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/varset.h"
36 #include "data/vector.h"
37 #include "libpspp/array.h"
38 #include "libpspp/assertion.h"
39 #include "libpspp/compiler.h"
40 #include "libpspp/hash-functions.h"
41 #include "libpspp/hmap.h"
42 #include "libpspp/i18n.h"
43 #include "libpspp/message.h"
44 #include "libpspp/misc.h"
45 #include "libpspp/pool.h"
46 #include "libpspp/str.h"
47 #include "libpspp/string-array.h"
48 #include "libpspp/ll.h"
50 #include "gl/intprops.h"
51 #include "gl/minmax.h"
52 #include "gl/xalloc.h"
53 #include "gl/xmemdup0.h"
56 #define _(msgid) gettext (msgid)
62 struct vardict_info *vars; /* Variables. */
63 size_t n_vars; /* Number of variables. */
64 size_t allocated_vars; /* Allocated space in 'vars'. */
65 struct caseproto *proto; /* Prototype for dictionary cases
67 struct hmap name_map; /* Variable index by name. */
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. */
81 struct varset **varsets; /* Variable sets. */
82 size_t n_varsets; /* Number of variable sets. */
84 /* Whether variable names must be valid identifiers. Normally, this is
85 true, but sometimes a dictionary is prepared for external use
86 (e.g. output to a CSV file) where names don't have to be valid. */
87 bool names_must_be_ids;
89 char *encoding; /* Character encoding of string data */
91 const struct dict_callbacks *callbacks; /* Callbacks on dictionary
93 void *cb_data ; /* Data passed to callbacks */
95 void (*changed) (struct dictionary *, void *); /* Generic change callback */
99 static void dict_unset_split_var (struct dictionary *, struct variable *, bool);
100 static void dict_unset_mrset_var (struct dictionary *, struct variable *);
101 static void dict_unset_varset_var (struct dictionary *, struct variable *);
103 /* Compares two double pointers to variables, which should point
104 to elements of a struct dictionary's `var' member array. */
106 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
108 struct variable *const *a = a_;
109 struct variable *const *b = b_;
111 return *a < *b ? -1 : *a > *b;
115 unindex_var (struct dictionary *d, struct vardict_info *vardict)
117 hmap_delete (&d->name_map, &vardict->name_node);
120 /* This function assumes that vardict->name_node.hash is valid, that is, that
121 its name has not changed since it was hashed (rename_var() updates this
122 hash along with the name itself). */
124 reindex_var (struct dictionary *d, struct vardict_info *vardict, bool skip_callbacks)
126 struct variable *old = (d->callbacks && d->callbacks->var_changed
127 ? var_clone (vardict->var)
130 struct variable *var = vardict->var;
131 var_set_vardict (var, vardict);
132 hmap_insert_fast (&d->name_map, &vardict->name_node,
133 vardict->name_node.hash);
135 if (! skip_callbacks)
137 if (d->changed) d->changed (d, d->changed_data);
140 d->callbacks->var_changed (d, var_get_dict_index (var), VAR_TRAIT_POSITION, old, d->cb_data);
146 /* Sets the case_index in V's vardict to CASE_INDEX. */
148 set_var_case_index (struct variable *v, int case_index)
150 var_get_vardict (v)->case_index = case_index;
153 /* Removes the dictionary variables with indexes from FROM to TO (exclusive)
156 unindex_vars (struct dictionary *d, size_t from, size_t to)
160 for (i = from; i < to; i++)
161 unindex_var (d, &d->vars[i]);
164 /* Re-sets the dict_index in the dictionary variables with
165 indexes from FROM to TO (exclusive). */
167 reindex_vars (struct dictionary *d, size_t from, size_t to, bool skip_callbacks)
171 for (i = from; i < to; i++)
172 reindex_var (d, &d->vars[i], skip_callbacks);
177 /* Returns the encoding for data in dictionary D. The return value is a
178 nonnull string that contains an IANA character set name. */
180 dict_get_encoding (const struct dictionary *d)
185 /* Checks whether UTF-8 string ID is an acceptable identifier in DICT's
186 encoding. Returns true if it is, otherwise an error message that the caller
188 char * WARN_UNUSED_RESULT
189 dict_id_is_valid__ (const struct dictionary *dict, const char *id)
191 if (!dict->names_must_be_ids)
193 return id_is_valid__ (id, dict->encoding);
197 error_to_bool (char *error)
208 /* Returns true if UTF-8 string ID is an acceptable identifier in DICT's
209 encoding, false otherwise. */
211 dict_id_is_valid (const struct dictionary *dict, const char *id)
213 return error_to_bool (dict_id_is_valid__ (dict, id));
217 dict_set_change_callback (struct dictionary *d,
218 void (*changed) (struct dictionary *, void*),
221 d->changed = changed;
222 d->changed_data = data;
225 /* Discards dictionary D's caseproto. (It will be regenerated
226 lazily, on demand.) */
228 invalidate_proto (struct dictionary *d)
230 caseproto_unref (d->proto);
234 /* Print a representation of dictionary D to stdout, for
235 debugging purposes. */
237 dict_dump (const struct dictionary *d)
240 for (i = 0 ; i < d->n_vars ; ++i)
242 const struct variable *v = d->vars[i].var;
243 printf ("Name: %s;\tdict_idx: %zu; case_idx: %zu\n",
245 var_get_dict_index (v),
246 var_get_case_index (v));
251 /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
252 the dictionary or any of the variables it contains are modified.
