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);
417 dict_get_split_type (const struct dictionary *d)
419 return d->split_type;
422 /* Sets N split vars SPLIT in dictionary D. */
424 dict_set_split_vars (struct dictionary *d,
425 struct variable *const *split, size_t n,
426 enum split_type type)
428 dict_set_split_vars__ (d, split, n, type, false);
432 dict_clear_split_vars (struct dictionary *d)
434 dict_set_split_vars (d, NULL, 0, SPLIT_LAYERED);
438 /* Deletes variable V from dictionary D and frees V.
440 This is a very bad idea if there might be any pointers to V
441 from outside D. In general, no variable in the active dataset's
442 dictionary should be deleted when any transformations are
443 active on the dictionary's dataset, because those
444 transformations might reference the deleted variable. The
445 safest time to delete a variable is just after a procedure has
446 been executed, as done by DELETE VARIABLES.
448 Pointers to V within D are not a problem, because
449 dict_delete_var() knows to remove V from split variables,
450 weights, filters, etc. */
452 dict_delete_var__ (struct dictionary *d, struct variable *v, bool skip_callbacks)
454 int dict_index = var_get_dict_index (v);
455 const int case_index = var_get_case_index (v);
457 assert (dict_contains_var (d, v));
459 dict_unset_split_var (d, v, skip_callbacks);
460 dict_unset_mrset_var (d, v);
463 dict_set_weight (d, NULL);
466 dict_set_filter (d, NULL);
468 dict_clear_vectors (d);
470 /* Remove V from var array. */
471 unindex_vars (d, dict_index, d->n_vars);
472 remove_element (d->vars, d->n_vars, sizeof *d->vars, dict_index);
475 /* Update dict_index for each affected variable. */
476 reindex_vars (d, dict_index, d->n_vars, skip_callbacks);
479 var_clear_vardict (v);
481 if (! skip_callbacks)
483 if (d->changed) d->changed (d, d->changed_data);
484 if (d->callbacks && d->callbacks->var_deleted)
485 d->callbacks->var_deleted (d, v, dict_index, case_index, d->cb_data);
488 invalidate_proto (d);
492 /* Deletes variable V from dictionary D and frees V.
494 This is a very bad idea if there might be any pointers to V
495 from outside D. In general, no variable in the active dataset's
496 dictionary should be deleted when any transformations are
497 active on the dictionary's dataset, because those
498 transformations might reference the deleted variable. The
499 safest time to delete a variable is just after a procedure has
500 been executed, as done by DELETE VARIABLES.
502 Pointers to V within D are not a problem, because
503 dict_delete_var() knows to remove V from split variables,
504 weights, filters, etc. */
506 dict_delete_var (struct dictionary *d, struct variable *v)
508 dict_delete_var__ (d, v, false);
512 /* Deletes the COUNT variables listed in VARS from D. This is
513 unsafe; see the comment on dict_delete_var() for details. */
515 dict_delete_vars (struct dictionary *d,
516 struct variable *const *vars, size_t count)
518 /* FIXME: this can be done in O(count) time, but this algorithm
520 assert (count == 0 || vars != NULL);
523 dict_delete_var (d, *vars++);
526 /* Deletes the COUNT variables in D starting at index IDX. This
527 is unsafe; see the comment on dict_delete_var() for
528 details. Deleting consecutive vars will result in less callbacks
529 compared to iterating over dict_delete_var.
530 A simple while loop over dict_delete_var will
531 produce (d->n_vars - IDX) * COUNT variable changed callbacks
532 plus COUNT variable delete callbacks.
533 This here produces d->n_vars - IDX variable changed callbacks
534 plus COUNT variable delete callbacks. */
536 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
538 assert (idx + count <= d->n_vars);
540 /* We need to store the variable and the corresponding case_index
541 for the delete callbacks later. We store them in a linked list.*/
544 struct variable *var;
547 struct ll_list list = LL_INITIALIZER (list);
549 for (size_t i = idx; i < idx + count; i++)
551 struct delvar *dv = xmalloc (sizeof (struct delvar));
553 struct variable *v = d->vars[i].var;
555 dict_unset_split_var (d, v, false);
556 dict_unset_mrset_var (d, v);
559 dict_set_weight (d, NULL);
562 dict_set_filter (d, NULL);
565 dv->case_index = var_get_case_index (v);
566 ll_push_tail (&list, (struct ll *)dv);
569 dict_clear_vectors (d);
571 /* Remove variables from var array. */
572 unindex_vars (d, idx, d->n_vars);
573 remove_range (d->vars, d->n_vars, sizeof *d->vars, idx, count);
576 /* Reindexing will result variable-changed callback */
577 reindex_vars (d, idx, d->n_vars, false);
579 invalidate_proto (d);
580 if (d->changed) d->changed (d, d->changed_data);
582 /* Now issue the variable delete callbacks and delete
583 the variables. The vardict is not valid at this point
584 anymore. That is the reason why we stored the
585 caseindex before reindexing. */
586 for (size_t vi = idx; vi < idx + count; vi++)
588 struct delvar *dv = (struct delvar *) ll_pop_head (&list);
589 var_clear_vardict (dv->var);
590 if (d->callbacks && d->callbacks->var_deleted)
591 d->callbacks->var_deleted (d, dv->var, vi, dv->case_index, d->cb_data);
597 /* Deletes scratch variables from dictionary D. */
599 dict_delete_scratch_vars (struct dictionary *d)
603 /* FIXME: this can be done in O(count) time, but this algorithm
605 for (i = 0; i < d->n_vars;)
606 if (var_get_dict_class (d->vars[i].var) == DC_SCRATCH)
607 dict_delete_var (d, d->vars[i].var);
614 /* Clears the contents from a dictionary without destroying the
615 dictionary itself. */
617 dict_clear__ (struct dictionary *d, bool skip_callbacks)
619 /* FIXME? Should we really clear case_limit, label, documents?
