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"
28 #include "cat-routines.h"
35 #include "value-labels.h"
39 #define _(msgid) gettext (msgid)
44 struct variable **var; /* Variables. */
45 size_t var_cnt, var_cap; /* Number of variables, capacity. */
46 struct hsh_table *name_tab; /* Variable index by name. */
47 int next_value_idx; /* Index of next `union value' to allocate. */
48 struct variable **split; /* SPLIT FILE vars. */
49 size_t split_cnt; /* SPLIT FILE count. */
50 struct variable *weight; /* WEIGHT variable. */
51 struct variable *filter; /* FILTER variable. */
52 int case_limit; /* Current case limit (N command). */
53 char *label; /* File label. */
54 char *documents; /* Documents, as a string. */
55 struct vector **vector; /* Vectors of variables. */
56 size_t vector_cnt; /* Number of vectors. */
59 /* Active file dictionary. */
60 struct dictionary *default_dict;
62 /* Creates and returns a new dictionary. */
66 struct dictionary *d = xmalloc (sizeof *d);
69 d->var_cnt = d->var_cap = 0;
70 d->name_tab = hsh_create (8, compare_var_names, hash_var_name, NULL, NULL);
71 d->next_value_idx = 0;
85 /* Creates and returns a (deep) copy of an existing
88 dict_clone (const struct dictionary *s)
97 for (i = 0; i < s->var_cnt; i++)
99 struct variable *sv = s->var[i];
100 struct variable *dv = dict_clone_var_assert (d, sv, sv->name);
101 var_set_short_name (dv, sv->short_name);
104 d->next_value_idx = s->next_value_idx;
106 d->split_cnt = s->split_cnt;
107 if (d->split_cnt > 0)
109 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
110 for (i = 0; i < d->split_cnt; i++)
111 d->split[i] = dict_lookup_var_assert (d, s->split[i]->name);
114 if (s->weight != NULL)
115 d->weight = dict_lookup_var_assert (d, s->weight->name);
117 if (s->filter != NULL)
118 d->filter = dict_lookup_var_assert (d, s->filter->name);
120 d->case_limit = s->case_limit;
121 dict_set_label (d, dict_get_label (s));
122 dict_set_documents (d, dict_get_documents (s));
124 d->vector_cnt = s->vector_cnt;
125 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
126 for (i = 0; i < s->vector_cnt; i++)
128 struct vector *sv = s->vector[i];
129 struct vector *dv = d->vector[i] = xmalloc (sizeof *dv);
133 strcpy (dv->name, sv->name);
135 dv->var = xnmalloc (dv->cnt, sizeof *dv->var);
136 for (j = 0; j < dv->cnt; j++)
137 dv->var[j] = d->var[sv->var[j]->index];
143 /* Clears the contents from a dictionary without destroying the
144 dictionary itself. */
146 dict_clear (struct dictionary *d)
148 /* FIXME? Should we really clear case_limit, label, documents?
