1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006, 2007, 2009, 2010 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 caseproto *proto; /* Prototype for dictionary cases
56 struct hsh_table *name_tab; /* Variable index by name. */
57 int next_value_idx; /* Index of next `union value' to allocate. */
58 const struct variable **split; /* SPLIT FILE vars. */
59 size_t split_cnt; /* SPLIT FILE count. */
60 struct variable *weight; /* WEIGHT variable. */
61 struct variable *filter; /* FILTER variable. */
62 casenumber case_limit; /* Current case limit (N command). */
63 char *label; /* File label. */
64 struct string documents; /* Documents, as a string. */
65 struct vector **vector; /* Vectors of variables. */
66 size_t vector_cnt; /* Number of vectors. */
67 struct attrset attributes; /* Custom attributes. */
69 char *encoding; /* Character encoding of string data */
71 const struct dict_callbacks *callbacks; /* Callbacks on dictionary
73 void *cb_data ; /* Data passed to callbacks */
75 void (*changed) (struct dictionary *, void *); /* Generic change callback */
81 dict_set_encoding (struct dictionary *d, const char *enc)
86 d->encoding = xstrdup (enc);
91 dict_get_encoding (const struct dictionary *d)
98 dict_set_change_callback (struct dictionary *d,
99 void (*changed) (struct dictionary *, void*),
102 d->changed = changed;
103 d->changed_data = data;
106 /* Discards dictionary D's caseproto. (It will be regenerated
107 lazily, on demand.) */
109 invalidate_proto (struct dictionary *d)
111 caseproto_unref (d->proto);
115 /* Print a representation of dictionary D to stdout, for
116 debugging purposes. */
118 dict_dump (const struct dictionary *d)
121 for (i = 0 ; i < d->var_cnt ; ++i )
123 const struct variable *v =
125 printf ("Name: %s;\tdict_idx: %zu; case_idx: %zu\n",
127 var_get_dict_index (v),
128 var_get_case_index (v));
133 /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
134 the dictionary or any of the variables it contains are modified.
135 Each callback will get passed CALLBACK_DATA.
136 Any callback may be NULL, in which case it'll be ignored.
139 dict_set_callbacks (struct dictionary *dict,
140 const struct dict_callbacks *callbacks,
143 dict->callbacks = callbacks;
144 dict->cb_data = callback_data;
147 /* Shallow copy the callbacks from SRC to DEST */
149 dict_copy_callbacks (struct dictionary *dest,
150 const struct dictionary *src)
152 dest->callbacks = src->callbacks;
153 dest->cb_data = src->cb_data;
156 /* Creates and returns a new dictionary. */
160 struct dictionary *d = xzalloc (sizeof *d);
162 d->name_tab = hsh_create (8, compare_vars_by_name, hash_var_by_name,
164 attrset_init (&d->attributes);
168 /* Creates and returns a (deep) copy of an existing
171 The new dictionary's case indexes are copied from the old
172 dictionary. If the new dictionary won't be used to access
173 cases produced with the old dictionary, then the new
174 dictionary's case indexes should be compacted with
175 dict_compact_values to save space. */
177 dict_clone (const struct dictionary *s)
179 struct dictionary *d;
186 for (i = 0; i < s->var_cnt; i++)
188 struct variable *sv = s->var[i];
189 struct variable *dv = dict_clone_var_assert (d, sv);
192 for (i = 0; i < var_get_short_name_cnt (sv); i++)
193 var_set_short_name (dv, i, var_get_short_name (sv, i));
195 var_get_vardict (dv)->case_index = var_get_vardict (sv)->case_index;
198 d->next_value_idx = s->next_value_idx;
200 d->split_cnt = s->split_cnt;
201 if (d->split_cnt > 0)
203 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
204 for (i = 0; i < d->split_cnt; i++)
205 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
208 if (s->weight != NULL)
209 dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight)));
211 if (s->filter != NULL)
212 dict_set_filter (d, dict_lookup_var_assert (d, var_get_name (s->filter)));
214 d->case_limit = s->case_limit;
215 dict_set_label (d, dict_get_label (s));
216 dict_set_documents (d, dict_get_documents (s));
218 d->vector_cnt = s->vector_cnt;
219 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
220 for (i = 0; i < s->vector_cnt; i++)
221 d->vector[i] = vector_clone (s->vector[i], s, d);
224 d->encoding = xstrdup (s->encoding);
226 dict_set_attributes (d, dict_get_attributes (s));
231 /* Clears the contents from a dictionary without destroying the
232 dictionary itself. */
234 dict_clear (struct dictionary *d)
236 /* FIXME? Should we really clear case_limit, label, documents?
