1 /* PSPP - computes sample statistics.
2 Copyright (C) 1997-9, 2000, 2006 Free Software Foundation, Inc.
3 Written by Ben Pfaff <blp@gnu.org>.
5 This program is free software; you can redistribute it and/or
6 modify it under the terms of the GNU General Public License as
7 published by the Free Software Foundation; either version 2 of the
8 License, or (at your option) any later version.
10 This program is distributed in the hope that it will be useful, but
11 WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 General Public License for more details.
15 You should have received a copy of the GNU General Public License
16 along with this program; if not, write to the Free Software
17 Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
21 #include "dictionary.h"
24 #include "algorithm.h"
28 #include "cat-routines.h"
34 #include "value-labels.h"
38 #define _(msgid) gettext (msgid)
43 struct variable **var; /* Variables. */
44 size_t var_cnt, var_cap; /* Number of variables, capacity. */
45 struct hsh_table *name_tab; /* Variable index by name. */
46 int next_value_idx; /* Index of next `union value' to allocate. */
47 struct variable **split; /* SPLIT FILE vars. */
48 size_t split_cnt; /* SPLIT FILE count. */
49 struct variable *weight; /* WEIGHT variable. */
50 struct variable *filter; /* FILTER variable. */
51 int case_limit; /* Current case limit (N command). */
52 char *label; /* File label. */
53 char *documents; /* Documents, as a string. */
54 struct vector **vector; /* Vectors of variables. */
55 size_t vector_cnt; /* Number of vectors. */
58 /* Creates and returns a new dictionary. */
62 struct dictionary *d = xmalloc (sizeof *d);
65 d->var_cnt = d->var_cap = 0;
66 d->name_tab = hsh_create (8, compare_var_names, hash_var_name, NULL, NULL);
67 d->next_value_idx = 0;
81 /* Creates and returns a (deep) copy of an existing
84 dict_clone (const struct dictionary *s)
93 for (i = 0; i < s->var_cnt; i++)
95 struct variable *sv = s->var[i];
96 struct variable *dv = dict_clone_var_assert (d, sv, sv->name);
97 var_set_short_name (dv, sv->short_name);
100 d->next_value_idx = s->next_value_idx;
102 d->split_cnt = s->split_cnt;
103 if (d->split_cnt > 0)
105 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
106 for (i = 0; i < d->split_cnt; i++)
107 d->split[i] = dict_lookup_var_assert (d, s->split[i]->name);
110 if (s->weight != NULL)
111 d->weight = dict_lookup_var_assert (d, s->weight->name);
113 if (s->filter != NULL)
114 d->filter = dict_lookup_var_assert (d, s->filter->name);
116 d->case_limit = s->case_limit;
117 dict_set_label (d, dict_get_label (s));
118 dict_set_documents (d, dict_get_documents (s));
120 d->vector_cnt = s->vector_cnt;
121 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
122 for (i = 0; i < s->vector_cnt; i++)
124 struct vector *sv = s->vector[i];
125 struct vector *dv = d->vector[i] = xmalloc (sizeof *dv);
129 strcpy (dv->name, sv->name);
131 dv->var = xnmalloc (dv->cnt, sizeof *dv->var);
132 for (j = 0; j < dv->cnt; j++)
133 dv->var[j] = d->var[sv->var[j]->index];
139 /* Clears the contents from a dictionary without destroying the
140 dictionary itself. */
142 dict_clear (struct dictionary *d)
144 /* FIXME? Should we really clear case_limit, label, documents?
