1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006, 2007 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 "dictionary.h"
24 #include <data/attributes.h>
25 #include <data/case.h>
26 #include <data/category.h>
27 #include <data/identifier.h>
28 #include <data/settings.h>
29 #include <data/value-labels.h>
30 #include <data/vardict.h>
31 #include <data/variable.h>
32 #include <data/vector.h>
33 #include <libpspp/array.h>
34 #include <libpspp/assertion.h>
35 #include <libpspp/compiler.h>
36 #include <libpspp/hash.h>
37 #include <libpspp/message.h>
38 #include <libpspp/misc.h>
39 #include <libpspp/pool.h>
40 #include <libpspp/str.h>
47 #define _(msgid) gettext (msgid)
52 struct variable **var; /* Variables. */
53 size_t var_cnt, var_cap; /* Number of variables, capacity. */
54 struct hsh_table *name_tab; /* Variable index by name. */
55 int next_value_idx; /* Index of next `union value' to allocate. */
56 const struct variable **split; /* SPLIT FILE vars. */
57 size_t split_cnt; /* SPLIT FILE count. */
58 struct variable *weight; /* WEIGHT variable. */
59 struct variable *filter; /* FILTER variable. */
60 casenumber case_limit; /* Current case limit (N command). */
61 char *label; /* File label. */
62 struct string documents; /* Documents, as a string. */
63 struct vector **vector; /* Vectors of variables. */
64 size_t vector_cnt; /* Number of vectors. */
65 struct attrset attributes; /* Custom attributes. */
66 const struct dict_callbacks *callbacks; /* Callbacks on dictionary
68 void *cb_data ; /* Data passed to callbacks */
70 void (*changed) (struct dictionary *, void *); /* Generic change callback */
75 dict_set_change_callback (struct dictionary *d,
76 void (*changed) (struct dictionary *, void*),
80 d->changed_data = data;
84 /* Print a representation of dictionary D to stdout, for
85 debugging purposes. */
87 dict_dump (const struct dictionary *d)
90 for (i = 0 ; i < d->var_cnt ; ++i )
92 const struct variable *v =
94 printf ("Name: %s;\tdict_idx: %d; case_idx: %d\n",
96 var_get_dict_index (v),
97 var_get_case_index (v));
102 /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
103 the dictionary or any of the variables it contains are modified.
104 Each callback will get passed CALLBACK_DATA.
105 Any callback may be NULL, in which case it'll be ignored.
108 dict_set_callbacks (struct dictionary *dict,
109 const struct dict_callbacks *callbacks,
112 dict->callbacks = callbacks;
113 dict->cb_data = callback_data;
116 /* Shallow copy the callbacks from SRC to DEST */
118 dict_copy_callbacks (struct dictionary *dest,
119 const struct dictionary *src)
121 dest->callbacks = src->callbacks;
122 dest->cb_data = src->cb_data;
125 /* Creates and returns a new dictionary. */
129 struct dictionary *d = xzalloc (sizeof *d);
131 d->name_tab = hsh_create (8, compare_vars_by_name, hash_var_by_name,
133 attrset_init (&d->attributes);
137 /* Creates and returns a (deep) copy of an existing
140 The new dictionary's case indexes are copied from the old
141 dictionary. If the new dictionary won't be used to access
142 cases produced with the old dictionary, then the new
143 dictionary's case indexes should be compacted with
144 dict_compact_values to save space. */
146 dict_clone (const struct dictionary *s)
148 struct dictionary *d;
155 for (i = 0; i < s->var_cnt; i++)
157 const struct vardict_info *svdi;
158 struct vardict_info dvdi;
159 struct variable *sv = s->var[i];
160 struct variable *dv = dict_clone_var_assert (d, sv, var_get_name (sv));
163 for (i = 0; i < var_get_short_name_cnt (sv); i++)
164 var_set_short_name (dv, i, var_get_short_name (sv, i));
166 svdi = var_get_vardict (sv);
169 var_set_vardict (dv, &dvdi);
172 d->next_value_idx = s->next_value_idx;
174 d->split_cnt = s->split_cnt;
175 if (d->split_cnt > 0)
177 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
178 for (i = 0; i < d->split_cnt; i++)
179 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
182 if (s->weight != NULL)
183 dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight)));
185 if (s->filter != NULL)
186 dict_set_filter (d, dict_lookup_var_assert (d, var_get_name (s->filter)));
188 d->case_limit = s->case_limit;
189 dict_set_label (d, dict_get_label (s));
190 dict_set_documents (d, dict_get_documents (s));
192 d->vector_cnt = s->vector_cnt;
193 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
194 for (i = 0; i < s->vector_cnt; i++)
195 d->vector[i] = vector_clone (s->vector[i], s, d);
197 dict_set_attributes (d, dict_get_attributes (s));
202 /* Clears the contents from a dictionary without destroying the
203 dictionary itself. */
205 dict_clear (struct dictionary *d)
207 /* FIXME? Should we really clear case_limit, label, documents?
