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 /* Checks whether UTF-8 string ID is an acceptable identifier in DICT's
183 encoding. Returns true if it is, otherwise an error message that the caller
185 char * WARN_UNUSED_RESULT
186 dict_id_is_valid__ (const struct dictionary *dict, const char *id)
188 if (!dict->names_must_be_ids)
190 return id_is_valid__ (id, dict->encoding);
194 error_to_bool (char *error)
205 /* Returns true if UTF-8 string ID is an acceptable identifier in DICT's
206 encoding, false otherwise. */
208 dict_id_is_valid (const struct dictionary *dict, const char *id)
210 return error_to_bool (dict_id_is_valid__ (dict, id));
214 dict_set_change_callback (struct dictionary *d,
215 void (*changed) (struct dictionary *, void*),
218 d->changed = changed;
219 d->changed_data = data;
222 /* Discards dictionary D's caseproto. (It will be regenerated
223 lazily, on demand.) */
225 invalidate_proto (struct dictionary *d)
227 caseproto_unref (d->proto);
231 /* Print a representation of dictionary D to stdout, for
232 debugging purposes. */
234 dict_dump (const struct dictionary *d)
237 for (i = 0 ; i < d->n_vars ; ++i)
239 const struct variable *v = d->vars[i].var;
240 printf ("Name: %s;\tdict_idx: %zu; case_idx: %zu\n",
242 var_get_dict_index (v),
243 var_get_case_index (v));
248 /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
249 the dictionary or any of the variables it contains are modified.
250 Each callback will get passed CALLBACK_DATA.
251 Any callback may be NULL, in which case it'll be ignored.
254 dict_set_callbacks (struct dictionary *dict,
255 const struct dict_callbacks *callbacks,
258 dict->callbacks = callbacks;
259 dict->cb_data = callback_data;
262 /* Shallow copy the callbacks from SRC to DEST */
264 dict_copy_callbacks (struct dictionary *dest,
265 const struct dictionary *src)
267 dest->callbacks = src->callbacks;
268 dest->cb_data = src->cb_data;
271 /* Creates and returns a new dictionary with the specified ENCODING. */
273 dict_create (const char *encoding)
275 struct dictionary *d = xmalloc (sizeof *d);
277 *d = (struct dictionary) {
278 .encoding = xstrdup (encoding),
279 .names_must_be_ids = true,
280 .name_map = HMAP_INITIALIZER (d->name_map),
281 .attributes = ATTRSET_INITIALIZER (d->attributes),
282 .split_type = SPLIT_NONE,
289 /* Creates and returns a (deep) copy of an existing
292 The new dictionary's case indexes are copied from the old
293 dictionary. If the new dictionary won't be used to access
294 cases produced with the old dictionary, then the new
295 dictionary's case indexes should be compacted with
296 dict_compact_values to save space.
298 Callbacks are not cloned. */
300 dict_clone (const struct dictionary *s)
302 struct dictionary *d;
305 d = dict_create (s->encoding);
306 dict_set_names_must_be_ids (d, dict_get_names_must_be_ids (s));
308 for (i = 0; i < s->n_vars; i++)
310 struct variable *sv = s->vars[i].var;
311 struct variable *dv = dict_clone_var_assert (d, sv);
313 for (size_t j = 0; j < var_get_n_short_names (sv); j++)
314 var_set_short_name (dv, j, var_get_short_name (sv, j));
316 var_get_vardict (dv)->case_index = var_get_vardict (sv)->case_index;
319 d->next_value_idx = s->next_value_idx;
321 d->n_splits = s->n_splits;
324 d->split = xnmalloc (d->n_splits, sizeof *d->split);
325 for (i = 0; i < d->n_splits; i++)
326 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
328 d->split_type = s->split_type;
330 if (s->weight != NULL)
331 dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight)));
333 if (s->filter != NULL)
334 dict_set_filter (d, dict_lookup_var_assert (d, var_get_name (s->filter)));
336 d->case_limit = s->case_limit;
337 dict_set_label (d, dict_get_label (s));
338 dict_set_documents (d, dict_get_documents (s));
340 d->n_vectors = s->n_vectors;
341 d->vector = xnmalloc (d->n_vectors, sizeof *d->vector);
342 for (i = 0; i < s->n_vectors; i++)
343 d->vector[i] = vector_clone (s->vector[i], s, d);
345 dict_set_attributes (d, dict_get_attributes (s));
347 for (i = 0; i < s->n_mrsets; i++)
349 const struct mrset *old = s->mrsets[i];
353 /* Clone old mrset, then replace vars from D by vars from S. */
354 new = mrset_clone (old);
355 for (j = 0; j < new->n_vars; j++)
356 new->vars[j] = dict_lookup_var_assert (d, var_get_name (new->vars[j]));
358 dict_add_mrset (d, new);
366 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
367 dict_get_n_splits() to determine how many SPLIT FILE vars
368 there are. Returns a null pointer if and only if there are no
370 const struct variable *const *
371 dict_get_split_vars (const struct dictionary *d)
376 /* Returns the number of SPLIT FILE vars. */
378 dict_get_n_splits (const struct dictionary *d)
383 /* Removes variable V, which must be in D, from D's set of split
386 dict_unset_split_var (struct dictionary *d, struct variable *v, bool skip_callbacks)
390 assert (dict_contains_var (d, v));
392 orig_count = d->n_splits;
393 d->n_splits = remove_equal (d->split, d->n_splits, sizeof *d->split,
394 &v, compare_var_ptrs, NULL);
395 if (orig_count != d->n_splits && !skip_callbacks)
397 if (d->changed) d->changed (d, d->changed_data);
398 /* We changed the set of split variables so invoke the
400 if (d->callbacks && d->callbacks->split_changed)
401 d->callbacks->split_changed (d, d->cb_data);
406 /* Sets N split vars SPLIT in dictionary D. N is silently capped to a maximum
409 dict_set_split_vars__ (struct dictionary *d,
410 struct variable *const *split, size_t n,
411 enum split_type type, bool skip_callbacks)
415 assert (n == 0 || split != NULL);
418 d->split_type = (n == 0 ? SPLIT_NONE
419 : type == SPLIT_NONE ? SPLIT_LAYERED
423 d->split = xnrealloc (d->split, n, sizeof *d->split) ;
424 memcpy (d->split, split, n * sizeof *d->split);
434 if (d->changed) d->changed (d, d->changed_data);
435 if (d->callbacks && d->callbacks->split_changed)
436 d->callbacks->split_changed (d, d->cb_data);
441 dict_get_split_type (const struct dictionary *d)
443 return d->split_type;
446 /* Sets N split vars SPLIT in dictionary D. */
448 dict_set_split_vars (struct dictionary *d,
449 struct variable *const *split, size_t n,
450 enum split_type type)
452 dict_set_split_vars__ (d, split, n, type, false);
456 dict_clear_split_vars (struct dictionary *d)
458 dict_set_split_vars (d, NULL, 0, SPLIT_NONE);
462 /* Deletes variable V from dictionary D and frees V.