237 Others are necessarily cleared by deleting all the variables.*/
240 while (d->var_cnt > 0 )
242 dict_delete_var (d, d->var[d->var_cnt - 1]);
247 d->var_cnt = d->var_cap = 0;
248 invalidate_proto (d);
249 hsh_clear (d->name_tab);
250 d->next_value_idx = 0;
251 dict_set_split_vars (d, NULL, 0);
252 dict_set_weight (d, NULL);
253 dict_set_filter (d, NULL);
257 ds_destroy (&d->documents);
258 dict_clear_vectors (d);
259 attrset_clear (&d->attributes);
262 /* Destroys the aux data for every variable in D, by calling
263 var_clear_aux() for each variable. */
265 dict_clear_aux (struct dictionary *d)
271 for (i = 0; i < d->var_cnt; i++)
272 var_clear_aux (d->var[i]);
275 /* Clears a dictionary and destroys it. */
277 dict_destroy (struct dictionary *d)
281 /* In general, we don't want callbacks occuring, if the dictionary
282 is being destroyed */
283 d->callbacks = NULL ;
286 hsh_destroy (d->name_tab);
287 attrset_destroy (&d->attributes);
292 /* Returns the number of variables in D. */
294 dict_get_var_cnt (const struct dictionary *d)
301 /* Returns the variable in D with dictionary index IDX, which
302 must be between 0 and the count returned by
303 dict_get_var_cnt(), exclusive. */
305 dict_get_var (const struct dictionary *d, size_t idx)
308 assert (idx < d->var_cnt);
313 /* Sets *VARS to an array of pointers to variables in D and *CNT
314 to the number of variables in *D. All variables are returned
315 except for those, if any, in the classes indicated by EXCLUDE.
316 (There is no point in putting DC_SYSTEM in EXCLUDE as
317 dictionaries never include system variables.) */
319 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
320 size_t *cnt, enum dict_class exclude)
322 dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude);
325 /* Sets *VARS to an array of pointers to variables in D and *CNT
326 to the number of variables in *D. All variables are returned
327 except for those, if any, in the classes indicated by EXCLUDE.
328 (There is no point in putting DC_SYSTEM in EXCLUDE as
329 dictionaries never include system variables.) */
331 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
332 size_t *cnt, enum dict_class exclude)
338 assert (vars != NULL);
339 assert (cnt != NULL);
340 assert (exclude == (exclude & DC_ALL));
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))
350 *vars = xnmalloc (count, sizeof **vars);
352 for (i = 0; i < d->var_cnt; i++)
354 enum dict_class class = var_get_dict_class (d->var[i]);
355 if (!(class & exclude))
356 (*vars)[(*cnt)++] = d->var[i];
358 assert (*cnt == count);
361 static struct variable *
362 add_var (struct dictionary *d, struct variable *v)
364 /* Add dictionary info to variable. */
365 struct vardict_info *vdi;
367 vdi = xmalloc (sizeof *vdi);
368 vdi->case_index = d->next_value_idx;
369 vdi->dict_index = d->var_cnt;
371 var_set_vardict (v, vdi);
373 /* Update dictionary. */
374 if (d->var_cnt >= d->var_cap)
376 d->var_cap = 8 + 2 * d->var_cap;
377 d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var);
379 d->var[d->var_cnt++] = v;
380 hsh_force_insert (d->name_tab, v);
382 if ( d->changed ) d->changed (d, d->changed_data);
383 if ( d->callbacks && d->callbacks->var_added )
384 d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
387 invalidate_proto (d);
392 /* Creates and returns a new variable in D with the given NAME
393 and WIDTH. Returns a null pointer if the given NAME would
394 duplicate that of an existing variable in the dictionary. */
396 dict_create_var (struct dictionary *d, const char *name, int width)
398 return (dict_lookup_var (d, name) == NULL
399 ? dict_create_var_assert (d, name, width)
403 /* Creates and returns a new variable in D with the given NAME
404 and WIDTH. Assert-fails if the given NAME would duplicate
405 that of an existing variable in the dictionary. */
407 dict_create_var_assert (struct dictionary *d, const char *name, int width)
409 assert (dict_lookup_var (d, name) == NULL);
410 return add_var (d, var_create (name, width));
413 /* Creates and returns a new variable in D, as a copy of existing variable
414 OLD_VAR, which need not be in D or in any dictionary. Returns a null
415 pointer if OLD_VAR's name would duplicate that of an existing variable in
418 dict_clone_var (struct dictionary *d, const struct variable *old_var)
420 return dict_clone_var_as (d, old_var, var_get_name (old_var));
423 /* Creates and returns a new variable in D, as a copy of existing variable
424 OLD_VAR, which need not be in D or in any dictionary. Assert-fails if
425 OLD_VAR's name would duplicate that of an existing variable in the
428 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var)
430 return dict_clone_var_as_assert (d, old_var, var_get_name (old_var));
433 /* Creates and returns a new variable in D with name NAME, as a copy of
434 existing variable OLD_VAR, which need not be in D or in any dictionary.