620 Others are necessarily cleared by deleting all the variables.*/
621 while (d->n_vars > 0)
623 dict_delete_var__ (d, d->vars[d->n_vars - 1].var, skip_callbacks);
628 d->n_vars = d->allocated_vars = 0;
629 invalidate_proto (d);
630 hmap_clear (&d->name_map);
631 d->next_value_idx = 0;
632 dict_set_split_vars__ (d, NULL, 0, SPLIT_LAYERED, skip_callbacks);
641 dict_set_weight (d, NULL);
642 dict_set_filter (d, NULL);
647 string_array_clear (&d->documents);
648 dict_clear_vectors (d);
649 attrset_clear (&d->attributes);
652 /* Clears the contents from a dictionary without destroying the
653 dictionary itself. */
655 dict_clear (struct dictionary *d)
657 dict_clear__ (d, false);
660 /* Clears a dictionary and destroys it. */
662 _dict_destroy (struct dictionary *d)
664 /* In general, we don't want callbacks occurring, if the dictionary
665 is being destroyed */
666 d->callbacks = NULL ;
668 dict_clear__ (d, true);
669 string_array_destroy (&d->documents);
670 hmap_destroy (&d->name_map);
671 attrset_destroy (&d->attributes);
672 dict_clear_mrsets (d);
678 dict_ref (struct dictionary *d)
685 dict_unref (struct dictionary *d)
690 assert (d->ref_cnt >= 0);
695 /* Returns the number of variables in D. */
697 dict_get_n_vars (const struct dictionary *d)
702 /* Returns the variable in D with dictionary index IDX, which
703 must be between 0 and the count returned by
704 dict_get_n_vars(), exclusive. */
706 dict_get_var (const struct dictionary *d, size_t idx)
708 assert (idx < d->n_vars);
710 return d->vars[idx].var;
713 /* Sets *VARS to an array of pointers to variables in D and *N
714 to the number of variables in *D. All variables are returned
715 except for those, if any, in the classes indicated by EXCLUDE.
716 (There is no point in putting DC_SYSTEM in EXCLUDE as
717 dictionaries never include system variables.) */
719 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
720 size_t *n, enum dict_class exclude)
722 dict_get_vars_mutable (d, (struct variable ***) vars, n, exclude);
725 /* Sets *VARS to an array of pointers to variables in D and *N
726 to the number of variables in *D. All variables are returned
727 except for those, if any, in the classes indicated by EXCLUDE.
728 (There is no point in putting DC_SYSTEM in EXCLUDE as
729 dictionaries never include system variables.) */
731 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
732 size_t *n, enum dict_class exclude)
737 assert (exclude == (exclude & DC_ALL));
740 for (i = 0; i < d->n_vars; i++)
742 enum dict_class class = var_get_dict_class (d->vars[i].var);
743 if (!(class & exclude))
747 *vars = xnmalloc (count, sizeof **vars);
749 for (i = 0; i < d->n_vars; i++)
751 enum dict_class class = var_get_dict_class (d->vars[i].var);
752 if (!(class & exclude))
753 (*vars)[(*n)++] = d->vars[i].var;
755 assert (*n == count);
758 static struct variable *
759 add_var_with_case_index (struct dictionary *d, struct variable *v,
762 struct vardict_info *vardict;
764 assert (case_index >= d->next_value_idx);
766 /* Update dictionary. */
767 if (d->n_vars >= d->allocated_vars)
771 d->vars = x2nrealloc (d->vars, &d->allocated_vars, sizeof *d->vars);
772 hmap_clear (&d->name_map);
773 for (i = 0; i < d->n_vars; i++)
775 var_set_vardict (d->vars[i].var, &d->vars[i]);
776 hmap_insert_fast (&d->name_map, &d->vars[i].name_node,
777 d->vars[i].name_node.hash);
781 vardict = &d->vars[d->n_vars++];
784 hmap_insert (&d->name_map, &vardict->name_node,
785 utf8_hash_case_string (var_get_name (v), 0));
786 vardict->case_index = case_index;
787 var_set_vardict (v, vardict);
789 if (d->changed) d->changed (d, d->changed_data);
790 if (d->callbacks && d->callbacks->var_added)
791 d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
793 invalidate_proto (d);
794 d->next_value_idx = case_index + 1;
799 static struct variable *
800 add_var (struct dictionary *d, struct variable *v)
802 return add_var_with_case_index (d, v, d->next_value_idx);
805 /* Creates and returns a new variable in D with the given NAME
806 and WIDTH. Returns a null pointer if the given NAME would
807 duplicate that of an existing variable in the dictionary. */
809 dict_create_var (struct dictionary *d, const char *name, int width)
811 return (dict_lookup_var (d, name) == NULL
812 ? dict_create_var_assert (d, name, width)
816 /* Creates and returns a new variable in D with the given NAME
817 and WIDTH. Assert-fails if the given NAME would duplicate
818 that of an existing variable in the dictionary. */
820 dict_create_var_assert (struct dictionary *d, const char *name, int width)
822 assert (dict_lookup_var (d, name) == NULL);
823 return add_var (d, var_create (name, width));
826 /* Creates and returns a new variable in D, as a copy of existing variable
