1 /* PSPP - computes sample statistics.
2 Copyright (C) 1997-9, 2000 Free Software Foundation, Inc.
3 Written by Ben Pfaff <blp@gnu.org>.
5 This program is free software; you can redistribute it and/or
6 modify it under the terms of the GNU General Public License as
7 published by the Free Software Foundation; either version 2 of the
8 License, or (at your option) any later version.
10 This program is distributed in the hope that it will be useful, but
11 WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 General Public License for more details.
15 You should have received a copy of the GNU General Public License
16 along with this program; if not, write to the Free Software
17 Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
23 #include "algorithm.h"
29 #include "value-labels.h"
35 struct variable **var; /* Variables. */
36 size_t var_cnt, var_cap; /* Number of variables, capacity. */
37 struct hsh_table *name_tab; /* Variable index by name. */
38 int next_value_idx; /* Index of next `union value' to allocate. */
39 struct variable **split; /* SPLIT FILE vars. */
40 size_t split_cnt; /* SPLIT FILE count. */
41 struct variable *weight; /* WEIGHT variable. */
42 struct variable *filter; /* FILTER variable. */
43 int case_limit; /* Current case limit (N command). */
44 char *label; /* File label. */
45 char *documents; /* Documents, as a string. */
46 struct vector **vector; /* Vectors of variables. */
47 size_t vector_cnt; /* Number of vectors. */
50 /* Creates and returns a new dictionary. */
54 struct dictionary *d = xmalloc (sizeof *d);
57 d->var_cnt = d->var_cap = 0;
58 d->name_tab = hsh_create (8, compare_variables, hash_variable, NULL, NULL);
59 d->next_value_idx = 0;
73 /* Creates and returns a (deep) copy of an existing
76 dict_clone (const struct dictionary *s)
84 for (i = 0; i < s->var_cnt; i++)
85 dict_clone_var (d, s->var[i], s->var[i]->name);
86 d->next_value_idx = s->next_value_idx;
88 d->split_cnt = s->split_cnt;
91 d->split = xmalloc (d->split_cnt * sizeof *d->split);
92 for (i = 0; i < d->split_cnt; i++)
93 d->split[i] = dict_lookup_var_assert (d, s->split[i]->name);
96 if (s->weight != NULL)
97 d->weight = dict_lookup_var_assert (d, s->weight->name);
99 if (s->filter != NULL)
100 d->filter = dict_lookup_var_assert (d, s->filter->name);
102 d->case_limit = s->case_limit;
103 dict_set_label (d, dict_get_label (s));
104 dict_set_documents (d, dict_get_documents (s));
106 for (i = 0; i < s->vector_cnt; i++)
107 dict_create_vector (d, s->vector[i]->name,
108 s->vector[i]->var, s->vector[i]->cnt);
113 /* Clears the contents from a dictionary without destroying the
114 dictionary itself. */
116 dict_clear (struct dictionary *d)
118 /* FIXME? Should we really clear case_limit, label, documents?
