1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006, 2007 Free Software Foundation, Inc.
4 This program is free software: you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation, either version 3 of the License, or
7 (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program. If not, see <http://www.gnu.org/licenses/>. */
19 #include "dictionary.h"
27 #include "value-labels.h"
31 #include <libpspp/alloc.h>
32 #include <libpspp/array.h>
33 #include <libpspp/compiler.h>
34 #include <libpspp/hash.h>
35 #include <libpspp/message.h>
36 #include <libpspp/misc.h>
37 #include <libpspp/pool.h>
38 #include <libpspp/str.h>
43 #define _(msgid) gettext (msgid)
48 struct variable **var; /* Variables. */
49 size_t var_cnt, var_cap; /* Number of variables, capacity. */
50 struct hsh_table *name_tab; /* Variable index by name. */
51 int next_value_idx; /* Index of next `union value' to allocate. */
52 const struct variable **split; /* SPLIT FILE vars. */
53 size_t split_cnt; /* SPLIT FILE count. */
54 struct variable *weight; /* WEIGHT variable. */
55 struct variable *filter; /* FILTER variable. */
56 size_t case_limit; /* Current case limit (N command). */
57 char *label; /* File label. */
58 struct string documents; /* Documents, as a string. */
59 struct vector **vector; /* Vectors of variables. */
60 size_t vector_cnt; /* Number of vectors. */
61 const struct dict_callbacks *callbacks; /* Callbacks on dictionary
63 void *cb_data ; /* Data passed to callbacks */
67 /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
68 the dictionary or any of the variables it contains are modified.
69 Each callback will get passed CALLBACK_DATA.
70 Any callback may be NULL, in which case it'll be ignored.
73 dict_set_callbacks (struct dictionary *dict,
74 const struct dict_callbacks *callbacks,
77 dict->callbacks = callbacks;
78 dict->cb_data = callback_data;
81 /* Shallow copy the callbacks from SRC to DEST */
83 dict_copy_callbacks (struct dictionary *dest,
84 const struct dictionary *src)
86 dest->callbacks = src->callbacks;
87 dest->cb_data = src->cb_data;
90 /* Creates and returns a new dictionary. */
94 struct dictionary *d = xzalloc (sizeof *d);
96 d->name_tab = hsh_create (8, compare_vars_by_name, hash_var_by_name,
101 /* Creates and returns a (deep) copy of an existing
104 dict_clone (const struct dictionary *s)
106 struct dictionary *d;
113 for (i = 0; i < s->var_cnt; i++)
115 struct variable *sv = s->var[i];
116 struct variable *dv = dict_clone_var_assert (d, sv, var_get_name (sv));
117 var_set_short_name (dv, var_get_short_name (sv));
120 d->next_value_idx = s->next_value_idx;
122 d->split_cnt = s->split_cnt;
123 if (d->split_cnt > 0)
125 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
126 for (i = 0; i < d->split_cnt; i++)
127 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
130 if (s->weight != NULL)
131 dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight)));
133 if (s->filter != NULL)
134 dict_set_filter (d, dict_lookup_var_assert (d, var_get_name (s->filter)));
136 d->case_limit = s->case_limit;
137 dict_set_label (d, dict_get_label (s));
138 dict_set_documents (d, dict_get_documents (s));
140 d->vector_cnt = s->vector_cnt;
141 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
142 for (i = 0; i < s->vector_cnt; i++)
143 d->vector[i] = vector_clone (s->vector[i], s, d);
148 /* Clears the contents from a dictionary without destroying the
149 dictionary itself. */
151 dict_clear (struct dictionary *d)
153 /* FIXME? Should we really clear case_limit, label, documents?