435 Returns a null pointer if the given NAME would duplicate that of an existing
436 variable in the dictionary. */
438 dict_clone_var_as (struct dictionary *d, const struct variable *old_var,
441 return (dict_lookup_var (d, name) == NULL
442 ? dict_clone_var_as_assert (d, old_var, name)
446 /* Creates and returns a new variable in D with name NAME, as a copy of
447 existing variable OLD_VAR, which need not be in D or in any dictionary.
448 Assert-fails if the given NAME would duplicate that of an existing variable
449 in the dictionary. */
451 dict_clone_var_as_assert (struct dictionary *d, const struct variable *old_var,
454 struct variable *new_var = var_clone (old_var);
455 assert (dict_lookup_var (d, name) == NULL);
456 var_set_name (new_var, name);
457 return add_var (d, new_var);
460 /* Returns the variable named NAME in D, or a null pointer if no
461 variable has that name. */
463 dict_lookup_var (const struct dictionary *d, const char *name)
465 struct variable *target ;
466 struct variable *result ;
468 if ( ! var_is_plausible_name (name, false))
471 target = var_create (name, 0);
472 result = hsh_find (d->name_tab, target);
473 var_destroy (target);
475 if ( result && var_has_vardict (result))
477 const struct vardict_info *vdi = var_get_vardict (result);
478 assert (vdi->dict == d);
484 /* Returns the variable named NAME in D. Assert-fails if no
485 variable has that name. */
487 dict_lookup_var_assert (const struct dictionary *d, const char *name)
489 struct variable *v = dict_lookup_var (d, name);
494 /* Returns true if variable V is in dictionary D,
497 dict_contains_var (const struct dictionary *d, const struct variable *v)
499 if (var_has_vardict (v))
501 const struct vardict_info *vdi = var_get_vardict (v);
502 return (vdi->dict_index >= 0
503 && vdi->dict_index < d->var_cnt
504 && d->var[vdi->dict_index] == v);
510 /* Compares two double pointers to variables, which should point
511 to elements of a struct dictionary's `var' member array. */
513 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
515 struct variable *const *a = a_;
516 struct variable *const *b = b_;
518 return *a < *b ? -1 : *a > *b;
521 /* Sets the dict_index in V's vardict to DICT_INDEX. */
523 set_var_dict_index (struct dictionary *d, struct variable *v, int dict_index)
525 var_get_vardict (v)->dict_index = dict_index;
527 if ( d->changed ) d->changed (d, d->changed_data);
528 if ( d->callbacks && d->callbacks->var_changed )
529 d->callbacks->var_changed (d, dict_index, d->cb_data);
532 /* Sets the case_index in V's vardict to CASE_INDEX. */
534 set_var_case_index (struct variable *v, int case_index)
536 var_get_vardict (v)->case_index = case_index;
539 /* Re-sets the dict_index in the dictionary variables with
540 indexes from FROM to TO (exclusive). */
542 reindex_vars (struct dictionary *d, size_t from, size_t to)
546 for (i = from; i < to; i++)
547 set_var_dict_index (d, d->var[i], i);
550 /* Deletes variable V from dictionary D and frees V.
552 This is a very bad idea if there might be any pointers to V
553 from outside D. In general, no variable in the active file's
554 dictionary should be deleted when any transformations are
555 active on the dictionary's dataset, because those
556 transformations might reference the deleted variable. The
557 safest time to delete a variable is just after a procedure has
558 been executed, as done by DELETE VARIABLES.
