1 /* PSPP - computes sample statistics.
2 Copyright (C) 1997-9, 2000 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"
32 #include "value-labels.h"
36 #define _(msgid) gettext (msgid)
41 struct variable **var; /* Variables. */
42 size_t var_cnt, var_cap; /* Number of variables, capacity. */
43 struct hsh_table *name_tab; /* Variable index by name. */
44 int next_value_idx; /* Index of next `union value' to allocate. */
45 struct variable **split; /* SPLIT FILE vars. */
46 size_t split_cnt; /* SPLIT FILE count. */
47 struct variable *weight; /* WEIGHT variable. */
48 struct variable *filter; /* FILTER variable. */
49 int case_limit; /* Current case limit (N command). */
50 char *label; /* File label. */
51 char *documents; /* Documents, as a string. */
52 struct vector **vector; /* Vectors of variables. */
53 size_t vector_cnt; /* Number of vectors. */
56 /* Creates and returns a new dictionary. */
60 struct dictionary *d = xmalloc (sizeof *d);
63 d->var_cnt = d->var_cap = 0;
64 d->name_tab = hsh_create (8, compare_var_names, hash_var_name, NULL, NULL);
65 d->next_value_idx = 0;
79 /* Creates and returns a (deep) copy of an existing
82 dict_clone (const struct dictionary *s)
91 for (i = 0; i < s->var_cnt; i++)
93 struct variable *sv = s->var[i];
94 struct variable *dv = dict_clone_var_assert (d, sv, sv->name);
95 var_set_short_name (dv, sv->short_name);
98 d->next_value_idx = s->next_value_idx;
100 d->split_cnt = s->split_cnt;
101 if (d->split_cnt > 0)
103 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
104 for (i = 0; i < d->split_cnt; i++)
105 d->split[i] = dict_lookup_var_assert (d, s->split[i]->name);
108 if (s->weight != NULL)
109 d->weight = dict_lookup_var_assert (d, s->weight->name);
111 if (s->filter != NULL)
112 d->filter = dict_lookup_var_assert (d, s->filter->name);
114 d->case_limit = s->case_limit;
115 dict_set_label (d, dict_get_label (s));
116 dict_set_documents (d, dict_get_documents (s));
118 d->vector_cnt = s->vector_cnt;
119 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
120 for (i = 0; i < s->vector_cnt; i++)
122 struct vector *sv = s->vector[i];
123 struct vector *dv = d->vector[i] = xmalloc (sizeof *dv);
127 strcpy (dv->name, sv->name);
129 dv->var = xnmalloc (dv->cnt, sizeof *dv->var);
130 for (j = 0; j < dv->cnt; j++)
131 dv->var[j] = d->var[sv->var[j]->index];
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 struct variable *v = d->var[i];
152 val_labs_destroy (v->val_labs);
158 d->var_cnt = d->var_cap = 0;
159 hsh_clear (d->name_tab);
160 d->next_value_idx = 0;
171 dict_clear_vectors (d);
174 /* Destroys the aux data for every variable in D, by calling
175 var_clear_aux() for each variable. */
177 dict_clear_aux (struct dictionary *d)
183 for (i = 0; i < d->var_cnt; i++)
184 var_clear_aux (d->var[i]);
187 /* Clears a dictionary and destroys it. */
189 dict_destroy (struct dictionary *d)
194 hsh_destroy (d->name_tab);
199 /* Returns the number of variables in D. */
201 dict_get_var_cnt (const struct dictionary *d)
208 /* Returns the variable in D with index IDX, which must be
209 between 0 and the count returned by dict_get_var_cnt(),
212 dict_get_var (const struct dictionary *d, size_t idx)
215 assert (idx < d->var_cnt);
220 /* Sets *VARS to an array of pointers to variables in D and *CNT
221 to the number of variables in *D. By default all variables
222 are returned, but bits may be set in EXCLUDE_CLASSES to
223 exclude ordinary, system, and/or scratch variables. */
225 dict_get_vars (const struct dictionary *d, struct variable ***vars,
226 size_t *cnt, unsigned exclude_classes)
232 assert (vars != NULL);
233 assert (cnt != NULL);
234 assert ((exclude_classes & ~((1u << DC_ORDINARY)
236 | (1u << DC_SCRATCH))) == 0);
239 for (i = 0; i < d->var_cnt; i++)
240 if (!(exclude_classes & (1u << dict_class_from_id (d->var[i]->name))))
243 *vars = xnmalloc (count, sizeof **vars);
245 for (i = 0; i < d->var_cnt; i++)
246 if (!(exclude_classes & (1u << dict_class_from_id (d->var[i]->name))))
247 (*vars)[(*cnt)++] = d->var[i];
248 assert (*cnt == count);
252 /* Creates and returns a new variable in D with the given NAME
253 and WIDTH. Returns a null pointer if the given NAME would
