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 var_clear_vardict (d->var[d->var_cnt - 1]);
160 var_destroy (d->var[d->var_cnt -1]);
164 if (d->callbacks && d->callbacks->var_deleted )
165 d->callbacks->var_deleted (d, d->var_cnt, d->cb_data);
169 d->var_cnt = d->var_cap = 0;
170 hsh_clear (d->name_tab);
171 d->next_value_idx = 0;
172 dict_set_split_vars (d, NULL, 0);
173 dict_set_weight (d, NULL);
174 dict_set_filter (d, NULL);
178 ds_destroy (&d->documents);
179 dict_clear_vectors (d);
182 /* Destroys the aux data for every variable in D, by calling
183 var_clear_aux() for each variable. */
185 dict_clear_aux (struct dictionary *d)
191 for (i = 0; i < d->var_cnt; i++)
192 var_clear_aux (d->var[i]);
195 /* Clears a dictionary and destroys it. */
197 dict_destroy (struct dictionary *d)
201 /* In general, we don't want callbacks occuring, if the dictionary
202 is being destroyed */
203 d->callbacks = NULL ;
206 hsh_destroy (d->name_tab);
211 /* Returns the number of variables in D. */
213 dict_get_var_cnt (const struct dictionary *d)
220 /* Returns the variable in D with dictionary index IDX, which
221 must be between 0 and the count returned by
222 dict_get_var_cnt(), exclusive. */
224 dict_get_var (const struct dictionary *d, size_t idx)
227 assert (idx < d->var_cnt);
233 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
234 size_t *cnt, unsigned exclude_classes)
236 dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude_classes);
239 /* Sets *VARS to an array of pointers to variables in D and *CNT
240 to the number of variables in *D. All variables are returned
241 if EXCLUDE_CLASSES is 0, or it may contain one or more of (1u
242 << DC_ORDINARY), (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to
243 exclude the corresponding type of variable. */
245 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
246 size_t *cnt, unsigned exclude_classes)
252 assert (vars != NULL);
253 assert (cnt != NULL);
254 assert ((exclude_classes & ~((1u << DC_ORDINARY)
256 | (1u << DC_SCRATCH))) == 0);
259 for (i = 0; i < d->var_cnt; i++)
261 enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
262 if (!(exclude_classes & (1u << class)))
266 *vars = xnmalloc (count, sizeof **vars);
268 for (i = 0; i < d->var_cnt; i++)
270 enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
271 if (!(exclude_classes & (1u << class)))
272 (*vars)[(*cnt)++] = d->var[i];
274 assert (*cnt == count);
277 static struct variable *
278 add_var (struct dictionary *d, struct variable *v)
280 /* Add dictionary info to variable. */
281 struct vardict_info vdi;
282 vdi.case_index = d->next_value_idx;
283 vdi.dict_index = d->var_cnt;
285 var_set_vardict (v, &vdi);
287 /* Update dictionary. */
288 if (d->var_cnt >= d->var_cap)
290 d->var_cap = 8 + 2 * d->var_cap;
291 d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var);
293 d->var[d->var_cnt++] = v;
294 hsh_force_insert (d->name_tab, v);
296 if ( d->callbacks && d->callbacks->var_added )
297 d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
299 d->next_value_idx += var_get_value_cnt (v);
304 /* Creates and returns a new variable in D with the given NAME
305 and WIDTH. Returns a null pointer if the given NAME would
306 duplicate that of an existing variable in the dictionary. */
308 dict_create_var (struct dictionary *d, const char *name, int width)
310 return (dict_lookup_var (d, name) == NULL
311 ? dict_create_var_assert (d, name, width)
315 /* Creates and returns a new variable in D with the given NAME
316 and WIDTH. Assert-fails if the given NAME would duplicate
317 that of an existing variable in the dictionary. */
