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)
81 d->encoding = strdup (enc);
85 dict_get_encoding (const struct dictionary *d)
92 dict_set_change_callback (struct dictionary *d,
93 void (*changed) (struct dictionary *, void*),
97 d->changed_data = data;
101 /* Print a representation of dictionary D to stdout, for
102 debugging purposes. */
104 dict_dump (const struct dictionary *d)
107 for (i = 0 ; i < d->var_cnt ; ++i )
109 const struct variable *v =
111 printf ("Name: %s;\tdict_idx: %d; case_idx: %d\n",
113 var_get_dict_index (v),
114 var_get_case_index (v));
119 /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
120 the dictionary or any of the variables it contains are modified.
121 Each callback will get passed CALLBACK_DATA.
122 Any callback may be NULL, in which case it'll be ignored.
125 dict_set_callbacks (struct dictionary *dict,
126 const struct dict_callbacks *callbacks,
129 dict->callbacks = callbacks;
130 dict->cb_data = callback_data;
133 /* Shallow copy the callbacks from SRC to DEST */
135 dict_copy_callbacks (struct dictionary *dest,
136 const struct dictionary *src)
138 dest->callbacks = src->callbacks;
139 dest->cb_data = src->cb_data;
142 /* Creates and returns a new dictionary. */
146 struct dictionary *d = xzalloc (sizeof *d);
148 d->name_tab = hsh_create (8, compare_vars_by_name, hash_var_by_name,
150 attrset_init (&d->attributes);
154 /* Creates and returns a (deep) copy of an existing
157 The new dictionary's case indexes are copied from the old
158 dictionary. If the new dictionary won't be used to access
159 cases produced with the old dictionary, then the new
160 dictionary's case indexes should be compacted with
161 dict_compact_values to save space. */
163 dict_clone (const struct dictionary *s)
165 struct dictionary *d;
172 for (i = 0; i < s->var_cnt; i++)
174 const struct vardict_info *svdi;
175 struct vardict_info dvdi;
176 struct variable *sv = s->var[i];
177 struct variable *dv = dict_clone_var_assert (d, sv, var_get_name (sv));
180 for (i = 0; i < var_get_short_name_cnt (sv); i++)
181 var_set_short_name (dv, i, var_get_short_name (sv, i));
183 svdi = var_get_vardict (sv);
186 var_set_vardict (dv, &dvdi);
189 d->next_value_idx = s->next_value_idx;
191 d->split_cnt = s->split_cnt;
192 if (d->split_cnt > 0)
194 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
195 for (i = 0; i < d->split_cnt; i++)
196 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
199 if (s->weight != NULL)
200 dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight)));
202 if (s->filter != NULL)
203 dict_set_filter (d, dict_lookup_var_assert (d, var_get_name (s->filter)));
205 d->case_limit = s->case_limit;
206 dict_set_label (d, dict_get_label (s));
207 dict_set_documents (d, dict_get_documents (s));
209 d->vector_cnt = s->vector_cnt;
210 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
211 for (i = 0; i < s->vector_cnt; i++)
212 d->vector[i] = vector_clone (s->vector[i], s, d);
214 d->encoding = strdup (s->encoding);
216 dict_set_attributes (d, dict_get_attributes (s));
221 /* Clears the contents from a dictionary without destroying the
222 dictionary itself. */
224 dict_clear (struct dictionary *d)
226 /* FIXME? Should we really clear case_limit, label, documents?
