1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006, 2007, 2009, 2010, 2011, 2012, 2013, 2014, 2015 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 "data/dictionary.h"
26 #include "data/attributes.h"
27 #include "data/case.h"
28 #include "data/identifier.h"
29 #include "data/mrset.h"
30 #include "data/settings.h"
31 #include "data/value-labels.h"
32 #include "data/vardict.h"
33 #include "data/variable.h"
34 #include "data/vector.h"
35 #include "libpspp/array.h"
36 #include "libpspp/assertion.h"
37 #include "libpspp/compiler.h"
38 #include "libpspp/hash-functions.h"
39 #include "libpspp/hmap.h"
40 #include "libpspp/i18n.h"
41 #include "libpspp/message.h"
42 #include "libpspp/misc.h"
43 #include "libpspp/pool.h"
44 #include "libpspp/str.h"
45 #include "libpspp/string-array.h"
47 #include "gl/intprops.h"
48 #include "gl/minmax.h"
49 #include "gl/xalloc.h"
50 #include "gl/xmemdup0.h"
53 #define _(msgid) gettext (msgid)
59 struct vardict_info *var; /* Variables. */
60 size_t var_cnt, var_cap; /* Number of variables, capacity. */
61 struct caseproto *proto; /* Prototype for dictionary cases
63 struct hmap name_map; /* Variable index by name. */
64 int next_value_idx; /* Index of next `union value' to allocate. */
65 const struct variable **split; /* SPLIT FILE vars. */
66 size_t split_cnt; /* SPLIT FILE count. */
67 struct variable *weight; /* WEIGHT variable. */
68 struct variable *filter; /* FILTER variable. */
69 casenumber case_limit; /* Current case limit (N command). */
70 char *label; /* File label. */
71 struct string_array documents; /* Documents. */
72 struct vector **vector; /* Vectors of variables. */
73 size_t vector_cnt; /* Number of vectors. */
74 struct attrset attributes; /* Custom attributes. */
75 struct mrset **mrsets; /* Multiple response sets. */
76 size_t n_mrsets; /* Number of multiple response sets. */
78 /* Whether variable names must be valid identifiers. Normally, this is
79 true, but sometimes a dictionary is prepared for external use
80 (e.g. output to a CSV file) where names don't have to be valid. */
81 bool names_must_be_ids;
83 char *encoding; /* Character encoding of string data */
85 const struct dict_callbacks *callbacks; /* Callbacks on dictionary
87 void *cb_data ; /* Data passed to callbacks */
89 void (*changed) (struct dictionary *, void *); /* Generic change callback */
93 static void dict_unset_split_var (struct dictionary *, struct variable *);
94 static void dict_unset_mrset_var (struct dictionary *, struct variable *);
96 /* Returns the encoding for data in dictionary D. The return value is a
97 nonnull string that contains an IANA character set name. */
99 dict_get_encoding (const struct dictionary *d)
104 /* Returns true if UTF-8 string ID is an acceptable identifier in DICT's
105 encoding, false otherwise. If ISSUE_ERROR is true, issues an explanatory
106 error message on failure. */
108 dict_id_is_valid (const struct dictionary *dict, const char *id,
111 return (!dict->names_must_be_ids
112 || id_is_valid (id, dict->encoding, issue_error));
116 dict_set_change_callback (struct dictionary *d,
117 void (*changed) (struct dictionary *, void*),
120 d->changed = changed;
121 d->changed_data = data;
124 /* Discards dictionary D's caseproto. (It will be regenerated
125 lazily, on demand.) */
127 invalidate_proto (struct dictionary *d)
129 caseproto_unref (d->proto);
133 /* Print a representation of dictionary D to stdout, for
134 debugging purposes. */
136 dict_dump (const struct dictionary *d)
139 for (i = 0 ; i < d->var_cnt ; ++i )
141 const struct variable *v = d->var[i].var;
142 printf ("Name: %s;\tdict_idx: %zu; case_idx: %zu\n",
144 var_get_dict_index (v),
145 var_get_case_index (v));
150 /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
151 the dictionary or any of the variables it contains are modified.
152 Each callback will get passed CALLBACK_DATA.
153 Any callback may be NULL, in which case it'll be ignored.
156 dict_set_callbacks (struct dictionary *dict,
157 const struct dict_callbacks *callbacks,
160 dict->callbacks = callbacks;
161 dict->cb_data = callback_data;
164 /* Shallow copy the callbacks from SRC to DEST */
166 dict_copy_callbacks (struct dictionary *dest,
167 const struct dictionary *src)
169 dest->callbacks = src->callbacks;
170 dest->cb_data = src->cb_data;
173 /* Creates and returns a new dictionary with the specified ENCODING. */
175 dict_create (const char *encoding)
177 struct dictionary *d = xzalloc (sizeof *d);
179 d->encoding = xstrdup (encoding);
180 d->names_must_be_ids = true;
181 hmap_init (&d->name_map);
182 attrset_init (&d->attributes);
189 dict_ref (struct dictionary *s)
195 /* Creates and returns a (deep) copy of an existing
198 The new dictionary's case indexes are copied from the old
199 dictionary. If the new dictionary won't be used to access
200 cases produced with the old dictionary, then the new
201 dictionary's case indexes should be compacted with
202 dict_compact_values to save space.
