1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006, 2007, 2009, 2010, 2011 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"
25 #include "data/attributes.h"
26 #include "data/case.h"
27 #include "data/identifier.h"
28 #include "data/mrset.h"
29 #include "data/settings.h"
30 #include "data/value-labels.h"
31 #include "data/vardict.h"
32 #include "data/variable.h"
33 #include "data/vector.h"
34 #include "libpspp/array.h"
35 #include "libpspp/assertion.h"
36 #include "libpspp/compiler.h"
37 #include "libpspp/hash-functions.h"
38 #include "libpspp/hmap.h"
39 #include "libpspp/message.h"
40 #include "libpspp/misc.h"
41 #include "libpspp/pool.h"
42 #include "libpspp/str.h"
44 #include "gl/intprops.h"
45 #include "gl/minmax.h"
46 #include "gl/xalloc.h"
49 #define _(msgid) gettext (msgid)
54 struct vardict_info *var; /* Variables. */
55 size_t var_cnt, var_cap; /* Number of variables, capacity. */
56 struct caseproto *proto; /* Prototype for dictionary cases
58 struct hmap name_map; /* Variable index by name. */
59 int next_value_idx; /* Index of next `union value' to allocate. */
60 const struct variable **split; /* SPLIT FILE vars. */
61 size_t split_cnt; /* SPLIT FILE count. */
62 struct variable *weight; /* WEIGHT variable. */
63 struct variable *filter; /* FILTER variable. */
64 casenumber case_limit; /* Current case limit (N command). */
65 char *label; /* File label. */
66 struct string documents; /* Documents, as a string. */
67 struct vector **vector; /* Vectors of variables. */
68 size_t vector_cnt; /* Number of vectors. */
69 struct attrset attributes; /* Custom attributes. */
70 struct mrset **mrsets; /* Multiple response sets. */
71 size_t n_mrsets; /* Number of multiple response sets. */
73 char *encoding; /* Character encoding of string data */
75 const struct dict_callbacks *callbacks; /* Callbacks on dictionary
77 void *cb_data ; /* Data passed to callbacks */
79 void (*changed) (struct dictionary *, void *); /* Generic change callback */
83 static void dict_unset_split_var (struct dictionary *, struct variable *);
84 static void dict_unset_mrset_var (struct dictionary *, struct variable *);
87 dict_set_encoding (struct dictionary *d, const char *enc)
92 d->encoding = xstrdup (enc);
97 dict_get_encoding (const struct dictionary *d)
104 dict_set_change_callback (struct dictionary *d,
105 void (*changed) (struct dictionary *, void*),
108 d->changed = changed;
109 d->changed_data = data;
112 /* Discards dictionary D's caseproto. (It will be regenerated
113 lazily, on demand.) */
115 invalidate_proto (struct dictionary *d)
117 caseproto_unref (d->proto);
121 /* Print a representation of dictionary D to stdout, for
122 debugging purposes. */
124 dict_dump (const struct dictionary *d)
127 for (i = 0 ; i < d->var_cnt ; ++i )
129 const struct variable *v = d->var[i].var;
130 printf ("Name: %s;\tdict_idx: %zu; case_idx: %zu\n",
132 var_get_dict_index (v),
133 var_get_case_index (v));
138 /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
139 the dictionary or any of the variables it contains are modified.
140 Each callback will get passed CALLBACK_DATA.
141 Any callback may be NULL, in which case it'll be ignored.
144 dict_set_callbacks (struct dictionary *dict,
145 const struct dict_callbacks *callbacks,
148 dict->callbacks = callbacks;
149 dict->cb_data = callback_data;
152 /* Shallow copy the callbacks from SRC to DEST */
154 dict_copy_callbacks (struct dictionary *dest,
155 const struct dictionary *src)
157 dest->callbacks = src->callbacks;
158 dest->cb_data = src->cb_data;
161 /* Creates and returns a new dictionary. */
165 struct dictionary *d = xzalloc (sizeof *d);
167 hmap_init (&d->name_map);
168 attrset_init (&d->attributes);
172 /* Creates and returns a (deep) copy of an existing
175 The new dictionary's case indexes are copied from the old
176 dictionary. If the new dictionary won't be used to access
177 cases produced with the old dictionary, then the new
178 dictionary's case indexes should be compacted with
179 dict_compact_values to save space. */
181 dict_clone (const struct dictionary *s)
183 struct dictionary *d;
188 for (i = 0; i < s->var_cnt; i++)
190 struct variable *sv = s->var[i].var;
191 struct variable *dv = dict_clone_var_assert (d, sv);
194 for (i = 0; i < var_get_short_name_cnt (sv); i++)
195 var_set_short_name (dv, i, var_get_short_name (sv, i));
197 var_get_vardict (dv)->case_index = var_get_vardict (sv)->case_index;
200 d->next_value_idx = s->next_value_idx;
202 d->split_cnt = s->split_cnt;
203 if (d->split_cnt > 0)
205 d->split = xnmalloc (d->split_cnt, sizeof *d->split);
206 for (i = 0; i < d->split_cnt; i++)
207 d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i]));
210 if (s->weight != NULL)
211 dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight)));
213 if (s->filter != NULL)
214 dict_set_filter (d, dict_lookup_var_assert (d, var_get_name (s->filter)));
216 d->case_limit = s->case_limit;
217 dict_set_label (d, dict_get_label (s));
218 dict_set_documents (d, dict_get_documents (s));
220 d->vector_cnt = s->vector_cnt;
221 d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector);
222 for (i = 0; i < s->vector_cnt; i++)
223 d->vector[i] = vector_clone (s->vector[i], s, d);
226 d->encoding = xstrdup (s->encoding);
228 dict_set_attributes (d, dict_get_attributes (s));
230 for (i = 0; i < s->n_mrsets; i++)
232 const struct mrset *old = s->mrsets[i];
236 /* Clone old mrset, then replace vars from D by vars from S. */
237 new = mrset_clone (old);
238 for (j = 0; j < new->n_vars; j++)
239 new->vars[j] = dict_lookup_var_assert (d, var_get_name (new->vars[j]));
241 dict_add_mrset (d, new);
247 /* Clears the contents from a dictionary without destroying the
248 dictionary itself. */
250 dict_clear (struct dictionary *d)
252 /* FIXME? Should we really clear case_limit, label, documents?