253 Each callback will get passed CALLBACK_DATA.
254 Any callback may be NULL, in which case it'll be ignored.
257 dict_set_callbacks (struct dictionary *dict,
258 const struct dict_callbacks *callbacks,
261 dict->callbacks = callbacks;
262 dict->cb_data = callback_data;
265 /* Shallow copy the callbacks from SRC to DEST */
267 dict_copy_callbacks (struct dictionary *dest,
268 const struct dictionary *src)
270 dest->callbacks = src->callbacks;
271 dest->cb_data = src->cb_data;
274 /* Creates and returns a new dictionary with the specified ENCODING. */
276 dict_create (const char *encoding)
278 struct dictionary *d = xmalloc (sizeof *d);
280 *d = (struct dictionary) {
281 .encoding = xstrdup (encoding),
282 .names_must_be_ids = true,
283 .name_map = HMAP_INITIALIZER (d->name_map),
284 .attributes = ATTRSET_INITIALIZER (d->attributes),
285 .split_type = SPLIT_NONE,
292 /* Creates and returns a (deep) copy of an existing
295 Callbacks are not cloned. */
297 dict_clone (const struct dictionary *s)
299 struct dictionary *d = dict_create (s->encoding);
300 dict_set_names_must_be_ids (d, dict_get_names_must_be_ids (s));
302 for (size_t i = 0; i < s->n_vars; i++)
304 struct variable *sv = s->vars[i].var;
305 struct variable *dv = dict_clone_var_assert (d, sv);
307 for (size_t j = 0; j < var_get_n_short_names (sv); j++)
308 var_set_short_name (dv, j, var_get_short_name (sv, j));
310 var_get_vardict (dv)->case_index = var_get_vardict (sv)->case_index;
313 d->n_splits = s->n_splits;
316 d->split = xnmalloc (d->n_splits, sizeof *d->split);
317 for (size_t i = 0; i < d->n_splits; i++)
318 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
320 d->split_type = s->split_type;
322 if (s->weight != NULL)
323 dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight)));
325 if (s->filter != NULL)
326 dict_set_filter (d, dict_lookup_var_assert (d, var_get_name (s->filter)));
328 d->case_limit = s->case_limit;
329 dict_set_label (d, dict_get_label (s));
330 dict_set_documents (d, dict_get_documents (s));
332 d->n_vectors = s->n_vectors;
333 d->vector = xnmalloc (d->n_vectors, sizeof *d->vector);
334 for (size_t i = 0; i < s->n_vectors; i++)
335 d->vector[i] = vector_clone (s->vector[i], s, d);
337 dict_set_attributes (d, dict_get_attributes (s));
339 for (size_t i = 0; i < s->n_mrsets; i++)
341 const struct mrset *old = s->mrsets[i];
345 /* Clone old mrset, then replace vars from D by vars from S. */
346 new = mrset_clone (old);
347 for (j = 0; j < new->n_vars; j++)
348 new->vars[j] = dict_lookup_var_assert (d, var_get_name (new->vars[j]));
350 dict_add_mrset (d, new);
353 for (size_t i = 0; i < s->n_varsets; i++)
355 const struct varset *old = s->varsets[i];
357 /* Clone old varset, then replace vars from D by vars from S. */
358 struct varset *new = varset_clone (old);
359 for (size_t j = 0; j < new->n_vars; j++)
360 new->vars[j] = dict_lookup_var_assert (d, var_get_name (new->vars[j]));
362 dict_add_varset (d, new);
368 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
369 dict_get_n_splits() to determine how many SPLIT FILE vars
370 there are. Returns a null pointer if and only if there are no
372 const struct variable *const *
373 dict_get_split_vars (const struct dictionary *d)
378 /* Returns the number of SPLIT FILE vars. */
380 dict_get_n_splits (const struct dictionary *d)
385 /* Removes variable V, which must be in D, from D's set of split
388 dict_unset_split_var (struct dictionary *d, struct variable *v, bool skip_callbacks)
392 assert (dict_contains_var (d, v));
394 orig_count = d->n_splits;
395 d->n_splits = remove_equal (d->split, d->n_splits, sizeof *d->split,
396 &v, compare_var_ptrs, NULL);
397 if (orig_count != d->n_splits && !skip_callbacks)
399 if (d->changed) d->changed (d, d->changed_data);
400 /* We changed the set of split variables so invoke the
402 if (d->callbacks && d->callbacks->split_changed)
403 d->callbacks->split_changed (d, d->cb_data);
408 /* Sets N split vars SPLIT in dictionary D. N is silently capped to a maximum
411 dict_set_split_vars__ (struct dictionary *d,
412 struct variable *const *split, size_t n,
413 enum split_type type, bool skip_callbacks)
417 assert (n == 0 || split != NULL);
420 d->split_type = (n == 0 ? SPLIT_NONE
421 : type == SPLIT_NONE ? SPLIT_LAYERED
425 d->split = xnrealloc (d->split, n, sizeof *d->split) ;
426 memcpy (d->split, split, n * sizeof *d->split);
436 if (d->changed) d->changed (d, d->changed_data);
437 if (d->callbacks && d->callbacks->split_changed)
438 d->callbacks->split_changed (d, d->cb_data);
443 dict_get_split_type (const struct dictionary *d)
445 return d->split_type;
448 /* Sets N split vars SPLIT in dictionary D. */
450 dict_set_split_vars (struct dictionary *d,
451 struct variable *const *split, size_t n,
452 enum split_type type)
454 dict_set_split_vars__ (d, split, n, type, false);
458 dict_clear_split_vars (struct dictionary *d)
460 dict_set_split_vars (d, NULL, 0, SPLIT_NONE);
464 /* Deletes variable V from dictionary D and frees V.
466 This is a very bad idea if there might be any pointers to V
467 from outside D. In general, no variable in the active dataset's
468 dictionary should be deleted when any transformations are
469 active on the dictionary's dataset, because those
470 transformations might reference the deleted variable. The
471 safest time to delete a variable is just after a procedure has
472 been executed, as done by DELETE VARIABLES.
474 Pointers to V within D are not a problem, because
475 dict_delete_var() knows to remove V from split variables,
476 weights, filters, etc. */
478 dict_delete_var__ (struct dictionary *d, struct variable *v, bool skip_callbacks)
480 int dict_index = var_get_dict_index (v);
481 const int case_index = var_get_case_index (v);
483 assert (dict_contains_var (d, v));
485 dict_unset_split_var (d, v, skip_callbacks);
486 dict_unset_mrset_var (d, v);
487 dict_unset_varset_var (d, v);
490 dict_set_weight (d, NULL);
493 dict_set_filter (d, NULL);
495 dict_clear_vectors (d);
497 /* Remove V from var array. */
498 unindex_vars (d, dict_index, d->n_vars);
499 remove_element (d->vars, d->n_vars, sizeof *d->vars, dict_index);
502 /* Update dict_index for each affected variable. */
503 reindex_vars (d, dict_index, d->n_vars, skip_callbacks);
506 var_clear_vardict (v);
508 if (! skip_callbacks)
510 if (d->changed) d->changed (d, d->changed_data);
511 if (d->callbacks && d->callbacks->var_deleted)
512 d->callbacks->var_deleted (d, v, dict_index, case_index, d->cb_data);
515 invalidate_proto (d);
519 /* Deletes variable V from dictionary D and frees V.
521 This is a very bad idea if there might be any pointers to V
522 from outside D. In general, no variable in the active dataset's
523 dictionary should be deleted when any transformations are
524 active on the dictionary's dataset, because those
525 transformations might reference the deleted variable. The
526 safest time to delete a variable is just after a procedure has
527 been executed, as done by DELETE VARIABLES.
