1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006, 2007, 2009, 2010, 2011, 2012, 2013, 2014,
3 2015, 2020 Free Software Foundation, Inc.
5 This program is free software: you can redistribute it and/or modify
6 it under the terms of the GNU General Public License as published by
7 the Free Software Foundation, either version 3 of the License, or
8 (at your option) any later version.
10 This program is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 GNU General Public License for more details.
15 You should have received a copy of the GNU General Public License
16 along with this program. If not, see <http://www.gnu.org/licenses/>. */
20 #include "data/dictionary.h"
27 #include "data/attributes.h"
28 #include "data/case.h"
29 #include "data/identifier.h"
30 #include "data/mrset.h"
31 #include "data/settings.h"
32 #include "data/value-labels.h"
33 #include "data/vardict.h"
34 #include "data/variable.h"
35 #include "data/vector.h"
36 #include "libpspp/array.h"
37 #include "libpspp/assertion.h"
38 #include "libpspp/compiler.h"
39 #include "libpspp/hash-functions.h"
40 #include "libpspp/hmap.h"
41 #include "libpspp/i18n.h"
42 #include "libpspp/message.h"
43 #include "libpspp/misc.h"
44 #include "libpspp/pool.h"
45 #include "libpspp/str.h"
46 #include "libpspp/string-array.h"
47 #include "libpspp/ll.h"
49 #include "gl/intprops.h"
50 #include "gl/minmax.h"
51 #include "gl/xalloc.h"
52 #include "gl/xmemdup0.h"
55 #define _(msgid) gettext (msgid)
61 struct vardict_info *vars; /* Variables. */
62 size_t n_vars; /* Number of variables. */
63 size_t allocated_vars; /* Allocated space in 'vars'. */
64 struct caseproto *proto; /* Prototype for dictionary cases
66 struct hmap name_map; /* Variable index by name. */
67 int next_value_idx; /* Index of next `union value' to allocate. */
68 const struct variable **split; /* SPLIT FILE vars. */
69 size_t n_splits; /* SPLIT FILE count. */
70 enum split_type split_type;
71 struct variable *weight; /* WEIGHT variable. */
72 struct variable *filter; /* FILTER variable. */
73 casenumber case_limit; /* Current case limit (N command). */
74 char *label; /* File label. */
75 struct string_array documents; /* Documents. */
76 struct vector **vector; /* Vectors of variables. */
77 size_t n_vectors; /* Number of vectors. */
78 struct attrset attributes; /* Custom attributes. */
79 struct mrset **mrsets; /* Multiple response sets. */
80 size_t n_mrsets; /* Number of multiple response sets. */
82 /* Whether variable names must be valid identifiers. Normally, this is
83 true, but sometimes a dictionary is prepared for external use
84 (e.g. output to a CSV file) where names don't have to be valid. */
85 bool names_must_be_ids;
87 char *encoding; /* Character encoding of string data */
89 const struct dict_callbacks *callbacks; /* Callbacks on dictionary
91 void *cb_data ; /* Data passed to callbacks */
93 void (*changed) (struct dictionary *, void *); /* Generic change callback */
97 static void dict_unset_split_var (struct dictionary *, struct variable *, bool);
98 static void dict_unset_mrset_var (struct dictionary *, struct variable *);
100 /* Compares two double pointers to variables, which should point
101 to elements of a struct dictionary's `var' member array. */
103 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
105 struct variable *const *a = a_;
106 struct variable *const *b = b_;
108 return *a < *b ? -1 : *a > *b;
112 unindex_var (struct dictionary *d, struct vardict_info *vardict)
114 hmap_delete (&d->name_map, &vardict->name_node);
117 /* This function assumes that vardict->name_node.hash is valid, that is, that
118 its name has not changed since it was hashed (rename_var() updates this
119 hash along with the name itself). */
121 reindex_var (struct dictionary *d, struct vardict_info *vardict, bool skip_callbacks)
123 struct variable *old = (d->callbacks && d->callbacks->var_changed
124 ? var_clone (vardict->var)
127 struct variable *var = vardict->var;
128 var_set_vardict (var, vardict);
129 hmap_insert_fast (&d->name_map, &vardict->name_node,
130 vardict->name_node.hash);
132 if (! skip_callbacks)
134 if (d->changed) d->changed (d, d->changed_data);
137 d->callbacks->var_changed (d, var_get_dict_index (var), VAR_TRAIT_POSITION, old, d->cb_data);
143 /* Sets the case_index in V's vardict to CASE_INDEX. */
145 set_var_case_index (struct variable *v, int case_index)
147 var_get_vardict (v)->case_index = case_index;
150 /* Removes the dictionary variables with indexes from FROM to TO (exclusive)
153 unindex_vars (struct dictionary *d, size_t from, size_t to)
157 for (i = from; i < to; i++)
158 unindex_var (d, &d->vars[i]);
161 /* Re-sets the dict_index in the dictionary variables with
162 indexes from FROM to TO (exclusive). */
164 reindex_vars (struct dictionary *d, size_t from, size_t to, bool skip_callbacks)
168 for (i = from; i < to; i++)
169 reindex_var (d, &d->vars[i], skip_callbacks);
174 /* Returns the encoding for data in dictionary D. The return value is a
175 nonnull string that contains an IANA character set name. */
177 dict_get_encoding (const struct dictionary *d)
182 /* Returns true if UTF-8 string ID is an acceptable identifier in DICT's
183 encoding, false otherwise. If ISSUE_ERROR is true, issues an explanatory
184 error message on failure. */
186 dict_id_is_valid (const struct dictionary *dict, const char *id,
189 return (!dict->names_must_be_ids
190 || id_is_valid (id, dict->encoding, issue_error));
194 dict_set_change_callback (struct dictionary *d,
195 void (*changed) (struct dictionary *, void*),
198 d->changed = changed;
199 d->changed_data = data;
202 /* Discards dictionary D's caseproto. (It will be regenerated
203 lazily, on demand.) */
205 invalidate_proto (struct dictionary *d)
207 caseproto_unref (d->proto);
211 /* Print a representation of dictionary D to stdout, for
212 debugging purposes. */
214 dict_dump (const struct dictionary *d)
217 for (i = 0 ; i < d->n_vars ; ++i)
219 const struct variable *v = d->vars[i].var;
220 printf ("Name: %s;\tdict_idx: %zu; case_idx: %zu\n",
222 var_get_dict_index (v),
223 var_get_case_index (v));
228 /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
229 the dictionary or any of the variables it contains are modified.