208 Others are necessarily cleared by deleting all the variables.*/
211 while (d->var_cnt > 0 )
213 dict_delete_var (d, d->var[d->var_cnt - 1]);
218 d->var_cnt = d->var_cap = 0;
219 hsh_clear (d->name_tab);
220 d->next_value_idx = 0;
221 dict_set_split_vars (d, NULL, 0);
222 dict_set_weight (d, NULL);
223 dict_set_filter (d, NULL);
227 ds_destroy (&d->documents);
228 dict_clear_vectors (d);
229 attrset_clear (&d->attributes);
232 /* Destroys the aux data for every variable in D, by calling
233 var_clear_aux() for each variable. */
235 dict_clear_aux (struct dictionary *d)
241 for (i = 0; i < d->var_cnt; i++)
242 var_clear_aux (d->var[i]);
245 /* Clears a dictionary and destroys it. */
247 dict_destroy (struct dictionary *d)
251 /* In general, we don't want callbacks occuring, if the dictionary
252 is being destroyed */
253 d->callbacks = NULL ;
256 hsh_destroy (d->name_tab);
257 attrset_destroy (&d->attributes);
262 /* Returns the number of variables in D. */
264 dict_get_var_cnt (const struct dictionary *d)
271 /* Returns the variable in D with dictionary index IDX, which
272 must be between 0 and the count returned by
273 dict_get_var_cnt(), exclusive. */
275 dict_get_var (const struct dictionary *d, size_t idx)
278 assert (idx < d->var_cnt);
283 /* Sets *VARS to an array of pointers to variables in D and *CNT
284 to the number of variables in *D. All variables are returned
285 except for those, if any, in the classes indicated by EXCLUDE.
286 (There is no point in putting DC_SYSTEM in EXCLUDE as
287 dictionaries never include system variables.) */
289 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
290 size_t *cnt, enum dict_class exclude)
292 dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude);
295 /* Sets *VARS to an array of pointers to variables in D and *CNT
296 to the number of variables in *D. All variables are returned
297 except for those, if any, in the classes indicated by EXCLUDE.
298 (There is no point in putting DC_SYSTEM in EXCLUDE as
299 dictionaries never include system variables.) */
301 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
302 size_t *cnt, enum dict_class exclude)
308 assert (vars != NULL);
309 assert (cnt != NULL);
310 assert (exclude == (exclude & DC_ALL));
313 for (i = 0; i < d->var_cnt; i++)
315 enum dict_class class = var_get_dict_class (d->var[i]);
316 if (!(class & exclude))
320 *vars = xnmalloc (count, sizeof **vars);
322 for (i = 0; i < d->var_cnt; i++)
324 enum dict_class class = var_get_dict_class (d->var[i]);
325 if (!(class & exclude))
326 (*vars)[(*cnt)++] = d->var[i];
328 assert (*cnt == count);
331 static struct variable *
332 add_var (struct dictionary *d, struct variable *v)
334 /* Add dictionary info to variable. */
335 struct vardict_info vdi;
336 vdi.case_index = d->next_value_idx;
337 vdi.dict_index = d->var_cnt;
339 var_set_vardict (v, &vdi);
341 /* Update dictionary. */
342 if (d->var_cnt >= d->var_cap)
344 d->var_cap = 8 + 2 * d->var_cap;
345 d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var);
347 d->var[d->var_cnt++] = v;
348 hsh_force_insert (d->name_tab, v);
350 if ( d->changed ) d->changed (d, d->changed_data);
351 if ( d->callbacks && d->callbacks->var_added )
352 d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
354 d->next_value_idx += var_get_value_cnt (v);
359 /* Creates and returns a new variable in D with the given NAME
360 and WIDTH. Returns a null pointer if the given NAME would
361 duplicate that of an existing variable in the dictionary. */
363 dict_create_var (struct dictionary *d, const char *name, int width)
365 return (dict_lookup_var (d, name) == NULL
366 ? dict_create_var_assert (d, name, width)
370 /* Creates and returns a new variable in D with the given NAME
371 and WIDTH. Assert-fails if the given NAME would duplicate
372 that of an existing variable in the dictionary. */