464 This is a very bad idea if there might be any pointers to V
465 from outside D. In general, no variable in the active dataset's
466 dictionary should be deleted when any transformations are
467 active on the dictionary's dataset, because those
468 transformations might reference the deleted variable. The
469 safest time to delete a variable is just after a procedure has
470 been executed, as done by DELETE VARIABLES.
472 Pointers to V within D are not a problem, because
473 dict_delete_var() knows to remove V from split variables,
474 weights, filters, etc. */
476 dict_delete_var__ (struct dictionary *d, struct variable *v, bool skip_callbacks)
478 int dict_index = var_get_dict_index (v);
479 const int case_index = var_get_case_index (v);
481 assert (dict_contains_var (d, v));
483 dict_unset_split_var (d, v, skip_callbacks);
484 dict_unset_mrset_var (d, v);
487 dict_set_weight (d, NULL);
490 dict_set_filter (d, NULL);
492 dict_clear_vectors (d);
494 /* Remove V from var array. */
495 unindex_vars (d, dict_index, d->n_vars);
496 remove_element (d->vars, d->n_vars, sizeof *d->vars, dict_index);
499 /* Update dict_index for each affected variable. */
500 reindex_vars (d, dict_index, d->n_vars, skip_callbacks);
503 var_clear_vardict (v);
505 if (! skip_callbacks)
507 if (d->changed) d->changed (d, d->changed_data);
508 if (d->callbacks && d->callbacks->var_deleted)
509 d->callbacks->var_deleted (d, v, dict_index, case_index, d->cb_data);
512 invalidate_proto (d);
516 /* Deletes variable V from dictionary D and frees V.
518 This is a very bad idea if there might be any pointers to V
519 from outside D. In general, no variable in the active dataset's
520 dictionary should be deleted when any transformations are
521 active on the dictionary's dataset, because those
522 transformations might reference the deleted variable. The
523 safest time to delete a variable is just after a procedure has
524 been executed, as done by DELETE VARIABLES.
526 Pointers to V within D are not a problem, because
527 dict_delete_var() knows to remove V from split variables,
528 weights, filters, etc. */
530 dict_delete_var (struct dictionary *d, struct variable *v)
532 dict_delete_var__ (d, v, false);
536 /* Deletes the COUNT variables listed in VARS from D. This is
537 unsafe; see the comment on dict_delete_var() for details. */
539 dict_delete_vars (struct dictionary *d,
540 struct variable *const *vars, size_t count)
542 /* FIXME: this can be done in O(count) time, but this algorithm
544 assert (count == 0 || vars != NULL);
547 dict_delete_var (d, *vars++);
550 /* Deletes the COUNT variables in D starting at index IDX. This
551 is unsafe; see the comment on dict_delete_var() for
552 details. Deleting consecutive vars will result in less callbacks
553 compared to iterating over dict_delete_var.
554 A simple while loop over dict_delete_var will
555 produce (d->n_vars - IDX) * COUNT variable changed callbacks
556 plus COUNT variable delete callbacks.
557 This here produces d->n_vars - IDX variable changed callbacks
558 plus COUNT variable delete callbacks. */
560 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
562 assert (idx + count <= d->n_vars);
564 /* We need to store the variable and the corresponding case_index
565 for the delete callbacks later. We store them in a linked list.*/
568 struct variable *var;
571 struct ll_list list = LL_INITIALIZER (list);
573 for (size_t i = idx; i < idx + count; i++)
575 struct delvar *dv = xmalloc (sizeof (struct delvar));
577 struct variable *v = d->vars[i].var;
579 dict_unset_split_var (d, v, false);
580 dict_unset_mrset_var (d, v);
583 dict_set_weight (d, NULL);
586 dict_set_filter (d, NULL);
589 dv->case_index = var_get_case_index (v);
590 ll_push_tail (&list, (struct ll *)dv);
593 dict_clear_vectors (d);
595 /* Remove variables from var array. */
596 unindex_vars (d, idx, d->n_vars);
597 remove_range (d->vars, d->n_vars, sizeof *d->vars, idx, count);
600 /* Reindexing will result variable-changed callback */
601 reindex_vars (d, idx, d->n_vars, false);
603 invalidate_proto (d);
604 if (d->changed) d->changed (d, d->changed_data);
606 /* Now issue the variable delete callbacks and delete
607 the variables. The vardict is not valid at this point
608 anymore. That is the reason why we stored the
609 caseindex before reindexing. */
610 for (size_t vi = idx; vi < idx + count; vi++)
612 struct delvar *dv = (struct delvar *) ll_pop_head (&list);
613 var_clear_vardict (dv->var);
614 if (d->callbacks && d->callbacks->var_deleted)
615 d->callbacks->var_deleted (d, dv->var, vi, dv->case_index, d->cb_data);
621 /* Deletes scratch variables from dictionary D. */
623 dict_delete_scratch_vars (struct dictionary *d)
627 /* FIXME: this can be done in O(count) time, but this algorithm
629 for (i = 0; i < d->n_vars;)
630 if (var_get_dict_class (d->vars[i].var) == DC_SCRATCH)
631 dict_delete_var (d, d->vars[i].var);
638 /* Clears the contents from a dictionary without destroying the
639 dictionary itself. */
641 dict_clear__ (struct dictionary *d, bool skip_callbacks)
643 /* FIXME? Should we really clear case_limit, label, documents?
