1 /* PSPP - computes sample statistics.
2 Copyright (C) 1997-9, 2000, 2006 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;
85 /* Shallow copy the callbacks from SRC to DEST */
87 dict_copy_callbacks (struct dictionary *dest,
88 const struct dictionary *src)
90 dest->callbacks = src->callbacks;
91 dest->cb_data = src->cb_data;
94 /* Creates and returns a new dictionary. */
98 struct dictionary *d = xzalloc (sizeof *d);
101 d->var_cnt = d->var_cap = 0;
102 d->name_tab = hsh_create (8, compare_vars_by_name, hash_var_by_name,
104 d->next_value_idx = 0;
118 /* Creates and returns a (deep) copy of an existing
121 dict_clone (const struct dictionary *s)
123 struct dictionary *d;
130 for (i = 0; i < s->var_cnt; i++)
132 struct variable *sv = s->var[i];
133 struct variable *dv = dict_clone_var_assert (d, sv, var_get_name (sv));
134 var_set_short_name (dv, var_get_short_name (sv));
137 d->next_value_idx = s->next_value_idx;
139 d->split_cnt = s->split_cnt;
140 if (d->split_cnt > 0)
142 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
143 for (i = 0; i < d->split_cnt; i++)
144 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
147 if (s->weight != NULL)
148 d->weight = dict_lookup_var_assert (d, var_get_name (s->weight));
150 if (s->filter != NULL)
151 d->filter = dict_lookup_var_assert (d, var_get_name (s->filter));
153 d->case_limit = s->case_limit;
154 dict_set_label (d, dict_get_label (s));
155 dict_set_documents (d, dict_get_documents (s));
157 d->vector_cnt = s->vector_cnt;
158 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
159 for (i = 0; i < s->vector_cnt; i++)
160 d->vector[i] = vector_clone (s->vector[i], s, d);
165 /* Clears the contents from a dictionary without destroying the
166 dictionary itself. */
168 dict_clear (struct dictionary *d)
170 /* FIXME? Should we really clear case_limit, label, documents?
171 Others are necessarily cleared by deleting all the variables.*/
176 for (i = 0; i < d->var_cnt; i++)
178 if (d->callbacks && d->callbacks->var_deleted )
179 d->callbacks->var_deleted (d, i, d->cb_data);
181 var_clear_vardict (d->var[i]);
182 var_destroy (d->var[i]);
186 d->var_cnt = d->var_cap = 0;
187 hsh_clear (d->name_tab);
188 d->next_value_idx = 0;
199 dict_clear_vectors (d);
202 /* Destroys the aux data for every variable in D, by calling
203 var_clear_aux() for each variable. */
205 dict_clear_aux (struct dictionary *d)
211 for (i = 0; i < d->var_cnt; i++)
212 var_clear_aux (d->var[i]);
215 /* Clears a dictionary and destroys it. */
217 dict_destroy (struct dictionary *d)
222 hsh_destroy (d->name_tab);
227 /* Returns the number of variables in D. */
229 dict_get_var_cnt (const struct dictionary *d)
236 /* Returns the variable in D with dictionary index IDX, which
237 must be between 0 and the count returned by
238 dict_get_var_cnt(), exclusive. */
240 dict_get_var (const struct dictionary *d, size_t idx)
243 assert (idx < d->var_cnt);
248 /* Sets *VARS to an array of pointers to variables in D and *CNT
249 to the number of variables in *D. All variables are returned
250 if EXCLUDE_CLASSES is 0, or it may contain one or more of (1u
251 << DC_ORDINARY), (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to
252 exclude the corresponding type of variable. */
254 dict_get_vars (const struct dictionary *d, struct variable ***vars,
255 size_t *cnt, unsigned exclude_classes)
261 assert (vars != NULL);
262 assert (cnt != NULL);
263 assert ((exclude_classes & ~((1u << DC_ORDINARY)
265 | (1u << DC_SCRATCH))) == 0);
268 for (i = 0; i < d->var_cnt; i++)
270 enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
271 if (!(exclude_classes & (1u << class)))
275 *vars = xnmalloc (count, sizeof **vars);
277 for (i = 0; i < d->var_cnt; i++)
279 enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
280 if (!(exclude_classes & (1u << class)))
281 (*vars)[(*cnt)++] = d->var[i];
283 assert (*cnt == count);
286 static struct variable *
287 add_var (struct dictionary *d, struct variable *v)
289 /* Add dictionary info to variable. */
290 struct vardict_info vdi;
291 vdi.case_index = d->next_value_idx;
292 vdi.dict_index = d->var_cnt;