319 dict_create_var_assert (struct dictionary *d, const char *name, int width)
321 assert (dict_lookup_var (d, name) == NULL);
322 return add_var (d, var_create (name, width));
325 /* Creates and returns a new variable in D with name NAME, as a
326 copy of existing variable OLD_VAR, which need not be in D or
327 in any dictionary. Returns a null pointer if the given NAME
328 would duplicate that of an existing variable in the
331 dict_clone_var (struct dictionary *d, const struct variable *old_var,
334 return (dict_lookup_var (d, name) == NULL
335 ? dict_clone_var_assert (d, old_var, name)
339 /* Creates and returns a new variable in D with name NAME, as a
340 copy of existing variable OLD_VAR, which need not be in D or
341 in any dictionary. Assert-fails if the given NAME would
342 duplicate that of an existing variable in the dictionary. */
344 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var,
347 struct variable *new_var = var_clone (old_var);
348 assert (dict_lookup_var (d, name) == NULL);
349 var_set_name (new_var, name);
350 return add_var (d, new_var);
353 /* Returns the variable named NAME in D, or a null pointer if no
354 variable has that name. */
356 dict_lookup_var (const struct dictionary *d, const char *name)
358 struct variable *target ;
359 struct variable *result ;
361 if ( ! var_is_valid_name (name, false))
364 target = var_create (name, 0);
365 result = hsh_find (d->name_tab, target);
366 var_destroy (target);
371 /* Returns the variable named NAME in D. Assert-fails if no
372 variable has that name. */
374 dict_lookup_var_assert (const struct dictionary *d, const char *name)
376 struct variable *v = dict_lookup_var (d, name);
381 /* Returns true if variable V is in dictionary D,
384 dict_contains_var (const struct dictionary *d, const struct variable *v)
386 if (var_has_vardict (v))
388 const struct vardict_info *vdi = var_get_vardict (v);
389 return (vdi->dict_index >= 0
390 && vdi->dict_index < d->var_cnt
391 && d->var[vdi->dict_index] == v);
397 /* Compares two double pointers to variables, which should point
398 to elements of a struct dictionary's `var' member array. */
400 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
402 struct variable *const *a = a_;
403 struct variable *const *b = b_;
405 return *a < *b ? -1 : *a > *b;
408 /* Sets the dict_index in V's vardict to DICT_INDEX. */
410 set_var_dict_index (struct variable *v, int dict_index)
412 struct vardict_info vdi = *var_get_vardict (v);
413 struct dictionary *d = vdi.dict;
414 vdi.dict_index = dict_index;
415 var_set_vardict (v, &vdi);
417 if ( d->callbacks && d->callbacks->var_changed )
418 d->callbacks->var_changed (d, dict_index, d->cb_data);
421 /* Sets the case_index in V's vardict to DICT_INDEX. */
423 set_var_case_index (struct variable *v, int case_index)
425 struct vardict_info vdi = *var_get_vardict (v);
426 vdi.case_index = case_index;
427 var_set_vardict (v, &vdi);
430 /* Re-sets the dict_index in the dictionary variables with
431 indexes from FROM to TO (exclusive). */
433 reindex_vars (struct dictionary *d, size_t from, size_t to)
437 for (i = from; i < to; i++)
438 set_var_dict_index (d->var[i], i);
441 /* Deletes variable V from dictionary D and frees V.
443 This is a very bad idea if there might be any pointers to V
444 from outside D. In general, no variable in the active file's
445 dictionary should be deleted when any transformations are
446 active on the dictionary's dataset, because those
447 transformations might reference the deleted variable. The
448 safest time to delete a variable is just after a procedure has
449 been executed, as done by MODIFY VARS.
