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 = xmalloc (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 = xmalloc (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 = xmalloc (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 = xmalloc (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 (strlen (name) >= 1);
264 assert (strlen (name) <= LONG_NAME_LEN);
266 assert (width >= 0 && width < 256);
268 /* Make sure there's not already a variable by that name. */
269 if (dict_lookup_var (d, name) != NULL)
272 /* Allocate and initialize variable. */
273 v = xmalloc (sizeof *v);
274 str_copy_trunc (v->name, sizeof v->name, name);
275 v->type = width == 0 ? NUMERIC : ALPHA;
277 v->fv = d->next_value_idx;
278 v->nv = width == 0 ? 1 : DIV_RND_UP (width, 8);
280 v->reinit = dict_class_from_id (v->name) != DC_SCRATCH;
281 v->index = d->var_cnt;
282 v->miss_type = MISSING_NONE;
283 if (v->type == NUMERIC)
286 v->alignment = ALIGN_RIGHT;
287 v->display_width = 8;
288 v->measure = MEASURE_SCALE;
292 v->print = make_output_format (FMT_A, v->width, 0);
293 v->alignment = ALIGN_LEFT;
294 v->display_width = 8;
295 v->measure = MEASURE_NOMINAL;
298 v->val_labs = val_labs_create (v->width);
300 var_clear_short_name (v);
304 /* Update dictionary. */
305 if (d->var_cnt >= d->var_cap)
307 d->var_cap = 8 + 2 * d->var_cap;
308 d->var = xrealloc (d->var, d->var_cap * sizeof *d->var);
310 d->var[v->index] = v;
312 hsh_force_insert (d->name_tab, v);
314 d->next_value_idx += v->nv;
319 /* Creates and returns a new variable in D with the given NAME
320 and WIDTH. Assert-fails if the given NAME would duplicate
321 that of an existing variable in the dictionary. */
323 dict_create_var_assert (struct dictionary *d, const char *name, int width)
325 struct variable *v = dict_create_var (d, name, width);
330 /* Creates and returns a new variable in D with name NAME, as a
331 copy of existing variable OV, which need not be in D or in any
332 dictionary. Returns a null pointer if the given NAME would
333 duplicate that of an existing variable in the dictionary. */
335 dict_clone_var (struct dictionary *d, const struct variable *ov,
342 assert (name != NULL);
344 assert (strlen (name) >= 1);
345 assert (strlen (name) <= LONG_NAME_LEN);
347 nv = dict_create_var (d, name, ov->width);
351 /* Copy most members not copied via dict_create_var().
352 short_name[] is intentionally not copied, because there is
353 no reason to give a new variable with potentially a new name
354 the same short name. */
356 nv->reinit = ov->reinit;
357 nv->miss_type = ov->miss_type;
358 memcpy (nv->missing, ov->missing, sizeof nv->missing);
359 nv->print = ov->print;
360 nv->write = ov->write;
361 val_labs_destroy (nv->val_labs);
362 nv->val_labs = val_labs_copy (ov->val_labs);
363 if (ov->label != NULL)
364 nv->label = xstrdup (ov->label);
365 nv->measure = ov->measure;
366 nv->display_width = ov->display_width;
367 nv->alignment = ov->alignment;
372 /* Creates and returns a new variable in D with name NAME, as a
373 copy of existing variable OV, which need not be in D or in any
374 dictionary. Assert-fails if the given NAME would duplicate
375 that of an existing variable in the dictionary. */
377 dict_clone_var_assert (struct dictionary *d, const struct variable *ov,
380 struct variable *v = dict_clone_var (d, ov, name);
385 /* Returns the variable named NAME in D, or a null pointer if no
386 variable has that name. */
388 dict_lookup_var (const struct dictionary *d, const char *name)
393 assert (name != NULL);
395 str_copy_trunc (v.name, sizeof v.name, name);
396 return hsh_find (d->name_tab, &v);
399 /* Returns the variable named NAME in D. Assert-fails if no
400 variable has that name. */
402 dict_lookup_var_assert (const struct dictionary *d, const char *name)
404 struct variable *v = dict_lookup_var (d, name);
409 /* Returns nonzero if variable V is in dictionary D. */
411 dict_contains_var (const struct dictionary *d, const struct variable *v)
416 return v->index >= 0 && v->index < d->var_cnt && d->var[v->index] == v;
419 /* Compares two double pointers to variables, which should point
420 to elements of a struct dictionary's `var' member array. */
422 compare_var_ptrs (const void *a_, const void *b_, void *aux UNUSED)
424 struct variable *const *a = a_;
425 struct variable *const *b = b_;
427 return *a < *b ? -1 : *a > *b;
430 /* Deletes variable V from dictionary D and frees V.
