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)
84 d->encoding = xstrdup (enc);
88 dict_get_encoding (const struct dictionary *d)
95 dict_set_change_callback (struct dictionary *d,
96 void (*changed) (struct dictionary *, void*),
100 d->changed_data = data;
103 /* Discards dictionary D's caseproto. (It will be regenerated
104 lazily, on demand.) */
106 invalidate_proto (struct dictionary *d)
108 caseproto_unref (d->proto);
112 /* Print a representation of dictionary D to stdout, for
113 debugging purposes. */
115 dict_dump (const struct dictionary *d)
118 for (i = 0 ; i < d->var_cnt ; ++i )
120 const struct variable *v =
122 printf ("Name: %s;\tdict_idx: %zu; case_idx: %zu\n",
124 var_get_dict_index (v),
125 var_get_case_index (v));
130 /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
131 the dictionary or any of the variables it contains are modified.
132 Each callback will get passed CALLBACK_DATA.
133 Any callback may be NULL, in which case it'll be ignored.
136 dict_set_callbacks (struct dictionary *dict,
137 const struct dict_callbacks *callbacks,
140 dict->callbacks = callbacks;
141 dict->cb_data = callback_data;
144 /* Shallow copy the callbacks from SRC to DEST */
146 dict_copy_callbacks (struct dictionary *dest,
147 const struct dictionary *src)
149 dest->callbacks = src->callbacks;
150 dest->cb_data = src->cb_data;
153 /* Creates and returns a new dictionary. */
157 struct dictionary *d = xzalloc (sizeof *d);
159 d->name_tab = hsh_create (8, compare_vars_by_name, hash_var_by_name,
161 attrset_init (&d->attributes);
165 /* Creates and returns a (deep) copy of an existing
168 The new dictionary's case indexes are copied from the old
169 dictionary. If the new dictionary won't be used to access
170 cases produced with the old dictionary, then the new
171 dictionary's case indexes should be compacted with
172 dict_compact_values to save space. */
174 dict_clone (const struct dictionary *s)
176 struct dictionary *d;
183 for (i = 0; i < s->var_cnt; i++)
185 const struct vardict_info *svdi;
186 struct vardict_info dvdi;
187 struct variable *sv = s->var[i];
188 struct variable *dv = dict_clone_var_assert (d, sv);
191 for (i = 0; i < var_get_short_name_cnt (sv); i++)
192 var_set_short_name (dv, i, var_get_short_name (sv, i));
194 svdi = var_get_vardict (sv);
197 var_set_vardict (dv, &dvdi);
200 d->next_value_idx = s->next_value_idx;
202 d->split_cnt = s->split_cnt;
203 if (d->split_cnt > 0)
205 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
206 for (i = 0; i < d->split_cnt; i++)
207 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
210 if (s->weight != NULL)
211 dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight)));
213 if (s->filter != NULL)
214 dict_set_filter (d, dict_lookup_var_assert (d, var_get_name (s->filter)));
216 d->case_limit = s->case_limit;
217 dict_set_label (d, dict_get_label (s));
218 dict_set_documents (d, dict_get_documents (s));
220 d->vector_cnt = s->vector_cnt;
221 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
222 for (i = 0; i < s->vector_cnt; i++)
223 d->vector[i] = vector_clone (s->vector[i], s, d);
226 d->encoding = xstrdup (s->encoding);
228 dict_set_attributes (d, dict_get_attributes (s));
233 /* Clears the contents from a dictionary without destroying the
234 dictionary itself. */
236 dict_clear (struct dictionary *d)
238 /* FIXME? Should we really clear case_limit, label, documents?