827 OLD_VAR, which need not be in D or in any dictionary. Returns a null
828 pointer if OLD_VAR's name would duplicate that of an existing variable in
831 dict_clone_var (struct dictionary *d, const struct variable *old_var)
833 return dict_clone_var_as (d, old_var, var_get_name (old_var));
836 /* Creates and returns a new variable in D, as a copy of existing variable
837 OLD_VAR, which need not be in D or in any dictionary. Assert-fails if
838 OLD_VAR's name would duplicate that of an existing variable in the
841 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var)
843 return dict_clone_var_as_assert (d, old_var, var_get_name (old_var));
846 /* Creates and returns a new variable in D with name NAME, as a copy of
847 existing variable OLD_VAR, which need not be in D or in any dictionary.
848 Returns a null pointer if the given NAME would duplicate that of an existing
849 variable in the dictionary. */
851 dict_clone_var_as (struct dictionary *d, const struct variable *old_var,
854 return (dict_lookup_var (d, name) == NULL
855 ? dict_clone_var_as_assert (d, old_var, name)
859 /* Creates and returns a new variable in D with name NAME, as a copy of
860 existing variable OLD_VAR, which need not be in D or in any dictionary.
861 Assert-fails if the given NAME would duplicate that of an existing variable
862 in the dictionary. */
864 dict_clone_var_as_assert (struct dictionary *d, const struct variable *old_var,
867 struct variable *new_var = var_clone (old_var);
868 assert (dict_lookup_var (d, name) == NULL);
869 var_set_name (new_var, name);
870 return add_var (d, new_var);
874 dict_clone_var_in_place_assert (struct dictionary *d,
875 const struct variable *old_var)
877 assert (dict_lookup_var (d, var_get_name (old_var)) == NULL);
878 return add_var_with_case_index (d, var_clone (old_var),
879 var_get_case_index (old_var));
882 /* Returns the variable named NAME in D, or a null pointer if no
883 variable has that name. */
885 dict_lookup_var (const struct dictionary *d, const char *name)
887 struct vardict_info *vardict;
889 HMAP_FOR_EACH_WITH_HASH (vardict, struct vardict_info, name_node,
890 utf8_hash_case_string (name, 0), &d->name_map)
892 struct variable *var = vardict->var;
893 if (!utf8_strcasecmp (var_get_name (var), name))
900 /* Returns the variable named NAME in D. Assert-fails if no
901 variable has that name. */
903 dict_lookup_var_assert (const struct dictionary *d, const char *name)
905 struct variable *v = dict_lookup_var (d, name);
910 /* Returns true if variable V is in dictionary D,
913 dict_contains_var (const struct dictionary *d, const struct variable *v)
915 return (var_has_vardict (v)
916 && vardict_get_dictionary (var_get_vardict (v)) == d);
919 /* Moves V to 0-based position IDX in D. Other variables in D,
920 if any, retain their relative positions. Runs in time linear
921 in the distance moved. */
923 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
925 size_t old_index = var_get_dict_index (v);
927 assert (new_index < d->n_vars);
929 unindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
930 move_element (d->vars, d->n_vars, sizeof *d->vars, old_index, new_index);
931 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1, false);
934 /* Reorders the variables in D, placing the COUNT variables
935 listed in ORDER in that order at the beginning of D. The
936 other variables in D, if any, retain their relative
939 dict_reorder_vars (struct dictionary *d,
940 struct variable *const *order, size_t count)
942 struct vardict_info *new_var;
945 assert (count == 0 || order != NULL);
946 assert (count <= d->n_vars);
948 new_var = xnmalloc (d->allocated_vars, sizeof *new_var);
950 /* Add variables in ORDER to new_var. */
951 for (i = 0; i < count; i++)
953 struct vardict_info *old_var;
955 assert (dict_contains_var (d, order[i]));
957 old_var = var_get_vardict (order[i]);
958 new_var[i] = *old_var;
959 old_var->dict = NULL;
962 /* Add remaining variables to new_var. */
963 for (i = 0; i < d->n_vars; i++)
964 if (d->vars[i].dict != NULL)
965 new_var[count++] = d->vars[i];
966 assert (count == d->n_vars);
968 /* Replace old vardicts by new ones. */
972 hmap_clear (&d->name_map);
973 reindex_vars (d, 0, d->n_vars, false);
976 /* Changes the name of variable V that is currently in a dictionary to
979 rename_var (struct variable *v, const char *new_name)
981 struct vardict_info *vardict = var_get_vardict (v);
982 var_clear_vardict (v);
983 var_set_name (v, new_name);
984 vardict->name_node.hash = utf8_hash_case_string (new_name, 0);
985 var_set_vardict (v, vardict);
988 /* Tries to changes the name of V in D to name NEW_NAME. Returns true if
989 successful, false if a variable (other than V) with the given name already
992 dict_try_rename_var (struct dictionary *d, struct variable *v,
