1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006, 2007, 2009, 2010, 2011, 2012, 2013, 2014, 2015 Free Software Foundation, Inc.
4 This program is free software: you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation, either version 3 of the License, or
7 (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program. If not, see <http://www.gnu.org/licenses/>. */
19 #include "data/dictionary.h"
26 #include "data/attributes.h"
27 #include "data/case.h"
28 #include "data/identifier.h"
29 #include "data/mrset.h"
30 #include "data/settings.h"
31 #include "data/value-labels.h"
32 #include "data/vardict.h"
33 #include "data/variable.h"
34 #include "data/vector.h"
35 #include "libpspp/array.h"
36 #include "libpspp/assertion.h"
37 #include "libpspp/compiler.h"
38 #include "libpspp/hash-functions.h"
39 #include "libpspp/hmap.h"
40 #include "libpspp/i18n.h"
41 #include "libpspp/message.h"
42 #include "libpspp/misc.h"
43 #include "libpspp/pool.h"
44 #include "libpspp/str.h"
45 #include "libpspp/string-array.h"
47 #include "gl/intprops.h"
48 #include "gl/minmax.h"
49 #include "gl/xalloc.h"
50 #include "gl/xmemdup0.h"
53 #define _(msgid) gettext (msgid)
58 struct vardict_info *var; /* Variables. */
59 size_t var_cnt, var_cap; /* Number of variables, capacity. */
60 struct caseproto *proto; /* Prototype for dictionary cases
62 struct hmap name_map; /* Variable index by name. */
63 int next_value_idx; /* Index of next `union value' to allocate. */
64 const struct variable **split; /* SPLIT FILE vars. */
65 size_t split_cnt; /* SPLIT FILE count. */
66 struct variable *weight; /* WEIGHT variable. */
67 struct variable *filter; /* FILTER variable. */
68 casenumber case_limit; /* Current case limit (N command). */
69 char *label; /* File label. */
70 struct string_array documents; /* Documents. */
71 struct vector **vector; /* Vectors of variables. */
72 size_t vector_cnt; /* Number of vectors. */
73 struct attrset attributes; /* Custom attributes. */
74 struct mrset **mrsets; /* Multiple response sets. */
75 size_t n_mrsets; /* Number of multiple response sets. */
77 char *encoding; /* Character encoding of string data */
79 const struct dict_callbacks *callbacks; /* Callbacks on dictionary
81 void *cb_data ; /* Data passed to callbacks */
83 void (*changed) (struct dictionary *, void *); /* Generic change callback */
87 static void dict_unset_split_var (struct dictionary *, struct variable *);
88 static void dict_unset_mrset_var (struct dictionary *, struct variable *);
90 /* Returns the encoding for data in dictionary D. The return value is a
91 nonnull string that contains an IANA character set name. */
93 dict_get_encoding (const struct dictionary *d)
98 /* Returns true if UTF-8 string ID is an acceptable identifier in DICT's
99 encoding, false otherwise. If ISSUE_ERROR is true, issues an explanatory
100 error message on failure. */
102 dict_id_is_valid (const struct dictionary *dict, const char *id,
105 return id_is_valid (id, dict->encoding, issue_error);
109 dict_set_change_callback (struct dictionary *d,
110 void (*changed) (struct dictionary *, void*),
113 d->changed = changed;
114 d->changed_data = data;
117 /* Discards dictionary D's caseproto. (It will be regenerated
118 lazily, on demand.) */
120 invalidate_proto (struct dictionary *d)
122 caseproto_unref (d->proto);
126 /* Print a representation of dictionary D to stdout, for
127 debugging purposes. */
129 dict_dump (const struct dictionary *d)
132 for (i = 0 ; i < d->var_cnt ; ++i )
134 const struct variable *v = d->var[i].var;
135 printf ("Name: %s;\tdict_idx: %zu; case_idx: %zu\n",
137 var_get_dict_index (v),
138 var_get_case_index (v));
143 /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
144 the dictionary or any of the variables it contains are modified.
145 Each callback will get passed CALLBACK_DATA.
146 Any callback may be NULL, in which case it'll be ignored.
149 dict_set_callbacks (struct dictionary *dict,
150 const struct dict_callbacks *callbacks,
153 dict->callbacks = callbacks;
154 dict->cb_data = callback_data;
157 /* Shallow copy the callbacks from SRC to DEST */
159 dict_copy_callbacks (struct dictionary *dest,
160 const struct dictionary *src)
162 dest->callbacks = src->callbacks;
163 dest->cb_data = src->cb_data;
166 /* Creates and returns a new dictionary with the specified ENCODING. */
168 dict_create (const char *encoding)
170 struct dictionary *d = xzalloc (sizeof *d);
172 d->encoding = xstrdup (encoding);
173 hmap_init (&d->name_map);
174 attrset_init (&d->attributes);
179 /* Creates and returns a (deep) copy of an existing
182 The new dictionary's case indexes are copied from the old
183 dictionary. If the new dictionary won't be used to access
184 cases produced with the old dictionary, then the new
185 dictionary's case indexes should be compacted with
186 dict_compact_values to save space.
