1 /* PSPP - computes sample statistics.
2 Copyright (C) 1997-9, 2000, 2006, 2007 Free Software Foundation, Inc.
4 This program is free software; you can redistribute it and/or
5 modify it under the terms of the GNU General Public License as
6 published by the Free Software Foundation; either version 2 of the
7 License, or (at your option) any later version.
9 This program is distributed in the hope that it will be useful, but
10 WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 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, write to the Free Software
16 Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
21 #include "dictionary.h"
27 #include "cat-routines.h"
30 #include "value-labels.h"
34 #include <libpspp/alloc.h>
35 #include <libpspp/array.h>
36 #include <libpspp/compiler.h>
37 #include <libpspp/hash.h>
38 #include <libpspp/message.h>
39 #include <libpspp/misc.h>
40 #include <libpspp/pool.h>
41 #include <libpspp/str.h>
46 #define _(msgid) gettext (msgid)
51 struct variable **var; /* Variables. */
52 size_t var_cnt, var_cap; /* Number of variables, capacity. */
53 struct hsh_table *name_tab; /* Variable index by name. */
54 int next_value_idx; /* Index of next `union value' to allocate. */
55 struct variable **split; /* SPLIT FILE vars. */
56 size_t split_cnt; /* SPLIT FILE count. */
57 struct variable *weight; /* WEIGHT variable. */
58 struct variable *filter; /* FILTER variable. */
59 size_t case_limit; /* Current case limit (N command). */
60 char *label; /* File label. */
61 char *documents; /* Documents, as a string. */
62 struct vector **vector; /* Vectors of variables. */
63 size_t vector_cnt; /* Number of vectors. */
64 const struct dict_callbacks *callbacks; /* Callbacks on dictionary
66 void *cb_data ; /* Data passed to callbacks */
70 /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
71 the dictionary or any of the variables it contains are modified.
72 Each callback will get passed CALLBACK_DATA.
73 Any callback may be NULL, in which case it'll be ignored.
76 dict_set_callbacks (struct dictionary *dict,
77 const struct dict_callbacks *callbacks,
80 dict->callbacks = callbacks;
81 dict->cb_data = callback_data;
84 /* Shallow copy the callbacks from SRC to DEST */
86 dict_copy_callbacks (struct dictionary *dest,
87 const struct dictionary *src)
89 dest->callbacks = src->callbacks;
90 dest->cb_data = src->cb_data;
93 /* Creates and returns a new dictionary. */
97 struct dictionary *d = xzalloc (sizeof *d);
99 d->name_tab = hsh_create (8, compare_vars_by_name, hash_var_by_name,
104 /* Creates and returns a (deep) copy of an existing
107 dict_clone (const struct dictionary *s)
109 struct dictionary *d;
116 for (i = 0; i < s->var_cnt; i++)
118 struct variable *sv = s->var[i];
119 struct variable *dv = dict_clone_var_assert (d, sv, var_get_name (sv));
120 var_set_short_name (dv, var_get_short_name (sv));
123 d->next_value_idx = s->next_value_idx;
125 d->split_cnt = s->split_cnt;
126 if (d->split_cnt > 0)
128 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
129 for (i = 0; i < d->split_cnt; i++)
130 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
133 if (s->weight != NULL)
134 dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight)));
136 if (s->filter != NULL)
137 dict_set_filter (d, dict_lookup_var_assert (d, var_get_name (s->filter)));
139 d->case_limit = s->case_limit;
140 dict_set_label (d, dict_get_label (s));
141 dict_set_documents (d, dict_get_documents (s));
143 d->vector_cnt = s->vector_cnt;
144 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
145 for (i = 0; i < s->vector_cnt; i++)
146 d->vector[i] = vector_clone (s->vector[i], s, d);
151 /* Clears the contents from a dictionary without destroying the
152 dictionary itself. */
154 dict_clear (struct dictionary *d)
156 /* FIXME? Should we really clear case_limit, label, documents?
