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
22 #include "dictionary.h"
28 #include "cat-routines.h"
31 #include "value-labels.h"
35 #include <libpspp/alloc.h>
36 #include <libpspp/array.h>
37 #include <libpspp/compiler.h>
38 #include <libpspp/hash.h>
39 #include <libpspp/message.h>
40 #include <libpspp/misc.h>
41 #include <libpspp/pool.h>
42 #include <libpspp/str.h>
47 #define _(msgid) gettext (msgid)
52 struct variable **var; /* Variables. */
53 size_t var_cnt, var_cap; /* Number of variables, capacity. */
54 struct hsh_table *name_tab; /* Variable index by name. */
55 int next_value_idx; /* Index of next `union value' to allocate. */
56 struct variable **split; /* SPLIT FILE vars. */
57 size_t split_cnt; /* SPLIT FILE count. */
58 struct variable *weight; /* WEIGHT variable. */
59 struct variable *filter; /* FILTER variable. */
60 size_t case_limit; /* Current case limit (N command). */
61 char *label; /* File label. */
62 char *documents; /* Documents, as a string. */
63 struct vector **vector; /* Vectors of variables. */
64 size_t vector_cnt; /* Number of vectors. */
67 /* Creates and returns a new dictionary. */
71 struct dictionary *d = xmalloc (sizeof *d);
74 d->var_cnt = d->var_cap = 0;
75 d->name_tab = hsh_create (8, compare_vars_by_name, hash_var_by_name,
77 d->next_value_idx = 0;
91 /* Creates and returns a (deep) copy of an existing
94 dict_clone (const struct dictionary *s)
103 for (i = 0; i < s->var_cnt; i++)
105 struct variable *sv = s->var[i];
106 struct variable *dv = dict_clone_var_assert (d, sv, var_get_name (sv));
107 var_set_short_name (dv, var_get_short_name (sv));
110 d->next_value_idx = s->next_value_idx;
112 d->split_cnt = s->split_cnt;
113 if (d->split_cnt > 0)
115 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
116 for (i = 0; i < d->split_cnt; i++)
117 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
120 if (s->weight != NULL)
121 d->weight = dict_lookup_var_assert (d, var_get_name (s->weight));
123 if (s->filter != NULL)
124 d->filter = dict_lookup_var_assert (d, var_get_name (s->filter));
126 d->case_limit = s->case_limit;
127 dict_set_label (d, dict_get_label (s));
128 dict_set_documents (d, dict_get_documents (s));
130 d->vector_cnt = s->vector_cnt;
131 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
132 for (i = 0; i < s->vector_cnt; i++)
133 d->vector[i] = vector_clone (s->vector[i], s, d);
138 /* Clears the contents from a dictionary without destroying the
139 dictionary itself. */
141 dict_clear (struct dictionary *d)
143 /* FIXME? Should we really clear case_limit, label, documents?
144 Others are necessarily cleared by deleting all the variables.*/
149 for (i = 0; i < d->var_cnt; i++)
151 var_clear_vardict (d->var[i]);
152 var_destroy (d->var[i]);
156 d->var_cnt = d->var_cap = 0;
157 hsh_clear (d->name_tab);
158 d->next_value_idx = 0;
169 dict_clear_vectors (d);
172 /* Destroys the aux data for every variable in D, by calling
173 var_clear_aux() for each variable. */
175 dict_clear_aux (struct dictionary *d)
181 for (i = 0; i < d->var_cnt; i++)
182 var_clear_aux (d->var[i]);
185 /* Clears a dictionary and destroys it. */
187 dict_destroy (struct dictionary *d)
192 hsh_destroy (d->name_tab);
197 /* Returns the number of variables in D. */
199 dict_get_var_cnt (const struct dictionary *d)
206 /* Returns the variable in D with dictionary index IDX, which
207 must be between 0 and the count returned by
208 dict_get_var_cnt(), exclusive. */
210 dict_get_var (const struct dictionary *d, size_t idx)
213 assert (idx < d->var_cnt);
218 /* Sets *VARS to an array of pointers to variables in D and *CNT
219 to the number of variables in *D. All variables are returned
220 if EXCLUDE_CLASSES is 0, or it may contain one or more of (1u
221 << DC_ORDINARY), (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to
222 exclude the corresponding type of variable. */
224 dict_get_vars (const struct dictionary *d, struct variable ***vars,
225 size_t *cnt, unsigned exclude_classes)
231 assert (vars != NULL);
232 assert (cnt != NULL);
233 assert ((exclude_classes & ~((1u << DC_ORDINARY)
235 | (1u << DC_SCRATCH))) == 0);
238 for (i = 0; i < d->var_cnt; i++)
240 enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
241 if (!(exclude_classes & (1u << class)))
245 *vars = xnmalloc (count, sizeof **vars);