154 Others are necessarily cleared by deleting all the variables.*/
157 while (d->var_cnt > 0 )
159 struct variable *v = d->var[d->var_cnt - 1];
160 int dict_index = var_get_dict_index (v);
161 int case_index = var_get_case_index (v);
162 int val_cnt = var_get_value_cnt (v);
164 var_clear_vardict (v);
169 if (d->callbacks && d->callbacks->var_deleted )
170 d->callbacks->var_deleted (d,
171 dict_index, case_index, val_cnt,
177 d->var_cnt = d->var_cap = 0;
178 hsh_clear (d->name_tab);
179 d->next_value_idx = 0;
180 dict_set_split_vars (d, NULL, 0);
181 dict_set_weight (d, NULL);
182 dict_set_filter (d, NULL);
186 ds_destroy (&d->documents);
187 dict_clear_vectors (d);
190 /* Destroys the aux data for every variable in D, by calling
191 var_clear_aux() for each variable. */
193 dict_clear_aux (struct dictionary *d)
199 for (i = 0; i < d->var_cnt; i++)
200 var_clear_aux (d->var[i]);
203 /* Clears a dictionary and destroys it. */
205 dict_destroy (struct dictionary *d)
209 /* In general, we don't want callbacks occuring, if the dictionary
210 is being destroyed */
211 d->callbacks = NULL ;
214 hsh_destroy (d->name_tab);
219 /* Returns the number of variables in D. */
221 dict_get_var_cnt (const struct dictionary *d)
228 /* Returns the variable in D with dictionary index IDX, which
229 must be between 0 and the count returned by
230 dict_get_var_cnt(), exclusive. */
232 dict_get_var (const struct dictionary *d, size_t idx)
235 assert (idx < d->var_cnt);
241 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
242 size_t *cnt, unsigned exclude_classes)
244 dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude_classes);
247 /* Sets *VARS to an array of pointers to variables in D and *CNT
248 to the number of variables in *D. All variables are returned
249 if EXCLUDE_CLASSES is 0, or it may contain one or more of (1u
250 << DC_ORDINARY), (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to
251 exclude the corresponding type of variable. */
253 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
254 size_t *cnt, unsigned exclude_classes)
260 assert (vars != NULL);
261 assert (cnt != NULL);
262 assert ((exclude_classes & ~((1u << DC_ORDINARY)
264 | (1u << DC_SCRATCH))) == 0);
267 for (i = 0; i < d->var_cnt; i++)
269 enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
270 if (!(exclude_classes & (1u << class)))
274 *vars = xnmalloc (count, sizeof **vars);
276 for (i = 0; i < d->var_cnt; i++)
278 enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
279 if (!(exclude_classes & (1u << class)))
280 (*vars)[(*cnt)++] = d->var[i];
282 assert (*cnt == count);
285 static struct variable *
286 add_var (struct dictionary *d, struct variable *v)
288 /* Add dictionary info to variable. */
289 struct vardict_info vdi;
290 vdi.case_index = d->next_value_idx;
291 vdi.dict_index = d->var_cnt;
293 var_set_vardict (v, &vdi);
295 /* Update dictionary. */
296 if (d->var_cnt >= d->var_cap)
298 d->var_cap = 8 + 2 * d->var_cap;
299 d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var);
301 d->var[d->var_cnt++] = v;
302 hsh_force_insert (d->name_tab, v);
304 if ( d->callbacks && d->callbacks->var_added )
305 d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
307 d->next_value_idx += var_get_value_cnt (v);
312 /* Creates and returns a new variable in D with the given NAME
313 and WIDTH. Returns a null pointer if the given NAME would
314 duplicate that of an existing variable in the dictionary. */
316 dict_create_var (struct dictionary *d, const char *name, int width)
318 return (dict_lookup_var (d, name) == NULL
319 ? dict_create_var_assert (d, name, width)
323 /* Creates and returns a new variable in D with the given NAME
324 and WIDTH. Assert-fails if the given NAME would duplicate
325 that of an existing variable in the dictionary. */
327 dict_create_var_assert (struct dictionary *d, const char *name, int width)
329 assert (dict_lookup_var (d, name) == NULL);
330 return add_var (d, var_create (name, width));
333 /* Creates and returns a new variable in D with name NAME, as a
334 copy of existing variable OLD_VAR, which need not be in D or
335 in any dictionary. Returns a null pointer if the given NAME
336 would duplicate that of an existing variable in the
339 dict_clone_var (struct dictionary *d, const struct variable *old_var,