529 Pointers to V within D are not a problem, because
530 dict_delete_var() knows to remove V from split variables,
531 weights, filters, etc. */
533 dict_delete_var (struct dictionary *d, struct variable *v)
535 dict_delete_var__ (d, v, false);
539 /* Deletes the COUNT variables listed in VARS from D. This is
540 unsafe; see the comment on dict_delete_var() for details. */
542 dict_delete_vars (struct dictionary *d,
543 struct variable *const *vars, size_t count)
545 /* FIXME: this can be done in O(count) time, but this algorithm
547 assert (count == 0 || vars != NULL);
550 dict_delete_var (d, *vars++);
553 /* Deletes the COUNT variables in D starting at index IDX. This
554 is unsafe; see the comment on dict_delete_var() for
555 details. Deleting consecutive vars will result in less callbacks
556 compared to iterating over dict_delete_var.
557 A simple while loop over dict_delete_var will
558 produce (d->n_vars - IDX) * COUNT variable changed callbacks
559 plus COUNT variable delete callbacks.
560 This here produces d->n_vars - IDX variable changed callbacks
561 plus COUNT variable delete callbacks. */
563 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
565 assert (idx + count <= d->n_vars);
567 /* We need to store the variable and the corresponding case_index
568 for the delete callbacks later. We store them in a linked list.*/
571 struct variable *var;
574 struct ll_list list = LL_INITIALIZER (list);
576 for (size_t i = idx; i < idx + count; i++)
578 struct delvar *dv = xmalloc (sizeof (struct delvar));
580 struct variable *v = d->vars[i].var;
582 dict_unset_split_var (d, v, false);
583 dict_unset_mrset_var (d, v);
584 dict_unset_varset_var (d, v);
587 dict_set_weight (d, NULL);
590 dict_set_filter (d, NULL);
593 dv->case_index = var_get_case_index (v);
594 ll_push_tail (&list, (struct ll *)dv);
597 dict_clear_vectors (d);
599 /* Remove variables from var array. */
600 unindex_vars (d, idx, d->n_vars);
601 remove_range (d->vars, d->n_vars, sizeof *d->vars, idx, count);
604 /* Reindexing will result variable-changed callback */
605 reindex_vars (d, idx, d->n_vars, false);
607 invalidate_proto (d);
608 if (d->changed) d->changed (d, d->changed_data);
610 /* Now issue the variable delete callbacks and delete
611 the variables. The vardict is not valid at this point
612 anymore. That is the reason why we stored the
613 caseindex before reindexing. */
614 for (size_t vi = idx; vi < idx + count; vi++)
616 struct delvar *dv = (struct delvar *) ll_pop_head (&list);
617 var_clear_vardict (dv->var);
618 if (d->callbacks && d->callbacks->var_deleted)
619 d->callbacks->var_deleted (d, dv->var, vi, dv->case_index, d->cb_data);
625 /* Deletes scratch variables from dictionary D. */
627 dict_delete_scratch_vars (struct dictionary *d)
631 /* FIXME: this can be done in O(count) time, but this algorithm
633 for (i = 0; i < d->n_vars;)
634 if (var_get_dict_class (d->vars[i].var) == DC_SCRATCH)
635 dict_delete_var (d, d->vars[i].var);
642 /* Clears the contents from a dictionary without destroying the
643 dictionary itself. */
645 dict_clear__ (struct dictionary *d, bool skip_callbacks)
647 /* FIXME? Should we really clear case_limit, label, documents?
648 Others are necessarily cleared by deleting all the variables.*/
649 while (d->n_vars > 0)
651 dict_delete_var__ (d, d->vars[d->n_vars - 1].var, skip_callbacks);
656 d->n_vars = d->allocated_vars = 0;
657 invalidate_proto (d);
658 hmap_clear (&d->name_map);
659 dict_set_split_vars__ (d, NULL, 0, SPLIT_NONE, skip_callbacks);
668 dict_set_weight (d, NULL);
669 dict_set_filter (d, NULL);
674 string_array_clear (&d->documents);
675 dict_clear_vectors (d);
676 attrset_clear (&d->attributes);
679 /* Clears the contents from a dictionary without destroying the
680 dictionary itself. */
682 dict_clear (struct dictionary *d)
684 dict_clear__ (d, false);
687 /* Clears a dictionary and destroys it. */
689 _dict_destroy (struct dictionary *d)
691 /* In general, we don't want callbacks occurring, if the dictionary
692 is being destroyed */
693 d->callbacks = NULL ;
695 dict_clear__ (d, true);
696 string_array_destroy (&d->documents);
697 hmap_destroy (&d->name_map);
698 attrset_destroy (&d->attributes);
699 dict_clear_mrsets (d);
700 dict_clear_varsets (d);
706 dict_ref (struct dictionary *d)
713 dict_unref (struct dictionary *d)
718 assert (d->ref_cnt >= 0);
723 /* Returns the number of variables in D. */
725 dict_get_n_vars (const struct dictionary *d)
730 /* Returns the variable in D with dictionary index IDX, which
731 must be between 0 and the count returned by
732 dict_get_n_vars(), exclusive. */
734 dict_get_var (const struct dictionary *d, size_t idx)
736 assert (idx < d->n_vars);
738 return d->vars[idx].var;
741 /* Sets *VARS to an array of pointers to variables in D and *N
742 to the number of variables in *D. All variables are returned
743 except for those, if any, in the classes indicated by EXCLUDE.
744 (There is no point in putting DC_SYSTEM in EXCLUDE as
745 dictionaries never include system variables.) */
747 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
748 size_t *n, enum dict_class exclude)
750 dict_get_vars_mutable (d, (struct variable ***) vars, n, exclude);
753 /* Sets *VARS to an array of pointers to variables in D and *N
754 to the number of variables in *D. All variables are returned
755 except for those, if any, in the classes indicated by EXCLUDE.