374 dict_create_var_assert (struct dictionary *d, const char *name, int width)
376 assert (dict_lookup_var (d, name) == NULL);
377 return add_var (d, var_create (name, width));
380 /* Creates and returns a new variable in D with name NAME, as a
381 copy of existing variable OLD_VAR, which need not be in D or
382 in any dictionary. Returns a null pointer if the given NAME
383 would duplicate that of an existing variable in the
386 dict_clone_var (struct dictionary *d, const struct variable *old_var,
389 return (dict_lookup_var (d, name) == NULL
390 ? dict_clone_var_assert (d, old_var, name)
394 /* Creates and returns a new variable in D with name NAME, as a
395 copy of existing variable OLD_VAR, which need not be in D or
396 in any dictionary. Assert-fails if the given NAME would
397 duplicate that of an existing variable in the dictionary. */
399 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var,
402 struct variable *new_var = var_clone (old_var);
403 assert (dict_lookup_var (d, name) == NULL);
404 var_set_name (new_var, name);
405 return add_var (d, new_var);
408 /* Returns the variable named NAME in D, or a null pointer if no
409 variable has that name. */
411 dict_lookup_var (const struct dictionary *d, const char *name)
413 struct variable *target ;
414 struct variable *result ;
416 if ( ! var_is_plausible_name (name, false))
419 target = var_create (name, 0);
420 result = hsh_find (d->name_tab, target);
421 var_destroy (target);
423 if ( result && var_has_vardict (result))
425 const struct vardict_info *vdi = var_get_vardict (result);
426 assert (vdi->dict == d);
432 /* Returns the variable named NAME in D. Assert-fails if no
433 variable has that name. */
435 dict_lookup_var_assert (const struct dictionary *d, const char *name)
437 struct variable *v = dict_lookup_var (d, name);
442 /* Returns true if variable V is in dictionary D,
445 dict_contains_var (const struct dictionary *d, const struct variable *v)
447 if (var_has_vardict (v))
449 const struct vardict_info *vdi = var_get_vardict (v);
450 return (vdi->dict_index >= 0
451 && vdi->dict_index < d->var_cnt
452 && d->var[vdi->dict_index] == v);
458 /* Compares two double pointers to variables, which should point
459 to elements of a struct dictionary's `var' member array. */
461 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
463 struct variable *const *a = a_;
464 struct variable *const *b = b_;
466 return *a < *b ? -1 : *a > *b;
469 /* Sets the dict_index in V's vardict to DICT_INDEX. */
471 set_var_dict_index (struct variable *v, int dict_index)
473 struct vardict_info vdi = *var_get_vardict (v);
474 struct dictionary *d = vdi.dict;
475 vdi.dict_index = dict_index;
476 var_set_vardict (v, &vdi);
478 if ( d->changed ) d->changed (d, d->changed_data);
479 if ( d->callbacks && d->callbacks->var_changed )
480 d->callbacks->var_changed (d, dict_index, d->cb_data);
483 /* Sets the case_index in V's vardict to CASE_INDEX. */
485 set_var_case_index (struct variable *v, int case_index)
487 struct vardict_info vdi = *var_get_vardict (v);
488 vdi.case_index = case_index;
489 var_set_vardict (v, &vdi);
492 /* Re-sets the dict_index in the dictionary variables with
493 indexes from FROM to TO (exclusive). */
495 reindex_vars (struct dictionary *d, size_t from, size_t to)
499 for (i = from; i < to; i++)
500 set_var_dict_index (d->var[i], i);
503 /* Deletes variable V from dictionary D and frees V.
505 This is a very bad idea if there might be any pointers to V
506 from outside D. In general, no variable in the active file's
507 dictionary should be deleted when any transformations are
508 active on the dictionary's dataset, because those
509 transformations might reference the deleted variable. The
510 safest time to delete a variable is just after a procedure has
511 been executed, as done by DELETE VARIABLES.
