1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006, 2007, 2009, 2010, 2011, 2012 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)
58 struct vardict_info *var; /* Variables. */
59 size_t var_cnt, var_cap; /* Number of variables, capacity. */
60 struct caseproto *proto; /* Prototype for dictionary cases
62 struct hmap name_map; /* Variable index by name. */
63 int next_value_idx; /* Index of next `union value' to allocate. */
64 const struct variable **split; /* SPLIT FILE vars. */
65 size_t split_cnt; /* SPLIT FILE count. */
66 struct variable *weight; /* WEIGHT variable. */
67 struct variable *filter; /* FILTER variable. */
68 casenumber case_limit; /* Current case limit (N command). */
69 char *label; /* File label. */
70 struct string_array documents; /* Documents. */
71 struct vector **vector; /* Vectors of variables. */
72 size_t vector_cnt; /* Number of vectors. */
73 struct attrset attributes; /* Custom attributes. */
74 struct mrset **mrsets; /* Multiple response sets. */
75 size_t n_mrsets; /* Number of multiple response sets. */
77 char *encoding; /* Character encoding of string data */
79 const struct dict_callbacks *callbacks; /* Callbacks on dictionary
81 void *cb_data ; /* Data passed to callbacks */
83 void (*changed) (struct dictionary *, void *); /* Generic change callback */
87 static void dict_unset_split_var (struct dictionary *, struct variable *);
88 static void dict_unset_mrset_var (struct dictionary *, struct variable *);
90 /* Returns the encoding for data in dictionary D. The return value is a
91 nonnull string that contains an IANA character set name. */
93 dict_get_encoding (const struct dictionary *d)
98 /* Returns true if UTF-8 string ID is an acceptable identifier in DICT's
99 encoding, false otherwise. If ISSUE_ERROR is true, issues an explanatory
100 error message on failure. */
102 dict_id_is_valid (const struct dictionary *dict, const char *id,
105 return id_is_valid (id, dict->encoding, issue_error);
109 dict_set_change_callback (struct dictionary *d,
110 void (*changed) (struct dictionary *, void*),
113 d->changed = changed;
114 d->changed_data = data;
117 /* Discards dictionary D's caseproto. (It will be regenerated
118 lazily, on demand.) */
120 invalidate_proto (struct dictionary *d)
122 caseproto_unref (d->proto);
126 /* Print a representation of dictionary D to stdout, for
127 debugging purposes. */
129 dict_dump (const struct dictionary *d)
132 for (i = 0 ; i < d->var_cnt ; ++i )
134 const struct variable *v = d->var[i].var;
135 printf ("Name: %s;\tdict_idx: %zu; case_idx: %zu\n",
137 var_get_dict_index (v),
138 var_get_case_index (v));
143 /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
144 the dictionary or any of the variables it contains are modified.
145 Each callback will get passed CALLBACK_DATA.
146 Any callback may be NULL, in which case it'll be ignored.
149 dict_set_callbacks (struct dictionary *dict,
150 const struct dict_callbacks *callbacks,
153 dict->callbacks = callbacks;
154 dict->cb_data = callback_data;
157 /* Shallow copy the callbacks from SRC to DEST */
159 dict_copy_callbacks (struct dictionary *dest,
160 const struct dictionary *src)
162 dest->callbacks = src->callbacks;
163 dest->cb_data = src->cb_data;
166 /* Creates and returns a new dictionary with the specified ENCODING. */
168 dict_create (const char *encoding)
170 struct dictionary *d = xzalloc (sizeof *d);
172 d->encoding = xstrdup (encoding);
173 hmap_init (&d->name_map);
174 attrset_init (&d->attributes);
179 /* Creates and returns a (deep) copy of an existing
182 The new dictionary's case indexes are copied from the old
183 dictionary. If the new dictionary won't be used to access
184 cases produced with the old dictionary, then the new
185 dictionary's case indexes should be compacted with
186 dict_compact_values to save space.
188 Callbacks are not cloned. */
190 dict_clone (const struct dictionary *s)
192 struct dictionary *d;
195 d = dict_create (s->encoding);
197 for (i = 0; i < s->var_cnt; i++)
199 struct variable *sv = s->var[i].var;
200 struct variable *dv = dict_clone_var_assert (d, sv);
203 for (i = 0; i < var_get_short_name_cnt (sv); i++)
204 var_set_short_name (dv, i, var_get_short_name (sv, i));
206 var_get_vardict (dv)->case_index = var_get_vardict (sv)->case_index;
209 d->next_value_idx = s->next_value_idx;
211 d->split_cnt = s->split_cnt;
212 if (d->split_cnt > 0)
214 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
215 for (i = 0; i < d->split_cnt; i++)
216 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
219 if (s->weight != NULL)
220 dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight)));
222 if (s->filter != NULL)
223 dict_set_filter (d, dict_lookup_var_assert (d, var_get_name (s->filter)));
225 d->case_limit = s->case_limit;
226 dict_set_label (d, dict_get_label (s));
227 dict_set_documents (d, dict_get_documents (s));
229 d->vector_cnt = s->vector_cnt;
230 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
231 for (i = 0; i < s->vector_cnt; i++)
232 d->vector[i] = vector_clone (s->vector[i], s, d);
234 dict_set_attributes (d, dict_get_attributes (s));
236 for (i = 0; i < s->n_mrsets; i++)
238 const struct mrset *old = s->mrsets[i];
242 /* Clone old mrset, then replace vars from D by vars from S. */
243 new = mrset_clone (old);
244 for (j = 0; j < new->n_vars; j++)
245 new->vars[j] = dict_lookup_var_assert (d, var_get_name (new->vars[j]));
247 dict_add_mrset (d, new);
253 /* Clears the contents from a dictionary without destroying the
254 dictionary itself. */
256 dict_clear (struct dictionary *d)
258 /* FIXME? Should we really clear case_limit, label, documents?