756 (There is no point in putting DC_SYSTEM in EXCLUDE as
757 dictionaries never include system variables.) */
759 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
760 size_t *n, enum dict_class exclude)
765 assert (exclude == (exclude & DC_ALL));
768 for (i = 0; i < d->n_vars; i++)
770 enum dict_class class = var_get_dict_class (d->vars[i].var);
771 if (!(class & exclude))
775 *vars = xnmalloc (count, sizeof **vars);
777 for (i = 0; i < d->n_vars; i++)
779 enum dict_class class = var_get_dict_class (d->vars[i].var);
780 if (!(class & exclude))
781 (*vars)[(*n)++] = d->vars[i].var;
783 assert (*n == count);
786 static struct variable *
787 add_var_with_case_index (struct dictionary *d, struct variable *v,
790 struct vardict_info *vardict;
792 /* Update dictionary. */
793 if (d->n_vars >= d->allocated_vars)
797 d->vars = x2nrealloc (d->vars, &d->allocated_vars, sizeof *d->vars);
798 hmap_clear (&d->name_map);
799 for (i = 0; i < d->n_vars; i++)
801 var_set_vardict (d->vars[i].var, &d->vars[i]);
802 hmap_insert_fast (&d->name_map, &d->vars[i].name_node,
803 d->vars[i].name_node.hash);
807 vardict = &d->vars[d->n_vars++];
810 hmap_insert (&d->name_map, &vardict->name_node,
811 utf8_hash_case_string (var_get_name (v), 0));
812 vardict->case_index = case_index;
813 var_set_vardict (v, vardict);
815 if (d->changed) d->changed (d, d->changed_data);
816 if (d->callbacks && d->callbacks->var_added)
817 d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
819 invalidate_proto (d);
824 static struct variable *
825 add_var (struct dictionary *d, struct variable *v)
827 return add_var_with_case_index (d, v, dict_get_n_vars (d));
830 /* Creates and returns a new variable in D with the given NAME
831 and WIDTH. Returns a null pointer if the given NAME would
832 duplicate that of an existing variable in the dictionary. */
834 dict_create_var (struct dictionary *d, const char *name, int width)
836 return (dict_lookup_var (d, name) == NULL
837 ? dict_create_var_assert (d, name, width)
841 /* Creates and returns a new variable in D with the given NAME
842 and WIDTH. Assert-fails if the given NAME would duplicate
843 that of an existing variable in the dictionary. */
845 dict_create_var_assert (struct dictionary *d, const char *name, int width)
847 assert (dict_lookup_var (d, name) == NULL);
848 return add_var (d, var_create (name, width));
851 /* Creates and returns a new variable in D, as a copy of existing variable
852 OLD_VAR, which need not be in D or in any dictionary. Returns a null
853 pointer if OLD_VAR's name would duplicate that of an existing variable in
856 dict_clone_var (struct dictionary *d, const struct variable *old_var)
858 return dict_clone_var_as (d, old_var, var_get_name (old_var));
861 /* Creates and returns a new variable in D, as a copy of existing variable
862 OLD_VAR, which need not be in D or in any dictionary. Assert-fails if
863 OLD_VAR's name would duplicate that of an existing variable in the
866 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var)
868 return dict_clone_var_as_assert (d, old_var, var_get_name (old_var));
871 /* Creates and returns a new variable in D with name NAME, as a copy of
872 existing variable OLD_VAR, which need not be in D or in any dictionary.
873 Returns a null pointer if the given NAME would duplicate that of an existing
874 variable in the dictionary. */
876 dict_clone_var_as (struct dictionary *d, const struct variable *old_var,
879 return (dict_lookup_var (d, name) == NULL
880 ? dict_clone_var_as_assert (d, old_var, name)
884 /* Creates and returns a new variable in D with name NAME, as a copy of
885 existing variable OLD_VAR, which need not be in D or in any dictionary.
886 Assert-fails if the given NAME would duplicate that of an existing variable
887 in the dictionary. */
889 dict_clone_var_as_assert (struct dictionary *d, const struct variable *old_var,
892 struct variable *new_var = var_clone (old_var);
893 assert (dict_lookup_var (d, name) == NULL);
894 var_set_name (new_var, name);
895 return add_var (d, new_var);
899 dict_clone_var_in_place_assert (struct dictionary *d,
900 const struct variable *old_var)
902 assert (dict_lookup_var (d, var_get_name (old_var)) == NULL);
903 return add_var_with_case_index (d, var_clone (old_var),
904 var_get_case_index (old_var));
907 /* Returns the variable named NAME in D, or a null pointer if no
908 variable has that name. */
910 dict_lookup_var (const struct dictionary *d, const char *name)
912 struct vardict_info *vardict;
914 HMAP_FOR_EACH_WITH_HASH (vardict, struct vardict_info, name_node,
915 utf8_hash_case_string (name, 0), &d->name_map)
917 struct variable *var = vardict->var;
918 if (!utf8_strcasecmp (var_get_name (var), name))
925 /* Returns the variable named NAME in D. Assert-fails if no
926 variable has that name. */
928 dict_lookup_var_assert (const struct dictionary *d, const char *name)
930 struct variable *v = dict_lookup_var (d, name);
935 /* Returns true if variable V is in dictionary D,
938 dict_contains_var (const struct dictionary *d, const struct variable *v)
940 return (var_has_vardict (v)
941 && vardict_get_dictionary (var_get_vardict (v)) == d);
944 /* Moves V to 0-based position IDX in D. Other variables in D,
945 if any, retain their relative positions. Runs in time linear
946 in the distance moved. */
948 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
950 size_t old_index = var_get_dict_index (v);
952 assert (new_index < d->n_vars);
954 unindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
955 move_element (d->vars, d->n_vars, sizeof *d->vars, old_index, new_index);
956 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1, false);
959 /* Reorders the variables in D, placing the COUNT variables
960 listed in ORDER in that order at the beginning of D. The
961 other variables in D, if any, retain their relative
964 dict_reorder_vars (struct dictionary *d,
965 struct variable *const *order, size_t count)
967 struct vardict_info *new_var;
970 assert (count == 0 || order != NULL);
971 assert (count <= d->n_vars);
973 new_var = xnmalloc (d->allocated_vars, sizeof *new_var);
975 /* Add variables in ORDER to new_var. */
976 for (i = 0; i < count; i++)
978 struct vardict_info *old_var;
980 assert (dict_contains_var (d, order[i]));
