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. */
67 char *encoding; /* Character encoding of string data */
69 const struct dict_callbacks *callbacks; /* Callbacks on dictionary
71 void *cb_data ; /* Data passed to callbacks */
73 void (*changed) (struct dictionary *, void *); /* Generic change callback */
79 dict_set_encoding (struct dictionary *d, const char *enc)
82 d->encoding = strdup (enc);
86 dict_get_encoding (const struct dictionary *d)
93 dict_set_change_callback (struct dictionary *d,
94 void (*changed) (struct dictionary *, void*),
98 d->changed_data = data;
102 /* Print a representation of dictionary D to stdout, for
103 debugging purposes. */
105 dict_dump (const struct dictionary *d)
108 for (i = 0 ; i < d->var_cnt ; ++i )
110 const struct variable *v =
112 printf ("Name: %s;\tdict_idx: %d; case_idx: %d\n",
114 var_get_dict_index (v),
115 var_get_case_index (v));
120 /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
121 the dictionary or any of the variables it contains are modified.
122 Each callback will get passed CALLBACK_DATA.
123 Any callback may be NULL, in which case it'll be ignored.
126 dict_set_callbacks (struct dictionary *dict,
127 const struct dict_callbacks *callbacks,
130 dict->callbacks = callbacks;
131 dict->cb_data = callback_data;
134 /* Shallow copy the callbacks from SRC to DEST */
136 dict_copy_callbacks (struct dictionary *dest,
137 const struct dictionary *src)
139 dest->callbacks = src->callbacks;
140 dest->cb_data = src->cb_data;
143 /* Creates and returns a new dictionary. */
147 struct dictionary *d = xzalloc (sizeof *d);
149 d->name_tab = hsh_create (8, compare_vars_by_name, hash_var_by_name,
151 attrset_init (&d->attributes);
155 /* Creates and returns a (deep) copy of an existing
158 The new dictionary's case indexes are copied from the old
159 dictionary. If the new dictionary won't be used to access
160 cases produced with the old dictionary, then the new
161 dictionary's case indexes should be compacted with
162 dict_compact_values to save space. */
164 dict_clone (const struct dictionary *s)
166 struct dictionary *d;
173 for (i = 0; i < s->var_cnt; i++)
175 const struct vardict_info *svdi;
176 struct vardict_info dvdi;
177 struct variable *sv = s->var[i];
178 struct variable *dv = dict_clone_var_assert (d, sv, var_get_name (sv));
181 for (i = 0; i < var_get_short_name_cnt (sv); i++)
182 var_set_short_name (dv, i, var_get_short_name (sv, i));
184 svdi = var_get_vardict (sv);
187 var_set_vardict (dv, &dvdi);
190 d->next_value_idx = s->next_value_idx;
192 d->split_cnt = s->split_cnt;
193 if (d->split_cnt > 0)
195 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
196 for (i = 0; i < d->split_cnt; i++)
197 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
200 if (s->weight != NULL)
201 dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight)));
203 if (s->filter != NULL)
204 dict_set_filter (d, dict_lookup_var_assert (d, var_get_name (s->filter)));
206 d->case_limit = s->case_limit;
207 dict_set_label (d, dict_get_label (s));
208 dict_set_documents (d, dict_get_documents (s));
210 d->vector_cnt = s->vector_cnt;
211 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
212 for (i = 0; i < s->vector_cnt; i++)
213 d->vector[i] = vector_clone (s->vector[i], s, d);
216 d->encoding = strdup (s->encoding);
218 dict_set_attributes (d, dict_get_attributes (s));
223 /* Clears the contents from a dictionary without destroying the
224 dictionary itself. */
226 dict_clear (struct dictionary *d)
228 /* FIXME? Should we really clear case_limit, label, documents?