254 duplicate that of an existing variable in the dictionary. */
256 dict_create_var (struct dictionary *d, const char *name, int width)
261 assert (name != NULL);
263 assert (width >= 0 && width < 256);
265 assert (var_is_valid_name(name,0));
267 /* Make sure there's not already a variable by that name. */
268 if (dict_lookup_var (d, name) != NULL)
271 /* Allocate and initialize variable. */
272 v = xmalloc (sizeof *v);
273 str_copy_trunc (v->name, sizeof v->name, name);
274 v->type = width == 0 ? NUMERIC : ALPHA;
276 v->fv = d->next_value_idx;
277 v->nv = width == 0 ? 1 : DIV_RND_UP (width, 8);
279 v->reinit = dict_class_from_id (v->name) != DC_SCRATCH;
280 v->index = d->var_cnt;
281 mv_init (&v->miss, width);
282 if (v->type == NUMERIC)
285 v->alignment = ALIGN_RIGHT;
286 v->display_width = 8;
287 v->measure = MEASURE_SCALE;
291 v->print = make_output_format (FMT_A, v->width, 0);
292 v->alignment = ALIGN_LEFT;
293 v->display_width = 8;
294 v->measure = MEASURE_NOMINAL;
297 v->val_labs = val_labs_create (v->width);
299 var_clear_short_name (v);
303 /* Update dictionary. */
304 if (d->var_cnt >= d->var_cap)
306 d->var_cap = 8 + 2 * d->var_cap;
307 d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var);
309 d->var[v->index] = v;
311 hsh_force_insert (d->name_tab, v);
313 d->next_value_idx += v->nv;
318 /* Creates and returns a new variable in D with the given NAME
319 and WIDTH. Assert-fails if the given NAME would duplicate
320 that of an existing variable in the dictionary. */
322 dict_create_var_assert (struct dictionary *d, const char *name, int width)
324 struct variable *v = dict_create_var (d, name, width);
329 /* Creates and returns a new variable in D with name NAME, as a
330 copy of existing variable OV, which need not be in D or in any
331 dictionary. Returns a null pointer if the given NAME would
332 duplicate that of an existing variable in the dictionary. */
334 dict_clone_var (struct dictionary *d, const struct variable *ov,
341 assert (name != NULL);
343 assert (strlen (name) >= 1);
344 assert (strlen (name) <= LONG_NAME_LEN);
346 nv = dict_create_var (d, name, ov->width);
350 /* Copy most members not copied via dict_create_var().
351 short_name[] is intentionally not copied, because there is
352 no reason to give a new variable with potentially a new name
353 the same short name. */
355 nv->reinit = ov->reinit;
356 mv_copy (&nv->miss, &ov->miss);
357 nv->print = ov->print;
358 nv->write = ov->write;
359 val_labs_destroy (nv->val_labs);
360 nv->val_labs = val_labs_copy (ov->val_labs);
361 if (ov->label != NULL)
362 nv->label = xstrdup (ov->label);
363 nv->measure = ov->measure;
364 nv->display_width = ov->display_width;
365 nv->alignment = ov->alignment;
370 /* Creates and returns a new variable in D with name NAME, as a
371 copy of existing variable OV, which need not be in D or in any
372 dictionary. Assert-fails if the given NAME would duplicate
373 that of an existing variable in the dictionary. */
375 dict_clone_var_assert (struct dictionary *d, const struct variable *ov,
378 struct variable *v = dict_clone_var (d, ov, name);
383 /* Returns the variable named NAME in D, or a null pointer if no
384 variable has that name. */
386 dict_lookup_var (const struct dictionary *d, const char *name)
391 assert (name != NULL);
393 str_copy_trunc (v.name, sizeof v.name, name);
394 return hsh_find (d->name_tab, &v);
397 /* Returns the variable named NAME in D. Assert-fails if no
398 variable has that name. */
400 dict_lookup_var_assert (const struct dictionary *d, const char *name)
402 struct variable *v = dict_lookup_var (d, name);
407 /* Returns nonzero if variable V is in dictionary D. */
409 dict_contains_var (const struct dictionary *d, const struct variable *v)
414 return v->index >= 0 && v->index < d->var_cnt && d->var[v->index] == v;
417 /* Compares two double pointers to variables, which should point
418 to elements of a struct dictionary's `var' member array. */
420 compare_var_ptrs (const void *a_, const void *b_, void *aux UNUSED)
422 struct variable *const *a = a_;
423 struct variable *const *b = b_;
425 return *a < *b ? -1 : *a > *b;
428 /* Deletes variable V from dictionary D and frees V.