982 old_var = var_get_vardict (order[i]);
983 new_var[i] = *old_var;
984 old_var->dict = NULL;
987 /* Add remaining variables to new_var. */
988 for (i = 0; i < d->n_vars; i++)
989 if (d->vars[i].dict != NULL)
990 new_var[count++] = d->vars[i];
991 assert (count == d->n_vars);
993 /* Replace old vardicts by new ones. */
997 hmap_clear (&d->name_map);
998 reindex_vars (d, 0, d->n_vars, false);
1001 /* Changes the name of variable V that is currently in a dictionary to
1004 rename_var (struct variable *v, const char *new_name)
1006 struct vardict_info *vardict = var_get_vardict (v);
1007 var_clear_vardict (v);
1008 var_set_name (v, new_name);
1009 vardict->name_node.hash = utf8_hash_case_string (new_name, 0);
1010 var_set_vardict (v, vardict);
1013 /* Tries to changes the name of V in D to name NEW_NAME. Returns true if
1014 successful, false if a variable (other than V) with the given name already
1017 dict_try_rename_var (struct dictionary *d, struct variable *v,
1018 const char *new_name)
1020 struct variable *conflict = dict_lookup_var (d, new_name);
1021 if (conflict && v != conflict)
1024 struct variable *old = var_clone (v);
1025 unindex_var (d, var_get_vardict (v));
1026 rename_var (v, new_name);
1027 reindex_var (d, var_get_vardict (v), false);
1029 if (settings_get_algorithm () == ENHANCED)
1030 var_clear_short_names (v);
1032 if (d->changed) d->changed (d, d->changed_data);
1033 if (d->callbacks && d->callbacks->var_changed)
1034 d->callbacks->var_changed (d, var_get_dict_index (v), VAR_TRAIT_NAME, old, d->cb_data);
1041 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
1042 a variable named NEW_NAME is already in D, except that
1043 NEW_NAME may be the same as V's existing name. */
1045 dict_rename_var (struct dictionary *d, struct variable *v,
1046 const char *new_name)
1048 bool ok UNUSED = dict_try_rename_var (d, v, new_name);
1052 /* Renames COUNT variables specified in VARS to the names given
1053 in NEW_NAMES within dictionary D. If the renaming would
1054 result in a duplicate variable name, returns false and stores a
1055 name that would be duplicated into *ERR_NAME (if ERR_NAME is
1056 non-null). Otherwise, the renaming is successful, and true
1059 dict_rename_vars (struct dictionary *d,
1060 struct variable **vars, char **new_names, size_t count,
1067 assert (count == 0 || vars != NULL);
1068 assert (count == 0 || new_names != NULL);
1070 /* Save the names of the variables to be renamed. */
1071 pool = pool_create ();
1072 old_names = pool_nalloc (pool, count, sizeof *old_names);
1073 for (i = 0; i < count; i++)
1074 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
1076 /* Remove the variables to be renamed from the name hash,
1078 for (i = 0; i < count; i++)
1080 unindex_var (d, var_get_vardict (vars[i]));
1081 rename_var (vars[i], new_names[i]);
1084 /* Add the renamed variables back into the name hash,
1085 checking for conflicts. */
1086 for (i = 0; i < count; i++)
1088 if (dict_lookup_var (d, var_get_name (vars[i])) != NULL)
1090 /* There is a name conflict.
1091 Back out all the name changes that have already
1092 taken place, and indicate failure. */
1093 size_t fail_idx = i;
1094 if (err_name != NULL)
1095 *err_name = new_names[i];
1097 for (i = 0; i < fail_idx; i++)
1098 unindex_var (d, var_get_vardict (vars[i]));
1100 for (i = 0; i < count; i++)
1102 rename_var (vars[i], old_names[i]);
1103 reindex_var (d, var_get_vardict (vars[i]), false);
1106 pool_destroy (pool);
1109 reindex_var (d, var_get_vardict (vars[i]), false);
1112 /* Clear short names. */
1113 if (settings_get_algorithm () == ENHANCED)
1114 for (i = 0; i < count; i++)
1115 var_clear_short_names (vars[i]);
1117 pool_destroy (pool);
1121 /* Returns true if a variable named NAME may be inserted in DICT;
1122 that is, if there is not already a variable with that name in
1123 DICT and if NAME is not a reserved word. (The caller's checks
1124 have already verified that NAME is otherwise acceptable as a
1127 var_name_is_insertable (const struct dictionary *dict, const char *name)
1129 return (dict_lookup_var (dict, name) == NULL
1130 && lex_id_to_token (ss_cstr (name)) == T_ID);
1134 make_hinted_name (const struct dictionary *dict, const char *hint)
1136 size_t hint_len = strlen (hint);
1137 bool dropped = false;
1142 if (hint_len > ID_MAX_LEN)
1143 hint_len = ID_MAX_LEN;
1145 /* The allocation size here is OK: characters that are copied directly fit
1146 OK, and characters that are not copied directly are replaced by a single
1147 '_' byte. If u8_mbtouc() replaces bad input by 0xfffd, then that will get
1148 replaced by '_' too. */
1149 root = rp = xmalloc (hint_len + 1);
1150 for (ofs = 0; ofs < hint_len; ofs += mblen)
1154 mblen = u8_mbtouc (&uc, CHAR_CAST (const uint8_t *, hint + ofs),
1157 ? lex_uc_is_id1 (uc) && uc != '$'
1158 : lex_uc_is_idn (uc))
1165 rp += u8_uctomb (CHAR_CAST (uint8_t *, rp), uc, 6);
1167 else if (rp != root)
1172 if (root[0] != '\0')
1174 unsigned long int i;
1176 if (var_name_is_insertable (dict, root))
1179 for (i = 0; i < ULONG_MAX; i++)
1181 char suffix[INT_BUFSIZE_BOUND (i) + 1];
1185 if (!str_format_26adic (i + 1, true, &suffix[1], sizeof suffix - 1))
1188 name = utf8_encoding_concat (root, suffix, dict->encoding, 64);
1189 if (var_name_is_insertable (dict, name))
1204 make_numeric_name (const struct dictionary *dict, unsigned long int *num_start)
1206 unsigned long int number;
1208 for (number = num_start != NULL ? MAX (*num_start, 1) : 1;
1212 char name[3 + INT_STRLEN_BOUND (number) + 1];
1214 sprintf (name, "VAR%03lu", number);
1215 if (dict_lookup_var (dict, name) == NULL)
1217 if (num_start != NULL)
1218 *num_start = number + 1;
1219 return xstrdup (name);
1227 /* Devises and returns a variable name unique within DICT. The variable name
1228 is owned by the caller, which must free it with free() when it is no longer
1231 HINT, if it is non-null, is used as a suggestion that will be
1232 modified for suitability as a variable name and for
1235 If HINT is null or entirely unsuitable, a name in the form
1236 "VAR%03d" will be generated, where the smallest unused integer
1237 value is used. If NUM_START is non-null, then its value is
1238 used as the minimum numeric value to check, and it is updated
1239 to the next value to be checked.
