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. */
66 /* Creates and returns a new dictionary. */
70 struct dictionary *d = xmalloc (sizeof *d);
73 d->var_cnt = d->var_cap = 0;
74 d->name_tab = hsh_create (8, compare_vars_by_name, hash_var_by_name,
76 d->next_value_idx = 0;
90 /* Creates and returns a (deep) copy of an existing
93 dict_clone (const struct dictionary *s)
102 for (i = 0; i < s->var_cnt; i++)
104 struct variable *sv = s->var[i];
105 struct variable *dv = dict_clone_var_assert (d, sv, var_get_name (sv));
106 var_set_short_name (dv, var_get_short_name (sv));
109 d->next_value_idx = s->next_value_idx;
111 d->split_cnt = s->split_cnt;
112 if (d->split_cnt > 0)
114 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
115 for (i = 0; i < d->split_cnt; i++)
116 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
119 if (s->weight != NULL)
120 d->weight = dict_lookup_var_assert (d, var_get_name (s->weight));
122 if (s->filter != NULL)
123 d->filter = dict_lookup_var_assert (d, var_get_name (s->filter));
125 d->case_limit = s->case_limit;
126 dict_set_label (d, dict_get_label (s));
127 dict_set_documents (d, dict_get_documents (s));
129 d->vector_cnt = s->vector_cnt;
130 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
131 for (i = 0; i < s->vector_cnt; i++)
132 d->vector[i] = vector_clone (s->vector[i], s, d);
137 /* Clears the contents from a dictionary without destroying the
138 dictionary itself. */
140 dict_clear (struct dictionary *d)
142 /* FIXME? Should we really clear case_limit, label, documents?
143 Others are necessarily cleared by deleting all the variables.*/
148 for (i = 0; i < d->var_cnt; i++)
150 var_clear_vardict (d->var[i]);
151 var_destroy (d->var[i]);
155 d->var_cnt = d->var_cap = 0;
156 hsh_clear (d->name_tab);
157 d->next_value_idx = 0;
168 dict_clear_vectors (d);
171 /* Destroys the aux data for every variable in D, by calling
172 var_clear_aux() for each variable. */
174 dict_clear_aux (struct dictionary *d)
180 for (i = 0; i < d->var_cnt; i++)
181 var_clear_aux (d->var[i]);
184 /* Clears a dictionary and destroys it. */
186 dict_destroy (struct dictionary *d)
191 hsh_destroy (d->name_tab);
196 /* Returns the number of variables in D. */
198 dict_get_var_cnt (const struct dictionary *d)
205 /* Returns the variable in D with dictionary index IDX, which
206 must be between 0 and the count returned by
207 dict_get_var_cnt(), exclusive. */
209 dict_get_var (const struct dictionary *d, size_t idx)
212 assert (idx < d->var_cnt);
217 /* Sets *VARS to an array of pointers to variables in D and *CNT
218 to the number of variables in *D. All variables are returned
219 if EXCLUDE_CLASSES is 0, or it may contain one or more of (1u
220 << DC_ORDINARY), (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to
221 exclude the corresponding type of variable. */
223 dict_get_vars (const struct dictionary *d, struct variable ***vars,
224 size_t *cnt, unsigned exclude_classes)
230 assert (vars != NULL);
231 assert (cnt != NULL);
232 assert ((exclude_classes & ~((1u << DC_ORDINARY)
234 | (1u << DC_SCRATCH))) == 0);
237 for (i = 0; i < d->var_cnt; i++)
239 enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
240 if (!(exclude_classes & (1u << class)))
244 *vars = xnmalloc (count, sizeof **vars);
246 for (i = 0; i < d->var_cnt; i++)
248 enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
249 if (!(exclude_classes & (1u << class)))
250 (*vars)[(*cnt)++] = d->var[i];
252 assert (*cnt == count);
255 static struct variable *
256 add_var (struct dictionary *d, struct variable *v)
258 /* Add dictionary info to variable. */
259 struct vardict_info vdi;
260 vdi.case_index = d->next_value_idx;
261 vdi.dict_index = d->var_cnt;
262 var_set_vardict (v, &vdi);
264 /* Update dictionary. */
265 if (d->var_cnt >= d->var_cap)
267 d->var_cap = 8 + 2 * d->var_cap;
268 d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var);