430 This is a very bad idea if there might be any pointers to V
431 from outside D. In general, no variable in default_dict
432 should be deleted when any transformations are active, because
433 those transformations might reference the deleted variable.
434 The safest time to delete a variable is just after a procedure
435 has been executed, as done by MODIFY VARS.
437 Pointers to V within D are not a problem, because
438 dict_delete_var() knows to remove V from split variables,
439 weights, filters, etc. */
441 dict_delete_var (struct dictionary *d, struct variable *v)
447 assert (dict_contains_var (d, v));
449 /* Delete aux data. */
452 /* Remove V from splits, weight, filter variables. */
453 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
454 &v, compare_var_ptrs, NULL);
459 dict_clear_vectors (d);
461 /* Remove V from var array. */
462 remove_element (d->var, d->var_cnt, sizeof *d->var, v->index);
466 for (i = v->index; i < d->var_cnt; i++)
467 d->var[i]->index = i;
469 /* Update name hash. */
470 hsh_force_delete (d->name_tab, v);
473 val_labs_destroy (v->val_labs);
478 /* Deletes the COUNT variables listed in VARS from D. This is
479 unsafe; see the comment on dict_delete_var() for details. */
481 dict_delete_vars (struct dictionary *d,
482 struct variable *const *vars, size_t count)
484 /* FIXME: this can be done in O(count) time, but this algorithm
487 assert (count == 0 || vars != NULL);
490 dict_delete_var (d, *vars++);
493 /* Deletes scratch variables from dictionary D. */
495 dict_delete_scratch_vars (struct dictionary *d)
499 /* FIXME: this can be done in O(count) time, but this algorithm
503 for (i = 0; i < d->var_cnt; )
504 if (dict_class_from_id (d->var[i]->name) == DC_SCRATCH)
505 dict_delete_var (d, d->var[i]);
510 /* Moves V to 0-based position IDX in D. Other variables in D,
511 if any, retain their relative positions. Runs in time linear
512 in the distance moved. */
514 dict_reorder_var (struct dictionary *d, struct variable *v,
517 size_t min_idx, max_idx;
522 assert (dict_contains_var (d, v));
523 assert (new_index < d->var_cnt);
525 move_element (d->var, d->var_cnt, sizeof *d->var, v->index, new_index);
527 min_idx = min (v->index, new_index);
528 max_idx = max (v->index, new_index);
529 for (i = min_idx; i <= max_idx; i++)
530 d->var[i]->index = i;
533 /* Reorders the variables in D, placing the COUNT variables
534 listed in ORDER in that order at the beginning of D. The
535 other variables in D, if any, retain their relative
538 dict_reorder_vars (struct dictionary *d,
539 struct variable *const *order, size_t count)
541 struct variable **new_var;
545 assert (count == 0 || order != NULL);
546 assert (count <= d->var_cnt);
548 new_var = xnmalloc (d->var_cnt, sizeof *new_var);
549 memcpy (new_var, order, count * sizeof *new_var);
550 for (i = 0; i < count; i++)
552 assert (d->var[order[i]->index] != NULL);
553 d->var[order[i]->index] = NULL;
556 for (i = 0; i < d->var_cnt; i++)
557 if (d->var[i] != NULL)
559 assert (count < d->var_cnt);
560 new_var[count] = d->var[i];
561 new_var[count]->index = count;
568 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
569 a variable named NEW_NAME is already in D, except that
570 NEW_NAME may be the same as V's existing name. */
572 dict_rename_var (struct dictionary *d, struct variable *v,
573 const char *new_name)
577 assert (new_name != NULL);
578 assert (var_is_valid_name (new_name, false));
579 assert (dict_contains_var (d, v));
580 assert (!compare_var_names (v->name, new_name, NULL)
581 || dict_lookup_var (d, new_name) == NULL);
583 hsh_force_delete (d->name_tab, v);
584 str_copy_trunc (v->name, sizeof v->name, new_name);
585 hsh_force_insert (d->name_tab, v);
587 if (get_algorithm () == ENHANCED)
588 var_clear_short_name (v);
591 /* Renames COUNT variables specified in VARS to the names given