342 return (dict_lookup_var (d, name) == NULL
343 ? dict_clone_var_assert (d, old_var, name)
347 /* Creates and returns a new variable in D with name NAME, as a
348 copy of existing variable OLD_VAR, which need not be in D or
349 in any dictionary. Assert-fails if the given NAME would
350 duplicate that of an existing variable in the dictionary. */
352 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var,
355 struct variable *new_var = var_clone (old_var);
356 assert (dict_lookup_var (d, name) == NULL);
357 var_set_name (new_var, name);
358 return add_var (d, new_var);
361 /* Returns the variable named NAME in D, or a null pointer if no
362 variable has that name. */
364 dict_lookup_var (const struct dictionary *d, const char *name)
366 struct variable *target ;
367 struct variable *result ;
369 if ( ! var_is_valid_name (name, false))
372 target = var_create (name, 0);
373 result = hsh_find (d->name_tab, target);
374 var_destroy (target);
379 /* Returns the variable named NAME in D. Assert-fails if no
380 variable has that name. */
382 dict_lookup_var_assert (const struct dictionary *d, const char *name)
384 struct variable *v = dict_lookup_var (d, name);
389 /* Returns true if variable V is in dictionary D,
392 dict_contains_var (const struct dictionary *d, const struct variable *v)
394 if (var_has_vardict (v))
396 const struct vardict_info *vdi = var_get_vardict (v);
397 return (vdi->dict_index >= 0
398 && vdi->dict_index < d->var_cnt
399 && d->var[vdi->dict_index] == v);
405 /* Compares two double pointers to variables, which should point
406 to elements of a struct dictionary's `var' member array. */
408 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
410 struct variable *const *a = a_;
411 struct variable *const *b = b_;
413 return *a < *b ? -1 : *a > *b;
416 /* Sets the dict_index in V's vardict to DICT_INDEX. */
418 set_var_dict_index (struct variable *v, int dict_index)
420 struct vardict_info vdi = *var_get_vardict (v);
421 struct dictionary *d = vdi.dict;
422 vdi.dict_index = dict_index;
423 var_set_vardict (v, &vdi);
425 if ( d->callbacks && d->callbacks->var_changed )
426 d->callbacks->var_changed (d, dict_index, d->cb_data);
429 /* Sets the case_index in V's vardict to CASE_INDEX. */
431 set_var_case_index (struct variable *v, int case_index)
433 struct vardict_info vdi = *var_get_vardict (v);
434 vdi.case_index = case_index;
435 var_set_vardict (v, &vdi);
438 /* Re-sets the dict_index in the dictionary variables with
439 indexes from FROM to TO (exclusive). */
441 reindex_vars (struct dictionary *d, size_t from, size_t to)
445 for (i = from; i < to; i++)
446 set_var_dict_index (d->var[i], i);
449 /* Deletes variable V from dictionary D and frees V.
451 This is a very bad idea if there might be any pointers to V
452 from outside D. In general, no variable in the active file's
453 dictionary should be deleted when any transformations are
454 active on the dictionary's dataset, because those
455 transformations might reference the deleted variable. The
456 safest time to delete a variable is just after a procedure has
457 been executed, as done by MODIFY VARS.
459 Pointers to V within D are not a problem, because
460 dict_delete_var() knows to remove V from split variables,
461 weights, filters, etc. */
463 dict_delete_var (struct dictionary *d, struct variable *v)
465 int dict_index = var_get_dict_index (v);
466 const int case_index = var_get_case_index (v);
467 const int val_cnt = var_get_value_cnt (v);
469 assert (dict_contains_var (d, v));
471 /* Delete aux data. */
474 dict_unset_split_var (d, v);
477 dict_set_weight (d, NULL);
480 dict_set_filter (d, NULL);
482 dict_clear_vectors (d);
484 /* Remove V from var array. */
485 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
488 /* Update dict_index for each affected variable. */
489 reindex_vars (d, dict_index, d->var_cnt);
491 /* Update name hash. */
492 hsh_force_delete (d->name_tab, v);
496 var_clear_vardict (v);
499 if (d->callbacks && d->callbacks->var_deleted )
500 d->callbacks->var_deleted (d, dict_index, case_index, val_cnt, d->cb_data);
503 /* Deletes the COUNT variables listed in VARS from D. This is
504 unsafe; see the comment on dict_delete_var() for details. */
