1 /* PSPP - computes sample statistics.
2 Copyright (C) 1997-9, 2000, 2006, 2007 Free Software Foundation, Inc.
4 This program is free software; you can redistribute it and/or
5 modify it under the terms of the GNU General Public License as
6 published by the Free Software Foundation; either version 2 of the
7 License, or (at your option) any later version.
9 This program is distributed in the hope that it will be useful, but
10 WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 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, write to the Free Software
16 Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
21 #include "dictionary.h"
27 #include "cat-routines.h"
30 #include "value-labels.h"
34 #include <libpspp/alloc.h>
35 #include <libpspp/array.h>
36 #include <libpspp/compiler.h>
37 #include <libpspp/hash.h>
38 #include <libpspp/message.h>
39 #include <libpspp/misc.h>
40 #include <libpspp/pool.h>
41 #include <libpspp/str.h>
46 #define _(msgid) gettext (msgid)
51 struct variable **var; /* Variables. */
52 size_t var_cnt, var_cap; /* Number of variables, capacity. */
53 struct hsh_table *name_tab; /* Variable index by name. */
54 int next_value_idx; /* Index of next `union value' to allocate. */
55 const struct variable **split; /* SPLIT FILE vars. */
56 size_t split_cnt; /* SPLIT FILE count. */
57 struct variable *weight; /* WEIGHT variable. */
58 struct variable *filter; /* FILTER variable. */
59 size_t case_limit; /* Current case limit (N command). */
60 char *label; /* File label. */
61 char *documents; /* Documents, as a string. */
62 struct vector **vector; /* Vectors of variables. */
63 size_t vector_cnt; /* Number of vectors. */
64 const struct dict_callbacks *callbacks; /* Callbacks on dictionary
66 void *cb_data ; /* Data passed to callbacks */
70 /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
71 the dictionary or any of the variables it contains are modified.
72 Each callback will get passed CALLBACK_DATA.
73 Any callback may be NULL, in which case it'll be ignored.
76 dict_set_callbacks (struct dictionary *dict,
77 const struct dict_callbacks *callbacks,
80 dict->callbacks = callbacks;
81 dict->cb_data = callback_data;
84 /* Shallow copy the callbacks from SRC to DEST */
86 dict_copy_callbacks (struct dictionary *dest,
87 const struct dictionary *src)
89 dest->callbacks = src->callbacks;
90 dest->cb_data = src->cb_data;
93 /* Creates and returns a new dictionary. */
97 struct dictionary *d = xzalloc (sizeof *d);
99 d->name_tab = hsh_create (8, compare_vars_by_name, hash_var_by_name,
104 /* Creates and returns a (deep) copy of an existing
107 dict_clone (const struct dictionary *s)
109 struct dictionary *d;
116 for (i = 0; i < s->var_cnt; i++)
118 struct variable *sv = s->var[i];
119 struct variable *dv = dict_clone_var_assert (d, sv, var_get_name (sv));
120 var_set_short_name (dv, var_get_short_name (sv));
123 d->next_value_idx = s->next_value_idx;
125 d->split_cnt = s->split_cnt;
126 if (d->split_cnt > 0)
128 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
129 for (i = 0; i < d->split_cnt; i++)
130 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
133 if (s->weight != NULL)
134 dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight)));
136 if (s->filter != NULL)
137 dict_set_filter (d, dict_lookup_var_assert (d, var_get_name (s->filter)));
139 d->case_limit = s->case_limit;
140 dict_set_label (d, dict_get_label (s));
141 dict_set_documents (d, dict_get_documents (s));
143 d->vector_cnt = s->vector_cnt;
144 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
145 for (i = 0; i < s->vector_cnt; i++)
146 d->vector[i] = vector_clone (s->vector[i], s, d);
151 /* Clears the contents from a dictionary without destroying the
152 dictionary itself. */
154 dict_clear (struct dictionary *d)
156 /* FIXME? Should we really clear case_limit, label, documents?