513 Pointers to V within D are not a problem, because
514 dict_delete_var() knows to remove V from split variables,
515 weights, filters, etc. */
517 dict_delete_var (struct dictionary *d, struct variable *v)
519 int dict_index = var_get_dict_index (v);
520 const int case_index = var_get_case_index (v);
521 const int val_cnt = var_get_value_cnt (v);
523 assert (dict_contains_var (d, v));
525 /* Delete aux data. */
528 dict_unset_split_var (d, v);
531 dict_set_weight (d, NULL);
534 dict_set_filter (d, NULL);
536 dict_clear_vectors (d);
538 /* Remove V from var array. */
539 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
542 /* Update dict_index for each affected variable. */
543 reindex_vars (d, dict_index, d->var_cnt);
545 /* Update name hash. */
546 hsh_force_delete (d->name_tab, v);
550 var_clear_vardict (v);
553 if ( d->changed ) d->changed (d, d->changed_data);
554 if (d->callbacks && d->callbacks->var_deleted )
555 d->callbacks->var_deleted (d, dict_index, case_index, val_cnt, d->cb_data);
558 /* Deletes the COUNT variables listed in VARS from D. This is
559 unsafe; see the comment on dict_delete_var() for details. */
561 dict_delete_vars (struct dictionary *d,
562 struct variable *const *vars, size_t count)
564 /* FIXME: this can be done in O(count) time, but this algorithm
567 assert (count == 0 || vars != NULL);
570 dict_delete_var (d, *vars++);
573 /* Deletes the COUNT variables in D starting at index IDX. This
574 is unsafe; see the comment on dict_delete_var() for
577 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
579 /* FIXME: this can be done in O(count) time, but this algorithm
581 assert (idx + count <= d->var_cnt);
584 dict_delete_var (d, d->var[idx]);
587 /* Deletes scratch variables from dictionary D. */
589 dict_delete_scratch_vars (struct dictionary *d)
593 /* FIXME: this can be done in O(count) time, but this algorithm
597 for (i = 0; i < d->var_cnt; )
598 if (var_get_dict_class (d->var[i]) == DC_SCRATCH)
599 dict_delete_var (d, d->var[i]);
604 /* Moves V to 0-based position IDX in D. Other variables in D,
605 if any, retain their relative positions. Runs in time linear
606 in the distance moved. */
608 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
610 size_t old_index = var_get_dict_index (v);
612 assert (new_index < d->var_cnt);
613 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
614 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
617 /* Reorders the variables in D, placing the COUNT variables
618 listed in ORDER in that order at the beginning of D. The
619 other variables in D, if any, retain their relative
622 dict_reorder_vars (struct dictionary *d,
623 struct variable *const *order, size_t count)
625 struct variable **new_var;
629 assert (count == 0 || order != NULL);
630 assert (count <= d->var_cnt);
632 new_var = xnmalloc (d->var_cnt, sizeof *new_var);
633 memcpy (new_var, order, count * sizeof *new_var);
634 for (i = 0; i < count; i++)
636 size_t index = var_get_dict_index (order[i]);
637 assert (d->var[index] == order[i]);
638 d->var[index] = NULL;
639 set_var_dict_index (order[i], i);
641 for (i = 0; i < d->var_cnt; i++)
642 if (d->var[i] != NULL)
644 assert (count < d->var_cnt);
645 new_var[count] = d->var[i];
646 set_var_dict_index (new_var[count], count);
653 /* Changes the name of variable V in dictionary D to NEW_NAME. */
655 rename_var (struct dictionary *d, struct variable *v, const char *new_name)
657 struct vardict_info vdi;
659 assert (dict_contains_var (d, v));
661 vdi = *var_get_vardict (v);
662 var_clear_vardict (v);
663 var_set_name (v, new_name);
664 var_set_vardict (v, &vdi);
667 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
668 a variable named NEW_NAME is already in D, except that
669 NEW_NAME may be the same as V's existing name. */
671 dict_rename_var (struct dictionary *d, struct variable *v,
672 const char *new_name)
674 assert (!strcasecmp (var_get_name (v), new_name)
675 || dict_lookup_var (d, new_name) == NULL);
677 hsh_force_delete (d->name_tab, v);
678 rename_var (d, v, new_name);
679 hsh_force_insert (d->name_tab, v);
681 if (settings_get_algorithm () == ENHANCED)
682 var_clear_short_names (v);
684 if ( d->changed ) d->changed (d, d->changed_data);
685 if ( d->callbacks && d->callbacks->var_changed )
686 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
689 /* Renames COUNT variables specified in VARS to the names given
690 in NEW_NAMES within dictionary D. If the renaming would
691 result in a duplicate variable name, returns false and stores a
692 name that would be duplicated into *ERR_NAME (if ERR_NAME is
693 non-null). Otherwise, the renaming is successful, and true
696 dict_rename_vars (struct dictionary *d,
697 struct variable **vars, char **new_names, size_t count,
704 assert (count == 0 || vars != NULL);
705 assert (count == 0 || new_names != NULL);
707 /* Save the names of the variables to be renamed. */
708 pool = pool_create ();
709 old_names = pool_nalloc (pool, count, sizeof *old_names);
710 for (i = 0; i < count; i++)
711 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
713 /* Remove the variables to be renamed from the name hash,
715 for (i = 0; i < count; i++)
717 hsh_force_delete (d->name_tab, vars[i]);
718 rename_var (d, vars[i], new_names[i]);
721 /* Add the renamed variables back into the name hash,
722 checking for conflicts. */
723 for (i = 0; i < count; i++)
724 if (hsh_insert (d->name_tab, vars[i]) != NULL)
726 /* There is a name conflict.