506 dict_delete_vars (struct dictionary *d,
507 struct variable *const *vars, size_t count)
509 /* FIXME: this can be done in O(count) time, but this algorithm
512 assert (count == 0 || vars != NULL);
515 dict_delete_var (d, *vars++);
518 /* Deletes the COUNT variables in D starting at index IDX. This
519 is unsafe; see the comment on dict_delete_var() for
522 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
524 /* FIXME: this can be done in O(count) time, but this algorithm
526 assert (idx + count <= d->var_cnt);
529 dict_delete_var (d, d->var[idx]);
532 /* Deletes scratch variables from dictionary D. */
534 dict_delete_scratch_vars (struct dictionary *d)
538 /* FIXME: this can be done in O(count) time, but this algorithm
542 for (i = 0; i < d->var_cnt; )
543 if (dict_class_from_id (var_get_name (d->var[i])) == DC_SCRATCH)
544 dict_delete_var (d, d->var[i]);
549 /* Moves V to 0-based position IDX in D. Other variables in D,
550 if any, retain their relative positions. Runs in time linear
551 in the distance moved. */
553 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
555 size_t old_index = var_get_dict_index (v);
557 assert (new_index < d->var_cnt);
558 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
559 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
562 /* Reorders the variables in D, placing the COUNT variables
563 listed in ORDER in that order at the beginning of D. The
564 other variables in D, if any, retain their relative
567 dict_reorder_vars (struct dictionary *d,
568 struct variable *const *order, size_t count)
570 struct variable **new_var;
574 assert (count == 0 || order != NULL);
575 assert (count <= d->var_cnt);
577 new_var = xnmalloc (d->var_cnt, sizeof *new_var);
578 memcpy (new_var, order, count * sizeof *new_var);
579 for (i = 0; i < count; i++)
581 size_t index = var_get_dict_index (order[i]);
582 assert (d->var[index] == order[i]);
583 d->var[index] = NULL;
584 set_var_dict_index (order[i], i);
586 for (i = 0; i < d->var_cnt; i++)
587 if (d->var[i] != NULL)
589 assert (count < d->var_cnt);
590 new_var[count] = d->var[i];
591 set_var_dict_index (new_var[count], count);
598 /* Changes the name of variable V in dictionary D to NEW_NAME. */
600 rename_var (struct dictionary *d, struct variable *v, const char *new_name)
602 struct vardict_info vdi;
604 assert (dict_contains_var (d, v));
606 vdi = *var_get_vardict (v);
607 var_clear_vardict (v);
608 var_set_name (v, new_name);
609 var_set_vardict (v, &vdi);
612 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
613 a variable named NEW_NAME is already in D, except that
614 NEW_NAME may be the same as V's existing name. */
616 dict_rename_var (struct dictionary *d, struct variable *v,
617 const char *new_name)
619 assert (!strcasecmp (var_get_name (v), new_name)
620 || dict_lookup_var (d, new_name) == NULL);
622 hsh_force_delete (d->name_tab, v);
623 rename_var (d, v, new_name);
624 hsh_force_insert (d->name_tab, v);
626 if (get_algorithm () == ENHANCED)
627 var_clear_short_name (v);
629 if ( d->callbacks && d->callbacks->var_changed )
630 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
633 /* Renames COUNT variables specified in VARS to the names given
634 in NEW_NAMES within dictionary D. If the renaming would
635 result in a duplicate variable name, returns false and stores a
636 name that would be duplicated into *ERR_NAME (if ERR_NAME is
637 non-null). Otherwise, the renaming is successful, and true
640 dict_rename_vars (struct dictionary *d,
641 struct variable **vars, char **new_names, size_t count,
648 assert (count == 0 || vars != NULL);
649 assert (count == 0 || new_names != NULL);
651 /* Save the names of the variables to be renamed. */
652 pool = pool_create ();
653 old_names = pool_nalloc (pool, count, sizeof *old_names);
654 for (i = 0; i < count; i++)
655 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
657 /* Remove the variables to be renamed from the name hash,
659 for (i = 0; i < count; i++)
661 hsh_force_delete (d->name_tab, vars[i]);
662 rename_var (d, vars[i], new_names[i]);
665 /* Add the renamed variables back into the name hash,
666 checking for conflicts. */
667 for (i = 0; i < count; i++)
668 if (hsh_insert (d->name_tab, vars[i]) != NULL)
670 /* There is a name conflict.