188 Callbacks are not cloned. */
190 dict_clone (const struct dictionary *s)
192 struct dictionary *d;
195 d = dict_create (s->encoding);
197 for (i = 0; i < s->var_cnt; i++)
199 struct variable *sv = s->var[i].var;
200 struct variable *dv = dict_clone_var_assert (d, sv);
203 for (i = 0; i < var_get_short_name_cnt (sv); i++)
204 var_set_short_name (dv, i, var_get_short_name (sv, i));
206 var_get_vardict (dv)->case_index = var_get_vardict (sv)->case_index;
209 d->next_value_idx = s->next_value_idx;
211 d->split_cnt = s->split_cnt;
212 if (d->split_cnt > 0)
214 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
215 for (i = 0; i < d->split_cnt; i++)
216 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
219 if (s->weight != NULL)
220 dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight)));
222 if (s->filter != NULL)
223 dict_set_filter (d, dict_lookup_var_assert (d, var_get_name (s->filter)));
225 d->case_limit = s->case_limit;
226 dict_set_label (d, dict_get_label (s));
227 dict_set_documents (d, dict_get_documents (s));
229 d->vector_cnt = s->vector_cnt;
230 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
231 for (i = 0; i < s->vector_cnt; i++)
232 d->vector[i] = vector_clone (s->vector[i], s, d);
234 dict_set_attributes (d, dict_get_attributes (s));
236 for (i = 0; i < s->n_mrsets; i++)
238 const struct mrset *old = s->mrsets[i];
242 /* Clone old mrset, then replace vars from D by vars from S. */
243 new = mrset_clone (old);
244 for (j = 0; j < new->n_vars; j++)
245 new->vars[j] = dict_lookup_var_assert (d, var_get_name (new->vars[j]));
247 dict_add_mrset (d, new);
253 /* Clears the contents from a dictionary without destroying the
254 dictionary itself. */
256 dict_clear (struct dictionary *d)
258 /* FIXME? Should we really clear case_limit, label, documents?
259 Others are necessarily cleared by deleting all the variables.*/
260 while (d->var_cnt > 0 )
262 dict_delete_var (d, d->var[d->var_cnt - 1].var);
267 d->var_cnt = d->var_cap = 0;
268 invalidate_proto (d);
269 hmap_clear (&d->name_map);
270 d->next_value_idx = 0;
271 dict_set_split_vars (d, NULL, 0);
272 dict_set_weight (d, NULL);
273 dict_set_filter (d, NULL);
277 string_array_clear (&d->documents);
278 dict_clear_vectors (d);
279 attrset_clear (&d->attributes);
282 /* Clears a dictionary and destroys it. */
284 dict_destroy (struct dictionary *d)
288 /* In general, we don't want callbacks occurring, if the dictionary
289 is being destroyed */
290 d->callbacks = NULL ;
293 string_array_destroy (&d->documents);
294 hmap_destroy (&d->name_map);
295 attrset_destroy (&d->attributes);
296 dict_clear_mrsets (d);
302 /* Returns the number of variables in D. */
304 dict_get_var_cnt (const struct dictionary *d)
309 /* Returns the variable in D with dictionary index IDX, which
310 must be between 0 and the count returned by
311 dict_get_var_cnt(), exclusive. */
313 dict_get_var (const struct dictionary *d, size_t idx)
315 assert (idx < d->var_cnt);
317 return d->var[idx].var;
320 /* Sets *VARS to an array of pointers to variables in D and *CNT
321 to the number of variables in *D. All variables are returned
322 except for those, if any, in the classes indicated by EXCLUDE.
323 (There is no point in putting DC_SYSTEM in EXCLUDE as
324 dictionaries never include system variables.) */
326 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
327 size_t *cnt, enum dict_class exclude)
329 dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude);
332 /* Sets *VARS to an array of pointers to variables in D and *CNT
333 to the number of variables in *D. All variables are returned
334 except for those, if any, in the classes indicated by EXCLUDE.
335 (There is no point in putting DC_SYSTEM in EXCLUDE as
336 dictionaries never include system variables.) */
338 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
339 size_t *cnt, enum dict_class exclude)
344 assert (exclude == (exclude & DC_ALL));
347 for (i = 0; i < d->var_cnt; i++)
349 enum dict_class class = var_get_dict_class (d->var[i].var);
350 if (!(class & exclude))
354 *vars = xnmalloc (count, sizeof **vars);
356 for (i = 0; i < d->var_cnt; i++)
358 enum dict_class class = var_get_dict_class (d->var[i].var);
359 if (!(class & exclude))
360 (*vars)[(*cnt)++] = d->var[i].var;
362 assert (*cnt == count);
365 static struct variable *
366 add_var_with_case_index (struct dictionary *d, struct variable *v,
369 struct vardict_info *vardict;
371 assert (case_index >= d->next_value_idx);
373 /* Update dictionary. */
374 if (d->var_cnt >= d->var_cap)
378 d->var = x2nrealloc (d->var, &d->var_cap, sizeof *d->var);
379 hmap_clear (&d->name_map);
380 for (i = 0; i < d->var_cnt; i++)
382 var_set_vardict (d->var[i].var, &d->var[i]);
383 hmap_insert_fast (&d->name_map, &d->var[i].name_node,
384 d->var[i].name_node.hash);
388 vardict = &d->var[d->var_cnt++];
391 hmap_insert (&d->name_map, &vardict->name_node,
392 utf8_hash_case_string (var_get_name (v), 0));
393 vardict->case_index = case_index;
394 var_set_vardict (v, vardict);
396 if ( d->changed ) d->changed (d, d->changed_data);
397 if ( d->callbacks && d->callbacks->var_added )
398 d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
400 invalidate_proto (d);
401 d->next_value_idx = case_index + 1;
406 static struct variable *
407 add_var (struct dictionary *d, struct variable *v)
409 return add_var_with_case_index (d, v, d->next_value_idx);
412 /* Creates and returns a new variable in D with the given NAME
413 and WIDTH. Returns a null pointer if the given NAME would
414 duplicate that of an existing variable in the dictionary. */
416 dict_create_var (struct dictionary *d, const char *name, int width)
418 return (dict_lookup_var (d, name) == NULL
419 ? dict_create_var_assert (d, name, width)
423 /* Creates and returns a new variable in D with the given NAME
424 and WIDTH. Assert-fails if the given NAME would duplicate
425 that of an existing variable in the dictionary. */
427 dict_create_var_assert (struct dictionary *d, const char *name, int width)
429 assert (dict_lookup_var (d, name) == NULL);
430 return add_var (d, var_create (name, width));
433 /* Creates and returns a new variable in D, as a copy of existing variable
434 OLD_VAR, which need not be in D or in any dictionary. Returns a null
435 pointer if OLD_VAR's name would duplicate that of an existing variable in
438 dict_clone_var (struct dictionary *d, const struct variable *old_var)
440 return dict_clone_var_as (d, old_var, var_get_name (old_var));
443 /* Creates and returns a new variable in D, as a copy of existing variable
444 OLD_VAR, which need not be in D or in any dictionary. Assert-fails if
445 OLD_VAR's name would duplicate that of an existing variable in the
448 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var)
450 return dict_clone_var_as_assert (d, old_var, var_get_name (old_var));
453 /* Creates and returns a new variable in D with name NAME, as a copy of
454 existing variable OLD_VAR, which need not be in D or in any dictionary.