149 Others are necessarily cleared by deleting all the variables.*/
154 for (i = 0; i < d->var_cnt; i++)
156 struct variable *v = d->var[i];
158 val_labs_destroy (v->val_labs);
164 d->var_cnt = d->var_cap = 0;
165 hsh_clear (d->name_tab);
166 d->next_value_idx = 0;
177 dict_clear_vectors (d);
180 /* Destroys the aux data for every variable in D, by calling
181 var_clear_aux() for each variable. */
183 dict_clear_aux (struct dictionary *d)
189 for (i = 0; i < d->var_cnt; i++)
190 var_clear_aux (d->var[i]);
193 /* Clears a dictionary and destroys it. */
195 dict_destroy (struct dictionary *d)
200 hsh_destroy (d->name_tab);
205 /* Returns the number of variables in D. */
207 dict_get_var_cnt (const struct dictionary *d)
214 /* Returns the variable in D with index IDX, which must be
215 between 0 and the count returned by dict_get_var_cnt(),
218 dict_get_var (const struct dictionary *d, size_t idx)
221 assert (idx < d->var_cnt);
226 /* Sets *VARS to an array of pointers to variables in D and *CNT
227 to the number of variables in *D. By default all variables
228 are returned, but bits may be set in EXCLUDE_CLASSES to
229 exclude ordinary, system, and/or scratch variables. */
231 dict_get_vars (const struct dictionary *d, struct variable ***vars,
232 size_t *cnt, unsigned exclude_classes)
238 assert (vars != NULL);
239 assert (cnt != NULL);
240 assert ((exclude_classes & ~((1u << DC_ORDINARY)
242 | (1u << DC_SCRATCH))) == 0);
245 for (i = 0; i < d->var_cnt; i++)
246 if (!(exclude_classes & (1u << dict_class_from_id (d->var[i]->name))))
249 *vars = xnmalloc (count, sizeof **vars);
251 for (i = 0; i < d->var_cnt; i++)
252 if (!(exclude_classes & (1u << dict_class_from_id (d->var[i]->name))))
253 (*vars)[(*cnt)++] = d->var[i];
254 assert (*cnt == count);
258 /* Creates and returns a new variable in D with the given NAME
259 and WIDTH. Returns a null pointer if the given NAME would
260 duplicate that of an existing variable in the dictionary. */
262 dict_create_var (struct dictionary *d, const char *name, int width)
267 assert (name != NULL);
269 assert (width >= 0 && width < 256);
271 assert (var_is_valid_name(name,0));
273 /* Make sure there's not already a variable by that name. */
274 if (dict_lookup_var (d, name) != NULL)
277 /* Allocate and initialize variable. */
278 v = xmalloc (sizeof *v);
279 str_copy_trunc (v->name, sizeof v->name, name);
280 v->type = width == 0 ? NUMERIC : ALPHA;
282 v->fv = d->next_value_idx;
283 v->nv = width == 0 ? 1 : DIV_RND_UP (width, 8);
285 v->reinit = dict_class_from_id (v->name) != DC_SCRATCH;
286 v->index = d->var_cnt;
287 mv_init (&v->miss, width);
288 if (v->type == NUMERIC)
291 v->alignment = ALIGN_RIGHT;
292 v->display_width = 8;
293 v->measure = MEASURE_SCALE;
297 v->print = make_output_format (FMT_A, v->width, 0);
298 v->alignment = ALIGN_LEFT;
299 v->display_width = 8;
300 v->measure = MEASURE_NOMINAL;
303 v->val_labs = val_labs_create (v->width);
305 var_clear_short_name (v);
310 /* Update dictionary. */
311 if (d->var_cnt >= d->var_cap)
313 d->var_cap = 8 + 2 * d->var_cap;
314 d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var);
316 d->var[v->index] = v;
318 hsh_force_insert (d->name_tab, v);
320 d->next_value_idx += v->nv;
325 /* Creates and returns a new variable in D with the given NAME
326 and WIDTH. Assert-fails if the given NAME would duplicate
327 that of an existing variable in the dictionary. */
329 dict_create_var_assert (struct dictionary *d, const char *name, int width)
331 struct variable *v = dict_create_var (d, name, width);
336 /* Creates and returns a new variable in D with name NAME, as a
337 copy of existing variable OV, which need not be in D or in any
338 dictionary. Returns a null pointer if the given NAME would
339 duplicate that of an existing variable in the dictionary. */
341 dict_clone_var (struct dictionary *d, const struct variable *ov,
348 assert (name != NULL);
350 assert (strlen (name) >= 1);
351 assert (strlen (name) <= LONG_NAME_LEN);
353 nv = dict_create_var (d, name, ov->width);
357 /* Copy most members not copied via dict_create_var().