727 Back out all the name changes that have already
728 taken place, and indicate failure. */
730 if (err_name != NULL)
731 *err_name = new_names[i];
733 for (i = 0; i < fail_idx; i++)
734 hsh_force_delete (d->name_tab, vars[i]);
736 for (i = 0; i < count; i++)
738 rename_var (d, vars[i], old_names[i]);
739 hsh_force_insert (d->name_tab, vars[i]);
746 /* Clear short names. */
747 if (settings_get_algorithm () == ENHANCED)
748 for (i = 0; i < count; i++)
749 var_clear_short_names (vars[i]);
755 /* Returns true if a variable named NAME may be inserted in DICT;
756 that is, if there is not already a variable with that name in
757 DICT and if NAME is not a reserved word. (The caller's checks
758 have already verified that NAME is otherwise acceptable as a
761 var_name_is_insertable (const struct dictionary *dict, const char *name)
763 return (dict_lookup_var (dict, name) == NULL
764 && lex_id_to_token (ss_cstr (name)) == T_ID);
768 make_hinted_name (const struct dictionary *dict, const char *hint,
769 char name[VAR_NAME_LEN + 1])
771 bool dropped = false;
774 for (cp = name; *hint && cp < name + VAR_NAME_LEN; hint++)
777 ? lex_is_id1 (*hint) && *hint != '$'
778 : lex_is_idn (*hint))
785 if (cp < name + VAR_NAME_LEN)
795 size_t len = strlen (name);
798 if (var_name_is_insertable (dict, name))
801 for (i = 0; i < ULONG_MAX; i++)
803 char suffix[INT_BUFSIZE_BOUND (i) + 1];
807 if (!str_format_26adic (i + 1, &suffix[1], sizeof suffix - 1))
810 ofs = MIN (VAR_NAME_LEN - strlen (suffix), len);
811 strcpy (&name[ofs], suffix);
813 if (var_name_is_insertable (dict, name))
822 make_numeric_name (const struct dictionary *dict, unsigned long int *num_start,
823 char name[VAR_NAME_LEN + 1])
825 unsigned long int number;
827 for (number = num_start != NULL ? MAX (*num_start, 1) : 1;
831 sprintf (name, "VAR%03lu", number);
832 if (dict_lookup_var (dict, name) == NULL)
834 if (num_start != NULL)
835 *num_start = number + 1;
840 if (num_start != NULL)
841 *num_start = ULONG_MAX;
846 /* Attempts to devise a variable name unique within DICT.
847 Returns true if successful, in which case the new variable
848 name is stored into NAME. Returns false if all names that can
849 be generated have already been taken. (Returning false is
850 quite unlikely: at least ULONG_MAX unique names can be
853 HINT, if it is non-null, is used as a suggestion that will be
854 modified for suitability as a variable name and for
857 If HINT is null or entirely unsuitable, a name in the form
858 "VAR%03d" will be generated, where the smallest unused integer
859 value is used. If NUM_START is non-null, then its value is
860 used as the minimum numeric value to check, and it is updated
861 to the next value to be checked.