671 Back out all the name changes that have already
672 taken place, and indicate failure. */
674 if (err_name != NULL)
675 *err_name = new_names[i];
677 for (i = 0; i < fail_idx; i++)
678 hsh_force_delete (d->name_tab, vars[i]);
680 for (i = 0; i < count; i++)
682 rename_var (d, vars[i], old_names[i]);
683 hsh_force_insert (d->name_tab, vars[i]);
690 /* Clear short names. */
691 if (get_algorithm () == ENHANCED)
692 for (i = 0; i < count; i++)
693 var_clear_short_name (vars[i]);
699 /* Returns the weighting variable in dictionary D, or a null
700 pointer if the dictionary is unweighted. */
702 dict_get_weight (const struct dictionary *d)
705 assert (d->weight == NULL || dict_contains_var (d, d->weight));
710 /* Returns the value of D's weighting variable in case C, except that a
711 negative weight is returned as 0. Returns 1 if the dictionary is
712 unweighted. Will warn about missing, negative, or zero values if
713 warn_on_invalid is true. The function will set warn_on_invalid to false
714 if an invalid weight is found. */
716 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
717 bool *warn_on_invalid)
722 if (d->weight == NULL)
726 double w = case_num (c, d->weight);
727 if (w < 0.0 || var_is_num_missing (d->weight, w, MV_ANY))
729 if ( w == 0.0 && warn_on_invalid != NULL && *warn_on_invalid ) {
730 *warn_on_invalid = false;
731 msg (SW, _("At least one case in the data file had a weight value "
732 "that was user-missing, system-missing, zero, or "
733 "negative. These case(s) were ignored."));
739 /* Sets the weighting variable of D to V, or turning off
740 weighting if V is a null pointer. */
742 dict_set_weight (struct dictionary *d, struct variable *v)
745 assert (v == NULL || dict_contains_var (d, v));
746 assert (v == NULL || var_is_numeric (v));
750 if ( d->callbacks && d->callbacks->weight_changed )
751 d->callbacks->weight_changed (d,
752 v ? var_get_dict_index (v) : -1,
756 /* Returns the filter variable in dictionary D (see cmd_filter())
757 or a null pointer if the dictionary is unfiltered. */
759 dict_get_filter (const struct dictionary *d)
762 assert (d->filter == NULL || dict_contains_var (d, d->filter));
767 /* Sets V as the filter variable for dictionary D. Passing a
768 null pointer for V turn off filtering. */
770 dict_set_filter (struct dictionary *d, struct variable *v)
773 assert (v == NULL || dict_contains_var (d, v));
777 if ( d->callbacks && d->callbacks->filter_changed )
778 d->callbacks->filter_changed (d,
779 v ? var_get_dict_index (v) : -1,
783 /* Returns the case limit for dictionary D, or zero if the number
784 of cases is unlimited. */
786 dict_get_case_limit (const struct dictionary *d)
790 return d->case_limit;
793 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
794 0 for CASE_LIMIT to indicate no limit. */
796 dict_set_case_limit (struct dictionary *d, size_t case_limit)
800 d->case_limit = case_limit;
803 /* Returns the case index of the next value to be added to D.
804 This value is the number of `union value's that need to be
805 allocated to store a case for dictionary D. */
807 dict_get_next_value_idx (const struct dictionary *d)
811 return d->next_value_idx;
814 /* Returns the number of bytes needed to store a case for
817 dict_get_case_size (const struct dictionary *d)
821 return sizeof (union value) * dict_get_next_value_idx (d);
824 /* Deletes scratch variables in dictionary D and reassigns values
825 so that fragmentation is eliminated. */
827 dict_compact_values (struct dictionary *d)
831 d->next_value_idx = 0;
832 for (i = 0; i < d->var_cnt; )
834 struct variable *v = d->var[i];
836 if (dict_class_from_id (var_get_name (v)) != DC_SCRATCH)
838 set_var_case_index (v, d->next_value_idx);
839 d->next_value_idx += var_get_value_cnt (v);
843 dict_delete_var (d, v);
849 Reassigns case indices for D, increasing each index above START by
853 dict_pad_values (struct dictionary *d, int start, int padding)
860 for (i = 0; i < d->var_cnt; ++i)
862 struct variable *v = d->var[i];
864 int index = var_get_case_index (v);
867 set_var_case_index (v, index + padding);
870 d->next_value_idx += padding;
874 /* Returns the number of values that would be used by a case if
875 dict_compact_values() were called. */
877 dict_get_compacted_value_cnt (const struct dictionary *d)
883 for (i = 0; i < d->var_cnt; i++)
884 if (dict_class_from_id (var_get_name (d->var[i])) != DC_SCRATCH)
885 cnt += var_get_value_cnt (d->var[i]);
889 /* Creates and returns an array mapping from a dictionary index
890 to the case index that the corresponding variable will have
891 after calling dict_compact_values(). Scratch variables
892 receive -1 for case index because dict_compact_values() will
895 dict_get_compacted_dict_index_to_case_index (const struct dictionary *d)
898 size_t next_value_idx;
901 map = xnmalloc (d->var_cnt, sizeof *map);
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 map[i] = next_value_idx;
910 next_value_idx += var_get_value_cnt (v);
918 /* Returns true if a case for dictionary D would be smaller after
919 compacting, false otherwise. Compacting a case eliminates
920 "holes" between values and after the last value. Holes are
921 created by deleting variables (or by scratch variables).
923 The return value may differ from whether compacting a case
924 from dictionary D would *change* the case: compacting could
925 rearrange values even if it didn't reduce space
928 dict_compacting_would_shrink (const struct dictionary *d)
930 return dict_get_compacted_value_cnt (d) < dict_get_next_value_idx (d);
933 /* Returns true if a case for dictionary D would change after
934 compacting, false otherwise. Compacting a case eliminates
935 "holes" between values and after the last value. Holes are
936 created by deleting variables (or by scratch variables).