227 Others are necessarily cleared by deleting all the variables.*/
230 while (d->var_cnt > 0 )
232 dict_delete_var (d, d->var[d->var_cnt - 1]);
237 d->var_cnt = d->var_cap = 0;
238 hsh_clear (d->name_tab);
239 d->next_value_idx = 0;
240 dict_set_split_vars (d, NULL, 0);
241 dict_set_weight (d, NULL);
242 dict_set_filter (d, NULL);
246 ds_destroy (&d->documents);
247 dict_clear_vectors (d);
248 attrset_clear (&d->attributes);
251 /* Destroys the aux data for every variable in D, by calling
252 var_clear_aux() for each variable. */
254 dict_clear_aux (struct dictionary *d)
260 for (i = 0; i < d->var_cnt; i++)
261 var_clear_aux (d->var[i]);
264 /* Clears a dictionary and destroys it. */
266 dict_destroy (struct dictionary *d)
270 /* In general, we don't want callbacks occuring, if the dictionary
271 is being destroyed */
272 d->callbacks = NULL ;
275 hsh_destroy (d->name_tab);
276 attrset_destroy (&d->attributes);
281 /* Returns the number of variables in D. */
283 dict_get_var_cnt (const struct dictionary *d)
290 /* Returns the variable in D with dictionary index IDX, which
291 must be between 0 and the count returned by
292 dict_get_var_cnt(), exclusive. */
294 dict_get_var (const struct dictionary *d, size_t idx)
297 assert (idx < d->var_cnt);
302 /* Sets *VARS to an array of pointers to variables in D and *CNT
303 to the number of variables in *D. All variables are returned
304 except for those, if any, in the classes indicated by EXCLUDE.
305 (There is no point in putting DC_SYSTEM in EXCLUDE as
306 dictionaries never include system variables.) */
308 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
309 size_t *cnt, enum dict_class exclude)
311 dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude);
314 /* Sets *VARS to an array of pointers to variables in D and *CNT
315 to the number of variables in *D. All variables are returned
316 except for those, if any, in the classes indicated by EXCLUDE.
317 (There is no point in putting DC_SYSTEM in EXCLUDE as
318 dictionaries never include system variables.) */
320 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
321 size_t *cnt, enum dict_class exclude)
327 assert (vars != NULL);
328 assert (cnt != NULL);
329 assert (exclude == (exclude & DC_ALL));
332 for (i = 0; i < d->var_cnt; i++)
334 enum dict_class class = var_get_dict_class (d->var[i]);
335 if (!(class & exclude))
339 *vars = xnmalloc (count, sizeof **vars);
341 for (i = 0; i < d->var_cnt; i++)
343 enum dict_class class = var_get_dict_class (d->var[i]);
344 if (!(class & exclude))
345 (*vars)[(*cnt)++] = d->var[i];
347 assert (*cnt == count);
350 static struct variable *
351 add_var (struct dictionary *d, struct variable *v)
353 /* Add dictionary info to variable. */
354 struct vardict_info vdi;
355 vdi.case_index = d->next_value_idx;
356 vdi.dict_index = d->var_cnt;
358 var_set_vardict (v, &vdi);
360 /* Update dictionary. */
361 if (d->var_cnt >= d->var_cap)
363 d->var_cap = 8 + 2 * d->var_cap;
364 d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var);
366 d->var[d->var_cnt++] = v;
367 hsh_force_insert (d->name_tab, v);
369 if ( d->changed ) d->changed (d, d->changed_data);
370 if ( d->callbacks && d->callbacks->var_added )
371 d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
373 d->next_value_idx += var_get_value_cnt (v);
378 /* Creates and returns a new variable in D with the given NAME
379 and WIDTH. Returns a null pointer if the given NAME would
380 duplicate that of an existing variable in the dictionary. */
382 dict_create_var (struct dictionary *d, const char *name, int width)
384 return (dict_lookup_var (d, name) == NULL
385 ? dict_create_var_assert (d, name, width)
389 /* Creates and returns a new variable in D with the given NAME
390 and WIDTH. Assert-fails if the given NAME would duplicate
391 that of an existing variable in the dictionary. */