157 Others are necessarily cleared by deleting all the variables.*/
160 while (d->var_cnt > 0 )
162 var_clear_vardict (d->var[d->var_cnt - 1]);
163 var_destroy (d->var[d->var_cnt -1]);
167 if (d->callbacks && d->callbacks->var_deleted )
168 d->callbacks->var_deleted (d, d->var_cnt, d->cb_data);
172 d->var_cnt = d->var_cap = 0;
173 hsh_clear (d->name_tab);
174 d->next_value_idx = 0;
175 dict_set_split_vars (d, NULL, 0);
176 dict_set_weight (d, NULL);
177 dict_set_filter (d, NULL);
183 dict_clear_vectors (d);
186 /* Destroys the aux data for every variable in D, by calling
187 var_clear_aux() for each variable. */
189 dict_clear_aux (struct dictionary *d)
195 for (i = 0; i < d->var_cnt; i++)
196 var_clear_aux (d->var[i]);
199 /* Clears a dictionary and destroys it. */
201 dict_destroy (struct dictionary *d)
205 /* In general, we don't want callbacks occuring, if the dictionary
206 is being destroyed */
207 d->callbacks = NULL ;
210 hsh_destroy (d->name_tab);
215 /* Returns the number of variables in D. */
217 dict_get_var_cnt (const struct dictionary *d)
224 /* Returns the variable in D with dictionary index IDX, which
225 must be between 0 and the count returned by
226 dict_get_var_cnt(), exclusive. */
228 dict_get_var (const struct dictionary *d, size_t idx)
231 assert (idx < d->var_cnt);
237 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
238 size_t *cnt, unsigned exclude_classes)
240 dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude_classes);
243 /* Sets *VARS to an array of pointers to variables in D and *CNT
244 to the number of variables in *D. All variables are returned
245 if EXCLUDE_CLASSES is 0, or it may contain one or more of (1u
246 << DC_ORDINARY), (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to
247 exclude the corresponding type of variable. */
249 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
250 size_t *cnt, unsigned exclude_classes)
256 assert (vars != NULL);
257 assert (cnt != NULL);
258 assert ((exclude_classes & ~((1u << DC_ORDINARY)
260 | (1u << DC_SCRATCH))) == 0);
263 for (i = 0; i < d->var_cnt; i++)
265 enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
266 if (!(exclude_classes & (1u << class)))
270 *vars = xnmalloc (count, sizeof **vars);
272 for (i = 0; i < d->var_cnt; i++)
274 enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
275 if (!(exclude_classes & (1u << class)))
276 (*vars)[(*cnt)++] = d->var[i];
278 assert (*cnt == count);
281 static struct variable *
282 add_var (struct dictionary *d, struct variable *v)
284 /* Add dictionary info to variable. */
285 struct vardict_info vdi;
286 vdi.case_index = d->next_value_idx;
287 vdi.dict_index = d->var_cnt;
289 var_set_vardict (v, &vdi);
291 /* Update dictionary. */
292 if (d->var_cnt >= d->var_cap)
294 d->var_cap = 8 + 2 * d->var_cap;
295 d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var);
297 d->var[d->var_cnt++] = v;
298 hsh_force_insert (d->name_tab, v);
300 if ( d->callbacks && d->callbacks->var_added )
301 d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
303 d->next_value_idx += var_get_value_cnt (v);
308 /* Creates and returns a new variable in D with the given NAME
309 and WIDTH. Returns a null pointer if the given NAME would
310 duplicate that of an existing variable in the dictionary. */
312 dict_create_var (struct dictionary *d, const char *name, int width)
314 return (dict_lookup_var (d, name) == NULL
315 ? dict_create_var_assert (d, name, width)
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 assert (dict_lookup_var (d, name) == NULL);
326 return add_var (d, var_create (name, width));
329 /* Creates and returns a new variable in D with name NAME, as a
330 copy of existing variable OLD_VAR, which need not be in D or
331 in any dictionary. Returns a null pointer if the given NAME
332 would duplicate that of an existing variable in the
335 dict_clone_var (struct dictionary *d, const struct variable *old_var,
338 return (dict_lookup_var (d, name) == NULL
339 ? dict_clone_var_assert (d, old_var, name)
343 /* Creates and returns a new variable in D with name NAME, as a