119 Others are necessarily cleared by deleting all the variables.*/
124 for (i = 0; i < d->var_cnt; i++)
126 struct variable *v = d->var[i];
127 val_labs_destroy (v->val_labs);
133 d->var_cnt = d->var_cap = 0;
134 hsh_clear (d->name_tab);
135 d->next_value_idx = 0;
146 dict_clear_vectors (d);
149 /* Clears a dictionary and destroys it. */
151 dict_destroy (struct dictionary *d)
156 hsh_destroy (d->name_tab);
161 /* Returns the number of variables in D. */
163 dict_get_var_cnt (const struct dictionary *d)
170 /* Returns the variable in D with index IDX, which must be
171 between 0 and the count returned by dict_get_var_cnt(),
174 dict_get_var (const struct dictionary *d, size_t idx)
177 assert (idx < d->var_cnt);
182 /* Sets *VARS to an array of pointers to variables in D and *CNT
183 to the number of variables in *D. By default all variables
184 are returned, but bits may be set in EXCLUDE_CLASSES to
185 exclude ordinary, system, and/or scratch variables. */
187 dict_get_vars (const struct dictionary *d, struct variable ***vars,
188 size_t *cnt, unsigned exclude_classes)
194 assert (vars != NULL);
195 assert (cnt != NULL);
196 assert ((exclude_classes & ~((1u << DC_ORDINARY)
198 | (1u << DC_SCRATCH))) == 0);
201 for (i = 0; i < d->var_cnt; i++)
202 if (!(exclude_classes & (1u << dict_class_from_id (d->var[i]->name))))
205 *vars = xmalloc (count * sizeof **vars);
207 for (i = 0; i < d->var_cnt; i++)
208 if (!(exclude_classes & (1u << dict_class_from_id (d->var[i]->name))))
209 (*vars)[(*cnt)++] = d->var[i];
210 assert (*cnt == count);
213 /* Creates and returns a new variable in D with the given NAME
214 and WIDTH. Returns a null pointer if the given NAME would
215 duplicate that of an existing variable in the dictionary. */
217 dict_create_var (struct dictionary *d, const char *name, int width)
222 assert (name != NULL);
223 assert (strlen (name) >= 1 && strlen (name) <= 8);
224 assert (width >= 0 && width < 256);
226 /* Make sure there's not already a variable by that name. */
227 if (dict_lookup_var (d, name) != NULL)
230 /* Allocate and initialize variable. */
231 v = xmalloc (sizeof *v);
232 strncpy (v->name, name, sizeof v->name);
234 v->index = d->var_cnt;
235 v->type = width == 0 ? NUMERIC : ALPHA;
237 v->fv = d->next_value_idx;
238 v->nv = width == 0 ? 1 : DIV_RND_UP (width, 8);
240 v->reinit = dict_class_from_id (name) != DC_SCRATCH;
241 v->miss_type = MISSING_NONE;
242 if (v->type == NUMERIC)
244 v->print.type = FMT_F;
250 v->print.type = FMT_A;
251 v->print.w = v->width;
255 v->val_labs = val_labs_create (v->width);
258 /* Update dictionary. */
259 if (d->var_cnt >= d->var_cap)
261 d->var_cap = 8 + 2 * d->var_cap;
262 d->var = xrealloc (d->var, d->var_cap * sizeof *d->var);
264 d->var[v->index] = v;
266 hsh_force_insert (d->name_tab, v);
267 d->next_value_idx += v->nv;
272 /* Creates and returns a new variable in D with the given NAME
273 and WIDTH. Assert-fails if the given NAME would duplicate
274 that of an existing variable in the dictionary. */
276 dict_create_var_assert (struct dictionary *d, const char *name, int width)
278 struct variable *v = dict_create_var (d, name, width);
283 /* Creates a new variable in D named NAME, as a copy of existing
284 variable OV, which need not be in D or in any dictionary. */
286 dict_clone_var (struct dictionary *d, const struct variable *ov,
293 assert (name != NULL);
294 assert (strlen (name) >= 1 && strlen (name) <= 8);
296 nv = dict_create_var (d, name, ov->width);
301 nv->reinit = ov->reinit;
302 nv->miss_type = ov->miss_type;
303 memcpy (nv->missing, ov->missing, sizeof nv->missing);
304 nv->print = ov->print;
305 nv->write = ov->write;
306 val_labs_destroy (nv->val_labs);
307 nv->val_labs = val_labs_copy (ov->val_labs);
308 if (ov->label != NULL)
309 nv->label = xstrdup (ov->label);
314 /* Changes the name of V in D to name NEW_NAME. Assert-fails if
315 a variable named NEW_NAME is already in D, except that
316 NEW_NAME may be the same as V's existing name. */
318 dict_rename_var (struct dictionary *d, struct variable *v,
319 const char *new_name)
323 assert (new_name != NULL);
324 assert (strlen (new_name) >= 1 && strlen (new_name) <= 8);
325 assert (dict_contains_var (d, v));
327 if (!strcmp (v->name, new_name))
330 assert (dict_lookup_var (d, new_name) == NULL);
332 hsh_force_delete (d->name_tab, v);
333 strncpy (v->name, new_name, sizeof v->name);
335 hsh_force_insert (d->name_tab, v);
338 /* Returns the variable named NAME in D, or a null pointer if no
339 variable has that name. */
341 dict_lookup_var (const struct dictionary *d, const char *name)
346 assert (name != NULL);
347 assert (strlen (name) >= 1 && strlen (name) <= 8);
349 strncpy (v.name, name, sizeof v.name);
352 return hsh_find (d->name_tab, &v);
355 /* Returns the variable named NAME in D. Assert-fails if no
356 variable has that name. */
358 dict_lookup_var_assert (const struct dictionary *d, const char *name)
360 struct variable *v = dict_lookup_var (d, name);
365 /* Returns nonzero if variable V is in dictionary D. */
367 dict_contains_var (const struct dictionary *d, const struct variable *v)
372 return v->index >= 0 && v->index < d->var_cnt && d->var[v->index] == v;
375 /* Compares two double pointers to variables, which should point
376 to elements of a struct dictionary's `var' member array. */
378 compare_variable_dblptrs (const void *a_, const void *b_, void *aux UNUSED)
380 struct variable *const *a = a_;
381 struct variable *const *b = b_;
391 /* Deletes variable V from dictionary D and frees V.