247 for (i = 0; i < d->var_cnt; i++)
249 enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
250 if (!(exclude_classes & (1u << class)))
251 (*vars)[(*cnt)++] = d->var[i];
253 assert (*cnt == count);
256 static struct variable *
257 add_var (struct dictionary *d, struct variable *v)
259 /* Add dictionary info to variable. */
260 struct vardict_info vdi;
261 vdi.case_index = d->next_value_idx;
262 vdi.dict_index = d->var_cnt;
263 var_set_vardict (v, &vdi);
265 /* Update dictionary. */
266 if (d->var_cnt >= d->var_cap)
268 d->var_cap = 8 + 2 * d->var_cap;
269 d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var);
271 d->var[d->var_cnt++] = v;
272 hsh_force_insert (d->name_tab, v);
274 d->next_value_idx += var_get_value_cnt (v);
279 /* Creates and returns a new variable in D with the given NAME
280 and WIDTH. Returns a null pointer if the given NAME would
281 duplicate that of an existing variable in the dictionary. */
283 dict_create_var (struct dictionary *d, const char *name, int width)
285 return (dict_lookup_var (d, name) == NULL
286 ? dict_create_var_assert (d, name, width)
290 /* Creates and returns a new variable in D with the given NAME
291 and WIDTH. Assert-fails if the given NAME would duplicate
292 that of an existing variable in the dictionary. */
294 dict_create_var_assert (struct dictionary *d, const char *name, int width)
296 assert (dict_lookup_var (d, name) == NULL);
297 return add_var (d, var_create (name, width));
300 /* Creates and returns a new variable in D with name NAME, as a
301 copy of existing variable OLD_VAR, which need not be in D or
302 in any dictionary. Returns a null pointer if the given NAME
303 would duplicate that of an existing variable in the
306 dict_clone_var (struct dictionary *d, const struct variable *old_var,
309 return (dict_lookup_var (d, name) == NULL
310 ? dict_clone_var_assert (d, old_var, name)
314 /* Creates and returns a new variable in D with name NAME, as a
315 copy of existing variable OLD_VAR, which need not be in D or
316 in any dictionary. Assert-fails if the given NAME would
317 duplicate that of an existing variable in the dictionary. */
319 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var,
322 struct variable *new_var = var_clone (old_var);
323 assert (dict_lookup_var (d, name) == NULL);
324 var_set_name (new_var, name);
325 return add_var (d, new_var);
328 /* Returns the variable named NAME in D, or a null pointer if no
329 variable has that name. */
331 dict_lookup_var (const struct dictionary *d, const char *name)
333 struct variable *target = var_create (name, 0);
334 struct variable *result = hsh_find (d->name_tab, target);
335 var_destroy (target);
339 /* Returns the variable named NAME in D. Assert-fails if no
340 variable has that name. */
342 dict_lookup_var_assert (const struct dictionary *d, const char *name)
344 struct variable *v = dict_lookup_var (d, name);
349 /* Returns true if variable V is in dictionary D,
352 dict_contains_var (const struct dictionary *d, const struct variable *v)
354 if (var_has_vardict (v))
356 const struct vardict_info *vdi = var_get_vardict (v);
357 return (vdi->dict_index >= 0
358 && vdi->dict_index < d->var_cnt
359 && d->var[vdi->dict_index] == v);
365 /* Compares two double pointers to variables, which should point
366 to elements of a struct dictionary's `var' member array. */
368 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
370 struct variable *const *a = a_;
371 struct variable *const *b = b_;
373 return *a < *b ? -1 : *a > *b;
376 /* Sets the dict_index in V's vardict to DICT_INDEX. */
378 set_var_dict_index (struct variable *v, int dict_index)
380 struct vardict_info vdi = *var_get_vardict (v);
381 vdi.dict_index = dict_index;
382 var_set_vardict (v, &vdi);
385 /* Sets the case_index in V's vardict to DICT_INDEX. */
387 set_var_case_index (struct variable *v, int case_index)
389 struct vardict_info vdi = *var_get_vardict (v);
390 vdi.case_index = case_index;
391 var_set_vardict (v, &vdi);
394 /* Re-sets the dict_index in the dictionary variables with
395 indexes from FROM to TO (exclusive). */
397 reindex_vars (struct dictionary *d, size_t from, size_t to)
401 for (i = from; i < to; i++)
402 set_var_dict_index (d->var[i], i);
405 /* Deletes variable V from dictionary D and frees V.