864 dict_make_unique_var_name (const struct dictionary *dict, const char *hint,
865 unsigned long int *num_start,
866 char name[VAR_NAME_LEN + 1])
868 return ((hint != NULL && make_hinted_name (dict, hint, name))
869 || make_numeric_name (dict, num_start, name));
872 /* Returns the weighting variable in dictionary D, or a null
873 pointer if the dictionary is unweighted. */
875 dict_get_weight (const struct dictionary *d)
878 assert (d->weight == NULL || dict_contains_var (d, d->weight));
883 /* Returns the value of D's weighting variable in case C, except
884 that a negative weight is returned as 0. Returns 1 if the
885 dictionary is unweighted. Will warn about missing, negative,
886 or zero values if *WARN_ON_INVALID is true. The function will
887 set *WARN_ON_INVALID to false if an invalid weight is
890 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
891 bool *warn_on_invalid)
896 if (d->weight == NULL)
900 double w = case_num (c, d->weight);
901 if (w < 0.0 || var_is_num_missing (d->weight, w, MV_ANY))
903 if ( w == 0.0 && warn_on_invalid != NULL && *warn_on_invalid ) {
904 *warn_on_invalid = false;
905 msg (SW, _("At least one case in the data file had a weight value "
906 "that was user-missing, system-missing, zero, or "
907 "negative. These case(s) were ignored."));
913 /* Sets the weighting variable of D to V, or turning off
914 weighting if V is a null pointer. */
916 dict_set_weight (struct dictionary *d, struct variable *v)
919 assert (v == NULL || dict_contains_var (d, v));
920 assert (v == NULL || var_is_numeric (v));
924 if (d->changed) d->changed (d, d->changed_data);
925 if ( d->callbacks && d->callbacks->weight_changed )
926 d->callbacks->weight_changed (d,
927 v ? var_get_dict_index (v) : -1,
931 /* Returns the filter variable in dictionary D (see cmd_filter())
932 or a null pointer if the dictionary is unfiltered. */
934 dict_get_filter (const struct dictionary *d)
937 assert (d->filter == NULL || dict_contains_var (d, d->filter));
942 /* Sets V as the filter variable for dictionary D. Passing a
943 null pointer for V turn off filtering. */
945 dict_set_filter (struct dictionary *d, struct variable *v)
948 assert (v == NULL || dict_contains_var (d, v));
949 assert (v == NULL || var_is_numeric (v));
953 if (d->changed) d->changed (d, d->changed_data);
954 if ( d->callbacks && d->callbacks->filter_changed )
955 d->callbacks->filter_changed (d,
956 v ? var_get_dict_index (v) : -1,
960 /* Returns the case limit for dictionary D, or zero if the number
961 of cases is unlimited. */
963 dict_get_case_limit (const struct dictionary *d)
967 return d->case_limit;
970 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
971 0 for CASE_LIMIT to indicate no limit. */
973 dict_set_case_limit (struct dictionary *d, casenumber case_limit)
977 d->case_limit = case_limit;
980 /* Returns the case index of the next value to be added to D.
981 This value is the number of `union value's that need to be
982 allocated to store a case for dictionary D. */
984 dict_get_next_value_idx (const struct dictionary *d)
988 return d->next_value_idx;
991 /* Returns the number of bytes needed to store a case for
994 dict_get_case_size (const struct dictionary *d)
998 return sizeof (union value) * dict_get_next_value_idx (d);
1001 /* Reassigns values in dictionary D so that fragmentation is
1004 dict_compact_values (struct dictionary *d)
1008 d->next_value_idx = 0;
1009 for (i = 0; i < d->var_cnt; i++)
1011 struct variable *v = d->var[i];
1012 set_var_case_index (v, d->next_value_idx);
1013 d->next_value_idx += var_get_value_cnt (v);
1018 Reassigns case indices for D, increasing each index above START by
1022 dict_pad_values (struct dictionary *d, int start, int padding)
1029 for (i = 0; i < d->var_cnt; ++i)
1031 struct variable *v = d->var[i];
1033 int index = var_get_case_index (v);
1035 if ( index >= start)
1036 set_var_case_index (v, index + padding);
1039 d->next_value_idx += padding;
1043 /* Returns the number of values occupied by the variables in
1044 dictionary D. All variables are considered if EXCLUDE_CLASSES
1045 is 0, or it may contain one or more of (1u << DC_ORDINARY),
1046 (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to exclude the
1047 corresponding type of variable.