229 Others are necessarily cleared by deleting all the variables.*/
232 while (d->var_cnt > 0 )
234 dict_delete_var (d, d->var[d->var_cnt - 1]);
239 d->var_cnt = d->var_cap = 0;
240 hsh_clear (d->name_tab);
241 d->next_value_idx = 0;
242 dict_set_split_vars (d, NULL, 0);
243 dict_set_weight (d, NULL);
244 dict_set_filter (d, NULL);
248 ds_destroy (&d->documents);
249 dict_clear_vectors (d);
250 attrset_clear (&d->attributes);
253 /* Destroys the aux data for every variable in D, by calling
254 var_clear_aux() for each variable. */
256 dict_clear_aux (struct dictionary *d)
262 for (i = 0; i < d->var_cnt; i++)
263 var_clear_aux (d->var[i]);
266 /* Clears a dictionary and destroys it. */
268 dict_destroy (struct dictionary *d)
272 /* In general, we don't want callbacks occuring, if the dictionary
273 is being destroyed */
274 d->callbacks = NULL ;
277 hsh_destroy (d->name_tab);
278 attrset_destroy (&d->attributes);
283 /* Returns the number of variables in D. */
285 dict_get_var_cnt (const struct dictionary *d)
292 /* Returns the variable in D with dictionary index IDX, which
293 must be between 0 and the count returned by
294 dict_get_var_cnt(), exclusive. */
296 dict_get_var (const struct dictionary *d, size_t idx)
299 assert (idx < d->var_cnt);
304 /* Sets *VARS to an array of pointers to variables in D and *CNT
305 to the number of variables in *D. All variables are returned
306 except for those, if any, in the classes indicated by EXCLUDE.
307 (There is no point in putting DC_SYSTEM in EXCLUDE as
308 dictionaries never include system variables.) */
310 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
311 size_t *cnt, enum dict_class exclude)
313 dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude);
316 /* Sets *VARS to an array of pointers to variables in D and *CNT
317 to the number of variables in *D. All variables are returned
318 except for those, if any, in the classes indicated by EXCLUDE.
319 (There is no point in putting DC_SYSTEM in EXCLUDE as
320 dictionaries never include system variables.) */
322 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
323 size_t *cnt, enum dict_class exclude)
329 assert (vars != NULL);
330 assert (cnt != NULL);
331 assert (exclude == (exclude & DC_ALL));
334 for (i = 0; i < d->var_cnt; i++)
336 enum dict_class class = var_get_dict_class (d->var[i]);
337 if (!(class & exclude))
341 *vars = xnmalloc (count, sizeof **vars);
343 for (i = 0; i < d->var_cnt; i++)
345 enum dict_class class = var_get_dict_class (d->var[i]);
346 if (!(class & exclude))
347 (*vars)[(*cnt)++] = d->var[i];
349 assert (*cnt == count);
352 static struct variable *
353 add_var (struct dictionary *d, struct variable *v)
355 /* Add dictionary info to variable. */
356 struct vardict_info vdi;
357 vdi.case_index = d->next_value_idx;
358 vdi.dict_index = d->var_cnt;
360 var_set_vardict (v, &vdi);
362 /* Update dictionary. */
363 if (d->var_cnt >= d->var_cap)
365 d->var_cap = 8 + 2 * d->var_cap;
366 d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var);
368 d->var[d->var_cnt++] = v;
369 hsh_force_insert (d->name_tab, v);
371 if ( d->changed ) d->changed (d, d->changed_data);
372 if ( d->callbacks && d->callbacks->var_added )
373 d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
375 d->next_value_idx += var_get_value_cnt (v);
380 /* Creates and returns a new variable in D with the given NAME
381 and WIDTH. Returns a null pointer if the given NAME would
382 duplicate that of an existing variable in the dictionary. */
384 dict_create_var (struct dictionary *d, const char *name, int width)
386 return (dict_lookup_var (d, name) == NULL
387 ? dict_create_var_assert (d, name, width)
391 /* Creates and returns a new variable in D with the given NAME
392 and WIDTH. Assert-fails if the given NAME would duplicate
393 that of an existing variable in the dictionary. */