1242 dict_make_unique_var_name (const struct dictionary *dict, const char *hint,
1243 unsigned long int *num_start)
1247 char *hinted_name = make_hinted_name (dict, hint);
1248 if (hinted_name != NULL)
1252 return make_numeric_name (dict, num_start);
1255 /* Returns whether variable names must be valid identifiers. Normally, this is
1256 true, but sometimes a dictionary is prepared for external use (e.g. output
1257 to a CSV file) where names don't have to be valid. */
1259 dict_get_names_must_be_ids (const struct dictionary *d)
1261 return d->names_must_be_ids;
1264 /* Sets whether variable names must be valid identifiers. Normally, this is
1265 true, but sometimes a dictionary is prepared for external use (e.g. output
1266 to a CSV file) where names don't have to be valid.
1268 Changing this setting from false to true doesn't make the dictionary check
1269 all the existing variable names, so it can cause an invariant violation. */
1271 dict_set_names_must_be_ids (struct dictionary *d, bool names_must_be_ids)
1273 d->names_must_be_ids = names_must_be_ids;
1276 /* Returns the weighting variable in dictionary D, or a null
1277 pointer if the dictionary is unweighted. */
1279 dict_get_weight (const struct dictionary *d)
1281 assert (d->weight == NULL || dict_contains_var (d, d->weight));
1286 /* Returns the value of D's weighting variable in case C, except
1287 that a negative or missing weight is returned as 0. Returns 1 if the
1288 dictionary is unweighted. Will warn about missing, negative,
1289 or zero values if *WARN_ON_INVALID is true. The function will
1290 set *WARN_ON_INVALID to false if an invalid weight is
1293 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
1294 bool *warn_on_invalid)
1298 if (d->weight == NULL)
1302 double w = case_num (c, d->weight);
1304 return var_force_valid_weight (d->weight, w, warn_on_invalid);
1308 /* Like dict_get_case_weight(), but additionally rounds each weight to the
1311 dict_get_rounded_case_weight (const struct dictionary *d,
1312 const struct ccase *c, bool *warn_on_invalid)
1314 return floor (dict_get_case_weight (d, c, warn_on_invalid) + 0.5);
1317 /* Returns the format to use for weights. */
1319 dict_get_weight_format (const struct dictionary *d)
1321 return d->weight ? var_get_print_format (d->weight) : F_8_0;
1324 /* Sets the weighting variable of D to V, or turning off
1325 weighting if V is a null pointer. */
1327 dict_set_weight (struct dictionary *d, struct variable *v)
1329 assert (v == NULL || dict_contains_var (d, v));
1330 assert (v == NULL || var_is_numeric (v));
1334 if (d->changed) d->changed (d, d->changed_data);
1335 if (d->callbacks && d->callbacks->weight_changed)
1336 d->callbacks->weight_changed (d,
1337 v ? var_get_dict_index (v) : -1,
1341 /* Returns the filter variable in dictionary D (see cmd_filter())
1342 or a null pointer if the dictionary is unfiltered. */
1344 dict_get_filter (const struct dictionary *d)
1346 assert (d->filter == NULL || dict_contains_var (d, d->filter));
1351 /* Sets V as the filter variable for dictionary D. Passing a
1352 null pointer for V turn off filtering. */
1354 dict_set_filter (struct dictionary *d, struct variable *v)
1356 assert (v == NULL || dict_contains_var (d, v));
1357 assert (v == NULL || var_is_numeric (v));
1361 if (d->changed) d->changed (d, d->changed_data);
1362 if (d->callbacks && d->callbacks->filter_changed)
1363 d->callbacks->filter_changed (d,
1364 v ? var_get_dict_index (v) : -1,
1368 /* Returns the case limit for dictionary D, or zero if the number
1369 of cases is unlimited. */
1371 dict_get_case_limit (const struct dictionary *d)
1373 return d->case_limit;
1376 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
1377 0 for CASE_LIMIT to indicate no limit. */
1379 dict_set_case_limit (struct dictionary *d, casenumber case_limit)
1381 d->case_limit = case_limit;
1384 /* Returns the prototype used for cases created by dictionary D. */
1385 const struct caseproto *
1386 dict_get_proto (const struct dictionary *d_)
1388 struct dictionary *d = CONST_CAST (struct dictionary *, d_);
1389 if (d->proto == NULL)
1391 short int *widths = xnmalloc (d->n_vars, sizeof *widths);
1392 for (size_t i = 0; i < d->n_vars; i++)
1394 for (size_t i = 0; i < d->n_vars; i++)
1396 const struct variable *var = d->vars[i].var;
1397 size_t case_idx = var_get_case_index (var);
1398 assert (case_idx < d->n_vars);
1399 assert (widths[case_idx] == -1);
1400 widths[case_idx] = var_get_width (var);
1403 d->proto = caseproto_from_widths (widths, d->n_vars);
1408 /* Reassigns values in dictionary D so that fragmentation is
1411 dict_compact_values (struct dictionary *d)
1413 for (size_t i = 0; i < d->n_vars; i++)
1415 struct variable *v = d->vars[i].var;
1416 set_var_case_index (v, i);
1418 invalidate_proto (d);
1421 /* Returns the number of values occupied by the variables in
1422 dictionary D. All variables are considered if EXCLUDE_CLASSES
1423 is 0, or it may contain one or more of DC_ORDINARY, DC_SYSTEM,
1424 or DC_SCRATCH to exclude the corresponding type of variable. */
1426 dict_count_values (const struct dictionary *d, unsigned int exclude_classes)
1428 assert (!(exclude_classes & ~DC_ALL));
1431 for (size_t i = 0; i < d->n_vars; i++)
1433 enum dict_class class = var_get_dict_class (d->vars[i].var);
1434 if (!(exclude_classes & class))
1440 /* Returns the case prototype that would result after deleting
1441 all variables from D that are not in one of the
1442 EXCLUDE_CLASSES and compacting the dictionary with
1445 The caller must unref the returned caseproto when it is no
1448 dict_get_compacted_proto (const struct dictionary *d,
1449 unsigned int exclude_classes)
1451 struct caseproto *proto;
1454 assert (!(exclude_classes & ~DC_ALL));
1456 proto = caseproto_create ();
1457 for (i = 0; i < d->n_vars; i++)
1459 struct variable *v = d->vars[i].var;
1460 if (!(exclude_classes & var_get_dict_class (v)))
1461 proto = caseproto_add_width (proto, var_get_width (v));