592 in NEW_NAMES within dictionary D. If the renaming would
593 result in a duplicate variable name, returns zero and stores a
594 name that would be duplicated into *ERR_NAME (if ERR_NAME is
595 non-null). Otherwise, the renaming is successful, and nonzero
598 dict_rename_vars (struct dictionary *d,
599 struct variable **vars, char **new_names,
600 size_t count, char **err_name)
607 assert (count == 0 || vars != NULL);
608 assert (count == 0 || new_names != NULL);
610 /* Remove the variables to be renamed from the name hash,
611 save their names, and rename them. */
612 old_names = xnmalloc (count, sizeof *old_names);
613 for (i = 0; i < count; i++)
615 assert (d->var[vars[i]->index] == vars[i]);
616 assert (var_is_valid_name (new_names[i], false));
617 hsh_force_delete (d->name_tab, vars[i]);
618 old_names[i] = xstrdup (vars[i]->name);
619 strcpy (vars[i]->name, new_names[i]);
622 /* Add the renamed variables back into the name hash,
623 checking for conflicts. */
624 for (i = 0; i < count; i++)
626 assert (new_names[i] != NULL);
627 assert (*new_names[i] != '\0');
628 assert (strlen (new_names[i]) >= 1);
629 assert (strlen (new_names[i]) <= LONG_NAME_LEN);
631 if (hsh_insert (d->name_tab, vars[i]) != NULL)
633 /* There is a name conflict.
634 Back out all the name changes that have already
635 taken place, and indicate failure. */
637 if (err_name != NULL)
638 *err_name = new_names[i];
640 for (i = 0; i < fail_idx; i++)
641 hsh_force_delete (d->name_tab, vars[i]);
643 for (i = 0; i < count; i++)
645 strcpy (vars[i]->name, old_names[i]);
646 hsh_force_insert (d->name_tab, vars[i]);
654 /* Clear short names. */
655 if (get_algorithm () == ENHANCED)
656 for (i = 0; i < count; i++)
657 var_clear_short_name (vars[i]);
660 /* Free the old names we kept around. */
661 for (i = 0; i < count; i++)
668 /* Returns the weighting variable in dictionary D, or a null
669 pointer if the dictionary is unweighted. */
671 dict_get_weight (const struct dictionary *d)
674 assert (d->weight == NULL || dict_contains_var (d, d->weight));
679 /* Returns the value of D's weighting variable in case C, except that a
680 negative weight is returned as 0. Returns 1 if the dictionary is
681 unweighted. Will warn about missing, negative, or zero values if
682 warn_on_invalid is nonzero. The function will set warn_on_invalid to zero
683 if an invalid weight is found. */
685 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
686 int *warn_on_invalid)
691 if (d->weight == NULL)
695 double w = case_num (c, d->weight->fv);
696 if (w < 0.0 || mv_is_num_missing (&d->weight->miss, w))
698 if ( w == 0.0 && *warn_on_invalid ) {
699 *warn_on_invalid = 0;
700 msg (SW, _("At least one case in the data file had a weight value "
701 "that was user-missing, system-missing, zero, or "
702 "negative. These case(s) were ignored."));
708 /* Sets the weighting variable of D to V, or turning off
709 weighting if V is a null pointer. */
711 dict_set_weight (struct dictionary *d, struct variable *v)
714 assert (v == NULL || dict_contains_var (d, v));
715 assert (v == NULL || v->type == NUMERIC);
720 /* Returns the filter variable in dictionary D (see cmd_filter())
721 or a null pointer if the dictionary is unfiltered. */
723 dict_get_filter (const struct dictionary *d)
726 assert (d->filter == NULL || dict_contains_var (d, d->filter));
731 /* Sets V as the filter variable for dictionary D. Passing a
732 null pointer for V turn off filtering. */
734 dict_set_filter (struct dictionary *d, struct variable *v)
737 assert (v == NULL || dict_contains_var (d, v));
742 /* Returns the case limit for dictionary D, or zero if the number
743 of cases is unlimited (see cmd_n()). */
745 dict_get_case_limit (const struct dictionary *d)
749 return d->case_limit;
752 /* Sets CASE_LIMIT as the case limit for dictionary D. Zero for