157 Others are necessarily cleared by deleting all the variables.*/
162 for (i = 0; i < d->var_cnt; i++)
164 if (d->callbacks && d->callbacks->var_deleted )
165 d->callbacks->var_deleted (d, i, d->cb_data);
167 var_clear_vardict (d->var[i]);
168 var_destroy (d->var[i]);
172 d->var_cnt = d->var_cap = 0;
173 hsh_clear (d->name_tab);
174 d->next_value_idx = 0;
175 dict_set_split_vars (d, NULL, 0);
176 dict_set_weight (d, NULL);
177 dict_set_filter (d, NULL);
183 dict_clear_vectors (d);
186 /* Destroys the aux data for every variable in D, by calling
187 var_clear_aux() for each variable. */
189 dict_clear_aux (struct dictionary *d)
195 for (i = 0; i < d->var_cnt; i++)
196 var_clear_aux (d->var[i]);
199 /* Clears a dictionary and destroys it. */
201 dict_destroy (struct dictionary *d)
205 /* In general, we don't want callbacks occuring, if the dictionary
206 is being destroyed */
207 d->callbacks = NULL ;
210 hsh_destroy (d->name_tab);
215 /* Returns the number of variables in D. */
217 dict_get_var_cnt (const struct dictionary *d)
224 /* Returns the variable in D with dictionary index IDX, which
225 must be between 0 and the count returned by
226 dict_get_var_cnt(), exclusive. */
228 dict_get_var (const struct dictionary *d, size_t idx)
231 assert (idx < d->var_cnt);
236 /* Sets *VARS to an array of pointers to variables in D and *CNT
237 to the number of variables in *D. All variables are returned
238 if EXCLUDE_CLASSES is 0, or it may contain one or more of (1u
239 << DC_ORDINARY), (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to
240 exclude the corresponding type of variable. */
242 dict_get_vars (const struct dictionary *d, struct variable ***vars,
243 size_t *cnt, unsigned exclude_classes)
249 assert (vars != NULL);
250 assert (cnt != NULL);
251 assert ((exclude_classes & ~((1u << DC_ORDINARY)
253 | (1u << DC_SCRATCH))) == 0);
256 for (i = 0; i < d->var_cnt; i++)
258 enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
259 if (!(exclude_classes & (1u << class)))
263 *vars = xnmalloc (count, sizeof **vars);
265 for (i = 0; i < d->var_cnt; i++)
267 enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
268 if (!(exclude_classes & (1u << class)))
269 (*vars)[(*cnt)++] = d->var[i];
271 assert (*cnt == count);
274 static struct variable *
275 add_var (struct dictionary *d, struct variable *v)
277 /* Add dictionary info to variable. */
278 struct vardict_info vdi;
279 vdi.case_index = d->next_value_idx;
280 vdi.dict_index = d->var_cnt;
282 var_set_vardict (v, &vdi);
284 /* Update dictionary. */
285 if (d->var_cnt >= d->var_cap)
287 d->var_cap = 8 + 2 * d->var_cap;
288 d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var);
290 d->var[d->var_cnt++] = v;
291 hsh_force_insert (d->name_tab, v);
293 if ( d->callbacks && d->callbacks->var_added )
294 d->callbacks->var_added (d, d->next_value_idx, d->cb_data);
296 d->next_value_idx += var_get_value_cnt (v);
301 /* Creates and returns a new variable in D with the given NAME
302 and WIDTH. Returns a null pointer if the given NAME would
303 duplicate that of an existing variable in the dictionary. */
305 dict_create_var (struct dictionary *d, const char *name, int width)
307 return (dict_lookup_var (d, name) == NULL
308 ? dict_create_var_assert (d, name, width)
312 /* Creates and returns a new variable in D with the given NAME
313 and WIDTH. Assert-fails if the given NAME would duplicate
314 that of an existing variable in the dictionary. */
316 dict_create_var_assert (struct dictionary *d, const char *name, int width)