407 This is a very bad idea if there might be any pointers to V
408 from outside D. In general, no variable in should be deleted when
409 any transformations are active on the dictionary's dataset, because
410 those transformations might reference the deleted variable.
411 The safest time to delete a variable is just after a procedure
412 has been executed, as done by MODIFY VARS.
414 Pointers to V within D are not a problem, because
415 dict_delete_var() knows to remove V from split variables,
416 weights, filters, etc. */
418 dict_delete_var (struct dictionary *d, struct variable *v)
420 int dict_index = var_get_dict_index (v);
422 assert (dict_contains_var (d, v));
424 /* Delete aux data. */
427 /* Remove V from splits, weight, filter variables. */
428 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
429 &v, compare_var_ptrs, NULL);
434 dict_clear_vectors (d);
436 /* Remove V from var array. */
437 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
440 /* Update dict_index for each affected variable. */
441 reindex_vars (d, dict_index, d->var_cnt);
443 /* Update name hash. */
444 hsh_force_delete (d->name_tab, v);
447 var_clear_vardict (v);
451 /* Deletes the COUNT variables listed in VARS from D. This is
452 unsafe; see the comment on dict_delete_var() for details. */
454 dict_delete_vars (struct dictionary *d,
455 struct variable *const *vars, size_t count)
457 /* FIXME: this can be done in O(count) time, but this algorithm
460 assert (count == 0 || vars != NULL);
463 dict_delete_var (d, *vars++);
466 /* Deletes the COUNT variables in D starting at index IDX. This
467 is unsafe; see the comment on dict_delete_var() for
470 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
472 /* FIXME: this can be done in O(count) time, but this algorithm
474 assert (idx + count <= d->var_cnt);
477 dict_delete_var (d, d->var[idx]);
480 /* Deletes scratch variables from dictionary D. */
482 dict_delete_scratch_vars (struct dictionary *d)
486 /* FIXME: this can be done in O(count) time, but this algorithm
490 for (i = 0; i < d->var_cnt; )
491 if (dict_class_from_id (var_get_name (d->var[i])) == DC_SCRATCH)
492 dict_delete_var (d, d->var[i]);
497 /* Moves V to 0-based position IDX in D. Other variables in D,
498 if any, retain their relative positions. Runs in time linear
499 in the distance moved. */
501 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
503 size_t old_index = var_get_dict_index (v);
505 assert (new_index < d->var_cnt);
506 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
507 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
510 /* Reorders the variables in D, placing the COUNT variables
511 listed in ORDER in that order at the beginning of D. The
512 other variables in D, if any, retain their relative
515 dict_reorder_vars (struct dictionary *d,
516 struct variable *const *order, size_t count)
518 struct variable **new_var;
522 assert (count == 0 || order != NULL);
523 assert (count <= d->var_cnt);
525 new_var = xnmalloc (d->var_cnt, sizeof *new_var);
526 memcpy (new_var, order, count * sizeof *new_var);
527 for (i = 0; i < count; i++)
529 size_t index = var_get_dict_index (order[i]);
530 assert (d->var[index] == order[i]);
531 d->var[index] = NULL;
532 set_var_dict_index (order[i], i);
534 for (i = 0; i < d->var_cnt; i++)
535 if (d->var[i] != NULL)
537 assert (count < d->var_cnt);
538 new_var[count] = d->var[i];
539 set_var_dict_index (new_var[count], count);
546 /* Changes the name of variable V in dictionary D to NEW_NAME. */
548 rename_var (struct dictionary *d, struct variable *v, const char *new_name)
550 struct vardict_info vdi;
552 assert (dict_contains_var (d, v));
554 vdi = *var_get_vardict (v);
555 var_clear_vardict (v);
556 var_set_name (v, new_name);
557 var_set_vardict (v, &vdi);
560 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
561 a variable named NEW_NAME is already in D, except that
562 NEW_NAME may be the same as V's existing name. */
564 dict_rename_var (struct dictionary *d, struct variable *v,
565 const char *new_name)