239 Others are necessarily cleared by deleting all the variables.*/
242 while (d->var_cnt > 0 )
244 dict_delete_var (d, d->var[d->var_cnt - 1]);
249 d->var_cnt = d->var_cap = 0;
250 invalidate_proto (d);
251 hsh_clear (d->name_tab);
252 d->next_value_idx = 0;
253 dict_set_split_vars (d, NULL, 0);
254 dict_set_weight (d, NULL);
255 dict_set_filter (d, NULL);
259 ds_destroy (&d->documents);
260 dict_clear_vectors (d);
261 attrset_clear (&d->attributes);
264 /* Destroys the aux data for every variable in D, by calling
265 var_clear_aux() for each variable. */
267 dict_clear_aux (struct dictionary *d)
273 for (i = 0; i < d->var_cnt; i++)
274 var_clear_aux (d->var[i]);
277 /* Clears a dictionary and destroys it. */
279 dict_destroy (struct dictionary *d)
283 /* In general, we don't want callbacks occuring, if the dictionary
284 is being destroyed */
285 d->callbacks = NULL ;
288 hsh_destroy (d->name_tab);
289 attrset_destroy (&d->attributes);
294 /* Returns the number of variables in D. */
296 dict_get_var_cnt (const struct dictionary *d)
303 /* Returns the variable in D with dictionary index IDX, which
304 must be between 0 and the count returned by
305 dict_get_var_cnt(), exclusive. */
307 dict_get_var (const struct dictionary *d, size_t idx)
310 assert (idx < d->var_cnt);
315 /* Sets *VARS to an array of pointers to variables in D and *CNT
316 to the number of variables in *D. All variables are returned
317 except for those, if any, in the classes indicated by EXCLUDE.
318 (There is no point in putting DC_SYSTEM in EXCLUDE as
319 dictionaries never include system variables.) */
321 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
322 size_t *cnt, enum dict_class exclude)
324 dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude);
327 /* Sets *VARS to an array of pointers to variables in D and *CNT
328 to the number of variables in *D. All variables are returned
329 except for those, if any, in the classes indicated by EXCLUDE.
330 (There is no point in putting DC_SYSTEM in EXCLUDE as
331 dictionaries never include system variables.) */
333 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
334 size_t *cnt, enum dict_class exclude)
340 assert (vars != NULL);
341 assert (cnt != NULL);
342 assert (exclude == (exclude & DC_ALL));
345 for (i = 0; i < d->var_cnt; i++)
347 enum dict_class class = var_get_dict_class (d->var[i]);
348 if (!(class & exclude))
352 *vars = xnmalloc (count, sizeof **vars);
354 for (i = 0; i < d->var_cnt; i++)
356 enum dict_class class = var_get_dict_class (d->var[i]);
357 if (!(class & exclude))
358 (*vars)[(*cnt)++] = d->var[i];
360 assert (*cnt == count);
363 static struct variable *
364 add_var (struct dictionary *d, struct variable *v)
366 /* Add dictionary info to variable. */
367 struct vardict_info 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 variable *v, int dict_index)
525 struct vardict_info vdi = *var_get_vardict (v);
526 struct dictionary *d = vdi.dict;
527 vdi.dict_index = dict_index;
528 var_set_vardict (v, &vdi);
530 if ( d->changed ) d->changed (d, d->changed_data);
531 if ( d->callbacks && d->callbacks->var_changed )
532 d->callbacks->var_changed (d, dict_index, d->cb_data);
535 /* Sets the case_index in V's vardict to CASE_INDEX. */
537 set_var_case_index (struct variable *v, int case_index)
539 struct vardict_info vdi = *var_get_vardict (v);
540 vdi.case_index = case_index;
541 var_set_vardict (v, &vdi);
544 /* Re-sets the dict_index in the dictionary variables with
545 indexes from FROM to TO (exclusive). */
547 reindex_vars (struct dictionary *d, size_t from, size_t to)
551 for (i = from; i < to; i++)
552 set_var_dict_index (d->var[i], i);
555 /* Deletes variable V from dictionary D and frees V.
