/* PSPP - a program for statistical analysis.
- Copyright (C) 1997-9, 2000, 2006 Free Software Foundation, Inc.
+ Copyright (C) 1997-9, 2000, 2006, 2009 Free Software Foundation, Inc.
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
#include <stdlib.h>
-
-#include "category.h"
-#include "data-out.h"
-#include "format.h"
-#include "dictionary.h"
-#include "identifier.h"
-#include "missing-values.h"
-#include "value-labels.h"
-#include "vardict.h"
+#include <data/attributes.h>
+#include <data/category.h>
+#include <data/data-out.h>
+#include <data/format.h>
+#include <data/dictionary.h>
+#include <data/identifier.h>
+#include <data/missing-values.h>
+#include <data/value-labels.h>
+#include <data/vardict.h>
#include <libpspp/misc.h>
-#include <libpspp/alloc.h>
#include <libpspp/assertion.h>
#include <libpspp/compiler.h>
#include <libpspp/hash.h>
#include <libpspp/message.h>
#include <libpspp/str.h>
+#include "minmax.h"
+#include "xalloc.h"
+
#include "gettext.h"
#define _(msgid) gettext (msgid)
struct variable
{
/* Dictionary information. */
- char name[LONG_NAME_LEN + 1]; /* Variable name. Mixed case. */
+ char name[VAR_NAME_LEN + 1]; /* Variable name. Mixed case. */
int width; /* 0 for numeric, otherwise string width. */
struct missing_values miss; /* Missing values. */
struct fmt_spec print; /* Default format for PRINT. */
/* Data for use by containing dictionary. */
struct vardict_info vardict;
- /* Short name, used only for system and portable file input
- and output. Upper case only. Short names are not necessarily
- unique. Any variable may have no short name, indicated by an
- empty string. */
- char short_name[SHORT_NAME_LEN + 1];
+ /* Used only for system and portable file input and output.
+ See short-names.h. */
+ char **short_names;
+ size_t short_name_cnt;
/* Each command may use these fields as needed. */
void *aux;
vectors with binary entries, so any variable of type ALPHA will
have its values stored here. */
struct cat_vals *obs_vals;
- };
-
-/* Returns true if VAR_TYPE is a valid variable type. */
-bool
-var_type_is_valid (enum var_type var_type)
-{
- return var_type == VAR_NUMERIC || var_type == VAR_STRING;
-}
-/* Returns the variable type for the given width. */
-enum var_type
-var_type_from_width (int width)
-{
- return width != 0 ? VAR_STRING : VAR_NUMERIC;
-}
+ /* Custom attributes. */
+ struct attrset attributes;
+ };
\f
/* Creates and returns a new variable with the given NAME and
WIDTH and other fields initialized to default values. The
var_create (const char *name, int width)
{
struct variable *v;
+ enum val_type type;
assert (width >= 0 && width <= MAX_STRING);
v->width = width;
mv_init (&v->miss, width);
v->leave = var_must_leave (v);
- if (var_is_numeric (v))
- {
- v->print = fmt_for_output (FMT_F, 8, 2);
- v->alignment = ALIGN_RIGHT;
- v->display_width = 8;
- v->measure = MEASURE_SCALE;
- }
- else
- {
- v->print = fmt_for_output (FMT_A, var_get_width (v), 0);
- v->alignment = ALIGN_LEFT;
- v->display_width = 8;
- v->measure = MEASURE_NOMINAL;
- }
- v->write = v->print;
+ type = val_type_from_width (width);
+ v->alignment = var_default_alignment (type);
+ v->measure = var_default_measure (type);
+ v->display_width = var_default_display_width (width);
+ v->print = v->write = var_default_formats (width);
v->val_labs = NULL;
v->label = NULL;
- var_clear_short_name (v);
+ v->short_names = NULL;
+ v->short_name_cnt = 0;
v->aux = NULL;
v->aux_dtor = NULL;
v->obs_vals = NULL;
+ attrset_init (&v->attributes);
return v;
}
var_set_display_width (new_var, var_get_display_width (old_var));
var_set_alignment (new_var, var_get_alignment (old_var));
var_set_leave (new_var, var_get_leave (old_var));
+ var_set_attributes (new_var, var_get_attributes (old_var));
return new_var;
}
+/* Create a variable of the specified WIDTH to be used for
+ internal calculations only. The variable is assigned a unique
+ dictionary index and a case index of CASE_IDX. */
+struct variable *
+var_create_internal (int case_idx, int width)
+{
+ struct variable *v = var_create ("$internal", width);
+ struct vardict_info vdi;
+ static int counter = INT_MAX / 2;
+
+ vdi.dict = NULL;
+ vdi.case_index = case_idx;
+ vdi.dict_index = counter++;
+ if (counter == INT_MAX)
+ counter = INT_MAX / 2;
+
+ var_set_vardict (v, &vdi);
+
+ return v;
+}
+
/* Destroys variable V.