344 copy of existing variable OLD_VAR, which need not be in D or
345 in any dictionary. Assert-fails if the given NAME would
346 duplicate that of an existing variable in the dictionary. */
348 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var,
351 struct variable *new_var = var_clone (old_var);
352 assert (dict_lookup_var (d, name) == NULL);
353 var_set_name (new_var, name);
354 return add_var (d, new_var);
357 /* Returns the variable named NAME in D, or a null pointer if no
358 variable has that name. */
360 dict_lookup_var (const struct dictionary *d, const char *name)
362 struct variable *target ;
363 struct variable *result ;
365 if ( ! var_is_valid_name (name, false))
368 target = var_create (name, 0);
369 result = hsh_find (d->name_tab, target);
370 var_destroy (target);
375 /* Returns the variable named NAME in D. Assert-fails if no
376 variable has that name. */
378 dict_lookup_var_assert (const struct dictionary *d, const char *name)
380 struct variable *v = dict_lookup_var (d, name);
385 /* Returns true if variable V is in dictionary D,
388 dict_contains_var (const struct dictionary *d, const struct variable *v)
390 if (var_has_vardict (v))
392 const struct vardict_info *vdi = var_get_vardict (v);
393 return (vdi->dict_index >= 0
394 && vdi->dict_index < d->var_cnt
395 && d->var[vdi->dict_index] == v);
401 /* Compares two double pointers to variables, which should point
402 to elements of a struct dictionary's `var' member array. */
404 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
406 struct variable *const *a = a_;
407 struct variable *const *b = b_;
409 return *a < *b ? -1 : *a > *b;
412 /* Sets the dict_index in V's vardict to DICT_INDEX. */
414 set_var_dict_index (struct variable *v, int dict_index)
416 struct vardict_info vdi = *var_get_vardict (v);
417 struct dictionary *d = vdi.dict;
418 vdi.dict_index = dict_index;
419 var_set_vardict (v, &vdi);
421 if ( d->callbacks && d->callbacks->var_changed )
422 d->callbacks->var_changed (d, dict_index, d->cb_data);
425 /* Sets the case_index in V's vardict to DICT_INDEX. */
427 set_var_case_index (struct variable *v, int case_index)
429 struct vardict_info vdi = *var_get_vardict (v);
430 vdi.case_index = case_index;
431 var_set_vardict (v, &vdi);
434 /* Re-sets the dict_index in the dictionary variables with
435 indexes from FROM to TO (exclusive). */
437 reindex_vars (struct dictionary *d, size_t from, size_t to)
441 for (i = from; i < to; i++)
442 set_var_dict_index (d->var[i], i);
445 /* Deletes variable V from dictionary D and frees V.
447 This is a very bad idea if there might be any pointers to V
448 from outside D. In general, no variable in should be deleted when
449 any transformations are active on the dictionary's dataset, because
450 those transformations might reference the deleted variable.
451 The safest time to delete a variable is just after a procedure
452 has been executed, as done by MODIFY VARS.
454 Pointers to V within D are not a problem, because
455 dict_delete_var() knows to remove V from split variables,
456 weights, filters, etc. */
458 dict_delete_var (struct dictionary *d, struct variable *v)
460 int dict_index = var_get_dict_index (v);
462 assert (dict_contains_var (d, v));
464 /* Delete aux data. */
467 dict_unset_split_var (d, v);
470 dict_set_weight (d, NULL);
473 dict_set_filter (d, NULL);
475 dict_clear_vectors (d);
477 /* Remove V from var array. */
478 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
481 /* Update dict_index for each affected variable. */
482 reindex_vars (d, dict_index, d->var_cnt);
484 /* Update name hash. */
485 hsh_force_delete (d->name_tab, v);
489 var_clear_vardict (v);
492 if (d->callbacks && d->callbacks->var_deleted )
493 d->callbacks->var_deleted (d, dict_index, d->cb_data);
496 /* Deletes the COUNT variables listed in VARS from D. This is
497 unsafe; see the comment on dict_delete_var() for details. */
499 dict_delete_vars (struct dictionary *d,
500 struct variable *const *vars, size_t count)
502 /* FIXME: this can be done in O(count) time, but this algorithm
505 assert (count == 0 || vars != NULL);