557 This is a very bad idea if there might be any pointers to V
558 from outside D. In general, no variable in the active file's
559 dictionary should be deleted when any transformations are
560 active on the dictionary's dataset, because those
561 transformations might reference the deleted variable. The
562 safest time to delete a variable is just after a procedure has
563 been executed, as done by DELETE VARIABLES.
565 Pointers to V within D are not a problem, because
566 dict_delete_var() knows to remove V from split variables,
567 weights, filters, etc. */
569 dict_delete_var (struct dictionary *d, struct variable *v)
571 int dict_index = var_get_dict_index (v);
572 const int case_index = var_get_case_index (v);
573 const int width = var_get_width (v);
575 assert (dict_contains_var (d, v));
577 /* Delete aux data. */
580 dict_unset_split_var (d, v);
583 dict_set_weight (d, NULL);
586 dict_set_filter (d, NULL);
588 dict_clear_vectors (d);
590 /* Remove V from var array. */
591 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
594 /* Update dict_index for each affected variable. */
595 reindex_vars (d, dict_index, d->var_cnt);
597 /* Update name hash. */
598 hsh_force_delete (d->name_tab, v);
602 var_clear_vardict (v);
605 if ( d->changed ) d->changed (d, d->changed_data);
607 invalidate_proto (d);
608 if (d->callbacks && d->callbacks->var_deleted )
609 d->callbacks->var_deleted (d, dict_index, case_index, width, d->cb_data);
612 /* Deletes the COUNT variables listed in VARS from D. This is
613 unsafe; see the comment on dict_delete_var() for details. */
615 dict_delete_vars (struct dictionary *d,
616 struct variable *const *vars, size_t count)
618 /* FIXME: this can be done in O(count) time, but this algorithm
621 assert (count == 0 || vars != NULL);
624 dict_delete_var (d, *vars++);
627 /* Deletes the COUNT variables in D starting at index IDX. This
628 is unsafe; see the comment on dict_delete_var() for
631 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
633 /* FIXME: this can be done in O(count) time, but this algorithm
635 assert (idx + count <= d->var_cnt);
638 dict_delete_var (d, d->var[idx]);
641 /* Deletes scratch variables from dictionary D. */
643 dict_delete_scratch_vars (struct dictionary *d)
647 /* FIXME: this can be done in O(count) time, but this algorithm
651 for (i = 0; i < d->var_cnt; )
652 if (var_get_dict_class (d->var[i]) == DC_SCRATCH)
653 dict_delete_var (d, d->var[i]);
658 /* Moves V to 0-based position IDX in D. Other variables in D,
659 if any, retain their relative positions. Runs in time linear
660 in the distance moved. */
662 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
664 size_t old_index = var_get_dict_index (v);
666 assert (new_index < d->var_cnt);
667 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
668 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
671 /* Reorders the variables in D, placing the COUNT variables
672 listed in ORDER in that order at the beginning of D. The
673 other variables in D, if any, retain their relative
676 dict_reorder_vars (struct dictionary *d,
677 struct variable *const *order, size_t count)
679 struct variable **new_var;
683 assert (count == 0 || order != NULL);
684 assert (count <= d->var_cnt);
686 new_var = xnmalloc (d->var_cnt, sizeof *new_var);
687 memcpy (new_var, order, count * sizeof *new_var);
688 for (i = 0; i < count; i++)
690 size_t index = var_get_dict_index (order[i]);
691 assert (d->var[index] == order[i]);
692 d->var[index] = NULL;
693 set_var_dict_index (order[i], i);
695 for (i = 0; i < d->var_cnt; i++)
696 if (d->var[i] != NULL)
698 assert (count < d->var_cnt);
699 new_var[count] = d->var[i];
700 set_var_dict_index (new_var[count], count);
707 /* Changes the name of variable V in dictionary D to NEW_NAME. */
709 rename_var (struct dictionary *d, struct variable *v, const char *new_name)
711 struct vardict_info vdi;
713 assert (dict_contains_var (d, v));
715 vdi = *var_get_vardict (v);