204 Callbacks are not cloned. */
206 dict_clone (const struct dictionary *s)
208 struct dictionary *d;
211 d = dict_create (s->encoding);
212 dict_set_names_must_be_ids (d, dict_get_names_must_be_ids (s));
214 for (i = 0; i < s->var_cnt; i++)
216 struct variable *sv = s->var[i].var;
217 struct variable *dv = dict_clone_var_assert (d, sv);
220 for (i = 0; i < var_get_short_name_cnt (sv); i++)
221 var_set_short_name (dv, i, var_get_short_name (sv, i));
223 var_get_vardict (dv)->case_index = var_get_vardict (sv)->case_index;
226 d->next_value_idx = s->next_value_idx;
228 d->split_cnt = s->split_cnt;
229 if (d->split_cnt > 0)
231 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
232 for (i = 0; i < d->split_cnt; i++)
233 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
236 if (s->weight != NULL)
237 dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight)));
239 if (s->filter != NULL)
240 dict_set_filter (d, dict_lookup_var_assert (d, var_get_name (s->filter)));
242 d->case_limit = s->case_limit;
243 dict_set_label (d, dict_get_label (s));
244 dict_set_documents (d, dict_get_documents (s));
246 d->vector_cnt = s->vector_cnt;
247 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
248 for (i = 0; i < s->vector_cnt; i++)
249 d->vector[i] = vector_clone (s->vector[i], s, d);
251 dict_set_attributes (d, dict_get_attributes (s));
253 for (i = 0; i < s->n_mrsets; i++)
255 const struct mrset *old = s->mrsets[i];
259 /* Clone old mrset, then replace vars from D by vars from S. */
260 new = mrset_clone (old);
261 for (j = 0; j < new->n_vars; j++)
262 new->vars[j] = dict_lookup_var_assert (d, var_get_name (new->vars[j]));
264 dict_add_mrset (d, new);
270 /* Clears the contents from a dictionary without destroying the
271 dictionary itself. */
273 dict_clear (struct dictionary *d)
275 /* FIXME? Should we really clear case_limit, label, documents?
276 Others are necessarily cleared by deleting all the variables.*/
277 while (d->var_cnt > 0 )
279 dict_delete_var (d, d->var[d->var_cnt - 1].var);
284 d->var_cnt = d->var_cap = 0;
285 invalidate_proto (d);
286 hmap_clear (&d->name_map);
287 d->next_value_idx = 0;
288 dict_set_split_vars (d, NULL, 0);
289 dict_set_weight (d, NULL);
290 dict_set_filter (d, NULL);
294 string_array_clear (&d->documents);
295 dict_clear_vectors (d);
296 attrset_clear (&d->attributes);
299 /* Clears a dictionary and destroys it. */
301 _dict_destroy (struct dictionary *d)
303 /* In general, we don't want callbacks occurring, if the dictionary
304 is being destroyed */
305 d->callbacks = NULL ;
308 string_array_destroy (&d->documents);
309 hmap_destroy (&d->name_map);
310 attrset_destroy (&d->attributes);
311 dict_clear_mrsets (d);
317 dict_unref (struct dictionary *d)
322 assert (d->ref_cnt >= 0);
327 /* Returns the number of variables in D. */
329 dict_get_var_cnt (const struct dictionary *d)
334 /* Returns the variable in D with dictionary index IDX, which
335 must be between 0 and the count returned by
336 dict_get_var_cnt(), exclusive. */
338 dict_get_var (const struct dictionary *d, size_t idx)
340 assert (idx < d->var_cnt);
342 return d->var[idx].var;
345 /* Sets *VARS to an array of pointers to variables in D and *CNT
346 to the number of variables in *D. All variables are returned
347 except for those, if any, in the classes indicated by EXCLUDE.
348 (There is no point in putting DC_SYSTEM in EXCLUDE as
349 dictionaries never include system variables.) */
351 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
352 size_t *cnt, enum dict_class exclude)
354 dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude);
357 /* Sets *VARS to an array of pointers to variables in D and *CNT
358 to the number of variables in *D. All variables are returned
359 except for those, if any, in the classes indicated by EXCLUDE.
360 (There is no point in putting DC_SYSTEM in EXCLUDE as
361 dictionaries never include system variables.) */
363 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
364 size_t *cnt, enum dict_class exclude)
369 assert (exclude == (exclude & DC_ALL));
372 for (i = 0; i < d->var_cnt; i++)
374 enum dict_class class = var_get_dict_class (d->var[i].var);
375 if (!(class & exclude))
379 *vars = xnmalloc (count, sizeof **vars);
381 for (i = 0; i < d->var_cnt; i++)
383 enum dict_class class = var_get_dict_class (d->var[i].var);
384 if (!(class & exclude))
385 (*vars)[(*cnt)++] = d->var[i].var;
387 assert (*cnt == count);
390 static struct variable *
391 add_var_with_case_index (struct dictionary *d, struct variable *v,
394 struct vardict_info *vardict;
396 assert (case_index >= d->next_value_idx);
398 /* Update dictionary. */
399 if (d->var_cnt >= d->var_cap)
403 d->var = x2nrealloc (d->var, &d->var_cap, sizeof *d->var);
404 hmap_clear (&d->name_map);
405 for (i = 0; i < d->var_cnt; i++)
407 var_set_vardict (d->var[i].var, &d->var[i]);
408 hmap_insert_fast (&d->name_map, &d->var[i].name_node,
409 d->var[i].name_node.hash);
413 vardict = &d->var[d->var_cnt++];
416 hmap_insert (&d->name_map, &vardict->name_node,
417 utf8_hash_case_string (var_get_name (v), 0));
418 vardict->case_index = case_index;
419 var_set_vardict (v, vardict);
421 if ( d->changed ) d->changed (d, d->changed_data);
422 if ( d->callbacks && d->callbacks->var_added )
423 d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
425 invalidate_proto (d);
426 d->next_value_idx = case_index + 1;
431 static struct variable *
432 add_var (struct dictionary *d, struct variable *v)
434 return add_var_with_case_index (d, v, d->next_value_idx);
437 /* Creates and returns a new variable in D with the given NAME
438 and WIDTH. Returns a null pointer if the given NAME would
439 duplicate that of an existing variable in the dictionary. */
441 dict_create_var (struct dictionary *d, const char *name, int width)
443 return (dict_lookup_var (d, name) == NULL
444 ? dict_create_var_assert (d, name, width)
448 /* Creates and returns a new variable in D with the given NAME
449 and WIDTH. Assert-fails if the given NAME would duplicate
450 that of an existing variable in the dictionary. */
452 dict_create_var_assert (struct dictionary *d, const char *name, int width)
454 assert (dict_lookup_var (d, name) == NULL);
455 return add_var (d, var_create (name, width));
458 /* Creates and returns a new variable in D, as a copy of existing variable
459 OLD_VAR, which need not be in D or in any dictionary. Returns a null
460 pointer if OLD_VAR's name would duplicate that of an existing variable in
463 dict_clone_var (struct dictionary *d, const struct variable *old_var)
465 return dict_clone_var_as (d, old_var, var_get_name (old_var));
468 /* Creates and returns a new variable in D, as a copy of existing variable
469 OLD_VAR, which need not be in D or in any dictionary. Assert-fails if
470 OLD_VAR's name would duplicate that of an existing variable in the
473 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var)
475 return dict_clone_var_as_assert (d, old_var, var_get_name (old_var));
478 /* Creates and returns a new variable in D with name NAME, as a copy of
479 existing variable OLD_VAR, which need not be in D or in any dictionary.