432 This is a very bad idea if there might be any pointers to V
433 from outside D. In general, no variable in default_dict
434 should be deleted when any transformations are active, because
435 those transformations might reference the deleted variable.
436 The safest time to delete a variable is just after a procedure
437 has been executed, as done by MODIFY VARS.
439 Pointers to V within D are not a problem, because
440 dict_delete_var() knows to remove V from split variables,
441 weights, filters, etc. */
443 dict_delete_var (struct dictionary *d, struct variable *v)
449 assert (dict_contains_var (d, v));
451 /* Delete aux data. */
454 /* Remove V from splits, weight, filter variables. */
455 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
456 &v, compare_var_ptrs, NULL);
461 dict_clear_vectors (d);
463 /* Remove V from var array. */
464 remove_element (d->var, d->var_cnt, sizeof *d->var, v->index);
468 for (i = v->index; i < d->var_cnt; i++)
469 d->var[i]->index = i;
471 /* Update name hash. */
472 hsh_force_delete (d->name_tab, v);
475 val_labs_destroy (v->val_labs);
480 /* Deletes the COUNT variables listed in VARS from D. This is
481 unsafe; see the comment on dict_delete_var() for details. */
483 dict_delete_vars (struct dictionary *d,
484 struct variable *const *vars, size_t count)
486 /* FIXME: this can be done in O(count) time, but this algorithm
489 assert (count == 0 || vars != NULL);
492 dict_delete_var (d, *vars++);
495 /* Moves V to 0-based position IDX in D. Other variables in D,
496 if any, retain their relative positions. Runs in time linear
497 in the distance moved. */
499 dict_reorder_var (struct dictionary *d, struct variable *v,
502 size_t min_idx, max_idx;
507 assert (dict_contains_var (d, v));
508 assert (new_index < d->var_cnt);
510 move_element (d->var, d->var_cnt, sizeof *d->var, v->index, new_index);
512 min_idx = min (v->index, new_index);
513 max_idx = max (v->index, new_index);
514 for (i = min_idx; i <= max_idx; i++)
515 d->var[i]->index = i;
518 /* Reorders the variables in D, placing the COUNT variables
519 listed in ORDER in that order at the beginning of D. The
520 other variables in D, if any, retain their relative
523 dict_reorder_vars (struct dictionary *d,
524 struct variable *const *order, size_t count)
526 struct variable **new_var;
530 assert (count == 0 || order != NULL);
531 assert (count <= d->var_cnt);
533 new_var = xmalloc (d->var_cnt * sizeof *new_var);
534 memcpy (new_var, order, count * sizeof *new_var);
535 for (i = 0; i < count; i++)
537 assert (d->var[order[i]->index] != NULL);
538 d->var[order[i]->index] = NULL;
541 for (i = 0; i < d->var_cnt; i++)
542 if (d->var[i] != NULL)
544 assert (count < d->var_cnt);
545 new_var[count] = d->var[i];
546 new_var[count]->index = count;
553 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
554 a variable named NEW_NAME is already in D, except that
555 NEW_NAME may be the same as V's existing name. */
557 dict_rename_var (struct dictionary *d, struct variable *v,
558 const char *new_name)
562 assert (new_name != NULL);
563 assert (var_is_valid_name (new_name, false));
564 assert (dict_contains_var (d, v));
565 assert (!compare_var_names (v->name, new_name, NULL)
566 || dict_lookup_var (d, new_name) == NULL);
568 hsh_force_delete (d->name_tab, v);
569 str_copy_trunc (v->name, sizeof v->name, new_name);
570 hsh_force_insert (d->name_tab, v);
572 if (get_algorithm () == ENHANCED)
573 var_clear_short_name (v);
576 /* Renames COUNT variables specified in VARS to the names given