393 dict_create_var_assert (struct dictionary *d, const char *name, int width)
395 assert (dict_lookup_var (d, name) == NULL);
396 return add_var (d, var_create (name, width));
399 /* Creates and returns a new variable in D with name NAME, as a
400 copy of existing variable OLD_VAR, which need not be in D or
401 in any dictionary. Returns a null pointer if the given NAME
402 would duplicate that of an existing variable in the
405 dict_clone_var (struct dictionary *d, const struct variable *old_var,
408 return (dict_lookup_var (d, name) == NULL
409 ? dict_clone_var_assert (d, old_var, name)
413 /* Creates and returns a new variable in D with name NAME, as a
414 copy of existing variable OLD_VAR, which need not be in D or
415 in any dictionary. Assert-fails if the given NAME would
416 duplicate that of an existing variable in the dictionary. */
418 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var,
421 struct variable *new_var = var_clone (old_var);
422 assert (dict_lookup_var (d, name) == NULL);
423 var_set_name (new_var, name);
424 return add_var (d, new_var);
427 /* Returns the variable named NAME in D, or a null pointer if no
428 variable has that name. */
430 dict_lookup_var (const struct dictionary *d, const char *name)
432 struct variable *target ;
433 struct variable *result ;
435 if ( ! var_is_plausible_name (name, false))
438 target = var_create (name, 0);
439 result = hsh_find (d->name_tab, target);
440 var_destroy (target);
442 if ( result && var_has_vardict (result))
444 const struct vardict_info *vdi = var_get_vardict (result);
445 assert (vdi->dict == d);
451 /* Returns the variable named NAME in D. Assert-fails if no
452 variable has that name. */
454 dict_lookup_var_assert (const struct dictionary *d, const char *name)
456 struct variable *v = dict_lookup_var (d, name);
461 /* Returns true if variable V is in dictionary D,
464 dict_contains_var (const struct dictionary *d, const struct variable *v)
466 if (var_has_vardict (v))
468 const struct vardict_info *vdi = var_get_vardict (v);
469 return (vdi->dict_index >= 0
470 && vdi->dict_index < d->var_cnt
471 && d->var[vdi->dict_index] == v);
477 /* Compares two double pointers to variables, which should point
478 to elements of a struct dictionary's `var' member array. */
480 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
482 struct variable *const *a = a_;
483 struct variable *const *b = b_;
485 return *a < *b ? -1 : *a > *b;
488 /* Sets the dict_index in V's vardict to DICT_INDEX. */
490 set_var_dict_index (struct variable *v, int dict_index)
492 struct vardict_info vdi = *var_get_vardict (v);
493 struct dictionary *d = vdi.dict;
494 vdi.dict_index = dict_index;
495 var_set_vardict (v, &vdi);
497 if ( d->changed ) d->changed (d, d->changed_data);
498 if ( d->callbacks && d->callbacks->var_changed )
499 d->callbacks->var_changed (d, dict_index, d->cb_data);
502 /* Sets the case_index in V's vardict to CASE_INDEX. */
504 set_var_case_index (struct variable *v, int case_index)
506 struct vardict_info vdi = *var_get_vardict (v);
507 vdi.case_index = case_index;
508 var_set_vardict (v, &vdi);
511 /* Re-sets the dict_index in the dictionary variables with
512 indexes from FROM to TO (exclusive). */
514 reindex_vars (struct dictionary *d, size_t from, size_t to)
518 for (i = from; i < to; i++)
519 set_var_dict_index (d->var[i], i);
522 /* Deletes variable V from dictionary D and frees V.
524 This is a very bad idea if there might be any pointers to V
525 from outside D. In general, no variable in the active file's
526 dictionary should be deleted when any transformations are
527 active on the dictionary's dataset, because those
528 transformations might reference the deleted variable. The
529 safest time to delete a variable is just after a procedure has
530 been executed, as done by DELETE VARIABLES.