318 assert (dict_lookup_var (d, name) == NULL);
319 return add_var (d, var_create (name, width));
322 /* Creates and returns a new variable in D with name NAME, as a
323 copy of existing variable OLD_VAR, which need not be in D or
324 in any dictionary. Returns a null pointer if the given NAME
325 would duplicate that of an existing variable in the
328 dict_clone_var (struct dictionary *d, const struct variable *old_var,
331 return (dict_lookup_var (d, name) == NULL
332 ? dict_clone_var_assert (d, old_var, name)
336 /* Creates and returns a new variable in D with name NAME, as a
337 copy of existing variable OLD_VAR, which need not be in D or
338 in any dictionary. Assert-fails if the given NAME would
339 duplicate that of an existing variable in the dictionary. */
341 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var,
344 struct variable *new_var = var_clone (old_var);
345 assert (dict_lookup_var (d, name) == NULL);
346 var_set_name (new_var, name);
347 return add_var (d, new_var);
350 /* Returns the variable named NAME in D, or a null pointer if no
351 variable has that name. */
353 dict_lookup_var (const struct dictionary *d, const char *name)
355 struct variable *target = var_create (name, 0);
356 struct variable *result = hsh_find (d->name_tab, target);
357 var_destroy (target);
361 /* Returns the variable named NAME in D. Assert-fails if no
362 variable has that name. */
364 dict_lookup_var_assert (const struct dictionary *d, const char *name)
366 struct variable *v = dict_lookup_var (d, name);
371 /* Returns true if variable V is in dictionary D,
374 dict_contains_var (const struct dictionary *d, const struct variable *v)
376 if (var_has_vardict (v))
378 const struct vardict_info *vdi = var_get_vardict (v);
379 return (vdi->dict_index >= 0
380 && vdi->dict_index < d->var_cnt
381 && d->var[vdi->dict_index] == v);
387 /* Compares two double pointers to variables, which should point
388 to elements of a struct dictionary's `var' member array. */
390 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
392 struct variable *const *a = a_;
393 struct variable *const *b = b_;
395 return *a < *b ? -1 : *a > *b;
398 /* Sets the dict_index in V's vardict to DICT_INDEX. */
400 set_var_dict_index (struct variable *v, int dict_index)
402 struct vardict_info vdi = *var_get_vardict (v);
403 struct dictionary *d = vdi.dict;
404 vdi.dict_index = dict_index;
405 var_set_vardict (v, &vdi);
407 if ( d->callbacks && d->callbacks->var_changed )
408 d->callbacks->var_changed (d, dict_index, d->cb_data);
411 /* Sets the case_index in V's vardict to DICT_INDEX. */
413 set_var_case_index (struct variable *v, int case_index)
415 struct vardict_info vdi = *var_get_vardict (v);
416 vdi.case_index = case_index;
417 var_set_vardict (v, &vdi);
420 /* Re-sets the dict_index in the dictionary variables with
421 indexes from FROM to TO (exclusive). */
423 reindex_vars (struct dictionary *d, size_t from, size_t to)
427 for (i = from; i < to; i++)
428 set_var_dict_index (d->var[i], i);
431 /* Deletes variable V from dictionary D and frees V.
433 This is a very bad idea if there might be any pointers to V
434 from outside D. In general, no variable in should be deleted when
435 any transformations are active on the dictionary's dataset, because
436 those transformations might reference the deleted variable.
437 The safest time to delete a variable is just after a procedure
438 has been executed, as done by MODIFY VARS.