644 Others are necessarily cleared by deleting all the variables.*/
645 while (d->n_vars > 0)
647 dict_delete_var__ (d, d->vars[d->n_vars - 1].var, skip_callbacks);
652 d->n_vars = d->allocated_vars = 0;
653 invalidate_proto (d);
654 hmap_clear (&d->name_map);
655 d->next_value_idx = 0;
656 dict_set_split_vars__ (d, NULL, 0, SPLIT_NONE, skip_callbacks);
665 dict_set_weight (d, NULL);
666 dict_set_filter (d, NULL);
671 string_array_clear (&d->documents);
672 dict_clear_vectors (d);
673 attrset_clear (&d->attributes);
676 /* Clears the contents from a dictionary without destroying the
677 dictionary itself. */
679 dict_clear (struct dictionary *d)
681 dict_clear__ (d, false);
684 /* Clears a dictionary and destroys it. */
686 _dict_destroy (struct dictionary *d)
688 /* In general, we don't want callbacks occurring, if the dictionary
689 is being destroyed */
690 d->callbacks = NULL ;
692 dict_clear__ (d, true);
693 string_array_destroy (&d->documents);
694 hmap_destroy (&d->name_map);
695 attrset_destroy (&d->attributes);
696 dict_clear_mrsets (d);
702 dict_ref (struct dictionary *d)
709 dict_unref (struct dictionary *d)
714 assert (d->ref_cnt >= 0);
719 /* Returns the number of variables in D. */
721 dict_get_n_vars (const struct dictionary *d)
726 /* Returns the variable in D with dictionary index IDX, which
727 must be between 0 and the count returned by
728 dict_get_n_vars(), exclusive. */
730 dict_get_var (const struct dictionary *d, size_t idx)
732 assert (idx < d->n_vars);
734 return d->vars[idx].var;
737 /* Sets *VARS to an array of pointers to variables in D and *N
738 to the number of variables in *D. All variables are returned
739 except for those, if any, in the classes indicated by EXCLUDE.
740 (There is no point in putting DC_SYSTEM in EXCLUDE as
741 dictionaries never include system variables.) */
743 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
744 size_t *n, enum dict_class exclude)
746 dict_get_vars_mutable (d, (struct variable ***) vars, n, exclude);
749 /* Sets *VARS to an array of pointers to variables in D and *N
750 to the number of variables in *D. All variables are returned
751 except for those, if any, in the classes indicated by EXCLUDE.
752 (There is no point in putting DC_SYSTEM in EXCLUDE as
753 dictionaries never include system variables.) */
755 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
756 size_t *n, enum dict_class exclude)
761 assert (exclude == (exclude & DC_ALL));
764 for (i = 0; i < d->n_vars; i++)
766 enum dict_class class = var_get_dict_class (d->vars[i].var);
767 if (!(class & exclude))
771 *vars = xnmalloc (count, sizeof **vars);
773 for (i = 0; i < d->n_vars; i++)
775 enum dict_class class = var_get_dict_class (d->vars[i].var);
776 if (!(class & exclude))
777 (*vars)[(*n)++] = d->vars[i].var;
779 assert (*n == count);
782 static struct variable *
783 add_var_with_case_index (struct dictionary *d, struct variable *v,
786 struct vardict_info *vardict;
788 assert (case_index >= d->next_value_idx);
790 /* Update dictionary. */
791 if (d->n_vars >= d->allocated_vars)
795 d->vars = x2nrealloc (d->vars, &d->allocated_vars, sizeof *d->vars);
796 hmap_clear (&d->name_map);
797 for (i = 0; i < d->n_vars; i++)
799 var_set_vardict (d->vars[i].var, &d->vars[i]);
800 hmap_insert_fast (&d->name_map, &d->vars[i].name_node,
801 d->vars[i].name_node.hash);
805 vardict = &d->vars[d->n_vars++];
808 hmap_insert (&d->name_map, &vardict->name_node,
809 utf8_hash_case_string (var_get_name (v), 0));
810 vardict->case_index = case_index;
811 var_set_vardict (v, vardict);
813 if (d->changed) d->changed (d, d->changed_data);
814 if (d->callbacks && d->callbacks->var_added)
815 d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
817 invalidate_proto (d);
818 d->next_value_idx = case_index + 1;
823 static struct variable *
824 add_var (struct dictionary *d, struct variable *v)
826 return add_var_with_case_index (d, v, d->next_value_idx);
829 /* Creates and returns a new variable in D with the given NAME
830 and WIDTH. Returns a null pointer if the given NAME would
831 duplicate that of an existing variable in the dictionary. */
833 dict_create_var (struct dictionary *d, const char *name, int width)
835 return (dict_lookup_var (d, name) == NULL
836 ? dict_create_var_assert (d, name, width)
840 /* Creates and returns a new variable in D with the given NAME
841 and WIDTH. Assert-fails if the given NAME would duplicate
842 that of an existing variable in the dictionary. */
844 dict_create_var_assert (struct dictionary *d, const char *name, int width)
846 assert (dict_lookup_var (d, name) == NULL);
847 return add_var (d, var_create (name, width));
850 /* Creates and returns a new variable in D, as a copy of existing variable
851 OLD_VAR, which need not be in D or in any dictionary. Returns a null
852 pointer if OLD_VAR's name would duplicate that of an existing variable in
855 dict_clone_var (struct dictionary *d, const struct variable *old_var)
857 return dict_clone_var_as (d, old_var, var_get_name (old_var));
860 /* Creates and returns a new variable in D, as a copy of existing variable
861 OLD_VAR, which need not be in D or in any dictionary. Assert-fails if
862 OLD_VAR's name would duplicate that of an existing variable in the
865 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var)
867 return dict_clone_var_as_assert (d, old_var, var_get_name (old_var));
870 /* Creates and returns a new variable in D with name NAME, as a copy of
871 existing variable OLD_VAR, which need not be in D or in any dictionary.