230 Each callback will get passed CALLBACK_DATA.
231 Any callback may be NULL, in which case it'll be ignored.
234 dict_set_callbacks (struct dictionary *dict,
235 const struct dict_callbacks *callbacks,
238 dict->callbacks = callbacks;
239 dict->cb_data = callback_data;
242 /* Shallow copy the callbacks from SRC to DEST */
244 dict_copy_callbacks (struct dictionary *dest,
245 const struct dictionary *src)
247 dest->callbacks = src->callbacks;
248 dest->cb_data = src->cb_data;
251 /* Creates and returns a new dictionary with the specified ENCODING. */
253 dict_create (const char *encoding)
255 struct dictionary *d = xmalloc (sizeof *d);
257 *d = (struct dictionary) {
258 .encoding = xstrdup (encoding),
259 .names_must_be_ids = true,
260 .name_map = HMAP_INITIALIZER (d->name_map),
261 .attributes = ATTRSET_INITIALIZER (d->attributes),
262 .split_type = SPLIT_LAYERED,
269 /* Creates and returns a (deep) copy of an existing
272 The new dictionary's case indexes are copied from the old
273 dictionary. If the new dictionary won't be used to access
274 cases produced with the old dictionary, then the new
275 dictionary's case indexes should be compacted with
276 dict_compact_values to save space.
278 Callbacks are not cloned. */
280 dict_clone (const struct dictionary *s)
282 struct dictionary *d;
285 d = dict_create (s->encoding);
286 dict_set_names_must_be_ids (d, dict_get_names_must_be_ids (s));
288 for (i = 0; i < s->n_vars; i++)
290 struct variable *sv = s->vars[i].var;
291 struct variable *dv = dict_clone_var_assert (d, sv);
294 for (i = 0; i < var_get_n_short_names (sv); i++)
295 var_set_short_name (dv, i, var_get_short_name (sv, i));
297 var_get_vardict (dv)->case_index = var_get_vardict (sv)->case_index;
300 d->next_value_idx = s->next_value_idx;
302 d->n_splits = s->n_splits;
305 d->split = xnmalloc (d->n_splits, sizeof *d->split);
306 for (i = 0; i < d->n_splits; i++)
307 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
309 d->split_type = s->split_type;
311 if (s->weight != NULL)
312 dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight)));
314 if (s->filter != NULL)
315 dict_set_filter (d, dict_lookup_var_assert (d, var_get_name (s->filter)));
317 d->case_limit = s->case_limit;
318 dict_set_label (d, dict_get_label (s));
319 dict_set_documents (d, dict_get_documents (s));
321 d->n_vectors = s->n_vectors;
322 d->vector = xnmalloc (d->n_vectors, sizeof *d->vector);
323 for (i = 0; i < s->n_vectors; i++)
324 d->vector[i] = vector_clone (s->vector[i], s, d);
326 dict_set_attributes (d, dict_get_attributes (s));
328 for (i = 0; i < s->n_mrsets; i++)
330 const struct mrset *old = s->mrsets[i];
334 /* Clone old mrset, then replace vars from D by vars from S. */
335 new = mrset_clone (old);
336 for (j = 0; j < new->n_vars; j++)
337 new->vars[j] = dict_lookup_var_assert (d, var_get_name (new->vars[j]));
339 dict_add_mrset (d, new);
347 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
348 dict_get_n_splits() to determine how many SPLIT FILE vars
349 there are. Returns a null pointer if and only if there are no
351 const struct variable *const *
352 dict_get_split_vars (const struct dictionary *d)
357 /* Returns the number of SPLIT FILE vars. */
359 dict_get_n_splits (const struct dictionary *d)
364 /* Removes variable V, which must be in D, from D's set of split
367 dict_unset_split_var (struct dictionary *d, struct variable *v, bool skip_callbacks)
371 assert (dict_contains_var (d, v));
373 orig_count = d->n_splits;
374 d->n_splits = remove_equal (d->split, d->n_splits, sizeof *d->split,
375 &v, compare_var_ptrs, NULL);
376 if (orig_count != d->n_splits && !skip_callbacks)
378 if (d->changed) d->changed (d, d->changed_data);
379 /* We changed the set of split variables so invoke the
381 if (d->callbacks && d->callbacks->split_changed)
382 d->callbacks->split_changed (d, d->cb_data);
387 /* Sets N split vars SPLIT in dictionary D. */
389 dict_set_split_vars__ (struct dictionary *d,
390 struct variable *const *split, size_t n,
391 enum split_type type, bool skip_callbacks)
393 assert (n == 0 || split != NULL);
396 d->split_type = type;
399 d->split = xnrealloc (d->split, n, sizeof *d->split) ;
400 memcpy (d->split, split, n * sizeof *d->split);
410 if (d->changed) d->changed (d, d->changed_data);
411 if (d->callbacks && d->callbacks->split_changed)
412 d->callbacks->split_changed (d, d->cb_data);
416 /* Sets N split vars SPLIT in dictionary D. */
418 dict_set_split_vars (struct dictionary *d,
419 struct variable *const *split, size_t n,
420 enum split_type type)
422 dict_set_split_vars__ (d, split, n, type, false);
426 dict_clear_split_vars (struct dictionary *d)
428 dict_set_split_vars (d, NULL, 0, SPLIT_LAYERED);
432 /* Deletes variable V from dictionary D and frees V.
434 This is a very bad idea if there might be any pointers to V
435 from outside D. In general, no variable in the active dataset's
436 dictionary should be deleted when any transformations are
437 active on the dictionary's dataset, because those
438 transformations might reference the deleted variable. The
439 safest time to delete a variable is just after a procedure has
440 been executed, as done by DELETE VARIABLES.
442 Pointers to V within D are not a problem, because
443 dict_delete_var() knows to remove V from split variables,
444 weights, filters, etc. */
446 dict_delete_var__ (struct dictionary *d, struct variable *v, bool skip_callbacks)
448 int dict_index = var_get_dict_index (v);
449 const int case_index = var_get_case_index (v);
451 assert (dict_contains_var (d, v));
453 dict_unset_split_var (d, v, skip_callbacks);
454 dict_unset_mrset_var (d, v);
457 dict_set_weight (d, NULL);
460 dict_set_filter (d, NULL);
462 dict_clear_vectors (d);
464 /* Remove V from var array. */
465 unindex_vars (d, dict_index, d->n_vars);
466 remove_element (d->vars, d->n_vars, sizeof *d->vars, dict_index);
469 /* Update dict_index for each affected variable. */
470 reindex_vars (d, dict_index, d->n_vars, skip_callbacks);
473 var_clear_vardict (v);
475 if (! skip_callbacks)
477 if (d->changed) d->changed (d, d->changed_data);
478 if (d->callbacks && d->callbacks->var_deleted)
479 d->callbacks->var_deleted (d, v, dict_index, case_index, d->cb_data);
482 invalidate_proto (d);
486 /* Deletes variable V from dictionary D and frees V.