393 This is a very bad idea if there might be any pointers to V
394 from outside D. In general, no variable in default_dict
395 should be deleted when any transformations are active, because
396 those transformations might reference the deleted variable.
397 The safest time to delete a variable is just after a procedure
398 has been executed, as done by MODIFY VARS.
400 Pointers to V within D are not a problem, because
401 dict_delete_var() knows to remove V from split variables,
402 weights, filters, etc. */
404 dict_delete_var (struct dictionary *d, struct variable *v)
410 assert (dict_contains_var (d, v));
411 assert (d->var[v->index] == v);
413 /* Remove v from splits, weight, filter variables. */
414 d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
416 compare_variable_dblptrs, NULL);
421 dict_clear_vectors (d);
423 /* Remove v from var array. */
425 memmove (d->var + v->index, d->var + v->index + 1,
426 (d->var_cnt - v->index) * sizeof *d->var);
429 for (i = v->index; i < d->var_cnt; i++)
430 d->var[i]->index = i;
432 /* Update name hash. */
433 hsh_force_delete (d->name_tab, v);
436 val_labs_destroy (v->val_labs);
441 /* Deletes the COUNT variables listed in VARS from D. This is
442 unsafe; see the comment on dict_delete_var() for details. */
444 dict_delete_vars (struct dictionary *d,
445 struct variable *const *vars, size_t count)
447 /* FIXME: this can be done in O(count) time, but this algorithm
450 assert (count == 0 || vars != NULL);
453 dict_delete_var (d, *vars++);
456 /* Reorders the variables in D, placing the COUNT variables
457 listed in ORDER in that order at the beginning of D. The
458 other variables in D, if any, retain their relative
461 dict_reorder_vars (struct dictionary *d,
462 struct variable *const *order, size_t count)
464 struct variable **new_var;
468 assert (count == 0 || order != NULL);
469 assert (count <= d->var_cnt);
471 new_var = xmalloc (d->var_cnt * sizeof *new_var);
472 memcpy (new_var, order, count * sizeof *new_var);
473 for (i = 0; i < count; i++)
475 assert (d->var[order[i]->index] != NULL);
476 d->var[order[i]->index] = NULL;
479 for (i = 0; i < d->var_cnt; i++)
480 if (d->var[i] != NULL)
482 assert (count < d->var_cnt);
483 new_var[count] = d->var[i];
484 new_var[count]->index = count;
491 /* Renames COUNT variables specified in VARS to the names given
492 in NEW_NAMES within dictionary D. If the renaming would
493 result in a duplicate variable name, returns zero and stores a
494 name that would be duplicated into *ERR_NAME (if ERR_NAME is
495 non-null). Otherwise, the renaming is successful, and nonzero
498 dict_rename_vars (struct dictionary *d,
499 struct variable **vars, char **new_names,
500 size_t count, char **err_name)
507 assert (count == 0 || vars != NULL);
508 assert (count == 0 || new_names != NULL);
510 old_names = xmalloc (count * sizeof *old_names);
511 for (i = 0; i < count; i++)
513 assert (d->var[vars[i]->index] == vars[i]);
514 hsh_force_delete (d->name_tab, vars[i]);
515 old_names[i] = xstrdup (vars[i]->name);
518 for (i = 0; i < count; i++)
520 assert (new_names[i] != NULL);
521 assert (*new_names[i] != '\0');
522 assert (strlen (new_names[i]) < 9);
523 strcpy (vars[i]->name, new_names[i]);
524 if (hsh_insert (d->name_tab, vars[i]) != NULL)
527 if (err_name != NULL)
528 *err_name = new_names[i];
530 for (i = 0; i < fail_idx; i++)
531 hsh_force_delete (d->name_tab, vars[i]);
533 for (i = 0; i < count; i++)
535 strcpy (vars[i]->name, old_names[i]);
536 hsh_force_insert (d->name_tab, vars[i]);