577 in NEW_NAMES within dictionary D. If the renaming would
578 result in a duplicate variable name, returns zero and stores a
579 name that would be duplicated into *ERR_NAME (if ERR_NAME is
580 non-null). Otherwise, the renaming is successful, and nonzero
583 dict_rename_vars (struct dictionary *d,
584 struct variable **vars, char **new_names,
585 size_t count, char **err_name)
592 assert (count == 0 || vars != NULL);
593 assert (count == 0 || new_names != NULL);
595 /* Remove the variables to be renamed from the name hash,
596 save their names, and rename them. */
597 old_names = xmalloc (count * sizeof *old_names);
598 for (i = 0; i < count; i++)
600 assert (d->var[vars[i]->index] == vars[i]);
601 assert (var_is_valid_name (new_names[i], false));
602 hsh_force_delete (d->name_tab, vars[i]);
603 old_names[i] = xstrdup (vars[i]->name);
604 strcpy (vars[i]->name, new_names[i]);
607 /* Add the renamed variables back into the name hash,
608 checking for conflicts. */
609 for (i = 0; i < count; i++)
611 assert (new_names[i] != NULL);
612 assert (*new_names[i] != '\0');
613 assert (strlen (new_names[i]) >= 1);
614 assert (strlen (new_names[i]) <= LONG_NAME_LEN);
616 if (hsh_insert (d->name_tab, vars[i]) != NULL)
618 /* There is a name conflict.
619 Back out all the name changes that have already
620 taken place, and indicate failure. */
622 if (err_name != NULL)
623 *err_name = new_names[i];
625 for (i = 0; i < fail_idx; i++)
626 hsh_force_delete (d->name_tab, vars[i]);
628 for (i = 0; i < count; i++)
630 strcpy (vars[i]->name, old_names[i]);
631 hsh_force_insert (d->name_tab, vars[i]);
639 /* Clear short names. */
640 if (get_algorithm () == ENHANCED)
641 for (i = 0; i < count; i++)
642 var_clear_short_name (vars[i]);
645 /* Free the old names we kept around. */
646 for (i = 0; i < count; i++)
653 /* Returns the weighting variable in dictionary D, or a null
654 pointer if the dictionary is unweighted. */
656 dict_get_weight (const struct dictionary *d)
659 assert (d->weight == NULL || dict_contains_var (d, d->weight));
664 /* Returns the value of D's weighting variable in case C, except that a
665 negative weight is returned as 0. Returns 1 if the dictionary is
666 unweighted. Will warn about missing, negative, or zero values if
667 warn_on_invalid is nonzero. The function will set warn_on_invalid to zero
668 if an invalid weight is found. */
670 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
671 int *warn_on_invalid)
676 if (d->weight == NULL)
680 double w = case_num (c, d->weight->fv);
681 if ( w < 0.0 || w == SYSMIS || is_num_user_missing(w, d->weight) )
683 if ( w == 0.0 && *warn_on_invalid ) {
684 *warn_on_invalid = 0;
685 msg (SW, _("At least one case in the data file had a weight value "
686 "that was user-missing, system-missing, zero, or "
687 "negative. These case(s) were ignored."));
693 /* Sets the weighting variable of D to V, or turning off
694 weighting if V is a null pointer. */
696 dict_set_weight (struct dictionary *d, struct variable *v)
699 assert (v == NULL || dict_contains_var (d, v));
700 assert (v == NULL || v->type == NUMERIC);
705 /* Returns the filter variable in dictionary D (see cmd_filter())
706 or a null pointer if the dictionary is unfiltered. */
708 dict_get_filter (const struct dictionary *d)
711 assert (d->filter == NULL || dict_contains_var (d, d->filter));
716 /* Sets V as the filter variable for dictionary D. Passing a
717 null pointer for V turn off filtering. */
719 dict_set_filter (struct dictionary *d, struct variable *v)
722 assert (v == NULL || dict_contains_var (d, v));