294 var_set_vardict (v, &vdi);
296 /* Update dictionary. */
297 if (d->var_cnt >= d->var_cap)
299 d->var_cap = 8 + 2 * d->var_cap;
300 d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var);
302 d->var[d->var_cnt++] = v;
303 hsh_force_insert (d->name_tab, v);
305 if ( d->callbacks && d->callbacks->var_added )
306 d->callbacks->var_added (d, d->next_value_idx, d->cb_data);
308 d->next_value_idx += var_get_value_cnt (v);
313 /* Creates and returns a new variable in D with the given NAME
314 and WIDTH. Returns a null pointer if the given NAME would
315 duplicate that of an existing variable in the dictionary. */
317 dict_create_var (struct dictionary *d, const char *name, int width)
319 return (dict_lookup_var (d, name) == NULL
320 ? dict_create_var_assert (d, name, width)
324 /* Creates and returns a new variable in D with the given NAME
325 and WIDTH. Assert-fails if the given NAME would duplicate
326 that of an existing variable in the dictionary. */
328 dict_create_var_assert (struct dictionary *d, const char *name, int width)
330 assert (dict_lookup_var (d, name) == NULL);
331 return add_var (d, var_create (name, width));
334 /* Creates and returns a new variable in D with name NAME, as a
335 copy of existing variable OLD_VAR, which need not be in D or
336 in any dictionary. Returns a null pointer if the given NAME
337 would duplicate that of an existing variable in the
340 dict_clone_var (struct dictionary *d, const struct variable *old_var,
343 return (dict_lookup_var (d, name) == NULL
344 ? dict_clone_var_assert (d, old_var, name)
348 /* Creates and returns a new variable in D with name NAME, as a
349 copy of existing variable OLD_VAR, which need not be in D or
350 in any dictionary. Assert-fails if the given NAME would
351 duplicate that of an existing variable in the dictionary. */
353 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var,
356 struct variable *new_var = var_clone (old_var);
357 assert (dict_lookup_var (d, name) == NULL);
358 var_set_name (new_var, name);
359 return add_var (d, new_var);
362 /* Returns the variable named NAME in D, or a null pointer if no
363 variable has that name. */
365 dict_lookup_var (const struct dictionary *d, const char *name)
367 struct variable *target = var_create (name, 0);
368 struct variable *result = hsh_find (d->name_tab, target);
369 var_destroy (target);
373 /* Returns the variable named NAME in D. Assert-fails if no
374 variable has that name. */
376 dict_lookup_var_assert (const struct dictionary *d, const char *name)
378 struct variable *v = dict_lookup_var (d, name);
383 /* Returns true if variable V is in dictionary D,
386 dict_contains_var (const struct dictionary *d, const struct variable *v)
388 if (var_has_vardict (v))
390 const struct vardict_info *vdi = var_get_vardict (v);
391 return (vdi->dict_index >= 0
392 && vdi->dict_index < d->var_cnt
393 && d->var[vdi->dict_index] == v);
399 /* Compares two double pointers to variables, which should point
400 to elements of a struct dictionary's `var' member array. */
402 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
404 struct variable *const *a = a_;
405 struct variable *const *b = b_;
407 return *a < *b ? -1 : *a > *b;
410 /* Sets the dict_index in V's vardict to DICT_INDEX. */
412 set_var_dict_index (struct variable *v, int dict_index)
414 struct vardict_info vdi = *var_get_vardict (v);
415 struct dictionary *d = vdi.dict;
416 vdi.dict_index = dict_index;
417 var_set_vardict (v, &vdi);
419 if ( d->callbacks && d->callbacks->var_changed )
420 d->callbacks->var_changed (d, dict_index, d->cb_data);
423 /* Sets the case_index in V's vardict to DICT_INDEX. */
425 set_var_case_index (struct variable *v, int case_index)
427 struct vardict_info vdi = *var_get_vardict (v);
428 vdi.case_index = case_index;
429 var_set_vardict (v, &vdi);
432 /* Re-sets the dict_index in the dictionary variables with
433 indexes from FROM to TO (exclusive). */
435 reindex_vars (struct dictionary *d, size_t from, size_t to)
439 for (i = from; i < to; i++)
440 set_var_dict_index (d->var[i], i);
443 /* Deletes variable V from dictionary D and frees V.