145 Others are necessarily cleared by deleting all the variables.*/
150 for (i = 0; i < d->var_cnt; i++)
152 struct variable *v = d->var[i];
154 val_labs_destroy (v->val_labs);
160 d->var_cnt = d->var_cap = 0;
161 hsh_clear (d->name_tab);
162 d->next_value_idx = 0;
173 dict_clear_vectors (d);
176 /* Destroys the aux data for every variable in D, by calling
177 var_clear_aux() for each variable. */
179 dict_clear_aux (struct dictionary *d)
185 for (i = 0; i < d->var_cnt; i++)
186 var_clear_aux (d->var[i]);
189 /* Clears a dictionary and destroys it. */
191 dict_destroy (struct dictionary *d)
196 hsh_destroy (d->name_tab);
201 /* Returns the number of variables in D. */
203 dict_get_var_cnt (const struct dictionary *d)
210 /* Returns the variable in D with index IDX, which must be
211 between 0 and the count returned by dict_get_var_cnt(),
214 dict_get_var (const struct dictionary *d, size_t idx)
217 assert (idx < d->var_cnt);
222 /* Sets *VARS to an array of pointers to variables in D and *CNT
223 to the number of variables in *D. By default all variables
224 are returned, but bits may be set in EXCLUDE_CLASSES to
225 exclude ordinary, system, and/or scratch variables. */
227 dict_get_vars (const struct dictionary *d, struct variable ***vars,
228 size_t *cnt, unsigned exclude_classes)
234 assert (vars != NULL);
235 assert (cnt != NULL);
236 assert ((exclude_classes & ~((1u << DC_ORDINARY)
238 | (1u << DC_SCRATCH))) == 0);
241 for (i = 0; i < d->var_cnt; i++)
242 if (!(exclude_classes & (1u << dict_class_from_id (d->var[i]->name))))
245 *vars = xnmalloc (count, sizeof **vars);
247 for (i = 0; i < d->var_cnt; i++)
248 if (!(exclude_classes & (1u << dict_class_from_id (d->var[i]->name))))
249 (*vars)[(*cnt)++] = d->var[i];
250 assert (*cnt == count);
254 /* Creates and returns a new variable in D with the given NAME
255 and WIDTH. Returns a null pointer if the given NAME would
256 duplicate that of an existing variable in the dictionary. */
258 dict_create_var (struct dictionary *d, const char *name, int width)
263 assert (name != NULL);
265 assert (width >= 0 && width < 256);
267 assert (var_is_valid_name(name,0));
269 /* Make sure there's not already a variable by that name. */
270 if (dict_lookup_var (d, name) != NULL)
273 /* Allocate and initialize variable. */
274 v = xmalloc (sizeof *v);
275 str_copy_trunc (v->name, sizeof v->name, name);
276 v->type = width == 0 ? NUMERIC : ALPHA;
278 v->fv = d->next_value_idx;
279 v->nv = width == 0 ? 1 : DIV_RND_UP (width, 8);
281 v->reinit = dict_class_from_id (v->name) != DC_SCRATCH;
282 v->index = d->var_cnt;
283 mv_init (&v->miss, width);
284 if (v->type == NUMERIC)
287 v->alignment = ALIGN_RIGHT;
288 v->display_width = 8;
289 v->measure = MEASURE_SCALE;
293 v->print = make_output_format (FMT_A, v->width, 0);
294 v->alignment = ALIGN_LEFT;
295 v->display_width = 8;
296 v->measure = MEASURE_NOMINAL;
299 v->val_labs = val_labs_create (v->width);
301 var_clear_short_name (v);
306 /* Update dictionary. */
307 if (d->var_cnt >= d->var_cap)
309 d->var_cap = 8 + 2 * d->var_cap;
310 d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var);
312 d->var[v->index] = v;
314 hsh_force_insert (d->name_tab, v);
316 d->next_value_idx += v->nv;
321 /* Creates and returns a new variable in D with the given NAME
322 and WIDTH. Assert-fails if the given NAME would duplicate
323 that of an existing variable in the dictionary. */
325 dict_create_var_assert (struct dictionary *d, const char *name, int width)
327 struct variable *v = dict_create_var (d, name, width);
332 /* Creates and returns a new variable in D with name NAME, as a
333 copy of existing variable OV, which need not be in D or in any
334 dictionary. Returns a null pointer if the given NAME would
335 duplicate that of an existing variable in the dictionary. */
337 dict_clone_var (struct dictionary *d, const struct variable *ov,
344 assert (name != NULL);
346 assert (strlen (name) >= 1);
347 assert (strlen (name) <= LONG_NAME_LEN);
349 nv = dict_create_var (d, name, ov->width);
353 /* Copy most members not copied via dict_create_var().