455 Returns a null pointer if the given NAME would duplicate that of an existing
456 variable in the dictionary. */
458 dict_clone_var_as (struct dictionary *d, const struct variable *old_var,
461 return (dict_lookup_var (d, name) == NULL
462 ? dict_clone_var_as_assert (d, old_var, name)
466 /* Creates and returns a new variable in D with name NAME, as a copy of
467 existing variable OLD_VAR, which need not be in D or in any dictionary.
468 Assert-fails if the given NAME would duplicate that of an existing variable
469 in the dictionary. */
471 dict_clone_var_as_assert (struct dictionary *d, const struct variable *old_var,
474 struct variable *new_var = var_clone (old_var);
475 assert (dict_lookup_var (d, name) == NULL);
476 var_set_name (new_var, name);
477 return add_var (d, new_var);
481 dict_clone_var_in_place_assert (struct dictionary *d,
482 const struct variable *old_var)
484 assert (dict_lookup_var (d, var_get_name (old_var)) == NULL);
485 return add_var_with_case_index (d, var_clone (old_var),
486 var_get_case_index (old_var));
489 /* Returns the variable named NAME in D, or a null pointer if no
490 variable has that name. */
492 dict_lookup_var (const struct dictionary *d, const char *name)
494 struct vardict_info *vardict;
496 HMAP_FOR_EACH_WITH_HASH (vardict, struct vardict_info, name_node,
497 utf8_hash_case_string (name, 0), &d->name_map)
499 struct variable *var = vardict->var;
500 if (!utf8_strcasecmp (var_get_name (var), name))
507 /* Returns the variable named NAME in D. Assert-fails if no
508 variable has that name. */
510 dict_lookup_var_assert (const struct dictionary *d, const char *name)
512 struct variable *v = dict_lookup_var (d, name);
517 /* Returns true if variable V is in dictionary D,
520 dict_contains_var (const struct dictionary *d, const struct variable *v)
522 return (var_has_vardict (v)
523 && vardict_get_dictionary (var_get_vardict (v)) == d);
526 /* Compares two double pointers to variables, which should point
527 to elements of a struct dictionary's `var' member array. */
529 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
531 struct variable *const *a = a_;
532 struct variable *const *b = b_;
534 return *a < *b ? -1 : *a > *b;
538 unindex_var (struct dictionary *d, struct vardict_info *vardict)
540 hmap_delete (&d->name_map, &vardict->name_node);
543 /* This function assumes that vardict->name_node.hash is valid, that is, that
544 its name has not changed since it was hashed (rename_var() updates this
545 hash along with the name itself). */
547 reindex_var (struct dictionary *d, struct vardict_info *vardict)
549 struct variable *old = (d->callbacks && d->callbacks->var_changed
550 ? var_clone (vardict->var)
553 struct variable *var = vardict->var;
554 var_set_vardict (var, vardict);
555 hmap_insert_fast (&d->name_map, &vardict->name_node,
556 vardict->name_node.hash);
558 if ( d->changed ) d->changed (d, d->changed_data);
561 d->callbacks->var_changed (d, var_get_dict_index (var), VAR_TRAIT_POSITION, old, d->cb_data);
566 /* Sets the case_index in V's vardict to CASE_INDEX. */
568 set_var_case_index (struct variable *v, int case_index)
570 var_get_vardict (v)->case_index = case_index;
573 /* Removes the dictionary variables with indexes from FROM to TO (exclusive)
576 unindex_vars (struct dictionary *d, size_t from, size_t to)
580 for (i = from; i < to; i++)
581 unindex_var (d, &d->var[i]);
584 /* Re-sets the dict_index in the dictionary variables with
585 indexes from FROM to TO (exclusive). */
587 reindex_vars (struct dictionary *d, size_t from, size_t to)
591 for (i = from; i < to; i++)
592 reindex_var (d, &d->var[i]);
595 /* Deletes variable V from dictionary D and frees V.
597 This is a very bad idea if there might be any pointers to V
598 from outside D. In general, no variable in the active dataset's
599 dictionary should be deleted when any transformations are
600 active on the dictionary's dataset, because those
601 transformations might reference the deleted variable. The
602 safest time to delete a variable is just after a procedure has
603 been executed, as done by DELETE VARIABLES.