445 This is a very bad idea if there might be any pointers to V
446 from outside D. In general, no variable in should be deleted when
447 any transformations are active on the dictionary's dataset, because
448 those transformations might reference the deleted variable.
449 The safest time to delete a variable is just after a procedure
450 has been executed, as done by MODIFY VARS.
452 Pointers to V within D are not a problem, because
453 dict_delete_var() knows to remove V from split variables,
454 weights, filters, etc. */
456 dict_delete_var (struct dictionary *d, struct variable *v)
458 int dict_index = var_get_dict_index (v);
460 assert (dict_contains_var (d, v));
462 /* Delete aux data. */
465 /* Remove V from splits, weight, filter variables. */
466 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
467 &v, compare_var_ptrs, NULL);
472 dict_clear_vectors (d);
474 /* Remove V from var array. */
475 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
478 /* Update dict_index for each affected variable. */
479 reindex_vars (d, dict_index, d->var_cnt);
481 /* Update name hash. */
482 hsh_force_delete (d->name_tab, v);
486 var_clear_vardict (v);
489 if (d->callbacks && d->callbacks->var_deleted )
490 d->callbacks->var_deleted (d, dict_index, d->cb_data);
493 /* Deletes the COUNT variables listed in VARS from D. This is
494 unsafe; see the comment on dict_delete_var() for details. */
496 dict_delete_vars (struct dictionary *d,
497 struct variable *const *vars, size_t count)
499 /* FIXME: this can be done in O(count) time, but this algorithm
502 assert (count == 0 || vars != NULL);
505 dict_delete_var (d, *vars++);
508 /* Deletes the COUNT variables in D starting at index IDX. This
509 is unsafe; see the comment on dict_delete_var() for
512 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
514 /* FIXME: this can be done in O(count) time, but this algorithm
516 assert (idx + count <= d->var_cnt);
519 dict_delete_var (d, d->var[idx]);
522 /* Deletes scratch variables from dictionary D. */
524 dict_delete_scratch_vars (struct dictionary *d)
528 /* FIXME: this can be done in O(count) time, but this algorithm
532 for (i = 0; i < d->var_cnt; )
533 if (dict_class_from_id (var_get_name (d->var[i])) == DC_SCRATCH)
534 dict_delete_var (d, d->var[i]);
539 /* Moves V to 0-based position IDX in D. Other variables in D,
540 if any, retain their relative positions. Runs in time linear
541 in the distance moved. */
543 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
545 size_t old_index = var_get_dict_index (v);
547 assert (new_index < d->var_cnt);
548 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
549 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
552 /* Reorders the variables in D, placing the COUNT variables
553 listed in ORDER in that order at the beginning of D. The
554 other variables in D, if any, retain their relative
557 dict_reorder_vars (struct dictionary *d,
558 struct variable *const *order, size_t count)
560 struct variable **new_var;
564 assert (count == 0 || order != NULL);
565 assert (count <= d->var_cnt);
567 new_var = xnmalloc (d->var_cnt, sizeof *new_var);
568 memcpy (new_var, order, count * sizeof *new_var);
569 for (i = 0; i < count; i++)
571 size_t index = var_get_dict_index (order[i]);
572 assert (d->var[index] == order[i]);
573 d->var[index] = NULL;
574 set_var_dict_index (order[i], i);
576 for (i = 0; i < d->var_cnt; i++)
577 if (d->var[i] != NULL)
579 assert (count < d->var_cnt);
580 new_var[count] = d->var[i];
581 set_var_dict_index (new_var[count], count);
588 /* Changes the name of variable V in dictionary D to NEW_NAME. */
590 rename_var (struct dictionary *d, struct variable *v, const char *new_name)
592 struct vardict_info vdi;
594 assert (dict_contains_var (d, v));
596 vdi = *var_get_vardict (v);
597 var_clear_vardict (v);
598 var_set_name (v, new_name);
599 var_set_vardict (v, &vdi);
602 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
603 a variable named NEW_NAME is already in D, except that
604 NEW_NAME may be the same as V's existing name. */
606 dict_rename_var (struct dictionary *d, struct variable *v,
607 const char *new_name)