395 dict_create_var_assert (struct dictionary *d, const char *name, int width)
397 assert (dict_lookup_var (d, name) == NULL);
398 return add_var (d, var_create (name, width));
401 /* Creates and returns a new variable in D with name NAME, as a
402 copy of existing variable OLD_VAR, which need not be in D or
403 in any dictionary. Returns a null pointer if the given NAME
404 would duplicate that of an existing variable in the
407 dict_clone_var (struct dictionary *d, const struct variable *old_var,
410 return (dict_lookup_var (d, name) == NULL
411 ? dict_clone_var_assert (d, old_var, name)
415 /* Creates and returns a new variable in D with name NAME, as a
416 copy of existing variable OLD_VAR, which need not be in D or
417 in any dictionary. Assert-fails if the given NAME would
418 duplicate that of an existing variable in the dictionary. */
420 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var,
423 struct variable *new_var = var_clone (old_var);
424 assert (dict_lookup_var (d, name) == NULL);
425 var_set_name (new_var, name);
426 return add_var (d, new_var);
429 /* Returns the variable named NAME in D, or a null pointer if no
430 variable has that name. */
432 dict_lookup_var (const struct dictionary *d, const char *name)
434 struct variable *target ;
435 struct variable *result ;
437 if ( ! var_is_plausible_name (name, false))
440 target = var_create (name, 0);
441 result = hsh_find (d->name_tab, target);
442 var_destroy (target);
444 if ( result && var_has_vardict (result))
446 const struct vardict_info *vdi = var_get_vardict (result);
447 assert (vdi->dict == d);
453 /* Returns the variable named NAME in D. Assert-fails if no
454 variable has that name. */
456 dict_lookup_var_assert (const struct dictionary *d, const char *name)
458 struct variable *v = dict_lookup_var (d, name);
463 /* Returns true if variable V is in dictionary D,
466 dict_contains_var (const struct dictionary *d, const struct variable *v)
468 if (var_has_vardict (v))
470 const struct vardict_info *vdi = var_get_vardict (v);
471 return (vdi->dict_index >= 0
472 && vdi->dict_index < d->var_cnt
473 && d->var[vdi->dict_index] == v);
479 /* Compares two double pointers to variables, which should point
480 to elements of a struct dictionary's `var' member array. */
482 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
484 struct variable *const *a = a_;
485 struct variable *const *b = b_;
487 return *a < *b ? -1 : *a > *b;
490 /* Sets the dict_index in V's vardict to DICT_INDEX. */
492 set_var_dict_index (struct variable *v, int dict_index)
494 struct vardict_info vdi = *var_get_vardict (v);
495 struct dictionary *d = vdi.dict;
496 vdi.dict_index = dict_index;
497 var_set_vardict (v, &vdi);
499 if ( d->changed ) d->changed (d, d->changed_data);
500 if ( d->callbacks && d->callbacks->var_changed )
501 d->callbacks->var_changed (d, dict_index, d->cb_data);
504 /* Sets the case_index in V's vardict to CASE_INDEX. */
506 set_var_case_index (struct variable *v, int case_index)
508 struct vardict_info vdi = *var_get_vardict (v);
509 vdi.case_index = case_index;
510 var_set_vardict (v, &vdi);
513 /* Re-sets the dict_index in the dictionary variables with
514 indexes from FROM to TO (exclusive). */
516 reindex_vars (struct dictionary *d, size_t from, size_t to)
520 for (i = from; i < to; i++)
521 set_var_dict_index (d->var[i], i);
524 /* Deletes variable V from dictionary D and frees V.
526 This is a very bad idea if there might be any pointers to V
527 from outside D. In general, no variable in the active file's
528 dictionary should be deleted when any transformations are
529 active on the dictionary's dataset, because those
530 transformations might reference the deleted variable. The
531 safest time to delete a variable is just after a procedure has
532 been executed, as done by DELETE VARIABLES.