259 Others are necessarily cleared by deleting all the variables.*/
260 while (d->var_cnt > 0 )
262 dict_delete_var (d, d->var[d->var_cnt - 1].var);
267 d->var_cnt = d->var_cap = 0;
268 invalidate_proto (d);
269 hmap_clear (&d->name_map);
270 d->next_value_idx = 0;
271 dict_set_split_vars (d, NULL, 0);
272 dict_set_weight (d, NULL);
273 dict_set_filter (d, NULL);
277 string_array_clear (&d->documents);
278 dict_clear_vectors (d);
279 attrset_clear (&d->attributes);
282 /* Destroys the aux data for every variable in D, by calling
283 var_clear_aux() for each variable. */
285 dict_clear_aux (struct dictionary *d)
289 for (i = 0; i < d->var_cnt; i++)
290 var_clear_aux (d->var[i].var);
293 /* Clears a dictionary and destroys it. */
295 dict_destroy (struct dictionary *d)
299 /* In general, we don't want callbacks occuring, if the dictionary
300 is being destroyed */
301 d->callbacks = NULL ;
304 hmap_destroy (&d->name_map);
305 attrset_destroy (&d->attributes);
306 dict_clear_mrsets (d);
312 /* Returns the number of variables in D. */
314 dict_get_var_cnt (const struct dictionary *d)
319 /* Returns the variable in D with dictionary index IDX, which
320 must be between 0 and the count returned by
321 dict_get_var_cnt(), exclusive. */
323 dict_get_var (const struct dictionary *d, size_t idx)
325 assert (idx < d->var_cnt);
327 return d->var[idx].var;
330 /* Sets *VARS to an array of pointers to variables in D and *CNT
331 to the number of variables in *D. All variables are returned
332 except for those, if any, in the classes indicated by EXCLUDE.
333 (There is no point in putting DC_SYSTEM in EXCLUDE as
334 dictionaries never include system variables.) */
336 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
337 size_t *cnt, enum dict_class exclude)
339 dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude);
342 /* Sets *VARS to an array of pointers to variables in D and *CNT
343 to the number of variables in *D. All variables are returned
344 except for those, if any, in the classes indicated by EXCLUDE.
345 (There is no point in putting DC_SYSTEM in EXCLUDE as
346 dictionaries never include system variables.) */
348 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
349 size_t *cnt, enum dict_class exclude)
354 assert (exclude == (exclude & DC_ALL));
357 for (i = 0; i < d->var_cnt; i++)
359 enum dict_class class = var_get_dict_class (d->var[i].var);
360 if (!(class & exclude))
364 *vars = xnmalloc (count, sizeof **vars);
366 for (i = 0; i < d->var_cnt; i++)
368 enum dict_class class = var_get_dict_class (d->var[i].var);
369 if (!(class & exclude))
370 (*vars)[(*cnt)++] = d->var[i].var;
372 assert (*cnt == count);
375 static struct variable *
376 add_var (struct dictionary *d, struct variable *v)
378 struct vardict_info *vardict;
380 /* Update dictionary. */
381 if (d->var_cnt >= d->var_cap)
385 d->var = x2nrealloc (d->var, &d->var_cap, sizeof *d->var);
386 hmap_clear (&d->name_map);
387 for (i = 0; i < d->var_cnt; i++)
389 var_set_vardict (d->var[i].var, &d->var[i]);
390 hmap_insert_fast (&d->name_map, &d->var[i].name_node,
391 d->var[i].name_node.hash);
395 vardict = &d->var[d->var_cnt++];
398 hmap_insert (&d->name_map, &vardict->name_node,
399 hash_case_string (var_get_name (v), 0));
400 vardict->case_index = d->next_value_idx;
401 var_set_vardict (v, vardict);
403 if ( d->changed ) d->changed (d, d->changed_data);
404 if ( d->callbacks && d->callbacks->var_added )
405 d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
408 invalidate_proto (d);
413 /* Creates and returns a new variable in D with the given NAME
414 and WIDTH. Returns a null pointer if the given NAME would
415 duplicate that of an existing variable in the dictionary. */
417 dict_create_var (struct dictionary *d, const char *name, int width)
419 return (dict_lookup_var (d, name) == NULL
420 ? dict_create_var_assert (d, name, width)
424 /* Creates and returns a new variable in D with the given NAME
425 and WIDTH. Assert-fails if the given NAME would duplicate
426 that of an existing variable in the dictionary. */
428 dict_create_var_assert (struct dictionary *d, const char *name, int width)
430 assert (dict_lookup_var (d, name) == NULL);
431 return add_var (d, var_create (name, width));
434 /* Creates and returns a new variable in D, as a copy of existing variable
435 OLD_VAR, which need not be in D or in any dictionary. Returns a null
436 pointer if OLD_VAR's name would duplicate that of an existing variable in
439 dict_clone_var (struct dictionary *d, const struct variable *old_var)
441 return dict_clone_var_as (d, old_var, var_get_name (old_var));
444 /* Creates and returns a new variable in D, as a copy of existing variable
445 OLD_VAR, which need not be in D or in any dictionary. Assert-fails if
446 OLD_VAR's name would duplicate that of an existing variable in the
449 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var)
451 return dict_clone_var_as_assert (d, old_var, var_get_name (old_var));
454 /* Creates and returns a new variable in D with name NAME, as a copy of
455 existing variable OLD_VAR, which need not be in D or in any dictionary.