532 Pointers to V within D are not a problem, because
533 dict_delete_var() knows to remove V from split variables,
534 weights, filters, etc. */
536 dict_delete_var (struct dictionary *d, struct variable *v)
538 int dict_index = var_get_dict_index (v);
539 const int case_index = var_get_case_index (v);
540 const int val_cnt = var_get_value_cnt (v);
542 assert (dict_contains_var (d, v));
544 /* Delete aux data. */
547 dict_unset_split_var (d, v);
550 dict_set_weight (d, NULL);
553 dict_set_filter (d, NULL);
555 dict_clear_vectors (d);
557 /* Remove V from var array. */
558 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
561 /* Update dict_index for each affected variable. */
562 reindex_vars (d, dict_index, d->var_cnt);
564 /* Update name hash. */
565 hsh_force_delete (d->name_tab, v);
569 var_clear_vardict (v);
572 if ( d->changed ) d->changed (d, d->changed_data);
573 if (d->callbacks && d->callbacks->var_deleted )
574 d->callbacks->var_deleted (d, dict_index, case_index, val_cnt, d->cb_data);
577 /* Deletes the COUNT variables listed in VARS from D. This is
578 unsafe; see the comment on dict_delete_var() for details. */
580 dict_delete_vars (struct dictionary *d,
581 struct variable *const *vars, size_t count)
583 /* FIXME: this can be done in O(count) time, but this algorithm
586 assert (count == 0 || vars != NULL);
589 dict_delete_var (d, *vars++);
592 /* Deletes the COUNT variables in D starting at index IDX. This
593 is unsafe; see the comment on dict_delete_var() for
596 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
598 /* FIXME: this can be done in O(count) time, but this algorithm
600 assert (idx + count <= d->var_cnt);
603 dict_delete_var (d, d->var[idx]);
606 /* Deletes scratch variables from dictionary D. */
608 dict_delete_scratch_vars (struct dictionary *d)
612 /* FIXME: this can be done in O(count) time, but this algorithm
616 for (i = 0; i < d->var_cnt; )
617 if (var_get_dict_class (d->var[i]) == DC_SCRATCH)
618 dict_delete_var (d, d->var[i]);
623 /* Moves V to 0-based position IDX in D. Other variables in D,
624 if any, retain their relative positions. Runs in time linear
625 in the distance moved. */
627 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
629 size_t old_index = var_get_dict_index (v);
631 assert (new_index < d->var_cnt);
632 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
633 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
636 /* Reorders the variables in D, placing the COUNT variables
637 listed in ORDER in that order at the beginning of D. The
638 other variables in D, if any, retain their relative
641 dict_reorder_vars (struct dictionary *d,
642 struct variable *const *order, size_t count)
644 struct variable **new_var;
648 assert (count == 0 || order != NULL);
649 assert (count <= d->var_cnt);
651 new_var = xnmalloc (d->var_cnt, sizeof *new_var);
652 memcpy (new_var, order, count * sizeof *new_var);
653 for (i = 0; i < count; i++)
655 size_t index = var_get_dict_index (order[i]);
656 assert (d->var[index] == order[i]);
657 d->var[index] = NULL;
658 set_var_dict_index (order[i], i);
660 for (i = 0; i < d->var_cnt; i++)
661 if (d->var[i] != NULL)
663 assert (count < d->var_cnt);
664 new_var[count] = d->var[i];
665 set_var_dict_index (new_var[count], count);
672 /* Changes the name of variable V in dictionary D to NEW_NAME. */
674 rename_var (struct dictionary *d, struct variable *v, const char *new_name)
676 struct vardict_info vdi;
678 assert (dict_contains_var (d, v));
680 vdi = *var_get_vardict (v);
681 var_clear_vardict (v);
682 var_set_name (v, new_name);
683 var_set_vardict (v, &vdi);
686 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
687 a variable named NEW_NAME is already in D, except that
688 NEW_NAME may be the same as V's existing name. */
690 dict_rename_var (struct dictionary *d, struct variable *v,
691 const char *new_name)
693 assert (!strcasecmp (var_get_name (v), new_name)
694 || dict_lookup_var (d, new_name) == NULL);
696 hsh_force_delete (d->name_tab, v);
697 rename_var (d, v, new_name);
698 hsh_force_insert (d->name_tab, v);
700 if (settings_get_algorithm () == ENHANCED)
701 var_clear_short_names (v);
703 if ( d->changed ) d->changed (d, d->changed_data);
704 if ( d->callbacks && d->callbacks->var_changed )
705 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
708 /* Renames COUNT variables specified in VARS to the names given
709 in NEW_NAMES within dictionary D. If the renaming would
710 result in a duplicate variable name, returns false and stores a
711 name that would be duplicated into *ERR_NAME (if ERR_NAME is
712 non-null). Otherwise, the renaming is successful, and true
715 dict_rename_vars (struct dictionary *d,
716 struct variable **vars, char **new_names, size_t count,
723 assert (count == 0 || vars != NULL);
724 assert (count == 0 || new_names != NULL);
726 /* Save the names of the variables to be renamed. */
727 pool = pool_create ();
728 old_names = pool_nalloc (pool, count, sizeof *old_names);
729 for (i = 0; i < count; i++)
730 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
732 /* Remove the variables to be renamed from the name hash,
734 for (i = 0; i < count; i++)
736 hsh_force_delete (d->name_tab, vars[i]);
737 rename_var (d, vars[i], new_names[i]);
740 /* Add the renamed variables back into the name hash,
741 checking for conflicts. */
742 for (i = 0; i < count; i++)
743 if (hsh_insert (d->name_tab, vars[i]) != NULL)
745 /* There is a name conflict.