872 Returns a null pointer if the given NAME would duplicate that of an existing
873 variable in the dictionary. */
875 dict_clone_var_as (struct dictionary *d, const struct variable *old_var,
878 return (dict_lookup_var (d, name) == NULL
879 ? dict_clone_var_as_assert (d, old_var, name)
883 /* Creates and returns a new variable in D with name NAME, as a copy of
884 existing variable OLD_VAR, which need not be in D or in any dictionary.
885 Assert-fails if the given NAME would duplicate that of an existing variable
886 in the dictionary. */
888 dict_clone_var_as_assert (struct dictionary *d, const struct variable *old_var,
891 struct variable *new_var = var_clone (old_var);
892 assert (dict_lookup_var (d, name) == NULL);
893 var_set_name (new_var, name);
894 return add_var (d, new_var);
898 dict_clone_var_in_place_assert (struct dictionary *d,
899 const struct variable *old_var)
901 assert (dict_lookup_var (d, var_get_name (old_var)) == NULL);
902 return add_var_with_case_index (d, var_clone (old_var),
903 var_get_case_index (old_var));
906 /* Returns the variable named NAME in D, or a null pointer if no
907 variable has that name. */
909 dict_lookup_var (const struct dictionary *d, const char *name)
911 struct vardict_info *vardict;
913 HMAP_FOR_EACH_WITH_HASH (vardict, struct vardict_info, name_node,
914 utf8_hash_case_string (name, 0), &d->name_map)
916 struct variable *var = vardict->var;
917 if (!utf8_strcasecmp (var_get_name (var), name))
924 /* Returns the variable named NAME in D. Assert-fails if no
925 variable has that name. */
927 dict_lookup_var_assert (const struct dictionary *d, const char *name)
929 struct variable *v = dict_lookup_var (d, name);
934 /* Returns true if variable V is in dictionary D,
937 dict_contains_var (const struct dictionary *d, const struct variable *v)
939 return (var_has_vardict (v)
940 && vardict_get_dictionary (var_get_vardict (v)) == d);
943 /* Moves V to 0-based position IDX in D. Other variables in D,
944 if any, retain their relative positions. Runs in time linear
945 in the distance moved. */
947 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
949 size_t old_index = var_get_dict_index (v);
951 assert (new_index < d->n_vars);
953 unindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
954 move_element (d->vars, d->n_vars, sizeof *d->vars, old_index, new_index);
955 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1, false);
958 /* Reorders the variables in D, placing the COUNT variables
959 listed in ORDER in that order at the beginning of D. The
960 other variables in D, if any, retain their relative
963 dict_reorder_vars (struct dictionary *d,
964 struct variable *const *order, size_t count)
966 struct vardict_info *new_var;
969 assert (count == 0 || order != NULL);
970 assert (count <= d->n_vars);
972 new_var = xnmalloc (d->allocated_vars, sizeof *new_var);
974 /* Add variables in ORDER to new_var. */
975 for (i = 0; i < count; i++)
977 struct vardict_info *old_var;
979 assert (dict_contains_var (d, order[i]));
981 old_var = var_get_vardict (order[i]);
982 new_var[i] = *old_var;
983 old_var->dict = NULL;
986 /* Add remaining variables to new_var. */
987 for (i = 0; i < d->n_vars; i++)
988 if (d->vars[i].dict != NULL)
989 new_var[count++] = d->vars[i];
990 assert (count == d->n_vars);
992 /* Replace old vardicts by new ones. */
996 hmap_clear (&d->name_map);
997 reindex_vars (d, 0, d->n_vars, false);
1000 /* Changes the name of variable V that is currently in a dictionary to
1003 rename_var (struct variable *v, const char *new_name)
1005 struct vardict_info *vardict = var_get_vardict (v);
1006 var_clear_vardict (v);
1007 var_set_name (v, new_name);
1008 vardict->name_node.hash = utf8_hash_case_string (new_name, 0);
1009 var_set_vardict (v, vardict);
1012 /* Tries to changes the name of V in D to name NEW_NAME. Returns true if
1013 successful, false if a variable (other than V) with the given name already
1016 dict_try_rename_var (struct dictionary *d, struct variable *v,
1017 const char *new_name)
1019 struct variable *conflict = dict_lookup_var (d, new_name);
1020 if (conflict && v != conflict)
1023 struct variable *old = var_clone (v);
1024 unindex_var (d, var_get_vardict (v));
1025 rename_var (v, new_name);
1026 reindex_var (d, var_get_vardict (v), false);
1028 if (settings_get_algorithm () == ENHANCED)
1029 var_clear_short_names (v);
1031 if (d->changed) d->changed (d, d->changed_data);
1032 if (d->callbacks && d->callbacks->var_changed)
1033 d->callbacks->var_changed (d, var_get_dict_index (v), VAR_TRAIT_NAME, old, d->cb_data);
1040 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
1041 a variable named NEW_NAME is already in D, except that
1042 NEW_NAME may be the same as V's existing name. */
1044 dict_rename_var (struct dictionary *d, struct variable *v,
1045 const char *new_name)
1047 bool ok UNUSED = dict_try_rename_var (d, v, new_name);
1051 /* Renames COUNT variables specified in VARS to the names given
1052 in NEW_NAMES within dictionary D. If the renaming would
1053 result in a duplicate variable name, returns false and stores a
1054 name that would be duplicated into *ERR_NAME (if ERR_NAME is
1055 non-null). Otherwise, the renaming is successful, and true
1058 dict_rename_vars (struct dictionary *d,
1059 struct variable **vars, char **new_names, size_t count,
1066 assert (count == 0 || vars != NULL);
1067 assert (count == 0 || new_names != NULL);
1069 /* Save the names of the variables to be renamed. */
1070 pool = pool_create ();
1071 old_names = pool_nalloc (pool, count, sizeof *old_names);
1072 for (i = 0; i < count; i++)
1073 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
1075 /* Remove the variables to be renamed from the name hash,
1077 for (i = 0; i < count; i++)
1079 unindex_var (d, var_get_vardict (vars[i]));
1080 rename_var (vars[i], new_names[i]);
1083 /* Add the renamed variables back into the name hash,
1084 checking for conflicts. */
1085 for (i = 0; i < count; i++)
1087 if (dict_lookup_var (d, var_get_name (vars[i])) != NULL)
1089 /* There is a name conflict.