354 short_name[] is intentionally not copied, because there is
355 no reason to give a new variable with potentially a new name
356 the same short name. */
358 nv->reinit = ov->reinit;
359 mv_copy (&nv->miss, &ov->miss);
360 nv->print = ov->print;
361 nv->write = ov->write;
362 val_labs_destroy (nv->val_labs);
363 nv->val_labs = val_labs_copy (ov->val_labs);
364 if (ov->label != NULL)
365 nv->label = xstrdup (ov->label);
366 nv->measure = ov->measure;
367 nv->display_width = ov->display_width;
368 nv->alignment = ov->alignment;
373 /* Creates and returns a new variable in D with name NAME, as a
374 copy of existing variable OV, which need not be in D or in any
375 dictionary. Assert-fails if the given NAME would duplicate
376 that of an existing variable in the dictionary. */
378 dict_clone_var_assert (struct dictionary *d, const struct variable *ov,
381 struct variable *v = dict_clone_var (d, ov, name);
386 /* Returns the variable named NAME in D, or a null pointer if no
387 variable has that name. */
389 dict_lookup_var (const struct dictionary *d, const char *name)
394 assert (name != NULL);
396 str_copy_trunc (v.name, sizeof v.name, name);
397 return hsh_find (d->name_tab, &v);
400 /* Returns the variable named NAME in D. Assert-fails if no
401 variable has that name. */
403 dict_lookup_var_assert (const struct dictionary *d, const char *name)
405 struct variable *v = dict_lookup_var (d, name);
410 /* Returns true if variable V is in dictionary D,
413 dict_contains_var (const struct dictionary *d, const struct variable *v)
418 return v->index >= 0 && v->index < d->var_cnt && d->var[v->index] == v;
421 /* Compares two double pointers to variables, which should point
422 to elements of a struct dictionary's `var' member array. */
424 compare_var_ptrs (const void *a_, const void *b_, void *aux UNUSED)
426 struct variable *const *a = a_;
427 struct variable *const *b = b_;
429 return *a < *b ? -1 : *a > *b;
432 /* Deletes variable V from dictionary D and frees V.
434 This is a very bad idea if there might be any pointers to V
435 from outside D. In general, no variable in default_dict
436 should be deleted when any transformations are active, because
437 those transformations might reference the deleted variable.
438 The safest time to delete a variable is just after a procedure
439 has been executed, as done by MODIFY VARS.
441 Pointers to V within D are not a problem, because
442 dict_delete_var() knows to remove V from split variables,
443 weights, filters, etc. */
445 dict_delete_var (struct dictionary *d, struct variable *v)
451 assert (dict_contains_var (d, v));
453 /* Delete aux data. */
456 /* Remove V from splits, weight, filter variables. */
457 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
458 &v, compare_var_ptrs, NULL);
463 dict_clear_vectors (d);
465 /* Remove V from var array. */
466 remove_element (d->var, d->var_cnt, sizeof *d->var, v->index);
470 for (i = v->index; i < d->var_cnt; i++)
471 d->var[i]->index = i;
473 /* Update name hash. */
474 hsh_force_delete (d->name_tab, v);
477 val_labs_destroy (v->val_labs);
478 cat_stored_values_destroy (v);
483 /* Deletes the COUNT variables listed in VARS from D. This is
484 unsafe; see the comment on dict_delete_var() for details. */
486 dict_delete_vars (struct dictionary *d,
487 struct variable *const *vars, size_t count)
489 /* FIXME: this can be done in O(count) time, but this algorithm
492 assert (count == 0 || vars != NULL);
495 dict_delete_var (d, *vars++);
498 /* Deletes scratch variables from dictionary D. */