456 Returns a null pointer if the given NAME would duplicate that of an existing
457 variable in the dictionary. */
459 dict_clone_var_as (struct dictionary *d, const struct variable *old_var,
462 return (dict_lookup_var (d, name) == NULL
463 ? dict_clone_var_as_assert (d, old_var, name)
467 /* Creates and returns a new variable in D with name NAME, as a copy of
468 existing variable OLD_VAR, which need not be in D or in any dictionary.
469 Assert-fails if the given NAME would duplicate that of an existing variable
470 in the dictionary. */
472 dict_clone_var_as_assert (struct dictionary *d, const struct variable *old_var,
475 struct variable *new_var = var_clone (old_var);
476 assert (dict_lookup_var (d, name) == NULL);
477 var_set_name (new_var, name);
478 return add_var (d, new_var);
481 /* Returns the variable named NAME in D, or a null pointer if no
482 variable has that name. */
484 dict_lookup_var (const struct dictionary *d, const char *name)
486 struct vardict_info *vardict;
488 HMAP_FOR_EACH_WITH_HASH (vardict, struct vardict_info, name_node,
489 hash_case_string (name, 0), &d->name_map)
491 struct variable *var = vardict->var;
492 if (!strcasecmp (var_get_name (var), name))
499 /* Returns the variable named NAME in D. Assert-fails if no
500 variable has that name. */
502 dict_lookup_var_assert (const struct dictionary *d, const char *name)
504 struct variable *v = dict_lookup_var (d, name);
509 /* Returns true if variable V is in dictionary D,
512 dict_contains_var (const struct dictionary *d, const struct variable *v)
514 return (var_has_vardict (v)
515 && vardict_get_dictionary (var_get_vardict (v)) == d);
518 /* Compares two double pointers to variables, which should point
519 to elements of a struct dictionary's `var' member array. */
521 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
523 struct variable *const *a = a_;
524 struct variable *const *b = b_;
526 return *a < *b ? -1 : *a > *b;
530 unindex_var (struct dictionary *d, struct vardict_info *vardict)
532 hmap_delete (&d->name_map, &vardict->name_node);
535 /* This function assumes that vardict->name_node.hash is valid, that is, that
536 its name has not changed since it was hashed (rename_var() updates this
537 hash along with the name itself). */
539 reindex_var (struct dictionary *d, struct vardict_info *vardict)
541 struct variable *var = vardict->var;
543 var_set_vardict (var, vardict);
544 hmap_insert_fast (&d->name_map, &vardict->name_node,
545 vardict->name_node.hash);
547 if ( d->changed ) d->changed (d, d->changed_data);
548 if ( d->callbacks && d->callbacks->var_changed )
549 d->callbacks->var_changed (d, var_get_dict_index (var), d->cb_data);
552 /* Sets the case_index in V's vardict to CASE_INDEX. */
554 set_var_case_index (struct variable *v, int case_index)
556 var_get_vardict (v)->case_index = case_index;
559 /* Removes the dictionary variables with indexes from FROM to TO (exclusive)
562 unindex_vars (struct dictionary *d, size_t from, size_t to)
566 for (i = from; i < to; i++)
567 unindex_var (d, &d->var[i]);
570 /* Re-sets the dict_index in the dictionary variables with
571 indexes from FROM to TO (exclusive). */
573 reindex_vars (struct dictionary *d, size_t from, size_t to)
577 for (i = from; i < to; i++)
578 reindex_var (d, &d->var[i]);
581 /* Deletes variable V from dictionary D and frees V.
583 This is a very bad idea if there might be any pointers to V
584 from outside D. In general, no variable in the active dataset's
585 dictionary should be deleted when any transformations are
586 active on the dictionary's dataset, because those
587 transformations might reference the deleted variable. The
588 safest time to delete a variable is just after a procedure has
589 been executed, as done by DELETE VARIABLES.