270 d->var[d->var_cnt++] = v;
271 hsh_force_insert (d->name_tab, v);
273 d->next_value_idx += var_get_value_cnt (v);
278 /* Creates and returns a new variable in D with the given NAME
279 and WIDTH. Returns a null pointer if the given NAME would
280 duplicate that of an existing variable in the dictionary. */
282 dict_create_var (struct dictionary *d, const char *name, int width)
284 return (dict_lookup_var (d, name) == NULL
285 ? dict_create_var_assert (d, name, width)
289 /* Creates and returns a new variable in D with the given NAME
290 and WIDTH. Assert-fails if the given NAME would duplicate
291 that of an existing variable in the dictionary. */
293 dict_create_var_assert (struct dictionary *d, const char *name, int width)
295 assert (dict_lookup_var (d, name) == NULL);
296 return add_var (d, var_create (name, width));
299 /* Creates and returns a new variable in D with name NAME, as a
300 copy of existing variable OLD_VAR, which need not be in D or
301 in any dictionary. Returns a null pointer if the given NAME
302 would duplicate that of an existing variable in the
305 dict_clone_var (struct dictionary *d, const struct variable *old_var,
308 return (dict_lookup_var (d, name) == NULL
309 ? dict_clone_var_assert (d, old_var, name)
313 /* Creates and returns a new variable in D with name NAME, as a
314 copy of existing variable OLD_VAR, which need not be in D or
315 in any dictionary. Assert-fails if the given NAME would
316 duplicate that of an existing variable in the dictionary. */
318 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var,
321 struct variable *new_var = var_clone (old_var);
322 assert (dict_lookup_var (d, name) == NULL);
323 var_set_name (new_var, name);
324 return add_var (d, new_var);
327 /* Returns the variable named NAME in D, or a null pointer if no
328 variable has that name. */
330 dict_lookup_var (const struct dictionary *d, const char *name)
332 struct variable *target = var_create (name, 0);
333 struct variable *result = hsh_find (d->name_tab, target);
334 var_destroy (target);
338 /* Returns the variable named NAME in D. Assert-fails if no
339 variable has that name. */
341 dict_lookup_var_assert (const struct dictionary *d, const char *name)
343 struct variable *v = dict_lookup_var (d, name);
348 /* Returns true if variable V is in dictionary D,
351 dict_contains_var (const struct dictionary *d, const struct variable *v)
353 if (var_has_vardict (v))
355 const struct vardict_info *vdi = var_get_vardict (v);
356 return (vdi->dict_index >= 0
357 && vdi->dict_index < d->var_cnt
358 && d->var[vdi->dict_index] == v);
364 /* Compares two double pointers to variables, which should point
365 to elements of a struct dictionary's `var' member array. */
367 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
369 struct variable *const *a = a_;
370 struct variable *const *b = b_;
372 return *a < *b ? -1 : *a > *b;
375 /* Sets the dict_index in V's vardict to DICT_INDEX. */
377 set_var_dict_index (struct variable *v, int dict_index)
379 struct vardict_info vdi = *var_get_vardict (v);
380 vdi.dict_index = dict_index;
381 var_set_vardict (v, &vdi);
384 /* Sets the case_index in V's vardict to DICT_INDEX. */
386 set_var_case_index (struct variable *v, int case_index)
388 struct vardict_info vdi = *var_get_vardict (v);
389 vdi.case_index = case_index;
390 var_set_vardict (v, &vdi);
393 /* Re-sets the dict_index in the dictionary variables with
394 indexes from FROM to TO (exclusive). */
396 reindex_vars (struct dictionary *d, size_t from, size_t to)
400 for (i = from; i < to; i++)
401 set_var_dict_index (d->var[i], i);
404 /* Deletes variable V from dictionary D and frees V.
406 This is a very bad idea if there might be any pointers to V
407 from outside D. In general, no variable in should be deleted when
408 any transformations are active on the dictionary's dataset, because
409 those transformations might reference the deleted variable.