567 assert (!strcasecmp (var_get_name (v), new_name)
568 || dict_lookup_var (d, new_name) == NULL);
570 hsh_force_delete (d->name_tab, v);
571 rename_var (d, v, new_name);
572 hsh_force_insert (d->name_tab, v);
574 if (get_algorithm () == ENHANCED)
575 var_clear_short_name (v);
578 /* Renames COUNT variables specified in VARS to the names given
579 in NEW_NAMES within dictionary D. If the renaming would
580 result in a duplicate variable name, returns false and stores a
581 name that would be duplicated into *ERR_NAME (if ERR_NAME is
582 non-null). Otherwise, the renaming is successful, and true
585 dict_rename_vars (struct dictionary *d,
586 struct variable **vars, char **new_names, size_t count,
593 assert (count == 0 || vars != NULL);
594 assert (count == 0 || new_names != NULL);
596 /* Save the names of the variables to be renamed. */
597 pool = pool_create ();
598 old_names = pool_nalloc (pool, count, sizeof *old_names);
599 for (i = 0; i < count; i++)
600 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
602 /* Remove the variables to be renamed from the name hash,
604 for (i = 0; i < count; i++)
606 hsh_force_delete (d->name_tab, vars[i]);
607 rename_var (d, vars[i], new_names[i]);
610 /* Add the renamed variables back into the name hash,
611 checking for conflicts. */
612 for (i = 0; i < count; i++)
613 if (hsh_insert (d->name_tab, vars[i]) != NULL)
615 /* There is a name conflict.
616 Back out all the name changes that have already
617 taken place, and indicate failure. */
619 if (err_name != NULL)
620 *err_name = new_names[i];
622 for (i = 0; i < fail_idx; i++)
623 hsh_force_delete (d->name_tab, vars[i]);
625 for (i = 0; i < count; i++)
627 rename_var (d, vars[i], old_names[i]);
628 hsh_force_insert (d->name_tab, vars[i]);
635 /* Clear short names. */
636 if (get_algorithm () == ENHANCED)
637 for (i = 0; i < count; i++)
638 var_clear_short_name (vars[i]);
644 /* Returns the weighting variable in dictionary D, or a null
645 pointer if the dictionary is unweighted. */
647 dict_get_weight (const struct dictionary *d)
650 assert (d->weight == NULL || dict_contains_var (d, d->weight));
655 /* Returns the value of D's weighting variable in case C, except that a
656 negative weight is returned as 0. Returns 1 if the dictionary is
657 unweighted. Will warn about missing, negative, or zero values if
658 warn_on_invalid is true. The function will set warn_on_invalid to false
659 if an invalid weight is found. */
661 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
662 bool *warn_on_invalid)
667 if (d->weight == NULL)
671 double w = case_num (c, d->weight);
672 if (w < 0.0 || var_is_num_missing (d->weight, w))
674 if ( w == 0.0 && *warn_on_invalid ) {
675 *warn_on_invalid = false;
676 msg (SW, _("At least one case in the data file had a weight value "
677 "that was user-missing, system-missing, zero, or "
678 "negative. These case(s) were ignored."));
684 /* Sets the weighting variable of D to V, or turning off
685 weighting if V is a null pointer. */
687 dict_set_weight (struct dictionary *d, struct variable *v)
690 assert (v == NULL || dict_contains_var (d, v));
691 assert (v == NULL || var_is_numeric (v));
696 /* Returns the filter variable in dictionary D (see cmd_filter())
697 or a null pointer if the dictionary is unfiltered. */
699 dict_get_filter (const struct dictionary *d)
702 assert (d->filter == NULL || dict_contains_var (d, d->filter));
707 /* Sets V as the filter variable for dictionary D. Passing a
708 null pointer for V turn off filtering. */
710 dict_set_filter (struct dictionary *d, struct variable *v)
713 assert (v == NULL || dict_contains_var (d, v));
718 /* Returns the case limit for dictionary D, or zero if the number
719 of cases is unlimited. */
721 dict_get_case_limit (const struct dictionary *d)
725 return d->case_limit;
728 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
729 0 for CASE_LIMIT to indicate no limit. */
731 dict_set_case_limit (struct dictionary *d, size_t case_limit)
735 d->case_limit = case_limit;
738 /* Returns the case index of the next value to be added to D.