591 Pointers to V within D are not a problem, because
592 dict_delete_var() knows to remove V from split variables,
593 weights, filters, etc. */
595 dict_delete_var (struct dictionary *d, struct variable *v)
597 int dict_index = var_get_dict_index (v);
598 const int case_index = var_get_case_index (v);
600 assert (dict_contains_var (d, v));
602 /* Delete aux data. */
605 dict_unset_split_var (d, v);
606 dict_unset_mrset_var (d, v);
609 dict_set_weight (d, NULL);
612 dict_set_filter (d, NULL);
614 dict_clear_vectors (d);
616 /* Remove V from var array. */
617 unindex_vars (d, dict_index, d->var_cnt);
618 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
621 /* Update dict_index for each affected variable. */
622 reindex_vars (d, dict_index, d->var_cnt);
625 var_clear_vardict (v);
627 if ( d->changed ) d->changed (d, d->changed_data);
629 invalidate_proto (d);
630 if (d->callbacks && d->callbacks->var_deleted )
631 d->callbacks->var_deleted (d, v, dict_index, case_index, d->cb_data);
636 /* Deletes the COUNT variables listed in VARS from D. This is
637 unsafe; see the comment on dict_delete_var() for details. */
639 dict_delete_vars (struct dictionary *d,
640 struct variable *const *vars, size_t count)
642 /* FIXME: this can be done in O(count) time, but this algorithm
644 assert (count == 0 || vars != NULL);
647 dict_delete_var (d, *vars++);
650 /* Deletes the COUNT variables in D starting at index IDX. This
651 is unsafe; see the comment on dict_delete_var() for
654 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
656 /* FIXME: this can be done in O(count) time, but this algorithm
658 assert (idx + count <= d->var_cnt);
661 dict_delete_var (d, d->var[idx].var);
664 /* Deletes scratch variables from dictionary D. */
666 dict_delete_scratch_vars (struct dictionary *d)
670 /* FIXME: this can be done in O(count) time, but this algorithm
672 for (i = 0; i < d->var_cnt; )
673 if (var_get_dict_class (d->var[i].var) == DC_SCRATCH)
674 dict_delete_var (d, d->var[i].var);
679 /* Moves V to 0-based position IDX in D. Other variables in D,
680 if any, retain their relative positions. Runs in time linear
681 in the distance moved. */
683 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
685 size_t old_index = var_get_dict_index (v);
687 assert (new_index < d->var_cnt);
689 unindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
690 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
691 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
694 /* Reorders the variables in D, placing the COUNT variables
695 listed in ORDER in that order at the beginning of D. The
696 other variables in D, if any, retain their relative
699 dict_reorder_vars (struct dictionary *d,
700 struct variable *const *order, size_t count)
702 struct vardict_info *new_var;
705 assert (count == 0 || order != NULL);
706 assert (count <= d->var_cnt);
708 new_var = xnmalloc (d->var_cap, sizeof *new_var);
710 /* Add variables in ORDER to new_var. */
711 for (i = 0; i < count; i++)
713 struct vardict_info *old_var;
715 assert (dict_contains_var (d, order[i]));
717 old_var = var_get_vardict (order[i]);
718 new_var[i] = *old_var;
719 old_var->dict = NULL;
722 /* Add remaining variables to new_var. */
723 for (i = 0; i < d->var_cnt; i++)
724 if (d->var[i].dict != NULL)
725 new_var[count++] = d->var[i];
726 assert (count == d->var_cnt);
728 /* Replace old vardicts by new ones. */
732 hmap_clear (&d->name_map);
733 reindex_vars (d, 0, d->var_cnt);
736 /* Changes the name of variable V that is currently in a dictionary to
739 rename_var (struct variable *v, const char *new_name)
741 struct vardict_info *vardict = var_get_vardict (v);
742 var_clear_vardict (v);
743 var_set_name (v, new_name);
744 vardict->name_node.hash = hash_case_string (new_name, 0);
745 var_set_vardict (v, vardict);
748 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
749 a variable named NEW_NAME is already in D, except that
750 NEW_NAME may be the same as V's existing name. */
752 dict_rename_var (struct dictionary *d, struct variable *v,
753 const char *new_name)
755 assert (!strcasecmp (var_get_name (v), new_name)
756 || dict_lookup_var (d, new_name) == NULL);
758 unindex_var (d, var_get_vardict (v));
759 rename_var (v, new_name);
760 reindex_var (d, var_get_vardict (v));
762 if (settings_get_algorithm () == ENHANCED)
763 var_clear_short_names (v);
765 if ( d->changed ) d->changed (d, d->changed_data);
766 if ( d->callbacks && d->callbacks->var_changed )
767 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
770 /* Renames COUNT variables specified in VARS to the names given
771 in NEW_NAMES within dictionary D. If the renaming would
772 result in a duplicate variable name, returns false and stores a
773 name that would be duplicated into *ERR_NAME (if ERR_NAME is
774 non-null). Otherwise, the renaming is successful, and true
777 dict_rename_vars (struct dictionary *d,
778 struct variable **vars, char **new_names, size_t count,
785 assert (count == 0 || vars != NULL);
786 assert (count == 0 || new_names != NULL);
788 /* Save the names of the variables to be renamed. */
789 pool = pool_create ();
790 old_names = pool_nalloc (pool, count, sizeof *old_names);
791 for (i = 0; i < count; i++)
792 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
794 /* Remove the variables to be renamed from the name hash,
796 for (i = 0; i < count; i++)
798 unindex_var (d, var_get_vardict (vars[i]));
799 rename_var (vars[i], new_names[i]);
802 /* Add the renamed variables back into the name hash,
803 checking for conflicts. */
804 for (i = 0; i < count; i++)
806 if (dict_lookup_var (d, var_get_name (vars[i])) != NULL)
808 /* There is a name conflict.