253 Others are necessarily cleared by deleting all the variables.*/
254 while (d->var_cnt > 0 )
256 dict_delete_var (d, d->var[d->var_cnt - 1].var);
261 d->var_cnt = d->var_cap = 0;
262 invalidate_proto (d);
263 hmap_clear (&d->name_map);
264 d->next_value_idx = 0;
265 dict_set_split_vars (d, NULL, 0);
266 dict_set_weight (d, NULL);
267 dict_set_filter (d, NULL);
271 ds_destroy (&d->documents);
272 dict_clear_vectors (d);
273 attrset_clear (&d->attributes);
276 /* Destroys the aux data for every variable in D, by calling
277 var_clear_aux() for each variable. */
279 dict_clear_aux (struct dictionary *d)
283 for (i = 0; i < d->var_cnt; i++)
284 var_clear_aux (d->var[i].var);
287 /* Clears a dictionary and destroys it. */
289 dict_destroy (struct dictionary *d)
293 /* In general, we don't want callbacks occuring, if the dictionary
294 is being destroyed */
295 d->callbacks = NULL ;
298 hmap_destroy (&d->name_map);
299 attrset_destroy (&d->attributes);
300 dict_clear_mrsets (d);
306 /* Returns the number of variables in D. */
308 dict_get_var_cnt (const struct dictionary *d)
313 /* Returns the variable in D with dictionary index IDX, which
314 must be between 0 and the count returned by
315 dict_get_var_cnt(), exclusive. */
317 dict_get_var (const struct dictionary *d, size_t idx)
319 assert (idx < d->var_cnt);
321 return d->var[idx].var;
324 /* Sets *VARS to an array of pointers to variables in D and *CNT
325 to the number of variables in *D. All variables are returned
326 except for those, if any, in the classes indicated by EXCLUDE.
327 (There is no point in putting DC_SYSTEM in EXCLUDE as
328 dictionaries never include system variables.) */
330 dict_get_vars (const struct dictionary *d, const struct variable ***vars,
331 size_t *cnt, enum dict_class exclude)
333 dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude);
336 /* Sets *VARS to an array of pointers to variables in D and *CNT
337 to the number of variables in *D. All variables are returned
338 except for those, if any, in the classes indicated by EXCLUDE.
339 (There is no point in putting DC_SYSTEM in EXCLUDE as
340 dictionaries never include system variables.) */
342 dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
343 size_t *cnt, enum dict_class exclude)
348 assert (exclude == (exclude & DC_ALL));
351 for (i = 0; i < d->var_cnt; i++)
353 enum dict_class class = var_get_dict_class (d->var[i].var);
354 if (!(class & exclude))
358 *vars = xnmalloc (count, sizeof **vars);
360 for (i = 0; i < d->var_cnt; i++)
362 enum dict_class class = var_get_dict_class (d->var[i].var);
363 if (!(class & exclude))
364 (*vars)[(*cnt)++] = d->var[i].var;
366 assert (*cnt == count);
369 static struct variable *
370 add_var (struct dictionary *d, struct variable *v)
372 struct vardict_info *vardict;
374 /* Update dictionary. */
375 if (d->var_cnt >= d->var_cap)
379 d->var = x2nrealloc (d->var, &d->var_cap, sizeof *d->var);
380 hmap_clear (&d->name_map);
381 for (i = 0; i < d->var_cnt; i++)
383 var_set_vardict (d->var[i].var, &d->var[i]);
384 hmap_insert_fast (&d->name_map, &d->var[i].name_node,
385 d->var[i].name_node.hash);
389 vardict = &d->var[d->var_cnt++];
392 hmap_insert (&d->name_map, &vardict->name_node,
393 hash_case_string (var_get_name (v), 0));
394 vardict->case_index = d->next_value_idx;
395 var_set_vardict (v, vardict);
397 if ( d->changed ) d->changed (d, d->changed_data);
398 if ( d->callbacks && d->callbacks->var_added )
399 d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
402 invalidate_proto (d);
407 /* Creates and returns a new variable in D with the given NAME
408 and WIDTH. Returns a null pointer if the given NAME would
409 duplicate that of an existing variable in the dictionary. */
411 dict_create_var (struct dictionary *d, const char *name, int width)
413 return (dict_lookup_var (d, name) == NULL
414 ? dict_create_var_assert (d, name, width)
418 /* Creates and returns a new variable in D with the given NAME
419 and WIDTH. Assert-fails if the given NAME would duplicate
420 that of an existing variable in the dictionary. */
422 dict_create_var_assert (struct dictionary *d, const char *name, int width)
424 assert (dict_lookup_var (d, name) == NULL);
425 return add_var (d, var_create (name, width));
428 /* Creates and returns a new variable in D, as a copy of existing variable
429 OLD_VAR, which need not be in D or in any dictionary. Returns a null
430 pointer if OLD_VAR's name would duplicate that of an existing variable in
433 dict_clone_var (struct dictionary *d, const struct variable *old_var)
435 return dict_clone_var_as (d, old_var, var_get_name (old_var));
438 /* Creates and returns a new variable in D, as a copy of existing variable
439 OLD_VAR, which need not be in D or in any dictionary. Assert-fails if
440 OLD_VAR's name would duplicate that of an existing variable in the
443 dict_clone_var_assert (struct dictionary *d, const struct variable *old_var)
445 return dict_clone_var_as_assert (d, old_var, var_get_name (old_var));
448 /* Creates and returns a new variable in D with name NAME, as a copy of
449 existing variable OLD_VAR, which need not be in D or in any dictionary.