440 Pointers to V within D are not a problem, because
441 dict_delete_var() knows to remove V from split variables,
442 weights, filters, etc. */
444 dict_delete_var (struct dictionary *d, struct variable *v)
446 int dict_index = var_get_dict_index (v);
448 assert (dict_contains_var (d, v));
450 /* Delete aux data. */
453 dict_unset_split_var (d, v);
456 dict_set_weight (d, NULL);
459 dict_set_filter (d, NULL);
461 dict_clear_vectors (d);
463 /* Remove V from var array. */
464 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
467 /* Update dict_index for each affected variable. */
468 reindex_vars (d, dict_index, d->var_cnt);
470 /* Update name hash. */
471 hsh_force_delete (d->name_tab, v);
475 var_clear_vardict (v);
478 if (d->callbacks && d->callbacks->var_deleted )
479 d->callbacks->var_deleted (d, dict_index, d->cb_data);
482 /* Deletes the COUNT variables listed in VARS from D. This is
483 unsafe; see the comment on dict_delete_var() for details. */
485 dict_delete_vars (struct dictionary *d,
486 struct variable *const *vars, size_t count)
488 /* FIXME: this can be done in O(count) time, but this algorithm
491 assert (count == 0 || vars != NULL);
494 dict_delete_var (d, *vars++);
497 /* Deletes the COUNT variables in D starting at index IDX. This
498 is unsafe; see the comment on dict_delete_var() for
501 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
503 /* FIXME: this can be done in O(count) time, but this algorithm
505 assert (idx + count <= d->var_cnt);
508 dict_delete_var (d, d->var[idx]);
511 /* Deletes scratch variables from dictionary D. */
513 dict_delete_scratch_vars (struct dictionary *d)
517 /* FIXME: this can be done in O(count) time, but this algorithm
521 for (i = 0; i < d->var_cnt; )
522 if (dict_class_from_id (var_get_name (d->var[i])) == DC_SCRATCH)
523 dict_delete_var (d, d->var[i]);
528 /* Moves V to 0-based position IDX in D. Other variables in D,
529 if any, retain their relative positions. Runs in time linear
530 in the distance moved. */
532 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
534 size_t old_index = var_get_dict_index (v);
536 assert (new_index < d->var_cnt);
537 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
538 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
541 /* Reorders the variables in D, placing the COUNT variables
542 listed in ORDER in that order at the beginning of D. The
543 other variables in D, if any, retain their relative
546 dict_reorder_vars (struct dictionary *d,
547 struct variable *const *order, size_t count)
549 struct variable **new_var;
553 assert (count == 0 || order != NULL);
554 assert (count <= d->var_cnt);
556 new_var = xnmalloc (d->var_cnt, sizeof *new_var);
557 memcpy (new_var, order, count * sizeof *new_var);
558 for (i = 0; i < count; i++)
560 size_t index = var_get_dict_index (order[i]);
561 assert (d->var[index] == order[i]);
562 d->var[index] = NULL;
563 set_var_dict_index (order[i], i);
565 for (i = 0; i < d->var_cnt; i++)
566 if (d->var[i] != NULL)
568 assert (count < d->var_cnt);
569 new_var[count] = d->var[i];
570 set_var_dict_index (new_var[count], count);
577 /* Changes the name of variable V in dictionary D to NEW_NAME. */
579 rename_var (struct dictionary *d, struct variable *v, const char *new_name)
581 struct vardict_info vdi;
583 assert (dict_contains_var (d, v));
585 vdi = *var_get_vardict (v);
586 var_clear_vardict (v);
587 var_set_name (v, new_name);
588 var_set_vardict (v, &vdi);
591 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
592 a variable named NEW_NAME is already in D, except that
593 NEW_NAME may be the same as V's existing name. */
595 dict_rename_var (struct dictionary *d, struct variable *v,
596 const char *new_name)
598 assert (!strcasecmp (var_get_name (v), new_name)
599 || dict_lookup_var (d, new_name) == NULL);
601 hsh_force_delete (d->name_tab, v);
602 rename_var (d, v, new_name);
603 hsh_force_insert (d->name_tab, v);
605 if (get_algorithm () == ENHANCED)
606 var_clear_short_name (v);
608 if ( d->callbacks && d->callbacks->var_changed )
609 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
612 /* Renames COUNT variables specified in VARS to the names given
613 in NEW_NAMES within dictionary D. If the renaming would
614 result in a duplicate variable name, returns false and stores a
615 name that would be duplicated into *ERR_NAME (if ERR_NAME is
616 non-null). Otherwise, the renaming is successful, and true
619 dict_rename_vars (struct dictionary *d,
620 struct variable **vars, char **new_names, size_t count,
627 assert (count == 0 || vars != NULL);
628 assert (count == 0 || new_names != NULL);
630 /* Save the names of the variables to be renamed. */
631 pool = pool_create ();
632 old_names = pool_nalloc (pool, count, sizeof *old_names);
633 for (i = 0; i < count; i++)
634 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
636 /* Remove the variables to be renamed from the name hash,
638 for (i = 0; i < count; i++)
640 hsh_force_delete (d->name_tab, vars[i]);
641 rename_var (d, vars[i], new_names[i]);
644 /* Add the renamed variables back into the name hash,
645 checking for conflicts. */
646 for (i = 0; i < count; i++)
647 if (hsh_insert (d->name_tab, vars[i]) != NULL)
649 /* There is a name conflict.