809 Back out all the name changes that have already
810 taken place, and indicate failure. */
812 if (err_name != NULL)
813 *err_name = new_names[i];
815 for (i = 0; i < fail_idx; i++)
816 unindex_var (d, var_get_vardict (vars[i]));
818 for (i = 0; i < count; i++)
820 rename_var (vars[i], old_names[i]);
821 reindex_var (d, var_get_vardict (vars[i]));
827 reindex_var (d, var_get_vardict (vars[i]));
830 /* Clear short names. */
831 if (settings_get_algorithm () == ENHANCED)
832 for (i = 0; i < count; i++)
833 var_clear_short_names (vars[i]);
839 /* Returns true if a variable named NAME may be inserted in DICT;
840 that is, if there is not already a variable with that name in
841 DICT and if NAME is not a reserved word. (The caller's checks
842 have already verified that NAME is otherwise acceptable as a
845 var_name_is_insertable (const struct dictionary *dict, const char *name)
847 return (dict_lookup_var (dict, name) == NULL
848 && lex_id_to_token (ss_cstr (name)) == T_ID);
852 make_hinted_name (const struct dictionary *dict, const char *hint)
854 size_t hint_len = strlen (hint);
855 bool dropped = false;
860 /* The allocation size here is OK: characters that are copied directly fit
861 OK, and characters that are not copied directly are replaced by a single
862 '_' byte. If u8_mbtouc() replaces bad input by 0xfffd, then that will get
863 replaced by '_' too. */
864 root = rp = xmalloc (hint_len + 1);
865 for (ofs = 0; ofs < hint_len; ofs += mblen)
869 mblen = u8_mbtouc (&uc, CHAR_CAST (const uint8_t *, hint + ofs),
872 ? lex_uc_is_id1 (uc) && uc != '$'
873 : lex_uc_is_idn (uc))
880 rp += u8_uctomb (CHAR_CAST (uint8_t *, rp), uc, 6);
891 if (var_name_is_insertable (dict, root))
894 for (i = 0; i < ULONG_MAX; i++)
896 char suffix[INT_BUFSIZE_BOUND (i) + 1];
900 if (!str_format_26adic (i + 1, &suffix[1], sizeof suffix - 1))
903 name = utf8_encoding_concat (root, suffix, dict->encoding, 64);
904 if (var_name_is_insertable (dict, name))
919 make_numeric_name (const struct dictionary *dict, unsigned long int *num_start)
921 unsigned long int number;
923 for (number = num_start != NULL ? MAX (*num_start, 1) : 1;
927 char name[3 + INT_STRLEN_BOUND (number) + 1];
929 sprintf (name, "VAR%03lu", number);
930 if (dict_lookup_var (dict, name) == NULL)
932 if (num_start != NULL)
933 *num_start = number + 1;
934 return xstrdup (name);
942 /* Devises and returns a variable name unique within DICT. The variable name
943 is owned by the caller, which must free it with free() when it is no longer
946 HINT, if it is non-null, is used as a suggestion that will be
947 modified for suitability as a variable name and for
950 If HINT is null or entirely unsuitable, a name in the form
951 "VAR%03d" will be generated, where the smallest unused integer
952 value is used. If NUM_START is non-null, then its value is
953 used as the minimum numeric value to check, and it is updated
954 to the next value to be checked.
957 dict_make_unique_var_name (const struct dictionary *dict, const char *hint,
958 unsigned long int *num_start)
962 char *hinted_name = make_hinted_name (dict, hint);
963 if (hinted_name != NULL)
966 return make_numeric_name (dict, num_start);
969 /* Returns the weighting variable in dictionary D, or a null
970 pointer if the dictionary is unweighted. */
972 dict_get_weight (const struct dictionary *d)
974 assert (d->weight == NULL || dict_contains_var (d, d->weight));
979 /* Returns the value of D's weighting variable in case C, except
980 that a negative weight is returned as 0. Returns 1 if the
981 dictionary is unweighted. Will warn about missing, negative,
982 or zero values if *WARN_ON_INVALID is true. The function will
983 set *WARN_ON_INVALID to false if an invalid weight is
986 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
987 bool *warn_on_invalid)
991 if (d->weight == NULL)
995 double w = case_num (c, d->weight);
996 if (w < 0.0 || var_is_num_missing (d->weight, w, MV_ANY))
998 if ( w == 0.0 && warn_on_invalid != NULL && *warn_on_invalid ) {
999 *warn_on_invalid = false;
1000 msg (SW, _("At least one case in the data file had a weight value "
1001 "that was user-missing, system-missing, zero, or "
1002 "negative. These case(s) were ignored."));
1008 /* Sets the weighting variable of D to V, or turning off
1009 weighting if V is a null pointer. */
1011 dict_set_weight (struct dictionary *d, struct variable *v)
1013 assert (v == NULL || dict_contains_var (d, v));
1014 assert (v == NULL || var_is_numeric (v));
1018 if (d->changed) d->changed (d, d->changed_data);
1019 if ( d->callbacks && d->callbacks->weight_changed )
1020 d->callbacks->weight_changed (d,
1021 v ? var_get_dict_index (v) : -1,
1025 /* Returns the filter variable in dictionary D (see cmd_filter())
1026 or a null pointer if the dictionary is unfiltered. */
1028 dict_get_filter (const struct dictionary *d)
1030 assert (d->filter == NULL || dict_contains_var (d, d->filter));
1035 /* Sets V as the filter variable for dictionary D. Passing a
1036 null pointer for V turn off filtering. */
1038 dict_set_filter (struct dictionary *d, struct variable *v)
1040 assert (v == NULL || dict_contains_var (d, v));
1041 assert (v == NULL || var_is_numeric (v));
1045 if (d->changed) d->changed (d, d->changed_data);
1046 if ( d->callbacks && d->callbacks->filter_changed )
1047 d->callbacks->filter_changed (d,
1048 v ? var_get_dict_index (v) : -1,
1052 /* Returns the case limit for dictionary D, or zero if the number
1053 of cases is unlimited. */
1055 dict_get_case_limit (const struct dictionary *d)
1057 return d->case_limit;
1060 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
1061 0 for CASE_LIMIT to indicate no limit. */
1063 dict_set_case_limit (struct dictionary *d, casenumber case_limit)
1065 d->case_limit = case_limit;
1068 /* Returns the prototype used for cases created by dictionary D. */
1069 const struct caseproto *
1070 dict_get_proto (const struct dictionary *d_)
1072 struct dictionary *d = CONST_CAST (struct dictionary *, d_);
1073 if (d->proto == NULL)
1077 d->proto = caseproto_create ();
1078 d->proto = caseproto_reserve (d->proto, d->var_cnt);
1079 for (i = 0; i < d->var_cnt; i++)
1080 d->proto = caseproto_set_width (d->proto,
1081 var_get_case_index (d->var[i].var),
1082 var_get_width (d->var[i].var));
1087 /* Returns the case index of the next value to be added to D.