488 This is a very bad idea if there might be any pointers to V
489 from outside D. In general, no variable in the active dataset's
490 dictionary should be deleted when any transformations are
491 active on the dictionary's dataset, because those
492 transformations might reference the deleted variable. The
493 safest time to delete a variable is just after a procedure has
494 been executed, as done by DELETE VARIABLES.
496 Pointers to V within D are not a problem, because
497 dict_delete_var() knows to remove V from split variables,
498 weights, filters, etc. */
500 dict_delete_var (struct dictionary *d, struct variable *v)
502 dict_delete_var__ (d, v, false);
506 /* Deletes the COUNT variables listed in VARS from D. This is
507 unsafe; see the comment on dict_delete_var() for details. */
509 dict_delete_vars (struct dictionary *d,
510 struct variable *const *vars, size_t count)
512 /* FIXME: this can be done in O(count) time, but this algorithm
514 assert (count == 0 || vars != NULL);
517 dict_delete_var (d, *vars++);
520 /* Deletes the COUNT variables in D starting at index IDX. This
521 is unsafe; see the comment on dict_delete_var() for
522 details. Deleting consecutive vars will result in less callbacks
523 compared to iterating over dict_delete_var.
524 A simple while loop over dict_delete_var will
525 produce (d->n_vars - IDX) * COUNT variable changed callbacks
526 plus COUNT variable delete callbacks.
527 This here produces d->n_vars - IDX variable changed callbacks
528 plus COUNT variable delete callbacks. */
530 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
532 assert (idx + count <= d->n_vars);
534 /* We need to store the variable and the corresponding case_index
535 for the delete callbacks later. We store them in a linked list.*/
538 struct variable *var;
541 struct ll_list list = LL_INITIALIZER (list);
543 for (size_t i = idx; i < idx + count; i++)
545 struct delvar *dv = xmalloc (sizeof (struct delvar));
547 struct variable *v = d->vars[i].var;
549 dict_unset_split_var (d, v, false);
550 dict_unset_mrset_var (d, v);
553 dict_set_weight (d, NULL);
556 dict_set_filter (d, NULL);
559 dv->case_index = var_get_case_index (v);
560 ll_push_tail (&list, (struct ll *)dv);
563 dict_clear_vectors (d);
565 /* Remove variables from var array. */
566 unindex_vars (d, idx, d->n_vars);
567 remove_range (d->vars, d->n_vars, sizeof *d->vars, idx, count);
570 /* Reindexing will result variable-changed callback */
571 reindex_vars (d, idx, d->n_vars, false);
573 invalidate_proto (d);
574 if (d->changed) d->changed (d, d->changed_data);
576 /* Now issue the variable delete callbacks and delete
577 the variables. The vardict is not valid at this point
578 anymore. That is the reason why we stored the
579 caseindex before reindexing. */
580 for (size_t vi = idx; vi < idx + count; vi++)
582 struct delvar *dv = (struct delvar *) ll_pop_head (&list);
583 var_clear_vardict (dv->var);
584 if (d->callbacks && d->callbacks->var_deleted)
585 d->callbacks->var_deleted (d, dv->var, vi, dv->case_index, d->cb_data);
591 /* Deletes scratch variables from dictionary D. */
593 dict_delete_scratch_vars (struct dictionary *d)
597 /* FIXME: this can be done in O(count) time, but this algorithm
599 for (i = 0; i < d->n_vars;)
600 if (var_get_dict_class (d->vars[i].var) == DC_SCRATCH)
601 dict_delete_var (d, d->vars[i].var);
608 /* Clears the contents from a dictionary without destroying the
609 dictionary itself. */
611 dict_clear__ (struct dictionary *d, bool skip_callbacks)
613 /* FIXME? Should we really clear case_limit, label, documents?
614 Others are necessarily cleared by deleting all the variables.*/
615 while (d->n_vars > 0)
617 dict_delete_var__ (d, d->vars[d->n_vars - 1].var, skip_callbacks);
622 d->n_vars = d->allocated_vars = 0;
623 invalidate_proto (d);
624 hmap_clear (&d->name_map);
625 d->next_value_idx = 0;
626 dict_set_split_vars__ (d, NULL, 0, SPLIT_LAYERED, skip_callbacks);
635 dict_set_weight (d, NULL);
636 dict_set_filter (d, NULL);
641 string_array_clear (&d->documents);
642 dict_clear_vectors (d);
643 attrset_clear (&d->attributes);
646 /* Clears the contents from a dictionary without destroying the
647 dictionary itself. */
649 dict_clear (struct dictionary *d)
651 dict_clear__ (d, false);
654 /* Clears a dictionary and destroys it. */
656 _dict_destroy (struct dictionary *d)
658 /* In general, we don't want callbacks occurring, if the dictionary
659 is being destroyed */
660 d->callbacks = NULL ;
662 dict_clear__ (d, true);
663 string_array_destroy (&d->documents);
664 hmap_destroy (&d->name_map);
665 attrset_destroy (&d->attributes);
666 dict_clear_mrsets (d);
672 dict_ref (struct dictionary *d)
679 dict_unref (struct dictionary *d)
684 assert (d->ref_cnt >= 0);
689 /* Returns the number of variables in D. */
691 dict_get_n_vars (const struct dictionary *d)
696 /* Returns the variable in D with dictionary index IDX, which
697 must be between 0 and the count returned by
698 dict_get_n_vars(), exclusive. */
700 dict_get_var (const struct dictionary *d, size_t idx)
702 assert (idx < d->n_vars);
704 return d->vars[idx].var;
707 /* Sets *VARS to an array of pointers to variables in D and *N
708 to the number of variables in *D. All variables are returned
709 except for those, if any, in the classes indicated by EXCLUDE.