534 Pointers to V within D are not a problem, because
535 dict_delete_var() knows to remove V from split variables,
536 weights, filters, etc. */
538 dict_delete_var (struct dictionary *d, struct variable *v)
540 int dict_index = var_get_dict_index (v);
541 const int case_index = var_get_case_index (v);
542 const int val_cnt = var_get_value_cnt (v);
544 assert (dict_contains_var (d, v));
546 /* Delete aux data. */
549 dict_unset_split_var (d, v);
552 dict_set_weight (d, NULL);
555 dict_set_filter (d, NULL);
557 dict_clear_vectors (d);
559 /* Remove V from var array. */
560 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
563 /* Update dict_index for each affected variable. */
564 reindex_vars (d, dict_index, d->var_cnt);
566 /* Update name hash. */
567 hsh_force_delete (d->name_tab, v);
571 var_clear_vardict (v);
574 if ( d->changed ) d->changed (d, d->changed_data);
575 if (d->callbacks && d->callbacks->var_deleted )
576 d->callbacks->var_deleted (d, dict_index, case_index, val_cnt, d->cb_data);
579 /* Deletes the COUNT variables listed in VARS from D. This is
580 unsafe; see the comment on dict_delete_var() for details. */
582 dict_delete_vars (struct dictionary *d,
583 struct variable *const *vars, size_t count)
585 /* FIXME: this can be done in O(count) time, but this algorithm
588 assert (count == 0 || vars != NULL);
591 dict_delete_var (d, *vars++);
594 /* Deletes the COUNT variables in D starting at index IDX. This
595 is unsafe; see the comment on dict_delete_var() for
598 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
600 /* FIXME: this can be done in O(count) time, but this algorithm
602 assert (idx + count <= d->var_cnt);
605 dict_delete_var (d, d->var[idx]);
608 /* Deletes scratch variables from dictionary D. */
610 dict_delete_scratch_vars (struct dictionary *d)
614 /* FIXME: this can be done in O(count) time, but this algorithm
618 for (i = 0; i < d->var_cnt; )
619 if (var_get_dict_class (d->var[i]) == DC_SCRATCH)
620 dict_delete_var (d, d->var[i]);
625 /* Moves V to 0-based position IDX in D. Other variables in D,
626 if any, retain their relative positions. Runs in time linear
627 in the distance moved. */
629 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
631 size_t old_index = var_get_dict_index (v);
633 assert (new_index < d->var_cnt);
634 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
635 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
638 /* Reorders the variables in D, placing the COUNT variables
639 listed in ORDER in that order at the beginning of D. The
640 other variables in D, if any, retain their relative
643 dict_reorder_vars (struct dictionary *d,
644 struct variable *const *order, size_t count)
646 struct variable **new_var;
650 assert (count == 0 || order != NULL);
651 assert (count <= d->var_cnt);
653 new_var = xnmalloc (d->var_cnt, sizeof *new_var);
654 memcpy (new_var, order, count * sizeof *new_var);
655 for (i = 0; i < count; i++)
657 size_t index = var_get_dict_index (order[i]);
658 assert (d->var[index] == order[i]);
659 d->var[index] = NULL;
660 set_var_dict_index (order[i], i);
662 for (i = 0; i < d->var_cnt; i++)
663 if (d->var[i] != NULL)
665 assert (count < d->var_cnt);
666 new_var[count] = d->var[i];
667 set_var_dict_index (new_var[count], count);
674 /* Changes the name of variable V in dictionary D to NEW_NAME. */
676 rename_var (struct dictionary *d, struct variable *v, const char *new_name)
678 struct vardict_info vdi;
680 assert (dict_contains_var (d, v));
682 vdi = *var_get_vardict (v);
683 var_clear_vardict (v);
684 var_set_name (v, new_name);
685 var_set_vardict (v, &vdi);
688 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
689 a variable named NEW_NAME is already in D, except that
690 NEW_NAME may be the same as V's existing name. */
692 dict_rename_var (struct dictionary *d, struct variable *v,
693 const char *new_name)
695 assert (!strcasecmp (var_get_name (v), new_name)
696 || dict_lookup_var (d, new_name) == NULL);
698 hsh_force_delete (d->name_tab, v);
699 rename_var (d, v, new_name);
700 hsh_force_insert (d->name_tab, v);
702 if (settings_get_algorithm () == ENHANCED)
703 var_clear_short_names (v);
705 if ( d->changed ) d->changed (d, d->changed_data);
706 if ( d->callbacks && d->callbacks->var_changed )
707 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
710 /* Renames COUNT variables specified in VARS to the names given
711 in NEW_NAMES within dictionary D. If the renaming would
712 result in a duplicate variable name, returns false and stores a
713 name that would be duplicated into *ERR_NAME (if ERR_NAME is
714 non-null). Otherwise, the renaming is successful, and true
717 dict_rename_vars (struct dictionary *d,
718 struct variable **vars, char **new_names, size_t count,
725 assert (count == 0 || vars != NULL);
726 assert (count == 0 || new_names != NULL);
728 /* Save the names of the variables to be renamed. */
729 pool = pool_create ();
730 old_names = pool_nalloc (pool, count, sizeof *old_names);
731 for (i = 0; i < count; i++)
732 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
734 /* Remove the variables to be renamed from the name hash,
736 for (i = 0; i < count; i++)
738 hsh_force_delete (d->name_tab, vars[i]);
739 rename_var (d, vars[i], new_names[i]);
742 /* Add the renamed variables back into the name hash,
743 checking for conflicts. */
744 for (i = 0; i < count; i++)
745 if (hsh_insert (d->name_tab, vars[i]) != NULL)
747 /* There is a name conflict.