451 Pointers to V within D are not a problem, because
452 dict_delete_var() knows to remove V from split variables,
453 weights, filters, etc. */
455 dict_delete_var (struct dictionary *d, struct variable *v)
457 int dict_index = var_get_dict_index (v);
459 assert (dict_contains_var (d, v));
461 /* Delete aux data. */
464 dict_unset_split_var (d, v);
467 dict_set_weight (d, NULL);
470 dict_set_filter (d, NULL);
472 dict_clear_vectors (d);
474 /* Remove V from var array. */
475 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
478 /* Update dict_index for each affected variable. */
479 reindex_vars (d, dict_index, d->var_cnt);
481 /* Update name hash. */
482 hsh_force_delete (d->name_tab, v);
486 var_clear_vardict (v);
489 if (d->callbacks && d->callbacks->var_deleted )
490 d->callbacks->var_deleted (d, dict_index, d->cb_data);
493 /* Deletes the COUNT variables listed in VARS from D. This is
494 unsafe; see the comment on dict_delete_var() for details. */
496 dict_delete_vars (struct dictionary *d,
497 struct variable *const *vars, size_t count)
499 /* FIXME: this can be done in O(count) time, but this algorithm
502 assert (count == 0 || vars != NULL);
505 dict_delete_var (d, *vars++);
508 /* Deletes the COUNT variables in D starting at index IDX. This
509 is unsafe; see the comment on dict_delete_var() for
512 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
514 /* FIXME: this can be done in O(count) time, but this algorithm
516 assert (idx + count <= d->var_cnt);
519 dict_delete_var (d, d->var[idx]);
522 /* Deletes scratch variables from dictionary D. */
524 dict_delete_scratch_vars (struct dictionary *d)
528 /* FIXME: this can be done in O(count) time, but this algorithm
532 for (i = 0; i < d->var_cnt; )
533 if (dict_class_from_id (var_get_name (d->var[i])) == DC_SCRATCH)
534 dict_delete_var (d, d->var[i]);
539 /* Moves V to 0-based position IDX in D. Other variables in D,
540 if any, retain their relative positions. Runs in time linear
541 in the distance moved. */
543 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
545 size_t old_index = var_get_dict_index (v);
547 assert (new_index < d->var_cnt);
548 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
549 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
552 /* Reorders the variables in D, placing the COUNT variables
553 listed in ORDER in that order at the beginning of D. The
554 other variables in D, if any, retain their relative
557 dict_reorder_vars (struct dictionary *d,
558 struct variable *const *order, size_t count)
560 struct variable **new_var;
564 assert (count == 0 || order != NULL);
565 assert (count <= d->var_cnt);
567 new_var = xnmalloc (d->var_cnt, sizeof *new_var);
568 memcpy (new_var, order, count * sizeof *new_var);
569 for (i = 0; i < count; i++)
571 size_t index = var_get_dict_index (order[i]);
572 assert (d->var[index] == order[i]);
573 d->var[index] = NULL;
574 set_var_dict_index (order[i], i);
576 for (i = 0; i < d->var_cnt; i++)
577 if (d->var[i] != NULL)
579 assert (count < d->var_cnt);
580 new_var[count] = d->var[i];
581 set_var_dict_index (new_var[count], count);
588 /* Changes the name of variable V in dictionary D to NEW_NAME. */
590 rename_var (struct dictionary *d, struct variable *v, const char *new_name)
592 struct vardict_info vdi;
594 assert (dict_contains_var (d, v));
596 vdi = *var_get_vardict (v);
597 var_clear_vardict (v);
598 var_set_name (v, new_name);
599 var_set_vardict (v, &vdi);
602 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
603 a variable named NEW_NAME is already in D, except that
604 NEW_NAME may be the same as V's existing name. */
606 dict_rename_var (struct dictionary *d, struct variable *v,
607 const char *new_name)
609 assert (!strcasecmp (var_get_name (v), new_name)
610 || dict_lookup_var (d, new_name) == NULL);
612 hsh_force_delete (d->name_tab, v);
613 rename_var (d, v, new_name);
614 hsh_force_insert (d->name_tab, v);
616 if (get_algorithm () == ENHANCED)
617 var_clear_short_name (v);
619 if ( d->callbacks && d->callbacks->var_changed )
620 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
623 /* Renames COUNT variables specified in VARS to the names given
624 in NEW_NAMES within dictionary D. If the renaming would
625 result in a duplicate variable name, returns false and stores a
626 name that would be duplicated into *ERR_NAME (if ERR_NAME is
627 non-null). Otherwise, the renaming is successful, and true
630 dict_rename_vars (struct dictionary *d,
631 struct variable **vars, char **new_names, size_t count,
638 assert (count == 0 || vars != NULL);
639 assert (count == 0 || new_names != NULL);
641 /* Save the names of the variables to be renamed. */
642 pool = pool_create ();
643 old_names = pool_nalloc (pool, count, sizeof *old_names);
644 for (i = 0; i < count; i++)
645 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
647 /* Remove the variables to be renamed from the name hash,
649 for (i = 0; i < count; i++)
651 hsh_force_delete (d->name_tab, vars[i]);
652 rename_var (d, vars[i], new_names[i]);
655 /* Add the renamed variables back into the name hash,
656 checking for conflicts. */
657 for (i = 0; i < count; i++)
658 if (hsh_insert (d->name_tab, vars[i]) != NULL)
660 /* There is a name conflict.