480 Returns a null pointer if the given NAME would duplicate that of an existing
481 variable in the dictionary. */
483 dict_clone_var_as (struct dictionary *d, const struct variable *old_var,
486 return (dict_lookup_var (d, name) == NULL
487 ? dict_clone_var_as_assert (d, old_var, name)
491 /* Creates and returns a new variable in D with name NAME, as a copy of
492 existing variable OLD_VAR, which need not be in D or in any dictionary.
493 Assert-fails if the given NAME would duplicate that of an existing variable
494 in the dictionary. */
496 dict_clone_var_as_assert (struct dictionary *d, const struct variable *old_var,
499 struct variable *new_var = var_clone (old_var);
500 assert (dict_lookup_var (d, name) == NULL);
501 var_set_name (new_var, name);
502 return add_var (d, new_var);
506 dict_clone_var_in_place_assert (struct dictionary *d,
507 const struct variable *old_var)
509 assert (dict_lookup_var (d, var_get_name (old_var)) == NULL);
510 return add_var_with_case_index (d, var_clone (old_var),
511 var_get_case_index (old_var));
514 /* Returns the variable named NAME in D, or a null pointer if no
515 variable has that name. */
517 dict_lookup_var (const struct dictionary *d, const char *name)
519 struct vardict_info *vardict;
521 HMAP_FOR_EACH_WITH_HASH (vardict, struct vardict_info, name_node,
522 utf8_hash_case_string (name, 0), &d->name_map)
524 struct variable *var = vardict->var;
525 if (!utf8_strcasecmp (var_get_name (var), name))
532 /* Returns the variable named NAME in D. Assert-fails if no
533 variable has that name. */
535 dict_lookup_var_assert (const struct dictionary *d, const char *name)
537 struct variable *v = dict_lookup_var (d, name);
542 /* Returns true if variable V is in dictionary D,
545 dict_contains_var (const struct dictionary *d, const struct variable *v)
547 return (var_has_vardict (v)
548 && vardict_get_dictionary (var_get_vardict (v)) == d);
551 /* Compares two double pointers to variables, which should point
552 to elements of a struct dictionary's `var' member array. */
554 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
556 struct variable *const *a = a_;
557 struct variable *const *b = b_;
559 return *a < *b ? -1 : *a > *b;
563 unindex_var (struct dictionary *d, struct vardict_info *vardict)
565 hmap_delete (&d->name_map, &vardict->name_node);
568 /* This function assumes that vardict->name_node.hash is valid, that is, that
569 its name has not changed since it was hashed (rename_var() updates this
570 hash along with the name itself). */
572 reindex_var (struct dictionary *d, struct vardict_info *vardict)
574 struct variable *old = (d->callbacks && d->callbacks->var_changed
575 ? var_clone (vardict->var)
578 struct variable *var = vardict->var;
579 var_set_vardict (var, vardict);
580 hmap_insert_fast (&d->name_map, &vardict->name_node,
581 vardict->name_node.hash);
583 if ( d->changed ) d->changed (d, d->changed_data);
586 d->callbacks->var_changed (d, var_get_dict_index (var), VAR_TRAIT_POSITION, old, d->cb_data);
591 /* Sets the case_index in V's vardict to CASE_INDEX. */
593 set_var_case_index (struct variable *v, int case_index)
595 var_get_vardict (v)->case_index = case_index;
598 /* Removes the dictionary variables with indexes from FROM to TO (exclusive)
601 unindex_vars (struct dictionary *d, size_t from, size_t to)
605 for (i = from; i < to; i++)
606 unindex_var (d, &d->var[i]);
609 /* Re-sets the dict_index in the dictionary variables with
610 indexes from FROM to TO (exclusive). */
612 reindex_vars (struct dictionary *d, size_t from, size_t to)
616 for (i = from; i < to; i++)
617 reindex_var (d, &d->var[i]);
620 /* Deletes variable V from dictionary D and frees V.
622 This is a very bad idea if there might be any pointers to V
623 from outside D. In general, no variable in the active dataset's
624 dictionary should be deleted when any transformations are
625 active on the dictionary's dataset, because those
626 transformations might reference the deleted variable. The
627 safest time to delete a variable is just after a procedure has
628 been executed, as done by DELETE VARIABLES.