1465 /* Returns the file label for D, or a null pointer if D is
1466 unlabeled (see cmd_file_label()). */
1468 dict_get_label (const struct dictionary *d)
1473 /* Sets D's file label to LABEL, truncating it to at most 60 bytes in D's
1476 Removes D's label if LABEL is null or the empty string. */
1478 dict_set_label (struct dictionary *d, const char *label)
1481 if (label == NULL || label[0] == '\0')
1484 d->label = utf8_encoding_trunc (label, d->encoding, 60);
1487 /* Returns the documents for D, as an UTF-8 encoded string_array. The
1488 return value is always nonnull; if there are no documents then the
1489 string_arary is empty.*/
1490 const struct string_array *
1491 dict_get_documents (const struct dictionary *d)
1493 return &d->documents;
1496 /* Replaces the documents for D by NEW_DOCS, a UTF-8 encoded string_array. */
1498 dict_set_documents (struct dictionary *d, const struct string_array *new_docs)
1500 /* Swap out the old documents, instead of destroying them immediately, to
1501 allow the new documents to include pointers into the old ones. */
1502 struct string_array old_docs = STRING_ARRAY_INITIALIZER;
1503 string_array_swap (&d->documents, &old_docs);
1505 for (size_t i = 0; i < new_docs->n; i++)
1506 dict_add_document_line (d, new_docs->strings[i], false);
1508 string_array_destroy (&old_docs);
1511 /* Replaces the documents for D by UTF-8 encoded string NEW_DOCS, dividing it
1512 into individual lines at new-line characters. Each line is truncated to at
1513 most DOC_LINE_LENGTH bytes in D's encoding. */
1515 dict_set_documents_string (struct dictionary *d, const char *new_docs)
1519 dict_clear_documents (d);
1520 for (s = new_docs; *s != '\0';)
1522 size_t len = strcspn (s, "\n");
1523 char *line = xmemdup0 (s, len);
1524 dict_add_document_line (d, line, false);
1533 /* Drops the documents from dictionary D. */
1535 dict_clear_documents (struct dictionary *d)
1537 string_array_clear (&d->documents);
1540 /* Appends the UTF-8 encoded LINE to the documents in D. LINE will be
1541 truncated so that it is no more than 80 bytes in the dictionary's
1542 encoding. If this causes some text to be lost, and ISSUE_WARNING is true,
1543 then a warning will be issued. */
1545 dict_add_document_line (struct dictionary *d, const char *line,
1551 trunc_len = utf8_encoding_trunc_len (line, d->encoding, DOC_LINE_LENGTH);
1552 truncated = line[trunc_len] != '\0';
1553 if (truncated && issue_warning)
1555 /* TRANSLATORS: "bytes" is correct, not characters due to UTF encoding */
1556 msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
1559 string_array_append_nocopy (&d->documents, xmemdup0 (line, trunc_len));
1564 /* Returns the number of document lines in dictionary D. */
1566 dict_get_document_n_lines (const struct dictionary *d)
1568 return d->documents.n;
1571 /* Returns document line number IDX in dictionary D. The caller must not
1572 modify or free the returned string. */
1574 dict_get_document_line (const struct dictionary *d, size_t idx)
1576 assert (idx < d->documents.n);
1577 return d->documents.strings[idx];
1580 /* Creates in D a vector named NAME that contains the N
1581 variables in VAR. Returns true if successful, or false if a
1582 vector named NAME already exists in D. */
1584 dict_create_vector (struct dictionary *d,
1586 struct variable **var, size_t n)
1589 for (size_t i = 0; i < n; i++)
1590 assert (dict_contains_var (d, var[i]));
1592 if (dict_lookup_vector (d, name) == NULL)
1594 d->vector = xnrealloc (d->vector, d->n_vectors + 1, sizeof *d->vector);
1595 d->vector[d->n_vectors++] = vector_create (name, var, n);
1602 /* Creates in D a vector named NAME that contains the N
1603 variables in VAR. A vector named NAME must not already exist
1606 dict_create_vector_assert (struct dictionary *d,
1608 struct variable **var, size_t n)
1610 assert (dict_lookup_vector (d, name) == NULL);
1611 dict_create_vector (d, name, var, n);
1614 /* Returns the vector in D with index IDX, which must be less
1615 than dict_get_n_vectors (D). */
1616 const struct vector *
1617 dict_get_vector (const struct dictionary *d, size_t idx)
1619 assert (idx < d->n_vectors);
1621 return d->vector[idx];
1624 /* Returns the number of vectors in D. */
1626 dict_get_n_vectors (const struct dictionary *d)
1628 return d->n_vectors;
1631 /* Looks up and returns the vector within D with the given
1633 const struct vector *
1634 dict_lookup_vector (const struct dictionary *d, const char *name)
1637 for (i = 0; i < d->n_vectors; i++)
1638 if (!utf8_strcasecmp (vector_get_name (d->vector[i]), name))
1639 return d->vector[i];
1643 /* Deletes all vectors from D. */
1645 dict_clear_vectors (struct dictionary *d)
1649 for (i = 0; i < d->n_vectors; i++)
1650 vector_destroy (d->vector[i]);
1657 /* Multiple response sets. */
1659 /* Returns the multiple response set in DICT with index IDX, which must be
1660 between 0 and the count returned by dict_get_n_mrsets(), exclusive. */
1661 const struct mrset *
1662 dict_get_mrset (const struct dictionary *dict, size_t idx)
1664 assert (idx < dict->n_mrsets);
1665 return dict->mrsets[idx];
1668 /* Returns the number of multiple response sets in DICT. */
1670 dict_get_n_mrsets (const struct dictionary *dict)
1672 return dict->n_mrsets;
1675 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1676 returns its index; otherwise, returns SIZE_MAX. */
1678 dict_lookup_mrset_idx (const struct dictionary *dict, const char *name)
1682 for (i = 0; i < dict->n_mrsets; i++)
1683 if (!utf8_strcasecmp (name, dict->mrsets[i]->name))
1689 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1690 returns it; otherwise, returns NULL. */
1691 const struct mrset *
1692 dict_lookup_mrset (const struct dictionary *dict, const char *name)
1694 size_t idx = dict_lookup_mrset_idx (dict, name);
1695 return idx != SIZE_MAX ? dict->mrsets[idx] : NULL;
1698 /* Adds MRSET to DICT, replacing any existing set with the same name. Returns
1699 true if a set was replaced, false if none existed with the specified name.