605 Pointers to V within D are not a problem, because
606 dict_delete_var() knows to remove V from split variables,
607 weights, filters, etc. */
609 dict_delete_var (struct dictionary *d, struct variable *v)
611 int dict_index = var_get_dict_index (v);
612 const int case_index = var_get_case_index (v);
614 assert (dict_contains_var (d, v));
616 dict_unset_split_var (d, v);
617 dict_unset_mrset_var (d, v);
620 dict_set_weight (d, NULL);
623 dict_set_filter (d, NULL);
625 dict_clear_vectors (d);
627 /* Remove V from var array. */
628 unindex_vars (d, dict_index, d->var_cnt);
629 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
632 /* Update dict_index for each affected variable. */
633 reindex_vars (d, dict_index, d->var_cnt);
636 var_clear_vardict (v);
638 if ( d->changed ) d->changed (d, d->changed_data);
640 invalidate_proto (d);
641 if (d->callbacks && d->callbacks->var_deleted )
642 d->callbacks->var_deleted (d, v, dict_index, case_index, d->cb_data);
647 /* Deletes the COUNT variables listed in VARS from D. This is
648 unsafe; see the comment on dict_delete_var() for details. */
650 dict_delete_vars (struct dictionary *d,
651 struct variable *const *vars, size_t count)
653 /* FIXME: this can be done in O(count) time, but this algorithm
655 assert (count == 0 || vars != NULL);
658 dict_delete_var (d, *vars++);
661 /* Deletes the COUNT variables in D starting at index IDX. This
662 is unsafe; see the comment on dict_delete_var() for
665 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
667 /* FIXME: this can be done in O(count) time, but this algorithm
669 assert (idx + count <= d->var_cnt);
672 dict_delete_var (d, d->var[idx].var);
675 /* Deletes scratch variables from dictionary D. */
677 dict_delete_scratch_vars (struct dictionary *d)
681 /* FIXME: this can be done in O(count) time, but this algorithm
683 for (i = 0; i < d->var_cnt; )
684 if (var_get_dict_class (d->var[i].var) == DC_SCRATCH)
685 dict_delete_var (d, d->var[i].var);
690 /* Moves V to 0-based position IDX in D. Other variables in D,
691 if any, retain their relative positions. Runs in time linear
692 in the distance moved. */
694 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
696 size_t old_index = var_get_dict_index (v);
698 assert (new_index < d->var_cnt);
700 unindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
701 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
702 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
705 /* Reorders the variables in D, placing the COUNT variables
706 listed in ORDER in that order at the beginning of D. The
707 other variables in D, if any, retain their relative
710 dict_reorder_vars (struct dictionary *d,
711 struct variable *const *order, size_t count)
713 struct vardict_info *new_var;
716 assert (count == 0 || order != NULL);
717 assert (count <= d->var_cnt);
719 new_var = xnmalloc (d->var_cap, sizeof *new_var);
721 /* Add variables in ORDER to new_var. */
722 for (i = 0; i < count; i++)
724 struct vardict_info *old_var;
726 assert (dict_contains_var (d, order[i]));
728 old_var = var_get_vardict (order[i]);
729 new_var[i] = *old_var;
730 old_var->dict = NULL;
733 /* Add remaining variables to new_var. */
734 for (i = 0; i < d->var_cnt; i++)
735 if (d->var[i].dict != NULL)
736 new_var[count++] = d->var[i];
737 assert (count == d->var_cnt);
739 /* Replace old vardicts by new ones. */
743 hmap_clear (&d->name_map);
744 reindex_vars (d, 0, d->var_cnt);
747 /* Changes the name of variable V that is currently in a dictionary to
750 rename_var (struct variable *v, const char *new_name)
752 struct vardict_info *vardict = var_get_vardict (v);
753 var_clear_vardict (v);
754 var_set_name (v, new_name);
755 vardict->name_node.hash = utf8_hash_case_string (new_name, 0);
756 var_set_vardict (v, vardict);
759 /* Tries to changes the name of V in D to name NEW_NAME. Returns true if
760 successful, false if a variable (other than V) with the given name already
763 dict_try_rename_var (struct dictionary *d, struct variable *v,
764 const char *new_name)
766 struct variable *conflict = dict_lookup_var (d, new_name);
767 if (conflict && v != conflict)
770 struct variable *old = var_clone (v);
771 unindex_var (d, var_get_vardict (v));
772 rename_var (v, new_name);
773 reindex_var (d, var_get_vardict (v));
775 if (settings_get_algorithm () == ENHANCED)
776 var_clear_short_names (v);
778 if ( d->changed ) d->changed (d, d->changed_data);
779 if ( d->callbacks && d->callbacks->var_changed )
780 d->callbacks->var_changed (d, var_get_dict_index (v), VAR_TRAIT_NAME, old, d->cb_data);
787 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
788 a variable named NEW_NAME is already in D, except that
789 NEW_NAME may be the same as V's existing name. */
791 dict_rename_var (struct dictionary *d, struct variable *v,
792 const char *new_name)
794 bool ok UNUSED = dict_try_rename_var (d, v, new_name);
798 /* Renames COUNT variables specified in VARS to the names given
799 in NEW_NAMES within dictionary D. If the renaming would
800 result in a duplicate variable name, returns false and stores a
801 name that would be duplicated into *ERR_NAME (if ERR_NAME is
802 non-null). Otherwise, the renaming is successful, and true
805 dict_rename_vars (struct dictionary *d,
806 struct variable **vars, char **new_names, size_t count,
813 assert (count == 0 || vars != NULL);
814 assert (count == 0 || new_names != NULL);
816 /* Save the names of the variables to be renamed. */
817 pool = pool_create ();
818 old_names = pool_nalloc (pool, count, sizeof *old_names);
819 for (i = 0; i < count; i++)
820 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
822 /* Remove the variables to be renamed from the name hash,
824 for (i = 0; i < count; i++)
826 unindex_var (d, var_get_vardict (vars[i]));
827 rename_var (vars[i], new_names[i]);
830 /* Add the renamed variables back into the name hash,
831 checking for conflicts. */
832 for (i = 0; i < count; i++)
834 if (dict_lookup_var (d, var_get_name (vars[i])) != NULL)
836 /* There is a name conflict.