450 Returns a null pointer if the given NAME would duplicate that of an existing
451 variable in the dictionary. */
453 dict_clone_var_as (struct dictionary *d, const struct variable *old_var,
456 return (dict_lookup_var (d, name) == NULL
457 ? dict_clone_var_as_assert (d, old_var, name)
461 /* Creates and returns a new variable in D with name NAME, as a copy of
462 existing variable OLD_VAR, which need not be in D or in any dictionary.
463 Assert-fails if the given NAME would duplicate that of an existing variable
464 in the dictionary. */
466 dict_clone_var_as_assert (struct dictionary *d, const struct variable *old_var,
469 struct variable *new_var = var_clone (old_var);
470 assert (dict_lookup_var (d, name) == NULL);
471 var_set_name (new_var, name);
472 return add_var (d, new_var);
475 /* Returns the variable named NAME in D, or a null pointer if no
476 variable has that name. */
478 dict_lookup_var (const struct dictionary *d, const char *name)
480 struct vardict_info *vardict;
482 HMAP_FOR_EACH_WITH_HASH (vardict, struct vardict_info, name_node,
483 hash_case_string (name, 0), &d->name_map)
485 struct variable *var = vardict->var;
486 if (!strcasecmp (var_get_name (var), name))
493 /* Returns the variable named NAME in D. Assert-fails if no
494 variable has that name. */
496 dict_lookup_var_assert (const struct dictionary *d, const char *name)
498 struct variable *v = dict_lookup_var (d, name);
503 /* Returns true if variable V is in dictionary D,
506 dict_contains_var (const struct dictionary *d, const struct variable *v)
508 return (var_has_vardict (v)
509 && vardict_get_dictionary (var_get_vardict (v)) == d);
512 /* Compares two double pointers to variables, which should point
513 to elements of a struct dictionary's `var' member array. */
515 compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED)
517 struct variable *const *a = a_;
518 struct variable *const *b = b_;
520 return *a < *b ? -1 : *a > *b;
524 unindex_var (struct dictionary *d, struct vardict_info *vardict)
526 hmap_delete (&d->name_map, &vardict->name_node);
529 /* This function assumes that vardict->name_node.hash is valid, that is, that
530 its name has not changed since it was hashed (rename_var() updates this
531 hash along with the name itself). */
533 reindex_var (struct dictionary *d, struct vardict_info *vardict)
535 struct variable *var = vardict->var;
537 var_set_vardict (var, vardict);
538 hmap_insert_fast (&d->name_map, &vardict->name_node,
539 vardict->name_node.hash);
541 if ( d->changed ) d->changed (d, d->changed_data);
542 if ( d->callbacks && d->callbacks->var_changed )
543 d->callbacks->var_changed (d, var_get_dict_index (var), d->cb_data);
546 /* Sets the case_index in V's vardict to CASE_INDEX. */
548 set_var_case_index (struct variable *v, int case_index)
550 var_get_vardict (v)->case_index = case_index;
553 /* Removes the dictionary variables with indexes from FROM to TO (exclusive)
556 unindex_vars (struct dictionary *d, size_t from, size_t to)
560 for (i = from; i < to; i++)
561 unindex_var (d, &d->var[i]);
564 /* Re-sets the dict_index in the dictionary variables with
565 indexes from FROM to TO (exclusive). */
567 reindex_vars (struct dictionary *d, size_t from, size_t to)
571 for (i = from; i < to; i++)
572 reindex_var (d, &d->var[i]);
575 /* Deletes variable V from dictionary D and frees V.
577 This is a very bad idea if there might be any pointers to V
578 from outside D. In general, no variable in the active file's
579 dictionary should be deleted when any transformations are
580 active on the dictionary's dataset, because those
581 transformations might reference the deleted variable. The
582 safest time to delete a variable is just after a procedure has
583 been executed, as done by DELETE VARIABLES.