650 Back out all the name changes that have already
651 taken place, and indicate failure. */
653 if (err_name != NULL)
654 *err_name = new_names[i];
656 for (i = 0; i < fail_idx; i++)
657 hsh_force_delete (d->name_tab, vars[i]);
659 for (i = 0; i < count; i++)
661 rename_var (d, vars[i], old_names[i]);
662 hsh_force_insert (d->name_tab, vars[i]);
669 /* Clear short names. */
670 if (get_algorithm () == ENHANCED)
671 for (i = 0; i < count; i++)
672 var_clear_short_name (vars[i]);
678 /* Returns the weighting variable in dictionary D, or a null
679 pointer if the dictionary is unweighted. */
681 dict_get_weight (const struct dictionary *d)
684 assert (d->weight == NULL || dict_contains_var (d, d->weight));
689 /* Returns the value of D's weighting variable in case C, except that a
690 negative weight is returned as 0. Returns 1 if the dictionary is
691 unweighted. Will warn about missing, negative, or zero values if
692 warn_on_invalid is true. The function will set warn_on_invalid to false
693 if an invalid weight is found. */
695 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
696 bool *warn_on_invalid)
701 if (d->weight == NULL)
705 double w = case_num (c, d->weight);
706 if (w < 0.0 || var_is_num_missing (d->weight, w, MV_ANY))
708 if ( w == 0.0 && *warn_on_invalid ) {
709 *warn_on_invalid = false;
710 msg (SW, _("At least one case in the data file had a weight value "
711 "that was user-missing, system-missing, zero, or "
712 "negative. These case(s) were ignored."));
718 /* Sets the weighting variable of D to V, or turning off
719 weighting if V is a null pointer. */
721 dict_set_weight (struct dictionary *d, struct variable *v)
724 assert (v == NULL || dict_contains_var (d, v));
725 assert (v == NULL || var_is_numeric (v));
729 if ( d->callbacks && d->callbacks->weight_changed )
730 d->callbacks->weight_changed (d,
731 v ? var_get_dict_index (v) : -1,
735 /* Returns the filter variable in dictionary D (see cmd_filter())
736 or a null pointer if the dictionary is unfiltered. */
738 dict_get_filter (const struct dictionary *d)
741 assert (d->filter == NULL || dict_contains_var (d, d->filter));
746 /* Sets V as the filter variable for dictionary D. Passing a
747 null pointer for V turn off filtering. */
749 dict_set_filter (struct dictionary *d, struct variable *v)
752 assert (v == NULL || dict_contains_var (d, v));
756 if ( d->callbacks && d->callbacks->filter_changed )
757 d->callbacks->filter_changed (d,
758 v ? var_get_dict_index (v) : -1,
762 /* Returns the case limit for dictionary D, or zero if the number
763 of cases is unlimited. */
765 dict_get_case_limit (const struct dictionary *d)
769 return d->case_limit;
772 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
773 0 for CASE_LIMIT to indicate no limit. */
775 dict_set_case_limit (struct dictionary *d, size_t case_limit)
779 d->case_limit = case_limit;
782 /* Returns the case index of the next value to be added to D.