358 short_name[] is intentionally not copied, because there is
359 no reason to give a new variable with potentially a new name
360 the same short name. */
362 nv->reinit = ov->reinit;
363 mv_copy (&nv->miss, &ov->miss);
364 nv->print = ov->print;
365 nv->write = ov->write;
366 val_labs_destroy (nv->val_labs);
367 nv->val_labs = val_labs_copy (ov->val_labs);
368 if (ov->label != NULL)
369 nv->label = xstrdup (ov->label);
370 nv->measure = ov->measure;
371 nv->display_width = ov->display_width;
372 nv->alignment = ov->alignment;
377 /* Creates and returns a new variable in D with name NAME, as a
378 copy of existing variable OV, which need not be in D or in any
379 dictionary. Assert-fails if the given NAME would duplicate
380 that of an existing variable in the dictionary. */
382 dict_clone_var_assert (struct dictionary *d, const struct variable *ov,
385 struct variable *v = dict_clone_var (d, ov, name);
390 /* Returns the variable named NAME in D, or a null pointer if no
391 variable has that name. */
393 dict_lookup_var (const struct dictionary *d, const char *name)
398 assert (name != NULL);
400 str_copy_trunc (v.name, sizeof v.name, name);
401 return hsh_find (d->name_tab, &v);
404 /* Returns the variable named NAME in D. Assert-fails if no
405 variable has that name. */
407 dict_lookup_var_assert (const struct dictionary *d, const char *name)
409 struct variable *v = dict_lookup_var (d, name);
414 /* Returns true if variable V is in dictionary D,
417 dict_contains_var (const struct dictionary *d, const struct variable *v)
422 return v->index >= 0 && v->index < d->var_cnt && d->var[v->index] == v;
425 /* Compares two double pointers to variables, which should point
426 to elements of a struct dictionary's `var' member array. */
428 compare_var_ptrs (const void *a_, const void *b_, void *aux UNUSED)
430 struct variable *const *a = a_;
431 struct variable *const *b = b_;
433 return *a < *b ? -1 : *a > *b;
436 /* Deletes variable V from dictionary D and frees V.
438 This is a very bad idea if there might be any pointers to V
439 from outside D. In general, no variable in default_dict
440 should be deleted when any transformations are active, because
441 those transformations might reference the deleted variable.
442 The safest time to delete a variable is just after a procedure
443 has been executed, as done by MODIFY VARS.
445 Pointers to V within D are not a problem, because
446 dict_delete_var() knows to remove V from split variables,
447 weights, filters, etc. */
449 dict_delete_var (struct dictionary *d, struct variable *v)
455 assert (dict_contains_var (d, v));
457 /* Delete aux data. */
460 /* Remove V from splits, weight, filter variables. */
461 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
462 &v, compare_var_ptrs, NULL);
467 dict_clear_vectors (d);
469 /* Remove V from var array. */
470 remove_element (d->var, d->var_cnt, sizeof *d->var, v->index);
474 for (i = v->index; i < d->var_cnt; i++)
475 d->var[i]->index = i;
477 /* Update name hash. */
478 hsh_force_delete (d->name_tab, v);
481 val_labs_destroy (v->val_labs);
482 cat_stored_values_destroy (v);
487 /* Deletes the COUNT variables listed in VARS from D. This is
488 unsafe; see the comment on dict_delete_var() for details. */
490 dict_delete_vars (struct dictionary *d,
491 struct variable *const *vars, size_t count)
493 /* FIXME: this can be done in O(count) time, but this algorithm
496 assert (count == 0 || vars != NULL);
499 dict_delete_var (d, *vars++);
502 /* Deletes scratch variables from dictionary D. */
504 dict_delete_scratch_vars (struct dictionary *d)
508 /* FIXME: this can be done in O(count) time, but this algorithm
512 for (i = 0; i < d->var_cnt; )
513 if (dict_class_from_id (d->var[i]->name) == DC_SCRATCH)
514 dict_delete_var (d, d->var[i]);
519 /* Moves V to 0-based position IDX in D. Other variables in D,
520 if any, retain their relative positions. Runs in time linear