710 (There is no point in putting DC_SYSTEM in EXCLUDE as
711 dictionaries never include system variables.) */
713 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
714 size_t *n, enum dict_class exclude)
716 dict_get_vars_mutable (d, (struct variable ***) vars, n, exclude);
719 /* Sets *VARS to an array of pointers to variables in D and *N
720 to the number of variables in *D. All variables are returned
721 except for those, if any, in the classes indicated by EXCLUDE.
722 (There is no point in putting DC_SYSTEM in EXCLUDE as
723 dictionaries never include system variables.) */
725 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
726 size_t *n, enum dict_class exclude)
731 assert (exclude == (exclude & DC_ALL));
734 for (i = 0; i < d->n_vars; i++)
736 enum dict_class class = var_get_dict_class (d->vars[i].var);
737 if (!(class & exclude))
741 *vars = xnmalloc (count, sizeof **vars);
743 for (i = 0; i < d->n_vars; i++)
745 enum dict_class class = var_get_dict_class (d->vars[i].var);
746 if (!(class & exclude))
747 (*vars)[(*n)++] = d->vars[i].var;
749 assert (*n == count);
752 static struct variable *
753 add_var_with_case_index (struct dictionary *d, struct variable *v,
756 struct vardict_info *vardict;
758 assert (case_index >= d->next_value_idx);
760 /* Update dictionary. */
761 if (d->n_vars >= d->allocated_vars)
765 d->vars = x2nrealloc (d->vars, &d->allocated_vars, sizeof *d->vars);
766 hmap_clear (&d->name_map);
767 for (i = 0; i < d->n_vars; i++)
769 var_set_vardict (d->vars[i].var, &d->vars[i]);
770 hmap_insert_fast (&d->name_map, &d->vars[i].name_node,
771 d->vars[i].name_node.hash);
775 vardict = &d->vars[d->n_vars++];
778 hmap_insert (&d->name_map, &vardict->name_node,
779 utf8_hash_case_string (var_get_name (v), 0));
780 vardict->case_index = case_index;
781 var_set_vardict (v, vardict);
783 if (d->changed) d->changed (d, d->changed_data);
784 if (d->callbacks && d->callbacks->var_added)
785 d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
787 invalidate_proto (d);
788 d->next_value_idx = case_index + 1;
793 static struct variable *
794 add_var (struct dictionary *d, struct variable *v)
796 return add_var_with_case_index (d, v, d->next_value_idx);
799 /* Creates and returns a new variable in D with the given NAME
800 and WIDTH. Returns a null pointer if the given NAME would
801 duplicate that of an existing variable in the dictionary. */
803 dict_create_var (struct dictionary *d, const char *name, int width)
805 return (dict_lookup_var (d, name) == NULL
806 ? dict_create_var_assert (d, name, width)
810 /* Creates and returns a new variable in D with the given NAME
811 and WIDTH. Assert-fails if the given NAME would duplicate
812 that of an existing variable in the dictionary. */
814 dict_create_var_assert (struct dictionary *d, const char *name, int width)
816 assert (dict_lookup_var (d, name) == NULL);
817 return add_var (d, var_create (name, width));
820 /* Creates and returns a new variable in D, as a copy of existing variable
821 OLD_VAR, which need not be in D or in any dictionary. Returns a null
822 pointer if OLD_VAR's name would duplicate that of an existing variable in
825 dict_clone_var (struct dictionary *d, const struct variable *old_var)
827 return dict_clone_var_as (d, old_var, var_get_name (old_var));
830 /* Creates and returns a new variable in D, as a copy of existing variable
831 OLD_VAR, which need not be in D or in any dictionary. Assert-fails if
832 OLD_VAR's name would duplicate that of an existing variable in the
835 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var)
837 return dict_clone_var_as_assert (d, old_var, var_get_name (old_var));
840 /* Creates and returns a new variable in D with name NAME, as a copy of
841 existing variable OLD_VAR, which need not be in D or in any dictionary.
842 Returns a null pointer if the given NAME would duplicate that of an existing
843 variable in the dictionary. */
845 dict_clone_var_as (struct dictionary *d, const struct variable *old_var,
848 return (dict_lookup_var (d, name) == NULL
849 ? dict_clone_var_as_assert (d, old_var, name)
853 /* Creates and returns a new variable in D with name NAME, as a copy of
854 existing variable OLD_VAR, which need not be in D or in any dictionary.
855 Assert-fails if the given NAME would duplicate that of an existing variable
856 in the dictionary. */
858 dict_clone_var_as_assert (struct dictionary *d, const struct variable *old_var,
861 struct variable *new_var = var_clone (old_var);
862 assert (dict_lookup_var (d, name) == NULL);
863 var_set_name (new_var, name);
864 return add_var (d, new_var);
868 dict_clone_var_in_place_assert (struct dictionary *d,
869 const struct variable *old_var)
871 assert (dict_lookup_var (d, var_get_name (old_var)) == NULL);
872 return add_var_with_case_index (d, var_clone (old_var),
873 var_get_case_index (old_var));
876 /* Returns the variable named NAME in D, or a null pointer if no
877 variable has that name. */
879 dict_lookup_var (const struct dictionary *d, const char *name)
881 struct vardict_info *vardict;
883 HMAP_FOR_EACH_WITH_HASH (vardict, struct vardict_info, name_node,
884 utf8_hash_case_string (name, 0), &d->name_map)
886 struct variable *var = vardict->var;
887 if (!utf8_strcasecmp (var_get_name (var), name))
894 /* Returns the variable named NAME in D. Assert-fails if no
895 variable has that name. */
897 dict_lookup_var_assert (const struct dictionary *d, const char *name)
899 struct variable *v = dict_lookup_var (d, name);
904 /* Returns true if variable V is in dictionary D,
907 dict_contains_var (const struct dictionary *d, const struct variable *v)
909 return (var_has_vardict (v)
910 && vardict_get_dictionary (var_get_vardict (v)) == d);
913 /* Moves V to 0-based position IDX in D. Other variables in D,
914 if any, retain their relative positions. Runs in time linear
915 in the distance moved. */
917 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
919 size_t old_index = var_get_dict_index (v);
921 assert (new_index < d->n_vars);
923 unindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
924 move_element (d->vars, d->n_vars, sizeof *d->vars, old_index, new_index);
925 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1, false);
928 /* Reorders the variables in D, placing the COUNT variables
929 listed in ORDER in that order at the beginning of D. The
930 other variables in D, if any, retain their relative
933 dict_reorder_vars (struct dictionary *d,
934 struct variable *const *order, size_t count)
936 struct vardict_info *new_var;