748 Back out all the name changes that have already
749 taken place, and indicate failure. */
751 if (err_name != NULL)
752 *err_name = new_names[i];
754 for (i = 0; i < fail_idx; i++)
755 hsh_force_delete (d->name_tab, vars[i]);
757 for (i = 0; i < count; i++)
759 rename_var (d, vars[i], old_names[i]);
760 hsh_force_insert (d->name_tab, vars[i]);
767 /* Clear short names. */
768 if (settings_get_algorithm () == ENHANCED)
769 for (i = 0; i < count; i++)
770 var_clear_short_names (vars[i]);
776 /* Returns true if a variable named NAME may be inserted in DICT;
777 that is, if there is not already a variable with that name in
778 DICT and if NAME is not a reserved word. (The caller's checks
779 have already verified that NAME is otherwise acceptable as a
782 var_name_is_insertable (const struct dictionary *dict, const char *name)
784 return (dict_lookup_var (dict, name) == NULL
785 && lex_id_to_token (ss_cstr (name)) == T_ID);
789 make_hinted_name (const struct dictionary *dict, const char *hint,
790 char name[VAR_NAME_LEN + 1])
792 bool dropped = false;
795 for (cp = name; *hint && cp < name + VAR_NAME_LEN; hint++)
798 ? lex_is_id1 (*hint) && *hint != '$'
799 : lex_is_idn (*hint))
806 if (cp < name + VAR_NAME_LEN)
816 size_t len = strlen (name);
819 if (var_name_is_insertable (dict, name))
822 for (i = 0; i < ULONG_MAX; i++)
824 char suffix[INT_BUFSIZE_BOUND (i) + 1];
828 if (!str_format_26adic (i + 1, &suffix[1], sizeof suffix - 1))
831 ofs = MIN (VAR_NAME_LEN - strlen (suffix), len);
832 strcpy (&name[ofs], suffix);
834 if (var_name_is_insertable (dict, name))
843 make_numeric_name (const struct dictionary *dict, unsigned long int *num_start,
844 char name[VAR_NAME_LEN + 1])
846 unsigned long int number;
848 for (number = num_start != NULL ? MAX (*num_start, 1) : 1;
852 sprintf (name, "VAR%03lu", number);
853 if (dict_lookup_var (dict, name) == NULL)
855 if (num_start != NULL)
856 *num_start = number + 1;
861 if (num_start != NULL)
862 *num_start = ULONG_MAX;
867 /* Attempts to devise a variable name unique within DICT.
868 Returns true if successful, in which case the new variable
869 name is stored into NAME. Returns false if all names that can
870 be generated have already been taken. (Returning false is
871 quite unlikely: at least ULONG_MAX unique names can be
874 HINT, if it is non-null, is used as a suggestion that will be
875 modified for suitability as a variable name and for
878 If HINT is null or entirely unsuitable, a name in the form
879 "VAR%03d" will be generated, where the smallest unused integer
880 value is used. If NUM_START is non-null, then its value is
881 used as the minimum numeric value to check, and it is updated
882 to the next value to be checked.