630 Pointers to V within D are not a problem, because
631 dict_delete_var() knows to remove V from split variables,
632 weights, filters, etc. */
634 dict_delete_var (struct dictionary *d, struct variable *v)
636 int dict_index = var_get_dict_index (v);
637 const int case_index = var_get_case_index (v);
639 assert (dict_contains_var (d, v));
641 dict_unset_split_var (d, v);
642 dict_unset_mrset_var (d, v);
645 dict_set_weight (d, NULL);
648 dict_set_filter (d, NULL);
650 dict_clear_vectors (d);
652 /* Remove V from var array. */
653 unindex_vars (d, dict_index, d->var_cnt);
654 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
657 /* Update dict_index for each affected variable. */
658 reindex_vars (d, dict_index, d->var_cnt);
661 var_clear_vardict (v);
663 if ( d->changed ) d->changed (d, d->changed_data);
665 invalidate_proto (d);
666 if (d->callbacks && d->callbacks->var_deleted )
667 d->callbacks->var_deleted (d, v, dict_index, case_index, d->cb_data);
672 /* Deletes the COUNT variables listed in VARS from D. This is
673 unsafe; see the comment on dict_delete_var() for details. */
675 dict_delete_vars (struct dictionary *d,
676 struct variable *const *vars, size_t count)
678 /* FIXME: this can be done in O(count) time, but this algorithm
680 assert (count == 0 || vars != NULL);
683 dict_delete_var (d, *vars++);
686 /* Deletes the COUNT variables in D starting at index IDX. This
687 is unsafe; see the comment on dict_delete_var() for
690 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
692 /* FIXME: this can be done in O(count) time, but this algorithm
694 assert (idx + count <= d->var_cnt);
697 dict_delete_var (d, d->var[idx].var);
700 /* Deletes scratch variables from dictionary D. */
702 dict_delete_scratch_vars (struct dictionary *d)
706 /* FIXME: this can be done in O(count) time, but this algorithm
708 for (i = 0; i < d->var_cnt; )
709 if (var_get_dict_class (d->var[i].var) == DC_SCRATCH)
710 dict_delete_var (d, d->var[i].var);
715 /* Moves V to 0-based position IDX in D. Other variables in D,
716 if any, retain their relative positions. Runs in time linear
717 in the distance moved. */
719 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
721 size_t old_index = var_get_dict_index (v);
723 assert (new_index < d->var_cnt);
725 unindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
726 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
727 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
730 /* Reorders the variables in D, placing the COUNT variables
731 listed in ORDER in that order at the beginning of D. The
732 other variables in D, if any, retain their relative
735 dict_reorder_vars (struct dictionary *d,
736 struct variable *const *order, size_t count)
738 struct vardict_info *new_var;
741 assert (count == 0 || order != NULL);
742 assert (count <= d->var_cnt);
744 new_var = xnmalloc (d->var_cap, sizeof *new_var);
746 /* Add variables in ORDER to new_var. */
747 for (i = 0; i < count; i++)
749 struct vardict_info *old_var;
751 assert (dict_contains_var (d, order[i]));
753 old_var = var_get_vardict (order[i]);
754 new_var[i] = *old_var;
755 old_var->dict = NULL;
758 /* Add remaining variables to new_var. */
759 for (i = 0; i < d->var_cnt; i++)
760 if (d->var[i].dict != NULL)
761 new_var[count++] = d->var[i];
762 assert (count == d->var_cnt);
764 /* Replace old vardicts by new ones. */
768 hmap_clear (&d->name_map);
769 reindex_vars (d, 0, d->var_cnt);
772 /* Changes the name of variable V that is currently in a dictionary to
775 rename_var (struct variable *v, const char *new_name)
777 struct vardict_info *vardict = var_get_vardict (v);
778 var_clear_vardict (v);
779 var_set_name (v, new_name);
780 vardict->name_node.hash = utf8_hash_case_string (new_name, 0);
781 var_set_vardict (v, vardict);
784 /* Tries to changes the name of V in D to name NEW_NAME. Returns true if
785 successful, false if a variable (other than V) with the given name already
788 dict_try_rename_var (struct dictionary *d, struct variable *v,
789 const char *new_name)
791 struct variable *conflict = dict_lookup_var (d, new_name);
792 if (conflict && v != conflict)
795 struct variable *old = var_clone (v);
796 unindex_var (d, var_get_vardict (v));
797 rename_var (v, new_name);
798 reindex_var (d, var_get_vardict (v));
800 if (settings_get_algorithm () == ENHANCED)
801 var_clear_short_names (v);
803 if ( d->changed ) d->changed (d, d->changed_data);
804 if ( d->callbacks && d->callbacks->var_changed )
805 d->callbacks->var_changed (d, var_get_dict_index (v), VAR_TRAIT_NAME, old, d->cb_data);
812 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
813 a variable named NEW_NAME is already in D, except that
814 NEW_NAME may be the same as V's existing name. */
816 dict_rename_var (struct dictionary *d, struct variable *v,
817 const char *new_name)
819 bool ok UNUSED = dict_try_rename_var (d, v, new_name);
823 /* Renames COUNT variables specified in VARS to the names given
824 in NEW_NAMES within dictionary D. If the renaming would
825 result in a duplicate variable name, returns false and stores a
826 name that would be duplicated into *ERR_NAME (if ERR_NAME is
827 non-null). Otherwise, the renaming is successful, and true
830 dict_rename_vars (struct dictionary *d,
831 struct variable **vars, char **new_names, size_t count,
838 assert (count == 0 || vars != NULL);
839 assert (count == 0 || new_names != NULL);
841 /* Save the names of the variables to be renamed. */
842 pool = pool_create ();
843 old_names = pool_nalloc (pool, count, sizeof *old_names);
844 for (i = 0; i < count; i++)
845 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
847 /* Remove the variables to be renamed from the name hash,
849 for (i = 0; i < count; i++)
851 unindex_var (d, var_get_vardict (vars[i]));
852 rename_var (vars[i], new_names[i]);
855 /* Add the renamed variables back into the name hash,
856 checking for conflicts. */
857 for (i = 0; i < count; i++)
859 if (dict_lookup_var (d, var_get_name (vars[i])) != NULL)
861 /* There is a name conflict.
