
279 name = _("active dataset");
281 handle = create_handle (NULL, xstrdup (name), FH_REF_DATASET, C_ENCODING);
286 /* Returns a set of default properties for a file handle. */
287 const struct fh_properties *
288 fh_default_properties (void)
290 #if defined _WIN32 || defined __WIN32__
291 #define DEFAULT_LINE_ENDS FH_END_CRLF
293 #define DEFAULT_LINE_ENDS FH_END_LF
296 static const struct fh_properties default_properties
297 = {FH_MODE_TEXT, DEFAULT_LINE_ENDS, 1024, 4, (char *) "Auto"};
298 return &default_properties;
301 /* Returns the identifier that may be used in syntax to name the
302 given HANDLE, which takes the form of a PSPP identifier. If
303 HANDLE has no identifier, returns a null pointer.
305 Return value is owned by the file handle.*/
307 fh_get_id (const struct file_handle *handle)
312 /* Returns a user-friendly string to identify the given HANDLE.
313 If HANDLE was created by referring to a file name, returns the
314 file name, enclosed in double quotes. Return value is owned
317 Useful for printing error messages about use of file handles. */
319 fh_get_name (const struct file_handle *handle)
324 /* Returns the type of object that HANDLE refers to. */
326 fh_get_referent (const struct file_handle *handle)
328 return handle->referent;
331 /* Returns the name of the file associated with HANDLE. */
333 fh_get_file_name (const struct file_handle *handle)
335 assert (handle->referent == FH_REF_FILE);
336 return handle->file_name;
340 /* Returns the character encoding of the name of the file associated with HANDLE. */
342 fh_get_file_name_encoding (const struct file_handle *handle)
344 assert (handle->referent == FH_REF_FILE);
345 return handle->file_name_encoding;
349 /* Returns the mode of HANDLE. */
351 fh_get_mode (const struct file_handle *handle)
353 assert (handle->referent == FH_REF_FILE);
357 /* Returns the line ends of HANDLE, which must be a handle associated with a
360 fh_get_line_ends (const struct file_handle *handle)
362 assert (handle->referent == FH_REF_FILE);
363 return handle->line_ends;
366 /* Returns the width of a logical record on HANDLE. */
368 fh_get_record_width (const struct file_handle *handle)
370 assert (handle->referent & (FH_REF_FILE | FH_REF_INLINE));
371 return handle->record_width;
374 /* Returns the number of characters per tab stop for HANDLE, or
375 zero if tabs are not to be expanded. Applicable only to
376 FH_MODE_TEXT files. */
378 fh_get_tab_width (const struct file_handle *handle)
380 assert (handle->referent & (FH_REF_FILE | FH_REF_INLINE));
381 return handle->tab_width;
384 /* Returns the encoding of characters read from HANDLE. */
386 fh_get_encoding (const struct file_handle *handle)
388 return handle->encoding;
391 /* Returns true if A and B refer to the same file or dataset, false
394 fh_equal (const struct file_handle *a, const struct file_handle *b)
396 if (a->referent != b->referent)
403 struct file_identity *a_id = fh_get_identity (a);
404 struct file_identity *b_id = fh_get_identity (b);
406 int cmp = fh_compare_file_identities (a_id, b_id);
408 fh_free_identity (a_id);
409 fh_free_identity (b_id);
418 return a->ds == b->ds;
425 /* Returns the dataset handle associated with HANDLE.
426 Applicable to only FH_REF_DATASET files. */
428 fh_get_dataset (const struct file_handle *handle)
430 assert (handle->referent == FH_REF_DATASET);
434 /* Returns the current default handle. */
436 fh_get_default_handle (void)
438 return default_handle ? default_handle : fh_inline_file ();
441 /* Sets NEW_DEFAULT_HANDLE as the default handle. */
443 fh_set_default_handle (struct file_handle *new_default_handle)
445 assert (new_default_handle == NULL
446 || (new_default_handle->referent & (FH_REF_INLINE | FH_REF_FILE)));
447 if (default_handle != NULL && default_handle != inline_file)
448 fh_unref (default_handle);
449 default_handle = new_default_handle;
450 if (default_handle != NULL)
451 default_handle = fh_ref (default_handle);
454 /* Information about a file handle's readers or writers. */
457 struct hmap_node node; /* hmap_node member. */
460 enum fh_referent referent; /* Type of underlying file. */
463 struct file_identity *file; /* FH_REF_FILE only. */
464 unsigned int unique_id; /* FH_REF_DATASET only. */
467 enum fh_access access; /* Type of file access. */
469 /* Number of openers. */
472 /* Applicable only when open_cnt > 0. */
473 bool exclusive; /* No other openers allowed? */
474 const char *type; /* Human-readable type of file. */
475 void *aux; /* Owner's auxiliary data. */
479 static void make_key (struct fh_lock *, const struct file_handle *,
481 static void free_key (struct fh_lock *);
482 static int compare_fh_locks (const struct fh_lock *a, const struct fh_lock *b);
483 static unsigned int hash_fh_lock (const struct fh_lock *lock);
485 /* Tries to lock handle H for the given kind of ACCESS and TYPE
486 of file. Returns a pointer to a struct fh_lock if successful,
487 otherwise a null pointer.