939 assert (count == 0 || order != NULL);
940 assert (count <= d->n_vars);
942 new_var = xnmalloc (d->allocated_vars, sizeof *new_var);
944 /* Add variables in ORDER to new_var. */
945 for (i = 0; i < count; i++)
947 struct vardict_info *old_var;
949 assert (dict_contains_var (d, order[i]));
951 old_var = var_get_vardict (order[i]);
952 new_var[i] = *old_var;
953 old_var->dict = NULL;
956 /* Add remaining variables to new_var. */
957 for (i = 0; i < d->n_vars; i++)
958 if (d->vars[i].dict != NULL)
959 new_var[count++] = d->vars[i];
960 assert (count == d->n_vars);
962 /* Replace old vardicts by new ones. */
966 hmap_clear (&d->name_map);
967 reindex_vars (d, 0, d->n_vars, false);
970 /* Changes the name of variable V that is currently in a dictionary to
973 rename_var (struct variable *v, const char *new_name)
975 struct vardict_info *vardict = var_get_vardict (v);
976 var_clear_vardict (v);
977 var_set_name (v, new_name);
978 vardict->name_node.hash = utf8_hash_case_string (new_name, 0);
979 var_set_vardict (v, vardict);
982 /* Tries to changes the name of V in D to name NEW_NAME. Returns true if
983 successful, false if a variable (other than V) with the given name already
986 dict_try_rename_var (struct dictionary *d, struct variable *v,
987 const char *new_name)
989 struct variable *conflict = dict_lookup_var (d, new_name);
990 if (conflict && v != conflict)
993 struct variable *old = var_clone (v);
994 unindex_var (d, var_get_vardict (v));
995 rename_var (v, new_name);
996 reindex_var (d, var_get_vardict (v), false);
998 if (settings_get_algorithm () == ENHANCED)
999 var_clear_short_names (v);
1001 if (d->changed) d->changed (d, d->changed_data);
1002 if (d->callbacks && d->callbacks->var_changed)
1003 d->callbacks->var_changed (d, var_get_dict_index (v), VAR_TRAIT_NAME, old, d->cb_data);
1010 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
1011 a variable named NEW_NAME is already in D, except that
1012 NEW_NAME may be the same as V's existing name. */
1014 dict_rename_var (struct dictionary *d, struct variable *v,
1015 const char *new_name)
1017 bool ok UNUSED = dict_try_rename_var (d, v, new_name);
1021 /* Renames COUNT variables specified in VARS to the names given
1022 in NEW_NAMES within dictionary D. If the renaming would
1023 result in a duplicate variable name, returns false and stores a
1024 name that would be duplicated into *ERR_NAME (if ERR_NAME is
1025 non-null). Otherwise, the renaming is successful, and true
1028 dict_rename_vars (struct dictionary *d,
1029 struct variable **vars, char **new_names, size_t count,
1036 assert (count == 0 || vars != NULL);
1037 assert (count == 0 || new_names != NULL);
1039 /* Save the names of the variables to be renamed. */
1040 pool = pool_create ();
1041 old_names = pool_nalloc (pool, count, sizeof *old_names);
1042 for (i = 0; i < count; i++)
1043 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
1045 /* Remove the variables to be renamed from the name hash,
1047 for (i = 0; i < count; i++)
1049 unindex_var (d, var_get_vardict (vars[i]));
1050 rename_var (vars[i], new_names[i]);
1053 /* Add the renamed variables back into the name hash,
1054 checking for conflicts. */
1055 for (i = 0; i < count; i++)
1057 if (dict_lookup_var (d, var_get_name (vars[i])) != NULL)
1059 /* There is a name conflict.
1060 Back out all the name changes that have already
1061 taken place, and indicate failure. */
1062 size_t fail_idx = i;
1063 if (err_name != NULL)
1064 *err_name = new_names[i];
1066 for (i = 0; i < fail_idx; i++)
1067 unindex_var (d, var_get_vardict (vars[i]));
1069 for (i = 0; i < count; i++)
1071 rename_var (vars[i], old_names[i]);
1072 reindex_var (d, var_get_vardict (vars[i]), false);
1075 pool_destroy (pool);
1078 reindex_var (d, var_get_vardict (vars[i]), false);
1081 /* Clear short names. */
1082 if (settings_get_algorithm () == ENHANCED)
1083 for (i = 0; i < count; i++)
1084 var_clear_short_names (vars[i]);
1086 pool_destroy (pool);
1090 /* Returns true if a variable named NAME may be inserted in DICT;
1091 that is, if there is not already a variable with that name in
1092 DICT and if NAME is not a reserved word. (The caller's checks
1093 have already verified that NAME is otherwise acceptable as a
1096 var_name_is_insertable (const struct dictionary *dict, const char *name)
1098 return (dict_lookup_var (dict, name) == NULL
1099 && lex_id_to_token (ss_cstr (name)) == T_ID);
1103 make_hinted_name (const struct dictionary *dict, const char *hint)
1105 size_t hint_len = strlen (hint);
1106 bool dropped = false;
1111 if (hint_len > ID_MAX_LEN)
1112 hint_len = ID_MAX_LEN;
1114 /* The allocation size here is OK: characters that are copied directly fit
1115 OK, and characters that are not copied directly are replaced by a single
1116 '_' byte. If u8_mbtouc() replaces bad input by 0xfffd, then that will get
1117 replaced by '_' too. */
1118 root = rp = xmalloc (hint_len + 1);
1119 for (ofs = 0; ofs < hint_len; ofs += mblen)
1123 mblen = u8_mbtouc (&uc, CHAR_CAST (const uint8_t *, hint + ofs),
1126 ? lex_uc_is_id1 (uc) && uc != '$'
1127 : lex_uc_is_idn (uc))
1134 rp += u8_uctomb (CHAR_CAST (uint8_t *, rp), uc, 6);
1136 else if (rp != root)
1141 if (root[0] != '\0')
1143 unsigned long int i;
1145 if (var_name_is_insertable (dict, root))
1148 for (i = 0; i < ULONG_MAX; i++)
1150 char suffix[INT_BUFSIZE_BOUND (i) + 1];
1154 if (!str_format_26adic (i + 1, true, &suffix[1], sizeof suffix - 1))
1157 name = utf8_encoding_concat (root, suffix, dict->encoding, 64);
1158 if (var_name_is_insertable (dict, name))
1173 make_numeric_name (const struct dictionary *dict, unsigned long int *num_start)
1175 unsigned long int number;
1177 for (number = num_start != NULL ? MAX (*num_start, 1) : 1;
1181 char name[3 + INT_STRLEN_BOUND (number) + 1];
1183 sprintf (name, "VAR%03lu", number);
1184 if (dict_lookup_var (dict, name) == NULL)
1186 if (num_start != NULL)
1187 *num_start = number + 1;
1188 return xstrdup (name);
1196 /* Devises and returns a variable name unique within DICT. The variable name
1197 is owned by the caller, which must free it with free() when it is no longer
1200 HINT, if it is non-null, is used as a suggestion that will be
1201 modified for suitability as a variable name and for
1204 If HINT is null or entirely unsuitable, a name in the form
1205 "VAR%03d" will be generated, where the smallest unused integer
1206 value is used. If NUM_START is non-null, then its value is
1207 used as the minimum numeric value to check, and it is updated
1208 to the next value to be checked.
