1 /* PSPP - a program for statistical analysis.
2 Copyright (C) 1997-9, 2000, 2006 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/>. */
26 #include "dictionary.h"
27 #include "identifier.h"
28 #include "missing-values.h"
29 #include "value-labels.h"
32 #include <libpspp/alloc.h>
33 #include <libpspp/assertion.h>
34 #include <libpspp/compiler.h>
35 #include <libpspp/hash.h>
36 #include <libpspp/message.h>
37 #include <libpspp/str.h>
40 #define _(msgid) gettext (msgid)
45 /* Dictionary information. */
46 char name[LONG_NAME_LEN + 1]; /* Variable name. Mixed case. */
47 int width; /* 0 for numeric, otherwise string width. */
48 struct missing_values miss; /* Missing values. */
49 struct fmt_spec print; /* Default format for PRINT. */
50 struct fmt_spec write; /* Default format for WRITE. */
51 struct val_labs *val_labs; /* Value labels. */
52 char *label; /* Variable label. */
54 /* GUI information. */
55 enum measure measure; /* Nominal, ordinal, or continuous. */
56 int display_width; /* Width of data editor column. */
57 enum alignment alignment; /* Alignment of data in GUI. */
59 /* Case information. */
60 bool leave; /* Leave value from case to case? */
62 /* Data for use by containing dictionary. */
63 struct vardict_info vardict;
65 /* Short name, used only for system and portable file input
66 and output. Upper case only. Short names are not necessarily
67 unique. Any variable may have no short name, indicated by an
69 char short_name[SHORT_NAME_LEN + 1];
71 /* Each command may use these fields as needed. */
73 void (*aux_dtor) (struct variable *);
75 /* Values of a categorical variable. Procedures need
76 vectors with binary entries, so any variable of type ALPHA will
77 have its values stored here. */
78 struct cat_vals *obs_vals;
81 /* Returns true if VAR_TYPE is a valid variable type. */
83 var_type_is_valid (enum var_type var_type)
85 return var_type == VAR_NUMERIC || var_type == VAR_STRING;
88 /* Returns the variable type for the given width. */
90 var_type_from_width (int width)
92 return width != 0 ? VAR_STRING : VAR_NUMERIC;
95 /* Creates and returns a new variable with the given NAME and
96 WIDTH and other fields initialized to default values. The
97 variable is not added to a dictionary; for that, use
98 dict_create_var instead. */
100 var_create (const char *name, int width)
104 assert (width >= 0 && width <= MAX_STRING);
106 v = xmalloc (sizeof *v);
107 v->vardict.dict_index = v->vardict.case_index = -1;
108 var_set_name (v, name);
110 mv_init (&v->miss, width);
111 v->leave = var_must_leave (v);
112 if (var_is_numeric (v))
114 v->print = fmt_for_output (FMT_F, 8, 2);
115 v->alignment = ALIGN_RIGHT;
116 v->display_width = 8;
117 v->measure = MEASURE_SCALE;
121 v->print = fmt_for_output (FMT_A, var_get_width (v), 0);
122 v->alignment = ALIGN_LEFT;
123 v->display_width = 8;
124 v->measure = MEASURE_NOMINAL;
129 var_clear_short_name (v);
137 /* Creates and returns a clone of OLD_VAR. Most properties of
138 the new variable are copied from OLD_VAR, except:
140 - The variable's short name is not copied, because there is
141 no reason to give a new variable with potentially a new
142 name the same short name.
144 - The new variable is not added to OLD_VAR's dictionary by
145 default. Use dict_clone_var, instead, to do that.
147 - Auxiliary data and obs_vals are not copied. */
149 var_clone (const struct variable *old_var)
151 struct variable *new_var = var_create (var_get_name (old_var),
152 var_get_width (old_var));
154 var_set_missing_values (new_var, var_get_missing_values (old_var));
155 var_set_print_format (new_var, var_get_print_format (old_var));
156 var_set_write_format (new_var, var_get_write_format (old_var));
157 var_set_value_labels (new_var, var_get_value_labels (old_var));
158 var_set_label (new_var, var_get_label (old_var));
159 var_set_measure (new_var, var_get_measure (old_var));
160 var_set_display_width (new_var, var_get_display_width (old_var));
161 var_set_alignment (new_var, var_get_alignment (old_var));
162 var_set_leave (new_var, var_get_leave (old_var));
167 /* Destroys variable V.