993 const char *new_name)
995 struct variable *conflict = dict_lookup_var (d, new_name);
996 if (conflict && v != conflict)
999 struct variable *old = var_clone (v);
1000 unindex_var (d, var_get_vardict (v));
1001 rename_var (v, new_name);
1002 reindex_var (d, var_get_vardict (v), false);
1004 if (settings_get_algorithm () == ENHANCED)
1005 var_clear_short_names (v);
1007 if (d->changed) d->changed (d, d->changed_data);
1008 if (d->callbacks && d->callbacks->var_changed)
1009 d->callbacks->var_changed (d, var_get_dict_index (v), VAR_TRAIT_NAME, old, d->cb_data);
1016 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
1017 a variable named NEW_NAME is already in D, except that
1018 NEW_NAME may be the same as V's existing name. */
1020 dict_rename_var (struct dictionary *d, struct variable *v,
1021 const char *new_name)
1023 bool ok UNUSED = dict_try_rename_var (d, v, new_name);
1027 /* Renames COUNT variables specified in VARS to the names given
1028 in NEW_NAMES within dictionary D. If the renaming would
1029 result in a duplicate variable name, returns false and stores a
1030 name that would be duplicated into *ERR_NAME (if ERR_NAME is
1031 non-null). Otherwise, the renaming is successful, and true
1034 dict_rename_vars (struct dictionary *d,
1035 struct variable **vars, char **new_names, size_t count,
1042 assert (count == 0 || vars != NULL);
1043 assert (count == 0 || new_names != NULL);
1045 /* Save the names of the variables to be renamed. */
1046 pool = pool_create ();
1047 old_names = pool_nalloc (pool, count, sizeof *old_names);
1048 for (i = 0; i < count; i++)
1049 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
1051 /* Remove the variables to be renamed from the name hash,
1053 for (i = 0; i < count; i++)
1055 unindex_var (d, var_get_vardict (vars[i]));
1056 rename_var (vars[i], new_names[i]);
1059 /* Add the renamed variables back into the name hash,
1060 checking for conflicts. */
1061 for (i = 0; i < count; i++)
1063 if (dict_lookup_var (d, var_get_name (vars[i])) != NULL)
1065 /* There is a name conflict.
1066 Back out all the name changes that have already
1067 taken place, and indicate failure. */
1068 size_t fail_idx = i;
1069 if (err_name != NULL)
1070 *err_name = new_names[i];
1072 for (i = 0; i < fail_idx; i++)
1073 unindex_var (d, var_get_vardict (vars[i]));
1075 for (i = 0; i < count; i++)
1077 rename_var (vars[i], old_names[i]);
1078 reindex_var (d, var_get_vardict (vars[i]), false);
1081 pool_destroy (pool);
1084 reindex_var (d, var_get_vardict (vars[i]), false);
1087 /* Clear short names. */
1088 if (settings_get_algorithm () == ENHANCED)
1089 for (i = 0; i < count; i++)
1090 var_clear_short_names (vars[i]);
1092 pool_destroy (pool);
1096 /* Returns true if a variable named NAME may be inserted in DICT;
1097 that is, if there is not already a variable with that name in
1098 DICT and if NAME is not a reserved word. (The caller's checks
1099 have already verified that NAME is otherwise acceptable as a
1102 var_name_is_insertable (const struct dictionary *dict, const char *name)
1104 return (dict_lookup_var (dict, name) == NULL
1105 && lex_id_to_token (ss_cstr (name)) == T_ID);
1109 make_hinted_name (const struct dictionary *dict, const char *hint)
1111 size_t hint_len = strlen (hint);
1112 bool dropped = false;
1117 if (hint_len > ID_MAX_LEN)
1118 hint_len = ID_MAX_LEN;
1120 /* The allocation size here is OK: characters that are copied directly fit
1121 OK, and characters that are not copied directly are replaced by a single
1122 '_' byte. If u8_mbtouc() replaces bad input by 0xfffd, then that will get
1123 replaced by '_' too. */
1124 root = rp = xmalloc (hint_len + 1);
1125 for (ofs = 0; ofs < hint_len; ofs += mblen)
1129 mblen = u8_mbtouc (&uc, CHAR_CAST (const uint8_t *, hint + ofs),
1132 ? lex_uc_is_id1 (uc) && uc != '$'
1133 : lex_uc_is_idn (uc))
1140 rp += u8_uctomb (CHAR_CAST (uint8_t *, rp), uc, 6);
1142 else if (rp != root)
1147 if (root[0] != '\0')
1149 unsigned long int i;
1151 if (var_name_is_insertable (dict, root))
1154 for (i = 0; i < ULONG_MAX; i++)
1156 char suffix[INT_BUFSIZE_BOUND (i) + 1];
1160 if (!str_format_26adic (i + 1, true, &suffix[1], sizeof suffix - 1))
1163 name = utf8_encoding_concat (root, suffix, dict->encoding, 64);
1164 if (var_name_is_insertable (dict, name))
1179 make_numeric_name (const struct dictionary *dict, unsigned long int *num_start)
1181 unsigned long int number;
1183 for (number = num_start != NULL ? MAX (*num_start, 1) : 1;
1187 char name[3 + INT_STRLEN_BOUND (number) + 1];
1189 sprintf (name, "VAR%03lu", number);
1190 if (dict_lookup_var (dict, name) == NULL)
1192 if (num_start != NULL)
1193 *num_start = number + 1;
1194 return xstrdup (name);