837 Back out all the name changes that have already
838 taken place, and indicate failure. */
840 if (err_name != NULL)
841 *err_name = new_names[i];
843 for (i = 0; i < fail_idx; i++)
844 unindex_var (d, var_get_vardict (vars[i]));
846 for (i = 0; i < count; i++)
848 rename_var (vars[i], old_names[i]);
849 reindex_var (d, var_get_vardict (vars[i]));
855 reindex_var (d, var_get_vardict (vars[i]));
858 /* Clear short names. */
859 if (settings_get_algorithm () == ENHANCED)
860 for (i = 0; i < count; i++)
861 var_clear_short_names (vars[i]);
867 /* Returns true if a variable named NAME may be inserted in DICT;
868 that is, if there is not already a variable with that name in
869 DICT and if NAME is not a reserved word. (The caller's checks
870 have already verified that NAME is otherwise acceptable as a
873 var_name_is_insertable (const struct dictionary *dict, const char *name)
875 return (dict_lookup_var (dict, name) == NULL
876 && lex_id_to_token (ss_cstr (name)) == T_ID);
880 make_hinted_name (const struct dictionary *dict, const char *hint)
882 size_t hint_len = strlen (hint);
883 bool dropped = false;
888 /* The allocation size here is OK: characters that are copied directly fit
889 OK, and characters that are not copied directly are replaced by a single
890 '_' byte. If u8_mbtouc() replaces bad input by 0xfffd, then that will get
891 replaced by '_' too. */
892 root = rp = xmalloc (hint_len + 1);
893 for (ofs = 0; ofs < hint_len; ofs += mblen)
897 mblen = u8_mbtouc (&uc, CHAR_CAST (const uint8_t *, hint + ofs),
900 ? lex_uc_is_id1 (uc) && uc != '$'
901 : lex_uc_is_idn (uc))
908 rp += u8_uctomb (CHAR_CAST (uint8_t *, rp), uc, 6);
919 if (var_name_is_insertable (dict, root))
922 for (i = 0; i < ULONG_MAX; i++)
924 char suffix[INT_BUFSIZE_BOUND (i) + 1];
928 if (!str_format_26adic (i + 1, true, &suffix[1], sizeof suffix - 1))
931 name = utf8_encoding_concat (root, suffix, dict->encoding, 64);
932 if (var_name_is_insertable (dict, name))
947 make_numeric_name (const struct dictionary *dict, unsigned long int *num_start)
949 unsigned long int number;
951 for (number = num_start != NULL ? MAX (*num_start, 1) : 1;
955 char name[3 + INT_STRLEN_BOUND (number) + 1];
957 sprintf (name, "VAR%03lu", number);
958 if (dict_lookup_var (dict, name) == NULL)
960 if (num_start != NULL)
961 *num_start = number + 1;
962 return xstrdup (name);
970 /* Devises and returns a variable name unique within DICT. The variable name
971 is owned by the caller, which must free it with free() when it is no longer
974 HINT, if it is non-null, is used as a suggestion that will be
975 modified for suitability as a variable name and for
978 If HINT is null or entirely unsuitable, a name in the form
979 "VAR%03d" will be generated, where the smallest unused integer
980 value is used. If NUM_START is non-null, then its value is
981 used as the minimum numeric value to check, and it is updated
982 to the next value to be checked.