1701 Ownership of MRSET is transferred to DICT. */
1703 dict_add_mrset (struct dictionary *dict, struct mrset *mrset)
1707 assert (mrset_ok (mrset, dict));
1709 idx = dict_lookup_mrset_idx (dict, mrset->name);
1710 if (idx == SIZE_MAX)
1712 dict->mrsets = xrealloc (dict->mrsets,
1713 (dict->n_mrsets + 1) * sizeof *dict->mrsets);
1714 dict->mrsets[dict->n_mrsets++] = mrset;
1719 mrset_destroy (dict->mrsets[idx]);
1720 dict->mrsets[idx] = mrset;
1725 /* Looks for a multiple response set in DICT named NAME. If found, removes it
1726 from DICT and returns true. If none is found, returns false without
1729 Deleting one multiple response set causes the indexes of other sets within
1732 dict_delete_mrset (struct dictionary *dict, const char *name)
1734 size_t idx = dict_lookup_mrset_idx (dict, name);
1735 if (idx != SIZE_MAX)
1737 mrset_destroy (dict->mrsets[idx]);
1738 dict->mrsets[idx] = dict->mrsets[--dict->n_mrsets];
1745 /* Deletes all multiple response sets from DICT. */
1747 dict_clear_mrsets (struct dictionary *dict)
1751 for (i = 0; i < dict->n_mrsets; i++)
1752 mrset_destroy (dict->mrsets[i]);
1753 free (dict->mrsets);
1754 dict->mrsets = NULL;
1758 /* Removes VAR, which must be in DICT, from DICT's multiple response sets. */
1760 dict_unset_mrset_var (struct dictionary *dict, struct variable *var)
1764 assert (dict_contains_var (dict, var));
1766 for (i = 0; i < dict->n_mrsets;)
1768 struct mrset *mrset = dict->mrsets[i];
1771 for (j = 0; j < mrset->n_vars;)
1772 if (mrset->vars[j] == var)
1773 remove_element (mrset->vars, mrset->n_vars--,
1774 sizeof *mrset->vars, j);
1778 if (mrset->n_vars < 2)
1780 mrset_destroy (mrset);
1781 dict->mrsets[i] = dict->mrsets[--dict->n_mrsets];
1789 /* Returns the variable set in DICT with index IDX, which must be between 0 and
1790 the count returned by dict_get_n_varsets(), exclusive. */
1791 const struct varset *
1792 dict_get_varset (const struct dictionary *dict, size_t idx)
1794 assert (idx < dict->n_varsets);
1795 return dict->varsets[idx];
1798 /* Returns the number of variable sets in DICT. */
1800 dict_get_n_varsets (const struct dictionary *dict)
1802 return dict->n_varsets;
1805 /* Looks for a variable set named NAME in DICT. If it finds one, returns its
1806 index; otherwise, returns SIZE_MAX. */
1808 dict_lookup_varset_idx (const struct dictionary *dict, const char *name)
1810 for (size_t i = 0; i < dict->n_varsets; i++)
1811 if (!utf8_strcasecmp (name, dict->varsets[i]->name))
1817 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1818 returns it; otherwise, returns NULL. */
1819 const struct varset *
1820 dict_lookup_varset (const struct dictionary *dict, const char *name)
1822 size_t idx = dict_lookup_varset_idx (dict, name);
1823 return idx != SIZE_MAX ? dict->varsets[idx] : NULL;
1826 /* Adds VARSET to DICT, replacing any existing set with the same name. Returns
1827 true if a set was replaced, false if none existed with the specified name.
1829 Ownership of VARSET is transferred to DICT. */
1831 dict_add_varset (struct dictionary *dict, struct varset *varset)
1833 size_t idx = dict_lookup_varset_idx (dict, varset->name);
1834 if (idx == SIZE_MAX)
1836 dict->varsets = xrealloc (dict->varsets,
1837 (dict->n_varsets + 1) * sizeof *dict->varsets);
1838 dict->varsets[dict->n_varsets++] = varset;
1843 varset_destroy (dict->varsets[idx]);
1844 dict->varsets[idx] = varset;
1849 /* Deletes all variable sets from DICT. */
1851 dict_clear_varsets (struct dictionary *dict)
1853 for (size_t i = 0; i < dict->n_varsets; i++)
1854 varset_destroy (dict->varsets[i]);
1855 free (dict->varsets);
1856 dict->varsets = NULL;
1857 dict->n_varsets = 0;
1860 /* Removes VAR, which must be in DICT, from DICT's multiple response sets. */
1862 dict_unset_varset_var (struct dictionary *dict, struct variable *var)
1864 assert (dict_contains_var (dict, var));
1866 for (size_t i = 0; i < dict->n_varsets; i++)
1868 struct varset *varset = dict->varsets[i];
1870 for (size_t j = 0; j < varset->n_vars;)
1871 if (varset->vars[j] == var)
1872 remove_element (varset->vars, varset->n_vars--,
1873 sizeof *varset->vars, j);
1879 /* Returns D's attribute set. The caller may examine or modify
1880 the attribute set, but must not destroy it. Destroying D or
1881 calling dict_set_attributes for D will also destroy D's
1884 dict_get_attributes (const struct dictionary *d)
1886 return CONST_CAST (struct attrset *, &d->attributes);
1889 /* Replaces D's attributes set by a copy of ATTRS. */
1891 dict_set_attributes (struct dictionary *d, const struct attrset *attrs)
1893 attrset_destroy (&d->attributes);
1894 attrset_clone (&d->attributes, attrs);
1897 /* Returns true if D has at least one attribute in its attribute
1898 set, false if D's attribute set is empty. */
1900 dict_has_attributes (const struct dictionary *d)
1902 return attrset_count (&d->attributes) > 0;
1905 /* Called from variable.c to notify the dictionary that some property (indicated
1906 by WHAT) of the variable has changed. OLDVAR is a copy of V as it existed
1907 prior to the change. OLDVAR is destroyed by this function.
1910 dict_var_changed (const struct variable *v, unsigned int what, struct variable *oldvar)
1912 if (var_has_vardict (v))
1914 const struct vardict_info *vardict = var_get_vardict (v);
1915 struct dictionary *d = vardict->dict;
1920 if (what & (VAR_TRAIT_WIDTH | VAR_TRAIT_POSITION))
1921 invalidate_proto (d);
1923 if (d->changed) d->changed (d, d->changed_data);
1924 if (d->callbacks && d->callbacks->var_changed)
1925 d->callbacks->var_changed (d, var_get_dict_index (v), what, oldvar, d->cb_data);
1933 vardict_get_dict_index (const struct vardict_info *vardict)
1935 return vardict - vardict->dict->vars;