V must not belong to a dictionary. If it does, use
dict_delete_var instead. */
{
if (v != NULL)
{
- assert (!var_has_vardict (v));
+ if (var_has_vardict (v))
+ {
+ const struct vardict_info *vdi = var_get_vardict (v);
+ assert (vdi->dict == NULL);
+ }
+ mv_destroy (&v->miss);
cat_stored_values_destroy (v->obs_vals);
+ var_clear_short_names (v);
var_clear_aux (v);
val_labs_destroy (v->val_labs);
var_clear_label (v);
msg (SE, _("Variable name cannot be empty string."));
return false;
}
- else if (length > LONG_NAME_LEN)
+ else if (length > VAR_NAME_LEN)
{
if (issue_error)
msg (SE, _("Variable name %s exceeds %d-character limit."),
- name, (int) LONG_NAME_LEN);
+ name, (int) VAR_NAME_LEN);
return false;
}
return true;
}
+/* Returns VAR's dictionary class. */
+enum dict_class
+var_get_dict_class (const struct variable *var)
+{
+ return dict_class_from_id (var->name);
+}
+
/* A hsh_compare_func that orders variables A and B by their
names. */
int
{
const struct variable *v = v_;
- return hsh_hash_case_string (v->name);
+ return hash_case_string (v->name, 0);
}
/* A hsh_compare_func that orders pointers to variables A and B
return strcasecmp (var_get_name (*a), var_get_name (*b));
}
+/* A hsh_compare_func that orders pointers to variables A and B
+ by their dictionary indexes. */
+int
+compare_var_ptrs_by_dict_index (const void *a_, const void *b_,
+ const void *aux UNUSED)
+{
+ struct variable *const *a = a_;
+ struct variable *const *b = b_;
+ size_t a_index = var_get_dict_index (*a);
+ size_t b_index = var_get_dict_index (*b);
+
+ return a_index < b_index ? -1 : a_index > b_index;
+}
+
/* A hsh_hash_func that hashes pointer to variable V based on its
name. */
unsigned
{
struct variable *const *v = v_;
- return hsh_hash_case_string (var_get_name (*v));
+ return hash_case_string (var_get_name (*v), 0);
}
\f
/* Returns the type of variable V. */
-enum var_type
+enum val_type
var_get_type (const struct variable *v)
{
- return var_type_from_width (v->width);
+ return val_type_from_width (v->width);
}
/* Returns the width of variable V. */
return v->width;
}
-/* Sets the width of V to WIDTH. */
+/* Changes the width of V to NEW_WIDTH.