862 Back out all the name changes that have already
863 taken place, and indicate failure. */
865 if (err_name != NULL)
866 *err_name = new_names[i];
868 for (i = 0; i < fail_idx; i++)
869 unindex_var (d, var_get_vardict (vars[i]));
871 for (i = 0; i < count; i++)
873 rename_var (vars[i], old_names[i]);
874 reindex_var (d, var_get_vardict (vars[i]));
880 reindex_var (d, var_get_vardict (vars[i]));
883 /* Clear short names. */
884 if (settings_get_algorithm () == ENHANCED)
885 for (i = 0; i < count; i++)
886 var_clear_short_names (vars[i]);
892 /* Returns true if a variable named NAME may be inserted in DICT;
893 that is, if there is not already a variable with that name in
894 DICT and if NAME is not a reserved word. (The caller's checks
895 have already verified that NAME is otherwise acceptable as a
898 var_name_is_insertable (const struct dictionary *dict, const char *name)
900 return (dict_lookup_var (dict, name) == NULL
901 && lex_id_to_token (ss_cstr (name)) == T_ID);
905 make_hinted_name (const struct dictionary *dict, const char *hint)
907 size_t hint_len = strlen (hint);
908 bool dropped = false;
913 /* The allocation size here is OK: characters that are copied directly fit
914 OK, and characters that are not copied directly are replaced by a single
915 '_' byte. If u8_mbtouc() replaces bad input by 0xfffd, then that will get
916 replaced by '_' too. */
917 root = rp = xmalloc (hint_len + 1);
918 for (ofs = 0; ofs < hint_len; ofs += mblen)
922 mblen = u8_mbtouc (&uc, CHAR_CAST (const uint8_t *, hint + ofs),
925 ? lex_uc_is_id1 (uc) && uc != '$'
926 : lex_uc_is_idn (uc))
933 rp += u8_uctomb (CHAR_CAST (uint8_t *, rp), uc, 6);
944 if (var_name_is_insertable (dict, root))
947 for (i = 0; i < ULONG_MAX; i++)
949 char suffix[INT_BUFSIZE_BOUND (i) + 1];
953 if (!str_format_26adic (i + 1, true, &suffix[1], sizeof suffix - 1))
956 name = utf8_encoding_concat (root, suffix, dict->encoding, 64);
957 if (var_name_is_insertable (dict, name))
972 make_numeric_name (const struct dictionary *dict, unsigned long int *num_start)
974 unsigned long int number;
976 for (number = num_start != NULL ? MAX (*num_start, 1) : 1;
980 char name[3 + INT_STRLEN_BOUND (number) + 1];
982 sprintf (name, "VAR%03lu", number);
983 if (dict_lookup_var (dict, name) == NULL)
985 if (num_start != NULL)
986 *num_start = number + 1;
987 return xstrdup (name);
995 /* Devises and returns a variable name unique within DICT. The variable name
996 is owned by the caller, which must free it with free() when it is no longer
999 HINT, if it is non-null, is used as a suggestion that will be
1000 modified for suitability as a variable name and for
1003 If HINT is null or entirely unsuitable, a name in the form
1004 "VAR%03d" will be generated, where the smallest unused integer
1005 value is used. If NUM_START is non-null, then its value is
1006 used as the minimum numeric value to check, and it is updated
1007 to the next value to be checked.
1010 dict_make_unique_var_name (const struct dictionary *dict, const char *hint,
1011 unsigned long int *num_start)
1015 char *hinted_name = make_hinted_name (dict, hint);
1016 if (hinted_name != NULL)
1019 return make_numeric_name (dict, num_start);
1022 /* Returns whether variable names must be valid identifiers. Normally, this is
1023 true, but sometimes a dictionary is prepared for external use (e.g. output
1024 to a CSV file) where names don't have to be valid. */
1026 dict_get_names_must_be_ids (const struct dictionary *d)
1028 return d->names_must_be_ids;
1031 /* Sets whether variable names must be valid identifiers. Normally, this is
1032 true, but sometimes a dictionary is prepared for external use (e.g. output
1033 to a CSV file) where names don't have to be valid.