508 dict_delete_var (d, *vars++);
511 /* Deletes the COUNT variables in D starting at index IDX. This
512 is unsafe; see the comment on dict_delete_var() for
515 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
517 /* FIXME: this can be done in O(count) time, but this algorithm
519 assert (idx + count <= d->var_cnt);
522 dict_delete_var (d, d->var[idx]);
525 /* Deletes scratch variables from dictionary D. */
527 dict_delete_scratch_vars (struct dictionary *d)
531 /* FIXME: this can be done in O(count) time, but this algorithm
535 for (i = 0; i < d->var_cnt; )
536 if (dict_class_from_id (var_get_name (d->var[i])) == DC_SCRATCH)
537 dict_delete_var (d, d->var[i]);
542 /* Moves V to 0-based position IDX in D. Other variables in D,
543 if any, retain their relative positions. Runs in time linear
544 in the distance moved. */
546 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
548 size_t old_index = var_get_dict_index (v);
550 assert (new_index < d->var_cnt);
551 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
552 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
555 /* Reorders the variables in D, placing the COUNT variables
556 listed in ORDER in that order at the beginning of D. The
557 other variables in D, if any, retain their relative
560 dict_reorder_vars (struct dictionary *d,
561 struct variable *const *order, size_t count)
563 struct variable **new_var;
567 assert (count == 0 || order != NULL);
568 assert (count <= d->var_cnt);
570 new_var = xnmalloc (d->var_cnt, sizeof *new_var);
571 memcpy (new_var, order, count * sizeof *new_var);
572 for (i = 0; i < count; i++)
574 size_t index = var_get_dict_index (order[i]);
575 assert (d->var[index] == order[i]);
576 d->var[index] = NULL;
577 set_var_dict_index (order[i], i);
579 for (i = 0; i < d->var_cnt; i++)
580 if (d->var[i] != NULL)
582 assert (count < d->var_cnt);
583 new_var[count] = d->var[i];
584 set_var_dict_index (new_var[count], count);
591 /* Changes the name of variable V in dictionary D to NEW_NAME. */
593 rename_var (struct dictionary *d, struct variable *v, const char *new_name)
595 struct vardict_info vdi;
597 assert (dict_contains_var (d, v));
599 vdi = *var_get_vardict (v);
600 var_clear_vardict (v);
601 var_set_name (v, new_name);
602 var_set_vardict (v, &vdi);
605 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
606 a variable named NEW_NAME is already in D, except that
607 NEW_NAME may be the same as V's existing name. */
609 dict_rename_var (struct dictionary *d, struct variable *v,
610 const char *new_name)
612 assert (!strcasecmp (var_get_name (v), new_name)
613 || dict_lookup_var (d, new_name) == NULL);
615 hsh_force_delete (d->name_tab, v);
616 rename_var (d, v, new_name);
617 hsh_force_insert (d->name_tab, v);
619 if (get_algorithm () == ENHANCED)
620 var_clear_short_name (v);
622 if ( d->callbacks && d->callbacks->var_changed )
623 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
626 /* Renames COUNT variables specified in VARS to the names given
627 in NEW_NAMES within dictionary D. If the renaming would
628 result in a duplicate variable name, returns false and stores a
629 name that would be duplicated into *ERR_NAME (if ERR_NAME is
630 non-null). Otherwise, the renaming is successful, and true
633 dict_rename_vars (struct dictionary *d,
634 struct variable **vars, char **new_names, size_t count,
641 assert (count == 0 || vars != NULL);
642 assert (count == 0 || new_names != NULL);
644 /* Save the names of the variables to be renamed. */
645 pool = pool_create ();
646 old_names = pool_nalloc (pool, count, sizeof *old_names);
647 for (i = 0; i < count; i++)
648 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
650 /* Remove the variables to be renamed from the name hash,
652 for (i = 0; i < count; i++)
654 hsh_force_delete (d->name_tab, vars[i]);
655 rename_var (d, vars[i], new_names[i]);
658 /* Add the renamed variables back into the name hash,
659 checking for conflicts. */
660 for (i = 0; i < count; i++)
661 if (hsh_insert (d->name_tab, vars[i]) != NULL)
663 /* There is a name conflict.