661 Back out all the name changes that have already
662 taken place, and indicate failure. */
664 if (err_name != NULL)
665 *err_name = new_names[i];
667 for (i = 0; i < fail_idx; i++)
668 hsh_force_delete (d->name_tab, vars[i]);
670 for (i = 0; i < count; i++)
672 rename_var (d, vars[i], old_names[i]);
673 hsh_force_insert (d->name_tab, vars[i]);
680 /* Clear short names. */
681 if (get_algorithm () == ENHANCED)
682 for (i = 0; i < count; i++)
683 var_clear_short_name (vars[i]);
689 /* Returns the weighting variable in dictionary D, or a null
690 pointer if the dictionary is unweighted. */
692 dict_get_weight (const struct dictionary *d)
695 assert (d->weight == NULL || dict_contains_var (d, d->weight));
700 /* Returns the value of D's weighting variable in case C, except that a
701 negative weight is returned as 0. Returns 1 if the dictionary is
702 unweighted. Will warn about missing, negative, or zero values if
703 warn_on_invalid is true. The function will set warn_on_invalid to false
704 if an invalid weight is found. */
706 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
707 bool *warn_on_invalid)
712 if (d->weight == NULL)
716 double w = case_num (c, d->weight);
717 if (w < 0.0 || var_is_num_missing (d->weight, w, MV_ANY))
719 if ( w == 0.0 && warn_on_invalid != NULL && *warn_on_invalid ) {
720 *warn_on_invalid = false;
721 msg (SW, _("At least one case in the data file had a weight value "
722 "that was user-missing, system-missing, zero, or "
723 "negative. These case(s) were ignored."));
729 /* Sets the weighting variable of D to V, or turning off
730 weighting if V is a null pointer. */
732 dict_set_weight (struct dictionary *d, struct variable *v)
735 assert (v == NULL || dict_contains_var (d, v));
736 assert (v == NULL || var_is_numeric (v));
740 if ( d->callbacks && d->callbacks->weight_changed )
741 d->callbacks->weight_changed (d,
742 v ? var_get_dict_index (v) : -1,
746 /* Returns the filter variable in dictionary D (see cmd_filter())
747 or a null pointer if the dictionary is unfiltered. */
749 dict_get_filter (const struct dictionary *d)
752 assert (d->filter == NULL || dict_contains_var (d, d->filter));
757 /* Sets V as the filter variable for dictionary D. Passing a
758 null pointer for V turn off filtering. */
760 dict_set_filter (struct dictionary *d, struct variable *v)
763 assert (v == NULL || dict_contains_var (d, v));
767 if ( d->callbacks && d->callbacks->filter_changed )
768 d->callbacks->filter_changed (d,
769 v ? var_get_dict_index (v) : -1,
773 /* Returns the case limit for dictionary D, or zero if the number
774 of cases is unlimited. */
776 dict_get_case_limit (const struct dictionary *d)
780 return d->case_limit;
783 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
784 0 for CASE_LIMIT to indicate no limit. */
786 dict_set_case_limit (struct dictionary *d, size_t case_limit)
790 d->case_limit = case_limit;
793 /* Returns the case index of the next value to be added to D.