985 dict_make_unique_var_name (const struct dictionary *dict, const char *hint,
986 unsigned long int *num_start)
990 char *hinted_name = make_hinted_name (dict, hint);
991 if (hinted_name != NULL)
994 return make_numeric_name (dict, num_start);
997 /* Returns the weighting variable in dictionary D, or a null
998 pointer if the dictionary is unweighted. */
1000 dict_get_weight (const struct dictionary *d)
1002 assert (d->weight == NULL || dict_contains_var (d, d->weight));
1007 /* Returns the value of D's weighting variable in case C, except
1008 that a negative weight is returned as 0. Returns 1 if the
1009 dictionary is unweighted. Will warn about missing, negative,
1010 or zero values if *WARN_ON_INVALID is true. The function will
1011 set *WARN_ON_INVALID to false if an invalid weight is
1014 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
1015 bool *warn_on_invalid)
1019 if (d->weight == NULL)
1023 double w = case_num (c, d->weight);
1025 return var_force_valid_weight (d->weight, w, warn_on_invalid);
1029 /* Sets the weighting variable of D to V, or turning off
1030 weighting if V is a null pointer. */
1032 dict_set_weight (struct dictionary *d, struct variable *v)
1034 assert (v == NULL || dict_contains_var (d, v));
1035 assert (v == NULL || var_is_numeric (v));
1039 if (d->changed) d->changed (d, d->changed_data);
1040 if ( d->callbacks && d->callbacks->weight_changed )
1041 d->callbacks->weight_changed (d,
1042 v ? var_get_dict_index (v) : -1,
1046 /* Returns the filter variable in dictionary D (see cmd_filter())
1047 or a null pointer if the dictionary is unfiltered. */
1049 dict_get_filter (const struct dictionary *d)
1051 assert (d->filter == NULL || dict_contains_var (d, d->filter));
1056 /* Sets V as the filter variable for dictionary D. Passing a
1057 null pointer for V turn off filtering. */
1059 dict_set_filter (struct dictionary *d, struct variable *v)
1061 assert (v == NULL || dict_contains_var (d, v));
1062 assert (v == NULL || var_is_numeric (v));
1066 if (d->changed) d->changed (d, d->changed_data);
1067 if ( d->callbacks && d->callbacks->filter_changed )
1068 d->callbacks->filter_changed (d,
1069 v ? var_get_dict_index (v) : -1,
1073 /* Returns the case limit for dictionary D, or zero if the number
1074 of cases is unlimited. */
1076 dict_get_case_limit (const struct dictionary *d)
1078 return d->case_limit;
1081 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
1082 0 for CASE_LIMIT to indicate no limit. */
1084 dict_set_case_limit (struct dictionary *d, casenumber case_limit)
1086 d->case_limit = case_limit;
1089 /* Returns the prototype used for cases created by dictionary D. */
1090 const struct caseproto *
1091 dict_get_proto (const struct dictionary *d_)
1093 struct dictionary *d = CONST_CAST (struct dictionary *, d_);
1094 if (d->proto == NULL)
1098 d->proto = caseproto_create ();
1099 d->proto = caseproto_reserve (d->proto, d->var_cnt);
1100 for (i = 0; i < d->var_cnt; i++)
1101 d->proto = caseproto_set_width (d->proto,
1102 var_get_case_index (d->var[i].var),
1103 var_get_width (d->var[i].var));
1108 /* Returns the case index of the next value to be added to D.
1109 This value is the number of `union value's that need to be
1110 allocated to store a case for dictionary D. */
1112 dict_get_next_value_idx (const struct dictionary *d)
1114 return d->next_value_idx;
1117 /* Returns the number of bytes needed to store a case for
1120 dict_get_case_size (const struct dictionary *d)
1122 return sizeof (union value) * dict_get_next_value_idx (d);
1125 /* Reassigns values in dictionary D so that fragmentation is
1128 dict_compact_values (struct dictionary *d)
1132 d->next_value_idx = 0;
1133 for (i = 0; i < d->var_cnt; i++)
1135 struct variable *v = d->var[i].var;
1136 set_var_case_index (v, d->next_value_idx++);
1138 invalidate_proto (d);
1141 /* Returns the number of values occupied by the variables in
1142 dictionary D. All variables are considered if EXCLUDE_CLASSES
1143 is 0, or it may contain one or more of (1u << DC_ORDINARY),
1144 (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to exclude the
1145 corresponding type of variable.
1147 The return value may be less than the number of values in one
1148 of dictionary D's cases (as returned by
1149 dict_get_next_value_idx) even if E is 0, because there may be
1150 gaps in D's cases due to deleted variables. */
1152 dict_count_values (const struct dictionary *d, unsigned int exclude_classes)
1157 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1159 | (1u << DC_SCRATCH))) == 0);
1162 for (i = 0; i < d->var_cnt; i++)
1164 enum dict_class class = var_get_dict_class (d->var[i].var);
1165 if (!(exclude_classes & (1u << class)))
1171 /* Returns the case prototype that would result after deleting
1172 all variables from D that are not in one of the
1173 EXCLUDE_CLASSES and compacting the dictionary with
1176 The caller must unref the returned caseproto when it is no
1179 dict_get_compacted_proto (const struct dictionary *d,
1180 unsigned int exclude_classes)
1182 struct caseproto *proto;
1185 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1187 | (1u << DC_SCRATCH))) == 0);
1189 proto = caseproto_create ();
1190 for (i = 0; i < d->var_cnt; i++)
1192 struct variable *v = d->var[i].var;
1193 if (!(exclude_classes & (1u << var_get_dict_class (v))))
1194 proto = caseproto_add_width (proto, var_get_width (v));
1199 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
1200 dict_get_split_cnt() to determine how many SPLIT FILE vars
1201 there are. Returns a null pointer if and only if there are no
1203 const struct variable *const *
1204 dict_get_split_vars (const struct dictionary *d)
1209 /* Returns the number of SPLIT FILE vars. */
1211 dict_get_split_cnt (const struct dictionary *d)
1213 return d->split_cnt;
1216 /* Removes variable V, which must be in D, from D's set of split
1219 dict_unset_split_var (struct dictionary *d, struct variable *v)
1223 assert (dict_contains_var (d, v));
1225 orig_count = d->split_cnt;
1226 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
1227 &v, compare_var_ptrs, NULL);
1228 if (orig_count != d->split_cnt)
1230 if (d->changed) d->changed (d, d->changed_data);
1231 /* We changed the set of split variables so invoke the