716 var_clear_vardict (v);
717 var_set_name (v, new_name);
718 var_set_vardict (v, &vdi);
721 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
722 a variable named NEW_NAME is already in D, except that
723 NEW_NAME may be the same as V's existing name. */
725 dict_rename_var (struct dictionary *d, struct variable *v,
726 const char *new_name)
728 assert (!strcasecmp (var_get_name (v), new_name)
729 || dict_lookup_var (d, new_name) == NULL);
731 hsh_force_delete (d->name_tab, v);
732 rename_var (d, v, new_name);
733 hsh_force_insert (d->name_tab, v);
735 if (settings_get_algorithm () == ENHANCED)
736 var_clear_short_names (v);
738 if ( d->changed ) d->changed (d, d->changed_data);
739 if ( d->callbacks && d->callbacks->var_changed )
740 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
743 /* Renames COUNT variables specified in VARS to the names given
744 in NEW_NAMES within dictionary D. If the renaming would
745 result in a duplicate variable name, returns false and stores a
746 name that would be duplicated into *ERR_NAME (if ERR_NAME is
747 non-null). Otherwise, the renaming is successful, and true
750 dict_rename_vars (struct dictionary *d,
751 struct variable **vars, char **new_names, size_t count,
758 assert (count == 0 || vars != NULL);
759 assert (count == 0 || new_names != NULL);
761 /* Save the names of the variables to be renamed. */
762 pool = pool_create ();
763 old_names = pool_nalloc (pool, count, sizeof *old_names);
764 for (i = 0; i < count; i++)
765 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
767 /* Remove the variables to be renamed from the name hash,
769 for (i = 0; i < count; i++)
771 hsh_force_delete (d->name_tab, vars[i]);
772 rename_var (d, vars[i], new_names[i]);
775 /* Add the renamed variables back into the name hash,
776 checking for conflicts. */
777 for (i = 0; i < count; i++)
778 if (hsh_insert (d->name_tab, vars[i]) != NULL)
780 /* There is a name conflict.
781 Back out all the name changes that have already
782 taken place, and indicate failure. */
784 if (err_name != NULL)
785 *err_name = new_names[i];
787 for (i = 0; i < fail_idx; i++)
788 hsh_force_delete (d->name_tab, vars[i]);
790 for (i = 0; i < count; i++)
792 rename_var (d, vars[i], old_names[i]);
793 hsh_force_insert (d->name_tab, vars[i]);
800 /* Clear short names. */
801 if (settings_get_algorithm () == ENHANCED)
802 for (i = 0; i < count; i++)
803 var_clear_short_names (vars[i]);
809 /* Returns true if a variable named NAME may be inserted in DICT;
810 that is, if there is not already a variable with that name in
811 DICT and if NAME is not a reserved word. (The caller's checks
812 have already verified that NAME is otherwise acceptable as a
815 var_name_is_insertable (const struct dictionary *dict, const char *name)
817 return (dict_lookup_var (dict, name) == NULL
818 && lex_id_to_token (ss_cstr (name)) == T_ID);
822 make_hinted_name (const struct dictionary *dict, const char *hint,
823 char name[VAR_NAME_LEN + 1])
825 bool dropped = false;
828 for (cp = name; *hint && cp < name + VAR_NAME_LEN; hint++)
831 ? lex_is_id1 (*hint) && *hint != '$'
832 : lex_is_idn (*hint))
839 if (cp < name + VAR_NAME_LEN)
849 size_t len = strlen (name);
852 if (var_name_is_insertable (dict, name))
855 for (i = 0; i < ULONG_MAX; i++)
857 char suffix[INT_BUFSIZE_BOUND (i) + 1];
861 if (!str_format_26adic (i + 1, &suffix[1], sizeof suffix - 1))
864 ofs = MIN (VAR_NAME_LEN - strlen (suffix), len);
865 strcpy (&name[ofs], suffix);
867 if (var_name_is_insertable (dict, name))
876 make_numeric_name (const struct dictionary *dict, unsigned long int *num_start,
877 char name[VAR_NAME_LEN + 1])
879 unsigned long int number;
881 for (number = num_start != NULL ? MAX (*num_start, 1) : 1;
885 sprintf (name, "VAR%03lu", number);
886 if (dict_lookup_var (dict, name) == NULL)
888 if (num_start != NULL)
889 *num_start = number + 1;
894 if (num_start != NULL)
895 *num_start = ULONG_MAX;
900 /* Attempts to devise a variable name unique within DICT.