783 This value is the number of `union value's that need to be
784 allocated to store a case for dictionary D. */
786 dict_get_next_value_idx (const struct dictionary *d)
790 return d->next_value_idx;
793 /* Returns the number of bytes needed to store a case for
796 dict_get_case_size (const struct dictionary *d)
800 return sizeof (union value) * dict_get_next_value_idx (d);
803 /* Deletes scratch variables in dictionary D and reassigns values
804 so that fragmentation is eliminated. */
806 dict_compact_values (struct dictionary *d)
810 d->next_value_idx = 0;
811 for (i = 0; i < d->var_cnt; )
813 struct variable *v = d->var[i];
815 if (dict_class_from_id (var_get_name (v)) != DC_SCRATCH)
817 set_var_case_index (v, d->next_value_idx);
818 d->next_value_idx += var_get_value_cnt (v);
822 dict_delete_var (d, v);
826 /* Returns the number of values that would be used by a case if
827 dict_compact_values() were called. */
829 dict_get_compacted_value_cnt (const struct dictionary *d)
835 for (i = 0; i < d->var_cnt; i++)
836 if (dict_class_from_id (var_get_name (d->var[i])) != DC_SCRATCH)
837 cnt += var_get_value_cnt (d->var[i]);
841 /* Creates and returns an array mapping from a dictionary index
842 to the case index that the corresponding variable will have
843 after calling dict_compact_values(). Scratch variables
844 receive -1 for case index because dict_compact_values() will
847 dict_get_compacted_dict_index_to_case_index (const struct dictionary *d)
850 size_t next_value_idx;
853 map = xnmalloc (d->var_cnt, sizeof *map);
855 for (i = 0; i < d->var_cnt; i++)
857 struct variable *v = d->var[i];
859 if (dict_class_from_id (var_get_name (v)) != DC_SCRATCH)
861 map[i] = next_value_idx;
862 next_value_idx += var_get_value_cnt (v);
870 /* Returns true if a case for dictionary D would be smaller after
871 compacting, false otherwise. Compacting a case eliminates
872 "holes" between values and after the last value. Holes are
873 created by deleting variables (or by scratch variables).
875 The return value may differ from whether compacting a case
876 from dictionary D would *change* the case: compacting could
877 rearrange values even if it didn't reduce space
880 dict_compacting_would_shrink (const struct dictionary *d)
882 return dict_get_compacted_value_cnt (d) < dict_get_next_value_idx (d);
885 /* Returns true if a case for dictionary D would change after
886 compacting, false otherwise. Compacting a case eliminates
887 "holes" between values and after the last value. Holes are
888 created by deleting variables (or by scratch variables).
890 The return value may differ from whether compacting a case
891 from dictionary D would *shrink* the case: compacting could
892 rearrange values without reducing space requirements. */
894 dict_compacting_would_change (const struct dictionary *d)
900 for (i = 0; i < dict_get_var_cnt (d); i++)
902 struct variable *v = dict_get_var (d, i);
903 if (var_get_case_index (v) != case_idx)
905 case_idx += var_get_value_cnt (v);
910 /* How to copy a contiguous range of values between cases. */
913 size_t src_idx; /* Starting value index in source case. */
914 size_t dst_idx; /* Starting value index in target case. */
915 size_t cnt; /* Number of values. */
918 /* How to compact a case. */
919 struct dict_compactor
921 struct copy_map *maps; /* Array of mappings. */
922 size_t map_cnt; /* Number of mappings. */
925 /* Creates and returns a dict_compactor that can be used to
926 compact cases for dictionary D.
928 Compacting a case eliminates "holes" between values and after
929 the last value. Holes are created by deleting variables (or
930 by scratch variables). */
931 struct dict_compactor *
932 dict_make_compactor (const struct dictionary *d)
934 struct dict_compactor *compactor;
935 struct copy_map *map;
936 size_t map_allocated;
940 compactor = xmalloc (sizeof *compactor);
941 compactor->maps = NULL;
942 compactor->map_cnt = 0;
947 for (i = 0; i < d->var_cnt; i++)
949 struct variable *v = d->var[i];
951 if (dict_class_from_id (var_get_name (v)) == DC_SCRATCH)
953 if (map != NULL && map->src_idx + map->cnt == var_get_case_index (v))