746 Back out all the name changes that have already
747 taken place, and indicate failure. */
749 if (err_name != NULL)
750 *err_name = new_names[i];
752 for (i = 0; i < fail_idx; i++)
753 hsh_force_delete (d->name_tab, vars[i]);
755 for (i = 0; i < count; i++)
757 rename_var (d, vars[i], old_names[i]);
758 hsh_force_insert (d->name_tab, vars[i]);
765 /* Clear short names. */
766 if (settings_get_algorithm () == ENHANCED)
767 for (i = 0; i < count; i++)
768 var_clear_short_names (vars[i]);
774 /* Returns true if a variable named NAME may be inserted in DICT;
775 that is, if there is not already a variable with that name in
776 DICT and if NAME is not a reserved word. (The caller's checks
777 have already verified that NAME is otherwise acceptable as a
780 var_name_is_insertable (const struct dictionary *dict, const char *name)
782 return (dict_lookup_var (dict, name) == NULL
783 && lex_id_to_token (ss_cstr (name)) == T_ID);
787 make_hinted_name (const struct dictionary *dict, const char *hint,
788 char name[VAR_NAME_LEN + 1])
790 bool dropped = false;
793 for (cp = name; *hint && cp < name + VAR_NAME_LEN; hint++)
796 ? lex_is_id1 (*hint) && *hint != '$'
797 : lex_is_idn (*hint))
804 if (cp < name + VAR_NAME_LEN)
814 size_t len = strlen (name);
817 if (var_name_is_insertable (dict, name))
820 for (i = 0; i < ULONG_MAX; i++)
822 char suffix[INT_BUFSIZE_BOUND (i) + 1];
826 if (!str_format_26adic (i + 1, &suffix[1], sizeof suffix - 1))
829 ofs = MIN (VAR_NAME_LEN - strlen (suffix), len);
830 strcpy (&name[ofs], suffix);
832 if (var_name_is_insertable (dict, name))
841 make_numeric_name (const struct dictionary *dict, unsigned long int *num_start,
842 char name[VAR_NAME_LEN + 1])
844 unsigned long int number;
846 for (number = num_start != NULL ? MAX (*num_start, 1) : 1;
850 sprintf (name, "VAR%03lu", number);
851 if (dict_lookup_var (dict, name) == NULL)
853 if (num_start != NULL)
854 *num_start = number + 1;
859 if (num_start != NULL)
860 *num_start = ULONG_MAX;
865 /* Attempts to devise a variable name unique within DICT.
866 Returns true if successful, in which case the new variable
867 name is stored into NAME. Returns false if all names that can
868 be generated have already been taken. (Returning false is
869 quite unlikely: at least ULONG_MAX unique names can be
872 HINT, if it is non-null, is used as a suggestion that will be
873 modified for suitability as a variable name and for
876 If HINT is null or entirely unsuitable, a name in the form
877 "VAR%03d" will be generated, where the smallest unused integer
878 value is used. If NUM_START is non-null, then its value is
879 used as the minimum numeric value to check, and it is updated
880 to the next value to be checked.