609 assert (!strcasecmp (var_get_name (v), new_name)
610 || dict_lookup_var (d, new_name) == NULL);
612 hsh_force_delete (d->name_tab, v);
613 rename_var (d, v, new_name);
614 hsh_force_insert (d->name_tab, v);
616 if (get_algorithm () == ENHANCED)
617 var_clear_short_name (v);
619 if ( d->callbacks && d->callbacks->var_changed )
620 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
623 /* Renames COUNT variables specified in VARS to the names given
624 in NEW_NAMES within dictionary D. If the renaming would
625 result in a duplicate variable name, returns false and stores a
626 name that would be duplicated into *ERR_NAME (if ERR_NAME is
627 non-null). Otherwise, the renaming is successful, and true
630 dict_rename_vars (struct dictionary *d,
631 struct variable **vars, char **new_names, size_t count,
638 assert (count == 0 || vars != NULL);
639 assert (count == 0 || new_names != NULL);
641 /* Save the names of the variables to be renamed. */
642 pool = pool_create ();
643 old_names = pool_nalloc (pool, count, sizeof *old_names);
644 for (i = 0; i < count; i++)
645 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
647 /* Remove the variables to be renamed from the name hash,
649 for (i = 0; i < count; i++)
651 hsh_force_delete (d->name_tab, vars[i]);
652 rename_var (d, vars[i], new_names[i]);
655 /* Add the renamed variables back into the name hash,
656 checking for conflicts. */
657 for (i = 0; i < count; i++)
658 if (hsh_insert (d->name_tab, vars[i]) != NULL)
660 /* There is a name conflict.
661 Back out all the name changes that have already
662 taken place, and indicate failure. */
664 if (err_name != NULL)
665 *err_name = new_names[i];
667 for (i = 0; i < fail_idx; i++)
668 hsh_force_delete (d->name_tab, vars[i]);
670 for (i = 0; i < count; i++)
672 rename_var (d, vars[i], old_names[i]);
673 hsh_force_insert (d->name_tab, vars[i]);
680 /* Clear short names. */
681 if (get_algorithm () == ENHANCED)
682 for (i = 0; i < count; i++)
683 var_clear_short_name (vars[i]);
689 /* Returns the weighting variable in dictionary D, or a null
690 pointer if the dictionary is unweighted. */
692 dict_get_weight (const struct dictionary *d)
695 assert (d->weight == NULL || dict_contains_var (d, d->weight));
700 /* Returns the value of D's weighting variable in case C, except that a
701 negative weight is returned as 0. Returns 1 if the dictionary is
702 unweighted. Will warn about missing, negative, or zero values if
703 warn_on_invalid is true. The function will set warn_on_invalid to false
704 if an invalid weight is found. */
706 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
707 bool *warn_on_invalid)
712 if (d->weight == NULL)
716 double w = case_num (c, d->weight);
717 if (w < 0.0 || var_is_num_missing (d->weight, w, MV_ANY))
719 if ( w == 0.0 && *warn_on_invalid ) {
720 *warn_on_invalid = false;
721 msg (SW, _("At least one case in the data file had a weight value "
722 "that was user-missing, system-missing, zero, or "
723 "negative. These case(s) were ignored."));
729 /* Sets the weighting variable of D to V, or turning off
730 weighting if V is a null pointer. */
732 dict_set_weight (struct dictionary *d, struct variable *v)
735 assert (v == NULL || dict_contains_var (d, v));
736 assert (v == NULL || var_is_numeric (v));
741 /* Returns the filter variable in dictionary D (see cmd_filter())
742 or a null pointer if the dictionary is unfiltered. */
744 dict_get_filter (const struct dictionary *d)
747 assert (d->filter == NULL || dict_contains_var (d, d->filter));
752 /* Sets V as the filter variable for dictionary D. Passing a
753 null pointer for V turn off filtering. */
755 dict_set_filter (struct dictionary *d, struct variable *v)
758 assert (v == NULL || dict_contains_var (d, v));
763 /* Returns the case limit for dictionary D, or zero if the number
764 of cases is unlimited. */
766 dict_get_case_limit (const struct dictionary *d)
770 return d->case_limit;
773 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
774 0 for CASE_LIMIT to indicate no limit. */
776 dict_set_case_limit (struct dictionary *d, size_t case_limit)
780 d->case_limit = case_limit;
783 /* Returns the case index of the next value to be added to D.