954 map->cnt += var_get_value_cnt (v);
957 if (compactor->map_cnt == map_allocated)
958 compactor->maps = x2nrealloc (compactor->maps, &map_allocated,
959 sizeof *compactor->maps);
960 map = &compactor->maps[compactor->map_cnt++];
961 map->src_idx = var_get_case_index (v);
962 map->dst_idx = value_idx;
963 map->cnt = var_get_value_cnt (v);
965 value_idx += var_get_value_cnt (v);
971 /* Compacts SRC by copying it to DST according to the scheme in
974 Compacting a case eliminates "holes" between values and after
975 the last value. Holes are created by deleting variables (or
976 by scratch variables). */
978 dict_compactor_compact (const struct dict_compactor *compactor,
979 struct ccase *dst, const struct ccase *src)
983 for (i = 0; i < compactor->map_cnt; i++)
985 const struct copy_map *map = &compactor->maps[i];
986 case_copy (dst, map->dst_idx, src, map->src_idx, map->cnt);
990 /* Destroys COMPACTOR. */
992 dict_compactor_destroy (struct dict_compactor *compactor)
994 if (compactor != NULL)
996 free (compactor->maps);
1001 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
1002 dict_get_split_cnt() to determine how many SPLIT FILE vars
1003 there are. Returns a null pointer if and only if there are no
1005 struct variable *const *
1006 dict_get_split_vars (const struct dictionary *d)
1013 /* Returns the number of SPLIT FILE vars. */
1015 dict_get_split_cnt (const struct dictionary *d)
1019 return d->split_cnt;
1022 /* Removes variable V from the set of split variables in dictionary D */
1024 dict_unset_split_var (struct dictionary *d,
1027 const int count = d->split_cnt;
1028 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
1029 &v, compare_var_ptrs, NULL);
1031 if ( count == d->split_cnt)
1034 if ( d->callbacks && d->callbacks->split_changed )
1035 d->callbacks->split_changed (d, d->cb_data);
1038 /* Sets CNT split vars SPLIT in dictionary D. */
1040 dict_set_split_vars (struct dictionary *d,
1041 struct variable *const *split, size_t cnt)
1044 assert (cnt == 0 || split != NULL);
1047 d->split = xnrealloc (d->split, cnt, sizeof *d->split);
1048 memcpy (d->split, split, cnt * sizeof *d->split);
1050 if ( d->callbacks && d->callbacks->split_changed )
1051 d->callbacks->split_changed (d, d->cb_data);
1054 /* Returns the file label for D, or a null pointer if D is
1055 unlabeled (see cmd_file_label()). */
1057 dict_get_label (const struct dictionary *d)
1064 /* Sets D's file label to LABEL, truncating it to a maximum of 60
1067 dict_set_label (struct dictionary *d, const char *label)
1074 else if (strlen (label) < 60)
1075 d->label = xstrdup (label);
1078 d->label = xmalloc (61);
1079 memcpy (d->label, label, 60);
1080 d->label[60] = '\0';
1084 /* Returns the documents for D, or a null pointer if D has no
1085 documents (see cmd_document()).. */
1087 dict_get_documents (const struct dictionary *d)
1091 return d->documents;
1094 /* Sets the documents for D to DOCUMENTS, or removes D's
1095 documents if DOCUMENT is a null pointer. */
1097 dict_set_documents (struct dictionary *d, const char *documents)
1101 free (d->documents);
1102 if (documents == NULL)
1103 d->documents = NULL;
1105 d->documents = xstrdup (documents);
1108 /* Creates in D a vector named NAME that contains the CNT
1109 variables in VAR. Returns true if successful, or false if a
1110 vector named NAME already exists in D. */
1112 dict_create_vector (struct dictionary *d,
1114 struct variable **var, size_t cnt)
1118 assert (var != NULL);
1120 for (i = 0; i < cnt; i++)
1121 assert (dict_contains_var (d, var[i]));
1123 if (dict_lookup_vector (d, name) == NULL)
1125 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1126 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1133 /* Returns the vector in D with index IDX, which must be less
1134 than dict_get_vector_cnt (D). */
1135 const struct vector *
1136 dict_get_vector (const struct dictionary *d, size_t idx)
1139 assert (idx < d->vector_cnt);
1141 return d->vector[idx];
1144 /* Returns the number of vectors in D. */
1146 dict_get_vector_cnt (const struct dictionary *d)
1150 return d->vector_cnt;