794 This value is the number of `union value's that need to be
795 allocated to store a case for dictionary D. */
797 dict_get_next_value_idx (const struct dictionary *d)
801 return d->next_value_idx;
804 /* Returns the number of bytes needed to store a case for
807 dict_get_case_size (const struct dictionary *d)
811 return sizeof (union value) * dict_get_next_value_idx (d);
814 /* Deletes scratch variables in dictionary D and reassigns values
815 so that fragmentation is eliminated. */
817 dict_compact_values (struct dictionary *d)
821 d->next_value_idx = 0;
822 for (i = 0; i < d->var_cnt; )
824 struct variable *v = d->var[i];
826 if (dict_class_from_id (var_get_name (v)) != DC_SCRATCH)
828 set_var_case_index (v, d->next_value_idx);
829 d->next_value_idx += var_get_value_cnt (v);
833 dict_delete_var (d, v);
837 /* Returns the number of values that would be used by a case if
838 dict_compact_values() were called. */
840 dict_get_compacted_value_cnt (const struct dictionary *d)
846 for (i = 0; i < d->var_cnt; i++)
847 if (dict_class_from_id (var_get_name (d->var[i])) != DC_SCRATCH)
848 cnt += var_get_value_cnt (d->var[i]);
852 /* Creates and returns an array mapping from a dictionary index
853 to the case index that the corresponding variable will have
854 after calling dict_compact_values(). Scratch variables
855 receive -1 for case index because dict_compact_values() will
858 dict_get_compacted_dict_index_to_case_index (const struct dictionary *d)
861 size_t next_value_idx;
864 map = xnmalloc (d->var_cnt, sizeof *map);
866 for (i = 0; i < d->var_cnt; i++)
868 struct variable *v = d->var[i];
870 if (dict_class_from_id (var_get_name (v)) != DC_SCRATCH)
872 map[i] = next_value_idx;
873 next_value_idx += var_get_value_cnt (v);
881 /* Returns true if a case for dictionary D would be smaller after
882 compacting, false otherwise. Compacting a case eliminates
883 "holes" between values and after the last value. Holes are
884 created by deleting variables (or by scratch variables).
886 The return value may differ from whether compacting a case
887 from dictionary D would *change* the case: compacting could
888 rearrange values even if it didn't reduce space
891 dict_compacting_would_shrink (const struct dictionary *d)
893 return dict_get_compacted_value_cnt (d) < dict_get_next_value_idx (d);
896 /* Returns true if a case for dictionary D would change after
897 compacting, false otherwise. Compacting a case eliminates
898 "holes" between values and after the last value. Holes are
899 created by deleting variables (or by scratch variables).
901 The return value may differ from whether compacting a case
902 from dictionary D would *shrink* the case: compacting could
903 rearrange values without reducing space requirements. */
905 dict_compacting_would_change (const struct dictionary *d)
911 for (i = 0; i < dict_get_var_cnt (d); i++)
913 struct variable *v = dict_get_var (d, i);
914 if (var_get_case_index (v) != case_idx)
916 case_idx += var_get_value_cnt (v);
921 /* How to copy a contiguous range of values between cases. */
924 size_t src_idx; /* Starting value index in source case. */
925 size_t dst_idx; /* Starting value index in target case. */
926 size_t cnt; /* Number of values. */
929 /* How to compact a case. */
930 struct dict_compactor
932 struct copy_map *maps; /* Array of mappings. */
933 size_t map_cnt; /* Number of mappings. */
936 /* Creates and returns a dict_compactor that can be used to
937 compact cases for dictionary D.