585 Pointers to V within D are not a problem, because
586 dict_delete_var() knows to remove V from split variables,
587 weights, filters, etc. */
589 dict_delete_var (struct dictionary *d, struct variable *v)
591 int dict_index = var_get_dict_index (v);
592 const int case_index = var_get_case_index (v);
593 const int width = var_get_width (v);
595 assert (dict_contains_var (d, v));
597 /* Delete aux data. */
600 dict_unset_split_var (d, v);
601 dict_unset_mrset_var (d, v);
604 dict_set_weight (d, NULL);
607 dict_set_filter (d, NULL);
609 dict_clear_vectors (d);
611 /* Remove V from var array. */
612 unindex_vars (d, dict_index, d->var_cnt);
613 remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
616 /* Update dict_index for each affected variable. */
617 reindex_vars (d, dict_index, d->var_cnt);
620 var_clear_vardict (v);
623 if ( d->changed ) d->changed (d, d->changed_data);
625 invalidate_proto (d);
626 if (d->callbacks && d->callbacks->var_deleted )
627 d->callbacks->var_deleted (d, dict_index, case_index, width, d->cb_data);
630 /* Deletes the COUNT variables listed in VARS from D. This is
631 unsafe; see the comment on dict_delete_var() for details. */
633 dict_delete_vars (struct dictionary *d,
634 struct variable *const *vars, size_t count)
636 /* FIXME: this can be done in O(count) time, but this algorithm
638 assert (count == 0 || vars != NULL);
641 dict_delete_var (d, *vars++);
644 /* Deletes the COUNT variables in D starting at index IDX. This
645 is unsafe; see the comment on dict_delete_var() for
648 dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count)
650 /* FIXME: this can be done in O(count) time, but this algorithm
652 assert (idx + count <= d->var_cnt);
655 dict_delete_var (d, d->var[idx].var);
658 /* Deletes scratch variables from dictionary D. */
660 dict_delete_scratch_vars (struct dictionary *d)
664 /* FIXME: this can be done in O(count) time, but this algorithm
666 for (i = 0; i < d->var_cnt; )
667 if (var_get_dict_class (d->var[i].var) == DC_SCRATCH)
668 dict_delete_var (d, d->var[i].var);
673 /* Moves V to 0-based position IDX in D. Other variables in D,
674 if any, retain their relative positions. Runs in time linear
675 in the distance moved. */
677 dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index)
679 size_t old_index = var_get_dict_index (v);
681 assert (new_index < d->var_cnt);
683 unindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
684 move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
685 reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
688 /* Reorders the variables in D, placing the COUNT variables
689 listed in ORDER in that order at the beginning of D. The
690 other variables in D, if any, retain their relative
693 dict_reorder_vars (struct dictionary *d,
694 struct variable *const *order, size_t count)
696 struct vardict_info *new_var;
699 assert (count == 0 || order != NULL);
700 assert (count <= d->var_cnt);
702 new_var = xnmalloc (d->var_cap, sizeof *new_var);
704 /* Add variables in ORDER to new_var. */
705 for (i = 0; i < count; i++)
707 struct vardict_info *old_var;
709 assert (dict_contains_var (d, order[i]));
711 old_var = var_get_vardict (order[i]);
712 new_var[i] = *old_var;
713 old_var->dict = NULL;
716 /* Add remaining variables to new_var. */
717 for (i = 0; i < d->var_cnt; i++)
718 if (d->var[i].dict != NULL)
719 new_var[count++] = d->var[i];
720 assert (count == d->var_cnt);
722 /* Replace old vardicts by new ones. */
726 hmap_clear (&d->name_map);
727 reindex_vars (d, 0, d->var_cnt);
730 /* Changes the name of variable V that is currently in a dictionary to
733 rename_var (struct variable *v, const char *new_name)
735 struct vardict_info *vardict = var_get_vardict (v);
736 var_clear_vardict (v);
737 var_set_name (v, new_name);
738 vardict->name_node.hash = hash_case_string (new_name, 0);
739 var_set_vardict (v, vardict);
742 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
743 a variable named NEW_NAME is already in D, except that
744 NEW_NAME may be the same as V's existing name. */
746 dict_rename_var (struct dictionary *d, struct variable *v,
747 const char *new_name)
749 assert (!strcasecmp (var_get_name (v), new_name)
750 || dict_lookup_var (d, new_name) == NULL);
752 unindex_var (d, var_get_vardict (v));
753 rename_var (v, new_name);
754 reindex_var (d, var_get_vardict (v));
756 if (settings_get_algorithm () == ENHANCED)
757 var_clear_short_names (v);
759 if ( d->changed ) d->changed (d, d->changed_data);
760 if ( d->callbacks && d->callbacks->var_changed )
761 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
764 /* Renames COUNT variables specified in VARS to the names given
765 in NEW_NAMES within dictionary D. If the renaming would
766 result in a duplicate variable name, returns false and stores a
767 name that would be duplicated into *ERR_NAME (if ERR_NAME is
768 non-null). Otherwise, the renaming is successful, and true
771 dict_rename_vars (struct dictionary *d,
772 struct variable **vars, char **new_names, size_t count,
779 assert (count == 0 || vars != NULL);
780 assert (count == 0 || new_names != NULL);
782 /* Save the names of the variables to be renamed. */
783 pool = pool_create ();
784 old_names = pool_nalloc (pool, count, sizeof *old_names);
785 for (i = 0; i < count; i++)
786 old_names[i] = pool_strdup (pool, var_get_name (vars[i]));
788 /* Remove the variables to be renamed from the name hash,
790 for (i = 0; i < count; i++)
792 unindex_var (d, var_get_vardict (vars[i]));
793 rename_var (vars[i], new_names[i]);
796 /* Add the renamed variables back into the name hash,
797 checking for conflicts. */
798 for (i = 0; i < count; i++)
800 if (dict_lookup_var (d, var_get_name (vars[i])) != NULL)
802 /* There is a name conflict.