521 in the distance moved. */
523 dict_reorder_var (struct dictionary *d, struct variable *v,
526 size_t min_idx, max_idx;
531 assert (dict_contains_var (d, v));
532 assert (new_index < d->var_cnt);
534 move_element (d->var, d->var_cnt, sizeof *d->var, v->index, new_index);
536 min_idx = min (v->index, new_index);
537 max_idx = max (v->index, new_index);
538 for (i = min_idx; i <= max_idx; i++)
539 d->var[i]->index = i;
542 /* Reorders the variables in D, placing the COUNT variables
543 listed in ORDER in that order at the beginning of D. The
544 other variables in D, if any, retain their relative
547 dict_reorder_vars (struct dictionary *d,
548 struct variable *const *order, size_t count)
550 struct variable **new_var;
554 assert (count == 0 || order != NULL);
555 assert (count <= d->var_cnt);
557 new_var = xnmalloc (d->var_cnt, sizeof *new_var);
558 memcpy (new_var, order, count * sizeof *new_var);
559 for (i = 0; i < count; i++)
561 assert (d->var[order[i]->index] != NULL);
562 d->var[order[i]->index] = NULL;
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 new_var[count]->index = count;
577 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
578 a variable named NEW_NAME is already in D, except that
579 NEW_NAME may be the same as V's existing name. */
581 dict_rename_var (struct dictionary *d, struct variable *v,
582 const char *new_name)
586 assert (new_name != NULL);
587 assert (var_is_valid_name (new_name, false));
588 assert (dict_contains_var (d, v));
589 assert (!compare_var_names (v->name, new_name, NULL)
590 || dict_lookup_var (d, new_name) == NULL);
592 hsh_force_delete (d->name_tab, v);
593 str_copy_trunc (v->name, sizeof v->name, new_name);
594 hsh_force_insert (d->name_tab, v);
596 if (get_algorithm () == ENHANCED)
597 var_clear_short_name (v);
600 /* Renames COUNT variables specified in VARS to the names given
601 in NEW_NAMES within dictionary D. If the renaming would
602 result in a duplicate variable name, returns false and stores a
603 name that would be duplicated into *ERR_NAME (if ERR_NAME is
604 non-null). Otherwise, the renaming is successful, and true
607 dict_rename_vars (struct dictionary *d,
608 struct variable **vars, char **new_names,
609 size_t count, char **err_name)
616 assert (count == 0 || vars != NULL);
617 assert (count == 0 || new_names != NULL);
619 /* Remove the variables to be renamed from the name hash,
620 save their names, and rename them. */
621 old_names = xnmalloc (count, sizeof *old_names);
622 for (i = 0; i < count; i++)
624 assert (d->var[vars[i]->index] == vars[i]);
625 assert (var_is_valid_name (new_names[i], false));
626 hsh_force_delete (d->name_tab, vars[i]);
627 old_names[i] = xstrdup (vars[i]->name);
628 strcpy (vars[i]->name, new_names[i]);
631 /* Add the renamed variables back into the name hash,
632 checking for conflicts. */
633 for (i = 0; i < count; i++)
635 assert (new_names[i] != NULL);
636 assert (*new_names[i] != '\0');
637 assert (strlen (new_names[i]) >= 1);
638 assert (strlen (new_names[i]) <= LONG_NAME_LEN);
640 if (hsh_insert (d->name_tab, vars[i]) != NULL)
642 /* There is a name conflict.
643 Back out all the name changes that have already
644 taken place, and indicate failure. */
646 if (err_name != NULL)
647 *err_name = new_names[i];
649 for (i = 0; i < fail_idx; i++)
650 hsh_force_delete (d->name_tab, vars[i]);
652 for (i = 0; i < count; i++)
654 strcpy (vars[i]->name, old_names[i]);
655 hsh_force_insert (d->name_tab, vars[i]);
663 /* Clear short names. */
664 if (get_algorithm () == ENHANCED)
665 for (i = 0; i < count; i++)
666 var_clear_short_name (vars[i]);
669 /* Free the old names we kept around. */
670 for (i = 0; i < count; i++)
677 /* Returns the weighting variable in dictionary D, or a null
678 pointer if the dictionary is unweighted. */
680 dict_get_weight (const struct dictionary *d)
683 assert (d->weight == NULL || dict_contains_var (d, d->weight));