664 Back out all the name changes that have already
665 taken place, and indicate failure. */
667 if (err_name != NULL)
668 *err_name = new_names[i];
670 for (i = 0; i < fail_idx; i++)
671 hsh_force_delete (d->name_tab, vars[i]);
673 for (i = 0; i < count; i++)
675 rename_var (d, vars[i], old_names[i]);
676 hsh_force_insert (d->name_tab, vars[i]);
683 /* Clear short names. */
684 if (get_algorithm () == ENHANCED)
685 for (i = 0; i < count; i++)
686 var_clear_short_name (vars[i]);
692 /* Returns the weighting variable in dictionary D, or a null
693 pointer if the dictionary is unweighted. */
695 dict_get_weight (const struct dictionary *d)
698 assert (d->weight == NULL || dict_contains_var (d, d->weight));
703 /* Returns the value of D's weighting variable in case C, except that a
704 negative weight is returned as 0. Returns 1 if the dictionary is
705 unweighted. Will warn about missing, negative, or zero values if
706 warn_on_invalid is true. The function will set warn_on_invalid to false
707 if an invalid weight is found. */
709 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
710 bool *warn_on_invalid)
715 if (d->weight == NULL)
719 double w = case_num (c, d->weight);
720 if (w < 0.0 || var_is_num_missing (d->weight, w, MV_ANY))
722 if ( w == 0.0 && *warn_on_invalid ) {
723 *warn_on_invalid = false;
724 msg (SW, _("At least one case in the data file had a weight value "
725 "that was user-missing, system-missing, zero, or "
726 "negative. These case(s) were ignored."));
732 /* Sets the weighting variable of D to V, or turning off
733 weighting if V is a null pointer. */
735 dict_set_weight (struct dictionary *d, struct variable *v)
738 assert (v == NULL || dict_contains_var (d, v));
739 assert (v == NULL || var_is_numeric (v));
743 if ( d->callbacks && d->callbacks->weight_changed )
744 d->callbacks->weight_changed (d,
745 v ? var_get_dict_index (v) : -1,
749 /* Returns the filter variable in dictionary D (see cmd_filter())
750 or a null pointer if the dictionary is unfiltered. */
752 dict_get_filter (const struct dictionary *d)
755 assert (d->filter == NULL || dict_contains_var (d, d->filter));
760 /* Sets V as the filter variable for dictionary D. Passing a
761 null pointer for V turn off filtering. */
763 dict_set_filter (struct dictionary *d, struct variable *v)
766 assert (v == NULL || dict_contains_var (d, v));
770 if ( d->callbacks && d->callbacks->filter_changed )
771 d->callbacks->filter_changed (d,
772 v ? var_get_dict_index (v) : -1,
776 /* Returns the case limit for dictionary D, or zero if the number
777 of cases is unlimited. */
779 dict_get_case_limit (const struct dictionary *d)
783 return d->case_limit;
786 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
787 0 for CASE_LIMIT to indicate no limit. */
789 dict_set_case_limit (struct dictionary *d, size_t case_limit)
793 d->case_limit = case_limit;
796 /* Returns the case index of the next value to be added to D.
797 This value is the number of `union value's that need to be
798 allocated to store a case for dictionary D. */
800 dict_get_next_value_idx (const struct dictionary *d)
804 return d->next_value_idx;
807 /* Returns the number of bytes needed to store a case for
810 dict_get_case_size (const struct dictionary *d)
814 return sizeof (union value) * dict_get_next_value_idx (d);
817 /* Deletes scratch variables in dictionary D and reassigns values
818 so that fragmentation is eliminated. */
820 dict_compact_values (struct dictionary *d)
824 d->next_value_idx = 0;
825 for (i = 0; i < d->var_cnt; )
827 struct variable *v = d->var[i];
829 if (dict_class_from_id (var_get_name (v)) != DC_SCRATCH)
831 set_var_case_index (v, d->next_value_idx);
832 d->next_value_idx += var_get_value_cnt (v);
836 dict_delete_var (d, v);
840 /* Returns the number of values that would be used by a case if
841 dict_compact_values() were called. */
843 dict_get_compacted_value_cnt (const struct dictionary *d)
849 for (i = 0; i < d->var_cnt; i++)
850 if (dict_class_from_id (var_get_name (d->var[i])) != DC_SCRATCH)
851 cnt += var_get_value_cnt (d->var[i]);
855 /* Creates and returns an array mapping from a dictionary index
856 to the case index that the corresponding variable will have
857 after calling dict_compact_values(). Scratch variables
858 receive -1 for case index because dict_compact_values() will
861 dict_get_compacted_dict_index_to_case_index (const struct dictionary *d)
864 size_t next_value_idx;
867 map = xnmalloc (d->var_cnt, sizeof *map);
869 for (i = 0; i < d->var_cnt; i++)
871 struct variable *v = d->var[i];
873 if (dict_class_from_id (var_get_name (v)) != DC_SCRATCH)
875 map[i] = next_value_idx;
876 next_value_idx += var_get_value_cnt (v);
884 /* Returns true if a case for dictionary D would be smaller after
885 compacting, false otherwise. Compacting a case eliminates
886 "holes" between values and after the last value. Holes are
887 created by deleting variables (or by scratch variables).