560 Pointers to V within D are not a problem, because
561 dict_delete_var() knows to remove V from split variables,
562 weights, filters, etc. */
564 dict_delete_var (struct dictionary *d, struct variable *v)
566 int dict_index = var_get_dict_index (v);
567 const int case_index = var_get_case_index (v);
568 const int width = var_get_width (v);
570 assert (dict_contains_var (d, v));
572 /* Delete aux data. */
575 dict_unset_split_var (d, v);
578 dict_set_weight (d, NULL);
581 dict_set_filter (d, NULL);
583 dict_clear_vectors (d);
585 /* Remove V from var array. */
586 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
589 /* Update dict_index for each affected variable. */
590 reindex_vars (d, dict_index, d->var_cnt);
592 /* Update name hash. */
593 hsh_force_delete (d->name_tab, v);
597 free (var_get_vardict (v));
598 var_clear_vardict (v);
601 if ( d->changed ) d->changed (d, d->changed_data);
603 invalidate_proto (d);
604 if (d->callbacks && d->callbacks->var_deleted )
605 d->callbacks->var_deleted (d, dict_index, case_index, width, d->cb_data);
608 /* Deletes the COUNT variables listed in VARS from D. This is
609 unsafe; see the comment on dict_delete_var() for details. */
611 dict_delete_vars (struct dictionary *d,
612 struct variable *const *vars, size_t count)
614 /* FIXME: this can be done in O(count) time, but this algorithm
617 assert (count == 0 || vars != NULL);
620 dict_delete_var (d, *vars++);
623 /* Deletes the COUNT variables in D starting at index IDX. This
624 is unsafe; see the comment on dict_delete_var() for
627 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
629 /* FIXME: this can be done in O(count) time, but this algorithm
631 assert (idx + count <= d->var_cnt);
634 dict_delete_var (d, d->var[idx]);
637 /* Deletes scratch variables from dictionary D. */
639 dict_delete_scratch_vars (struct dictionary *d)
643 /* FIXME: this can be done in O(count) time, but this algorithm
647 for (i = 0; i < d->var_cnt; )
648 if (var_get_dict_class (d->var[i]) == DC_SCRATCH)
649 dict_delete_var (d, d->var[i]);
654 /* Moves V to 0-based position IDX in D. Other variables in D,
655 if any, retain their relative positions. Runs in time linear
656 in the distance moved. */
658 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
660 size_t old_index = var_get_dict_index (v);
662 assert (new_index < d->var_cnt);
663 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
664 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
667 /* Reorders the variables in D, placing the COUNT variables
668 listed in ORDER in that order at the beginning of D. The
669 other variables in D, if any, retain their relative
672 dict_reorder_vars (struct dictionary *d,
673 struct variable *const *order, size_t count)
675 struct variable **new_var;
679 assert (count == 0 || order != NULL);
680 assert (count <= d->var_cnt);
682 new_var = xnmalloc (d->var_cap, sizeof *new_var);
683 memcpy (new_var, order, count * sizeof *new_var);
684 for (i = 0; i < count; i++)
686 size_t index = var_get_dict_index (order[i]);
687 assert (d->var[index] == order[i]);
688 d->var[index] = NULL;
689 set_var_dict_index (d, order[i], i);
691 for (i = 0; i < d->var_cnt; i++)
692 if (d->var[i] != NULL)
694 assert (count < d->var_cnt);
695 new_var[count] = d->var[i];
696 set_var_dict_index (d, new_var[count], count);
703 /* Changes the name of variable V in dictionary D to NEW_NAME. */
705 rename_var (struct dictionary *d, struct variable *v, const char *new_name)
707 struct vardict_info *vardict;
709 assert (dict_contains_var (d, v));
711 vardict = var_get_vardict (v);
712 var_clear_vardict (v);
713 var_set_name (v, new_name);
714 var_set_vardict (v, vardict);
717 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
718 a variable named NEW_NAME is already in D, except that
719 NEW_NAME may be the same as V's existing name. */
721 dict_rename_var (struct dictionary *d, struct variable *v,
722 const char *new_name)
724 assert (!strcasecmp (var_get_name (v), new_name)
725 || dict_lookup_var (d, new_name) == NULL);
727 hsh_force_delete (d->name_tab, v);
728 rename_var (d, v, new_name);
729 hsh_force_insert (d->name_tab, v);
731 if (settings_get_algorithm () == ENHANCED)
732 var_clear_short_names (v);
734 if ( d->changed ) d->changed (d, d->changed_data);
735 if ( d->callbacks && d->callbacks->var_changed )
736 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
739 /* Renames COUNT variables specified in VARS to the names given
740 in NEW_NAMES within dictionary D. If the renaming would
741 result in a duplicate variable name, returns false and stores a
742 name that would be duplicated into *ERR_NAME (if ERR_NAME is
743 non-null). Otherwise, the renaming is successful, and true
746 dict_rename_vars (struct dictionary *d,
747 struct variable **vars, char **new_names, size_t count,
754 assert (count == 0 || vars != NULL);
755 assert (count == 0 || new_names != NULL);
757 /* Save the names of the variables to be renamed. */
758 pool = pool_create ();
759 old_names = pool_nalloc (pool, count, sizeof *old_names);
760 for (i = 0; i < count; i++)
761 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
763 /* Remove the variables to be renamed from the name hash,
765 for (i = 0; i < count; i++)
767 hsh_force_delete (d->name_tab, vars[i]);
768 rename_var (d, vars[i], new_names[i]);
771 /* Add the renamed variables back into the name hash,
772 checking for conflicts. */
773 for (i = 0; i < count; i++)
774 if (hsh_insert (d->name_tab, vars[i]) != NULL)
776 /* There is a name conflict.