688 /* Returns the value of D's weighting variable in case C, except that a
689 negative weight is returned as 0. Returns 1 if the dictionary is
690 unweighted. Will warn about missing, negative, or zero values if
691 warn_on_invalid is nonzero. The function will set warn_on_invalid to zero
692 if an invalid weight is found. */
694 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
695 int *warn_on_invalid)
700 if (d->weight == NULL)
704 double w = case_num (c, d->weight->fv);
705 if (w < 0.0 || mv_is_num_missing (&d->weight->miss, w))
707 if ( w == 0.0 && *warn_on_invalid ) {
708 *warn_on_invalid = 0;
709 msg (SW, _("At least one case in the data file had a weight value "
710 "that was user-missing, system-missing, zero, or "
711 "negative. These case(s) were ignored."));
717 /* Sets the weighting variable of D to V, or turning off
718 weighting if V is a null pointer. */
720 dict_set_weight (struct dictionary *d, struct variable *v)
723 assert (v == NULL || dict_contains_var (d, v));
724 assert (v == NULL || v->type == NUMERIC);
729 /* Returns the filter variable in dictionary D (see cmd_filter())
730 or a null pointer if the dictionary is unfiltered. */
732 dict_get_filter (const struct dictionary *d)
735 assert (d->filter == NULL || dict_contains_var (d, d->filter));
740 /* Sets V as the filter variable for dictionary D. Passing a
741 null pointer for V turn off filtering. */
743 dict_set_filter (struct dictionary *d, struct variable *v)
746 assert (v == NULL || dict_contains_var (d, v));
751 /* Returns the case limit for dictionary D, or zero if the number
752 of cases is unlimited (see cmd_n()). */
754 dict_get_case_limit (const struct dictionary *d)
758 return d->case_limit;
761 /* Sets CASE_LIMIT as the case limit for dictionary D. Zero for
762 CASE_LIMIT indicates no limit. */
764 dict_set_case_limit (struct dictionary *d, int case_limit)
767 assert (case_limit >= 0);
769 d->case_limit = case_limit;
772 /* Returns the index of the next value to be added to D. This
773 value is the number of `union value's that need to be
774 allocated to store a case for dictionary D. */
776 dict_get_next_value_idx (const struct dictionary *d)
780 return d->next_value_idx;
783 /* Returns the number of bytes needed to store a case for
786 dict_get_case_size (const struct dictionary *d)
790 return sizeof (union value) * dict_get_next_value_idx (d);
793 /* Deletes scratch variables in dictionary D and reassigns values
794 so that fragmentation is eliminated. */
796 dict_compact_values (struct dictionary *d)
800 d->next_value_idx = 0;
801 for (i = 0; i < d->var_cnt; )
803 struct variable *v = d->var[i];
805 if (dict_class_from_id (v->name) != DC_SCRATCH)
807 v->fv = d->next_value_idx;
808 d->next_value_idx += v->nv;
812 dict_delete_var (d, v);
816 /* Returns the number of values that would be used by a case if
817 dict_compact_values() were called. */
819 dict_get_compacted_value_cnt (const struct dictionary *d)
825 for (i = 0; i < d->var_cnt; i++)
826 if (dict_class_from_id (d->var[i]->name) != DC_SCRATCH)
827 cnt += d->var[i]->nv;
831 /* Creates and returns an array mapping from a dictionary index
832 to the `fv' that the corresponding variable will have after
833 calling dict_compact_values(). Scratch variables receive -1
834 for `fv' because dict_compact_values() will delete them. */
836 dict_get_compacted_idx_to_fv (const struct dictionary *d)
839 size_t next_value_idx;
842 idx_to_fv = xnmalloc (d->var_cnt, sizeof *idx_to_fv);
844 for (i = 0; i < d->var_cnt; i++)
846 struct variable *v = d->var[i];
848 if (dict_class_from_id (v->name) != DC_SCRATCH)
850 idx_to_fv[i] = next_value_idx;
851 next_value_idx += v->nv;
859 /* Returns true if a case for dictionary D would be smaller after
860 compaction, false otherwise. Compacting a case eliminates
861 "holes" between values and after the last value. Holes are
862 created by deleting variables (or by scratch variables).