753 CASE_LIMIT indicates no limit. */
755 dict_set_case_limit (struct dictionary *d, int case_limit)
758 assert (case_limit >= 0);
760 d->case_limit = case_limit;
763 /* Returns the index of the next value to be added to D. This
764 value is the number of `union value's that need to be
765 allocated to store a case for dictionary D. */
767 dict_get_next_value_idx (const struct dictionary *d)
771 return d->next_value_idx;
774 /* Returns the number of bytes needed to store a case for
777 dict_get_case_size (const struct dictionary *d)
781 return sizeof (union value) * dict_get_next_value_idx (d);
784 /* Deletes scratch variables in dictionary D and reassigns values
785 so that fragmentation is eliminated. */
787 dict_compact_values (struct dictionary *d)
791 d->next_value_idx = 0;
792 for (i = 0; i < d->var_cnt; )
794 struct variable *v = d->var[i];
796 if (dict_class_from_id (v->name) != DC_SCRATCH)
798 v->fv = d->next_value_idx;
799 d->next_value_idx += v->nv;
803 dict_delete_var (d, v);
807 /* Copies values from SRC, which represents a case arranged
808 according to dictionary D, to DST, which represents a case
809 arranged according to the dictionary that will be produced by
810 dict_compact_values(D). */
812 dict_compact_case (const struct dictionary *d,
813 struct ccase *dst, const struct ccase *src)
819 for (i = 0; i < d->var_cnt; i++)
821 struct variable *v = d->var[i];
823 if (dict_class_from_id (v->name) != DC_SCRATCH)
825 case_copy (dst, value_idx, src, v->fv, v->nv);
831 /* Returns the number of values that would be used by a case if
832 dict_compact_values() were called. */
834 dict_get_compacted_value_cnt (const struct dictionary *d)
840 for (i = 0; i < d->var_cnt; i++)
841 if (dict_class_from_id (d->var[i]->name) != DC_SCRATCH)
842 cnt += d->var[i]->nv;
846 /* Creates and returns an array mapping from a dictionary index
847 to the `fv' that the corresponding variable will have after
848 calling dict_compact_values(). Scratch variables receive -1
849 for `fv' because dict_compact_values() will delete them. */
851 dict_get_compacted_idx_to_fv (const struct dictionary *d)
854 size_t next_value_idx;
857 idx_to_fv = xnmalloc (d->var_cnt, sizeof *idx_to_fv);
859 for (i = 0; i < d->var_cnt; i++)
861 struct variable *v = d->var[i];
863 if (dict_class_from_id (v->name) != DC_SCRATCH)
865 idx_to_fv[i] = next_value_idx;
866 next_value_idx += v->nv;
874 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
875 dict_get_split_cnt() to determine how many SPLIT FILE vars
876 there are. Returns a null pointer if and only if there are no
878 struct variable *const *
879 dict_get_split_vars (const struct dictionary *d)
886 /* Returns the number of SPLIT FILE vars. */
888 dict_get_split_cnt (const struct dictionary *d)
895 /* Sets CNT split vars SPLIT in dictionary D. */
897 dict_set_split_vars (struct dictionary *d,
898 struct variable *const *split, size_t cnt)
901 assert (cnt == 0 || split != NULL);
904 d->split = xnrealloc (d->split, cnt, sizeof *d->split);
905 memcpy (d->split, split, cnt * sizeof *d->split);
908 /* Returns the file label for D, or a null pointer if D is
909 unlabeled (see cmd_file_label()). */
911 dict_get_label (const struct dictionary *d)
918 /* Sets D's file label to LABEL, truncating it to a maximum of 60
921 dict_set_label (struct dictionary *d, const char *label)
928 else if (strlen (label) < 60)
929 d->label = xstrdup (label);
932 d->label = xmalloc (61);
933 memcpy (d->label, label, 60);
938 /* Returns the documents for D, or a null pointer if D has no
939 documents (see cmd_document()).. */
941 dict_get_documents (const struct dictionary *d)
948 /* Sets the documents for D to DOCUMENTS, or removes D's
949 documents if DOCUMENT is a null pointer. */
951 dict_set_documents (struct dictionary *d, const char *documents)