1211 dict_make_unique_var_name (const struct dictionary *dict, const char *hint,
1212 unsigned long int *num_start)
1216 char *hinted_name = make_hinted_name (dict, hint);
1217 if (hinted_name != NULL)
1221 return make_numeric_name (dict, num_start);
1224 /* Returns whether variable names must be valid identifiers. Normally, this is
1225 true, but sometimes a dictionary is prepared for external use (e.g. output
1226 to a CSV file) where names don't have to be valid. */
1228 dict_get_names_must_be_ids (const struct dictionary *d)
1230 return d->names_must_be_ids;
1233 /* Sets whether variable names must be valid identifiers. Normally, this is
1234 true, but sometimes a dictionary is prepared for external use (e.g. output
1235 to a CSV file) where names don't have to be valid.
1237 Changing this setting from false to true doesn't make the dictionary check
1238 all the existing variable names, so it can cause an invariant violation. */
1240 dict_set_names_must_be_ids (struct dictionary *d, bool names_must_be_ids)
1242 d->names_must_be_ids = names_must_be_ids;
1245 /* Returns the weighting variable in dictionary D, or a null
1246 pointer if the dictionary is unweighted. */
1248 dict_get_weight (const struct dictionary *d)
1250 assert (d->weight == NULL || dict_contains_var (d, d->weight));
1255 /* Returns the value of D's weighting variable in case C, except
1256 that a negative weight is returned as 0. Returns 1 if the
1257 dictionary is unweighted. Will warn about missing, negative,
1258 or zero values if *WARN_ON_INVALID is true. The function will
1259 set *WARN_ON_INVALID to false if an invalid weight is
1262 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
1263 bool *warn_on_invalid)
1267 if (d->weight == NULL)
1271 double w = case_num (c, d->weight);
1273 return var_force_valid_weight (d->weight, w, warn_on_invalid);
1277 /* Returns the format to use for weights. */
1278 const struct fmt_spec *
1279 dict_get_weight_format (const struct dictionary *d)
1281 return d->weight ? var_get_print_format (d->weight) : &F_8_0;
1284 /* Sets the weighting variable of D to V, or turning off
1285 weighting if V is a null pointer. */
1287 dict_set_weight (struct dictionary *d, struct variable *v)
1289 assert (v == NULL || dict_contains_var (d, v));
1290 assert (v == NULL || var_is_numeric (v));
1294 if (d->changed) d->changed (d, d->changed_data);
1295 if (d->callbacks && d->callbacks->weight_changed)
1296 d->callbacks->weight_changed (d,
1297 v ? var_get_dict_index (v) : -1,
1301 /* Returns the filter variable in dictionary D (see cmd_filter())
1302 or a null pointer if the dictionary is unfiltered. */
1304 dict_get_filter (const struct dictionary *d)
1306 assert (d->filter == NULL || dict_contains_var (d, d->filter));
1311 /* Sets V as the filter variable for dictionary D. Passing a
1312 null pointer for V turn off filtering. */
1314 dict_set_filter (struct dictionary *d, struct variable *v)
1316 assert (v == NULL || dict_contains_var (d, v));
1317 assert (v == NULL || var_is_numeric (v));
1321 if (d->changed) d->changed (d, d->changed_data);
1322 if (d->callbacks && d->callbacks->filter_changed)
1323 d->callbacks->filter_changed (d,
1324 v ? var_get_dict_index (v) : -1,
1328 /* Returns the case limit for dictionary D, or zero if the number
1329 of cases is unlimited. */
1331 dict_get_case_limit (const struct dictionary *d)
1333 return d->case_limit;
1336 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
1337 0 for CASE_LIMIT to indicate no limit. */
1339 dict_set_case_limit (struct dictionary *d, casenumber case_limit)
1341 d->case_limit = case_limit;
1344 /* Returns the prototype used for cases created by dictionary D. */
1345 const struct caseproto *
1346 dict_get_proto (const struct dictionary *d_)
1348 struct dictionary *d = CONST_CAST (struct dictionary *, d_);
1349 if (d->proto == NULL)
1353 d->proto = caseproto_create ();
1354 d->proto = caseproto_reserve (d->proto, d->n_vars);
1355 for (i = 0; i < d->n_vars; i++)
1356 d->proto = caseproto_set_width (d->proto,
1357 var_get_case_index (d->vars[i].var),
1358 var_get_width (d->vars[i].var));
1363 /* Returns the case index of the next value to be added to D.
1364 This value is the number of `union value's that need to be
1365 allocated to store a case for dictionary D. */
1367 dict_get_next_value_idx (const struct dictionary *d)
1369 return d->next_value_idx;
1372 /* Returns the number of bytes needed to store a case for
1375 dict_get_case_size (const struct dictionary *d)
1377 return sizeof (union value) * dict_get_next_value_idx (d);
1380 /* Reassigns values in dictionary D so that fragmentation is
1383 dict_compact_values (struct dictionary *d)
1387 d->next_value_idx = 0;
1388 for (i = 0; i < d->n_vars; i++)
1390 struct variable *v = d->vars[i].var;
1391 set_var_case_index (v, d->next_value_idx++);
1393 invalidate_proto (d);
1396 /* Returns the number of values occupied by the variables in
1397 dictionary D. All variables are considered if EXCLUDE_CLASSES
1398 is 0, or it may contain one or more of (1u << DC_ORDINARY),
1399 (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to exclude the
1400 corresponding type of variable.