168 V must not belong to a dictionary. If it does, use
169 dict_delete_var instead. */
171 var_destroy (struct variable *v)
175 assert (!var_has_vardict (v));
176 cat_stored_values_destroy (v->obs_vals);
178 val_labs_destroy (v->val_labs);
184 /* Variable names. */
186 /* Return variable V's name. */
188 var_get_name (const struct variable *v)
193 /* Sets V's name to NAME.
194 Do not use this function for a variable in a dictionary. Use
195 dict_rename_var instead. */
197 var_set_name (struct variable *v, const char *name)
199 assert (v->vardict.dict_index == -1);
200 assert (var_is_plausible_name (name, false));
202 str_copy_trunc (v->name, sizeof v->name, name);
203 dict_var_changed (v);
206 /* Returns true if NAME is an acceptable name for a variable,
207 false otherwise. If ISSUE_ERROR is true, issues an
208 explanatory error message on failure. */
210 var_is_valid_name (const char *name, bool issue_error)
215 assert (name != NULL);
217 /* Note that strlen returns number of BYTES, not the number of
219 length = strlen (name);
221 plausible = var_is_plausible_name(name, issue_error);
227 if (!lex_is_id1 (name[0]))
230 msg (SE, _("Character `%c' (in %s) may not appear "
231 "as the first character in a variable name."),
237 for (i = 0; i < length; i++)
239 if (!lex_is_idn (name[i]))
242 msg (SE, _("Character `%c' (in %s) may not appear in "
252 /* Returns true if NAME is an plausible name for a variable,
253 false otherwise. If ISSUE_ERROR is true, issues an
254 explanatory error message on failure.
255 This function makes no use of LC_CTYPE.
258 var_is_plausible_name (const char *name, bool issue_error)
262 assert (name != NULL);
264 /* Note that strlen returns number of BYTES, not the number of
266 length = strlen (name);
270 msg (SE, _("Variable name cannot be empty string."));
273 else if (length > LONG_NAME_LEN)
276 msg (SE, _("Variable name %s exceeds %d-character limit."),
277 name, (int) LONG_NAME_LEN);
281 if (lex_id_to_token (ss_cstr (name)) != T_ID)
284 msg (SE, _("`%s' may not be used as a variable name because it "
285 "is a reserved word."), name);
292 /* A hsh_compare_func that orders variables A and B by their
295 compare_vars_by_name (const void *a_, const void *b_, const void *aux UNUSED)
297 const struct variable *a = a_;
298 const struct variable *b = b_;
300 return strcasecmp (a->name, b->name);
303 /* A hsh_hash_func that hashes variable V based on its name. */
305 hash_var_by_name (const void *v_, const void *aux UNUSED)
307 const struct variable *v = v_;
309 return hsh_hash_case_string (v->name);
312 /* A hsh_compare_func that orders pointers to variables A and B
315 compare_var_ptrs_by_name (const void *a_, const void *b_,
316 const void *aux UNUSED)
318 struct variable *const *a = a_;
319 struct variable *const *b = b_;
321 return strcasecmp (var_get_name (*a), var_get_name (*b));
324 /* A hsh_hash_func that hashes pointer to variable V based on its
327 hash_var_ptr_by_name (const void *v_, const void *aux UNUSED)
329 struct variable *const *v = v_;
331 return hsh_hash_case_string (var_get_name (*v));
334 /* Returns the type of variable V. */
336 var_get_type (const struct variable *v)
338 return var_type_from_width (v->width);
341 /* Returns the width of variable V. */
343 var_get_width (const struct variable *v)
348 /* Sets the width of V to WIDTH. */
350 var_set_width (struct variable *v, int new_width)
352 enum var_type new_type = var_type_from_width (new_width);
354 if (mv_is_resizable (&v->miss, new_width))
355 mv_resize (&v->miss, new_width);
357 mv_init (&v->miss, new_width);
359 if (v->val_labs != NULL)
361 if (val_labs_can_set_width (v->val_labs, new_width))
362 val_labs_set_width (v->val_labs, new_width);
365 val_labs_destroy (v->val_labs);
370 if (var_get_type (v) != new_type)
372 v->print = (new_type == VAR_NUMERIC
373 ? fmt_for_output (FMT_F, 8, 2)
374 : fmt_for_output (FMT_A, new_width, 0));
377 else if (new_type == VAR_STRING)
379 v->print.w = v->print.type == FMT_AHEX ? new_width * 2 : new_width;
380 v->write.w = v->write.type == FMT_AHEX ? new_width * 2 : new_width;
383 v->width = new_width;
385 dict_var_changed (v);
388 /* Returns true if variable V is numeric, false otherwise. */
390 var_is_numeric (const struct variable *v)
392 return var_get_type (v) == VAR_NUMERIC;
395 /* Returns true if variable V is a string variable, false
398 var_is_alpha (const struct variable *v)
400 return var_get_type (v) == VAR_STRING;
403 /* Returns true if variable V is a short string variable, false
406 var_is_short_string (const struct variable *v)
408 return v->width > 0 && v->width <= MAX_SHORT_STRING;
411 /* Returns true if variable V is a long string variable, false
414 var_is_long_string (const struct variable *v)
416 return v->width > MAX_SHORT_STRING;
419 /* Returns the number of "union value"s need to store a value of
422 var_get_value_cnt (const struct variable *v)
424 return value_cnt_from_width (v->width);
427 /* Returns variable V's missing values. */
428 const struct missing_values *
429 var_get_missing_values (const struct variable *v)
434 /* Sets variable V's missing values to MISS, which must be of V's
435 width or at least resizable to V's width.