1233 if (d->callbacks && d->callbacks->split_changed)
1234 d->callbacks->split_changed (d, d->cb_data);
1238 /* Sets CNT split vars SPLIT in dictionary D. */
1240 dict_set_split_vars (struct dictionary *d,
1241 struct variable *const *split, size_t cnt)
1243 assert (cnt == 0 || split != NULL);
1248 d->split = xnrealloc (d->split, cnt, sizeof *d->split) ;
1249 memcpy (d->split, split, cnt * sizeof *d->split);
1257 if (d->changed) d->changed (d, d->changed_data);
1258 if ( d->callbacks && d->callbacks->split_changed )
1259 d->callbacks->split_changed (d, d->cb_data);
1262 /* Returns the file label for D, or a null pointer if D is
1263 unlabeled (see cmd_file_label()). */
1265 dict_get_label (const struct dictionary *d)
1270 /* Sets D's file label to LABEL, truncating it to at most 60 bytes in D's
1273 Removes D's label if LABEL is null or the empty string. */
1275 dict_set_label (struct dictionary *d, const char *label)
1278 if (label == NULL || label[0] == '\0')
1281 d->label = utf8_encoding_trunc (label, d->encoding, 60);
1284 /* Returns the documents for D, as an UTF-8 encoded string_array. The
1285 return value is always nonnull; if there are no documents then the
1286 string_arary is empty.*/
1287 const struct string_array *
1288 dict_get_documents (const struct dictionary *d)
1290 return &d->documents;
1293 /* Replaces the documents for D by NEW_DOCS, a UTF-8 encoded string_array. */
1295 dict_set_documents (struct dictionary *d, const struct string_array *new_docs)
1299 dict_clear_documents (d);
1301 for (i = 0; i < new_docs->n; i++)
1302 dict_add_document_line (d, new_docs->strings[i], false);
1305 /* Replaces the documents for D by UTF-8 encoded string NEW_DOCS, dividing it
1306 into individual lines at new-line characters. Each line is truncated to at
1307 most DOC_LINE_LENGTH bytes in D's encoding. */
1309 dict_set_documents_string (struct dictionary *d, const char *new_docs)
1313 dict_clear_documents (d);
1314 for (s = new_docs; *s != '\0'; )
1316 size_t len = strcspn (s, "\n");
1317 char *line = xmemdup0 (s, len);
1318 dict_add_document_line (d, line, false);
1327 /* Drops the documents from dictionary D. */
1329 dict_clear_documents (struct dictionary *d)
1331 string_array_clear (&d->documents);
1334 /* Appends the UTF-8 encoded LINE to the documents in D. LINE will be
1335 truncated so that it is no more than 80 bytes in the dictionary's
1336 encoding. If this causes some text to be lost, and ISSUE_WARNING is true,
1337 then a warning will be issued. */
1339 dict_add_document_line (struct dictionary *d, const char *line,
1345 trunc_len = utf8_encoding_trunc_len (line, d->encoding, DOC_LINE_LENGTH);
1346 truncated = line[trunc_len] != '\0';
1347 if (truncated && issue_warning)
1349 /* Note to translators: "bytes" is correct, not characters */
1350 msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
1353 string_array_append_nocopy (&d->documents, xmemdup0 (line, trunc_len));
1358 /* Returns the number of document lines in dictionary D. */
1360 dict_get_document_line_cnt (const struct dictionary *d)
1362 return d->documents.n;
1365 /* Returns document line number IDX in dictionary D. The caller must not
1366 modify or free the returned string. */
1368 dict_get_document_line (const struct dictionary *d, size_t idx)
1370 assert (idx < d->documents.n);
1371 return d->documents.strings[idx];
1374 /* Creates in D a vector named NAME that contains the CNT
1375 variables in VAR. Returns true if successful, or false if a
1376 vector named NAME already exists in D. */
1378 dict_create_vector (struct dictionary *d,
1380 struct variable **var, size_t cnt)
1385 for (i = 0; i < cnt; i++)
1386 assert (dict_contains_var (d, var[i]));
1388 if (dict_lookup_vector (d, name) == NULL)
1390 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1391 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1398 /* Creates in D a vector named NAME that contains the CNT
1399 variables in VAR. A vector named NAME must not already exist
1402 dict_create_vector_assert (struct dictionary *d,
1404 struct variable **var, size_t cnt)
1406 assert (dict_lookup_vector (d, name) == NULL);
1407 dict_create_vector (d, name, var, cnt);
1410 /* Returns the vector in D with index IDX, which must be less
1411 than dict_get_vector_cnt (D). */
1412 const struct vector *
1413 dict_get_vector (const struct dictionary *d, size_t idx)
1415 assert (idx < d->vector_cnt);
1417 return d->vector[idx];
1420 /* Returns the number of vectors in D. */
1422 dict_get_vector_cnt (const struct dictionary *d)
1424 return d->vector_cnt;
1427 /* Looks up and returns the vector within D with the given
1429 const struct vector *
1430 dict_lookup_vector (const struct dictionary *d, const char *name)
1433 for (i = 0; i < d->vector_cnt; i++)
1434 if (!utf8_strcasecmp (vector_get_name (d->vector[i]), name))
1435 return d->vector[i];
1439 /* Deletes all vectors from D. */
1441 dict_clear_vectors (struct dictionary *d)
1445 for (i = 0; i < d->vector_cnt; i++)
1446 vector_destroy (d->vector[i]);
1453 /* Multiple response sets. */
1455 /* Returns the multiple response set in DICT with index IDX, which must be
1456 between 0 and the count returned by dict_get_n_mrsets(), exclusive. */
1457 const struct mrset *
1458 dict_get_mrset (const struct dictionary *dict, size_t idx)
1460 assert (idx < dict->n_mrsets);
1461 return dict->mrsets[idx];
1464 /* Returns the number of multiple response sets in DICT. */
1466 dict_get_n_mrsets (const struct dictionary *dict)
1468 return dict->n_mrsets;
1471 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1472 returns its index; otherwise, returns SIZE_MAX. */
1474 dict_lookup_mrset_idx (const struct dictionary *dict, const char *name)
1478 for (i = 0; i < dict->n_mrsets; i++)
1479 if (!utf8_strcasecmp (name, dict->mrsets[i]->name))
1485 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1486 returns it; otherwise, returns NULL. */
1487 const struct mrset *
1488 dict_lookup_mrset (const struct dictionary *dict, const char *name)
1490 size_t idx = dict_lookup_mrset_idx (dict, name);
1491 return idx != SIZE_MAX ? dict->mrsets[idx] : NULL;
1494 /* Adds MRSET to DICT, replacing any existing set with the same name. Returns
1495 true if a set was replaced, false if none existed with the specified name.