1035 Changing this setting from false to true doesn't make the dictionary check
1036 all the existing variable names, so it can cause an invariant violation. */
1038 dict_set_names_must_be_ids (struct dictionary *d, bool names_must_be_ids)
1040 d->names_must_be_ids = names_must_be_ids;
1043 /* Returns the weighting variable in dictionary D, or a null
1044 pointer if the dictionary is unweighted. */
1046 dict_get_weight (const struct dictionary *d)
1048 assert (d->weight == NULL || dict_contains_var (d, d->weight));
1053 /* Returns the value of D's weighting variable in case C, except
1054 that a negative weight is returned as 0. Returns 1 if the
1055 dictionary is unweighted. Will warn about missing, negative,
1056 or zero values if *WARN_ON_INVALID is true. The function will
1057 set *WARN_ON_INVALID to false if an invalid weight is
1060 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
1061 bool *warn_on_invalid)
1065 if (d->weight == NULL)
1069 double w = case_num (c, d->weight);
1071 return var_force_valid_weight (d->weight, w, warn_on_invalid);
1075 /* Returns the format to use for weights. */
1076 const struct fmt_spec *
1077 dict_get_weight_format (const struct dictionary *d)
1079 return d->weight ? var_get_print_format (d->weight) : &F_8_0;
1082 /* Sets the weighting variable of D to V, or turning off
1083 weighting if V is a null pointer. */
1085 dict_set_weight (struct dictionary *d, struct variable *v)
1087 assert (v == NULL || dict_contains_var (d, v));
1088 assert (v == NULL || var_is_numeric (v));
1092 if (d->changed) d->changed (d, d->changed_data);
1093 if ( d->callbacks && d->callbacks->weight_changed )
1094 d->callbacks->weight_changed (d,
1095 v ? var_get_dict_index (v) : -1,
1099 /* Returns the filter variable in dictionary D (see cmd_filter())
1100 or a null pointer if the dictionary is unfiltered. */
1102 dict_get_filter (const struct dictionary *d)
1104 assert (d->filter == NULL || dict_contains_var (d, d->filter));
1109 /* Sets V as the filter variable for dictionary D. Passing a
1110 null pointer for V turn off filtering. */
1112 dict_set_filter (struct dictionary *d, struct variable *v)
1114 assert (v == NULL || dict_contains_var (d, v));
1115 assert (v == NULL || var_is_numeric (v));
1119 if (d->changed) d->changed (d, d->changed_data);
1120 if ( d->callbacks && d->callbacks->filter_changed )
1121 d->callbacks->filter_changed (d,
1122 v ? var_get_dict_index (v) : -1,
1126 /* Returns the case limit for dictionary D, or zero if the number
1127 of cases is unlimited. */
1129 dict_get_case_limit (const struct dictionary *d)
1131 return d->case_limit;
1134 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
1135 0 for CASE_LIMIT to indicate no limit. */
1137 dict_set_case_limit (struct dictionary *d, casenumber case_limit)
1139 d->case_limit = case_limit;
1142 /* Returns the prototype used for cases created by dictionary D. */
1143 const struct caseproto *
1144 dict_get_proto (const struct dictionary *d_)
1146 struct dictionary *d = CONST_CAST (struct dictionary *, d_);
1147 if (d->proto == NULL)
1151 d->proto = caseproto_create ();
1152 d->proto = caseproto_reserve (d->proto, d->var_cnt);
1153 for (i = 0; i < d->var_cnt; i++)
1154 d->proto = caseproto_set_width (d->proto,
1155 var_get_case_index (d->var[i].var),
1156 var_get_width (d->var[i].var));
1161 /* Returns the case index of the next value to be added to D.
1162 This value is the number of `union value's that need to be
1163 allocated to store a case for dictionary D. */
1165 dict_get_next_value_idx (const struct dictionary *d)
1167 return d->next_value_idx;
1170 /* Returns the number of bytes needed to store a case for
1173 dict_get_case_size (const struct dictionary *d)
1175 return sizeof (union value) * dict_get_next_value_idx (d);
1178 /* Reassigns values in dictionary D so that fragmentation is
1181 dict_compact_values (struct dictionary *d)
1185 d->next_value_idx = 0;
1186 for (i = 0; i < d->var_cnt; i++)
1188 struct variable *v = d->var[i].var;
1189 set_var_case_index (v, d->next_value_idx++);
1191 invalidate_proto (d);
1194 /* Returns the number of values occupied by the variables in
1195 dictionary D. All variables are considered if EXCLUDE_CLASSES
1196 is 0, or it may contain one or more of (1u << DC_ORDINARY),
1197 (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to exclude the
1198 corresponding type of variable.
1200 The return value may be less than the number of values in one
1201 of dictionary D's cases (as returned by
1202 dict_get_next_value_idx) even if E is 0, because there may be
1203 gaps in D's cases due to deleted variables. */
1205 dict_count_values (const struct dictionary *d, unsigned int exclude_classes)
1210 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1212 | (1u << DC_SCRATCH))) == 0);
1215 for (i = 0; i < d->var_cnt; i++)
1217 enum dict_class class = var_get_dict_class (d->var[i].var);
1218 if (!(exclude_classes & (1u << class)))
1224 /* Returns the case prototype that would result after deleting
1225 all variables from D that are not in one of the
1226 EXCLUDE_CLASSES and compacting the dictionary with
1229 The caller must unref the returned caseproto when it is no
1232 dict_get_compacted_proto (const struct dictionary *d,
1233 unsigned int exclude_classes)
1235 struct caseproto *proto;
1238 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1240 | (1u << DC_SCRATCH))) == 0);
1242 proto = caseproto_create ();
1243 for (i = 0; i < d->var_cnt; i++)
1245 struct variable *v = d->var[i].var;
1246 if (!(exclude_classes & (1u << var_get_dict_class (v))))
1247 proto = caseproto_add_width (proto, var_get_width (v));
1252 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
1253 dict_get_split_cnt() to determine how many SPLIT FILE vars
1254 there are. Returns a null pointer if and only if there are no
1256 const struct variable *const *
1257 dict_get_split_vars (const struct dictionary *d)
1262 /* Returns the number of SPLIT FILE vars. */
1264 dict_get_split_cnt (const struct dictionary *d)
1266 return d->split_cnt;
1269 /* Removes variable V, which must be in D, from D's set of split
1272 dict_unset_split_var (struct dictionary *d, struct variable *v)
1276 assert (dict_contains_var (d, v));
1278 orig_count = d->split_cnt;
1279 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
1280 &v, compare_var_ptrs, NULL);
1281 if (orig_count != d->split_cnt)