410 The safest time to delete a variable is just after a procedure
411 has been executed, as done by MODIFY VARS.
413 Pointers to V within D are not a problem, because
414 dict_delete_var() knows to remove V from split variables,
415 weights, filters, etc. */
417 dict_delete_var (struct dictionary *d, struct variable *v)
419 int dict_index = var_get_dict_index (v);
421 assert (dict_contains_var (d, v));
423 /* Delete aux data. */
426 /* Remove V from splits, weight, filter variables. */
427 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
428 &v, compare_var_ptrs, NULL);
433 dict_clear_vectors (d);
435 /* Remove V from var array. */
436 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
439 /* Update dict_index for each affected variable. */
440 reindex_vars (d, dict_index, d->var_cnt);
442 /* Update name hash. */
443 hsh_force_delete (d->name_tab, v);
446 var_clear_vardict (v);
450 /* Deletes the COUNT variables listed in VARS from D. This is
451 unsafe; see the comment on dict_delete_var() for details. */
453 dict_delete_vars (struct dictionary *d,
454 struct variable *const *vars, size_t count)
456 /* FIXME: this can be done in O(count) time, but this algorithm
459 assert (count == 0 || vars != NULL);
462 dict_delete_var (d, *vars++);
465 /* Deletes the COUNT variables in D starting at index IDX. This
466 is unsafe; see the comment on dict_delete_var() for
469 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
471 /* FIXME: this can be done in O(count) time, but this algorithm
473 assert (idx + count <= d->var_cnt);
476 dict_delete_var (d, d->var[idx]);
479 /* Deletes scratch variables from dictionary D. */
481 dict_delete_scratch_vars (struct dictionary *d)
485 /* FIXME: this can be done in O(count) time, but this algorithm
489 for (i = 0; i < d->var_cnt; )
490 if (dict_class_from_id (var_get_name (d->var[i])) == DC_SCRATCH)
491 dict_delete_var (d, d->var[i]);
496 /* Moves V to 0-based position IDX in D. Other variables in D,
497 if any, retain their relative positions. Runs in time linear
498 in the distance moved. */
500 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
502 size_t old_index = var_get_dict_index (v);
504 assert (new_index < d->var_cnt);
505 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
506 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
509 /* Reorders the variables in D, placing the COUNT variables
510 listed in ORDER in that order at the beginning of D. The
511 other variables in D, if any, retain their relative
514 dict_reorder_vars (struct dictionary *d,
515 struct variable *const *order, size_t count)
517 struct variable **new_var;
521 assert (count == 0 || order != NULL);
522 assert (count <= d->var_cnt);
524 new_var = xnmalloc (d->var_cnt, sizeof *new_var);
525 memcpy (new_var, order, count * sizeof *new_var);
526 for (i = 0; i < count; i++)
528 size_t index = var_get_dict_index (order[i]);
529 assert (d->var[index] == order[i]);
530 d->var[index] = NULL;
531 set_var_dict_index (order[i], i);
533 for (i = 0; i < d->var_cnt; i++)
534 if (d->var[i] != NULL)
536 assert (count < d->var_cnt);
537 new_var[count] = d->var[i];
538 set_var_dict_index (new_var[count], count);
545 /* Changes the name of variable V in dictionary D to NEW_NAME. */
547 rename_var (struct dictionary *d, struct variable *v, const char *new_name)
549 struct vardict_info vdi;
551 assert (dict_contains_var (d, v));
553 vdi = *var_get_vardict (v);
554 var_clear_vardict (v);
555 var_set_name (v, new_name);
556 var_set_vardict (v, &vdi);
559 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
560 a variable named NEW_NAME is already in D, except that
561 NEW_NAME may be the same as V's existing name. */
563 dict_rename_var (struct dictionary *d, struct variable *v,
564 const char *new_name)
566 assert (!strcasecmp (var_get_name (v), new_name)
567 || dict_lookup_var (d, new_name) == NULL);
569 hsh_force_delete (d->name_tab, v);
570 rename_var (d, v, new_name);