784 This value is the number of `union value's that need to be
785 allocated to store a case for dictionary D. */
787 dict_get_next_value_idx (const struct dictionary *d)
791 return d->next_value_idx;
794 /* Returns the number of bytes needed to store a case for
797 dict_get_case_size (const struct dictionary *d)
801 return sizeof (union value) * dict_get_next_value_idx (d);
804 /* Deletes scratch variables in dictionary D and reassigns values
805 so that fragmentation is eliminated. */
807 dict_compact_values (struct dictionary *d)
811 d->next_value_idx = 0;
812 for (i = 0; i < d->var_cnt; )
814 struct variable *v = d->var[i];
816 if (dict_class_from_id (var_get_name (v)) != DC_SCRATCH)
818 set_var_case_index (v, d->next_value_idx);
819 d->next_value_idx += var_get_value_cnt (v);
823 dict_delete_var (d, v);
827 /* Returns the number of values that would be used by a case if
828 dict_compact_values() were called. */
830 dict_get_compacted_value_cnt (const struct dictionary *d)
836 for (i = 0; i < d->var_cnt; i++)
837 if (dict_class_from_id (var_get_name (d->var[i])) != DC_SCRATCH)
838 cnt += var_get_value_cnt (d->var[i]);
842 /* Creates and returns an array mapping from a dictionary index
843 to the case index that the corresponding variable will have
844 after calling dict_compact_values(). Scratch variables
845 receive -1 for case index because dict_compact_values() will
848 dict_get_compacted_dict_index_to_case_index (const struct dictionary *d)
851 size_t next_value_idx;
854 map = xnmalloc (d->var_cnt, sizeof *map);
856 for (i = 0; i < d->var_cnt; i++)
858 struct variable *v = d->var[i];
860 if (dict_class_from_id (var_get_name (v)) != DC_SCRATCH)
862 map[i] = next_value_idx;
863 next_value_idx += var_get_value_cnt (v);
871 /* Returns true if a case for dictionary D would be smaller after
872 compacting, false otherwise. Compacting a case eliminates
873 "holes" between values and after the last value. Holes are
874 created by deleting variables (or by scratch variables).
876 The return value may differ from whether compacting a case
877 from dictionary D would *change* the case: compacting could
878 rearrange values even if it didn't reduce space
881 dict_compacting_would_shrink (const struct dictionary *d)
883 return dict_get_compacted_value_cnt (d) < dict_get_next_value_idx (d);
886 /* Returns true if a case for dictionary D would change after
887 compacting, false otherwise. Compacting a case eliminates
888 "holes" between values and after the last value. Holes are
889 created by deleting variables (or by scratch variables).
891 The return value may differ from whether compacting a case
892 from dictionary D would *shrink* the case: compacting could
893 rearrange values without reducing space requirements. */
895 dict_compacting_would_change (const struct dictionary *d)
901 for (i = 0; i < dict_get_var_cnt (d); i++)
903 struct variable *v = dict_get_var (d, i);
904 if (var_get_case_index (v) != case_idx)
906 case_idx += var_get_value_cnt (v);
911 /* How to copy a contiguous range of values between cases. */
914 size_t src_idx; /* Starting value index in source case. */
915 size_t dst_idx; /* Starting value index in target case. */
916 size_t cnt; /* Number of values. */
919 /* How to compact a case. */
920 struct dict_compactor
922 struct copy_map *maps; /* Array of mappings. */
923 size_t map_cnt; /* Number of mappings. */
926 /* Creates and returns a dict_compactor that can be used to
927 compact cases for dictionary D.