939 Compacting a case eliminates "holes" between values and after
940 the last value. Holes are created by deleting variables (or
941 by scratch variables). */
942 struct dict_compactor *
943 dict_make_compactor (const struct dictionary *d)
945 struct dict_compactor *compactor;
946 struct copy_map *map;
947 size_t map_allocated;
951 compactor = xmalloc (sizeof *compactor);
952 compactor->maps = NULL;
953 compactor->map_cnt = 0;
958 for (i = 0; i < d->var_cnt; i++)
960 struct variable *v = d->var[i];
962 if (dict_class_from_id (var_get_name (v)) == DC_SCRATCH)
964 if (map != NULL && map->src_idx + map->cnt == var_get_case_index (v))
965 map->cnt += var_get_value_cnt (v);
968 if (compactor->map_cnt == map_allocated)
969 compactor->maps = x2nrealloc (compactor->maps, &map_allocated,
970 sizeof *compactor->maps);
971 map = &compactor->maps[compactor->map_cnt++];
972 map->src_idx = var_get_case_index (v);
973 map->dst_idx = value_idx;
974 map->cnt = var_get_value_cnt (v);
976 value_idx += var_get_value_cnt (v);
982 /* Compacts SRC by copying it to DST according to the scheme in
985 Compacting a case eliminates "holes" between values and after
986 the last value. Holes are created by deleting variables (or
987 by scratch variables). */
989 dict_compactor_compact (const struct dict_compactor *compactor,
990 struct ccase *dst, const struct ccase *src)
994 for (i = 0; i < compactor->map_cnt; i++)
996 const struct copy_map *map = &compactor->maps[i];
997 case_copy (dst, map->dst_idx, src, map->src_idx, map->cnt);
1001 /* Destroys COMPACTOR. */
1003 dict_compactor_destroy (struct dict_compactor *compactor)
1005 if (compactor != NULL)
1007 free (compactor->maps);
1012 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
1013 dict_get_split_cnt() to determine how many SPLIT FILE vars
1014 there are. Returns a null pointer if and only if there are no
1016 const struct variable *const *
1017 dict_get_split_vars (const struct dictionary *d)
1024 /* Returns the number of SPLIT FILE vars. */
1026 dict_get_split_cnt (const struct dictionary *d)
1030 return d->split_cnt;
1033 /* Removes variable V from the set of split variables in dictionary D */
1035 dict_unset_split_var (struct dictionary *d,
1038 const int count = d->split_cnt;
1039 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
1040 &v, compare_var_ptrs, NULL);
1042 if ( count == d->split_cnt)
1045 if ( d->callbacks && d->callbacks->split_changed )
1046 d->callbacks->split_changed (d, d->cb_data);
1049 /* Sets CNT split vars SPLIT in dictionary D. */
1051 dict_set_split_vars (struct dictionary *d,
1052 struct variable *const *split, size_t cnt)
1055 assert (cnt == 0 || split != NULL);
1058 d->split = cnt > 0 ? xnrealloc (d->split, cnt, sizeof *d->split) : NULL;
1059 memcpy (d->split, split, cnt * sizeof *d->split);
1061 if ( d->callbacks && d->callbacks->split_changed )
1062 d->callbacks->split_changed (d, d->cb_data);
1065 /* Returns the file label for D, or a null pointer if D is
1066 unlabeled (see cmd_file_label()). */
1068 dict_get_label (const struct dictionary *d)
1075 /* Sets D's file label to LABEL, truncating it to a maximum of 60
1078 dict_set_label (struct dictionary *d, const char *label)
1085 else if (strlen (label) < 60)
1086 d->label = xstrdup (label);
1089 d->label = xmalloc (61);
1090 memcpy (d->label, label, 60);
1091 d->label[60] = '\0';
1095 /* Returns the documents for D, or a null pointer if D has no
1096 documents. If the return value is nonnull, then the string
1097 will be an exact multiple of DOC_LINE_LENGTH bytes in length,
1098 with each segment corresponding to one line. */
1100 dict_get_documents (const struct dictionary *d)