777 Back out all the name changes that have already
778 taken place, and indicate failure. */
780 if (err_name != NULL)
781 *err_name = new_names[i];
783 for (i = 0; i < fail_idx; i++)
784 hsh_force_delete (d->name_tab, vars[i]);
786 for (i = 0; i < count; i++)
788 rename_var (d, vars[i], old_names[i]);
789 hsh_force_insert (d->name_tab, vars[i]);
796 /* Clear short names. */
797 if (settings_get_algorithm () == ENHANCED)
798 for (i = 0; i < count; i++)
799 var_clear_short_names (vars[i]);
805 /* Returns true if a variable named NAME may be inserted in DICT;
806 that is, if there is not already a variable with that name in
807 DICT and if NAME is not a reserved word. (The caller's checks
808 have already verified that NAME is otherwise acceptable as a
811 var_name_is_insertable (const struct dictionary *dict, const char *name)
813 return (dict_lookup_var (dict, name) == NULL
814 && lex_id_to_token (ss_cstr (name)) == T_ID);
818 make_hinted_name (const struct dictionary *dict, const char *hint,
819 char name[VAR_NAME_LEN + 1])
821 bool dropped = false;
824 for (cp = name; *hint && cp < name + VAR_NAME_LEN; hint++)
827 ? lex_is_id1 (*hint) && *hint != '$'
828 : lex_is_idn (*hint))
835 if (cp < name + VAR_NAME_LEN)
845 size_t len = strlen (name);
848 if (var_name_is_insertable (dict, name))
851 for (i = 0; i < ULONG_MAX; i++)
853 char suffix[INT_BUFSIZE_BOUND (i) + 1];
857 if (!str_format_26adic (i + 1, &suffix[1], sizeof suffix - 1))
860 ofs = MIN (VAR_NAME_LEN - strlen (suffix), len);
861 strcpy (&name[ofs], suffix);
863 if (var_name_is_insertable (dict, name))
872 make_numeric_name (const struct dictionary *dict, unsigned long int *num_start,
873 char name[VAR_NAME_LEN + 1])
875 unsigned long int number;
877 for (number = num_start != NULL ? MAX (*num_start, 1) : 1;
881 sprintf (name, "VAR%03lu", number);
882 if (dict_lookup_var (dict, name) == NULL)
884 if (num_start != NULL)
885 *num_start = number + 1;
890 if (num_start != NULL)
891 *num_start = ULONG_MAX;
896 /* Attempts to devise a variable name unique within DICT.
897 Returns true if successful, in which case the new variable
898 name is stored into NAME. Returns false if all names that can
899 be generated have already been taken. (Returning false is
900 quite unlikely: at least ULONG_MAX unique names can be
903 HINT, if it is non-null, is used as a suggestion that will be
904 modified for suitability as a variable name and for
907 If HINT is null or entirely unsuitable, a name in the form
908 "VAR%03d" will be generated, where the smallest unused integer
909 value is used. If NUM_START is non-null, then its value is
910 used as the minimum numeric value to check, and it is updated
911 to the next value to be checked.