883 dict_make_unique_var_name (const struct dictionary *dict, const char *hint,
884 unsigned long int *num_start,
885 char name[VAR_NAME_LEN + 1])
887 return ((hint != NULL && make_hinted_name (dict, hint, name))
888 || make_numeric_name (dict, num_start, name));
891 /* Returns the weighting variable in dictionary D, or a null
892 pointer if the dictionary is unweighted. */
894 dict_get_weight (const struct dictionary *d)
897 assert (d->weight == NULL || dict_contains_var (d, d->weight));
902 /* Returns the value of D's weighting variable in case C, except
903 that a negative weight is returned as 0. Returns 1 if the
904 dictionary is unweighted. Will warn about missing, negative,
905 or zero values if *WARN_ON_INVALID is true. The function will
906 set *WARN_ON_INVALID to false if an invalid weight is
909 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
910 bool *warn_on_invalid)
915 if (d->weight == NULL)
919 double w = case_num (c, d->weight);
920 if (w < 0.0 || var_is_num_missing (d->weight, w, MV_ANY))
922 if ( w == 0.0 && warn_on_invalid != NULL && *warn_on_invalid ) {
923 *warn_on_invalid = false;
924 msg (SW, _("At least one case in the data file had a weight value "
925 "that was user-missing, system-missing, zero, or "
926 "negative. These case(s) were ignored."));
932 /* Sets the weighting variable of D to V, or turning off
933 weighting if V is a null pointer. */
935 dict_set_weight (struct dictionary *d, struct variable *v)
938 assert (v == NULL || dict_contains_var (d, v));
939 assert (v == NULL || var_is_numeric (v));
943 if (d->changed) d->changed (d, d->changed_data);
944 if ( d->callbacks && d->callbacks->weight_changed )
945 d->callbacks->weight_changed (d,
946 v ? var_get_dict_index (v) : -1,
950 /* Returns the filter variable in dictionary D (see cmd_filter())
951 or a null pointer if the dictionary is unfiltered. */
953 dict_get_filter (const struct dictionary *d)
956 assert (d->filter == NULL || dict_contains_var (d, d->filter));
961 /* Sets V as the filter variable for dictionary D. Passing a
962 null pointer for V turn off filtering. */
964 dict_set_filter (struct dictionary *d, struct variable *v)
967 assert (v == NULL || dict_contains_var (d, v));
968 assert (v == NULL || var_is_numeric (v));
972 if (d->changed) d->changed (d, d->changed_data);
973 if ( d->callbacks && d->callbacks->filter_changed )
974 d->callbacks->filter_changed (d,
975 v ? var_get_dict_index (v) : -1,
979 /* Returns the case limit for dictionary D, or zero if the number
980 of cases is unlimited. */
982 dict_get_case_limit (const struct dictionary *d)
986 return d->case_limit;
989 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
990 0 for CASE_LIMIT to indicate no limit. */
992 dict_set_case_limit (struct dictionary *d, casenumber case_limit)
996 d->case_limit = case_limit;
999 /* Returns the case index of the next value to be added to D.
1000 This value is the number of `union value's that need to be
1001 allocated to store a case for dictionary D. */
1003 dict_get_next_value_idx (const struct dictionary *d)
1007 return d->next_value_idx;
1010 /* Returns the number of bytes needed to store a case for
1013 dict_get_case_size (const struct dictionary *d)
1017 return sizeof (union value) * dict_get_next_value_idx (d);
1020 /* Reassigns values in dictionary D so that fragmentation is
1023 dict_compact_values (struct dictionary *d)
1027 d->next_value_idx = 0;
1028 for (i = 0; i < d->var_cnt; i++)
1030 struct variable *v = d->var[i];
1031 set_var_case_index (v, d->next_value_idx);
1032 d->next_value_idx += var_get_value_cnt (v);
1037 Reassigns case indices for D, increasing each index above START by
1041 dict_pad_values (struct dictionary *d, int start, int padding)
1048 for (i = 0; i < d->var_cnt; ++i)
1050 struct variable *v = d->var[i];
1052 int index = var_get_case_index (v);
1054 if ( index >= start)
1055 set_var_case_index (v, index + padding);
1058 d->next_value_idx += padding;
1062 /* Returns the number of values occupied by the variables in
1063 dictionary D. All variables are considered if EXCLUDE_CLASSES
1064 is 0, or it may contain one or more of (1u << DC_ORDINARY),
1065 (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to exclude the
1066 corresponding type of variable.