864 The return value may differ from whether compacting a case
865 from dictionary D would *change* the case: compaction could
866 rearrange values even if it didn't reduce space
869 dict_needs_compaction (const struct dictionary *d)
871 return dict_get_compacted_value_cnt (d) < dict_get_next_value_idx (d);
874 /* How to copy a contiguous range of values between cases. */
877 size_t src_idx; /* Starting value index in source case. */
878 size_t dst_idx; /* Starting value index in target case. */
879 size_t cnt; /* Number of values. */
882 /* How to compact a case. */
883 struct dict_compactor
885 struct copy_map *maps; /* Array of mappings. */
886 size_t map_cnt; /* Number of mappings. */
889 /* Creates and returns a dict_compactor that can be used to
890 compact cases for dictionary D.
892 Compacting a case eliminates "holes" between values and after
893 the last value. Holes are created by deleting variables (or
894 by scratch variables). */
895 struct dict_compactor *
896 dict_make_compactor (const struct dictionary *d)
898 struct dict_compactor *compactor;
899 struct copy_map *map;
900 size_t map_allocated;
904 compactor = xmalloc (sizeof *compactor);
905 compactor->maps = NULL;
906 compactor->map_cnt = 0;
911 for (i = 0; i < d->var_cnt; i++)
913 struct variable *v = d->var[i];
915 if (dict_class_from_id (v->name) == DC_SCRATCH)
917 if (map != NULL && map->src_idx + map->cnt == v->fv)
921 if (compactor->map_cnt == map_allocated)
922 compactor->maps = x2nrealloc (compactor->maps, &map_allocated,
923 sizeof *compactor->maps);
924 map = &compactor->maps[compactor->map_cnt++];
925 map->src_idx = v->fv;
926 map->dst_idx = value_idx;
935 /* Compacts SRC by copying it to DST according to the scheme in
938 Compacting a case eliminates "holes" between values and after
939 the last value. Holes are created by deleting variables (or
940 by scratch variables). */
942 dict_compactor_compact (const struct dict_compactor *compactor,
943 struct ccase *dst, const struct ccase *src)
947 for (i = 0; i < compactor->map_cnt; i++)
949 const struct copy_map *map = &compactor->maps[i];
950 case_copy (dst, map->dst_idx, src, map->src_idx, map->cnt);
954 /* Destroys COMPACTOR. */
956 dict_compactor_destroy (struct dict_compactor *compactor)
958 if (compactor != NULL)
960 free (compactor->maps);
965 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
966 dict_get_split_cnt() to determine how many SPLIT FILE vars
967 there are. Returns a null pointer if and only if there are no
969 struct variable *const *
970 dict_get_split_vars (const struct dictionary *d)
977 /* Returns the number of SPLIT FILE vars. */
979 dict_get_split_cnt (const struct dictionary *d)
986 /* Sets CNT split vars SPLIT in dictionary D. */
988 dict_set_split_vars (struct dictionary *d,
989 struct variable *const *split, size_t cnt)
992 assert (cnt == 0 || split != NULL);
995 d->split = xnrealloc (d->split, cnt, sizeof *d->split);
996 memcpy (d->split, split, cnt * sizeof *d->split);
999 /* Returns the file label for D, or a null pointer if D is
1000 unlabeled (see cmd_file_label()). */
1002 dict_get_label (const struct dictionary *d)
1009 /* Sets D's file label to LABEL, truncating it to a maximum of 60
1012 dict_set_label (struct dictionary *d, const char *label)
1019 else if (strlen (label) < 60)
1020 d->label = xstrdup (label);
1023 d->label = xmalloc (61);
1024 memcpy (d->label, label, 60);
1025 d->label[60] = '\0';
1029 /* Returns the documents for D, or a null pointer if D has no
1030 documents (see cmd_document()).. */
1032 dict_get_documents (const struct dictionary *d)
1036 return d->documents;
1039 /* Sets the documents for D to DOCUMENTS, or removes D's
1040 documents if DOCUMENT is a null pointer. */
1042 dict_set_documents (struct dictionary *d, const char *documents)
1046 free (d->documents);
1047 if (documents == NULL)