803 Back out all the name changes that have already
804 taken place, and indicate failure. */
806 if (err_name != NULL)
807 *err_name = new_names[i];
809 for (i = 0; i < fail_idx; i++)
810 unindex_var (d, var_get_vardict (vars[i]));
812 for (i = 0; i < count; i++)
814 rename_var (vars[i], old_names[i]);
815 reindex_var (d, var_get_vardict (vars[i]));
821 reindex_var (d, var_get_vardict (vars[i]));
824 /* Clear short names. */
825 if (settings_get_algorithm () == ENHANCED)
826 for (i = 0; i < count; i++)
827 var_clear_short_names (vars[i]);
833 /* Returns true if a variable named NAME may be inserted in DICT;
834 that is, if there is not already a variable with that name in
835 DICT and if NAME is not a reserved word. (The caller's checks
836 have already verified that NAME is otherwise acceptable as a
839 var_name_is_insertable (const struct dictionary *dict, const char *name)
841 return (dict_lookup_var (dict, name) == NULL
842 && lex_id_to_token (ss_cstr (name)) == T_ID);
846 make_hinted_name (const struct dictionary *dict, const char *hint)
848 char name[VAR_NAME_LEN + 1];
849 bool dropped = false;
852 for (cp = name; *hint && cp < name + VAR_NAME_LEN; hint++)
855 ? lex_is_id1 (*hint) && *hint != '$'
856 : lex_is_idn (*hint))
863 if (cp < name + VAR_NAME_LEN)
873 size_t len = strlen (name);
876 if (var_name_is_insertable (dict, name))
877 return xstrdup (name);
879 for (i = 0; i < ULONG_MAX; i++)
881 char suffix[INT_BUFSIZE_BOUND (i) + 1];
885 if (!str_format_26adic (i + 1, &suffix[1], sizeof suffix - 1))
888 ofs = MIN (VAR_NAME_LEN - strlen (suffix), len);
889 strcpy (&name[ofs], suffix);
891 if (var_name_is_insertable (dict, name))
892 return xstrdup (name);
900 make_numeric_name (const struct dictionary *dict, unsigned long int *num_start)
902 unsigned long int number;
904 for (number = num_start != NULL ? MAX (*num_start, 1) : 1;
908 char name[3 + INT_STRLEN_BOUND (number) + 1];
910 sprintf (name, "VAR%03lu", number);
911 if (dict_lookup_var (dict, name) == NULL)
913 if (num_start != NULL)
914 *num_start = number + 1;
915 return xstrdup (name);
923 /* Devises and returns a variable name unique within DICT. The variable name
924 is owned by the caller, which must free it with free() when it is no longer
927 HINT, if it is non-null, is used as a suggestion that will be
928 modified for suitability as a variable name and for
931 If HINT is null or entirely unsuitable, a name in the form
932 "VAR%03d" will be generated, where the smallest unused integer
933 value is used. If NUM_START is non-null, then its value is
934 used as the minimum numeric value to check, and it is updated
935 to the next value to be checked.
938 dict_make_unique_var_name (const struct dictionary *dict, const char *hint,
939 unsigned long int *num_start)
943 char *hinted_name = make_hinted_name (dict, hint);
944 if (hinted_name != NULL)
947 return make_numeric_name (dict, num_start);
950 /* Returns the weighting variable in dictionary D, or a null
951 pointer if the dictionary is unweighted. */
953 dict_get_weight (const struct dictionary *d)
955 assert (d->weight == NULL || dict_contains_var (d, d->weight));
960 /* Returns the value of D's weighting variable in case C, except
961 that a negative weight is returned as 0. Returns 1 if the
962 dictionary is unweighted. Will warn about missing, negative,
963 or zero values if *WARN_ON_INVALID is true. The function will
964 set *WARN_ON_INVALID to false if an invalid weight is
967 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
968 bool *warn_on_invalid)
972 if (d->weight == NULL)
976 double w = case_num (c, d->weight);
977 if (w < 0.0 || var_is_num_missing (d->weight, w, MV_ANY))
979 if ( w == 0.0 && warn_on_invalid != NULL && *warn_on_invalid ) {
980 *warn_on_invalid = false;
981 msg (SW, _("At least one case in the data file had a weight value "
982 "that was user-missing, system-missing, zero, or "
983 "negative. These case(s) were ignored."));
989 /* Sets the weighting variable of D to V, or turning off
990 weighting if V is a null pointer. */
992 dict_set_weight (struct dictionary *d, struct variable *v)
994 assert (v == NULL || dict_contains_var (d, v));
995 assert (v == NULL || var_is_numeric (v));
999 if (d->changed) d->changed (d, d->changed_data);
1000 if ( d->callbacks && d->callbacks->weight_changed )
1001 d->callbacks->weight_changed (d,
1002 v ? var_get_dict_index (v) : -1,
1006 /* Returns the filter variable in dictionary D (see cmd_filter())
1007 or a null pointer if the dictionary is unfiltered. */
1009 dict_get_filter (const struct dictionary *d)
1011 assert (d->filter == NULL || dict_contains_var (d, d->filter));
1016 /* Sets V as the filter variable for dictionary D. Passing a
1017 null pointer for V turn off filtering. */
1019 dict_set_filter (struct dictionary *d, struct variable *v)
1021 assert (v == NULL || dict_contains_var (d, v));
1022 assert (v == NULL || var_is_numeric (v));
1026 if (d->changed) d->changed (d, d->changed_data);
1027 if ( d->callbacks && d->callbacks->filter_changed )
1028 d->callbacks->filter_changed (d,
1029 v ? var_get_dict_index (v) : -1,
1033 /* Returns the case limit for dictionary D, or zero if the number
1034 of cases is unlimited. */
1036 dict_get_case_limit (const struct dictionary *d)
1038 return d->case_limit;
1041 /* Sets CASE_LIMIT as the case limit for dictionary D. Use
1042 0 for CASE_LIMIT to indicate no limit. */
1044 dict_set_case_limit (struct dictionary *d, casenumber case_limit)
1046 d->case_limit = case_limit;
1049 /* Returns the prototype used for cases created by dictionary D. */
1050 const struct caseproto *
1051 dict_get_proto (const struct dictionary *d_)
1053 struct dictionary *d = CONST_CAST (struct dictionary *, d_);
1054 if (d->proto == NULL)
1058 d->proto = caseproto_create ();
1059 d->proto = caseproto_reserve (d->proto, d->var_cnt);
1060 for (i = 0; i < d->var_cnt; i++)
1061 d->proto = caseproto_set_width (d->proto,
1062 var_get_case_index (d->var[i].var),
1063 var_get_width (d->var[i].var));
1068 /* Returns the case index of the next value to be added to D.