1088 This value is the number of `union value's that need to be
1089 allocated to store a case for dictionary D. */
1091 dict_get_next_value_idx (const struct dictionary *d)
1093 return d->next_value_idx;
1096 /* Returns the number of bytes needed to store a case for
1099 dict_get_case_size (const struct dictionary *d)
1101 return sizeof (union value) * dict_get_next_value_idx (d);
1104 /* Reassigns values in dictionary D so that fragmentation is
1107 dict_compact_values (struct dictionary *d)
1111 d->next_value_idx = 0;
1112 for (i = 0; i < d->var_cnt; i++)
1114 struct variable *v = d->var[i].var;
1115 set_var_case_index (v, d->next_value_idx++);
1117 invalidate_proto (d);
1120 /* Returns the number of values occupied by the variables in
1121 dictionary D. All variables are considered if EXCLUDE_CLASSES
1122 is 0, or it may contain one or more of (1u << DC_ORDINARY),
1123 (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to exclude the
1124 corresponding type of variable.
1126 The return value may be less than the number of values in one
1127 of dictionary D's cases (as returned by
1128 dict_get_next_value_idx) even if E is 0, because there may be
1129 gaps in D's cases due to deleted variables. */
1131 dict_count_values (const struct dictionary *d, unsigned int exclude_classes)
1136 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1138 | (1u << DC_SCRATCH))) == 0);
1141 for (i = 0; i < d->var_cnt; i++)
1143 enum dict_class class = var_get_dict_class (d->var[i].var);
1144 if (!(exclude_classes & (1u << class)))
1150 /* Returns the case prototype that would result after deleting
1151 all variables from D that are not in one of the
1152 EXCLUDE_CLASSES and compacting the dictionary with
1155 The caller must unref the returned caseproto when it is no
1158 dict_get_compacted_proto (const struct dictionary *d,
1159 unsigned int exclude_classes)
1161 struct caseproto *proto;
1164 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1166 | (1u << DC_SCRATCH))) == 0);
1168 proto = caseproto_create ();
1169 for (i = 0; i < d->var_cnt; i++)
1171 struct variable *v = d->var[i].var;
1172 if (!(exclude_classes & (1u << var_get_dict_class (v))))
1173 proto = caseproto_add_width (proto, var_get_width (v));
1178 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
1179 dict_get_split_cnt() to determine how many SPLIT FILE vars
1180 there are. Returns a null pointer if and only if there are no
1182 const struct variable *const *
1183 dict_get_split_vars (const struct dictionary *d)
1188 /* Returns the number of SPLIT FILE vars. */
1190 dict_get_split_cnt (const struct dictionary *d)
1192 return d->split_cnt;
1195 /* Removes variable V, which must be in D, from D's set of split
1198 dict_unset_split_var (struct dictionary *d, struct variable *v)
1202 assert (dict_contains_var (d, v));
1204 orig_count = d->split_cnt;
1205 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
1206 &v, compare_var_ptrs, NULL);
1207 if (orig_count != d->split_cnt)
1209 if (d->changed) d->changed (d, d->changed_data);
1210 /* We changed the set of split variables so invoke the
1212 if (d->callbacks && d->callbacks->split_changed)
1213 d->callbacks->split_changed (d, d->cb_data);
1217 /* Sets CNT split vars SPLIT in dictionary D. */
1219 dict_set_split_vars (struct dictionary *d,
1220 struct variable *const *split, size_t cnt)
1222 assert (cnt == 0 || split != NULL);
1227 d->split = xnrealloc (d->split, cnt, sizeof *d->split) ;
1228 memcpy (d->split, split, cnt * sizeof *d->split);
1236 if (d->changed) d->changed (d, d->changed_data);
1237 if ( d->callbacks && d->callbacks->split_changed )
1238 d->callbacks->split_changed (d, d->cb_data);
1241 /* Returns the file label for D, or a null pointer if D is
1242 unlabeled (see cmd_file_label()). */
1244 dict_get_label (const struct dictionary *d)
1249 /* Sets D's file label to LABEL, truncating it to at most 60 bytes in D's
1252 Removes D's label if LABEL is null or the empty string. */
1254 dict_set_label (struct dictionary *d, const char *label)
1257 if (label == NULL || label[0] == '\0')
1260 d->label = utf8_encoding_trunc (label, d->encoding, 60);
1263 /* Returns the documents for D, as an UTF-8 encoded string_array. The
1264 return value is always nonnull; if there are no documents then the
1265 string_arary is empty.*/
1266 const struct string_array *
1267 dict_get_documents (const struct dictionary *d)
1269 return &d->documents;
1272 /* Replaces the documents for D by NEW_DOCS, a UTF-8 encoded string_array. */
1274 dict_set_documents (struct dictionary *d, const struct string_array *new_docs)
1278 dict_clear_documents (d);
1280 for (i = 0; i < new_docs->n; i++)