885 dict_make_unique_var_name (const struct dictionary *dict, const char *hint,
886 unsigned long int *num_start,
887 char name[VAR_NAME_LEN + 1])
889 return ((hint != NULL && make_hinted_name (dict, hint, name))
890 || make_numeric_name (dict, num_start, name));
893 /* Returns the weighting variable in dictionary D, or a null
894 pointer if the dictionary is unweighted. */
896 dict_get_weight (const struct dictionary *d)
899 assert (d->weight == NULL || dict_contains_var (d, d->weight));
904 /* Returns the value of D's weighting variable in case C, except
905 that a negative weight is returned as 0. Returns 1 if the
906 dictionary is unweighted. Will warn about missing, negative,
907 or zero values if *WARN_ON_INVALID is true. The function will
908 set *WARN_ON_INVALID to false if an invalid weight is
911 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
912 bool *warn_on_invalid)
917 if (d->weight == NULL)
921 double w = case_num (c, d->weight);
922 if (w < 0.0 || var_is_num_missing (d->weight, w, MV_ANY))
924 if ( w == 0.0 && warn_on_invalid != NULL && *warn_on_invalid ) {
925 *warn_on_invalid = false;
926 msg (SW, _("At least one case in the data file had a weight value "
927 "that was user-missing, system-missing, zero, or "
928 "negative. These case(s) were ignored."));
934 /* Sets the weighting variable of D to V, or turning off
935 weighting if V is a null pointer. */
937 dict_set_weight (struct dictionary *d, struct variable *v)
940 assert (v == NULL || dict_contains_var (d, v));
941 assert (v == NULL || var_is_numeric (v));
945 if (d->changed) d->changed (d, d->changed_data);
946 if ( d->callbacks && d->callbacks->weight_changed )
947 d->callbacks->weight_changed (d,
948 v ? var_get_dict_index (v) : -1,
952 /* Returns the filter variable in dictionary D (see cmd_filter())
953 or a null pointer if the dictionary is unfiltered. */
955 dict_get_filter (const struct dictionary *d)
958 assert (d->filter == NULL || dict_contains_var (d, d->filter));
963 /* Sets V as the filter variable for dictionary D. Passing a
964 null pointer for V turn off filtering. */
966 dict_set_filter (struct dictionary *d, struct variable *v)
969 assert (v == NULL || dict_contains_var (d, v));
970 assert (v == NULL || var_is_numeric (v));
974 if (d->changed) d->changed (d, d->changed_data);
975 if ( d->callbacks && d->callbacks->filter_changed )
976 d->callbacks->filter_changed (d,
977 v ? var_get_dict_index (v) : -1,
981 /* Returns the case limit for dictionary D, or zero if the number
982 of cases is unlimited. */
984 dict_get_case_limit (const struct dictionary *d)
988 return d->case_limit;
991 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
992 0 for CASE_LIMIT to indicate no limit. */
994 dict_set_case_limit (struct dictionary *d, casenumber case_limit)
998 d->case_limit = case_limit;
1001 /* Returns the case index of the next value to be added to D.
1002 This value is the number of `union value's that need to be
1003 allocated to store a case for dictionary D. */
1005 dict_get_next_value_idx (const struct dictionary *d)
1009 return d->next_value_idx;
1012 /* Returns the number of bytes needed to store a case for
1015 dict_get_case_size (const struct dictionary *d)
1019 return sizeof (union value) * dict_get_next_value_idx (d);
1022 /* Reassigns values in dictionary D so that fragmentation is
1025 dict_compact_values (struct dictionary *d)
1029 d->next_value_idx = 0;
1030 for (i = 0; i < d->var_cnt; i++)
1032 struct variable *v = d->var[i];
1033 set_var_case_index (v, d->next_value_idx);
1034 d->next_value_idx += var_get_value_cnt (v);
1039 Reassigns case indices for D, increasing each index above START by
1043 dict_pad_values (struct dictionary *d, int start, int padding)
1050 for (i = 0; i < d->var_cnt; ++i)
1052 struct variable *v = d->var[i];
1054 int index = var_get_case_index (v);
1056 if ( index >= start)
1057 set_var_case_index (v, index + padding);
1060 d->next_value_idx += padding;
1064 /* Returns the number of values occupied by the variables in
1065 dictionary D. All variables are considered if EXCLUDE_CLASSES
1066 is 0, or it may contain one or more of (1u << DC_ORDINARY),
1067 (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to exclude the
1068 corresponding type of variable.