1497 Ownership of MRSET is transferred to DICT. */
1499 dict_add_mrset (struct dictionary *dict, struct mrset *mrset)
1503 assert (mrset_ok (mrset, dict));
1505 idx = dict_lookup_mrset_idx (dict, mrset->name);
1506 if (idx == SIZE_MAX)
1508 dict->mrsets = xrealloc (dict->mrsets,
1509 (dict->n_mrsets + 1) * sizeof *dict->mrsets);
1510 dict->mrsets[dict->n_mrsets++] = mrset;
1515 mrset_destroy (dict->mrsets[idx]);
1516 dict->mrsets[idx] = mrset;
1521 /* Looks for a multiple response set in DICT named NAME. If found, removes it
1522 from DICT and returns true. If none is found, returns false without
1525 Deleting one multiple response set causes the indexes of other sets within
1528 dict_delete_mrset (struct dictionary *dict, const char *name)
1530 size_t idx = dict_lookup_mrset_idx (dict, name);
1531 if (idx != SIZE_MAX)
1533 mrset_destroy (dict->mrsets[idx]);
1534 dict->mrsets[idx] = dict->mrsets[--dict->n_mrsets];
1541 /* Deletes all multiple response sets from DICT. */
1543 dict_clear_mrsets (struct dictionary *dict)
1547 for (i = 0; i < dict->n_mrsets; i++)
1548 mrset_destroy (dict->mrsets[i]);
1549 free (dict->mrsets);
1550 dict->mrsets = NULL;
1554 /* Removes VAR, which must be in DICT, from DICT's multiple response sets. */
1556 dict_unset_mrset_var (struct dictionary *dict, struct variable *var)
1560 assert (dict_contains_var (dict, var));
1562 for (i = 0; i < dict->n_mrsets; )
1564 struct mrset *mrset = dict->mrsets[i];
1567 for (j = 0; j < mrset->n_vars; )
1568 if (mrset->vars[j] == var)
1569 remove_element (mrset->vars, mrset->n_vars--,
1570 sizeof *mrset->vars, j);
1574 if (mrset->n_vars < 2)
1576 mrset_destroy (mrset);
1577 dict->mrsets[i] = dict->mrsets[--dict->n_mrsets];
1584 /* Returns D's attribute set. The caller may examine or modify
1585 the attribute set, but must not destroy it. Destroying D or
1586 calling dict_set_attributes for D will also destroy D's
1589 dict_get_attributes (const struct dictionary *d)
1591 return CONST_CAST (struct attrset *, &d->attributes);
1594 /* Replaces D's attributes set by a copy of ATTRS. */
1596 dict_set_attributes (struct dictionary *d, const struct attrset *attrs)
1598 attrset_destroy (&d->attributes);
1599 attrset_clone (&d->attributes, attrs);
1602 /* Returns true if D has at least one attribute in its attribute
1603 set, false if D's attribute set is empty. */
1605 dict_has_attributes (const struct dictionary *d)
1607 return attrset_count (&d->attributes) > 0;
1610 /* Called from variable.c to notify the dictionary that some property (indicated
1611 by WHAT) of the variable has changed. OLDVAR is a copy of V as it existed
1612 prior to the change. OLDVAR is destroyed by this function.
1615 dict_var_changed (const struct variable *v, unsigned int what, struct variable *oldvar)
1617 if ( var_has_vardict (v))
1619 const struct vardict_info *vardict = var_get_vardict (v);
1620 struct dictionary *d = vardict->dict;
1625 if (d->changed ) d->changed (d, d->changed_data);
1626 if ( d->callbacks && d->callbacks->var_changed )
1627 d->callbacks->var_changed (d, var_get_dict_index (v), what, oldvar, d->cb_data);
1629 var_destroy (oldvar);
1634 /* Dictionary used to contain "internal variables". */
1635 static struct dictionary *internal_dict;
1637 /* Create a variable of the specified WIDTH to be used for internal
1638 calculations only. The variable is assigned case index CASE_IDX. */
1640 dict_create_internal_var (int case_idx, int width)
1642 if (internal_dict == NULL)
1643 internal_dict = dict_create ("UTF-8");
1647 static int counter = INT_MAX / 2;
1648 struct variable *var;
1651 if (++counter == INT_MAX)
1652 counter = INT_MAX / 2;
1654 sprintf (name, "$internal%d", counter);
1655 var = dict_create_var (internal_dict, name, width);
1658 set_var_case_index (var, case_idx);
1664 /* Destroys VAR, which must have been created with
1665 dict_create_internal_var(). */
1667 dict_destroy_internal_var (struct variable *var)
1671 dict_delete_var (internal_dict, var);
1673 /* Destroy internal_dict if it has no variables left, just so that
1674 valgrind --leak-check --show-reachable won't show internal_dict. */
1675 if (dict_get_var_cnt (internal_dict) == 0)
1677 dict_destroy (internal_dict);
1678 internal_dict = NULL;
1684 vardict_get_dict_index (const struct vardict_info *vardict)
1686 return vardict - vardict->dict->var;