1281 dict_add_document_line (d, new_docs->strings[i], false);
1284 /* Replaces the documents for D by UTF-8 encoded string NEW_DOCS, dividing it
1285 into individual lines at new-line characters. Each line is truncated to at
1286 most DOC_LINE_LENGTH bytes in D's encoding. */
1288 dict_set_documents_string (struct dictionary *d, const char *new_docs)
1292 dict_clear_documents (d);
1293 for (s = new_docs; *s != '\0'; )
1295 size_t len = strcspn (s, "\n");
1296 char *line = xmemdup0 (s, len);
1297 dict_add_document_line (d, line, false);
1306 /* Drops the documents from dictionary D. */
1308 dict_clear_documents (struct dictionary *d)
1310 string_array_clear (&d->documents);
1313 /* Appends the UTF-8 encoded LINE to the documents in D. LINE will be
1314 truncated so that it is no more than 80 bytes in the dictionary's
1315 encoding. If this causes some text to be lost, and ISSUE_WARNING is true,
1316 then a warning will be issued. */
1318 dict_add_document_line (struct dictionary *d, const char *line,
1324 trunc_len = utf8_encoding_trunc_len (line, d->encoding, DOC_LINE_LENGTH);
1325 truncated = line[trunc_len] != '\0';
1326 if (truncated && issue_warning)
1328 /* Note to translators: "bytes" is correct, not characters */
1329 msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
1332 string_array_append_nocopy (&d->documents, xmemdup0 (line, trunc_len));
1337 /* Returns the number of document lines in dictionary D. */
1339 dict_get_document_line_cnt (const struct dictionary *d)
1341 return d->documents.n;
1344 /* Returns document line number IDX in dictionary D. The caller must not
1345 modify or free the returned string. */
1347 dict_get_document_line (const struct dictionary *d, size_t idx)
1349 assert (idx < d->documents.n);
1350 return d->documents.strings[idx];
1353 /* Creates in D a vector named NAME that contains the CNT
1354 variables in VAR. Returns true if successful, or false if a
1355 vector named NAME already exists in D. */
1357 dict_create_vector (struct dictionary *d,
1359 struct variable **var, size_t cnt)
1364 for (i = 0; i < cnt; i++)
1365 assert (dict_contains_var (d, var[i]));
1367 if (dict_lookup_vector (d, name) == NULL)
1369 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1370 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1377 /* Creates in D a vector named NAME that contains the CNT
1378 variables in VAR. A vector named NAME must not already exist
1381 dict_create_vector_assert (struct dictionary *d,
1383 struct variable **var, size_t cnt)
1385 assert (dict_lookup_vector (d, name) == NULL);
1386 dict_create_vector (d, name, var, cnt);
1389 /* Returns the vector in D with index IDX, which must be less
1390 than dict_get_vector_cnt (D). */
1391 const struct vector *
1392 dict_get_vector (const struct dictionary *d, size_t idx)
1394 assert (idx < d->vector_cnt);
1396 return d->vector[idx];
1399 /* Returns the number of vectors in D. */
1401 dict_get_vector_cnt (const struct dictionary *d)
1403 return d->vector_cnt;
1406 /* Looks up and returns the vector within D with the given
1408 const struct vector *
1409 dict_lookup_vector (const struct dictionary *d, const char *name)
1412 for (i = 0; i < d->vector_cnt; i++)
1413 if (!strcasecmp (vector_get_name (d->vector[i]), name))
1414 return d->vector[i];
1418 /* Deletes all vectors from D. */
1420 dict_clear_vectors (struct dictionary *d)
1424 for (i = 0; i < d->vector_cnt; i++)
1425 vector_destroy (d->vector[i]);
1432 /* Multiple response sets. */
1434 /* Returns the multiple response set in DICT with index IDX, which must be
1435 between 0 and the count returned by dict_get_n_mrsets(), exclusive. */
1436 const struct mrset *
1437 dict_get_mrset (const struct dictionary *dict, size_t idx)
1439 assert (idx < dict->n_mrsets);
1440 return dict->mrsets[idx];
1443 /* Returns the number of multiple response sets in DICT. */
1445 dict_get_n_mrsets (const struct dictionary *dict)
1447 return dict->n_mrsets;
1450 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1451 returns its index; otherwise, returns SIZE_MAX. */
1453 dict_lookup_mrset_idx (const struct dictionary *dict, const char *name)
1457 for (i = 0; i < dict->n_mrsets; i++)
1458 if (!strcasecmp (name, dict->mrsets[i]->name))
1464 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1465 returns it; otherwise, returns NULL. */
1466 const struct mrset *
1467 dict_lookup_mrset (const struct dictionary *dict, const char *name)
1469 size_t idx = dict_lookup_mrset_idx (dict, name);
1470 return idx != SIZE_MAX ? dict->mrsets[idx] : NULL;
1473 /* Adds MRSET to DICT, replacing any existing set with the same name. Returns
1474 true if a set was replaced, false if none existed with the specified name.