938 The return value may differ from whether compacting a case
939 from dictionary D would *shrink* the case: compacting could
940 rearrange values without reducing space requirements. */
942 dict_compacting_would_change (const struct dictionary *d)
948 for (i = 0; i < dict_get_var_cnt (d); i++)
950 struct variable *v = dict_get_var (d, i);
951 if (var_get_case_index (v) != case_idx)
953 case_idx += var_get_value_cnt (v);
958 /* How to copy a contiguous range of values between cases. */
961 size_t src_idx; /* Starting value index in source case. */
962 size_t dst_idx; /* Starting value index in target case. */
963 size_t cnt; /* Number of values. */
966 /* How to compact a case. */
967 struct dict_compactor
969 struct copy_map *maps; /* Array of mappings. */
970 size_t map_cnt; /* Number of mappings. */
973 /* Creates and returns a dict_compactor that can be used to
974 compact cases for dictionary D.
976 Compacting a case eliminates "holes" between values and after
977 the last value. Holes are created by deleting variables (or
978 by scratch variables). */
979 struct dict_compactor *
980 dict_make_compactor (const struct dictionary *d)
982 struct dict_compactor *compactor;
983 struct copy_map *map;
984 size_t map_allocated;
988 compactor = xmalloc (sizeof *compactor);
989 compactor->maps = NULL;
990 compactor->map_cnt = 0;
995 for (i = 0; i < d->var_cnt; i++)
997 struct variable *v = d->var[i];
999 if (dict_class_from_id (var_get_name (v)) == DC_SCRATCH)
1001 if (map != NULL && map->src_idx + map->cnt == var_get_case_index (v))
1002 map->cnt += var_get_value_cnt (v);
1005 if (compactor->map_cnt == map_allocated)
1006 compactor->maps = x2nrealloc (compactor->maps, &map_allocated,
1007 sizeof *compactor->maps);
1008 map = &compactor->maps[compactor->map_cnt++];
1009 map->src_idx = var_get_case_index (v);
1010 map->dst_idx = value_idx;
1011 map->cnt = var_get_value_cnt (v);
1013 value_idx += var_get_value_cnt (v);
1019 /* Compacts SRC by copying it to DST according to the scheme in
1022 Compacting a case eliminates "holes" between values and after
1023 the last value. Holes are created by deleting variables (or
1024 by scratch variables). */
1026 dict_compactor_compact (const struct dict_compactor *compactor,
1027 struct ccase *dst, const struct ccase *src)
1031 for (i = 0; i < compactor->map_cnt; i++)
1033 const struct copy_map *map = &compactor->maps[i];
1034 case_copy (dst, map->dst_idx, src, map->src_idx, map->cnt);
1038 /* Destroys COMPACTOR. */
1040 dict_compactor_destroy (struct dict_compactor *compactor)
1042 if (compactor != NULL)
1044 free (compactor->maps);
1049 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
1050 dict_get_split_cnt() to determine how many SPLIT FILE vars
1051 there are. Returns a null pointer if and only if there are no
1053 const struct variable *const *
1054 dict_get_split_vars (const struct dictionary *d)
1061 /* Returns the number of SPLIT FILE vars. */
1063 dict_get_split_cnt (const struct dictionary *d)
1067 return d->split_cnt;
1070 /* Removes variable V from the set of split variables in dictionary D */
1072 dict_unset_split_var (struct dictionary *d,
1075 const int count = d->split_cnt;
1076 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
1077 &v, compare_var_ptrs, NULL);
1079 if ( count == d->split_cnt)
1082 if ( d->callbacks && d->callbacks->split_changed )
1083 d->callbacks->split_changed (d, d->cb_data);
1086 /* Sets CNT split vars SPLIT in dictionary D. */
1088 dict_set_split_vars (struct dictionary *d,
1089 struct variable *const *split, size_t cnt)
1092 assert (cnt == 0 || split != NULL);
1095 d->split = cnt > 0 ? xnrealloc (d->split, cnt, sizeof *d->split) : NULL;
1096 memcpy (d->split, split, cnt * sizeof *d->split);
1098 if ( d->callbacks && d->callbacks->split_changed )
1099 d->callbacks->split_changed (d, d->cb_data);
1102 /* Returns the file label for D, or a null pointer if D is
1103 unlabeled (see cmd_file_label()). */
1105 dict_get_label (const struct dictionary *d)
1112 /* Sets D's file label to LABEL, truncating it to a maximum of 60
1115 dict_set_label (struct dictionary *d, const char *label)