1202 /* Devises and returns a variable name unique within DICT. The variable name
1203 is owned by the caller, which must free it with free() when it is no longer
1206 HINT, if it is non-null, is used as a suggestion that will be
1207 modified for suitability as a variable name and for
1210 If HINT is null or entirely unsuitable, a name in the form
1211 "VAR%03d" will be generated, where the smallest unused integer
1212 value is used. If NUM_START is non-null, then its value is
1213 used as the minimum numeric value to check, and it is updated
1214 to the next value to be checked.
1217 dict_make_unique_var_name (const struct dictionary *dict, const char *hint,
1218 unsigned long int *num_start)
1222 char *hinted_name = make_hinted_name (dict, hint);
1223 if (hinted_name != NULL)
1227 return make_numeric_name (dict, num_start);
1230 /* Returns whether variable names must be valid identifiers. Normally, this is
1231 true, but sometimes a dictionary is prepared for external use (e.g. output
1232 to a CSV file) where names don't have to be valid. */
1234 dict_get_names_must_be_ids (const struct dictionary *d)
1236 return d->names_must_be_ids;
1239 /* Sets whether variable names must be valid identifiers. Normally, this is
1240 true, but sometimes a dictionary is prepared for external use (e.g. output
1241 to a CSV file) where names don't have to be valid.
1243 Changing this setting from false to true doesn't make the dictionary check
1244 all the existing variable names, so it can cause an invariant violation. */
1246 dict_set_names_must_be_ids (struct dictionary *d, bool names_must_be_ids)
1248 d->names_must_be_ids = names_must_be_ids;
1251 /* Returns the weighting variable in dictionary D, or a null
1252 pointer if the dictionary is unweighted. */
1254 dict_get_weight (const struct dictionary *d)
1256 assert (d->weight == NULL || dict_contains_var (d, d->weight));
1261 /* Returns the value of D's weighting variable in case C, except
1262 that a negative weight is returned as 0. Returns 1 if the
1263 dictionary is unweighted. Will warn about missing, negative,
1264 or zero values if *WARN_ON_INVALID is true. The function will
1265 set *WARN_ON_INVALID to false if an invalid weight is
1268 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
1269 bool *warn_on_invalid)
1273 if (d->weight == NULL)
1277 double w = case_num (c, d->weight);
1279 return var_force_valid_weight (d->weight, w, warn_on_invalid);
1283 /* Returns the format to use for weights. */
1284 const struct fmt_spec *
1285 dict_get_weight_format (const struct dictionary *d)
1287 return d->weight ? var_get_print_format (d->weight) : &F_8_0;
1290 /* Sets the weighting variable of D to V, or turning off
1291 weighting if V is a null pointer. */
1293 dict_set_weight (struct dictionary *d, struct variable *v)
1295 assert (v == NULL || dict_contains_var (d, v));
1296 assert (v == NULL || var_is_numeric (v));
1300 if (d->changed) d->changed (d, d->changed_data);
1301 if (d->callbacks && d->callbacks->weight_changed)
1302 d->callbacks->weight_changed (d,
1303 v ? var_get_dict_index (v) : -1,
1307 /* Returns the filter variable in dictionary D (see cmd_filter())
1308 or a null pointer if the dictionary is unfiltered. */
1310 dict_get_filter (const struct dictionary *d)
1312 assert (d->filter == NULL || dict_contains_var (d, d->filter));
1317 /* Sets V as the filter variable for dictionary D. Passing a
1318 null pointer for V turn off filtering. */
1320 dict_set_filter (struct dictionary *d, struct variable *v)
1322 assert (v == NULL || dict_contains_var (d, v));
1323 assert (v == NULL || var_is_numeric (v));
1327 if (d->changed) d->changed (d, d->changed_data);
1328 if (d->callbacks && d->callbacks->filter_changed)
1329 d->callbacks->filter_changed (d,
1330 v ? var_get_dict_index (v) : -1,
1334 /* Returns the case limit for dictionary D, or zero if the number
1335 of cases is unlimited. */
1337 dict_get_case_limit (const struct dictionary *d)
1339 return d->case_limit;
1342 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
1343 0 for CASE_LIMIT to indicate no limit. */
1345 dict_set_case_limit (struct dictionary *d, casenumber case_limit)
1347 d->case_limit = case_limit;
1350 /* Returns the prototype used for cases created by dictionary D. */
1351 const struct caseproto *
1352 dict_get_proto (const struct dictionary *d_)
1354 struct dictionary *d = CONST_CAST (struct dictionary *, d_);
1355 if (d->proto == NULL)
1359 d->proto = caseproto_create ();
1360 d->proto = caseproto_reserve (d->proto, d->n_vars);
1361 for (i = 0; i < d->n_vars; i++)
1362 d->proto = caseproto_set_width (d->proto,
1363 var_get_case_index (d->vars[i].var),
1364 var_get_width (d->vars[i].var));
1369 /* Returns the case index of the next value to be added to D.