1476 Ownership of MRSET is transferred to DICT. */
1478 dict_add_mrset (struct dictionary *dict, struct mrset *mrset)
1482 assert (mrset_ok (mrset, dict));
1484 idx = dict_lookup_mrset_idx (dict, mrset->name);
1485 if (idx == SIZE_MAX)
1487 dict->mrsets = xrealloc (dict->mrsets,
1488 (dict->n_mrsets + 1) * sizeof *dict->mrsets);
1489 dict->mrsets[dict->n_mrsets++] = mrset;
1494 mrset_destroy (dict->mrsets[idx]);
1495 dict->mrsets[idx] = mrset;
1500 /* Looks for a multiple response set in DICT named NAME. If found, removes it
1501 from DICT and returns true. If none is found, returns false without
1504 Deleting one multiple response set causes the indexes of other sets within
1507 dict_delete_mrset (struct dictionary *dict, const char *name)
1509 size_t idx = dict_lookup_mrset_idx (dict, name);
1510 if (idx != SIZE_MAX)
1512 mrset_destroy (dict->mrsets[idx]);
1513 dict->mrsets[idx] = dict->mrsets[--dict->n_mrsets];
1520 /* Deletes all multiple response sets from DICT. */
1522 dict_clear_mrsets (struct dictionary *dict)
1526 for (i = 0; i < dict->n_mrsets; i++)
1527 mrset_destroy (dict->mrsets[i]);
1528 free (dict->mrsets);
1529 dict->mrsets = NULL;
1533 /* Removes VAR, which must be in DICT, from DICT's multiple response sets. */
1535 dict_unset_mrset_var (struct dictionary *dict, struct variable *var)
1539 assert (dict_contains_var (dict, var));
1541 for (i = 0; i < dict->n_mrsets; )
1543 struct mrset *mrset = dict->mrsets[i];
1546 for (j = 0; j < mrset->n_vars; )
1547 if (mrset->vars[j] == var)
1548 remove_element (mrset->vars, mrset->n_vars--,
1549 sizeof *mrset->vars, j);
1553 if (mrset->n_vars < 2)
1555 mrset_destroy (mrset);
1556 dict->mrsets[i] = dict->mrsets[--dict->n_mrsets];
1563 /* Returns D's attribute set. The caller may examine or modify
1564 the attribute set, but must not destroy it. Destroying D or
1565 calling dict_set_attributes for D will also destroy D's
1568 dict_get_attributes (const struct dictionary *d)
1570 return CONST_CAST (struct attrset *, &d->attributes);
1573 /* Replaces D's attributes set by a copy of ATTRS. */
1575 dict_set_attributes (struct dictionary *d, const struct attrset *attrs)
1577 attrset_destroy (&d->attributes);
1578 attrset_clone (&d->attributes, attrs);
1581 /* Returns true if D has at least one attribute in its attribute
1582 set, false if D's attribute set is empty. */
1584 dict_has_attributes (const struct dictionary *d)
1586 return attrset_count (&d->attributes) > 0;
1589 /* Called from variable.c to notify the dictionary that some property of
1590 the variable has changed */
1592 dict_var_changed (const struct variable *v)
1594 if ( var_has_vardict (v))
1596 const struct vardict_info *vardict = var_get_vardict (v);
1597 struct dictionary *d = vardict->dict;
1602 if (d->changed ) d->changed (d, d->changed_data);
1603 if ( d->callbacks && d->callbacks->var_changed )
1604 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
1609 /* Called from variable.c to notify the dictionary that the variable's width
1612 dict_var_resized (const struct variable *v, int old_width)
1614 if ( var_has_vardict (v))
1616 const struct vardict_info *vardict = var_get_vardict (v);
1617 struct dictionary *d;
1621 if (d->changed) d->changed (d, d->changed_data);
1623 invalidate_proto (d);
1624 if ( d->callbacks && d->callbacks->var_resized )
1625 d->callbacks->var_resized (d, var_get_dict_index (v), old_width,
1630 /* Called from variable.c to notify the dictionary that the variable's display width
1633 dict_var_display_width_changed (const struct variable *v)
1635 if ( var_has_vardict (v))
1637 const struct vardict_info *vardict = var_get_vardict (v);
1638 struct dictionary *d;
1642 if (d->changed) d->changed (d, d->changed_data);
1643 if ( d->callbacks && d->callbacks->var_display_width_changed )
1644 d->callbacks->var_display_width_changed (d, var_get_dict_index (v), d->cb_data);
1648 /* Dictionary used to contain "internal variables". */
1649 static struct dictionary *internal_dict;
1651 /* Create a variable of the specified WIDTH to be used for internal
1652 calculations only. The variable is assigned case index CASE_IDX. */
1654 dict_create_internal_var (int case_idx, int width)
1656 if (internal_dict == NULL)
1657 internal_dict = dict_create ("UTF-8");
1661 static int counter = INT_MAX / 2;
1662 struct variable *var;
1665 if (++counter == INT_MAX)
1666 counter = INT_MAX / 2;
1668 sprintf (name, "$internal%d", counter);
1669 var = dict_create_var (internal_dict, name, width);
1672 set_var_case_index (var, case_idx);
1678 /* Destroys VAR, which must have been created with
1679 dict_create_internal_var(). */
1681 dict_destroy_internal_var (struct variable *var)
1685 dict_delete_var (internal_dict, var);
1687 /* Destroy internal_dict if it has no variables left, just so that
1688 valgrind --leak-check --show-reachable won't show internal_dict. */
1689 if (dict_get_var_cnt (internal_dict) == 0)
1691 dict_destroy (internal_dict);
1692 internal_dict = NULL;
1698 vardict_get_dict_index (const struct vardict_info *vardict)
1700 return vardict - vardict->dict->var;