571 hsh_force_insert (d->name_tab, v);
573 if (get_algorithm () == ENHANCED)
574 var_clear_short_name (v);
577 /* Renames COUNT variables specified in VARS to the names given
578 in NEW_NAMES within dictionary D. If the renaming would
579 result in a duplicate variable name, returns false and stores a
580 name that would be duplicated into *ERR_NAME (if ERR_NAME is
581 non-null). Otherwise, the renaming is successful, and true
584 dict_rename_vars (struct dictionary *d,
585 struct variable **vars, char **new_names, size_t count,
592 assert (count == 0 || vars != NULL);
593 assert (count == 0 || new_names != NULL);
595 /* Save the names of the variables to be renamed. */
596 pool = pool_create ();
597 old_names = pool_nalloc (pool, count, sizeof *old_names);
598 for (i = 0; i < count; i++)
599 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
601 /* Remove the variables to be renamed from the name hash,
603 for (i = 0; i < count; i++)
605 hsh_force_delete (d->name_tab, vars[i]);
606 rename_var (d, vars[i], new_names[i]);
609 /* Add the renamed variables back into the name hash,
610 checking for conflicts. */
611 for (i = 0; i < count; i++)
612 if (hsh_insert (d->name_tab, vars[i]) != NULL)
614 /* There is a name conflict.
615 Back out all the name changes that have already
616 taken place, and indicate failure. */
618 if (err_name != NULL)
619 *err_name = new_names[i];
621 for (i = 0; i < fail_idx; i++)
622 hsh_force_delete (d->name_tab, vars[i]);
624 for (i = 0; i < count; i++)
626 rename_var (d, vars[i], old_names[i]);
627 hsh_force_insert (d->name_tab, vars[i]);
634 /* Clear short names. */
635 if (get_algorithm () == ENHANCED)
636 for (i = 0; i < count; i++)
637 var_clear_short_name (vars[i]);
643 /* Returns the weighting variable in dictionary D, or a null
644 pointer if the dictionary is unweighted. */
646 dict_get_weight (const struct dictionary *d)
649 assert (d->weight == NULL || dict_contains_var (d, d->weight));
654 /* Returns the value of D's weighting variable in case C, except that a
655 negative weight is returned as 0. Returns 1 if the dictionary is
656 unweighted. Will warn about missing, negative, or zero values if
657 warn_on_invalid is true. The function will set warn_on_invalid to false
658 if an invalid weight is found. */
660 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
661 bool *warn_on_invalid)
666 if (d->weight == NULL)
670 double w = case_num (c, d->weight);
671 if (w < 0.0 || var_is_num_missing (d->weight, w))
673 if ( w == 0.0 && *warn_on_invalid ) {
674 *warn_on_invalid = false;
675 msg (SW, _("At least one case in the data file had a weight value "
676 "that was user-missing, system-missing, zero, or "
677 "negative. These case(s) were ignored."));
683 /* Sets the weighting variable of D to V, or turning off
684 weighting if V is a null pointer. */
686 dict_set_weight (struct dictionary *d, struct variable *v)
689 assert (v == NULL || dict_contains_var (d, v));
690 assert (v == NULL || var_is_numeric (v));
695 /* Returns the filter variable in dictionary D (see cmd_filter())
696 or a null pointer if the dictionary is unfiltered. */
698 dict_get_filter (const struct dictionary *d)
701 assert (d->filter == NULL || dict_contains_var (d, d->filter));
706 /* Sets V as the filter variable for dictionary D. Passing a
707 null pointer for V turn off filtering. */
709 dict_set_filter (struct dictionary *d, struct variable *v)
712 assert (v == NULL || dict_contains_var (d, v));
717 /* Returns the case limit for dictionary D, or zero if the number
718 of cases is unlimited. */
720 dict_get_case_limit (const struct dictionary *d)
724 return d->case_limit;
727 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
728 0 for CASE_LIMIT to indicate no limit. */
730 dict_set_case_limit (struct dictionary *d, size_t case_limit)
734 d->case_limit = case_limit;
737 /* Returns the case index of the next value to be added to D.