1402 The return value may be less than the number of values in one
1403 of dictionary D's cases (as returned by
1404 dict_get_next_value_idx) even if E is 0, because there may be
1405 gaps in D's cases due to deleted variables. */
1407 dict_count_values (const struct dictionary *d, unsigned int exclude_classes)
1409 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1411 | (1u << DC_SCRATCH))) == 0);
1414 for (size_t i = 0; i < d->n_vars; i++)
1416 enum dict_class class = var_get_dict_class (d->vars[i].var);
1417 if (!(exclude_classes & (1u << class)))
1423 /* Returns the case prototype that would result after deleting
1424 all variables from D that are not in one of the
1425 EXCLUDE_CLASSES and compacting the dictionary with
1428 The caller must unref the returned caseproto when it is no
1431 dict_get_compacted_proto (const struct dictionary *d,
1432 unsigned int exclude_classes)
1434 struct caseproto *proto;
1437 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1439 | (1u << DC_SCRATCH))) == 0);
1441 proto = caseproto_create ();
1442 for (i = 0; i < d->n_vars; i++)
1444 struct variable *v = d->vars[i].var;
1445 if (!(exclude_classes & (1u << var_get_dict_class (v))))
1446 proto = caseproto_add_width (proto, var_get_width (v));
1450 /* Returns the file label for D, or a null pointer if D is
1451 unlabeled (see cmd_file_label()). */
1453 dict_get_label (const struct dictionary *d)
1458 /* Sets D's file label to LABEL, truncating it to at most 60 bytes in D's
1461 Removes D's label if LABEL is null or the empty string. */
1463 dict_set_label (struct dictionary *d, const char *label)
1466 if (label == NULL || label[0] == '\0')
1469 d->label = utf8_encoding_trunc (label, d->encoding, 60);
1472 /* Returns the documents for D, as an UTF-8 encoded string_array. The
1473 return value is always nonnull; if there are no documents then the
1474 string_arary is empty.*/
1475 const struct string_array *
1476 dict_get_documents (const struct dictionary *d)
1478 return &d->documents;
1481 /* Replaces the documents for D by NEW_DOCS, a UTF-8 encoded string_array. */
1483 dict_set_documents (struct dictionary *d, const struct string_array *new_docs)
1485 /* Swap out the old documents, instead of destroying them immediately, to
1486 allow the new documents to include pointers into the old ones. */
1487 struct string_array old_docs = STRING_ARRAY_INITIALIZER;
1488 string_array_swap (&d->documents, &old_docs);
1490 for (size_t i = 0; i < new_docs->n; i++)
1491 dict_add_document_line (d, new_docs->strings[i], false);
1493 string_array_destroy (&old_docs);
1496 /* Replaces the documents for D by UTF-8 encoded string NEW_DOCS, dividing it
1497 into individual lines at new-line characters. Each line is truncated to at
1498 most DOC_LINE_LENGTH bytes in D's encoding. */
1500 dict_set_documents_string (struct dictionary *d, const char *new_docs)
1504 dict_clear_documents (d);
1505 for (s = new_docs; *s != '\0';)
1507 size_t len = strcspn (s, "\n");
1508 char *line = xmemdup0 (s, len);
1509 dict_add_document_line (d, line, false);
1518 /* Drops the documents from dictionary D. */
1520 dict_clear_documents (struct dictionary *d)
1522 string_array_clear (&d->documents);
1525 /* Appends the UTF-8 encoded LINE to the documents in D. LINE will be
1526 truncated so that it is no more than 80 bytes in the dictionary's
1527 encoding. If this causes some text to be lost, and ISSUE_WARNING is true,
1528 then a warning will be issued. */
1530 dict_add_document_line (struct dictionary *d, const char *line,
1536 trunc_len = utf8_encoding_trunc_len (line, d->encoding, DOC_LINE_LENGTH);
1537 truncated = line[trunc_len] != '\0';
1538 if (truncated && issue_warning)
1540 /* TRANSLATORS: "bytes" is correct, not characters due to UTF encoding */
1541 msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
1544 string_array_append_nocopy (&d->documents, xmemdup0 (line, trunc_len));
1549 /* Returns the number of document lines in dictionary D. */
1551 dict_get_document_n_lines (const struct dictionary *d)
1553 return d->documents.n;
1556 /* Returns document line number IDX in dictionary D. The caller must not
1557 modify or free the returned string. */
1559 dict_get_document_line (const struct dictionary *d, size_t idx)
1561 assert (idx < d->documents.n);
1562 return d->documents.strings[idx];
1565 /* Creates in D a vector named NAME that contains the N
1566 variables in VAR. Returns true if successful, or false if a
1567 vector named NAME already exists in D. */
1569 dict_create_vector (struct dictionary *d,
1571 struct variable **var, size_t n)
1574 for (size_t i = 0; i < n; i++)
1575 assert (dict_contains_var (d, var[i]));
1577 if (dict_lookup_vector (d, name) == NULL)
1579 d->vector = xnrealloc (d->vector, d->n_vectors + 1, sizeof *d->vector);
1580 d->vector[d->n_vectors++] = vector_create (name, var, n);
1587 /* Creates in D a vector named NAME that contains the N
1588 variables in VAR. A vector named NAME must not already exist
1591 dict_create_vector_assert (struct dictionary *d,
1593 struct variable **var, size_t n)
1595 assert (dict_lookup_vector (d, name) == NULL);
1596 dict_create_vector (d, name, var, n);
1599 /* Returns the vector in D with index IDX, which must be less
1600 than dict_get_n_vectors (D). */
1601 const struct vector *
1602 dict_get_vector (const struct dictionary *d, size_t idx)
1604 assert (idx < d->n_vectors);
1606 return d->vector[idx];
1609 /* Returns the number of vectors in D. */
1611 dict_get_n_vectors (const struct dictionary *d)
1613 return d->n_vectors;
1616 /* Looks up and returns the vector within D with the given
1618 const struct vector *
1619 dict_lookup_vector (const struct dictionary *d, const char *name)
1622 for (i = 0; i < d->n_vectors; i++)
1623 if (!utf8_strcasecmp (vector_get_name (d->vector[i]), name))
1624 return d->vector[i];
1628 /* Deletes all vectors from D. */
1630 dict_clear_vectors (struct dictionary *d)
1634 for (i = 0; i < d->n_vectors; i++)
1635 vector_destroy (d->vector[i]);
1642 /* Multiple response sets. */
1644 /* Returns the multiple response set in DICT with index IDX, which must be
1645 between 0 and the count returned by dict_get_n_mrsets(), exclusive. */
1646 const struct mrset *
1647 dict_get_mrset (const struct dictionary *dict, size_t idx)
1649 assert (idx < dict->n_mrsets);
1650 return dict->mrsets[idx];
1653 /* Returns the number of multiple response sets in DICT. */
1655 dict_get_n_mrsets (const struct dictionary *dict)
1657 return dict->n_mrsets;
1660 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1661 returns its index; otherwise, returns SIZE_MAX. */
1663 dict_lookup_mrset_idx (const struct dictionary *dict, const char *name)
1667 for (i = 0; i < dict->n_mrsets; i++)
1668 if (!utf8_strcasecmp (name, dict->mrsets[i]->name))
1674 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1675 returns it; otherwise, returns NULL. */
1676 const struct mrset *
1677 dict_lookup_mrset (const struct dictionary *dict, const char *name)
1679 size_t idx = dict_lookup_mrset_idx (dict, name);
1680 return idx != SIZE_MAX ? dict->mrsets[idx] : NULL;
1683 /* Adds MRSET to DICT, replacing any existing set with the same name. Returns
1684 true if a set was replaced, false if none existed with the specified name.