1122 else if (strlen (label) < 60)
1123 d->label = xstrdup (label);
1126 d->label = xmalloc (61);
1127 memcpy (d->label, label, 60);
1128 d->label[60] = '\0';
1132 /* Returns the documents for D, or a null pointer if D has no
1133 documents. If the return value is nonnull, then the string
1134 will be an exact multiple of DOC_LINE_LENGTH bytes in length,
1135 with each segment corresponding to one line. */
1137 dict_get_documents (const struct dictionary *d)
1139 return ds_is_empty (&d->documents) ? NULL : ds_cstr (&d->documents);
1142 /* Sets the documents for D to DOCUMENTS, or removes D's
1143 documents if DOCUMENT is a null pointer. If DOCUMENTS is
1144 nonnull, then it should be an exact multiple of
1145 DOC_LINE_LENGTH bytes in length, with each segment
1146 corresponding to one line. */
1148 dict_set_documents (struct dictionary *d, const char *documents)
1152 ds_assign_cstr (&d->documents, documents != NULL ? documents : "");
1154 /* In case the caller didn't get it quite right, pad out the
1155 final line with spaces. */
1156 remainder = ds_length (&d->documents) % DOC_LINE_LENGTH;
1158 ds_put_char_multiple (&d->documents, ' ', DOC_LINE_LENGTH - remainder);
1161 /* Drops the documents from dictionary D. */
1163 dict_clear_documents (struct dictionary *d)
1165 ds_clear (&d->documents);
1168 /* Appends LINE to the documents in D. LINE will be truncated or
1169 padded on the right with spaces to make it exactly
1170 DOC_LINE_LENGTH bytes long. */
1172 dict_add_document_line (struct dictionary *d, const char *line)
1174 if (strlen (line) > DOC_LINE_LENGTH)
1176 /* Note to translators: "bytes" is correct, not characters */
1177 msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
1179 buf_copy_str_rpad (ds_put_uninit (&d->documents, DOC_LINE_LENGTH),
1180 DOC_LINE_LENGTH, line);
1183 /* Returns the number of document lines in dictionary D. */
1185 dict_get_document_line_cnt (const struct dictionary *d)
1187 return ds_length (&d->documents) / DOC_LINE_LENGTH;
1190 /* Copies document line number IDX from dictionary D into
1191 LINE, trimming off any trailing white space. */
1193 dict_get_document_line (const struct dictionary *d,
1194 size_t idx, struct string *line)
1196 assert (idx < dict_get_document_line_cnt (d));
1197 ds_assign_substring (line, ds_substr (&d->documents, idx * DOC_LINE_LENGTH,
1199 ds_rtrim (line, ss_cstr (CC_SPACES));
1202 /* Creates in D a vector named NAME that contains the CNT
1203 variables in VAR. Returns true if successful, or false if a
1204 vector named NAME already exists in D. */
1206 dict_create_vector (struct dictionary *d,
1208 struct variable **var, size_t cnt)
1212 assert (var != NULL);
1214 for (i = 0; i < cnt; i++)
1215 assert (dict_contains_var (d, var[i]));
1217 if (dict_lookup_vector (d, name) == NULL)
1219 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1220 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1227 /* Creates in D a vector named NAME that contains the CNT
1228 variables in VAR. A vector named NAME must not already exist
1231 dict_create_vector_assert (struct dictionary *d,
1233 struct variable **var, size_t cnt)
1235 assert (dict_lookup_vector (d, name) == NULL);
1236 dict_create_vector (d, name, var, cnt);
1239 /* Returns the vector in D with index IDX, which must be less
1240 than dict_get_vector_cnt (D). */
1241 const struct vector *
1242 dict_get_vector (const struct dictionary *d, size_t idx)
1245 assert (idx < d->vector_cnt);
1247 return d->vector[idx];
1250 /* Returns the number of vectors in D. */
1252 dict_get_vector_cnt (const struct dictionary *d)
1256 return d->vector_cnt;
1259 /* Looks up and returns the vector within D with the given
1261 const struct vector *
1262 dict_lookup_vector (const struct dictionary *d, const char *name)
1265 for (i = 0; i < d->vector_cnt; i++)
1266 if (!strcasecmp (vector_get_name (d->vector[i]), name))
1267 return d->vector[i];
1271 /* Deletes all vectors from D. */
1273 dict_clear_vectors (struct dictionary *d)
1277 for (i = 0; i < d->vector_cnt; i++)
1278 vector_destroy (d->vector[i]);
1285 /* Compares two strings. */