738 This value is the number of `union value's that need to be
739 allocated to store a case for dictionary D. */
741 dict_get_next_value_idx (const struct dictionary *d)
745 return d->next_value_idx;
748 /* Returns the number of bytes needed to store a case for
751 dict_get_case_size (const struct dictionary *d)
755 return sizeof (union value) * dict_get_next_value_idx (d);
758 /* Deletes scratch variables in dictionary D and reassigns values
759 so that fragmentation is eliminated. */
761 dict_compact_values (struct dictionary *d)
765 d->next_value_idx = 0;
766 for (i = 0; i < d->var_cnt; )
768 struct variable *v = d->var[i];
770 if (dict_class_from_id (var_get_name (v)) != DC_SCRATCH)
772 set_var_case_index (v, d->next_value_idx);
773 d->next_value_idx += var_get_value_cnt (v);
777 dict_delete_var (d, v);
781 /* Returns the number of values that would be used by a case if
782 dict_compact_values() were called. */
784 dict_get_compacted_value_cnt (const struct dictionary *d)
790 for (i = 0; i < d->var_cnt; i++)
791 if (dict_class_from_id (var_get_name (d->var[i])) != DC_SCRATCH)
792 cnt += var_get_value_cnt (d->var[i]);
796 /* Creates and returns an array mapping from a dictionary index
797 to the case index that the corresponding variable will have
798 after calling dict_compact_values(). Scratch variables
799 receive -1 for case index because dict_compact_values() will
802 dict_get_compacted_dict_index_to_case_index (const struct dictionary *d)
805 size_t next_value_idx;
808 map = xnmalloc (d->var_cnt, sizeof *map);
810 for (i = 0; i < d->var_cnt; i++)
812 struct variable *v = d->var[i];
814 if (dict_class_from_id (var_get_name (v)) != DC_SCRATCH)
816 map[i] = next_value_idx;
817 next_value_idx += var_get_value_cnt (v);
825 /* Returns true if a case for dictionary D would be smaller after
826 compacting, false otherwise. Compacting a case eliminates
827 "holes" between values and after the last value. Holes are
828 created by deleting variables (or by scratch variables).
830 The return value may differ from whether compacting a case
831 from dictionary D would *change* the case: compacting could
832 rearrange values even if it didn't reduce space
835 dict_compacting_would_shrink (const struct dictionary *d)
837 return dict_get_compacted_value_cnt (d) < dict_get_next_value_idx (d);
840 /* Returns true if a case for dictionary D would change after
841 compacting, false otherwise. Compacting a case eliminates
842 "holes" between values and after the last value. Holes are
843 created by deleting variables (or by scratch variables).
845 The return value may differ from whether compacting a case
846 from dictionary D would *shrink* the case: compacting could
847 rearrange values without reducing space requirements. */
849 dict_compacting_would_change (const struct dictionary *d)
855 for (i = 0; i < dict_get_var_cnt (d); i++)
857 struct variable *v = dict_get_var (d, i);
858 if (var_get_case_index (v) != case_idx)
860 case_idx += var_get_value_cnt (v);
865 /* How to copy a contiguous range of values between cases. */
868 size_t src_idx; /* Starting value index in source case. */
869 size_t dst_idx; /* Starting value index in target case. */
870 size_t cnt; /* Number of values. */
873 /* How to compact a case. */
874 struct dict_compactor
876 struct copy_map *maps; /* Array of mappings. */
877 size_t map_cnt; /* Number of mappings. */
880 /* Creates and returns a dict_compactor that can be used to
881 compact cases for dictionary D.