901 Returns true if successful, in which case the new variable
902 name is stored into NAME. Returns false if all names that can
903 be generated have already been taken. (Returning false is
904 quite unlikely: at least ULONG_MAX unique names can be
907 HINT, if it is non-null, is used as a suggestion that will be
908 modified for suitability as a variable name and for
911 If HINT is null or entirely unsuitable, a name in the form
912 "VAR%03d" will be generated, where the smallest unused integer
913 value is used. If NUM_START is non-null, then its value is
914 used as the minimum numeric value to check, and it is updated
915 to the next value to be checked.
918 dict_make_unique_var_name (const struct dictionary *dict, const char *hint,
919 unsigned long int *num_start,
920 char name[VAR_NAME_LEN + 1])
922 return ((hint != NULL && make_hinted_name (dict, hint, name))
923 || make_numeric_name (dict, num_start, name));
926 /* Returns the weighting variable in dictionary D, or a null
927 pointer if the dictionary is unweighted. */
929 dict_get_weight (const struct dictionary *d)
932 assert (d->weight == NULL || dict_contains_var (d, d->weight));
937 /* Returns the value of D's weighting variable in case C, except
938 that a negative weight is returned as 0. Returns 1 if the
939 dictionary is unweighted. Will warn about missing, negative,
940 or zero values if *WARN_ON_INVALID is true. The function will
941 set *WARN_ON_INVALID to false if an invalid weight is
944 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
945 bool *warn_on_invalid)
950 if (d->weight == NULL)
954 double w = case_num (c, d->weight);
955 if (w < 0.0 || var_is_num_missing (d->weight, w, MV_ANY))
957 if ( w == 0.0 && warn_on_invalid != NULL && *warn_on_invalid ) {
958 *warn_on_invalid = false;
959 msg (SW, _("At least one case in the data file had a weight value "
960 "that was user-missing, system-missing, zero, or "
961 "negative. These case(s) were ignored."));
967 /* Sets the weighting variable of D to V, or turning off
968 weighting if V is a null pointer. */
970 dict_set_weight (struct dictionary *d, struct variable *v)
973 assert (v == NULL || dict_contains_var (d, v));
974 assert (v == NULL || var_is_numeric (v));
978 if (d->changed) d->changed (d, d->changed_data);
979 if ( d->callbacks && d->callbacks->weight_changed )
980 d->callbacks->weight_changed (d,
981 v ? var_get_dict_index (v) : -1,
985 /* Returns the filter variable in dictionary D (see cmd_filter())
986 or a null pointer if the dictionary is unfiltered. */
988 dict_get_filter (const struct dictionary *d)
991 assert (d->filter == NULL || dict_contains_var (d, d->filter));
996 /* Sets V as the filter variable for dictionary D. Passing a
997 null pointer for V turn off filtering. */
999 dict_set_filter (struct dictionary *d, struct variable *v)
1002 assert (v == NULL || dict_contains_var (d, v));
1003 assert (v == NULL || var_is_numeric (v));
1007 if (d->changed) d->changed (d, d->changed_data);
1008 if ( d->callbacks && d->callbacks->filter_changed )
1009 d->callbacks->filter_changed (d,
1010 v ? var_get_dict_index (v) : -1,
1014 /* Returns the case limit for dictionary D, or zero if the number
1015 of cases is unlimited. */
1017 dict_get_case_limit (const struct dictionary *d)
1021 return d->case_limit;
1024 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
1025 0 for CASE_LIMIT to indicate no limit. */
1027 dict_set_case_limit (struct dictionary *d, casenumber case_limit)
1031 d->case_limit = case_limit;
1034 /* Returns the prototype used for cases created by dictionary D. */
1035 const struct caseproto *
1036 dict_get_proto (const struct dictionary *d_)
1038 struct dictionary *d = CONST_CAST (struct dictionary *, d_);
1039 if (d->proto == NULL)
1043 d->proto = caseproto_create ();
1044 d->proto = caseproto_reserve (d->proto, d->var_cnt);
1045 for (i = 0; i < d->var_cnt; i++)
1046 d->proto = caseproto_set_width (d->proto,
1047 var_get_case_index (d->var[i]),
1048 var_get_width (d->var[i]));
1053 /* Returns the case index of the next value to be added to D.