889 The return value may differ from whether compacting a case
890 from dictionary D would *change* the case: compacting could
891 rearrange values even if it didn't reduce space
894 dict_compacting_would_shrink (const struct dictionary *d)
896 return dict_get_compacted_value_cnt (d) < dict_get_next_value_idx (d);
899 /* Returns true if a case for dictionary D would change after
900 compacting, false otherwise. Compacting a case eliminates
901 "holes" between values and after the last value. Holes are
902 created by deleting variables (or by scratch variables).
904 The return value may differ from whether compacting a case
905 from dictionary D would *shrink* the case: compacting could
906 rearrange values without reducing space requirements. */
908 dict_compacting_would_change (const struct dictionary *d)
914 for (i = 0; i < dict_get_var_cnt (d); i++)
916 struct variable *v = dict_get_var (d, i);
917 if (var_get_case_index (v) != case_idx)
919 case_idx += var_get_value_cnt (v);
924 /* How to copy a contiguous range of values between cases. */
927 size_t src_idx; /* Starting value index in source case. */
928 size_t dst_idx; /* Starting value index in target case. */
929 size_t cnt; /* Number of values. */
932 /* How to compact a case. */
933 struct dict_compactor
935 struct copy_map *maps; /* Array of mappings. */
936 size_t map_cnt; /* Number of mappings. */
939 /* Creates and returns a dict_compactor that can be used to
940 compact cases for dictionary D.
942 Compacting a case eliminates "holes" between values and after
943 the last value. Holes are created by deleting variables (or
944 by scratch variables). */
945 struct dict_compactor *
946 dict_make_compactor (const struct dictionary *d)
948 struct dict_compactor *compactor;
949 struct copy_map *map;
950 size_t map_allocated;
954 compactor = xmalloc (sizeof *compactor);
955 compactor->maps = NULL;
956 compactor->map_cnt = 0;
961 for (i = 0; i < d->var_cnt; i++)
963 struct variable *v = d->var[i];
965 if (dict_class_from_id (var_get_name (v)) == DC_SCRATCH)
967 if (map != NULL && map->src_idx + map->cnt == var_get_case_index (v))
968 map->cnt += var_get_value_cnt (v);
971 if (compactor->map_cnt == map_allocated)
972 compactor->maps = x2nrealloc (compactor->maps, &map_allocated,
973 sizeof *compactor->maps);
974 map = &compactor->maps[compactor->map_cnt++];
975 map->src_idx = var_get_case_index (v);
976 map->dst_idx = value_idx;
977 map->cnt = var_get_value_cnt (v);
979 value_idx += var_get_value_cnt (v);
985 /* Compacts SRC by copying it to DST according to the scheme in
988 Compacting a case eliminates "holes" between values and after
989 the last value. Holes are created by deleting variables (or
990 by scratch variables). */
992 dict_compactor_compact (const struct dict_compactor *compactor,
993 struct ccase *dst, const struct ccase *src)
997 for (i = 0; i < compactor->map_cnt; i++)
999 const struct copy_map *map = &compactor->maps[i];
1000 case_copy (dst, map->dst_idx, src, map->src_idx, map->cnt);
1004 /* Destroys COMPACTOR. */
1006 dict_compactor_destroy (struct dict_compactor *compactor)
1008 if (compactor != NULL)
1010 free (compactor->maps);
1015 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
1016 dict_get_split_cnt() to determine how many SPLIT FILE vars
1017 there are. Returns a null pointer if and only if there are no
1019 const struct variable *const *
1020 dict_get_split_vars (const struct dictionary *d)
1027 /* Returns the number of SPLIT FILE vars. */
1029 dict_get_split_cnt (const struct dictionary *d)
1033 return d->split_cnt;