883 Compacting a case eliminates "holes" between values and after
884 the last value. Holes are created by deleting variables (or
885 by scratch variables). */
886 struct dict_compactor *
887 dict_make_compactor (const struct dictionary *d)
889 struct dict_compactor *compactor;
890 struct copy_map *map;
891 size_t map_allocated;
895 compactor = xmalloc (sizeof *compactor);
896 compactor->maps = NULL;
897 compactor->map_cnt = 0;
902 for (i = 0; i < d->var_cnt; i++)
904 struct variable *v = d->var[i];
906 if (dict_class_from_id (var_get_name (v)) == DC_SCRATCH)
908 if (map != NULL && map->src_idx + map->cnt == var_get_case_index (v))
909 map->cnt += var_get_value_cnt (v);
912 if (compactor->map_cnt == map_allocated)
913 compactor->maps = x2nrealloc (compactor->maps, &map_allocated,
914 sizeof *compactor->maps);
915 map = &compactor->maps[compactor->map_cnt++];
916 map->src_idx = var_get_case_index (v);
917 map->dst_idx = value_idx;
918 map->cnt = var_get_value_cnt (v);
920 value_idx += var_get_value_cnt (v);
926 /* Compacts SRC by copying it to DST according to the scheme in
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). */
933 dict_compactor_compact (const struct dict_compactor *compactor,
934 struct ccase *dst, const struct ccase *src)
938 for (i = 0; i < compactor->map_cnt; i++)
940 const struct copy_map *map = &compactor->maps[i];
941 case_copy (dst, map->dst_idx, src, map->src_idx, map->cnt);
945 /* Destroys COMPACTOR. */
947 dict_compactor_destroy (struct dict_compactor *compactor)
949 if (compactor != NULL)
951 free (compactor->maps);
956 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
957 dict_get_split_cnt() to determine how many SPLIT FILE vars
958 there are. Returns a null pointer if and only if there are no
960 struct variable *const *
961 dict_get_split_vars (const struct dictionary *d)
968 /* Returns the number of SPLIT FILE vars. */
970 dict_get_split_cnt (const struct dictionary *d)
977 /* Sets CNT split vars SPLIT in dictionary D. */
979 dict_set_split_vars (struct dictionary *d,
980 struct variable *const *split, size_t cnt)
983 assert (cnt == 0 || split != NULL);
986 d->split = xnrealloc (d->split, cnt, sizeof *d->split);
987 memcpy (d->split, split, cnt * sizeof *d->split);
990 /* Returns the file label for D, or a null pointer if D is
991 unlabeled (see cmd_file_label()). */
993 dict_get_label (const struct dictionary *d)
1000 /* Sets D's file label to LABEL, truncating it to a maximum of 60
1003 dict_set_label (struct dictionary *d, const char *label)
1010 else if (strlen (label) < 60)
1011 d->label = xstrdup (label);
1014 d->label = xmalloc (61);
1015 memcpy (d->label, label, 60);
1016 d->label[60] = '\0';
1020 /* Returns the documents for D, or a null pointer if D has no
1021 documents (see cmd_document()).. */
1023 dict_get_documents (const struct dictionary *d)
1027 return d->documents;
1030 /* Sets the documents for D to DOCUMENTS, or removes D's
1031 documents if DOCUMENT is a null pointer. */
1033 dict_set_documents (struct dictionary *d, const char *documents)
1037 free (d->documents);
1038 if (documents == NULL)
1039 d->documents = NULL;
1041 d->documents = xstrdup (documents);
1044 /* Creates in D a vector named NAME that contains the CNT
1045 variables in VAR. Returns true if successful, or false if a
1046 vector named NAME already exists in D. */
1048 dict_create_vector (struct dictionary *d,
1050 struct variable **var, size_t cnt)
1054 assert (var != NULL);
1056 for (i = 0; i < cnt; i++)
1057 assert (dict_contains_var (d, var[i]));
1059 if (dict_lookup_vector (d, name) == NULL)
1061 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1062 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1069 /* Returns the vector in D with index IDX, which must be less
1070 than dict_get_vector_cnt (D). */
1071 const struct vector *
1072 dict_get_vector (const struct dictionary *d, size_t idx)
1075 assert (idx < d->vector_cnt);