1054 This value is the number of `union value's that need to be
1055 allocated to store a case for dictionary D. */
1057 dict_get_next_value_idx (const struct dictionary *d)
1061 return d->next_value_idx;
1064 /* Returns the number of bytes needed to store a case for
1067 dict_get_case_size (const struct dictionary *d)
1071 return sizeof (union value) * dict_get_next_value_idx (d);
1074 /* Reassigns values in dictionary D so that fragmentation is
1077 dict_compact_values (struct dictionary *d)
1081 d->next_value_idx = 0;
1082 for (i = 0; i < d->var_cnt; i++)
1084 struct variable *v = d->var[i];
1085 set_var_case_index (v, d->next_value_idx++);
1087 invalidate_proto (d);
1090 /* Returns the number of values occupied by the variables in
1091 dictionary D. All variables are considered if EXCLUDE_CLASSES
1092 is 0, or it may contain one or more of (1u << DC_ORDINARY),
1093 (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to exclude the
1094 corresponding type of variable.
1096 The return value may be less than the number of values in one
1097 of dictionary D's cases (as returned by
1098 dict_get_next_value_idx) even if E is 0, because there may be
1099 gaps in D's cases due to deleted variables. */
1101 dict_count_values (const struct dictionary *d, unsigned int exclude_classes)
1106 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1108 | (1u << DC_SCRATCH))) == 0);
1111 for (i = 0; i < d->var_cnt; i++)
1113 enum dict_class class = var_get_dict_class (d->var[i]);
1114 if (!(exclude_classes & (1u << class)))
1120 /* Returns the case prototype that would result after deleting
1121 all variables from D that are not in one of the
1122 EXCLUDE_CLASSES and compacting the dictionary with
1125 The caller must unref the returned caseproto when it is no
1128 dict_get_compacted_proto (const struct dictionary *d,
1129 unsigned int exclude_classes)
1131 struct caseproto *proto;
1134 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1136 | (1u << DC_SCRATCH))) == 0);
1138 proto = caseproto_create ();
1139 for (i = 0; i < d->var_cnt; i++)
1141 struct variable *v = d->var[i];
1142 if (!(exclude_classes & (1u << var_get_dict_class (v))))
1143 proto = caseproto_add_width (proto, var_get_width (v));
1148 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
1149 dict_get_split_cnt() to determine how many SPLIT FILE vars
1150 there are. Returns a null pointer if and only if there are no
1152 const struct variable *const *
1153 dict_get_split_vars (const struct dictionary *d)
1160 /* Returns the number of SPLIT FILE vars. */
1162 dict_get_split_cnt (const struct dictionary *d)
1166 return d->split_cnt;
1169 /* Removes variable V, which must be in D, from D's set of split
1172 dict_unset_split_var (struct dictionary *d, struct variable *v)
1176 assert (dict_contains_var (d, v));
1178 orig_count = d->split_cnt;
1179 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
1180 &v, compare_var_ptrs, NULL);
1181 if (orig_count != d->split_cnt)
1183 if (d->changed) d->changed (d, d->changed_data);
1184 /* We changed the set of split variables so invoke the
1186 if (d->callbacks && d->callbacks->split_changed)
1187 d->callbacks->split_changed (d, d->cb_data);
1191 /* Sets CNT split vars SPLIT in dictionary D. */
1193 dict_set_split_vars (struct dictionary *d,
1194 struct variable *const *split, size_t cnt)