1069 This value is the number of `union value's that need to be
1070 allocated to store a case for dictionary D. */
1072 dict_get_next_value_idx (const struct dictionary *d)
1074 return d->next_value_idx;
1077 /* Returns the number of bytes needed to store a case for
1080 dict_get_case_size (const struct dictionary *d)
1082 return sizeof (union value) * dict_get_next_value_idx (d);
1085 /* Reassigns values in dictionary D so that fragmentation is
1088 dict_compact_values (struct dictionary *d)
1092 d->next_value_idx = 0;
1093 for (i = 0; i < d->var_cnt; i++)
1095 struct variable *v = d->var[i].var;
1096 set_var_case_index (v, d->next_value_idx++);
1098 invalidate_proto (d);
1101 /* Returns the number of values occupied by the variables in
1102 dictionary D. All variables are considered if EXCLUDE_CLASSES
1103 is 0, or it may contain one or more of (1u << DC_ORDINARY),
1104 (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to exclude the
1105 corresponding type of variable.
1107 The return value may be less than the number of values in one
1108 of dictionary D's cases (as returned by
1109 dict_get_next_value_idx) even if E is 0, because there may be
1110 gaps in D's cases due to deleted variables. */
1112 dict_count_values (const struct dictionary *d, unsigned int exclude_classes)
1117 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1119 | (1u << DC_SCRATCH))) == 0);
1122 for (i = 0; i < d->var_cnt; i++)
1124 enum dict_class class = var_get_dict_class (d->var[i].var);
1125 if (!(exclude_classes & (1u << class)))
1131 /* Returns the case prototype that would result after deleting
1132 all variables from D that are not in one of the
1133 EXCLUDE_CLASSES and compacting the dictionary with
1136 The caller must unref the returned caseproto when it is no
1139 dict_get_compacted_proto (const struct dictionary *d,
1140 unsigned int exclude_classes)
1142 struct caseproto *proto;
1145 assert ((exclude_classes & ~((1u << DC_ORDINARY)
1147 | (1u << DC_SCRATCH))) == 0);
1149 proto = caseproto_create ();
1150 for (i = 0; i < d->var_cnt; i++)
1152 struct variable *v = d->var[i].var;
1153 if (!(exclude_classes & (1u << var_get_dict_class (v))))
1154 proto = caseproto_add_width (proto, var_get_width (v));
1159 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
1160 dict_get_split_cnt() to determine how many SPLIT FILE vars
1161 there are. Returns a null pointer if and only if there are no
1163 const struct variable *const *
1164 dict_get_split_vars (const struct dictionary *d)
1169 /* Returns the number of SPLIT FILE vars. */
1171 dict_get_split_cnt (const struct dictionary *d)
1173 return d->split_cnt;
1176 /* Removes variable V, which must be in D, from D's set of split
1179 dict_unset_split_var (struct dictionary *d, struct variable *v)
1183 assert (dict_contains_var (d, v));
1185 orig_count = d->split_cnt;
1186 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
1187 &v, compare_var_ptrs, NULL);
1188 if (orig_count != d->split_cnt)
1190 if (d->changed) d->changed (d, d->changed_data);
1191 /* We changed the set of split variables so invoke the
1193 if (d->callbacks && d->callbacks->split_changed)
1194 d->callbacks->split_changed (d, d->cb_data);
1198 /* Sets CNT split vars SPLIT in dictionary D. */
1200 dict_set_split_vars (struct dictionary *d,
1201 struct variable *const *split, size_t cnt)
1203 assert (cnt == 0 || split != NULL);
1208 d->split = xnrealloc (d->split, cnt, sizeof *d->split) ;
1209 memcpy (d->split, split, cnt * sizeof *d->split);
1217 if (d->changed) d->changed (d, d->changed_data);
1218 if ( d->callbacks && d->callbacks->split_changed )
1219 d->callbacks->split_changed (d, d->cb_data);
1222 /* Returns the file label for D, or a null pointer if D is
1223 unlabeled (see cmd_file_label()). */
1225 dict_get_label (const struct dictionary *d)
1230 /* Sets D's file label to LABEL, truncating it to a maximum of 60
1233 dict_set_label (struct dictionary *d, const char *label)
1236 d->label = label != NULL ? xstrndup (label, 60) : NULL;
1239 /* Returns the documents for D, or a null pointer if D has no
1240 documents. If the return value is nonnull, then the string
1241 will be an exact multiple of DOC_LINE_LENGTH bytes in length,
1242 with each segment corresponding to one line. */
1244 dict_get_documents (const struct dictionary *d)
1246 return ds_is_empty (&d->documents) ? NULL : ds_cstr (&d->documents);
1249 /* Sets the documents for D to DOCUMENTS, or removes D's
1250 documents if DOCUMENT is a null pointer. If DOCUMENTS is
1251 nonnull, then it should be an exact multiple of