500 dict_delete_scratch_vars (struct dictionary *d)
504 /* FIXME: this can be done in O(count) time, but this algorithm
508 for (i = 0; i < d->var_cnt; )
509 if (dict_class_from_id (d->var[i]->name) == DC_SCRATCH)
510 dict_delete_var (d, d->var[i]);
515 /* Moves V to 0-based position IDX in D. Other variables in D,
516 if any, retain their relative positions. Runs in time linear
517 in the distance moved. */
519 dict_reorder_var (struct dictionary *d, struct variable *v,
522 size_t min_idx, max_idx;
527 assert (dict_contains_var (d, v));
528 assert (new_index < d->var_cnt);
530 move_element (d->var, d->var_cnt, sizeof *d->var, v->index, new_index);
532 min_idx = min (v->index, new_index);
533 max_idx = max (v->index, new_index);
534 for (i = min_idx; i <= max_idx; i++)
535 d->var[i]->index = i;
538 /* Reorders the variables in D, placing the COUNT variables
539 listed in ORDER in that order at the beginning of D. The
540 other variables in D, if any, retain their relative
543 dict_reorder_vars (struct dictionary *d,
544 struct variable *const *order, size_t count)
546 struct variable **new_var;
550 assert (count == 0 || order != NULL);
551 assert (count <= d->var_cnt);
553 new_var = xnmalloc (d->var_cnt, sizeof *new_var);
554 memcpy (new_var, order, count * sizeof *new_var);
555 for (i = 0; i < count; i++)
557 assert (d->var[order[i]->index] != NULL);
558 d->var[order[i]->index] = NULL;
561 for (i = 0; i < d->var_cnt; i++)
562 if (d->var[i] != NULL)
564 assert (count < d->var_cnt);
565 new_var[count] = d->var[i];
566 new_var[count]->index = count;
573 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
574 a variable named NEW_NAME is already in D, except that
575 NEW_NAME may be the same as V's existing name. */
577 dict_rename_var (struct dictionary *d, struct variable *v,
578 const char *new_name)
582 assert (new_name != NULL);
583 assert (var_is_valid_name (new_name, false));
584 assert (dict_contains_var (d, v));
585 assert (!compare_var_names (v->name, new_name, NULL)
586 || dict_lookup_var (d, new_name) == NULL);
588 hsh_force_delete (d->name_tab, v);
589 str_copy_trunc (v->name, sizeof v->name, new_name);
590 hsh_force_insert (d->name_tab, v);
592 if (get_algorithm () == ENHANCED)
593 var_clear_short_name (v);
596 /* Renames COUNT variables specified in VARS to the names given
597 in NEW_NAMES within dictionary D. If the renaming would
598 result in a duplicate variable name, returns false and stores a
599 name that would be duplicated into *ERR_NAME (if ERR_NAME is
600 non-null). Otherwise, the renaming is successful, and true
603 dict_rename_vars (struct dictionary *d,
604 struct variable **vars, char **new_names,
605 size_t count, char **err_name)
612 assert (count == 0 || vars != NULL);
613 assert (count == 0 || new_names != NULL);
615 /* Remove the variables to be renamed from the name hash,
616 save their names, and rename them. */
617 old_names = xnmalloc (count, sizeof *old_names);
618 for (i = 0; i < count; i++)
620 assert (d->var[vars[i]->index] == vars[i]);
621 assert (var_is_valid_name (new_names[i], false));
622 hsh_force_delete (d->name_tab, vars[i]);
623 old_names[i] = xstrdup (vars[i]->name);
624 strcpy (vars[i]->name, new_names[i]);
627 /* Add the renamed variables back into the name hash,
628 checking for conflicts. */
629 for (i = 0; i < count; i++)
631 assert (new_names[i] != NULL);
632 assert (*new_names[i] != '\0');
633 assert (strlen (new_names[i]) >= 1);
634 assert (strlen (new_names[i]) <= LONG_NAME_LEN);
636 if (hsh_insert (d->name_tab, vars[i]) != NULL)
638 /* There is a name conflict.