739 This value is the number of `union value's that need to be
740 allocated to store a case for dictionary D. */
742 dict_get_next_value_idx (const struct dictionary *d)
746 return d->next_value_idx;
749 /* Returns the number of bytes needed to store a case for
752 dict_get_case_size (const struct dictionary *d)
756 return sizeof (union value) * dict_get_next_value_idx (d);
759 /* Deletes scratch variables in dictionary D and reassigns values
760 so that fragmentation is eliminated. */
762 dict_compact_values (struct dictionary *d)
766 d->next_value_idx = 0;
767 for (i = 0; i < d->var_cnt; )
769 struct variable *v = d->var[i];
771 if (dict_class_from_id (var_get_name (v)) != DC_SCRATCH)
773 set_var_case_index (v, d->next_value_idx);
774 d->next_value_idx += var_get_value_cnt (v);
778 dict_delete_var (d, v);
782 /* Returns the number of values that would be used by a case if
783 dict_compact_values() were called. */
785 dict_get_compacted_value_cnt (const struct dictionary *d)
791 for (i = 0; i < d->var_cnt; i++)
792 if (dict_class_from_id (var_get_name (d->var[i])) != DC_SCRATCH)
793 cnt += var_get_value_cnt (d->var[i]);
797 /* Creates and returns an array mapping from a dictionary index
798 to the case index that the corresponding variable will have
799 after calling dict_compact_values(). Scratch variables
800 receive -1 for case index because dict_compact_values() will
803 dict_get_compacted_dict_index_to_case_index (const struct dictionary *d)
806 size_t next_value_idx;
809 map = xnmalloc (d->var_cnt, sizeof *map);
811 for (i = 0; i < d->var_cnt; i++)
813 struct variable *v = d->var[i];
815 if (dict_class_from_id (var_get_name (v)) != DC_SCRATCH)
817 map[i] = next_value_idx;
818 next_value_idx += var_get_value_cnt (v);
826 /* Returns true if a case for dictionary D would be smaller after
827 compacting, false otherwise. Compacting a case eliminates
828 "holes" between values and after the last value. Holes are
829 created by deleting variables (or by scratch variables).
831 The return value may differ from whether compacting a case
832 from dictionary D would *change* the case: compacting could
833 rearrange values even if it didn't reduce space
836 dict_compacting_would_shrink (const struct dictionary *d)
838 return dict_get_compacted_value_cnt (d) < dict_get_next_value_idx (d);
841 /* Returns true if a case for dictionary D would change after
842 compacting, false otherwise. Compacting a case eliminates
843 "holes" between values and after the last value. Holes are
844 created by deleting variables (or by scratch variables).
846 The return value may differ from whether compacting a case
847 from dictionary D would *shrink* the case: compacting could
848 rearrange values without reducing space requirements. */
850 dict_compacting_would_change (const struct dictionary *d)
856 for (i = 0; i < dict_get_var_cnt (d); i++)
858 struct variable *v = dict_get_var (d, i);
859 if (var_get_case_index (v) != case_idx)
861 case_idx += var_get_value_cnt (v);
866 /* How to copy a contiguous range of values between cases. */
869 size_t src_idx; /* Starting value index in source case. */
870 size_t dst_idx; /* Starting value index in target case. */
871 size_t cnt; /* Number of values. */
874 /* How to compact a case. */
875 struct dict_compactor
877 struct copy_map *maps; /* Array of mappings. */
878 size_t map_cnt; /* Number of mappings. */
881 /* Creates and returns a dict_compactor that can be used to
882 compact cases for dictionary D.