1252 DOC_LINE_LENGTH bytes in length, with each segment
1253 corresponding to one line. */
1255 dict_set_documents (struct dictionary *d, const char *documents)
1259 ds_assign_cstr (&d->documents, documents != NULL ? documents : "");
1261 /* In case the caller didn't get it quite right, pad out the
1262 final line with spaces. */
1263 remainder = ds_length (&d->documents) % DOC_LINE_LENGTH;
1265 ds_put_byte_multiple (&d->documents, ' ', DOC_LINE_LENGTH - remainder);
1268 /* Drops the documents from dictionary D. */
1270 dict_clear_documents (struct dictionary *d)
1272 ds_clear (&d->documents);
1275 /* Appends LINE to the documents in D. LINE will be truncated or
1276 padded on the right with spaces to make it exactly
1277 DOC_LINE_LENGTH bytes long. */
1279 dict_add_document_line (struct dictionary *d, const char *line)
1281 if (strlen (line) > DOC_LINE_LENGTH)
1283 /* Note to translators: "bytes" is correct, not characters */
1284 msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
1286 buf_copy_str_rpad (ds_put_uninit (&d->documents, DOC_LINE_LENGTH),
1287 DOC_LINE_LENGTH, line, ' ');
1290 /* Returns the number of document lines in dictionary D. */
1292 dict_get_document_line_cnt (const struct dictionary *d)
1294 return ds_length (&d->documents) / DOC_LINE_LENGTH;
1297 /* Copies document line number IDX from dictionary D into
1298 LINE, trimming off any trailing white space. */
1300 dict_get_document_line (const struct dictionary *d,
1301 size_t idx, struct string *line)
1303 assert (idx < dict_get_document_line_cnt (d));
1304 ds_assign_substring (line, ds_substr (&d->documents, idx * DOC_LINE_LENGTH,
1306 ds_rtrim (line, ss_cstr (CC_SPACES));
1309 /* Creates in D a vector named NAME that contains the CNT
1310 variables in VAR. Returns true if successful, or false if a
1311 vector named NAME already exists in D. */
1313 dict_create_vector (struct dictionary *d,
1315 struct variable **var, size_t cnt)
1320 for (i = 0; i < cnt; i++)
1321 assert (dict_contains_var (d, var[i]));
1323 if (dict_lookup_vector (d, name) == NULL)
1325 d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector);
1326 d->vector[d->vector_cnt++] = vector_create (name, var, cnt);
1333 /* Creates in D a vector named NAME that contains the CNT
1334 variables in VAR. A vector named NAME must not already exist
1337 dict_create_vector_assert (struct dictionary *d,
1339 struct variable **var, size_t cnt)
1341 assert (dict_lookup_vector (d, name) == NULL);
1342 dict_create_vector (d, name, var, cnt);
1345 /* Returns the vector in D with index IDX, which must be less
1346 than dict_get_vector_cnt (D). */
1347 const struct vector *
1348 dict_get_vector (const struct dictionary *d, size_t idx)
1350 assert (idx < d->vector_cnt);
1352 return d->vector[idx];
1355 /* Returns the number of vectors in D. */
1357 dict_get_vector_cnt (const struct dictionary *d)
1359 return d->vector_cnt;
1362 /* Looks up and returns the vector within D with the given
1364 const struct vector *
1365 dict_lookup_vector (const struct dictionary *d, const char *name)
1368 for (i = 0; i < d->vector_cnt; i++)
1369 if (!strcasecmp (vector_get_name (d->vector[i]), name))
1370 return d->vector[i];
1374 /* Deletes all vectors from D. */
1376 dict_clear_vectors (struct dictionary *d)
1380 for (i = 0; i < d->vector_cnt; i++)
1381 vector_destroy (d->vector[i]);
1388 /* Multiple response sets. */
1390 /* Returns the multiple response set in DICT with index IDX, which must be
1391 between 0 and the count returned by dict_get_n_mrsets(), exclusive. */
1392 const struct mrset *
1393 dict_get_mrset (const struct dictionary *dict, size_t idx)
1395 assert (idx < dict->n_mrsets);
1396 return dict->mrsets[idx];
1399 /* Returns the number of multiple response sets in DICT. */
1401 dict_get_n_mrsets (const struct dictionary *dict)
1403 return dict->n_mrsets;
1406 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1407 returns its index; otherwise, returns SIZE_MAX. */
1409 dict_lookup_mrset_idx (const struct dictionary *dict, const char *name)
1413 for (i = 0; i < dict->n_mrsets; i++)
1414 if (!strcasecmp (name, dict->mrsets[i]->name))
1420 /* Looks for a multiple response set named NAME in DICT. If it finds one,
1421 returns it; otherwise, returns NULL. */
1422 const struct mrset *
1423 dict_lookup_mrset (const struct dictionary *dict, const char *name)
1425 size_t idx = dict_lookup_mrset_idx (dict, name);
1426 return idx != SIZE_MAX ? dict->mrsets[idx] : NULL;
1429 /* Adds MRSET to DICT, replacing any existing set with the same name. Returns
1430 true if a set was replaced, false if none existed with the specified name.