1090 Back out all the name changes that have already
1091 taken place, and indicate failure. */
1092 size_t fail_idx = i;
1093 if (err_name != NULL)
1094 *err_name = new_names[i];
1096 for (i = 0; i < fail_idx; i++)
1097 unindex_var (d, var_get_vardict (vars[i]));
1099 for (i = 0; i < count; i++)
1101 rename_var (vars[i], old_names[i]);
1102 reindex_var (d, var_get_vardict (vars[i]), false);
1105 pool_destroy (pool);
1108 reindex_var (d, var_get_vardict (vars[i]), false);
1111 /* Clear short names. */
1112 if (settings_get_algorithm () == ENHANCED)
1113 for (i = 0; i < count; i++)
1114 var_clear_short_names (vars[i]);
1116 pool_destroy (pool);
1120 /* Returns true if a variable named NAME may be inserted in DICT;
1121 that is, if there is not already a variable with that name in
1122 DICT and if NAME is not a reserved word. (The caller's checks
1123 have already verified that NAME is otherwise acceptable as a
1126 var_name_is_insertable (const struct dictionary *dict, const char *name)
1128 return (dict_lookup_var (dict, name) == NULL
1129 && lex_id_to_token (ss_cstr (name)) == T_ID);
1133 make_hinted_name (const struct dictionary *dict, const char *hint)
1135 size_t hint_len = strlen (hint);
1136 bool dropped = false;
1141 if (hint_len > ID_MAX_LEN)
1142 hint_len = ID_MAX_LEN;
1144 /* The allocation size here is OK: characters that are copied directly fit
1145 OK, and characters that are not copied directly are replaced by a single
1146 '_' byte. If u8_mbtouc() replaces bad input by 0xfffd, then that will get
1147 replaced by '_' too. */
1148 root = rp = xmalloc (hint_len + 1);
1149 for (ofs = 0; ofs < hint_len; ofs += mblen)
1153 mblen = u8_mbtouc (&uc, CHAR_CAST (const uint8_t *, hint + ofs),
1156 ? lex_uc_is_id1 (uc) && uc != '$'
1157 : lex_uc_is_idn (uc))
1164 rp += u8_uctomb (CHAR_CAST (uint8_t *, rp), uc, 6);
1166 else if (rp != root)
1171 if (root[0] != '\0')
1173 unsigned long int i;
1175 if (var_name_is_insertable (dict, root))
1178 for (i = 0; i < ULONG_MAX; i++)
1180 char suffix[INT_BUFSIZE_BOUND (i) + 1];
1184 if (!str_format_26adic (i + 1, true, &suffix[1], sizeof suffix - 1))
1187 name = utf8_encoding_concat (root, suffix, dict->encoding, 64);
1188 if (var_name_is_insertable (dict, name))
1203 make_numeric_name (const struct dictionary *dict, unsigned long int *num_start)
1205 unsigned long int number;
1207 for (number = num_start != NULL ? MAX (*num_start, 1) : 1;
1211 char name[3 + INT_STRLEN_BOUND (number) + 1];
1213 sprintf (name, "VAR%03lu", number);
1214 if (dict_lookup_var (dict, name) == NULL)
1216 if (num_start != NULL)
1217 *num_start = number + 1;
1218 return xstrdup (name);
1226 /* Devises and returns a variable name unique within DICT. The variable name
1227 is owned by the caller, which must free it with free() when it is no longer
1230 HINT, if it is non-null, is used as a suggestion that will be
1231 modified for suitability as a variable name and for
1234 If HINT is null or entirely unsuitable, a name in the form
1235 "VAR%03d" will be generated, where the smallest unused integer
1236 value is used. If NUM_START is non-null, then its value is
1237 used as the minimum numeric value to check, and it is updated
1238 to the next value to be checked.
1241 dict_make_unique_var_name (const struct dictionary *dict, const char *hint,
1242 unsigned long int *num_start)
1246 char *hinted_name = make_hinted_name (dict, hint);
1247 if (hinted_name != NULL)
1251 return make_numeric_name (dict, num_start);
1254 /* Returns whether variable names must be valid identifiers. Normally, this is
1255 true, but sometimes a dictionary is prepared for external use (e.g. output
1256 to a CSV file) where names don't have to be valid. */
1258 dict_get_names_must_be_ids (const struct dictionary *d)
1260 return d->names_must_be_ids;
1263 /* Sets whether variable names must be valid identifiers. Normally, this is
1264 true, but sometimes a dictionary is prepared for external use (e.g. output
1265 to a CSV file) where names don't have to be valid.