1102 return ds_is_empty (&d->documents) ? NULL : ds_cstr (&d->documents);
1105 /* Sets the documents for D to DOCUMENTS, or removes D's
1106 documents if DOCUMENT is a null pointer. If DOCUMENTS is
1107 nonnull, then it should be an exact multiple of
1108 DOC_LINE_LENGTH bytes in length, with each segment
1109 corresponding to one line. */
1111 dict_set_documents (struct dictionary *d, const char *documents)
1115 ds_assign_cstr (&d->documents, documents != NULL ? documents : "");
1117 /* In case the caller didn't get it quite right, pad out the
1118 final line with spaces. */
1119 remainder = ds_length (&d->documents) % DOC_LINE_LENGTH;
1121 ds_put_char_multiple (&d->documents, ' ', DOC_LINE_LENGTH - remainder);
1124 /* Drops the documents from dictionary D. */
1126 dict_clear_documents (struct dictionary *d)
1128 ds_clear (&d->documents);
1131 /* Appends LINE to the documents in D. LINE will be truncated or
1132 padded on the right with spaces to make it exactly
1133 DOC_LINE_LENGTH bytes long. */
1135 dict_add_document_line (struct dictionary *d, const char *line)
1137 if (strlen (line) > DOC_LINE_LENGTH)
1139 /* Note to translators: "bytes" is correct, not characters */
1140 msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
1142 buf_copy_str_rpad (ds_put_uninit (&d->documents, DOC_LINE_LENGTH),
1143 DOC_LINE_LENGTH, line);
1146 /* Returns the number of document lines in dictionary D. */
1148 dict_get_document_line_cnt (const struct dictionary *d)
1150 return ds_length (&d->documents) / DOC_LINE_LENGTH;
1153 /* Copies document line number IDX from dictionary D into
1154 LINE, trimming off any trailing white space. */
1156 dict_get_document_line (const struct dictionary *d,
1157 size_t idx, struct string *line)
1159 assert (idx < dict_get_document_line_cnt (d));
1160 ds_assign_substring (line, ds_substr (&d->documents, idx * DOC_LINE_LENGTH,
1162 ds_rtrim (line, ss_cstr (CC_SPACES));
1165 /* Creates in D a vector named NAME that contains the CNT
1166 variables in VAR. Returns true if successful, or false if a
1167 vector named NAME already exists in D. */
1169 dict_create_vector (struct dictionary *d,
1171 struct variable **var, size_t cnt)
1175 assert (var != NULL);
1177 for (i = 0; i < cnt; i++)
1178 assert (dict_contains_var (d, var[i]));
1180 if (dict_lookup_vector (d, name) == NULL)
1182 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1183 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1190 /* Creates in D a vector named NAME that contains the CNT
1191 variables in VAR. A vector named NAME must not already exist
1194 dict_create_vector_assert (struct dictionary *d,
1196 struct variable **var, size_t cnt)
1198 assert (dict_lookup_vector (d, name) == NULL);
1199 dict_create_vector (d, name, var, cnt);
1202 /* Returns the vector in D with index IDX, which must be less
1203 than dict_get_vector_cnt (D). */
1204 const struct vector *
1205 dict_get_vector (const struct dictionary *d, size_t idx)
1208 assert (idx < d->vector_cnt);
1210 return d->vector[idx];
1213 /* Returns the number of vectors in D. */
1215 dict_get_vector_cnt (const struct dictionary *d)
1219 return d->vector_cnt;
1222 /* Looks up and returns the vector within D with the given
1224 const struct vector *
1225 dict_lookup_vector (const struct dictionary *d, const char *name)
1228 for (i = 0; i < d->vector_cnt; i++)
1229 if (!strcasecmp (vector_get_name (d->vector[i]), name))
1230 return d->vector[i];
1234 /* Deletes all vectors from D. */
1236 dict_clear_vectors (struct dictionary *d)
1240 for (i = 0; i < d->vector_cnt; i++)
1241 vector_destroy (d->vector[i]);
1248 /* Compares two strings. */