1686 Ownership of MRSET is transferred to DICT. */
1688 dict_add_mrset (struct dictionary *dict, struct mrset *mrset)
1692 assert (mrset_ok (mrset, dict));
1694 idx = dict_lookup_mrset_idx (dict, mrset->name);
1695 if (idx == SIZE_MAX)
1697 dict->mrsets = xrealloc (dict->mrsets,
1698 (dict->n_mrsets + 1) * sizeof *dict->mrsets);
1699 dict->mrsets[dict->n_mrsets++] = mrset;
1704 mrset_destroy (dict->mrsets[idx]);
1705 dict->mrsets[idx] = mrset;
1710 /* Looks for a multiple response set in DICT named NAME. If found, removes it
1711 from DICT and returns true. If none is found, returns false without
1714 Deleting one multiple response set causes the indexes of other sets within
1717 dict_delete_mrset (struct dictionary *dict, const char *name)
1719 size_t idx = dict_lookup_mrset_idx (dict, name);
1720 if (idx != SIZE_MAX)
1722 mrset_destroy (dict->mrsets[idx]);
1723 dict->mrsets[idx] = dict->mrsets[--dict->n_mrsets];
1730 /* Deletes all multiple response sets from DICT. */
1732 dict_clear_mrsets (struct dictionary *dict)
1736 for (i = 0; i < dict->n_mrsets; i++)
1737 mrset_destroy (dict->mrsets[i]);
1738 free (dict->mrsets);
1739 dict->mrsets = NULL;
1743 /* Removes VAR, which must be in DICT, from DICT's multiple response sets. */
1745 dict_unset_mrset_var (struct dictionary *dict, struct variable *var)
1749 assert (dict_contains_var (dict, var));
1751 for (i = 0; i < dict->n_mrsets;)
1753 struct mrset *mrset = dict->mrsets[i];
1756 for (j = 0; j < mrset->n_vars;)
1757 if (mrset->vars[j] == var)
1758 remove_element (mrset->vars, mrset->n_vars--,
1759 sizeof *mrset->vars, j);
1763 if (mrset->n_vars < 2)
1765 mrset_destroy (mrset);
1766 dict->mrsets[i] = dict->mrsets[--dict->n_mrsets];
1773 /* Returns D's attribute set. The caller may examine or modify
1774 the attribute set, but must not destroy it. Destroying D or
1775 calling dict_set_attributes for D will also destroy D's
1778 dict_get_attributes (const struct dictionary *d)
1780 return CONST_CAST (struct attrset *, &d->attributes);
1783 /* Replaces D's attributes set by a copy of ATTRS. */
1785 dict_set_attributes (struct dictionary *d, const struct attrset *attrs)
1787 attrset_destroy (&d->attributes);
1788 attrset_clone (&d->attributes, attrs);
1791 /* Returns true if D has at least one attribute in its attribute
1792 set, false if D's attribute set is empty. */
1794 dict_has_attributes (const struct dictionary *d)
1796 return attrset_count (&d->attributes) > 0;
1799 /* Called from variable.c to notify the dictionary that some property (indicated
1800 by WHAT) of the variable has changed. OLDVAR is a copy of V as it existed
1801 prior to the change. OLDVAR is destroyed by this function.
1804 dict_var_changed (const struct variable *v, unsigned int what, struct variable *oldvar)
1806 if (var_has_vardict (v))
1808 const struct vardict_info *vardict = var_get_vardict (v);
1809 struct dictionary *d = vardict->dict;
1814 if (what & (VAR_TRAIT_WIDTH | VAR_TRAIT_POSITION))
1815 invalidate_proto (d);
1817 if (d->changed) d->changed (d, d->changed_data);
1818 if (d->callbacks && d->callbacks->var_changed)
1819 d->callbacks->var_changed (d, var_get_dict_index (v), what, oldvar, d->cb_data);
1826 /* Dictionary used to contain "internal variables". */
1827 static struct dictionary *internal_dict;
1829 /* Create a variable of the specified WIDTH to be used for internal
1830 calculations only. The variable is assigned case index CASE_IDX. */
1832 dict_create_internal_var (int case_idx, int width)
1834 if (internal_dict == NULL)
1835 internal_dict = dict_create ("UTF-8");
1839 static int counter = INT_MAX / 2;
1840 struct variable *var;
1843 if (++counter == INT_MAX)
1844 counter = INT_MAX / 2;
1846 sprintf (name, "$internal%d", counter);
1847 var = dict_create_var (internal_dict, name, width);
1850 set_var_case_index (var, case_idx);
1856 /* Destroys VAR, which must have been created with
1857 dict_create_internal_var(). */
1859 dict_destroy_internal_var (struct variable *var)
1863 dict_delete_var (internal_dict, var);
1865 /* Destroy internal_dict if it has no variables left, just so that
1866 valgrind --leak-check --show-reachable won't show internal_dict. */
1867 if (dict_get_n_vars (internal_dict) == 0)
1869 dict_unref (internal_dict);
1870 internal_dict = NULL;
1876 vardict_get_dict_index (const struct vardict_info *vardict)
1878 return vardict - vardict->dict->vars;