1068 The return value may be less than the number of values in one
1069 of dictionary D's cases (as returned by
1070 dict_get_next_value_idx) even if E is 0, because there may be
1071 gaps in D's cases due to deleted variables. */
1073 dict_count_values (const struct dictionary *d, unsigned int exclude_classes)
1078 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1080 | (1u << DC_SCRATCH))) == 0);
1083 for (i = 0; i < d->var_cnt; i++)
1085 enum dict_class class = var_get_dict_class (d->var[i]);
1086 if (!(exclude_classes & (1u << class)))
1087 cnt += var_get_value_cnt (d->var[i]);
1092 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
1093 dict_get_split_cnt() to determine how many SPLIT FILE vars
1094 there are. Returns a null pointer if and only if there are no
1096 const struct variable *const *
1097 dict_get_split_vars (const struct dictionary *d)
1104 /* Returns the number of SPLIT FILE vars. */
1106 dict_get_split_cnt (const struct dictionary *d)
1110 return d->split_cnt;
1113 /* Removes variable V, which must be in D, from D's set of split
1116 dict_unset_split_var (struct dictionary *d, struct variable *v)
1120 assert (dict_contains_var (d, v));
1122 orig_count = d->split_cnt;
1123 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
1124 &v, compare_var_ptrs, NULL);
1125 if (orig_count != d->split_cnt)
1127 if (d->changed) d->changed (d, d->changed_data);
1128 /* We changed the set of split variables so invoke the
1130 if (d->callbacks && d->callbacks->split_changed)
1131 d->callbacks->split_changed (d, d->cb_data);
1135 /* Sets CNT split vars SPLIT in dictionary D. */
1137 dict_set_split_vars (struct dictionary *d,
1138 struct variable *const *split, size_t cnt)
1141 assert (cnt == 0 || split != NULL);
1146 d->split = xnrealloc (d->split, cnt, sizeof *d->split) ;
1147 memcpy (d->split, split, cnt * sizeof *d->split);
1155 if (d->changed) d->changed (d, d->changed_data);
1156 if ( d->callbacks && d->callbacks->split_changed )
1157 d->callbacks->split_changed (d, d->cb_data);
1160 /* Returns the file label for D, or a null pointer if D is
1161 unlabeled (see cmd_file_label()). */
1163 dict_get_label (const struct dictionary *d)
1170 /* Sets D's file label to LABEL, truncating it to a maximum of 60
1173 dict_set_label (struct dictionary *d, const char *label)
1178 d->label = label != NULL ? xstrndup (label, 60) : NULL;
1181 /* Returns the documents for D, or a null pointer if D has no
1182 documents. If the return value is nonnull, then the string
1183 will be an exact multiple of DOC_LINE_LENGTH bytes in length,
1184 with each segment corresponding to one line. */
1186 dict_get_documents (const struct dictionary *d)
1188 return ds_is_empty (&d->documents) ? NULL : ds_cstr (&d->documents);
1191 /* Sets the documents for D to DOCUMENTS, or removes D's
1192 documents if DOCUMENT is a null pointer. If DOCUMENTS is
1193 nonnull, then it should be an exact multiple of
1194 DOC_LINE_LENGTH bytes in length, with each segment
1195 corresponding to one line. */
1197 dict_set_documents (struct dictionary *d, const char *documents)
1201 ds_assign_cstr (&d->documents, documents != NULL ? documents : "");
1203 /* In case the caller didn't get it quite right, pad out the
1204 final line with spaces. */
1205 remainder = ds_length (&d->documents) % DOC_LINE_LENGTH;
1207 ds_put_char_multiple (&d->documents, ' ', DOC_LINE_LENGTH - remainder);
1210 /* Drops the documents from dictionary D. */
1212 dict_clear_documents (struct dictionary *d)
1214 ds_clear (&d->documents);
1217 /* Appends LINE to the documents in D. LINE will be truncated or
1218 padded on the right with spaces to make it exactly
1219 DOC_LINE_LENGTH bytes long. */
1221 dict_add_document_line (struct dictionary *d, const char *line)
1223 if (strlen (line) > DOC_LINE_LENGTH)
1225 /* Note to translators: "bytes" is correct, not characters */
1226 msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
1228 buf_copy_str_rpad (ds_put_uninit (&d->documents, DOC_LINE_LENGTH),
1229 DOC_LINE_LENGTH, line);
1232 /* Returns the number of document lines in dictionary D. */
1234 dict_get_document_line_cnt (const struct dictionary *d)
1236 return ds_length (&d->documents) / DOC_LINE_LENGTH;