639 Back out all the name changes that have already
640 taken place, and indicate failure. */
642 if (err_name != NULL)
643 *err_name = new_names[i];
645 for (i = 0; i < fail_idx; i++)
646 hsh_force_delete (d->name_tab, vars[i]);
648 for (i = 0; i < count; i++)
650 strcpy (vars[i]->name, old_names[i]);
651 hsh_force_insert (d->name_tab, vars[i]);
659 /* Clear short names. */
660 if (get_algorithm () == ENHANCED)
661 for (i = 0; i < count; i++)
662 var_clear_short_name (vars[i]);
665 /* Free the old names we kept around. */
666 for (i = 0; i < count; i++)
673 /* Returns the weighting variable in dictionary D, or a null
674 pointer if the dictionary is unweighted. */
676 dict_get_weight (const struct dictionary *d)
679 assert (d->weight == NULL || dict_contains_var (d, d->weight));
684 /* Returns the value of D's weighting variable in case C, except that a
685 negative weight is returned as 0. Returns 1 if the dictionary is
686 unweighted. Will warn about missing, negative, or zero values if
687 warn_on_invalid is nonzero. The function will set warn_on_invalid to zero
688 if an invalid weight is found. */
690 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
691 int *warn_on_invalid)
696 if (d->weight == NULL)
700 double w = case_num (c, d->weight->fv);
701 if (w < 0.0 || mv_is_num_missing (&d->weight->miss, w))
703 if ( w == 0.0 && *warn_on_invalid ) {
704 *warn_on_invalid = 0;
705 msg (SW, _("At least one case in the data file had a weight value "
706 "that was user-missing, system-missing, zero, or "
707 "negative. These case(s) were ignored."));
713 /* Sets the weighting variable of D to V, or turning off
714 weighting if V is a null pointer. */
716 dict_set_weight (struct dictionary *d, struct variable *v)
719 assert (v == NULL || dict_contains_var (d, v));
720 assert (v == NULL || v->type == NUMERIC);
725 /* Returns the filter variable in dictionary D (see cmd_filter())
726 or a null pointer if the dictionary is unfiltered. */
728 dict_get_filter (const struct dictionary *d)
731 assert (d->filter == NULL || dict_contains_var (d, d->filter));
736 /* Sets V as the filter variable for dictionary D. Passing a
737 null pointer for V turn off filtering. */
739 dict_set_filter (struct dictionary *d, struct variable *v)
742 assert (v == NULL || dict_contains_var (d, v));
747 /* Returns the case limit for dictionary D, or zero if the number
748 of cases is unlimited (see cmd_n()). */
750 dict_get_case_limit (const struct dictionary *d)
754 return d->case_limit;
757 /* Sets CASE_LIMIT as the case limit for dictionary D. Zero for
758 CASE_LIMIT indicates no limit. */
760 dict_set_case_limit (struct dictionary *d, int case_limit)
763 assert (case_limit >= 0);
765 d->case_limit = case_limit;
768 /* Returns the index of the next value to be added to D. This
769 value is the number of `union value's that need to be
770 allocated to store a case for dictionary D. */
772 dict_get_next_value_idx (const struct dictionary *d)
776 return d->next_value_idx;
779 /* Returns the number of bytes needed to store a case for
782 dict_get_case_size (const struct dictionary *d)
786 return sizeof (union value) * dict_get_next_value_idx (d);
789 /* Deletes scratch variables in dictionary D and reassigns values
790 so that fragmentation is eliminated. */
792 dict_compact_values (struct dictionary *d)
796 d->next_value_idx = 0;
797 for (i = 0; i < d->var_cnt; )
799 struct variable *v = d->var[i];
801 if (dict_class_from_id (v->name) != DC_SCRATCH)
803 v->fv = d->next_value_idx;
804 d->next_value_idx += v->nv;
808 dict_delete_var (d, v);
812 /* Returns the number of values that would be used by a case if
813 dict_compact_values() were called. */
815 dict_get_compacted_value_cnt (const struct dictionary *d)
821 for (i = 0; i < d->var_cnt; i++)
822 if (dict_class_from_id (d->var[i]->name) != DC_SCRATCH)
823 cnt += d->var[i]->nv;
827 /* Creates and returns an array mapping from a dictionary index
828 to the `fv' that the corresponding variable will have after
829 calling dict_compact_values(). Scratch variables receive -1
830 for `fv' because dict_compact_values() will delete them. */
832 dict_get_compacted_idx_to_fv (const struct dictionary *d)
835 size_t next_value_idx;
838 idx_to_fv = xnmalloc (d->var_cnt, sizeof *idx_to_fv);
840 for (i = 0; i < d->var_cnt; i++)
842 struct variable *v = d->var[i];
844 if (dict_class_from_id (v->name) != DC_SCRATCH)
846 idx_to_fv[i] = next_value_idx;
847 next_value_idx += v->nv;
855 /* Returns true if a case for dictionary D would be smaller after
856 compaction, false otherwise. Compacting a case eliminates
857 "holes" between values and after the last value. Holes are
858 created by deleting variables (or by scratch variables).