884 Compacting a case eliminates "holes" between values and after
885 the last value. Holes are created by deleting variables (or
886 by scratch variables). */
887 struct dict_compactor *
888 dict_make_compactor (const struct dictionary *d)
890 struct dict_compactor *compactor;
891 struct copy_map *map;
892 size_t map_allocated;
896 compactor = xmalloc (sizeof *compactor);
897 compactor->maps = NULL;
898 compactor->map_cnt = 0;
903 for (i = 0; i < d->var_cnt; i++)
905 struct variable *v = d->var[i];
907 if (dict_class_from_id (var_get_name (v)) == DC_SCRATCH)
909 if (map != NULL && map->src_idx + map->cnt == var_get_case_index (v))
910 map->cnt += var_get_value_cnt (v);
913 if (compactor->map_cnt == map_allocated)
914 compactor->maps = x2nrealloc (compactor->maps, &map_allocated,
915 sizeof *compactor->maps);
916 map = &compactor->maps[compactor->map_cnt++];
917 map->src_idx = var_get_case_index (v);
918 map->dst_idx = value_idx;
919 map->cnt = var_get_value_cnt (v);
921 value_idx += var_get_value_cnt (v);
927 /* Compacts SRC by copying it to DST according to the scheme in
930 Compacting a case eliminates "holes" between values and after
931 the last value. Holes are created by deleting variables (or
932 by scratch variables). */
934 dict_compactor_compact (const struct dict_compactor *compactor,
935 struct ccase *dst, const struct ccase *src)
939 for (i = 0; i < compactor->map_cnt; i++)
941 const struct copy_map *map = &compactor->maps[i];
942 case_copy (dst, map->dst_idx, src, map->src_idx, map->cnt);
946 /* Destroys COMPACTOR. */
948 dict_compactor_destroy (struct dict_compactor *compactor)
950 if (compactor != NULL)
952 free (compactor->maps);
957 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
958 dict_get_split_cnt() to determine how many SPLIT FILE vars
959 there are. Returns a null pointer if and only if there are no
961 struct variable *const *
962 dict_get_split_vars (const struct dictionary *d)
969 /* Returns the number of SPLIT FILE vars. */
971 dict_get_split_cnt (const struct dictionary *d)
978 /* Sets CNT split vars SPLIT in dictionary D. */
980 dict_set_split_vars (struct dictionary *d,
981 struct variable *const *split, size_t cnt)
984 assert (cnt == 0 || split != NULL);
987 d->split = xnrealloc (d->split, cnt, sizeof *d->split);
988 memcpy (d->split, split, cnt * sizeof *d->split);
991 /* Returns the file label for D, or a null pointer if D is
992 unlabeled (see cmd_file_label()). */
994 dict_get_label (const struct dictionary *d)
1001 /* Sets D's file label to LABEL, truncating it to a maximum of 60
1004 dict_set_label (struct dictionary *d, const char *label)
1011 else if (strlen (label) < 60)
1012 d->label = xstrdup (label);
1015 d->label = xmalloc (61);
1016 memcpy (d->label, label, 60);
1017 d->label[60] = '\0';
1021 /* Returns the documents for D, or a null pointer if D has no
1022 documents (see cmd_document()).. */
1024 dict_get_documents (const struct dictionary *d)
1028 return d->documents;
1031 /* Sets the documents for D to DOCUMENTS, or removes D's
1032 documents if DOCUMENT is a null pointer. */
1034 dict_set_documents (struct dictionary *d, const char *documents)
1038 free (d->documents);
1039 if (documents == NULL)
1040 d->documents = NULL;
1042 d->documents = xstrdup (documents);
1045 /* Creates in D a vector named NAME that contains the CNT
1046 variables in VAR. Returns true if successful, or false if a
1047 vector named NAME already exists in D. */
1049 dict_create_vector (struct dictionary *d,
1051 struct variable **var, size_t cnt)
1055 assert (var != NULL);
1057 for (i = 0; i < cnt; i++)
1058 assert (dict_contains_var (d, var[i]));
1060 if (dict_lookup_vector (d, name) == NULL)
1062 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1063 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1070 /* Returns the vector in D with index IDX, which must be less
1071 than dict_get_vector_cnt (D). */
1072 const struct vector *
1073 dict_get_vector (const struct dictionary *d, size_t idx)