1370 This value is the number of `union value's that need to be
1371 allocated to store a case for dictionary D. */
1373 dict_get_next_value_idx (const struct dictionary *d)
1375 return d->next_value_idx;
1378 /* Returns the number of bytes needed to store a case for
1381 dict_get_case_size (const struct dictionary *d)
1383 return sizeof (union value) * dict_get_next_value_idx (d);
1386 /* Reassigns values in dictionary D so that fragmentation is
1389 dict_compact_values (struct dictionary *d)
1393 d->next_value_idx = 0;
1394 for (i = 0; i < d->n_vars; i++)
1396 struct variable *v = d->vars[i].var;
1397 set_var_case_index (v, d->next_value_idx++);
1399 invalidate_proto (d);
1402 /* Returns the number of values occupied by the variables in
1403 dictionary D. All variables are considered if EXCLUDE_CLASSES
1404 is 0, or it may contain one or more of (1u << DC_ORDINARY),
1405 (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to exclude the
1406 corresponding type of variable.
1408 The return value may be less than the number of values in one
1409 of dictionary D's cases (as returned by
1410 dict_get_next_value_idx) even if E is 0, because there may be
1411 gaps in D's cases due to deleted variables. */
1413 dict_count_values (const struct dictionary *d, unsigned int exclude_classes)
1415 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1417 | (1u << DC_SCRATCH))) == 0);
1420 for (size_t i = 0; i < d->n_vars; i++)
1422 enum dict_class class = var_get_dict_class (d->vars[i].var);
1423 if (!(exclude_classes & (1u << class)))
1429 /* Returns the case prototype that would result after deleting
1430 all variables from D that are not in one of the
1431 EXCLUDE_CLASSES and compacting the dictionary with
1434 The caller must unref the returned caseproto when it is no
1437 dict_get_compacted_proto (const struct dictionary *d,
1438 unsigned int exclude_classes)
1440 struct caseproto *proto;
1443 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1445 | (1u << DC_SCRATCH))) == 0);
1447 proto = caseproto_create ();
1448 for (i = 0; i < d->n_vars; i++)
1450 struct variable *v = d->vars[i].var;
1451 if (!(exclude_classes & (1u << var_get_dict_class (v))))
1452 proto = caseproto_add_width (proto, var_get_width (v));
1456 /* Returns the file label for D, or a null pointer if D is
1457 unlabeled (see cmd_file_label()). */
1459 dict_get_label (const struct dictionary *d)
1464 /* Sets D's file label to LABEL, truncating it to at most 60 bytes in D's
1467 Removes D's label if LABEL is null or the empty string. */
1469 dict_set_label (struct dictionary *d, const char *label)
1472 if (label == NULL || label[0] == '\0')
1475 d->label = utf8_encoding_trunc (label, d->encoding, 60);
1478 /* Returns the documents for D, as an UTF-8 encoded string_array. The
1479 return value is always nonnull; if there are no documents then the
1480 string_arary is empty.*/
1481 const struct string_array *
1482 dict_get_documents (const struct dictionary *d)
1484 return &d->documents;
1487 /* Replaces the documents for D by NEW_DOCS, a UTF-8 encoded string_array. */
1489 dict_set_documents (struct dictionary *d, const struct string_array *new_docs)
1491 /* Swap out the old documents, instead of destroying them immediately, to
1492 allow the new documents to include pointers into the old ones. */
1493 struct string_array old_docs = STRING_ARRAY_INITIALIZER;
1494 string_array_swap (&d->documents, &old_docs);
1496 for (size_t i = 0; i < new_docs->n; i++)
1497 dict_add_document_line (d, new_docs->strings[i], false);
1499 string_array_destroy (&old_docs);
1502 /* Replaces the documents for D by UTF-8 encoded string NEW_DOCS, dividing it
1503 into individual lines at new-line characters. Each line is truncated to at
1504 most DOC_LINE_LENGTH bytes in D's encoding. */
1506 dict_set_documents_string (struct dictionary *d, const char *new_docs)
1510 dict_clear_documents (d);
1511 for (s = new_docs; *s != '\0';)
1513 size_t len = strcspn (s, "\n");
1514 char *line = xmemdup0 (s, len);
1515 dict_add_document_line (d, line, false);
1524 /* Drops the documents from dictionary D. */