860 The return value may differ from whether compacting a case
861 from dictionary D would *change* the case: compaction could
862 rearrange values even if it didn't reduce space
865 dict_needs_compaction (const struct dictionary *d)
867 return dict_get_compacted_value_cnt (d) < dict_get_next_value_idx (d);
870 /* How to copy a contiguous range of values between cases. */
873 size_t src_idx; /* Starting value index in source case. */
874 size_t dst_idx; /* Starting value index in target case. */
875 size_t cnt; /* Number of values. */
878 /* How to compact a case. */
879 struct dict_compactor
881 struct copy_map *maps; /* Array of mappings. */
882 size_t map_cnt; /* Number of mappings. */
885 /* Creates and returns a dict_compactor that can be used to
886 compact cases for dictionary D.
888 Compacting a case eliminates "holes" between values and after
889 the last value. Holes are created by deleting variables (or
890 by scratch variables). */
891 struct dict_compactor *
892 dict_make_compactor (const struct dictionary *d)
894 struct dict_compactor *compactor;
895 struct copy_map *map;
896 size_t map_allocated;
900 compactor = xmalloc (sizeof *compactor);
901 compactor->maps = NULL;
902 compactor->map_cnt = 0;
907 for (i = 0; i < d->var_cnt; i++)
909 struct variable *v = d->var[i];
911 if (dict_class_from_id (v->name) == DC_SCRATCH)
913 if (map != NULL && map->src_idx + map->cnt == v->fv)
917 if (compactor->map_cnt == map_allocated)
918 compactor->maps = x2nrealloc (compactor->maps, &map_allocated,
919 sizeof *compactor->maps);
920 map = &compactor->maps[compactor->map_cnt++];
921 map->src_idx = v->fv;
922 map->dst_idx = value_idx;
931 /* Compacts SRC by copying it to DST according to the scheme in
934 Compacting a case eliminates "holes" between values and after
935 the last value. Holes are created by deleting variables (or
936 by scratch variables). */
938 dict_compactor_compact (const struct dict_compactor *compactor,
939 struct ccase *dst, const struct ccase *src)
943 for (i = 0; i < compactor->map_cnt; i++)
945 const struct copy_map *map = &compactor->maps[i];
946 case_copy (dst, map->dst_idx, src, map->src_idx, map->cnt);
950 /* Destroys COMPACTOR. */
952 dict_compactor_destroy (struct dict_compactor *compactor)
954 if (compactor != NULL)
956 free (compactor->maps);
961 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
962 dict_get_split_cnt() to determine how many SPLIT FILE vars
963 there are. Returns a null pointer if and only if there are no
965 struct variable *const *
966 dict_get_split_vars (const struct dictionary *d)
973 /* Returns the number of SPLIT FILE vars. */
975 dict_get_split_cnt (const struct dictionary *d)
982 /* Sets CNT split vars SPLIT in dictionary D. */
984 dict_set_split_vars (struct dictionary *d,
985 struct variable *const *split, size_t cnt)
988 assert (cnt == 0 || split != NULL);
991 d->split = xnrealloc (d->split, cnt, sizeof *d->split);
992 memcpy (d->split, split, cnt * sizeof *d->split);
995 /* Returns the file label for D, or a null pointer if D is
996 unlabeled (see cmd_file_label()). */
998 dict_get_label (const struct dictionary *d)
1005 /* Sets D's file label to LABEL, truncating it to a maximum of 60
1008 dict_set_label (struct dictionary *d, const char *label)
1015 else if (strlen (label) < 60)
1016 d->label = xstrdup (label);
1019 d->label = xmalloc (61);
1020 memcpy (d->label, label, 60);
1021 d->label[60] = '\0';
1025 /* Returns the documents for D, or a null pointer if D has no
1026 documents (see cmd_document()).. */
1028 dict_get_documents (const struct dictionary *d)
1032 return d->documents;
1035 /* Sets the documents for D to DOCUMENTS, or removes D's
1036 documents if DOCUMENT is a null pointer. */
1038 dict_set_documents (struct dictionary *d, const char *documents)
1042 free (d->documents);
1043 if (documents == NULL)