436 If MISS is null, then V's missing values, if any, are
439 var_set_missing_values (struct variable *v, const struct missing_values *miss)
443 assert (mv_is_resizable (miss, v->width));
444 mv_copy (&v->miss, miss);
445 mv_resize (&v->miss, v->width);
448 mv_init (&v->miss, v->width);
450 dict_var_changed (v);
453 /* Sets variable V to have no user-missing values. */
455 var_clear_missing_values (struct variable *v)
457 var_set_missing_values (v, NULL);
460 /* Returns true if V has any user-missing values,
463 var_has_missing_values (const struct variable *v)
465 return !mv_is_empty (&v->miss);
468 /* Returns true if VALUE is in the given CLASS of missing values
469 in V, false otherwise. */
471 var_is_value_missing (const struct variable *v, const union value *value,
474 return mv_is_value_missing (&v->miss, value, class);
477 /* Returns true if D is in the given CLASS of missing values in
479 V must be a numeric variable. */
481 var_is_num_missing (const struct variable *v, double d, enum mv_class class)
483 return mv_is_num_missing (&v->miss, d, class);
486 /* Returns true if S[] is a missing value for V, false otherwise.
487 S[] must contain exactly as many characters as V's width.
488 V must be a string variable. */
490 var_is_str_missing (const struct variable *v, const char s[],
493 return mv_is_str_missing (&v->miss, s, class);
496 /* Returns variable V's value labels,
497 possibly a null pointer if it has none. */
498 const struct val_labs *
499 var_get_value_labels (const struct variable *v)
504 /* Returns true if variable V has at least one value label. */
506 var_has_value_labels (const struct variable *v)
508 return val_labs_count (v->val_labs) > 0;
511 /* Sets variable V's value labels to a copy of VLS,
512 which must have a width equal to V's width or one that can be
513 changed to V's width.
514 If VLS is null, then V's value labels, if any, are removed. */
516 var_set_value_labels (struct variable *v, const struct val_labs *vls)
518 val_labs_destroy (v->val_labs);
523 assert (val_labs_can_set_width (vls, v->width));
524 v->val_labs = val_labs_copy (vls);
525 val_labs_set_width (v->val_labs, v->width);
526 dict_var_changed (v);
530 /* Makes sure that V has a set of value labels,
531 by assigning one to it if necessary. */
533 alloc_value_labels (struct variable *v)
535 assert (!var_is_long_string (v));
536 if (v->val_labs == NULL)
537 v->val_labs = val_labs_create (v->width);
540 /* Attempts to add a value label with the given VALUE and LABEL
541 to V. Returns true if successful, false if VALUE has an
543 V must not be a long string variable. */
545 var_add_value_label (struct variable *v,
546 const union value *value, const char *label)
548 alloc_value_labels (v);
549 return val_labs_add (v->val_labs, *value, label);
552 /* Adds or replaces a value label with the given VALUE and LABEL
554 V must not be a long string variable. */
556 var_replace_value_label (struct variable *v,
557 const union value *value, const char *label)
559 alloc_value_labels (v);
560 val_labs_replace (v->val_labs, *value, label);
563 /* Removes V's value labels, if any. */
565 var_clear_value_labels (struct variable *v)
567 var_set_value_labels (v, NULL);
570 /* Returns the label associated with VALUE for variable V,
571 or a null pointer if none. */
573 var_lookup_value_label (const struct variable *v, const union value *value)
575 return val_labs_find (v->val_labs, *value);
578 /* Get a string representing VALUE for variable V.