1076 assert (idx < d->vector_cnt);
1078 return d->vector[idx];
1081 /* Returns the number of vectors in D. */
1083 dict_get_vector_cnt (const struct dictionary *d)
1087 return d->vector_cnt;
1090 /* Looks up and returns the vector within D with the given
1092 const struct vector *
1093 dict_lookup_vector (const struct dictionary *d, const char *name)
1096 for (i = 0; i < d->vector_cnt; i++)
1097 if (!strcasecmp (vector_get_name (d->vector[i]), name))
1098 return d->vector[i];
1102 /* Deletes all vectors from D. */
1104 dict_clear_vectors (struct dictionary *d)
1108 for (i = 0; i < d->vector_cnt; i++)
1109 vector_destroy (d->vector[i]);
1116 /* Compares two strings. */
1118 compare_strings (const void *a, const void *b, const void *aux UNUSED)
1120 return strcmp (a, b);
1123 /* Hashes a string. */
1125 hash_string (const void *s, const void *aux UNUSED)
1127 return hsh_hash_string (s);
1131 /* Sets V's short name to BASE, followed by a suffix of the form
1132 _A, _B, _C, ..., _AA, _AB, etc. according to the value of
1133 SUFFIX_NUMBER. Truncates BASE as necessary to fit. */
1135 set_var_short_name_suffix (struct variable *v, const char *base,
1138 char suffix[SHORT_NAME_LEN + 1];
1139 char short_name[SHORT_NAME_LEN + 1];
1144 assert (suffix_number >= 0);
1146 /* Set base name. */
1147 var_set_short_name (v, base);
1149 /* Compose suffix. */
1150 start = end = suffix + sizeof suffix - 1;
1154 *--start = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"[suffix_number % 26];
1155 if (start <= suffix + 1)
1156 msg (SE, _("Variable suffix too large."));
1157 suffix_number /= 26;
1159 while (suffix_number > 0);
1162 /* Append suffix to V's short name. */
1163 str_copy_trunc (short_name, sizeof short_name, base);
1165 if (len + strlen (short_name) > SHORT_NAME_LEN)
1166 ofs = SHORT_NAME_LEN - len;
1168 ofs = strlen (short_name);
1169 strcpy (short_name + ofs, start);
1172 var_set_short_name (v, short_name);
1175 /* Assigns a valid, unique short_name[] to each variable in D.
1176 Each variable whose actual name is short has highest priority
1177 for that short name. Otherwise, variables with an existing
1178 short_name[] have the next highest priority for a given short
1179 name; if it is already taken, then the variable is treated as
1180 if short_name[] had been empty. Otherwise, long names are
1181 truncated to form short names. If that causes conflicts,
1182 variables are renamed as PREFIX_A, PREFIX_B, and so on. */
1184 dict_assign_short_names (struct dictionary *d)
1186 struct hsh_table *short_names;
1189 /* Give variables whose names are short the corresponding short
1190 names, and clear short_names[] that conflict with a variable
1192 for (i = 0; i < d->var_cnt; i++)
1194 struct variable *v = d->var[i];
1195 const char *short_name = var_get_short_name (v);
1196 if (strlen (var_get_name (v)) <= SHORT_NAME_LEN)
1197 var_set_short_name (v, var_get_name (v));
1198 else if (short_name != NULL && dict_lookup_var (d, short_name) != NULL)
1199 var_clear_short_name (v);
1202 /* Each variable with an assigned short_name[] now gets it
1203 unless there is a conflict. */
1204 short_names = hsh_create (d->var_cnt, compare_strings, hash_string,
1206 for (i = 0; i < d->var_cnt; i++)
1208 struct variable *v = d->var[i];
1209 const char *name = var_get_short_name (v);
1210 if (name != NULL && hsh_insert (short_names, (char *) name) != NULL)
1211 var_clear_short_name (v);
1214 /* Now assign short names to remaining variables. */
1215 for (i = 0; i < d->var_cnt; i++)
1217 struct variable *v = d->var[i];
1218 const char *name = var_get_short_name (v);
1221 /* Form initial short_name from the variable name, then
1222 try _A, _B, ... _AA, _AB, etc., if needed.*/
1227 var_set_short_name (v, var_get_name (v));
1229 set_var_short_name_suffix (v, var_get_name (v), trial - 1);
1233 while (hsh_insert (short_names, (char *) var_get_short_name (v))
1238 /* Get rid of hash table. */
1239 hsh_destroy (short_names);