1049 The return value may be less than the number of values in one
1050 of dictionary D's cases (as returned by
1051 dict_get_next_value_idx) even if E is 0, because there may be
1052 gaps in D's cases due to deleted variables. */
1054 dict_count_values (const struct dictionary *d, unsigned int exclude_classes)
1059 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1061 | (1u << DC_SCRATCH))) == 0);
1064 for (i = 0; i < d->var_cnt; i++)
1066 enum dict_class class = var_get_dict_class (d->var[i]);
1067 if (!(exclude_classes & (1u << class)))
1068 cnt += var_get_value_cnt (d->var[i]);
1073 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
1074 dict_get_split_cnt() to determine how many SPLIT FILE vars
1075 there are. Returns a null pointer if and only if there are no
1077 const struct variable *const *
1078 dict_get_split_vars (const struct dictionary *d)
1085 /* Returns the number of SPLIT FILE vars. */
1087 dict_get_split_cnt (const struct dictionary *d)
1091 return d->split_cnt;
1094 /* Removes variable V, which must be in D, from D's set of split
1097 dict_unset_split_var (struct dictionary *d, struct variable *v)
1101 assert (dict_contains_var (d, v));
1103 orig_count = d->split_cnt;
1104 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
1105 &v, compare_var_ptrs, NULL);
1106 if (orig_count != d->split_cnt)
1108 if (d->changed) d->changed (d, d->changed_data);
1109 /* We changed the set of split variables so invoke the
1111 if (d->callbacks && d->callbacks->split_changed)
1112 d->callbacks->split_changed (d, d->cb_data);
1116 /* Sets CNT split vars SPLIT in dictionary D. */
1118 dict_set_split_vars (struct dictionary *d,
1119 struct variable *const *split, size_t cnt)
1122 assert (cnt == 0 || split != NULL);
1127 d->split = xnrealloc (d->split, cnt, sizeof *d->split) ;
1128 memcpy (d->split, split, cnt * sizeof *d->split);
1136 if (d->changed) d->changed (d, d->changed_data);
1137 if ( d->callbacks && d->callbacks->split_changed )
1138 d->callbacks->split_changed (d, d->cb_data);
1141 /* Returns the file label for D, or a null pointer if D is
1142 unlabeled (see cmd_file_label()). */
1144 dict_get_label (const struct dictionary *d)
1151 /* Sets D's file label to LABEL, truncating it to a maximum of 60
1154 dict_set_label (struct dictionary *d, const char *label)
1159 d->label = label != NULL ? xstrndup (label, 60) : NULL;
1162 /* Returns the documents for D, or a null pointer if D has no
1163 documents. If the return value is nonnull, then the string
1164 will be an exact multiple of DOC_LINE_LENGTH bytes in length,
1165 with each segment corresponding to one line. */
1167 dict_get_documents (const struct dictionary *d)
1169 return ds_is_empty (&d->documents) ? NULL : ds_cstr (&d->documents);
1172 /* Sets the documents for D to DOCUMENTS, or removes D's
1173 documents if DOCUMENT is a null pointer. If DOCUMENTS is
1174 nonnull, then it should be an exact multiple of
1175 DOC_LINE_LENGTH bytes in length, with each segment
1176 corresponding to one line. */
1178 dict_set_documents (struct dictionary *d, const char *documents)
1182 ds_assign_cstr (&d->documents, documents != NULL ? documents : "");
1184 /* In case the caller didn't get it quite right, pad out the
1185 final line with spaces. */
1186 remainder = ds_length (&d->documents) % DOC_LINE_LENGTH;
1188 ds_put_char_multiple (&d->documents, ' ', DOC_LINE_LENGTH - remainder);
1191 /* Drops the documents from dictionary D. */
1193 dict_clear_documents (struct dictionary *d)
1195 ds_clear (&d->documents);
1198 /* Appends LINE to the documents in D. LINE will be truncated or
1199 padded on the right with spaces to make it exactly
1200 DOC_LINE_LENGTH bytes long. */
1202 dict_add_document_line (struct dictionary *d, const char *line)
1204 if (strlen (line) > DOC_LINE_LENGTH)
1206 /* Note to translators: "bytes" is correct, not characters */
1207 msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
1209 buf_copy_str_rpad (ds_put_uninit (&d->documents, DOC_LINE_LENGTH),
1210 DOC_LINE_LENGTH, line);
1213 /* Returns the number of document lines in dictionary D. */
1215 dict_get_document_line_cnt (const struct dictionary *d)
1217 return ds_length (&d->documents) / DOC_LINE_LENGTH;