1283 if (d->changed) d->changed (d, d->changed_data);
1284 /* We changed the set of split variables so invoke the
1286 if (d->callbacks && d->callbacks->split_changed)
1287 d->callbacks->split_changed (d, d->cb_data);
1291 /* Sets CNT split vars SPLIT in dictionary D. */
1293 dict_set_split_vars (struct dictionary *d,
1294 struct variable *const *split, size_t cnt)
1296 assert (cnt == 0 || split != NULL);
1301 d->split = xnrealloc (d->split, cnt, sizeof *d->split) ;
1302 memcpy (d->split, split, cnt * sizeof *d->split);
1310 if (d->changed) d->changed (d, d->changed_data);
1311 if ( d->callbacks && d->callbacks->split_changed )
1312 d->callbacks->split_changed (d, d->cb_data);
1315 /* Returns the file label for D, or a null pointer if D is
1316 unlabeled (see cmd_file_label()). */
1318 dict_get_label (const struct dictionary *d)
1323 /* Sets D's file label to LABEL, truncating it to at most 60 bytes in D's
1326 Removes D's label if LABEL is null or the empty string. */
1328 dict_set_label (struct dictionary *d, const char *label)
1331 if (label == NULL || label[0] == '\0')
1334 d->label = utf8_encoding_trunc (label, d->encoding, 60);
1337 /* Returns the documents for D, as an UTF-8 encoded string_array. The
1338 return value is always nonnull; if there are no documents then the
1339 string_arary is empty.*/
1340 const struct string_array *
1341 dict_get_documents (const struct dictionary *d)
1343 return &d->documents;
1346 /* Replaces the documents for D by NEW_DOCS, a UTF-8 encoded string_array. */
1348 dict_set_documents (struct dictionary *d, const struct string_array *new_docs)
1352 dict_clear_documents (d);
1354 for (i = 0; i < new_docs->n; i++)
1355 dict_add_document_line (d, new_docs->strings[i], false);
1358 /* Replaces the documents for D by UTF-8 encoded string NEW_DOCS, dividing it
1359 into individual lines at new-line characters. Each line is truncated to at
1360 most DOC_LINE_LENGTH bytes in D's encoding. */
1362 dict_set_documents_string (struct dictionary *d, const char *new_docs)
1366 dict_clear_documents (d);
1367 for (s = new_docs; *s != '\0'; )
1369 size_t len = strcspn (s, "\n");
1370 char *line = xmemdup0 (s, len);
1371 dict_add_document_line (d, line, false);
1380 /* Drops the documents from dictionary D. */
1382 dict_clear_documents (struct dictionary *d)
1384 string_array_clear (&d->documents);
1387 /* Appends the UTF-8 encoded LINE to the documents in D. LINE will be
1388 truncated so that it is no more than 80 bytes in the dictionary's
1389 encoding. If this causes some text to be lost, and ISSUE_WARNING is true,
1390 then a warning will be issued. */
1392 dict_add_document_line (struct dictionary *d, const char *line,
1398 trunc_len = utf8_encoding_trunc_len (line, d->encoding, DOC_LINE_LENGTH);
1399 truncated = line[trunc_len] != '\0';
1400 if (truncated && issue_warning)
1402 /* Note to translators: "bytes" is correct, not characters */
1403 msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
1406 string_array_append_nocopy (&d->documents, xmemdup0 (line, trunc_len));
1411 /* Returns the number of document lines in dictionary D. */
1413 dict_get_document_line_cnt (const struct dictionary *d)
1415 return d->documents.n;
1418 /* Returns document line number IDX in dictionary D. The caller must not
1419 modify or free the returned string. */
1421 dict_get_document_line (const struct dictionary *d, size_t idx)
1423 assert (idx < d->documents.n);
1424 return d->documents.strings[idx];
1427 /* Creates in D a vector named NAME that contains the CNT
1428 variables in VAR. Returns true if successful, or false if a
1429 vector named NAME already exists in D. */
1431 dict_create_vector (struct dictionary *d,
1433 struct variable **var, size_t cnt)
1438 for (i = 0; i < cnt; i++)
1439 assert (dict_contains_var (d, var[i]));
1441 if (dict_lookup_vector (d, name) == NULL)
1443 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1444 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1451 /* Creates in D a vector named NAME that contains the CNT
1452 variables in VAR. A vector named NAME must not already exist
1455 dict_create_vector_assert (struct dictionary *d,
1457 struct variable **var, size_t cnt)
1459 assert (dict_lookup_vector (d, name) == NULL);
1460 dict_create_vector (d, name, var, cnt);
1463 /* Returns the vector in D with index IDX, which must be less
1464 than dict_get_vector_cnt (D). */
1465 const struct vector *
1466 dict_get_vector (const struct dictionary *d, size_t idx)
1468 assert (idx < d->vector_cnt);
1470 return d->vector[idx];
1473 /* Returns the number of vectors in D. */
1475 dict_get_vector_cnt (const struct dictionary *d)
1477 return d->vector_cnt;
1480 /* Looks up and returns the vector within D with the given
1482 const struct vector *
1483 dict_lookup_vector (const struct dictionary *d, const char *name)
1486 for (i = 0; i < d->vector_cnt; i++)
1487 if (!utf8_strcasecmp (vector_get_name (d->vector[i]), name))
1488 return d->vector[i];
1492 /* Deletes all vectors from D. */
1494 dict_clear_vectors (struct dictionary *d)
1498 for (i = 0; i < d->vector_cnt; i++)
1499 vector_destroy (d->vector[i]);
1506 /* Multiple response sets. */
1508 /* Returns the multiple response set in DICT with index IDX, which must be
1509 between 0 and the count returned by dict_get_n_mrsets(), exclusive. */
1510 const struct mrset *
1511 dict_get_mrset (const struct dictionary *dict, size_t idx)
1513 assert (idx < dict->n_mrsets);
1514 return dict->mrsets[idx];
1517 /* Returns the number of multiple response sets in DICT. */
1519 dict_get_n_mrsets (const struct dictionary *dict)
1521 return dict->n_mrsets;
1524 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1525 returns its index; otherwise, returns SIZE_MAX. */
1527 dict_lookup_mrset_idx (const struct dictionary *dict, const char *name)
1531 for (i = 0; i < dict->n_mrsets; i++)
1532 if (!utf8_strcasecmp (name, dict->mrsets[i]->name))
1538 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1539 returns it; otherwise, returns NULL. */
1540 const struct mrset *
1541 dict_lookup_mrset (const struct dictionary *dict, const char *name)
1543 size_t idx = dict_lookup_mrset_idx (dict, name);
1544 return idx != SIZE_MAX ? dict->mrsets[idx] : NULL;
1547 /* Adds MRSET to DICT, replacing any existing set with the same name. Returns
1548 true if a set was replaced, false if none existed with the specified name.