1239 /* Copies document line number IDX from dictionary D into
1240 LINE, trimming off any trailing white space. */
1242 dict_get_document_line (const struct dictionary *d,
1243 size_t idx, struct string *line)
1245 assert (idx < dict_get_document_line_cnt (d));
1246 ds_assign_substring (line, ds_substr (&d->documents, idx * DOC_LINE_LENGTH,
1248 ds_rtrim (line, ss_cstr (CC_SPACES));
1251 /* Creates in D a vector named NAME that contains the CNT
1252 variables in VAR. Returns true if successful, or false if a
1253 vector named NAME already exists in D. */
1255 dict_create_vector (struct dictionary *d,
1257 struct variable **var, size_t cnt)
1261 assert (var != NULL);
1263 for (i = 0; i < cnt; i++)
1264 assert (dict_contains_var (d, var[i]));
1266 if (dict_lookup_vector (d, name) == NULL)
1268 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1269 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1276 /* Creates in D a vector named NAME that contains the CNT
1277 variables in VAR. A vector named NAME must not already exist
1280 dict_create_vector_assert (struct dictionary *d,
1282 struct variable **var, size_t cnt)
1284 assert (dict_lookup_vector (d, name) == NULL);
1285 dict_create_vector (d, name, var, cnt);
1288 /* Returns the vector in D with index IDX, which must be less
1289 than dict_get_vector_cnt (D). */
1290 const struct vector *
1291 dict_get_vector (const struct dictionary *d, size_t idx)
1294 assert (idx < d->vector_cnt);
1296 return d->vector[idx];
1299 /* Returns the number of vectors in D. */
1301 dict_get_vector_cnt (const struct dictionary *d)
1305 return d->vector_cnt;
1308 /* Looks up and returns the vector within D with the given
1310 const struct vector *
1311 dict_lookup_vector (const struct dictionary *d, const char *name)
1314 for (i = 0; i < d->vector_cnt; i++)
1315 if (!strcasecmp (vector_get_name (d->vector[i]), name))
1316 return d->vector[i];
1320 /* Deletes all vectors from D. */
1322 dict_clear_vectors (struct dictionary *d)
1326 for (i = 0; i < d->vector_cnt; i++)
1327 vector_destroy (d->vector[i]);
1334 /* Returns D's attribute set. The caller may examine or modify
1335 the attribute set, but must not destroy it. Destroying D or
1336 calling dict_set_attributes for D will also destroy D's
1339 dict_get_attributes (const struct dictionary *d)
1341 return (struct attrset *) &d->attributes;
1344 /* Replaces D's attributes set by a copy of ATTRS. */
1346 dict_set_attributes (struct dictionary *d, const struct attrset *attrs)
1348 attrset_destroy (&d->attributes);
1349 attrset_clone (&d->attributes, attrs);
1352 /* Returns true if D has at least one attribute in its attribute
1353 set, false if D's attribute set is empty. */
1355 dict_has_attributes (const struct dictionary *d)
1357 return attrset_count (&d->attributes) > 0;
1360 /* Called from variable.c to notify the dictionary that some property of
1361 the variable has changed */
1363 dict_var_changed (const struct variable *v)
1365 if ( var_has_vardict (v))
1367 const struct vardict_info *vdi = var_get_vardict (v);
1368 struct dictionary *d = vdi->dict;
1373 if (d->changed ) d->changed (d, d->changed_data);
1374 if ( d->callbacks && d->callbacks->var_changed )
1375 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
1380 /* Called from variable.c to notify the dictionary that the variable's width
1383 dict_var_resized (const struct variable *v, int delta)
1385 if ( var_has_vardict (v))
1387 const struct vardict_info *vdi = var_get_vardict (v);
1388 struct dictionary *d;
1392 dict_pad_values (d, var_get_case_index(v) + 1, delta);
1394 if (d->changed) d->changed (d, d->changed_data);
1395 if ( d->callbacks && d->callbacks->var_resized )
1396 d->callbacks->var_resized (d, var_get_dict_index (v), delta, d->cb_data);
1400 /* Called from variable.c to notify the dictionary that the variable's display width
1403 dict_var_display_width_changed (const struct variable *v)
1405 if ( var_has_vardict (v))
1407 const struct vardict_info *vdi = var_get_vardict (v);
1408 struct dictionary *d;
1412 if (d->changed) d->changed (d, d->changed_data);
1413 if ( d->callbacks && d->callbacks->var_display_width_changed )
1414 d->callbacks->var_display_width_changed (d, var_get_dict_index (v), d->cb_data);