1197 assert (cnt == 0 || split != NULL);
1202 d->split = xnrealloc (d->split, cnt, sizeof *d->split) ;
1203 memcpy (d->split, split, cnt * sizeof *d->split);
1211 if (d->changed) d->changed (d, d->changed_data);
1212 if ( d->callbacks && d->callbacks->split_changed )
1213 d->callbacks->split_changed (d, d->cb_data);
1216 /* Returns the file label for D, or a null pointer if D is
1217 unlabeled (see cmd_file_label()). */
1219 dict_get_label (const struct dictionary *d)
1226 /* Sets D's file label to LABEL, truncating it to a maximum of 60
1229 dict_set_label (struct dictionary *d, const char *label)
1234 d->label = label != NULL ? xstrndup (label, 60) : NULL;
1237 /* Returns the documents for D, or a null pointer if D has no
1238 documents. If the return value is nonnull, then the string
1239 will be an exact multiple of DOC_LINE_LENGTH bytes in length,
1240 with each segment corresponding to one line. */
1242 dict_get_documents (const struct dictionary *d)
1244 return ds_is_empty (&d->documents) ? NULL : ds_cstr (&d->documents);
1247 /* Sets the documents for D to DOCUMENTS, or removes D's
1248 documents if DOCUMENT is a null pointer. If DOCUMENTS is
1249 nonnull, then it should be an exact multiple of
1250 DOC_LINE_LENGTH bytes in length, with each segment
1251 corresponding to one line. */
1253 dict_set_documents (struct dictionary *d, const char *documents)
1257 ds_assign_cstr (&d->documents, documents != NULL ? documents : "");
1259 /* In case the caller didn't get it quite right, pad out the
1260 final line with spaces. */
1261 remainder = ds_length (&d->documents) % DOC_LINE_LENGTH;
1263 ds_put_char_multiple (&d->documents, ' ', DOC_LINE_LENGTH - remainder);
1266 /* Drops the documents from dictionary D. */
1268 dict_clear_documents (struct dictionary *d)
1270 ds_clear (&d->documents);
1273 /* Appends LINE to the documents in D. LINE will be truncated or
1274 padded on the right with spaces to make it exactly
1275 DOC_LINE_LENGTH bytes long. */
1277 dict_add_document_line (struct dictionary *d, const char *line)
1279 if (strlen (line) > DOC_LINE_LENGTH)
1281 /* Note to translators: "bytes" is correct, not characters */
1282 msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
1284 buf_copy_str_rpad (ds_put_uninit (&d->documents, DOC_LINE_LENGTH),
1285 DOC_LINE_LENGTH, line, ' ');
1288 /* Returns the number of document lines in dictionary D. */
1290 dict_get_document_line_cnt (const struct dictionary *d)
1292 return ds_length (&d->documents) / DOC_LINE_LENGTH;
1295 /* Copies document line number IDX from dictionary D into
1296 LINE, trimming off any trailing white space. */
1298 dict_get_document_line (const struct dictionary *d,
1299 size_t idx, struct string *line)
1301 assert (idx < dict_get_document_line_cnt (d));
1302 ds_assign_substring (line, ds_substr (&d->documents, idx * DOC_LINE_LENGTH,
1304 ds_rtrim (line, ss_cstr (CC_SPACES));
1307 /* Creates in D a vector named NAME that contains the CNT
1308 variables in VAR. Returns true if successful, or false if a
1309 vector named NAME already exists in D. */
1311 dict_create_vector (struct dictionary *d,
1313 struct variable **var, size_t cnt)
1317 assert (var != NULL);
1319 for (i = 0; i < cnt; i++)
1320 assert (dict_contains_var (d, var[i]));
1322 if (dict_lookup_vector (d, name) == NULL)