956 if (documents == NULL)
959 d->documents = xstrdup (documents);
962 /* Creates in D a vector named NAME that contains CNT variables
963 VAR (see cmd_vector()). Returns nonzero if successful, or
964 zero if a vector named NAME already exists in D. */
966 dict_create_vector (struct dictionary *d,
968 struct variable **var, size_t cnt)
970 struct vector *vector;
974 assert (name != NULL);
975 assert (var_is_valid_name (name, false));
976 assert (var != NULL);
979 if (dict_lookup_vector (d, name) != NULL)
982 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
983 vector = d->vector[d->vector_cnt] = xmalloc (sizeof *vector);
984 vector->idx = d->vector_cnt++;
985 str_copy_trunc (vector->name, sizeof vector->name, name);
986 vector->var = xnmalloc (cnt, sizeof *var);
987 for (i = 0; i < cnt; i++)
989 assert (dict_contains_var (d, var[i]));
990 vector->var[i] = var[i];
997 /* Returns the vector in D with index IDX, which must be less
998 than dict_get_vector_cnt (D). */
999 const struct vector *
1000 dict_get_vector (const struct dictionary *d, size_t idx)
1003 assert (idx < d->vector_cnt);
1005 return d->vector[idx];
1008 /* Returns the number of vectors in D. */
1010 dict_get_vector_cnt (const struct dictionary *d)
1014 return d->vector_cnt;
1017 /* Looks up and returns the vector within D with the given
1019 const struct vector *
1020 dict_lookup_vector (const struct dictionary *d, const char *name)
1025 assert (name != NULL);
1027 for (i = 0; i < d->vector_cnt; i++)
1028 if (!strcasecmp (d->vector[i]->name, name))
1029 return d->vector[i];
1033 /* Deletes all vectors from D. */
1035 dict_clear_vectors (struct dictionary *d)
1041 for (i = 0; i < d->vector_cnt; i++)
1043 free (d->vector[i]->var);
1044 free (d->vector[i]);
1051 /* Compares two strings. */
1053 compare_strings (const void *a, const void *b, void *aux UNUSED)
1055 return strcmp (a, b);
1058 /* Hashes a string. */
1060 hash_string (const void *s, void *aux UNUSED)
1062 return hsh_hash_string (s);
1065 /* Assigns a valid, unique short_name[] to each variable in D.
1066 Each variable whose actual name is short has highest priority
1067 for that short name. Otherwise, variables with an existing
1068 short_name[] have the next highest priority for a given short
1069 name; if it is already taken, then the variable is treated as
1070 if short_name[] had been empty. Otherwise, long names are
1071 truncated to form short names. If that causes conflicts,
1072 variables are renamed as PREFIX_A, PREFIX_B, and so on. */
1074 dict_assign_short_names (struct dictionary *d)
1076 struct hsh_table *short_names;
1079 /* Give variables whose names are short the corresponding short
1080 names, and clear short_names[] that conflict with a variable
1082 for (i = 0; i < d->var_cnt; i++)
1084 struct variable *v = d->var[i];
1085 if (strlen (v->name) <= SHORT_NAME_LEN)
1086 var_set_short_name (v, v->name);
1087 else if (dict_lookup_var (d, v->short_name) != NULL)
1088 var_clear_short_name (v);
1091 /* Each variable with an assigned short_name[] now gets it
1092 unless there is a conflict. */
1093 short_names = hsh_create (d->var_cnt, compare_strings, hash_string,
1095 for (i = 0; i < d->var_cnt; i++)
1097 struct variable *v = d->var[i];
1098 if (v->short_name[0] && hsh_insert (short_names, v->short_name) != NULL)
1099 var_clear_short_name (v);
1102 /* Now assign short names to remaining variables. */
1103 for (i = 0; i < d->var_cnt; i++)
1105 struct variable *v = d->var[i];
1106 if (v->short_name[0] == '\0')
1110 /* Form initial short_name. */
1111 var_set_short_name (v, v->name);
1113 /* Try _A, _B, ... _AA, _AB, etc., if needed. */
1114 for (sfx = 0; hsh_insert (short_names, v->short_name) != NULL; sfx++)
1115 var_set_short_name_suffix (v, v->name, sfx);
1119 /* Get rid of hash table. */
1120 hsh_destroy (short_names);