1550 Ownership of MRSET is transferred to DICT. */
1552 dict_add_mrset (struct dictionary *dict, struct mrset *mrset)
1556 assert (mrset_ok (mrset, dict));
1558 idx = dict_lookup_mrset_idx (dict, mrset->name);
1559 if (idx == SIZE_MAX)
1561 dict->mrsets = xrealloc (dict->mrsets,
1562 (dict->n_mrsets + 1) * sizeof *dict->mrsets);
1563 dict->mrsets[dict->n_mrsets++] = mrset;
1568 mrset_destroy (dict->mrsets[idx]);
1569 dict->mrsets[idx] = mrset;
1574 /* Looks for a multiple response set in DICT named NAME. If found, removes it
1575 from DICT and returns true. If none is found, returns false without
1578 Deleting one multiple response set causes the indexes of other sets within
1581 dict_delete_mrset (struct dictionary *dict, const char *name)
1583 size_t idx = dict_lookup_mrset_idx (dict, name);
1584 if (idx != SIZE_MAX)
1586 mrset_destroy (dict->mrsets[idx]);
1587 dict->mrsets[idx] = dict->mrsets[--dict->n_mrsets];
1594 /* Deletes all multiple response sets from DICT. */
1596 dict_clear_mrsets (struct dictionary *dict)
1600 for (i = 0; i < dict->n_mrsets; i++)
1601 mrset_destroy (dict->mrsets[i]);
1602 free (dict->mrsets);
1603 dict->mrsets = NULL;
1607 /* Removes VAR, which must be in DICT, from DICT's multiple response sets. */
1609 dict_unset_mrset_var (struct dictionary *dict, struct variable *var)
1613 assert (dict_contains_var (dict, var));
1615 for (i = 0; i < dict->n_mrsets; )
1617 struct mrset *mrset = dict->mrsets[i];
1620 for (j = 0; j < mrset->n_vars; )
1621 if (mrset->vars[j] == var)
1622 remove_element (mrset->vars, mrset->n_vars--,
1623 sizeof *mrset->vars, j);
1627 if (mrset->n_vars < 2)
1629 mrset_destroy (mrset);
1630 dict->mrsets[i] = dict->mrsets[--dict->n_mrsets];
1637 /* Returns D's attribute set. The caller may examine or modify
1638 the attribute set, but must not destroy it. Destroying D or
1639 calling dict_set_attributes for D will also destroy D's
1642 dict_get_attributes (const struct dictionary *d)
1644 return CONST_CAST (struct attrset *, &d->attributes);
1647 /* Replaces D's attributes set by a copy of ATTRS. */
1649 dict_set_attributes (struct dictionary *d, const struct attrset *attrs)
1651 attrset_destroy (&d->attributes);
1652 attrset_clone (&d->attributes, attrs);
1655 /* Returns true if D has at least one attribute in its attribute
1656 set, false if D's attribute set is empty. */
1658 dict_has_attributes (const struct dictionary *d)
1660 return attrset_count (&d->attributes) > 0;
1663 /* Called from variable.c to notify the dictionary that some property (indicated
1664 by WHAT) of the variable has changed. OLDVAR is a copy of V as it existed
1665 prior to the change. OLDVAR is destroyed by this function.
1668 dict_var_changed (const struct variable *v, unsigned int what, struct variable *oldvar)
1670 if ( var_has_vardict (v))
1672 const struct vardict_info *vardict = var_get_vardict (v);
1673 struct dictionary *d = vardict->dict;
1678 if (d->changed ) d->changed (d, d->changed_data);
1679 if ( d->callbacks && d->callbacks->var_changed )
1680 d->callbacks->var_changed (d, var_get_dict_index (v), what, oldvar, d->cb_data);
1682 var_destroy (oldvar);
1687 /* Dictionary used to contain "internal variables". */
1688 static struct dictionary *internal_dict;
1690 /* Create a variable of the specified WIDTH to be used for internal
1691 calculations only. The variable is assigned case index CASE_IDX. */
1693 dict_create_internal_var (int case_idx, int width)
1695 if (internal_dict == NULL)
1696 internal_dict = dict_create ("UTF-8");
1700 static int counter = INT_MAX / 2;
1701 struct variable *var;
1704 if (++counter == INT_MAX)
1705 counter = INT_MAX / 2;
1707 sprintf (name, "$internal%d", counter);
1708 var = dict_create_var (internal_dict, name, width);
1711 set_var_case_index (var, case_idx);
1717 /* Destroys VAR, which must have been created with
1718 dict_create_internal_var(). */
1720 dict_destroy_internal_var (struct variable *var)
1724 dict_delete_var (internal_dict, var);
1726 /* Destroy internal_dict if it has no variables left, just so that
1727 valgrind --leak-check --show-reachable won't show internal_dict. */
1728 if (dict_get_var_cnt (internal_dict) == 0)
1730 dict_unref (internal_dict);
1731 internal_dict = NULL;
1737 vardict_get_dict_index (const struct vardict_info *vardict)
1739 return vardict - vardict->dict->var;