1526 dict_clear_documents (struct dictionary *d)
1528 string_array_clear (&d->documents);
1531 /* Appends the UTF-8 encoded LINE to the documents in D. LINE will be
1532 truncated so that it is no more than 80 bytes in the dictionary's
1533 encoding. If this causes some text to be lost, and ISSUE_WARNING is true,
1534 then a warning will be issued. */
1536 dict_add_document_line (struct dictionary *d, const char *line,
1542 trunc_len = utf8_encoding_trunc_len (line, d->encoding, DOC_LINE_LENGTH);
1543 truncated = line[trunc_len] != '\0';
1544 if (truncated && issue_warning)
1546 /* TRANSLATORS: "bytes" is correct, not characters due to UTF encoding */
1547 msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
1550 string_array_append_nocopy (&d->documents, xmemdup0 (line, trunc_len));
1555 /* Returns the number of document lines in dictionary D. */
1557 dict_get_document_n_lines (const struct dictionary *d)
1559 return d->documents.n;
1562 /* Returns document line number IDX in dictionary D. The caller must not
1563 modify or free the returned string. */
1565 dict_get_document_line (const struct dictionary *d, size_t idx)
1567 assert (idx < d->documents.n);
1568 return d->documents.strings[idx];
1571 /* Creates in D a vector named NAME that contains the N
1572 variables in VAR. Returns true if successful, or false if a
1573 vector named NAME already exists in D. */
1575 dict_create_vector (struct dictionary *d,
1577 struct variable **var, size_t n)
1580 for (size_t i = 0; i < n; i++)
1581 assert (dict_contains_var (d, var[i]));
1583 if (dict_lookup_vector (d, name) == NULL)
1585 d->vector = xnrealloc (d->vector, d->n_vectors + 1, sizeof *d->vector);
1586 d->vector[d->n_vectors++] = vector_create (name, var, n);
1593 /* Creates in D a vector named NAME that contains the N
1594 variables in VAR. A vector named NAME must not already exist
1597 dict_create_vector_assert (struct dictionary *d,
1599 struct variable **var, size_t n)
1601 assert (dict_lookup_vector (d, name) == NULL);
1602 dict_create_vector (d, name, var, n);
1605 /* Returns the vector in D with index IDX, which must be less
1606 than dict_get_n_vectors (D). */
1607 const struct vector *
1608 dict_get_vector (const struct dictionary *d, size_t idx)
1610 assert (idx < d->n_vectors);
1612 return d->vector[idx];
1615 /* Returns the number of vectors in D. */
1617 dict_get_n_vectors (const struct dictionary *d)
1619 return d->n_vectors;
1622 /* Looks up and returns the vector within D with the given
1624 const struct vector *
1625 dict_lookup_vector (const struct dictionary *d, const char *name)
1628 for (i = 0; i < d->n_vectors; i++)
1629 if (!utf8_strcasecmp (vector_get_name (d->vector[i]), name))
1630 return d->vector[i];
1634 /* Deletes all vectors from D. */
1636 dict_clear_vectors (struct dictionary *d)
1640 for (i = 0; i < d->n_vectors; i++)
1641 vector_destroy (d->vector[i]);
1648 /* Multiple response sets. */
1650 /* Returns the multiple response set in DICT with index IDX, which must be
1651 between 0 and the count returned by dict_get_n_mrsets(), exclusive. */
1652 const struct mrset *
1653 dict_get_mrset (const struct dictionary *dict, size_t idx)
1655 assert (idx < dict->n_mrsets);
1656 return dict->mrsets[idx];
1659 /* Returns the number of multiple response sets in DICT. */
1661 dict_get_n_mrsets (const struct dictionary *dict)
1663 return dict->n_mrsets;
1666 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1667 returns its index; otherwise, returns SIZE_MAX. */
1669 dict_lookup_mrset_idx (const struct dictionary *dict, const char *name)
1673 for (i = 0; i < dict->n_mrsets; i++)
1674 if (!utf8_strcasecmp (name, dict->mrsets[i]->name))
1680 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1681 returns it; otherwise, returns NULL. */
1682 const struct mrset *
1683 dict_lookup_mrset (const struct dictionary *dict, const char *name)
1685 size_t idx = dict_lookup_mrset_idx (dict, name);
1686 return idx != SIZE_MAX ? dict->mrsets[idx] : NULL;
1689 /* Adds MRSET to DICT, replacing any existing set with the same name. Returns
1690 true if a set was replaced, false if none existed with the specified name.