1153 /* Looks up and returns the vector within D with the given
1155 const struct vector *
1156 dict_lookup_vector (const struct dictionary *d, const char *name)
1159 for (i = 0; i < d->vector_cnt; i++)
1160 if (!strcasecmp (vector_get_name (d->vector[i]), name))
1161 return d->vector[i];
1165 /* Deletes all vectors from D. */
1167 dict_clear_vectors (struct dictionary *d)
1171 for (i = 0; i < d->vector_cnt; i++)
1172 vector_destroy (d->vector[i]);
1179 /* Compares two strings. */
1181 compare_strings (const void *a, const void *b, const void *aux UNUSED)
1183 return strcmp (a, b);
1186 /* Hashes a string. */
1188 hash_string (const void *s, const void *aux UNUSED)
1190 return hsh_hash_string (s);
1194 /* Sets V's short name to BASE, followed by a suffix of the form
1195 _A, _B, _C, ..., _AA, _AB, etc. according to the value of
1196 SUFFIX_NUMBER. Truncates BASE as necessary to fit. */
1198 set_var_short_name_suffix (struct variable *v, const char *base,
1201 char suffix[SHORT_NAME_LEN + 1];
1202 char short_name[SHORT_NAME_LEN + 1];
1207 assert (suffix_number >= 0);
1209 /* Set base name. */
1210 var_set_short_name (v, base);
1212 /* Compose suffix. */
1213 start = end = suffix + sizeof suffix - 1;
1217 *--start = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"[suffix_number % 26];
1218 if (start <= suffix + 1)
1219 msg (SE, _("Variable suffix too large."));
1220 suffix_number /= 26;
1222 while (suffix_number > 0);
1225 /* Append suffix to V's short name. */
1226 str_copy_trunc (short_name, sizeof short_name, base);
1228 if (len + strlen (short_name) > SHORT_NAME_LEN)
1229 ofs = SHORT_NAME_LEN - len;
1231 ofs = strlen (short_name);
1232 strcpy (short_name + ofs, start);
1235 var_set_short_name (v, short_name);
1238 /* Assigns a valid, unique short_name[] to each variable in D.
1239 Each variable whose actual name is short has highest priority
1240 for that short name. Otherwise, variables with an existing
1241 short_name[] have the next highest priority for a given short
1242 name; if it is already taken, then the variable is treated as
1243 if short_name[] had been empty. Otherwise, long names are
1244 truncated to form short names. If that causes conflicts,
1245 variables are renamed as PREFIX_A, PREFIX_B, and so on. */
1247 dict_assign_short_names (struct dictionary *d)
1249 struct hsh_table *short_names;
1252 /* Give variables whose names are short the corresponding short
1253 names, and clear short_names[] that conflict with a variable
1255 for (i = 0; i < d->var_cnt; i++)
1257 struct variable *v = d->var[i];
1258 const char *short_name = var_get_short_name (v);
1259 if (strlen (var_get_name (v)) <= SHORT_NAME_LEN)
1260 var_set_short_name (v, var_get_name (v));
1261 else if (short_name != NULL && dict_lookup_var (d, short_name) != NULL)
1262 var_clear_short_name (v);
1265 /* Each variable with an assigned short_name[] now gets it
1266 unless there is a conflict. */
1267 short_names = hsh_create (d->var_cnt, compare_strings, hash_string,
1269 for (i = 0; i < d->var_cnt; i++)
1271 struct variable *v = d->var[i];
1272 const char *name = var_get_short_name (v);
1273 if (name != NULL && hsh_insert (short_names, (char *) name) != NULL)
1274 var_clear_short_name (v);
1277 /* Now assign short names to remaining variables. */
1278 for (i = 0; i < d->var_cnt; i++)
1280 struct variable *v = d->var[i];
1281 const char *name = var_get_short_name (v);
1284 /* Form initial short_name from the variable name, then
1285 try _A, _B, ... _AA, _AB, etc., if needed.*/
1290 var_set_short_name (v, var_get_name (v));
1292 set_var_short_name_suffix (v, var_get_name (v), trial - 1);
1296 while (hsh_insert (short_names, (char *) var_get_short_name (v))
1301 /* Get rid of hash table. */
1302 hsh_destroy (short_names);
1306 /* Called from variable.c to notify the dictionary that some property of
1307 the variable has changed */
1309 dict_var_changed (const struct variable *v)
1311 if ( var_has_vardict (v))
1313 const struct vardict_info *vdi = var_get_vardict (v);
1314 struct dictionary *d;
1318 if ( d->callbacks && d->callbacks->var_changed )
1319 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);