489 H's referent type must be one of the bits in MASK. The caller
490 must verify this ahead of time; we simply assert it here.
492 TYPE is the sort of file, e.g. "system file". Only one type
493 of access is allowed on a given file at a time for reading,
494 and similarly for writing. If successful, a reference to TYPE
495 is retained, so it should probably be a string literal.
497 TYPE should be marked with N_() in the caller: that is, the
498 caller should not translate it with gettext, but fh_lock will
501 ACCESS specifies whether the lock is for reading or writing.
502 EXCLUSIVE is true to require exclusive access, false to allow
503 sharing with other accessors. Exclusive read access precludes
504 other readers, but not writers; exclusive write access
505 precludes other writers, but not readers. A sharable read or
506 write lock precludes reader or writers, respectively, of a
509 A lock may be associated with auxiliary data. See
510 fh_lock_get_aux and fh_lock_set_aux for more details. */
512 fh_lock (struct file_handle *h, enum fh_referent mask UNUSED,
513 const char *type, enum fh_access access, bool exclusive)
515 struct fh_lock *key = NULL;
517 struct fh_lock *lock = NULL;
518 bool found_lock = false;
520 assert ((fh_get_referent (h) & mask) != 0);
521 assert (access == FH_ACC_READ || access == FH_ACC_WRITE);
523 key = xmalloc (sizeof *key);
525 make_key (key, h, access);
528 key->exclusive = exclusive;
532 hash = hash_fh_lock (key);
534 HMAP_FOR_EACH_WITH_HASH (lock, struct fh_lock, node, hash, &locks)
536 if (0 == compare_fh_locks (lock, key))
545 if (strcmp (lock->type, type))
547 if (access == FH_ACC_READ)
548 msg (SE, _("Can't read from %s as a %s because it is "
549 "already being read as a %s."),
550 fh_get_name (h), gettext (type), gettext (lock->type));
552 msg (SE, _("Can't write to %s as a %s because it is "
553 "already being written as a %s."),
554 fh_get_name (h), gettext (type), gettext (lock->type));
557 else if (exclusive || lock->exclusive)
559 msg (SE, _("Can't re-open %s as a %s."),
560 fh_get_name (h), gettext (type));
571 hmap_insert (&locks, &key->node, hash);
573 HMAP_FOR_EACH_WITH_HASH (lock, struct fh_lock, node, hash, &locks)
575 if (0 == compare_fh_locks (lock, key))
587 /* Releases LOCK that was acquired with fh_lock.
588 Returns true if LOCK is still locked, because other clients
591 Returns false if LOCK has now been destroyed. In this case
592 the caller must ensure that any auxiliary data associated with
593 LOCK is destroyed, to avoid a memory leak. The caller must
594 obtain a pointer to the auxiliary data, e.g. via
595 fh_lock_get_aux *before* calling fh_unlock (because it yields
596 undefined behavior to call fh_lock_get_aux on a destroyed
599 fh_unlock (struct fh_lock *lock)
603 assert (lock->open_cnt > 0);
604 if (--lock->open_cnt == 0)
606 hmap_delete (&locks, &lock->node);
615 /* Returns auxiliary data for LOCK.