1048 d->documents = NULL;
1050 d->documents = xstrdup (documents);
1053 /* Creates in D a vector named NAME that contains CNT variables
1054 VAR (see cmd_vector()). Returns true if successful, or
1055 false if a vector named NAME already exists in D. */
1057 dict_create_vector (struct dictionary *d,
1059 struct variable **var, size_t cnt)
1061 struct vector *vector;
1065 assert (name != NULL);
1066 assert (var_is_valid_name (name, false));
1067 assert (var != NULL);
1070 if (dict_lookup_vector (d, name) != NULL)
1073 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1074 vector = d->vector[d->vector_cnt] = xmalloc (sizeof *vector);
1075 vector->idx = d->vector_cnt++;
1076 str_copy_trunc (vector->name, sizeof vector->name, name);
1077 vector->var = xnmalloc (cnt, sizeof *var);
1078 for (i = 0; i < cnt; i++)
1080 assert (dict_contains_var (d, var[i]));
1081 vector->var[i] = var[i];
1088 /* Returns the vector in D with index IDX, which must be less
1089 than dict_get_vector_cnt (D). */
1090 const struct vector *
1091 dict_get_vector (const struct dictionary *d, size_t idx)
1094 assert (idx < d->vector_cnt);
1096 return d->vector[idx];
1099 /* Returns the number of vectors in D. */
1101 dict_get_vector_cnt (const struct dictionary *d)
1105 return d->vector_cnt;
1108 /* Looks up and returns the vector within D with the given
1110 const struct vector *
1111 dict_lookup_vector (const struct dictionary *d, const char *name)
1116 assert (name != NULL);
1118 for (i = 0; i < d->vector_cnt; i++)
1119 if (!strcasecmp (d->vector[i]->name, name))
1120 return d->vector[i];
1124 /* Deletes all vectors from D. */
1126 dict_clear_vectors (struct dictionary *d)
1132 for (i = 0; i < d->vector_cnt; i++)
1134 free (d->vector[i]->var);
1135 free (d->vector[i]);
1142 /* Compares two strings. */
1144 compare_strings (const void *a, const void *b, void *aux UNUSED)
1146 return strcmp (a, b);
1149 /* Hashes a string. */
1151 hash_string (const void *s, void *aux UNUSED)
1153 return hsh_hash_string (s);
1156 /* Assigns a valid, unique short_name[] to each variable in D.
1157 Each variable whose actual name is short has highest priority
1158 for that short name. Otherwise, variables with an existing
1159 short_name[] have the next highest priority for a given short
1160 name; if it is already taken, then the variable is treated as
1161 if short_name[] had been empty. Otherwise, long names are
1162 truncated to form short names. If that causes conflicts,
1163 variables are renamed as PREFIX_A, PREFIX_B, and so on. */
1165 dict_assign_short_names (struct dictionary *d)
1167 struct hsh_table *short_names;
1170 /* Give variables whose names are short the corresponding short
1171 names, and clear short_names[] that conflict with a variable
1173 for (i = 0; i < d->var_cnt; i++)
1175 struct variable *v = d->var[i];
1176 if (strlen (v->name) <= SHORT_NAME_LEN)
1177 var_set_short_name (v, v->name);
1178 else if (dict_lookup_var (d, v->short_name) != NULL)
1179 var_clear_short_name (v);
1182 /* Each variable with an assigned short_name[] now gets it
1183 unless there is a conflict. */
1184 short_names = hsh_create (d->var_cnt, compare_strings, hash_string,
1186 for (i = 0; i < d->var_cnt; i++)
1188 struct variable *v = d->var[i];
1189 if (v->short_name[0] && hsh_insert (short_names, v->short_name) != NULL)
1190 var_clear_short_name (v);
1193 /* Now assign short names to remaining variables. */
1194 for (i = 0; i < d->var_cnt; i++)
1196 struct variable *v = d->var[i];
1197 if (v->short_name[0] == '\0')
1201 /* Form initial short_name. */
1202 var_set_short_name (v, v->name);
1204 /* Try _A, _B, ... _AA, _AB, etc., if needed. */
1205 for (sfx = 0; hsh_insert (short_names, v->short_name) != NULL; sfx++)
1206 var_set_short_name_suffix (v, v->name, sfx);
1210 /* Get rid of hash table. */
1211 hsh_destroy (short_names);