1220 /* Copies document line number IDX from dictionary D into
1221 LINE, trimming off any trailing white space. */
1223 dict_get_document_line (const struct dictionary *d,
1224 size_t idx, struct string *line)
1226 assert (idx < dict_get_document_line_cnt (d));
1227 ds_assign_substring (line, ds_substr (&d->documents, idx * DOC_LINE_LENGTH,
1229 ds_rtrim (line, ss_cstr (CC_SPACES));
1232 /* Creates in D a vector named NAME that contains the CNT
1233 variables in VAR. Returns true if successful, or false if a
1234 vector named NAME already exists in D. */
1236 dict_create_vector (struct dictionary *d,
1238 struct variable **var, size_t cnt)
1242 assert (var != NULL);
1244 for (i = 0; i < cnt; i++)
1245 assert (dict_contains_var (d, var[i]));
1247 if (dict_lookup_vector (d, name) == NULL)
1249 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1250 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1257 /* Creates in D a vector named NAME that contains the CNT
1258 variables in VAR. A vector named NAME must not already exist
1261 dict_create_vector_assert (struct dictionary *d,
1263 struct variable **var, size_t cnt)
1265 assert (dict_lookup_vector (d, name) == NULL);
1266 dict_create_vector (d, name, var, cnt);
1269 /* Returns the vector in D with index IDX, which must be less
1270 than dict_get_vector_cnt (D). */
1271 const struct vector *
1272 dict_get_vector (const struct dictionary *d, size_t idx)
1275 assert (idx < d->vector_cnt);
1277 return d->vector[idx];
1280 /* Returns the number of vectors in D. */
1282 dict_get_vector_cnt (const struct dictionary *d)
1286 return d->vector_cnt;
1289 /* Looks up and returns the vector within D with the given
1291 const struct vector *
1292 dict_lookup_vector (const struct dictionary *d, const char *name)
1295 for (i = 0; i < d->vector_cnt; i++)
1296 if (!strcasecmp (vector_get_name (d->vector[i]), name))
1297 return d->vector[i];
1301 /* Deletes all vectors from D. */
1303 dict_clear_vectors (struct dictionary *d)
1307 for (i = 0; i < d->vector_cnt; i++)
1308 vector_destroy (d->vector[i]);
1315 /* Returns D's attribute set. The caller may examine or modify
1316 the attribute set, but must not destroy it. Destroying D or
1317 calling dict_set_attributes for D will also destroy D's
1320 dict_get_attributes (const struct dictionary *d)
1322 return (struct attrset *) &d->attributes;
1325 /* Replaces D's attributes set by a copy of ATTRS. */
1327 dict_set_attributes (struct dictionary *d, const struct attrset *attrs)
1329 attrset_destroy (&d->attributes);
1330 attrset_clone (&d->attributes, attrs);
1333 /* Returns true if D has at least one attribute in its attribute
1334 set, false if D's attribute set is empty. */
1336 dict_has_attributes (const struct dictionary *d)
1338 return attrset_count (&d->attributes) > 0;
1341 /* Called from variable.c to notify the dictionary that some property of
1342 the variable has changed */
1344 dict_var_changed (const struct variable *v)
1346 if ( var_has_vardict (v))
1348 const struct vardict_info *vdi = var_get_vardict (v);
1349 struct dictionary *d = vdi->dict;
1351 if (d->changed ) d->changed (d, d->changed_data);
1352 if ( d->callbacks && d->callbacks->var_changed )
1353 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
1358 /* Called from variable.c to notify the dictionary that the variable's width
1361 dict_var_resized (const struct variable *v, int delta)
1363 if ( var_has_vardict (v))
1365 const struct vardict_info *vdi = var_get_vardict (v);
1366 struct dictionary *d;
1370 dict_pad_values (d, var_get_case_index(v) + 1, delta);
1372 if (d->changed) d->changed (d, d->changed_data);
1373 if ( d->callbacks && d->callbacks->var_resized )
1374 d->callbacks->var_resized (d, var_get_dict_index (v), delta, d->cb_data);
1378 /* Called from variable.c to notify the dictionary that the variable's display width
1381 dict_var_display_width_changed (const struct variable *v)
1383 if ( var_has_vardict (v))
1385 const struct vardict_info *vdi = var_get_vardict (v);
1386 struct dictionary *d;
1390 if (d->changed) d->changed (d, d->changed_data);
1391 if ( d->callbacks && d->callbacks->var_display_width_changed )
1392 d->callbacks->var_display_width_changed (d, var_get_dict_index (v), d->cb_data);