1432 Ownership of MRSET is transferred to DICT. */
1434 dict_add_mrset (struct dictionary *dict, struct mrset *mrset)
1438 assert (mrset_ok (mrset, dict));
1440 idx = dict_lookup_mrset_idx (dict, mrset->name);
1441 if (idx == SIZE_MAX)
1443 dict->mrsets = xrealloc (dict->mrsets,
1444 (dict->n_mrsets + 1) * sizeof *dict->mrsets);
1445 dict->mrsets[dict->n_mrsets++] = mrset;
1450 mrset_destroy (dict->mrsets[idx]);
1451 dict->mrsets[idx] = mrset;
1456 /* Looks for a multiple response set in DICT named NAME. If found, removes it
1457 from DICT and returns true. If none is found, returns false without
1460 Deleting one multiple response set causes the indexes of other sets within
1463 dict_delete_mrset (struct dictionary *dict, const char *name)
1465 size_t idx = dict_lookup_mrset_idx (dict, name);
1466 if (idx != SIZE_MAX)
1468 mrset_destroy (dict->mrsets[idx]);
1469 dict->mrsets[idx] = dict->mrsets[--dict->n_mrsets];
1476 /* Deletes all multiple response sets from DICT. */
1478 dict_clear_mrsets (struct dictionary *dict)
1482 for (i = 0; i < dict->n_mrsets; i++)
1483 mrset_destroy (dict->mrsets[i]);
1484 free (dict->mrsets);
1485 dict->mrsets = NULL;
1489 /* Removes VAR, which must be in DICT, from DICT's multiple response sets. */
1491 dict_unset_mrset_var (struct dictionary *dict, struct variable *var)
1495 assert (dict_contains_var (dict, var));
1497 for (i = 0; i < dict->n_mrsets; )
1499 struct mrset *mrset = dict->mrsets[i];
1502 for (j = 0; j < mrset->n_vars; )
1503 if (mrset->vars[j] == var)
1504 remove_element (mrset->vars, mrset->n_vars--,
1505 sizeof *mrset->vars, j);
1509 if (mrset->n_vars < 2)
1511 mrset_destroy (mrset);
1512 dict->mrsets[i] = dict->mrsets[--dict->n_mrsets];
1519 /* Returns D's attribute set. The caller may examine or modify
1520 the attribute set, but must not destroy it. Destroying D or
1521 calling dict_set_attributes for D will also destroy D's
1524 dict_get_attributes (const struct dictionary *d)
1526 return CONST_CAST (struct attrset *, &d->attributes);
1529 /* Replaces D's attributes set by a copy of ATTRS. */
1531 dict_set_attributes (struct dictionary *d, const struct attrset *attrs)
1533 attrset_destroy (&d->attributes);
1534 attrset_clone (&d->attributes, attrs);
1537 /* Returns true if D has at least one attribute in its attribute
1538 set, false if D's attribute set is empty. */
1540 dict_has_attributes (const struct dictionary *d)
1542 return attrset_count (&d->attributes) > 0;
1545 /* Called from variable.c to notify the dictionary that some property of
1546 the variable has changed */
1548 dict_var_changed (const struct variable *v)
1550 if ( var_has_vardict (v))
1552 const struct vardict_info *vardict = var_get_vardict (v);
1553 struct dictionary *d = vardict->dict;
1558 if (d->changed ) d->changed (d, d->changed_data);
1559 if ( d->callbacks && d->callbacks->var_changed )
1560 d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
1565 /* Called from variable.c to notify the dictionary that the variable's width
1568 dict_var_resized (const struct variable *v, int old_width)
1570 if ( var_has_vardict (v))
1572 const struct vardict_info *vardict = var_get_vardict (v);
1573 struct dictionary *d;
1577 if (d->changed) d->changed (d, d->changed_data);
1579 invalidate_proto (d);
1580 if ( d->callbacks && d->callbacks->var_resized )
1581 d->callbacks->var_resized (d, var_get_dict_index (v), old_width,
1586 /* Called from variable.c to notify the dictionary that the variable's display width
1589 dict_var_display_width_changed (const struct variable *v)
1591 if ( var_has_vardict (v))
1593 const struct vardict_info *vardict = var_get_vardict (v);
1594 struct dictionary *d;
1598 if (d->changed) d->changed (d, d->changed_data);
1599 if ( d->callbacks && d->callbacks->var_display_width_changed )
1600 d->callbacks->var_display_width_changed (d, var_get_dict_index (v), d->cb_data);
1604 /* Dictionary used to contain "internal variables". */
1605 static struct dictionary *internal_dict;
1607 /* Create a variable of the specified WIDTH to be used for internal
1608 calculations only. The variable is assigned case index CASE_IDX. */
1610 dict_create_internal_var (int case_idx, int width)
1612 if (internal_dict == NULL)
1613 internal_dict = dict_create ();
1617 static int counter = INT_MAX / 2;
1618 struct variable *var;
1621 if (++counter == INT_MAX)
1622 counter = INT_MAX / 2;
1624 sprintf (name, "$internal%d", counter);
1625 var = dict_create_var (internal_dict, name, width);
1628 set_var_case_index (var, case_idx);
1634 /* Destroys VAR, which must have been created with
1635 dict_create_internal_var(). */
1637 dict_destroy_internal_var (struct variable *var)
1641 dict_delete_var (internal_dict, var);
1643 /* Destroy internal_dict if it has no variables left, just so that
1644 valgrind --leak-check --show-reachable won't show internal_dict. */
1645 if (dict_get_var_cnt (internal_dict) == 0)
1647 dict_destroy (internal_dict);
1648 internal_dict = NULL;
1654 vardict_get_dict_index (const struct vardict_info *vardict)
1656 return vardict - vardict->dict->var;