1324 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1325 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1332 /* Creates in D a vector named NAME that contains the CNT
1333 variables in VAR. A vector named NAME must not already exist
1336 dict_create_vector_assert (struct dictionary *d,
1338 struct variable **var, size_t cnt)
1340 assert (dict_lookup_vector (d, name) == NULL);
1341 dict_create_vector (d, name, var, cnt);
1344 /* Returns the vector in D with index IDX, which must be less
1345 than dict_get_vector_cnt (D). */
1346 const struct vector *
1347 dict_get_vector (const struct dictionary *d, size_t idx)
1350 assert (idx < d->vector_cnt);
1352 return d->vector[idx];
1355 /* Returns the number of vectors in D. */
1357 dict_get_vector_cnt (const struct dictionary *d)
1361 return d->vector_cnt;
1364 /* Looks up and returns the vector within D with the given
1366 const struct vector *
1367 dict_lookup_vector (const struct dictionary *d, const char *name)
1370 for (i = 0; i < d->vector_cnt; i++)
1371 if (!strcasecmp (vector_get_name (d->vector[i]), name))
1372 return d->vector[i];
1376 /* Deletes all vectors from D. */
1378 dict_clear_vectors (struct dictionary *d)
1382 for (i = 0; i < d->vector_cnt; i++)
1383 vector_destroy (d->vector[i]);
1390 /* Returns D's attribute set. The caller may examine or modify
1391 the attribute set, but must not destroy it. Destroying D or
1392 calling dict_set_attributes for D will also destroy D's
1395 dict_get_attributes (const struct dictionary *d)
1397 return CONST_CAST (struct attrset *, &d->attributes);
1400 /* Replaces D's attributes set by a copy of ATTRS. */
1402 dict_set_attributes (struct dictionary *d, const struct attrset *attrs)
1404 attrset_destroy (&d->attributes);
1405 attrset_clone (&d->attributes, attrs);
1408 /* Returns true if D has at least one attribute in its attribute
1409 set, false if D's attribute set is empty. */
1411 dict_has_attributes (const struct dictionary *d)
1413 return attrset_count (&d->attributes) > 0;
1416 /* Called from variable.c to notify the dictionary that some property of
1417 the variable has changed */
1419 dict_var_changed (const struct variable *v)
1421 if ( var_has_vardict (v))
1423 const struct vardict_info *vdi = var_get_vardict (v);
1424 struct dictionary *d = vdi->dict;
1429 if (d->changed ) d->changed (d, d->changed_data);
1430 if ( d->callbacks && d->callbacks->var_changed )
1431 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
1436 /* Called from variable.c to notify the dictionary that the variable's width
1439 dict_var_resized (const struct variable *v, int old_width)
1441 if ( var_has_vardict (v))
1443 const struct vardict_info *vdi = var_get_vardict (v);
1444 struct dictionary *d;
1448 if (d->changed) d->changed (d, d->changed_data);
1450 invalidate_proto (d);
1451 if ( d->callbacks && d->callbacks->var_resized )
1452 d->callbacks->var_resized (d, var_get_dict_index (v), old_width,
1457 /* Called from variable.c to notify the dictionary that the variable's display width
1460 dict_var_display_width_changed (const struct variable *v)
1462 if ( var_has_vardict (v))
1464 const struct vardict_info *vdi = var_get_vardict (v);
1465 struct dictionary *d;
1469 if (d->changed) d->changed (d, d->changed_data);
1470 if ( d->callbacks && d->callbacks->var_display_width_changed )
1471 d->callbacks->var_display_width_changed (d, var_get_dict_index (v), d->cb_data);