617 Auxiliary data is shared by every client that holds LOCK (for
618 an exclusive lock, this is a single client). To avoid leaks,
619 auxiliary data must be released before LOCK is destroyed. */
621 fh_lock_get_aux (const struct fh_lock *lock)
626 /* Sets the auxiliary data for LOCK to AUX. */
628 fh_lock_set_aux (struct fh_lock *lock, void *aux)
633 /* Returns true if HANDLE is locked for the given type of ACCESS,
636 fh_is_locked (const struct file_handle *handle, enum fh_access access)
639 const struct fh_lock *k = NULL;
640 bool is_locked = false;
643 make_key (&key, handle, access);
645 hash = hash_fh_lock (&key);
648 HMAP_FOR_EACH_WITH_HASH (k, struct fh_lock, node, hash, &locks)
650 if (0 == compare_fh_locks (k, &key))
662 /* Initializes the key fields in LOCK for looking up or inserting
663 handle H for the given kind of ACCESS. */
665 make_key (struct fh_lock *lock, const struct file_handle *h,
666 enum fh_access access)
668 lock->referent = fh_get_referent (h);
669 lock->access = access;
670 if (lock->referent == FH_REF_FILE)
671 lock->u.file = fh_get_identity (h);
672 else if (lock->referent == FH_REF_DATASET)
673 lock->u.unique_id = dataset_seqno (fh_get_dataset (h));
676 /* Frees the key fields in LOCK. */
678 free_key (struct fh_lock *lock)
680 if (lock->referent == FH_REF_FILE)
681 fh_free_identity (lock->u.file);
684 /* Compares the key fields in struct fh_lock objects A and B and
685 returns a strcmp()-type result. */
687 compare_fh_locks (const struct fh_lock *a, const struct fh_lock *b)
689 if (a->referent != b->referent)
690 return a->referent < b->referent ? -1 : 1;
691 else if (a->access != b->access)
692 return a->access < b->access ? -1 : 1;
693 else if (a->referent == FH_REF_FILE)
694 return fh_compare_file_identities (a->u.file, b->u.file);
695 else if (a->referent == FH_REF_DATASET)
696 return (a->u.unique_id < b->u.unique_id ? -1
697 : a->u.unique_id > b->u.unique_id);
702 /* Returns a hash value for LOCK. */
704 hash_fh_lock (const struct fh_lock *lock)
707 if (lock->referent == FH_REF_FILE)
708 basis = fh_hash_identity (lock->u.file);
709 else if (lock->referent == FH_REF_DATASET)
710 basis = lock->u.unique_id;
713 return hash_int ((lock->referent << 3) | lock->access, basis);
721 /* A file's identity:
723 - For a file that exists, this is its device and inode.
725 - For a file that does not exist, but which has a directory
726 name that exists, this is the device and inode of the
727 directory, plus the file's base name.
729 - For a file that does not exist and has a nonexistent
730 directory, this is the file name.
732 Windows doesn't have inode numbers, so we just use the name
736 unsigned long long device; /* Device number. */
737 unsigned long long inode; /* Inode number. */
738 char *name; /* File name, where needed, otherwise NULL. */
741 /* Returns a pointer to a dynamically allocated structure whose
742 value can be used to tell whether two files are actually the
743 same file. The caller is responsible for freeing the structure with
744 fh_free_identity() when finished. */
745 static struct file_identity *
746 fh_get_identity (const struct file_handle *fh)
748 struct file_identity *identity = xmalloc (sizeof *identity);
750 const char *file_name = fh_get_file_name (fh);
752 #if !(defined _WIN32 || defined __WIN32__)
754 if (lstat (file_name, &s) == 0)
756 identity->device = s.st_dev;
757 identity->inode = s.st_ino;
758 identity->name = NULL;
762 char *dir = dir_name (file_name);
763 if (last_component (file_name) != NULL && stat (dir, &s) == 0)
765 identity->device = s.st_dev;
766 identity->inode = s.st_ino;
767 identity->name = base_name (file_name);
771 identity->device = 0;
773 identity->name = xstrdup (file_name);
779 HANDLE h = CreateFile (file_name, GENERIC_READ, 0, NULL, OPEN_EXISTING, FILE_ATTRIBUTE_READONLY, NULL);
780 if (h != INVALID_HANDLE_VALUE)
782 BY_HANDLE_FILE_INFORMATION fi;
783 ok = GetFileInformationByHandle (h, &fi);
786 identity->device = fi.dwVolumeSerialNumber;
787 identity->inode = fi.nFileIndexHigh;
788 identity->inode <<= (sizeof fi.nFileIndexLow) * CHAR_BIT;
789 identity->inode |= fi.nFileIndexLow;
797 identity->device = 0;
801 size_t pathlen = 255;
806 cname = xrealloc (cname, bufsize);
807 pathlen = GetFullPathName (file_name, bufsize, cname, NULL);
809 while (pathlen > bufsize);
810 identity->name = xstrdup (cname);
812 str_lowercase (identity->name);
819 /* Frees IDENTITY obtained from fh_get_identity(). */
821 fh_free_identity (struct file_identity *identity)
823 if (identity != NULL)
825 free (identity->name);
830 /* Compares A and B, returning a strcmp()-type result. */
832 fh_compare_file_identities (const struct file_identity *a,
833 const struct file_identity *b)
835 if (a->device != b->device)
836 return a->device < b->device ? -1 : 1;
837 else if (a->inode != b->inode)
838 return a->inode < b->inode ? -1 : 1;
839 else if (a->name != NULL)
840 return b->name != NULL ? strcmp (a->name, b->name) : 1;
842 return b->name != NULL ? -1 : 0;
845 /* Returns a hash value for IDENTITY. */
847 fh_hash_identity (const struct file_identity *identity)
849 unsigned int hash = hash_int (identity->device, identity->inode);
850 if (identity->name != NULL)
851 hash = hash_string (identity->name, hash);