1287 compare_strings (const void *a, const void *b, const void *aux UNUSED)
1289 return strcmp (a, b);
1292 /* Hashes a string. */
1294 hash_string (const void *s, const void *aux UNUSED)
1296 return hsh_hash_string (s);
1300 /* Sets V's short name to BASE, followed by a suffix of the form
1301 _A, _B, _C, ..., _AA, _AB, etc. according to the value of
1302 SUFFIX_NUMBER. Truncates BASE as necessary to fit. */
1304 set_var_short_name_suffix (struct variable *v, const char *base,
1307 char suffix[SHORT_NAME_LEN + 1];
1308 char short_name[SHORT_NAME_LEN + 1];
1313 assert (suffix_number >= 0);
1315 /* Set base name. */
1316 var_set_short_name (v, base);
1318 /* Compose suffix. */
1319 start = end = suffix + sizeof suffix - 1;
1323 *--start = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"[suffix_number % 26];
1324 if (start <= suffix + 1)
1325 msg (SE, _("Variable suffix too large."));
1326 suffix_number /= 26;
1328 while (suffix_number > 0);
1331 /* Append suffix to V's short name. */
1332 str_copy_trunc (short_name, sizeof short_name, base);
1334 if (len + strlen (short_name) > SHORT_NAME_LEN)
1335 ofs = SHORT_NAME_LEN - len;
1337 ofs = strlen (short_name);
1338 strcpy (short_name + ofs, start);
1341 var_set_short_name (v, short_name);
1344 /* Assigns a valid, unique short_name[] to each variable in D.
1345 Each variable whose actual name is short has highest priority
1346 for that short name. Otherwise, variables with an existing
1347 short_name[] have the next highest priority for a given short
1348 name; if it is already taken, then the variable is treated as
1349 if short_name[] had been empty. Otherwise, long names are
1350 truncated to form short names. If that causes conflicts,
1351 variables are renamed as PREFIX_A, PREFIX_B, and so on. */
1353 dict_assign_short_names (struct dictionary *d)
1355 struct hsh_table *short_names;
1358 /* Give variables whose names are short the corresponding short
1359 names, and clear short_names[] that conflict with a variable
1361 for (i = 0; i < d->var_cnt; i++)
1363 struct variable *v = d->var[i];
1364 const char *short_name = var_get_short_name (v);
1365 if (strlen (var_get_name (v)) <= SHORT_NAME_LEN)
1366 var_set_short_name (v, var_get_name (v));
1367 else if (short_name != NULL && dict_lookup_var (d, short_name) != NULL)
1368 var_clear_short_name (v);
1371 /* Each variable with an assigned short_name[] now gets it
1372 unless there is a conflict. */
1373 short_names = hsh_create (d->var_cnt, compare_strings, hash_string,
1375 for (i = 0; i < d->var_cnt; i++)
1377 struct variable *v = d->var[i];
1378 const char *name = var_get_short_name (v);
1379 if (name != NULL && hsh_insert (short_names, (char *) name) != NULL)
1380 var_clear_short_name (v);
1383 /* Now assign short names to remaining variables. */
1384 for (i = 0; i < d->var_cnt; i++)
1386 struct variable *v = d->var[i];
1387 const char *name = var_get_short_name (v);
1390 /* Form initial short_name from the variable name, then
1391 try _A, _B, ... _AA, _AB, etc., if needed.*/
1396 var_set_short_name (v, var_get_name (v));
1398 set_var_short_name_suffix (v, var_get_name (v), trial - 1);
1402 while (hsh_insert (short_names, (char *) var_get_short_name (v))
1407 /* Get rid of hash table. */
1408 hsh_destroy (short_names);
1412 /* Called from variable.c to notify the dictionary that some property of
1413 the variable has changed */
1415 dict_var_changed (const struct variable *v)
1417 if ( var_has_vardict (v))
1419 const struct vardict_info *vdi = var_get_vardict (v);
1420 struct dictionary *d;
1424 if ( d->callbacks && d->callbacks->var_changed )
1425 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
1430 /* Called from variable.c to notify the dictionary that the variable's width
1433 dict_var_resized (const struct variable *v, int delta)
1435 if ( var_has_vardict (v))
1437 const struct vardict_info *vdi = var_get_vardict (v);
1438 struct dictionary *d;
1442 dict_pad_values (d, var_get_case_index(v) + 1, delta);
1444 if ( d->callbacks && d->callbacks->var_resized )
1445 d->callbacks->var_resized (d, var_get_dict_index (v), delta, d->cb_data);