1692 Ownership of MRSET is transferred to DICT. */
1694 dict_add_mrset (struct dictionary *dict, struct mrset *mrset)
1698 assert (mrset_ok (mrset, dict));
1700 idx = dict_lookup_mrset_idx (dict, mrset->name);
1701 if (idx == SIZE_MAX)
1703 dict->mrsets = xrealloc (dict->mrsets,
1704 (dict->n_mrsets + 1) * sizeof *dict->mrsets);
1705 dict->mrsets[dict->n_mrsets++] = mrset;
1710 mrset_destroy (dict->mrsets[idx]);
1711 dict->mrsets[idx] = mrset;
1716 /* Looks for a multiple response set in DICT named NAME. If found, removes it
1717 from DICT and returns true. If none is found, returns false without
1720 Deleting one multiple response set causes the indexes of other sets within
1723 dict_delete_mrset (struct dictionary *dict, const char *name)
1725 size_t idx = dict_lookup_mrset_idx (dict, name);
1726 if (idx != SIZE_MAX)
1728 mrset_destroy (dict->mrsets[idx]);
1729 dict->mrsets[idx] = dict->mrsets[--dict->n_mrsets];
1736 /* Deletes all multiple response sets from DICT. */
1738 dict_clear_mrsets (struct dictionary *dict)
1742 for (i = 0; i < dict->n_mrsets; i++)
1743 mrset_destroy (dict->mrsets[i]);
1744 free (dict->mrsets);
1745 dict->mrsets = NULL;
1749 /* Removes VAR, which must be in DICT, from DICT's multiple response sets. */
1751 dict_unset_mrset_var (struct dictionary *dict, struct variable *var)
1755 assert (dict_contains_var (dict, var));
1757 for (i = 0; i < dict->n_mrsets;)
1759 struct mrset *mrset = dict->mrsets[i];
1762 for (j = 0; j < mrset->n_vars;)
1763 if (mrset->vars[j] == var)
1764 remove_element (mrset->vars, mrset->n_vars--,
1765 sizeof *mrset->vars, j);
1769 if (mrset->n_vars < 2)
1771 mrset_destroy (mrset);
1772 dict->mrsets[i] = dict->mrsets[--dict->n_mrsets];
1779 /* Returns D's attribute set. The caller may examine or modify
1780 the attribute set, but must not destroy it. Destroying D or
1781 calling dict_set_attributes for D will also destroy D's
1784 dict_get_attributes (const struct dictionary *d)
1786 return CONST_CAST (struct attrset *, &d->attributes);
1789 /* Replaces D's attributes set by a copy of ATTRS. */
1791 dict_set_attributes (struct dictionary *d, const struct attrset *attrs)
1793 attrset_destroy (&d->attributes);
1794 attrset_clone (&d->attributes, attrs);
1797 /* Returns true if D has at least one attribute in its attribute
1798 set, false if D's attribute set is empty. */
1800 dict_has_attributes (const struct dictionary *d)
1802 return attrset_count (&d->attributes) > 0;
1805 /* Called from variable.c to notify the dictionary that some property (indicated
1806 by WHAT) of the variable has changed. OLDVAR is a copy of V as it existed
1807 prior to the change. OLDVAR is destroyed by this function.
1810 dict_var_changed (const struct variable *v, unsigned int what, struct variable *oldvar)
1812 if (var_has_vardict (v))
1814 const struct vardict_info *vardict = var_get_vardict (v);
1815 struct dictionary *d = vardict->dict;
1820 if (what & (VAR_TRAIT_WIDTH | VAR_TRAIT_POSITION))
1821 invalidate_proto (d);
1823 if (d->changed) d->changed (d, d->changed_data);
1824 if (d->callbacks && d->callbacks->var_changed)
1825 d->callbacks->var_changed (d, var_get_dict_index (v), what, oldvar, d->cb_data);
1832 /* Dictionary used to contain "internal variables". */
1833 static struct dictionary *internal_dict;
1835 /* Create a variable of the specified WIDTH to be used for internal
1836 calculations only. The variable is assigned case index CASE_IDX. */
1838 dict_create_internal_var (int case_idx, int width)
1840 if (internal_dict == NULL)
1841 internal_dict = dict_create ("UTF-8");
1845 static int counter = INT_MAX / 2;
1846 struct variable *var;
1849 if (++counter == INT_MAX)
1850 counter = INT_MAX / 2;
1852 sprintf (name, "$internal%d", counter);
1853 var = dict_create_var (internal_dict, name, width);
1856 set_var_case_index (var, case_idx);
1862 /* Destroys VAR, which must have been created with
1863 dict_create_internal_var(). */
1865 dict_destroy_internal_var (struct variable *var)
1869 dict_delete_var (internal_dict, var);
1871 /* Destroy internal_dict if it has no variables left, just so that
1872 valgrind --leak-check --show-reachable won't show internal_dict. */
1873 if (dict_get_n_vars (internal_dict) == 0)
1875 dict_unref (internal_dict);
1876 internal_dict = NULL;
1882 vardict_get_dict_index (const struct vardict_info *vardict)
1884 return vardict - vardict->dict->vars;