1077 return d->vector[idx];
1080 /* Returns the number of vectors in D. */
1082 dict_get_vector_cnt (const struct dictionary *d)
1086 return d->vector_cnt;
1089 /* Looks up and returns the vector within D with the given
1091 const struct vector *
1092 dict_lookup_vector (const struct dictionary *d, const char *name)
1095 for (i = 0; i < d->vector_cnt; i++)
1096 if (!strcasecmp (vector_get_name (d->vector[i]), name))
1097 return d->vector[i];
1101 /* Deletes all vectors from D. */
1103 dict_clear_vectors (struct dictionary *d)
1107 for (i = 0; i < d->vector_cnt; i++)
1108 vector_destroy (d->vector[i]);
1115 /* Compares two strings. */
1117 compare_strings (const void *a, const void *b, const void *aux UNUSED)
1119 return strcmp (a, b);
1122 /* Hashes a string. */
1124 hash_string (const void *s, const void *aux UNUSED)
1126 return hsh_hash_string (s);
1130 /* Sets V's short name to BASE, followed by a suffix of the form
1131 _A, _B, _C, ..., _AA, _AB, etc. according to the value of
1132 SUFFIX_NUMBER. Truncates BASE as necessary to fit. */
1134 set_var_short_name_suffix (struct variable *v, const char *base,
1137 char suffix[SHORT_NAME_LEN + 1];
1138 char short_name[SHORT_NAME_LEN + 1];
1143 assert (suffix_number >= 0);
1145 /* Set base name. */
1146 var_set_short_name (v, base);
1148 /* Compose suffix. */
1149 start = end = suffix + sizeof suffix - 1;
1153 *--start = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"[suffix_number % 26];
1154 if (start <= suffix + 1)
1155 msg (SE, _("Variable suffix too large."));
1156 suffix_number /= 26;
1158 while (suffix_number > 0);
1161 /* Append suffix to V's short name. */
1162 str_copy_trunc (short_name, sizeof short_name, base);
1164 if (len + strlen (short_name) > SHORT_NAME_LEN)
1165 ofs = SHORT_NAME_LEN - len;
1167 ofs = strlen (short_name);
1168 strcpy (short_name + ofs, start);
1171 var_set_short_name (v, short_name);
1174 /* Assigns a valid, unique short_name[] to each variable in D.
1175 Each variable whose actual name is short has highest priority
1176 for that short name. Otherwise, variables with an existing
1177 short_name[] have the next highest priority for a given short
1178 name; if it is already taken, then the variable is treated as
1179 if short_name[] had been empty. Otherwise, long names are
1180 truncated to form short names. If that causes conflicts,
1181 variables are renamed as PREFIX_A, PREFIX_B, and so on. */
1183 dict_assign_short_names (struct dictionary *d)
1185 struct hsh_table *short_names;
1188 /* Give variables whose names are short the corresponding short
1189 names, and clear short_names[] that conflict with a variable
1191 for (i = 0; i < d->var_cnt; i++)
1193 struct variable *v = d->var[i];
1194 const char *short_name = var_get_short_name (v);
1195 if (strlen (var_get_name (v)) <= SHORT_NAME_LEN)
1196 var_set_short_name (v, var_get_name (v));
1197 else if (short_name != NULL && dict_lookup_var (d, short_name) != NULL)
1198 var_clear_short_name (v);
1201 /* Each variable with an assigned short_name[] now gets it
1202 unless there is a conflict. */
1203 short_names = hsh_create (d->var_cnt, compare_strings, hash_string,
1205 for (i = 0; i < d->var_cnt; i++)
1207 struct variable *v = d->var[i];
1208 const char *name = var_get_short_name (v);
1209 if (name != NULL && hsh_insert (short_names, (char *) name) != NULL)
1210 var_clear_short_name (v);
1213 /* Now assign short names to remaining variables. */
1214 for (i = 0; i < d->var_cnt; i++)
1216 struct variable *v = d->var[i];
1217 const char *name = var_get_short_name (v);
1220 /* Form initial short_name from the variable name, then
1221 try _A, _B, ... _AA, _AB, etc., if needed.*/
1226 var_set_short_name (v, var_get_name (v));
1228 set_var_short_name_suffix (v, var_get_name (v), trial - 1);
1232 while (hsh_insert (short_names, (char *) var_get_short_name (v))
1237 /* Get rid of hash table. */
1238 hsh_destroy (short_names);