544 for (i = 0; i < count; i++)
551 /* Returns the weighting variable in dictionary D, or a null
552 pointer if the dictionary is unweighted. */
554 dict_get_weight (const struct dictionary *d)
557 assert (d->weight == NULL || dict_contains_var (d, d->weight));
562 /* Returns the value of D's weighting variable in case C, except that a
563 negative weight is returned as 0. Returns 1 if the dictionary is
564 unweighted. Will warn about missing, negative, or zero values if
565 warn_on_invalid is nonzero. The function will set warn_on_invalid to zero
566 if an invalid weight is found. */
568 dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
569 int *warn_on_invalid)
574 if (d->weight == NULL)
578 double w = case_num (c, d->weight->fv);
579 if ( w < 0.0 || w == SYSMIS || is_num_user_missing(w, d->weight) )
581 if ( w == 0.0 && *warn_on_invalid ) {
582 *warn_on_invalid = 0;
583 msg (SW, _("At least one case in the data file had a weight value "
584 "that was user-missing, system-missing, zero, or "
585 "negative. These case(s) were ignored."));
591 /* Sets the weighting variable of D to V, or turning off
592 weighting if V is a null pointer. */
594 dict_set_weight (struct dictionary *d, struct variable *v)
597 assert (v == NULL || dict_contains_var (d, v));
598 assert (v == NULL || v->type == NUMERIC);
603 /* Returns the filter variable in dictionary D (see cmd_filter())
604 or a null pointer if the dictionary is unfiltered. */
606 dict_get_filter (const struct dictionary *d)
609 assert (d->filter == NULL || dict_contains_var (d, d->filter));
614 /* Sets V as the filter variable for dictionary D. Passing a
615 null pointer for V turn off filtering. */
617 dict_set_filter (struct dictionary *d, struct variable *v)
620 assert (v == NULL || dict_contains_var (d, v));
625 /* Returns the case limit for dictionary D, or zero if the number
626 of cases is unlimited (see cmd_n()). */
628 dict_get_case_limit (const struct dictionary *d)
632 return d->case_limit;
635 /* Sets CASE_LIMIT as the case limit for dictionary D. Zero for
636 CASE_LIMIT indicates no limit. */
638 dict_set_case_limit (struct dictionary *d, int case_limit)
641 assert (case_limit >= 0);
643 d->case_limit = case_limit;
646 /* Returns the index of the next value to be added to D. This
647 value is the number of `union value's that need to be
648 allocated to store a case for dictionary D. */
650 dict_get_next_value_idx (const struct dictionary *d)
654 return d->next_value_idx;
657 /* Returns the number of bytes needed to store a case for
660 dict_get_case_size (const struct dictionary *d)
664 return sizeof (union value) * dict_get_next_value_idx (d);
667 /* Deletes scratch variables in dictionary D and reassigns values
668 so that fragmentation is eliminated. */
670 dict_compact_values (struct dictionary *d)
674 d->next_value_idx = 0;
675 for (i = 0; i < d->var_cnt; )
677 struct variable *v = d->var[i];
679 if (dict_class_from_id (v->name) != DC_SCRATCH)
681 v->fv = d->next_value_idx;
682 d->next_value_idx += v->nv;
686 dict_delete_var (default_dict, v);
690 /* Returns the number of values that would be used by a case if
691 dict_compact_values() were called. */
693 dict_get_compacted_value_cnt (const struct dictionary *d)
699 for (i = 0; i < d->var_cnt; i++)
700 if (dict_class_from_id (d->var[i]->name) != DC_SCRATCH)
701 cnt += d->var[i]->nv;
705 /* Creates and returns an array mapping from a dictionary index
706 to the `fv' that the corresponding variable will have after