929 Compacting a case eliminates "holes" between values and after
930 the last value. Holes are created by deleting variables (or
931 by scratch variables). */
932 struct dict_compactor *
933 dict_make_compactor (const struct dictionary *d)
935 struct dict_compactor *compactor;
936 struct copy_map *map;
937 size_t map_allocated;
941 compactor = xmalloc (sizeof *compactor);
942 compactor->maps = NULL;
943 compactor->map_cnt = 0;
948 for (i = 0; i < d->var_cnt; i++)
950 struct variable *v = d->var[i];
952 if (dict_class_from_id (var_get_name (v)) == DC_SCRATCH)
954 if (map != NULL && map->src_idx + map->cnt == var_get_case_index (v))
955 map->cnt += var_get_value_cnt (v);
958 if (compactor->map_cnt == map_allocated)
959 compactor->maps = x2nrealloc (compactor->maps, &map_allocated,
960 sizeof *compactor->maps);
961 map = &compactor->maps[compactor->map_cnt++];
962 map->src_idx = var_get_case_index (v);
963 map->dst_idx = value_idx;
964 map->cnt = var_get_value_cnt (v);
966 value_idx += var_get_value_cnt (v);
972 /* Compacts SRC by copying it to DST according to the scheme in
975 Compacting a case eliminates "holes" between values and after
976 the last value. Holes are created by deleting variables (or
977 by scratch variables). */
979 dict_compactor_compact (const struct dict_compactor *compactor,
980 struct ccase *dst, const struct ccase *src)
984 for (i = 0; i < compactor->map_cnt; i++)
986 const struct copy_map *map = &compactor->maps[i];
987 case_copy (dst, map->dst_idx, src, map->src_idx, map->cnt);
991 /* Destroys COMPACTOR. */
993 dict_compactor_destroy (struct dict_compactor *compactor)
995 if (compactor != NULL)
997 free (compactor->maps);
1002 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
1003 dict_get_split_cnt() to determine how many SPLIT FILE vars
1004 there are. Returns a null pointer if and only if there are no
1006 struct variable *const *
1007 dict_get_split_vars (const struct dictionary *d)
1014 /* Returns the number of SPLIT FILE vars. */
1016 dict_get_split_cnt (const struct dictionary *d)
1020 return d->split_cnt;
1023 /* Sets CNT split vars SPLIT in dictionary D. */
1025 dict_set_split_vars (struct dictionary *d,
1026 struct variable *const *split, size_t cnt)
1029 assert (cnt == 0 || split != NULL);
1032 d->split = xnrealloc (d->split, cnt, sizeof *d->split);
1033 memcpy (d->split, split, cnt * sizeof *d->split);
1036 /* Returns the file label for D, or a null pointer if D is
1037 unlabeled (see cmd_file_label()). */
1039 dict_get_label (const struct dictionary *d)
1046 /* Sets D's file label to LABEL, truncating it to a maximum of 60
1049 dict_set_label (struct dictionary *d, const char *label)
1056 else if (strlen (label) < 60)
1057 d->label = xstrdup (label);
1060 d->label = xmalloc (61);
1061 memcpy (d->label, label, 60);
1062 d->label[60] = '\0';
1066 /* Returns the documents for D, or a null pointer if D has no
1067 documents (see cmd_document()).. */
1069 dict_get_documents (const struct dictionary *d)
1073 return d->documents;
1076 /* Sets the documents for D to DOCUMENTS, or removes D's
1077 documents if DOCUMENT is a null pointer. */
1079 dict_set_documents (struct dictionary *d, const char *documents)
1083 free (d->documents);
1084 if (documents == NULL)
1085 d->documents = NULL;
1087 d->documents = xstrdup (documents);
1090 /* Creates in D a vector named NAME that contains the CNT
1091 variables in VAR. Returns true if successful, or false if a
1092 vector named NAME already exists in D. */
1094 dict_create_vector (struct dictionary *d,
1096 struct variable **var, size_t cnt)
1100 assert (var != NULL);
1102 for (i = 0; i < cnt; i++)
1103 assert (dict_contains_var (d, var[i]));
1105 if (dict_lookup_vector (d, name) == NULL)
1107 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1108 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1115 /* Returns the vector in D with index IDX, which must be less
1116 than dict_get_vector_cnt (D). */
1117 const struct vector *
1118 dict_get_vector (const struct dictionary *d, size_t idx)
1121 assert (idx < d->vector_cnt);
1123 return d->vector[idx];
1126 /* Returns the number of vectors in D. */
1128 dict_get_vector_cnt (const struct dictionary *d)