1267 Changing this setting from false to true doesn't make the dictionary check
1268 all the existing variable names, so it can cause an invariant violation. */
1270 dict_set_names_must_be_ids (struct dictionary *d, bool names_must_be_ids)
1272 d->names_must_be_ids = names_must_be_ids;
1275 /* Returns the weighting variable in dictionary D, or a null
1276 pointer if the dictionary is unweighted. */
1278 dict_get_weight (const struct dictionary *d)
1280 assert (d->weight == NULL || dict_contains_var (d, d->weight));
1285 /* Returns the value of D's weighting variable in case C, except
1286 that a negative or missing weight is returned as 0. Returns 1 if the
1287 dictionary is unweighted. Will warn about missing, negative,
1288 or zero values if *WARN_ON_INVALID is true. The function will
1289 set *WARN_ON_INVALID to false if an invalid weight is
1292 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
1293 bool *warn_on_invalid)
1297 if (d->weight == NULL)
1301 double w = case_num (c, d->weight);
1303 return var_force_valid_weight (d->weight, w, warn_on_invalid);
1307 /* Like dict_get_case_weight(), but additionally rounds each weight to the
1310 dict_get_rounded_case_weight (const struct dictionary *d,
1311 const struct ccase *c, bool *warn_on_invalid)
1313 return floor (dict_get_case_weight (d, c, warn_on_invalid) + 0.5);
1316 /* Returns the format to use for weights. */
1318 dict_get_weight_format (const struct dictionary *d)
1320 return d->weight ? var_get_print_format (d->weight) : F_8_0;
1323 /* Sets the weighting variable of D to V, or turning off
1324 weighting if V is a null pointer. */
1326 dict_set_weight (struct dictionary *d, struct variable *v)
1328 assert (v == NULL || dict_contains_var (d, v));
1329 assert (v == NULL || var_is_numeric (v));
1333 if (d->changed) d->changed (d, d->changed_data);
1334 if (d->callbacks && d->callbacks->weight_changed)
1335 d->callbacks->weight_changed (d,
1336 v ? var_get_dict_index (v) : -1,
1340 /* Returns the filter variable in dictionary D (see cmd_filter())
1341 or a null pointer if the dictionary is unfiltered. */
1343 dict_get_filter (const struct dictionary *d)
1345 assert (d->filter == NULL || dict_contains_var (d, d->filter));
1350 /* Sets V as the filter variable for dictionary D. Passing a
1351 null pointer for V turn off filtering. */
1353 dict_set_filter (struct dictionary *d, struct variable *v)
1355 assert (v == NULL || dict_contains_var (d, v));
1356 assert (v == NULL || var_is_numeric (v));
1360 if (d->changed) d->changed (d, d->changed_data);
1361 if (d->callbacks && d->callbacks->filter_changed)
1362 d->callbacks->filter_changed (d,
1363 v ? var_get_dict_index (v) : -1,
1367 /* Returns the case limit for dictionary D, or zero if the number
1368 of cases is unlimited. */
1370 dict_get_case_limit (const struct dictionary *d)
1372 return d->case_limit;
1375 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
1376 0 for CASE_LIMIT to indicate no limit. */
1378 dict_set_case_limit (struct dictionary *d, casenumber case_limit)
1380 d->case_limit = case_limit;
1383 /* Returns the prototype used for cases created by dictionary D. */
1384 const struct caseproto *
1385 dict_get_proto (const struct dictionary *d_)
1387 struct dictionary *d = CONST_CAST (struct dictionary *, d_);
1388 if (d->proto == NULL)
1392 d->proto = caseproto_create ();
1393 d->proto = caseproto_reserve (d->proto, d->n_vars);
1394 for (i = 0; i < d->n_vars; i++)
1395 d->proto = caseproto_set_width (d->proto,
1396 var_get_case_index (d->vars[i].var),
1397 var_get_width (d->vars[i].var));
1402 /* Returns the case index of the next value to be added to D.
1403 This value is the number of `union value's that need to be
1404 allocated to store a case for dictionary D. */
1406 dict_get_next_value_idx (const struct dictionary *d)
1408 return d->next_value_idx;
1411 /* Returns the number of bytes needed to store a case for
1414 dict_get_case_size (const struct dictionary *d)
1416 return sizeof (union value) * dict_get_next_value_idx (d);
1419 /* Reassigns values in dictionary D so that fragmentation is
1422 dict_compact_values (struct dictionary *d)
1426 d->next_value_idx = 0;
1427 for (i = 0; i < d->n_vars; i++)
1429 struct variable *v = d->vars[i].var;
1430 set_var_case_index (v, d->next_value_idx++);
1432 invalidate_proto (d);
1435 /* Returns the number of values occupied by the variables in
1436 dictionary D. All variables are considered if EXCLUDE_CLASSES
1437 is 0, or it may contain one or more of DC_ORDINARY, DC_SYSTEM,
1438 or DC_SCRATCH to exclude the corresponding type of variable.