1044 d->documents = NULL;
1046 d->documents = xstrdup (documents);
1049 /* Creates in D a vector named NAME that contains CNT variables
1050 VAR (see cmd_vector()). Returns true if successful, or
1051 false if a vector named NAME already exists in D. */
1053 dict_create_vector (struct dictionary *d,
1055 struct variable **var, size_t cnt)
1057 struct vector *vector;
1061 assert (name != NULL);
1062 assert (var_is_valid_name (name, false));
1063 assert (var != NULL);
1066 if (dict_lookup_vector (d, name) != NULL)
1069 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1070 vector = d->vector[d->vector_cnt] = xmalloc (sizeof *vector);
1071 vector->idx = d->vector_cnt++;
1072 str_copy_trunc (vector->name, sizeof vector->name, name);
1073 vector->var = xnmalloc (cnt, sizeof *var);
1074 for (i = 0; i < cnt; i++)
1076 assert (dict_contains_var (d, var[i]));
1077 vector->var[i] = var[i];
1084 /* Returns the vector in D with index IDX, which must be less
1085 than dict_get_vector_cnt (D). */
1086 const struct vector *
1087 dict_get_vector (const struct dictionary *d, size_t idx)
1090 assert (idx < d->vector_cnt);
1092 return d->vector[idx];
1095 /* Returns the number of vectors in D. */
1097 dict_get_vector_cnt (const struct dictionary *d)
1101 return d->vector_cnt;
1104 /* Looks up and returns the vector within D with the given
1106 const struct vector *
1107 dict_lookup_vector (const struct dictionary *d, const char *name)
1112 assert (name != NULL);
1114 for (i = 0; i < d->vector_cnt; i++)
1115 if (!strcasecmp (d->vector[i]->name, name))
1116 return d->vector[i];
1120 /* Deletes all vectors from D. */
1122 dict_clear_vectors (struct dictionary *d)
1128 for (i = 0; i < d->vector_cnt; i++)
1130 free (d->vector[i]->var);
1131 free (d->vector[i]);
1138 /* Compares two strings. */
1140 compare_strings (const void *a, const void *b, void *aux UNUSED)
1142 return strcmp (a, b);
1145 /* Hashes a string. */
1147 hash_string (const void *s, void *aux UNUSED)
1149 return hsh_hash_string (s);
1152 /* Assigns a valid, unique short_name[] to each variable in D.
1153 Each variable whose actual name is short has highest priority
1154 for that short name. Otherwise, variables with an existing
1155 short_name[] have the next highest priority for a given short
1156 name; if it is already taken, then the variable is treated as
1157 if short_name[] had been empty. Otherwise, long names are
1158 truncated to form short names. If that causes conflicts,
1159 variables are renamed as PREFIX_A, PREFIX_B, and so on. */
1161 dict_assign_short_names (struct dictionary *d)
1163 struct hsh_table *short_names;
1166 /* Give variables whose names are short the corresponding short
1167 names, and clear short_names[] that conflict with a variable
1169 for (i = 0; i < d->var_cnt; i++)
1171 struct variable *v = d->var[i];
1172 if (strlen (v->name) <= SHORT_NAME_LEN)
1173 var_set_short_name (v, v->name);
1174 else if (dict_lookup_var (d, v->short_name) != NULL)
1175 var_clear_short_name (v);
1178 /* Each variable with an assigned short_name[] now gets it
1179 unless there is a conflict. */
1180 short_names = hsh_create (d->var_cnt, compare_strings, hash_string,
1182 for (i = 0; i < d->var_cnt; i++)
1184 struct variable *v = d->var[i];
1185 if (v->short_name[0] && hsh_insert (short_names, v->short_name) != NULL)
1186 var_clear_short_name (v);
1189 /* Now assign short names to remaining variables. */
1190 for (i = 0; i < d->var_cnt; i++)
1192 struct variable *v = d->var[i];
1193 if (v->short_name[0] == '\0')
1197 /* Form initial short_name. */
1198 var_set_short_name (v, v->name);
1200 /* Try _A, _B, ... _AA, _AB, etc., if needed. */
1201 for (sfx = 0; hsh_insert (short_names, v->short_name) != NULL; sfx++)
1202 var_set_short_name_suffix (v, v->name, sfx);
1206 /* Get rid of hash table. */
1207 hsh_destroy (short_names);