914 dict_make_unique_var_name (const struct dictionary *dict, const char *hint,
915 unsigned long int *num_start,
916 char name[VAR_NAME_LEN + 1])
918 return ((hint != NULL && make_hinted_name (dict, hint, name))
919 || make_numeric_name (dict, num_start, name));
922 /* Returns the weighting variable in dictionary D, or a null
923 pointer if the dictionary is unweighted. */
925 dict_get_weight (const struct dictionary *d)
928 assert (d->weight == NULL || dict_contains_var (d, d->weight));
933 /* Returns the value of D's weighting variable in case C, except
934 that a negative weight is returned as 0. Returns 1 if the
935 dictionary is unweighted. Will warn about missing, negative,
936 or zero values if *WARN_ON_INVALID is true. The function will
937 set *WARN_ON_INVALID to false if an invalid weight is
940 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
941 bool *warn_on_invalid)
946 if (d->weight == NULL)
950 double w = case_num (c, d->weight);
951 if (w < 0.0 || var_is_num_missing (d->weight, w, MV_ANY))
953 if ( w == 0.0 && warn_on_invalid != NULL && *warn_on_invalid ) {
954 *warn_on_invalid = false;
955 msg (SW, _("At least one case in the data file had a weight value "
956 "that was user-missing, system-missing, zero, or "
957 "negative. These case(s) were ignored."));
963 /* Sets the weighting variable of D to V, or turning off
964 weighting if V is a null pointer. */
966 dict_set_weight (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->weight_changed )
976 d->callbacks->weight_changed (d,
977 v ? var_get_dict_index (v) : -1,
981 /* Returns the filter variable in dictionary D (see cmd_filter())
982 or a null pointer if the dictionary is unfiltered. */
984 dict_get_filter (const struct dictionary *d)
987 assert (d->filter == NULL || dict_contains_var (d, d->filter));
992 /* Sets V as the filter variable for dictionary D. Passing a
993 null pointer for V turn off filtering. */
995 dict_set_filter (struct dictionary *d, struct variable *v)
998 assert (v == NULL || dict_contains_var (d, v));
999 assert (v == NULL || var_is_numeric (v));
1003 if (d->changed) d->changed (d, d->changed_data);
1004 if ( d->callbacks && d->callbacks->filter_changed )
1005 d->callbacks->filter_changed (d,
1006 v ? var_get_dict_index (v) : -1,
1010 /* Returns the case limit for dictionary D, or zero if the number
1011 of cases is unlimited. */
1013 dict_get_case_limit (const struct dictionary *d)
1017 return d->case_limit;
1020 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
1021 0 for CASE_LIMIT to indicate no limit. */
1023 dict_set_case_limit (struct dictionary *d, casenumber case_limit)
1027 d->case_limit = case_limit;
1030 /* Returns the prototype used for cases created by dictionary D. */
1031 const struct caseproto *
1032 dict_get_proto (const struct dictionary *d_)
1034 struct dictionary *d = CONST_CAST (struct dictionary *, d_);
1035 if (d->proto == NULL)
1039 d->proto = caseproto_create ();
1040 d->proto = caseproto_reserve (d->proto, d->var_cnt);
1041 for (i = 0; i < d->var_cnt; i++)
1042 d->proto = caseproto_set_width (d->proto,
1043 var_get_case_index (d->var[i]),
1044 var_get_width (d->var[i]));
1049 /* Returns the case index of the next value to be added to D.
1050 This value is the number of `union value's that need to be
1051 allocated to store a case for dictionary D. */
1053 dict_get_next_value_idx (const struct dictionary *d)
1057 return d->next_value_idx;
1060 /* Returns the number of bytes needed to store a case for
1063 dict_get_case_size (const struct dictionary *d)
1067 return sizeof (union value) * dict_get_next_value_idx (d);
1070 /* Reassigns values in dictionary D so that fragmentation is
1073 dict_compact_values (struct dictionary *d)
1077 d->next_value_idx = 0;
1078 for (i = 0; i < d->var_cnt; i++)
1080 struct variable *v = d->var[i];
1081 set_var_case_index (v, d->next_value_idx++);
1083 invalidate_proto (d);
1086 /* Returns the number of values occupied by the variables in
1087 dictionary D. All variables are considered if EXCLUDE_CLASSES
1088 is 0, or it may contain one or more of (1u << DC_ORDINARY),
1089 (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to exclude the
1090 corresponding type of variable.
1092 The return value may be less than the number of values in one
1093 of dictionary D's cases (as returned by
1094 dict_get_next_value_idx) even if E is 0, because there may be
1095 gaps in D's cases due to deleted variables. */
1097 dict_count_values (const struct dictionary *d, unsigned int exclude_classes)
1102 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1104 | (1u << DC_SCRATCH))) == 0);
1107 for (i = 0; i < d->var_cnt; i++)
1109 enum dict_class class = var_get_dict_class (d->var[i]);
1110 if (!(exclude_classes & (1u << class)))