579 That is, if VALUE has a label, return that label,
580 otherwise format VALUE and return the formatted string. */
582 var_get_value_name (const struct variable *v, const union value *value)
584 const char *name = var_lookup_value_label (v, value);
587 static char buf[MAX_STRING + 1];
588 data_out (value, &v->print, buf);
589 buf[v->print.w] = '\0';
595 /* Print and write formats. */
597 /* Returns V's print format specification. */
598 const struct fmt_spec *
599 var_get_print_format (const struct variable *v)
604 /* Sets V's print format specification to PRINT, which must be a
605 valid format specification for outputting a variable of V's
608 var_set_print_format (struct variable *v, const struct fmt_spec *print)
610 assert (fmt_check_width_compat (print, v->width));
612 dict_var_changed (v);
615 /* Returns V's write format specification. */
616 const struct fmt_spec *
617 var_get_write_format (const struct variable *v)
622 /* Sets V's write format specification to WRITE, which must be a
623 valid format specification for outputting a variable of V's
626 var_set_write_format (struct variable *v, const struct fmt_spec *write)
628 assert (fmt_check_width_compat (write, v->width));
630 dict_var_changed (v);
633 /* Sets V's print and write format specifications to FORMAT,
634 which must be a valid format specification for outputting a
635 variable of V's width. */
637 var_set_both_formats (struct variable *v, const struct fmt_spec *format)
639 var_set_print_format (v, format);
640 var_set_write_format (v, format);
643 /* Return a string representing this variable, in the form most
644 appropriate from a human factors perspective, that is, its
645 variable label if it has one, otherwise its name. */
647 var_to_string (const struct variable *v)
649 return v->label != NULL ? v->label : v->name;
652 /* Returns V's variable label, or a null pointer if it has none. */
654 var_get_label (const struct variable *v)
659 /* Sets V's variable label to LABEL, stripping off leading and
660 trailing white space and truncating to 255 characters.
661 If LABEL is a null pointer or if LABEL is an empty string
662 (after stripping white space), then V's variable label (if
665 var_set_label (struct variable *v, const char *label)
672 struct substring s = ss_cstr (label);
673 ss_trim (&s, ss_cstr (CC_SPACES));
674 ss_truncate (&s, 255);
675 if (!ss_is_empty (s))
676 v->label = ss_xstrdup (s);
677 dict_var_changed (v);
681 /* Removes any variable label from V. */
683 var_clear_label (struct variable *v)
685 var_set_label (v, NULL);
688 /* Returns true if V has a variable V,
691 var_has_label (const struct variable *v)
693 return v->label != NULL;
696 /* Returns true if M is a valid variable measurement level,
699 measure_is_valid (enum measure m)
701 return m == MEASURE_NOMINAL || m == MEASURE_ORDINAL || m == MEASURE_SCALE;
704 /* Returns V's measurement level. */
706 var_get_measure (const struct variable *v)
711 /* Sets V's measurement level to MEASURE. */
713 var_set_measure (struct variable *v, enum measure measure)
715 assert (measure_is_valid (measure));
716 v->measure = measure;
717 dict_var_changed (v);
720 /* Returns V's display width, which applies only to GUIs. */
722 var_get_display_width (const struct variable *v)
724 return v->display_width;
730 /* Sets V's display width to DISPLAY_WIDTH. */
732 var_set_display_width (struct variable *v, int display_width)
734 v->display_width = display_width;
735 dict_var_changed (v);
738 /* Returns true if A is a valid alignment,
741 alignment_is_valid (enum alignment a)
743 return a == ALIGN_LEFT || a == ALIGN_RIGHT || a == ALIGN_CENTRE;
746 /* Returns V's display alignment, which applies only to GUIs. */
748 var_get_alignment (const struct variable *v)
753 /* Sets V's display alignment to ALIGNMENT. */
755 var_set_alignment (struct variable *v, enum alignment alignment)
757 assert (alignment_is_valid (alignment));
758 v->alignment = alignment;
759 dict_var_changed (v);
762 /* Whether variables' values should be preserved from case to
765 /* Returns true if variable V's value should be left from case to
766 case, instead of being reset to 0, system-missing, or blanks. */
768 var_get_leave (const struct variable *v)
773 /* Sets V's leave setting to LEAVE. */
775 var_set_leave (struct variable *v, bool leave)
777 assert (leave || !var_must_leave (v));
779 dict_var_changed (v);
782 /* Returns true if V must be left from case to case,
783 false if it can be set either way. */
785 var_must_leave (const struct variable *v)
787 return dict_class_from_id (v->name) == DC_SCRATCH;
790 /* Returns V's short name, if it has one, or a null pointer
793 Short names are used only for system and portable file input
794 and output. They are upper-case only, not necessarily unique,
795 and limited to SHORT_NAME_LEN characters (plus a null
796 terminator). Any variable may have no short name, indicated
797 by returning a null pointer. */
799 var_get_short_name (const struct variable *v)
801 return v->short_name[0] != '\0' ? v->short_name : NULL;
804 /* Sets V's short_name to SHORT_NAME, truncating it to
805 SHORT_NAME_LEN characters and converting it to uppercase in
806 the process. Specifying a null pointer for SHORT_NAME clears
807 the variable's short name. */
809 var_set_short_name (struct variable *v, const char *short_name)
812 assert (short_name == NULL || var_is_plausible_name (short_name, false));
814 if (short_name != NULL)
816 str_copy_trunc (v->short_name, sizeof v->short_name, short_name);
817 str_uppercase (v->short_name);
820 v->short_name[0] = '\0';
821 dict_var_changed (v);
824 /* Clears V's short name. */
826 var_clear_short_name (struct variable *v)
830 v->short_name[0] = '\0';
833 /* Relationship with dictionary. */
835 /* Returns V's index within its dictionary, the value
836 for which "dict_get_var (dict, index)" will return V.