727 /* Returns the case limit for dictionary D, or zero if the number
728 of cases is unlimited (see cmd_n()). */
730 dict_get_case_limit (const struct dictionary *d)
734 return d->case_limit;
737 /* Sets CASE_LIMIT as the case limit for dictionary D. Zero for
738 CASE_LIMIT indicates no limit. */
740 dict_set_case_limit (struct dictionary *d, int case_limit)
743 assert (case_limit >= 0);
745 d->case_limit = case_limit;
748 /* Returns the index of the next value to be added to D. This
749 value is the number of `union value's that need to be
750 allocated to store a case for dictionary D. */
752 dict_get_next_value_idx (const struct dictionary *d)
756 return d->next_value_idx;
759 /* Returns the number of bytes needed to store a case for
762 dict_get_case_size (const struct dictionary *d)
766 return sizeof (union value) * dict_get_next_value_idx (d);
769 /* Deletes scratch variables in dictionary D and reassigns values
770 so that fragmentation is eliminated. */
772 dict_compact_values (struct dictionary *d)
776 d->next_value_idx = 0;
777 for (i = 0; i < d->var_cnt; )
779 struct variable *v = d->var[i];
781 if (dict_class_from_id (v->name) != DC_SCRATCH)
783 v->fv = d->next_value_idx;
784 d->next_value_idx += v->nv;
788 dict_delete_var (d, v);
792 /* Copies values from SRC, which represents a case arranged
793 according to dictionary D, to DST, which represents a case
794 arranged according to the dictionary that will be produced by
795 dict_compact_values(D). */
797 dict_compact_case (const struct dictionary *d,
798 struct ccase *dst, const struct ccase *src)
804 for (i = 0; i < d->var_cnt; i++)
806 struct variable *v = d->var[i];
808 if (dict_class_from_id (v->name) != DC_SCRATCH)
810 case_copy (dst, value_idx, src, v->fv, v->nv);
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 = xmalloc (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 the SPLIT FILE vars (see cmd_split_file()). Call
860 dict_get_split_cnt() to determine how many SPLIT FILE vars
861 there are. Returns a null pointer if and only if there are no
863 struct variable *const *
864 dict_get_split_vars (const struct dictionary *d)
871 /* Returns the number of SPLIT FILE vars. */
873 dict_get_split_cnt (const struct dictionary *d)
880 /* Sets CNT split vars SPLIT in dictionary D. */
882 dict_set_split_vars (struct dictionary *d,
883 struct variable *const *split, size_t cnt)
886 assert (cnt == 0 || split != NULL);
889 d->split = xrealloc (d->split, cnt * sizeof *d->split);
890 memcpy (d->split, split, cnt * sizeof *d->split);
893 /* Returns the file label for D, or a null pointer if D is
894 unlabeled (see cmd_file_label()). */
896 dict_get_label (const struct dictionary *d)
903 /* Sets D's file label to LABEL, truncating it to a maximum of 60
906 dict_set_label (struct dictionary *d, const char *label)
913 else if (strlen (label) < 60)
914 d->label = xstrdup (label);
917 d->label = xmalloc (61);
918 memcpy (d->label, label, 60);
923 /* Returns the documents for D, or a null pointer if D has no
924 documents (see cmd_document()).. */
926 dict_get_documents (const struct dictionary *d)
933 /* Sets the documents for D to DOCUMENTS, or removes D's
934 documents if DOCUMENT is a null pointer. */
936 dict_set_documents (struct dictionary *d, const char *documents)
941 if (documents == NULL)
944 d->documents = xstrdup (documents);
947 /* Creates in D a vector named NAME that contains CNT variables
948 VAR (see cmd_vector()). Returns nonzero if successful, or
949 zero if a vector named NAME already exists in D. */
951 dict_create_vector (struct dictionary *d,