1116 /* Returns the case prototype that would result after deleting
1117 all variables from D that are not in one of the
1118 EXCLUDE_CLASSES and compacting the dictionary with
1121 The caller must unref the returned caseproto when it is no
1124 dict_get_compacted_proto (const struct dictionary *d,
1125 unsigned int exclude_classes)
1127 struct caseproto *proto;
1130 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1132 | (1u << DC_SCRATCH))) == 0);
1134 proto = caseproto_create ();
1135 for (i = 0; i < d->var_cnt; i++)
1137 struct variable *v = d->var[i];
1138 if (!(exclude_classes & (1u << var_get_dict_class (v))))
1139 proto = caseproto_add_width (proto, var_get_width (v));
1144 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
1145 dict_get_split_cnt() to determine how many SPLIT FILE vars
1146 there are. Returns a null pointer if and only if there are no
1148 const struct variable *const *
1149 dict_get_split_vars (const struct dictionary *d)
1156 /* Returns the number of SPLIT FILE vars. */
1158 dict_get_split_cnt (const struct dictionary *d)
1162 return d->split_cnt;
1165 /* Removes variable V, which must be in D, from D's set of split
1168 dict_unset_split_var (struct dictionary *d, struct variable *v)
1172 assert (dict_contains_var (d, v));
1174 orig_count = d->split_cnt;
1175 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
1176 &v, compare_var_ptrs, NULL);
1177 if (orig_count != d->split_cnt)
1179 if (d->changed) d->changed (d, d->changed_data);
1180 /* We changed the set of split variables so invoke the
1182 if (d->callbacks && d->callbacks->split_changed)
1183 d->callbacks->split_changed (d, d->cb_data);
1187 /* Sets CNT split vars SPLIT in dictionary D. */
1189 dict_set_split_vars (struct dictionary *d,
1190 struct variable *const *split, size_t cnt)
1193 assert (cnt == 0 || split != NULL);
1198 d->split = xnrealloc (d->split, cnt, sizeof *d->split) ;
1199 memcpy (d->split, split, cnt * sizeof *d->split);
1207 if (d->changed) d->changed (d, d->changed_data);
1208 if ( d->callbacks && d->callbacks->split_changed )
1209 d->callbacks->split_changed (d, d->cb_data);
1212 /* Returns the file label for D, or a null pointer if D is
1213 unlabeled (see cmd_file_label()). */
1215 dict_get_label (const struct dictionary *d)
1222 /* Sets D's file label to LABEL, truncating it to a maximum of 60
1225 dict_set_label (struct dictionary *d, const char *label)
1230 d->label = label != NULL ? xstrndup (label, 60) : NULL;
1233 /* Returns the documents for D, or a null pointer if D has no
1234 documents. If the return value is nonnull, then the string
1235 will be an exact multiple of DOC_LINE_LENGTH bytes in length,
1236 with each segment corresponding to one line. */
1238 dict_get_documents (const struct dictionary *d)
1240 return ds_is_empty (&d->documents) ? NULL : ds_cstr (&d->documents);
1243 /* Sets the documents for D to DOCUMENTS, or removes D's
1244 documents if DOCUMENT is a null pointer. If DOCUMENTS is
1245 nonnull, then it should be an exact multiple of
1246 DOC_LINE_LENGTH bytes in length, with each segment
1247 corresponding to one line. */
1249 dict_set_documents (struct dictionary *d, const char *documents)
1253 ds_assign_cstr (&d->documents, documents != NULL ? documents : "");
1255 /* In case the caller didn't get it quite right, pad out the
1256 final line with spaces. */
1257 remainder = ds_length (&d->documents) % DOC_LINE_LENGTH;
1259 ds_put_char_multiple (&d->documents, ' ', DOC_LINE_LENGTH - remainder);
1262 /* Drops the documents from dictionary D. */
1264 dict_clear_documents (struct dictionary *d)
1266 ds_clear (&d->documents);
1269 /* Appends LINE to the documents in D. LINE will be truncated or
1270 padded on the right with spaces to make it exactly
1271 DOC_LINE_LENGTH bytes long. */
1273 dict_add_document_line (struct dictionary *d, const char *line)
1275 if (strlen (line) > DOC_LINE_LENGTH)
1277 /* Note to translators: "bytes" is correct, not characters */
1278 msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
1280 buf_copy_str_rpad (ds_put_uninit (&d->documents, DOC_LINE_LENGTH),
1281 DOC_LINE_LENGTH, line, ' ');
1284 /* Returns the number of document lines in dictionary D. */
1286 dict_get_document_line_cnt (const struct dictionary *d)
1288 return ds_length (&d->documents) / DOC_LINE_LENGTH;
1291 /* Copies document line number IDX from dictionary D into
1292 LINE, trimming off any trailing white space. */
1294 dict_get_document_line (const struct dictionary *d,
1295 size_t idx, struct string *line)
1297 assert (idx < dict_get_document_line_cnt (d));
1298 ds_assign_substring (line, ds_substr (&d->documents, idx * DOC_LINE_LENGTH,
1300 ds_rtrim (line, ss_cstr (CC_SPACES));
1303 /* Creates in D a vector named NAME that contains the CNT
1304 variables in VAR. Returns true if successful, or false if a