1070 The return value may be less than the number of values in one
1071 of dictionary D's cases (as returned by
1072 dict_get_next_value_idx) even if E is 0, because there may be
1073 gaps in D's cases due to deleted variables. */
1075 dict_count_values (const struct dictionary *d, unsigned int exclude_classes)
1080 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1082 | (1u << DC_SCRATCH))) == 0);
1085 for (i = 0; i < d->var_cnt; i++)
1087 enum dict_class class = var_get_dict_class (d->var[i]);
1088 if (!(exclude_classes & (1u << class)))
1089 cnt += var_get_value_cnt (d->var[i]);
1094 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
1095 dict_get_split_cnt() to determine how many SPLIT FILE vars
1096 there are. Returns a null pointer if and only if there are no
1098 const struct variable *const *
1099 dict_get_split_vars (const struct dictionary *d)
1106 /* Returns the number of SPLIT FILE vars. */
1108 dict_get_split_cnt (const struct dictionary *d)
1112 return d->split_cnt;
1115 /* Removes variable V, which must be in D, from D's set of split
1118 dict_unset_split_var (struct dictionary *d, struct variable *v)
1122 assert (dict_contains_var (d, v));
1124 orig_count = d->split_cnt;
1125 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
1126 &v, compare_var_ptrs, NULL);
1127 if (orig_count != d->split_cnt)
1129 if (d->changed) d->changed (d, d->changed_data);
1130 /* We changed the set of split variables so invoke the
1132 if (d->callbacks && d->callbacks->split_changed)
1133 d->callbacks->split_changed (d, d->cb_data);
1137 /* Sets CNT split vars SPLIT in dictionary D. */
1139 dict_set_split_vars (struct dictionary *d,
1140 struct variable *const *split, size_t cnt)
1143 assert (cnt == 0 || split != NULL);
1148 d->split = xnrealloc (d->split, cnt, sizeof *d->split) ;
1149 memcpy (d->split, split, cnt * sizeof *d->split);
1157 if (d->changed) d->changed (d, d->changed_data);
1158 if ( d->callbacks && d->callbacks->split_changed )
1159 d->callbacks->split_changed (d, d->cb_data);
1162 /* Returns the file label for D, or a null pointer if D is
1163 unlabeled (see cmd_file_label()). */
1165 dict_get_label (const struct dictionary *d)
1172 /* Sets D's file label to LABEL, truncating it to a maximum of 60
1175 dict_set_label (struct dictionary *d, const char *label)
1180 d->label = label != NULL ? xstrndup (label, 60) : NULL;
1183 /* Returns the documents for D, or a null pointer if D has no
1184 documents. If the return value is nonnull, then the string
1185 will be an exact multiple of DOC_LINE_LENGTH bytes in length,
1186 with each segment corresponding to one line. */
1188 dict_get_documents (const struct dictionary *d)
1190 return ds_is_empty (&d->documents) ? NULL : ds_cstr (&d->documents);
1193 /* Sets the documents for D to DOCUMENTS, or removes D's
1194 documents if DOCUMENT is a null pointer. If DOCUMENTS is
1195 nonnull, then it should be an exact multiple of
1196 DOC_LINE_LENGTH bytes in length, with each segment
1197 corresponding to one line. */
1199 dict_set_documents (struct dictionary *d, const char *documents)
1203 ds_assign_cstr (&d->documents, documents != NULL ? documents : "");
1205 /* In case the caller didn't get it quite right, pad out the
1206 final line with spaces. */
1207 remainder = ds_length (&d->documents) % DOC_LINE_LENGTH;
1209 ds_put_char_multiple (&d->documents, ' ', DOC_LINE_LENGTH - remainder);
1212 /* Drops the documents from dictionary D. */
1214 dict_clear_documents (struct dictionary *d)
1216 ds_clear (&d->documents);
1219 /* Appends LINE to the documents in D. LINE will be truncated or
1220 padded on the right with spaces to make it exactly
1221 DOC_LINE_LENGTH bytes long. */
1223 dict_add_document_line (struct dictionary *d, const char *line)
1225 if (strlen (line) > DOC_LINE_LENGTH)
1227 /* Note to translators: "bytes" is correct, not characters */
1228 msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
1230 buf_copy_str_rpad (ds_put_uninit (&d->documents, DOC_LINE_LENGTH),
1231 DOC_LINE_LENGTH, line);
1234 /* Returns the number of document lines in dictionary D. */
1236 dict_get_document_line_cnt (const struct dictionary *d)
1238 return ds_length (&d->documents) / DOC_LINE_LENGTH;