1440 The return value may be less than the number of values in one
1441 of dictionary D's cases (as returned by
1442 dict_get_next_value_idx) even if E is 0, because there may be
1443 gaps in D's cases due to deleted variables. */
1445 dict_count_values (const struct dictionary *d, unsigned int exclude_classes)
1447 assert (!(exclude_classes & ~DC_ALL));
1450 for (size_t i = 0; i < d->n_vars; i++)
1452 enum dict_class class = var_get_dict_class (d->vars[i].var);
1453 if (!(exclude_classes & class))
1459 /* Returns the case prototype that would result after deleting
1460 all variables from D that are not in one of the
1461 EXCLUDE_CLASSES and compacting the dictionary with
1464 The caller must unref the returned caseproto when it is no
1467 dict_get_compacted_proto (const struct dictionary *d,
1468 unsigned int exclude_classes)
1470 struct caseproto *proto;
1473 assert (!(exclude_classes & ~DC_ALL));
1475 proto = caseproto_create ();
1476 for (i = 0; i < d->n_vars; i++)
1478 struct variable *v = d->vars[i].var;
1479 if (!(exclude_classes & var_get_dict_class (v)))
1480 proto = caseproto_add_width (proto, var_get_width (v));
1484 /* Returns the file label for D, or a null pointer if D is
1485 unlabeled (see cmd_file_label()). */
1487 dict_get_label (const struct dictionary *d)
1492 /* Sets D's file label to LABEL, truncating it to at most 60 bytes in D's
1495 Removes D's label if LABEL is null or the empty string. */
1497 dict_set_label (struct dictionary *d, const char *label)
1500 if (label == NULL || label[0] == '\0')
1503 d->label = utf8_encoding_trunc (label, d->encoding, 60);
1506 /* Returns the documents for D, as an UTF-8 encoded string_array. The
1507 return value is always nonnull; if there are no documents then the
1508 string_arary is empty.*/
1509 const struct string_array *
1510 dict_get_documents (const struct dictionary *d)
1512 return &d->documents;
1515 /* Replaces the documents for D by NEW_DOCS, a UTF-8 encoded string_array. */
1517 dict_set_documents (struct dictionary *d, const struct string_array *new_docs)
1519 /* Swap out the old documents, instead of destroying them immediately, to
1520 allow the new documents to include pointers into the old ones. */
1521 struct string_array old_docs = STRING_ARRAY_INITIALIZER;
1522 string_array_swap (&d->documents, &old_docs);
1524 for (size_t i = 0; i < new_docs->n; i++)
1525 dict_add_document_line (d, new_docs->strings[i], false);
1527 string_array_destroy (&old_docs);
1530 /* Replaces the documents for D by UTF-8 encoded string NEW_DOCS, dividing it
1531 into individual lines at new-line characters. Each line is truncated to at
1532 most DOC_LINE_LENGTH bytes in D's encoding. */
1534 dict_set_documents_string (struct dictionary *d, const char *new_docs)
1538 dict_clear_documents (d);
1539 for (s = new_docs; *s != '\0';)
1541 size_t len = strcspn (s, "\n");
1542 char *line = xmemdup0 (s, len);
1543 dict_add_document_line (d, line, false);
1552 /* Drops the documents from dictionary D. */
1554 dict_clear_documents (struct dictionary *d)
1556 string_array_clear (&d->documents);
1559 /* Appends the UTF-8 encoded LINE to the documents in D. LINE will be
1560 truncated so that it is no more than 80 bytes in the dictionary's
1561 encoding. If this causes some text to be lost, and ISSUE_WARNING is true,
1562 then a warning will be issued. */
1564 dict_add_document_line (struct dictionary *d, const char *line,
1570 trunc_len = utf8_encoding_trunc_len (line, d->encoding, DOC_LINE_LENGTH);
1571 truncated = line[trunc_len] != '\0';
1572 if (truncated && issue_warning)
1574 /* TRANSLATORS: "bytes" is correct, not characters due to UTF encoding */
1575 msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
1578 string_array_append_nocopy (&d->documents, xmemdup0 (line, trunc_len));
1583 /* Returns the number of document lines in dictionary D. */
1585 dict_get_document_n_lines (const struct dictionary *d)
1587 return d->documents.n;
1590 /* Returns document line number IDX in dictionary D. The caller must not
1591 modify or free the returned string. */
1593 dict_get_document_line (const struct dictionary *d, size_t idx)
1595 assert (idx < d->documents.n);
1596 return d->documents.strings[idx];
1599 /* Creates in D a vector named NAME that contains the N
1600 variables in VAR. Returns true if successful, or false if a
1601 vector named NAME already exists in D. */
1603 dict_create_vector (struct dictionary *d,
1605 struct variable **var, size_t n)
1608 for (size_t i = 0; i < n; i++)
1609 assert (dict_contains_var (d, var[i]));
1611 if (dict_lookup_vector (d, name) == NULL)
1613 d->vector = xnrealloc (d->vector, d->n_vectors + 1, sizeof *d->vector);
1614 d->vector[d->n_vectors++] = vector_create (name, var, n);
1621 /* Creates in D a vector named NAME that contains the N
1622 variables in VAR. A vector named NAME must not already exist
1625 dict_create_vector_assert (struct dictionary *d,
1627 struct variable **var, size_t n)
1629 assert (dict_lookup_vector (d, name) == NULL);
1630 dict_create_vector (d, name, var, n);
1633 /* Returns the vector in D with index IDX, which must be less
1634 than dict_get_n_vectors (D). */
1635 const struct vector *
1636 dict_get_vector (const struct dictionary *d, size_t idx)
1638 assert (idx < d->n_vectors);
1640 return d->vector[idx];
1643 /* Returns the number of vectors in D. */
1645 dict_get_n_vectors (const struct dictionary *d)
1647 return d->n_vectors;
1650 /* Looks up and returns the vector within D with the given
1652 const struct vector *
1653 dict_lookup_vector (const struct dictionary *d, const char *name)
1656 for (i = 0; i < d->n_vectors; i++)
1657 if (!utf8_strcasecmp (vector_get_name (d->vector[i]), name))
1658 return d->vector[i];
1662 /* Deletes all vectors from D. */
1664 dict_clear_vectors (struct dictionary *d)
1668 for (i = 0; i < d->n_vectors; i++)
1669 vector_destroy (d->vector[i]);
1676 /* Multiple response sets. */
1678 /* Returns the multiple response set in DICT with index IDX, which must be
1679 between 0 and the count returned by dict_get_n_mrsets(), exclusive. */
1680 const struct mrset *
1681 dict_get_mrset (const struct dictionary *dict, size_t idx)
1683 assert (idx < dict->n_mrsets);
1684 return dict->mrsets[idx];
1687 /* Returns the number of multiple response sets in DICT. */
1689 dict_get_n_mrsets (const struct dictionary *dict)
1691 return dict->n_mrsets;
1694 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1695 returns its index; otherwise, returns SIZE_MAX. */
1697 dict_lookup_mrset_idx (const struct dictionary *dict, const char *name)
1701 for (i = 0; i < dict->n_mrsets; i++)
1702 if (!utf8_strcasecmp (name, dict->mrsets[i]->name))
1708 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1709 returns it; otherwise, returns NULL. */
1710 const struct mrset *
1711 dict_lookup_mrset (const struct dictionary *dict, const char *name)
1713 size_t idx = dict_lookup_mrset_idx (dict, name);
1714 return idx != SIZE_MAX ? dict->mrsets[idx] : NULL;
1717 /* Adds MRSET to DICT, replacing any existing set with the same name. Returns
1718 true if a set was replaced, false if none existed with the specified name.