837 V must be in a dictionary. */
839 var_get_dict_index (const struct variable *v)
841 assert (v->vardict.dict_index != -1);
842 return v->vardict.dict_index;
845 /* Returns V's index within the case represented by its
846 dictionary, that is, the value for which "case_data_idx (case,
847 index)" will return the data for V in that case.
848 V must be in a dictionary. */
850 var_get_case_index (const struct variable *v)
852 assert (v->vardict.case_index != -1);
853 return v->vardict.case_index;
856 /* Returns V's auxiliary data, or a null pointer if none has been
859 var_get_aux (const struct variable *v)
864 /* Assign auxiliary data AUX to variable V, which must not
865 already have auxiliary data. Before V's auxiliary data is
866 cleared, AUX_DTOR(V) will be called. (var_dtor_free, below,
867 may be appropriate for use as AUX_DTOR.) */
869 var_attach_aux (const struct variable *v_,
870 void *aux, void (*aux_dtor) (struct variable *))
872 struct variable *v = (struct variable *) v_ ; /* cast away const */
873 assert (v->aux == NULL);
874 assert (aux != NULL);
876 v->aux_dtor = aux_dtor;
880 /* Remove auxiliary data, if any, from V, and return it, without
881 calling any associated destructor. */
883 var_detach_aux (struct variable *v)
886 assert (aux != NULL);
891 /* Clears auxiliary data, if any, from V, and calls any
892 associated destructor. */
894 var_clear_aux (struct variable *v)
899 if (v->aux_dtor != NULL)
905 /* This function is appropriate for use an auxiliary data
906 destructor (passed as AUX_DTOR to var_attach_aux()) for the
907 case where the auxiliary data should be passed to free(). */
909 var_dtor_free (struct variable *v)
914 /* Observed categorical values. */
916 /* Returns V's observed categorical values,
917 which V must have. */
919 var_get_obs_vals (const struct variable *v)
921 assert (v->obs_vals != NULL);
925 /* Sets V's observed categorical values to CAT_VALS. */
927 var_set_obs_vals (const struct variable *v_, struct cat_vals *cat_vals)
929 struct variable *v = (struct variable *) v_ ; /* cast away const */
930 cat_stored_values_destroy (v->obs_vals);
931 v->obs_vals = cat_vals;
934 /* Returns true if V has observed categorical values,
937 var_has_obs_vals (const struct variable *v)
939 return v->obs_vals != NULL;
942 /* Returns the dictionary class corresponding to a variable named
945 dict_class_from_id (const char *name)
958 /* Returns the name of dictionary class DICT_CLASS. */
960 dict_class_to_name (enum dict_class dict_class)
965 return _("ordinary");
975 /* Returns V's vardict structure. */
976 const struct vardict_info *
977 var_get_vardict (const struct variable *v)
979 assert (var_has_vardict (v));
983 /* Sets V's vardict data to VARDICT. */
985 var_set_vardict (struct variable *v, const struct vardict_info *vardict)
987 assert (vardict->dict_index >= 0);
988 assert (vardict->case_index >= 0);
989 v->vardict = *vardict;
992 /* Returns true if V has vardict data. */
994 var_has_vardict (const struct variable *v)
996 return v->vardict.dict_index != -1;
999 /* Clears V's vardict data. */
1001 var_clear_vardict (struct variable *v)
1003 v->vardict.dict_index = v->vardict.case_index = -1;