1036 /* Removes variable V from the set of split variables in dictionary D */
1038 dict_unset_split_var (struct dictionary *d,
1041 const int count = d->split_cnt;
1042 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
1043 &v, compare_var_ptrs, NULL);
1045 if ( count == d->split_cnt)
1048 if ( d->callbacks && d->callbacks->split_changed )
1049 d->callbacks->split_changed (d, d->cb_data);
1052 /* Sets CNT split vars SPLIT in dictionary D. */
1054 dict_set_split_vars (struct dictionary *d,
1055 struct variable *const *split, size_t cnt)
1058 assert (cnt == 0 || split != NULL);
1061 d->split = xnrealloc (d->split, cnt, sizeof *d->split);
1062 memcpy (d->split, split, cnt * sizeof *d->split);
1064 if ( d->callbacks && d->callbacks->split_changed )
1065 d->callbacks->split_changed (d, d->cb_data);
1068 /* Returns the file label for D, or a null pointer if D is
1069 unlabeled (see cmd_file_label()). */
1071 dict_get_label (const struct dictionary *d)
1078 /* Sets D's file label to LABEL, truncating it to a maximum of 60
1081 dict_set_label (struct dictionary *d, const char *label)
1088 else if (strlen (label) < 60)
1089 d->label = xstrdup (label);
1092 d->label = xmalloc (61);
1093 memcpy (d->label, label, 60);
1094 d->label[60] = '\0';
1098 /* Returns the documents for D, or a null pointer if D has no
1099 documents (see cmd_document()).. */
1101 dict_get_documents (const struct dictionary *d)
1105 return d->documents;
1108 /* Sets the documents for D to DOCUMENTS, or removes D's
1109 documents if DOCUMENT is a null pointer. */
1111 dict_set_documents (struct dictionary *d, const char *documents)
1115 free (d->documents);
1116 if (documents == NULL)
1117 d->documents = NULL;
1119 d->documents = xstrdup (documents);
1122 /* Creates in D a vector named NAME that contains the CNT
1123 variables in VAR. Returns true if successful, or false if a
1124 vector named NAME already exists in D. */
1126 dict_create_vector (struct dictionary *d,
1128 struct variable **var, size_t cnt)
1132 assert (var != NULL);
1134 for (i = 0; i < cnt; i++)
1135 assert (dict_contains_var (d, var[i]));
1137 if (dict_lookup_vector (d, name) == NULL)
1139 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1140 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1147 /* Creates in D a vector named NAME that contains the CNT
1148 variables in VAR. A vector named NAME must not already exist
1151 dict_create_vector_assert (struct dictionary *d,
1153 struct variable **var, size_t cnt)
1155 assert (dict_lookup_vector (d, name) == NULL);
1156 dict_create_vector (d, name, var, cnt);
1159 /* Returns the vector in D with index IDX, which must be less
1160 than dict_get_vector_cnt (D). */
1161 const struct vector *
1162 dict_get_vector (const struct dictionary *d, size_t idx)
1165 assert (idx < d->vector_cnt);
1167 return d->vector[idx];
1170 /* Returns the number of vectors in D. */
1172 dict_get_vector_cnt (const struct dictionary *d)
1176 return d->vector_cnt;
1179 /* Looks up and returns the vector within D with the given
1181 const struct vector *
1182 dict_lookup_vector (const struct dictionary *d, const char *name)
1185 for (i = 0; i < d->vector_cnt; i++)
1186 if (!strcasecmp (vector_get_name (d->vector[i]), name))
1187 return d->vector[i];
1191 /* Deletes all vectors from D. */
1193 dict_clear_vectors (struct dictionary *d)
1197 for (i = 0; i < d->vector_cnt; i++)
1198 vector_destroy (d->vector[i]);
1205 /* Compares two strings. */
1207 compare_strings (const void *a, const void *b, const void *aux UNUSED)
1209 return strcmp (a, b);