1720 Ownership of MRSET is transferred to DICT. */
1722 dict_add_mrset (struct dictionary *dict, struct mrset *mrset)
1726 assert (mrset_ok (mrset, dict));
1728 idx = dict_lookup_mrset_idx (dict, mrset->name);
1729 if (idx == SIZE_MAX)
1731 dict->mrsets = xrealloc (dict->mrsets,
1732 (dict->n_mrsets + 1) * sizeof *dict->mrsets);
1733 dict->mrsets[dict->n_mrsets++] = mrset;
1738 mrset_destroy (dict->mrsets[idx]);
1739 dict->mrsets[idx] = mrset;
1744 /* Looks for a multiple response set in DICT named NAME. If found, removes it
1745 from DICT and returns true. If none is found, returns false without
1748 Deleting one multiple response set causes the indexes of other sets within
1751 dict_delete_mrset (struct dictionary *dict, const char *name)
1753 size_t idx = dict_lookup_mrset_idx (dict, name);
1754 if (idx != SIZE_MAX)
1756 mrset_destroy (dict->mrsets[idx]);
1757 dict->mrsets[idx] = dict->mrsets[--dict->n_mrsets];
1764 /* Deletes all multiple response sets from DICT. */
1766 dict_clear_mrsets (struct dictionary *dict)
1770 for (i = 0; i < dict->n_mrsets; i++)
1771 mrset_destroy (dict->mrsets[i]);
1772 free (dict->mrsets);
1773 dict->mrsets = NULL;
1777 /* Removes VAR, which must be in DICT, from DICT's multiple response sets. */
1779 dict_unset_mrset_var (struct dictionary *dict, struct variable *var)
1783 assert (dict_contains_var (dict, var));
1785 for (i = 0; i < dict->n_mrsets;)
1787 struct mrset *mrset = dict->mrsets[i];
1790 for (j = 0; j < mrset->n_vars;)
1791 if (mrset->vars[j] == var)
1792 remove_element (mrset->vars, mrset->n_vars--,
1793 sizeof *mrset->vars, j);
1797 if (mrset->n_vars < 2)
1799 mrset_destroy (mrset);
1800 dict->mrsets[i] = dict->mrsets[--dict->n_mrsets];
1807 /* Returns D's attribute set. The caller may examine or modify
1808 the attribute set, but must not destroy it. Destroying D or
1809 calling dict_set_attributes for D will also destroy D's
1812 dict_get_attributes (const struct dictionary *d)
1814 return CONST_CAST (struct attrset *, &d->attributes);
1817 /* Replaces D's attributes set by a copy of ATTRS. */
1819 dict_set_attributes (struct dictionary *d, const struct attrset *attrs)
1821 attrset_destroy (&d->attributes);
1822 attrset_clone (&d->attributes, attrs);
1825 /* Returns true if D has at least one attribute in its attribute
1826 set, false if D's attribute set is empty. */
1828 dict_has_attributes (const struct dictionary *d)
1830 return attrset_count (&d->attributes) > 0;
1833 /* Called from variable.c to notify the dictionary that some property (indicated
1834 by WHAT) of the variable has changed. OLDVAR is a copy of V as it existed
1835 prior to the change. OLDVAR is destroyed by this function.
1838 dict_var_changed (const struct variable *v, unsigned int what, struct variable *oldvar)
1840 if (var_has_vardict (v))
1842 const struct vardict_info *vardict = var_get_vardict (v);
1843 struct dictionary *d = vardict->dict;
1848 if (what & (VAR_TRAIT_WIDTH | VAR_TRAIT_POSITION))
1849 invalidate_proto (d);
1851 if (d->changed) d->changed (d, d->changed_data);
1852 if (d->callbacks && d->callbacks->var_changed)
1853 d->callbacks->var_changed (d, var_get_dict_index (v), what, oldvar, d->cb_data);
1860 /* Dictionary used to contain "internal variables". */
1861 static struct dictionary *internal_dict;
1863 /* Create a variable of the specified WIDTH to be used for internal
1864 calculations only. The variable is assigned case index CASE_IDX. */
1866 dict_create_internal_var (int case_idx, int width)
1868 if (internal_dict == NULL)
1869 internal_dict = dict_create ("UTF-8");
1873 static int counter = INT_MAX / 2;
1874 struct variable *var;
1877 if (++counter == INT_MAX)
1878 counter = INT_MAX / 2;
1880 sprintf (name, "$internal%d", counter);
1881 var = dict_create_var (internal_dict, name, width);
1884 set_var_case_index (var, case_idx);
1890 /* Destroys VAR, which must have been created with
1891 dict_create_internal_var(). */
1893 dict_destroy_internal_var (struct variable *var)
1897 dict_delete_var (internal_dict, var);
1899 /* Destroy internal_dict if it has no variables left, just so that
1900 valgrind --leak-check --show-reachable won't show internal_dict. */
1901 if (dict_get_n_vars (internal_dict) == 0)
1903 dict_unref (internal_dict);
1904 internal_dict = NULL;
1910 vardict_get_dict_index (const struct vardict_info *vardict)
1912 return vardict - vardict->dict->vars;