1250 compare_strings (const void *a, const void *b, const void *aux UNUSED)
1252 return strcmp (a, b);
1255 /* Hashes a string. */
1257 hash_string (const void *s, const void *aux UNUSED)
1259 return hsh_hash_string (s);
1263 /* Sets V's short name to BASE, followed by a suffix of the form
1264 _A, _B, _C, ..., _AA, _AB, etc. according to the value of
1265 SUFFIX_NUMBER. Truncates BASE as necessary to fit. */
1267 set_var_short_name_suffix (struct variable *v, const char *base,
1270 char suffix[SHORT_NAME_LEN + 1];
1271 char short_name[SHORT_NAME_LEN + 1];
1276 assert (suffix_number >= 0);
1278 /* Set base name. */
1279 var_set_short_name (v, base);
1281 /* Compose suffix. */
1282 start = end = suffix + sizeof suffix - 1;
1286 *--start = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"[suffix_number % 26];
1287 if (start <= suffix + 1)
1288 msg (SE, _("Variable suffix too large."));
1289 suffix_number /= 26;
1291 while (suffix_number > 0);
1294 /* Append suffix to V's short name. */
1295 str_copy_trunc (short_name, sizeof short_name, base);
1297 if (len + strlen (short_name) > SHORT_NAME_LEN)
1298 ofs = SHORT_NAME_LEN - len;
1300 ofs = strlen (short_name);
1301 strcpy (short_name + ofs, start);
1304 var_set_short_name (v, short_name);
1307 /* Assigns a valid, unique short_name[] to each variable in D.
1308 Each variable whose actual name is short has highest priority
1309 for that short name. Otherwise, variables with an existing
1310 short_name[] have the next highest priority for a given short
1311 name; if it is already taken, then the variable is treated as
1312 if short_name[] had been empty. Otherwise, long names are
1313 truncated to form short names. If that causes conflicts,
1314 variables are renamed as PREFIX_A, PREFIX_B, and so on. */
1316 dict_assign_short_names (struct dictionary *d)
1318 struct hsh_table *short_names;
1321 /* Give variables whose names are short the corresponding short
1322 names, and clear short_names[] that conflict with a variable
1324 for (i = 0; i < d->var_cnt; i++)
1326 struct variable *v = d->var[i];
1327 const char *short_name = var_get_short_name (v);
1328 if (strlen (var_get_name (v)) <= SHORT_NAME_LEN)
1329 var_set_short_name (v, var_get_name (v));
1330 else if (short_name != NULL && dict_lookup_var (d, short_name) != NULL)
1331 var_clear_short_name (v);
1334 /* Each variable with an assigned short_name[] now gets it
1335 unless there is a conflict. */
1336 short_names = hsh_create (d->var_cnt, compare_strings, hash_string,
1338 for (i = 0; i < d->var_cnt; i++)
1340 struct variable *v = d->var[i];
1341 const char *name = var_get_short_name (v);
1342 if (name != NULL && hsh_insert (short_names, (char *) name) != NULL)
1343 var_clear_short_name (v);
1346 /* Now assign short names to remaining variables. */
1347 for (i = 0; i < d->var_cnt; i++)
1349 struct variable *v = d->var[i];
1350 const char *name = var_get_short_name (v);
1353 /* Form initial short_name from the variable name, then
1354 try _A, _B, ... _AA, _AB, etc., if needed.*/
1359 var_set_short_name (v, var_get_name (v));
1361 set_var_short_name_suffix (v, var_get_name (v), trial - 1);
1365 while (hsh_insert (short_names, (char *) var_get_short_name (v))
1370 /* Get rid of hash table. */
1371 hsh_destroy (short_names);
1375 /* Called from variable.c to notify the dictionary that some property of
1376 the variable has changed */
1378 dict_var_changed (const struct variable *v)
1380 if ( var_has_vardict (v))
1382 const struct vardict_info *vdi = var_get_vardict (v);
1383 struct dictionary *d;
1387 if ( d->callbacks && d->callbacks->var_changed )
1388 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);