1305 vector named NAME already exists in D. */
1307 dict_create_vector (struct dictionary *d,
1309 struct variable **var, size_t cnt)
1313 assert (var != NULL);
1315 for (i = 0; i < cnt; i++)
1316 assert (dict_contains_var (d, var[i]));
1318 if (dict_lookup_vector (d, name) == NULL)
1320 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1321 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1328 /* Creates in D a vector named NAME that contains the CNT
1329 variables in VAR. A vector named NAME must not already exist
1332 dict_create_vector_assert (struct dictionary *d,
1334 struct variable **var, size_t cnt)
1336 assert (dict_lookup_vector (d, name) == NULL);
1337 dict_create_vector (d, name, var, cnt);
1340 /* Returns the vector in D with index IDX, which must be less
1341 than dict_get_vector_cnt (D). */
1342 const struct vector *
1343 dict_get_vector (const struct dictionary *d, size_t idx)
1346 assert (idx < d->vector_cnt);
1348 return d->vector[idx];
1351 /* Returns the number of vectors in D. */
1353 dict_get_vector_cnt (const struct dictionary *d)
1357 return d->vector_cnt;
1360 /* Looks up and returns the vector within D with the given
1362 const struct vector *
1363 dict_lookup_vector (const struct dictionary *d, const char *name)
1366 for (i = 0; i < d->vector_cnt; i++)
1367 if (!strcasecmp (vector_get_name (d->vector[i]), name))
1368 return d->vector[i];
1372 /* Deletes all vectors from D. */
1374 dict_clear_vectors (struct dictionary *d)
1378 for (i = 0; i < d->vector_cnt; i++)
1379 vector_destroy (d->vector[i]);
1386 /* Returns D's attribute set. The caller may examine or modify
1387 the attribute set, but must not destroy it. Destroying D or
1388 calling dict_set_attributes for D will also destroy D's
1391 dict_get_attributes (const struct dictionary *d)
1393 return CONST_CAST (struct attrset *, &d->attributes);
1396 /* Replaces D's attributes set by a copy of ATTRS. */
1398 dict_set_attributes (struct dictionary *d, const struct attrset *attrs)
1400 attrset_destroy (&d->attributes);
1401 attrset_clone (&d->attributes, attrs);
1404 /* Returns true if D has at least one attribute in its attribute
1405 set, false if D's attribute set is empty. */
1407 dict_has_attributes (const struct dictionary *d)
1409 return attrset_count (&d->attributes) > 0;
1412 /* Called from variable.c to notify the dictionary that some property of
1413 the variable has changed */
1415 dict_var_changed (const struct variable *v)
1417 if ( var_has_vardict (v))
1419 const struct vardict_info *vdi = var_get_vardict (v);
1420 struct dictionary *d = vdi->dict;
1425 if (d->changed ) d->changed (d, d->changed_data);
1426 if ( d->callbacks && d->callbacks->var_changed )
1427 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
1432 /* Called from variable.c to notify the dictionary that the variable's width
1435 dict_var_resized (const struct variable *v, int old_width)
1437 if ( var_has_vardict (v))
1439 const struct vardict_info *vdi = var_get_vardict (v);
1440 struct dictionary *d;
1444 if (d->changed) d->changed (d, d->changed_data);
1446 invalidate_proto (d);
1447 if ( d->callbacks && d->callbacks->var_resized )
1448 d->callbacks->var_resized (d, var_get_dict_index (v), old_width,
1453 /* Called from variable.c to notify the dictionary that the variable's display width
1456 dict_var_display_width_changed (const struct variable *v)
1458 if ( var_has_vardict (v))
1460 const struct vardict_info *vdi = var_get_vardict (v);
1461 struct dictionary *d;
1465 if (d->changed) d->changed (d, d->changed_data);
1466 if ( d->callbacks && d->callbacks->var_display_width_changed )
1467 d->callbacks->var_display_width_changed (d, var_get_dict_index (v), d->cb_data);
1471 /* Dictionary used to contain "internal variables". */
1472 static struct dictionary *internal_dict;
1474 /* Create a variable of the specified WIDTH to be used for internal
1475 calculations only. The variable is assigned case index CASE_IDX. */
1477 dict_create_internal_var (int case_idx, int width)
1479 if (internal_dict == NULL)
1480 internal_dict = dict_create ();
1484 static int counter = INT_MAX / 2;
1485 struct variable *var;
1488 if (++counter == INT_MAX)
1489 counter = INT_MAX / 2;
1491 sprintf (name, "$internal%d", counter);
1492 var = dict_create_var (internal_dict, name, width);
1495 set_var_case_index (var, case_idx);
1501 /* Destroys VAR, which must have been created with
1502 dict_create_internal_var(). */
1504 dict_destroy_internal_var (struct variable *var)
1508 dict_delete_var (internal_dict, var);
1510 /* Destroy internal_dict if it has no variables left, just so that
1511 valgrind --leak-check --show-reachable won't show internal_dict. */
1512 if (dict_get_var_cnt (internal_dict) == 0)
1514 dict_destroy (internal_dict);
1515 internal_dict = NULL;