1212 /* Hashes a string. */
1214 hash_string (const void *s, const void *aux UNUSED)
1216 return hsh_hash_string (s);
1220 /* Sets V's short name to BASE, followed by a suffix of the form
1221 _A, _B, _C, ..., _AA, _AB, etc. according to the value of
1222 SUFFIX_NUMBER. Truncates BASE as necessary to fit. */
1224 set_var_short_name_suffix (struct variable *v, const char *base,
1227 char suffix[SHORT_NAME_LEN + 1];
1228 char short_name[SHORT_NAME_LEN + 1];
1233 assert (suffix_number >= 0);
1235 /* Set base name. */
1236 var_set_short_name (v, base);
1238 /* Compose suffix. */
1239 start = end = suffix + sizeof suffix - 1;
1243 *--start = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"[suffix_number % 26];
1244 if (start <= suffix + 1)
1245 msg (SE, _("Variable suffix too large."));
1246 suffix_number /= 26;
1248 while (suffix_number > 0);
1251 /* Append suffix to V's short name. */
1252 str_copy_trunc (short_name, sizeof short_name, base);
1254 if (len + strlen (short_name) > SHORT_NAME_LEN)
1255 ofs = SHORT_NAME_LEN - len;
1257 ofs = strlen (short_name);
1258 strcpy (short_name + ofs, start);
1261 var_set_short_name (v, short_name);
1264 /* Assigns a valid, unique short_name[] to each variable in D.
1265 Each variable whose actual name is short has highest priority
1266 for that short name. Otherwise, variables with an existing
1267 short_name[] have the next highest priority for a given short
1268 name; if it is already taken, then the variable is treated as
1269 if short_name[] had been empty. Otherwise, long names are
1270 truncated to form short names. If that causes conflicts,
1271 variables are renamed as PREFIX_A, PREFIX_B, and so on. */
1273 dict_assign_short_names (struct dictionary *d)
1275 struct hsh_table *short_names;
1278 /* Give variables whose names are short the corresponding short
1279 names, and clear short_names[] that conflict with a variable
1281 for (i = 0; i < d->var_cnt; i++)
1283 struct variable *v = d->var[i];
1284 const char *short_name = var_get_short_name (v);
1285 if (strlen (var_get_name (v)) <= SHORT_NAME_LEN)
1286 var_set_short_name (v, var_get_name (v));
1287 else if (short_name != NULL && dict_lookup_var (d, short_name) != NULL)
1288 var_clear_short_name (v);
1291 /* Each variable with an assigned short_name[] now gets it
1292 unless there is a conflict. */
1293 short_names = hsh_create (d->var_cnt, compare_strings, hash_string,
1295 for (i = 0; i < d->var_cnt; i++)
1297 struct variable *v = d->var[i];
1298 const char *name = var_get_short_name (v);
1299 if (name != NULL && hsh_insert (short_names, (char *) name) != NULL)
1300 var_clear_short_name (v);
1303 /* Now assign short names to remaining variables. */
1304 for (i = 0; i < d->var_cnt; i++)
1306 struct variable *v = d->var[i];
1307 const char *name = var_get_short_name (v);
1310 /* Form initial short_name from the variable name, then
1311 try _A, _B, ... _AA, _AB, etc., if needed.*/
1316 var_set_short_name (v, var_get_name (v));
1318 set_var_short_name_suffix (v, var_get_name (v), trial - 1);
1322 while (hsh_insert (short_names, (char *) var_get_short_name (v))
1327 /* Get rid of hash table. */
1328 hsh_destroy (short_names);
1332 /* Called from variable.c to notify the dictionary that some property of
1333 the variable has changed */
1335 dict_var_changed (const struct variable *v)
1337 if ( var_has_vardict (v))
1339 const struct vardict_info *vdi = var_get_vardict (v);
1340 struct dictionary *d;
1344 if ( d->callbacks && d->callbacks->var_changed )
1345 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);