707 calling dict_compact_values(). Scratch variables receive -1
708 for `fv' because dict_compact_values() will delete them. */
710 dict_get_compacted_idx_to_fv (const struct dictionary *d)
713 size_t next_value_idx;
716 idx_to_fv = xmalloc (d->var_cnt * sizeof *idx_to_fv);
718 for (i = 0; i < d->var_cnt; i++)
720 struct variable *v = d->var[i];
722 if (dict_class_from_id (v->name) != DC_SCRATCH)
724 idx_to_fv[i] = next_value_idx;
725 next_value_idx += v->nv;
733 /* Returns the SPLIT FILE vars (see cmd_split_file()). Call
734 dict_get_split_cnt() to determine how many SPLIT FILE vars
735 there are. Returns a null pointer if and only if there are no
737 struct variable *const *
738 dict_get_split_vars (const struct dictionary *d)
745 /* Returns the number of SPLIT FILE vars. */
747 dict_get_split_cnt (const struct dictionary *d)
754 /* Sets CNT split vars SPLIT in dictionary D. */
756 dict_set_split_vars (struct dictionary *d,
757 struct variable *const *split, size_t cnt)
760 assert (cnt == 0 || split != NULL);
763 d->split = xrealloc (d->split, cnt * sizeof *d->split);
764 memcpy (d->split, split, cnt * sizeof *d->split);
767 /* Returns the file label for D, or a null pointer if D is
768 unlabeled (see cmd_file_label()). */
770 dict_get_label (const struct dictionary *d)
777 /* Sets D's file label to LABEL, truncating it to a maximum of 60
780 dict_set_label (struct dictionary *d, const char *label)
787 else if (strlen (label) < 60)
788 d->label = xstrdup (label);
791 d->label = xmalloc (61);
792 memcpy (d->label, label, 60);
797 /* Returns the documents for D, or a null pointer if D has no
798 documents (see cmd_document()).. */
800 dict_get_documents (const struct dictionary *d)
807 /* Sets the documents for D to DOCUMENTS, or removes D's
808 documents if DOCUMENT is a null pointer. */
810 dict_set_documents (struct dictionary *d, const char *documents)
815 if (documents == NULL)
818 d->documents = xstrdup (documents);
821 /* Creates in D a vector named NAME that contains CNT variables
822 VAR (see cmd_vector()). Returns nonzero if successful, or
823 zero if a vector named NAME already exists in D. */
825 dict_create_vector (struct dictionary *d,
827 struct variable **var, size_t cnt)
829 struct vector *vector;
832 assert (name != NULL);
833 assert (strlen (name) > 0 && strlen (name) < 9);
834 assert (var != NULL);
837 if (dict_lookup_vector (d, name) != NULL)
840 d->vector = xrealloc (d->vector, (d->vector_cnt + 1) * sizeof *d->vector);
841 vector = d->vector[d->vector_cnt] = xmalloc (sizeof *vector);
842 vector->idx = d->vector_cnt++;
843 strncpy (vector->name, name, 8);
844 vector->name[8] = '\0';
845 vector->var = xmalloc (cnt * sizeof *var);
846 memcpy (vector->var, var, cnt * sizeof *var);
852 /* Returns the vector in D with index IDX, which must be less
853 than dict_get_vector_cnt (D). */
854 const struct vector *
855 dict_get_vector (const struct dictionary *d, size_t idx)
858 assert (idx < d->vector_cnt);
860 return d->vector[idx];
863 /* Returns the number of vectors in D. */
865 dict_get_vector_cnt (const struct dictionary *d)
869 return d->vector_cnt;
872 /* Looks up and returns the vector within D with the given
874 const struct vector *
875 dict_lookup_vector (const struct dictionary *d, const char *name)
880 assert (name != NULL);
882 for (i = 0; i < d->vector_cnt; i++)
883 if (!strcmp (d->vector[i]->name, name))
888 /* Deletes all vectors from D. */
890 dict_clear_vectors (struct dictionary *d)
896 for (i = 0; i < d->vector_cnt; i++)
898 free (d->vector[i]->var);