1132 return d->vector_cnt;
1135 /* Looks up and returns the vector within D with the given
1137 const struct vector *
1138 dict_lookup_vector (const struct dictionary *d, const char *name)
1141 for (i = 0; i < d->vector_cnt; i++)
1142 if (!strcasecmp (vector_get_name (d->vector[i]), name))
1143 return d->vector[i];
1147 /* Deletes all vectors from D. */
1149 dict_clear_vectors (struct dictionary *d)
1153 for (i = 0; i < d->vector_cnt; i++)
1154 vector_destroy (d->vector[i]);
1161 /* Compares two strings. */
1163 compare_strings (const void *a, const void *b, const void *aux UNUSED)
1165 return strcmp (a, b);
1168 /* Hashes a string. */
1170 hash_string (const void *s, const void *aux UNUSED)
1172 return hsh_hash_string (s);
1176 /* Sets V's short name to BASE, followed by a suffix of the form
1177 _A, _B, _C, ..., _AA, _AB, etc. according to the value of
1178 SUFFIX_NUMBER. Truncates BASE as necessary to fit. */
1180 set_var_short_name_suffix (struct variable *v, const char *base,
1183 char suffix[SHORT_NAME_LEN + 1];
1184 char short_name[SHORT_NAME_LEN + 1];
1189 assert (suffix_number >= 0);
1191 /* Set base name. */
1192 var_set_short_name (v, base);
1194 /* Compose suffix. */
1195 start = end = suffix + sizeof suffix - 1;
1199 *--start = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"[suffix_number % 26];
1200 if (start <= suffix + 1)
1201 msg (SE, _("Variable suffix too large."));
1202 suffix_number /= 26;
1204 while (suffix_number > 0);
1207 /* Append suffix to V's short name. */
1208 str_copy_trunc (short_name, sizeof short_name, base);
1210 if (len + strlen (short_name) > SHORT_NAME_LEN)
1211 ofs = SHORT_NAME_LEN - len;
1213 ofs = strlen (short_name);
1214 strcpy (short_name + ofs, start);
1217 var_set_short_name (v, short_name);
1220 /* Assigns a valid, unique short_name[] to each variable in D.
1221 Each variable whose actual name is short has highest priority
1222 for that short name. Otherwise, variables with an existing
1223 short_name[] have the next highest priority for a given short
1224 name; if it is already taken, then the variable is treated as
1225 if short_name[] had been empty. Otherwise, long names are
1226 truncated to form short names. If that causes conflicts,
1227 variables are renamed as PREFIX_A, PREFIX_B, and so on. */
1229 dict_assign_short_names (struct dictionary *d)
1231 struct hsh_table *short_names;
1234 /* Give variables whose names are short the corresponding short
1235 names, and clear short_names[] that conflict with a variable
1237 for (i = 0; i < d->var_cnt; i++)
1239 struct variable *v = d->var[i];
1240 const char *short_name = var_get_short_name (v);
1241 if (strlen (var_get_name (v)) <= SHORT_NAME_LEN)
1242 var_set_short_name (v, var_get_name (v));
1243 else if (short_name != NULL && dict_lookup_var (d, short_name) != NULL)
1244 var_clear_short_name (v);
1247 /* Each variable with an assigned short_name[] now gets it
1248 unless there is a conflict. */
1249 short_names = hsh_create (d->var_cnt, compare_strings, hash_string,
1251 for (i = 0; i < d->var_cnt; i++)
1253 struct variable *v = d->var[i];
1254 const char *name = var_get_short_name (v);
1255 if (name != NULL && hsh_insert (short_names, (char *) name) != NULL)
1256 var_clear_short_name (v);
1259 /* Now assign short names to remaining variables. */
1260 for (i = 0; i < d->var_cnt; i++)
1262 struct variable *v = d->var[i];
1263 const char *name = var_get_short_name (v);
1266 /* Form initial short_name from the variable name, then
1267 try _A, _B, ... _AA, _AB, etc., if needed.*/
1272 var_set_short_name (v, var_get_name (v));
1274 set_var_short_name_suffix (v, var_get_name (v), trial - 1);
1278 while (hsh_insert (short_names, (char *) var_get_short_name (v))
1283 /* Get rid of hash table. */
1284 hsh_destroy (short_names);
1288 /* Called from variable.c to notify the dictionary that some property of
1289 the variable has changed */
1291 dict_var_changed (const struct variable *v)
1293 if ( var_has_vardict (v))
1295 const struct vardict_info *vdi = var_get_vardict (v);
1296 struct dictionary *d;
1300 if ( d->callbacks && d->callbacks->var_changed )
1301 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);