953 struct variable **var, size_t cnt)
955 struct vector *vector;
959 assert (name != NULL);
960 assert (var_is_valid_name (name, false));
961 assert (var != NULL);
964 if (dict_lookup_vector (d, name) != NULL)
967 d->vector = xrealloc (d->vector, (d->vector_cnt + 1) * sizeof *d->vector);
968 vector = d->vector[d->vector_cnt] = xmalloc (sizeof *vector);
969 vector->idx = d->vector_cnt++;
970 str_copy_trunc (vector->name, sizeof vector->name, name);
971 vector->var = xmalloc (cnt * sizeof *var);
972 for (i = 0; i < cnt; i++)
974 assert (dict_contains_var (d, var[i]));
975 vector->var[i] = var[i];
982 /* Returns the vector in D with index IDX, which must be less
983 than dict_get_vector_cnt (D). */
984 const struct vector *
985 dict_get_vector (const struct dictionary *d, size_t idx)
988 assert (idx < d->vector_cnt);
990 return d->vector[idx];
993 /* Returns the number of vectors in D. */
995 dict_get_vector_cnt (const struct dictionary *d)
999 return d->vector_cnt;
1002 /* Looks up and returns the vector within D with the given
1004 const struct vector *
1005 dict_lookup_vector (const struct dictionary *d, const char *name)
1010 assert (name != NULL);
1012 for (i = 0; i < d->vector_cnt; i++)
1013 if (!strcasecmp (d->vector[i]->name, name))
1014 return d->vector[i];
1018 /* Deletes all vectors from D. */
1020 dict_clear_vectors (struct dictionary *d)
1026 for (i = 0; i < d->vector_cnt; i++)
1028 free (d->vector[i]->var);
1029 free (d->vector[i]);
1036 /* Compares two strings. */
1038 compare_strings (const void *a, const void *b, void *aux UNUSED)
1040 return strcmp (a, b);
1043 /* Hashes a string. */
1045 hash_string (const void *s, void *aux UNUSED)
1047 return hsh_hash_string (s);
1050 /* Assigns a valid, unique short_name[] to each variable in D.
1051 Each variable whose actual name is short has highest priority
1052 for that short name. Otherwise, variables with an existing
1053 short_name[] have the next highest priority for a given short
1054 name; if it is already taken, then the variable is treated as
1055 if short_name[] had been empty. Otherwise, long names are
1056 truncated to form short names. If that causes conflicts,
1057 variables are renamed as PREFIX_A, PREFIX_B, and so on. */
1059 dict_assign_short_names (struct dictionary *d)
1061 struct hsh_table *short_names;
1064 /* Give variables whose names are short the corresponding short
1065 names, and clear short_names[] that conflict with a variable
1067 for (i = 0; i < d->var_cnt; i++)
1069 struct variable *v = d->var[i];
1070 if (strlen (v->name) <= SHORT_NAME_LEN)
1071 var_set_short_name (v, v->name);
1072 else if (dict_lookup_var (d, v->short_name) != NULL)
1073 var_clear_short_name (v);
1076 /* Each variable with an assigned short_name[] now gets it
1077 unless there is a conflict. */
1078 short_names = hsh_create (d->var_cnt, compare_strings, hash_string,
1080 for (i = 0; i < d->var_cnt; i++)
1082 struct variable *v = d->var[i];
1083 if (v->short_name[0] && hsh_insert (short_names, v->short_name) != NULL)
1084 var_clear_short_name (v);
1087 /* Now assign short names to remaining variables. */
1088 for (i = 0; i < d->var_cnt; i++)
1090 struct variable *v = d->var[i];
1091 if (v->short_name[0] == '\0')
1095 /* Form initial short_name. */
1096 var_set_short_name (v, v->name);
1098 /* Try _A, _B, ... _AA, _AB, etc., if needed. */
1099 for (sfx = 0; hsh_insert (short_names, v->short_name) != NULL; sfx++)
1100 var_set_short_name_suffix (v, v->name, sfx);
1104 /* Get rid of hash table. */
1105 hsh_destroy (short_names);