1241 /* Copies document line number IDX from dictionary D into
1242 LINE, trimming off any trailing white space. */
1244 dict_get_document_line (const struct dictionary *d,
1245 size_t idx, struct string *line)
1247 assert (idx < dict_get_document_line_cnt (d));
1248 ds_assign_substring (line, ds_substr (&d->documents, idx * DOC_LINE_LENGTH,
1250 ds_rtrim (line, ss_cstr (CC_SPACES));
1253 /* Creates in D a vector named NAME that contains the CNT
1254 variables in VAR. Returns true if successful, or false if a
1255 vector named NAME already exists in D. */
1257 dict_create_vector (struct dictionary *d,
1259 struct variable **var, size_t cnt)
1263 assert (var != NULL);
1265 for (i = 0; i < cnt; i++)
1266 assert (dict_contains_var (d, var[i]));
1268 if (dict_lookup_vector (d, name) == NULL)
1270 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1271 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1278 /* Creates in D a vector named NAME that contains the CNT
1279 variables in VAR. A vector named NAME must not already exist
1282 dict_create_vector_assert (struct dictionary *d,
1284 struct variable **var, size_t cnt)
1286 assert (dict_lookup_vector (d, name) == NULL);
1287 dict_create_vector (d, name, var, cnt);
1290 /* Returns the vector in D with index IDX, which must be less
1291 than dict_get_vector_cnt (D). */
1292 const struct vector *
1293 dict_get_vector (const struct dictionary *d, size_t idx)
1296 assert (idx < d->vector_cnt);
1298 return d->vector[idx];
1301 /* Returns the number of vectors in D. */
1303 dict_get_vector_cnt (const struct dictionary *d)
1307 return d->vector_cnt;
1310 /* Looks up and returns the vector within D with the given
1312 const struct vector *
1313 dict_lookup_vector (const struct dictionary *d, const char *name)
1316 for (i = 0; i < d->vector_cnt; i++)
1317 if (!strcasecmp (vector_get_name (d->vector[i]), name))
1318 return d->vector[i];
1322 /* Deletes all vectors from D. */
1324 dict_clear_vectors (struct dictionary *d)
1328 for (i = 0; i < d->vector_cnt; i++)
1329 vector_destroy (d->vector[i]);
1336 /* Returns D's attribute set. The caller may examine or modify
1337 the attribute set, but must not destroy it. Destroying D or
1338 calling dict_set_attributes for D will also destroy D's
1341 dict_get_attributes (const struct dictionary *d)
1343 return (struct attrset *) &d->attributes;
1346 /* Replaces D's attributes set by a copy of ATTRS. */
1348 dict_set_attributes (struct dictionary *d, const struct attrset *attrs)
1350 attrset_destroy (&d->attributes);
1351 attrset_clone (&d->attributes, attrs);
1354 /* Returns true if D has at least one attribute in its attribute
1355 set, false if D's attribute set is empty. */
1357 dict_has_attributes (const struct dictionary *d)
1359 return attrset_count (&d->attributes) > 0;
1362 /* Called from variable.c to notify the dictionary that some property of
1363 the variable has changed */
1365 dict_var_changed (const struct variable *v)
1367 if ( var_has_vardict (v))
1369 const struct vardict_info *vdi = var_get_vardict (v);
1370 struct dictionary *d = vdi->dict;
1375 if (d->changed ) d->changed (d, d->changed_data);
1376 if ( d->callbacks && d->callbacks->var_changed )
1377 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
1382 /* Called from variable.c to notify the dictionary that the variable's width
1385 dict_var_resized (const struct variable *v, int delta)
1387 if ( var_has_vardict (v))
1389 const struct vardict_info *vdi = var_get_vardict (v);
1390 struct dictionary *d;
1394 dict_pad_values (d, var_get_case_index(v) + 1, delta);
1396 if (d->changed) d->changed (d, d->changed_data);
1397 if ( d->callbacks && d->callbacks->var_resized )
1398 d->callbacks->var_resized (d, var_get_dict_index (v), delta, d->cb_data);
1402 /* Called from variable.c to notify the dictionary that the variable's display width
1405 dict_var_display_width_changed (const struct variable *v)
1407 if ( var_has_vardict (v))
1409 const struct vardict_info *vdi = var_get_vardict (v);
1410 struct dictionary *d;
1414 if (d->changed) d->changed (d, d->changed_data);
1415 if ( d->callbacks && d->callbacks->var_display_width_changed )
1416 d->callbacks->var_display_width_changed (d, var_get_dict_index (v), d->cb_data);