+ This function should be used cautiously. */
void
var_set_width (struct variable *v, int new_width)
{
const int old_width = v->width;
- enum var_type new_type = var_type_from_width (new_width);
+
+ if (old_width == new_width)
+ return;
if (mv_is_resizable (&v->miss, new_width))
mv_resize (&v->miss, new_width);
else
- mv_init (&v->miss, new_width);
+ {
+ mv_destroy (&v->miss);
+ mv_init (&v->miss, new_width);
+ }
if (v->val_labs != NULL)
{
}
}
- if (var_get_type (v) != new_type)
- {
- v->print = (new_type == VAR_NUMERIC
- ? fmt_for_output (FMT_F, 8, 2)
- : fmt_for_output (FMT_A, new_width, 0));
- v->write = v->print;
- }
- else if (new_type == VAR_STRING)
- {
- v->print.w = v->print.type == FMT_AHEX ? new_width * 2 : new_width;
- v->write.w = v->write.type == FMT_AHEX ? new_width * 2 : new_width;
- }
+ fmt_resize (&v->print, new_width);
+ fmt_resize (&v->write, new_width);
v->width = new_width;
-
- {
- const int old_val_count = value_cnt_from_width (old_width);
- const int new_val_count = value_cnt_from_width (new_width);
-
- if ( old_val_count != new_val_count)
- dict_var_resized (v, new_val_count - old_val_count);
- }
-
+ dict_var_resized (v, old_width);
dict_var_changed (v);
}
bool
var_is_numeric (const struct variable *v)
{
- return var_get_type (v) == VAR_NUMERIC;
+ return var_get_type (v) == VAL_NUMERIC;
}
/* Returns true if variable V is a string variable, false
bool
var_is_alpha (const struct variable *v)
{
- return var_get_type (v) == VAR_STRING;
-}
-
-/* Returns true if variable V is a short string variable, false
- otherwise. */
-bool
-var_is_short_string (const struct variable *v)
-{
- return v->width > 0 && v->width <= MAX_SHORT_STRING;
-}
-
-/* Returns true if variable V is a long string variable, false
- otherwise. */
-bool
-var_is_long_string (const struct variable *v)
-{
- return v->width > MAX_SHORT_STRING;
-}
-
-/* Returns the number of "union value"s need to store a value of
- variable V. */
-size_t
-var_get_value_cnt (const struct variable *v)
-{
- return value_cnt_from_width (v->width);
+ return var_get_type (v) == VAL_STRING;
}
\f
/* Returns variable V's missing values. */
if (miss != NULL)
{
assert (mv_is_resizable (miss, v->width));
+ mv_destroy (&v->miss);
mv_copy (&v->miss, miss);
mv_resize (&v->miss, v->width);
}
else
- mv_init (&v->miss, v->width);
+ mv_clear (&v->miss);
dict_var_changed (v);
}
if (vls != NULL)
{
assert (val_labs_can_set_width (vls, v->width));
- v->val_labs = val_labs_copy (vls);
+ v->val_labs = val_labs_clone (vls);
val_labs_set_width (v->val_labs, v->width);
dict_var_changed (v);
}
static void
alloc_value_labels (struct variable *v)
{
- assert (!var_is_long_string (v));
if (v->val_labs == NULL)
v->val_labs = val_labs_create (v->width);
}
/* Attempts to add a value label with the given VALUE and LABEL
to V. Returns true if successful, false if VALUE has an
- existing label.
- V must not be a long string variable. */
+ existing label or if V is a long string variable. */
bool
var_add_value_label (struct variable *v,
const union value *value, const char *label)
{
alloc_value_labels (v);
- return val_labs_add (v->val_labs, *value, label);
+ return val_labs_add (v->val_labs, value, label);
}
/* Adds or replaces a value label with the given VALUE and LABEL
to V.
- V must not be a long string variable. */
+ Has no effect if V is a long string variable. */
void
var_replace_value_label (struct variable *v,
const union value *value, const char *label)
{
alloc_value_labels (v);
- val_labs_replace (v->val_labs, *value, label);
+ val_labs_replace (v->val_labs, value, label);
}
/* Removes V's value labels, if any. */
const char *
var_lookup_value_label (const struct variable *v, const union value *value)
{
- return val_labs_find (v->val_labs, *value);
+ return val_labs_find (v->val_labs, value);
}
-/* Get a string representing VALUE for variable V.
- That is, if VALUE has a label, return that label,
- otherwise format VALUE and return the formatted string. */
-const char *
-var_get_value_name (const struct variable *v, const union value *value)
+/* Append STR with a string representing VALUE for variable V.
+ That is, if VALUE has a label, append that label,
+ otherwise format VALUE and append the formatted string.
+ STR must be a pointer to an initialised struct string.
+*/
+void
+var_append_value_name (const struct variable *v, const union value *value,
+ struct string *str)
{
const char *name = var_lookup_value_label (v, value);
if (name == NULL)
{
- static char buf[MAX_STRING + 1];
- data_out (value, &v->print, buf);
- buf[v->print.w] = '\0';
- name = buf;
+ char *s = data_out (value, &v->print);
+ ds_put_cstr (str, s);
+ free (s);
}
- return name;
+ else
+ ds_put_cstr (str, name);
}
\f
/* Print and write formats. */
}
/* Sets V's print format specification to PRINT, which must be a
- valid format specification for outputting a variable of V's
- width. */
+ valid format specification for a variable of V's width
+ (ordinarily an output format, but input formats are not
+ rejected). */
void
var_set_print_format (struct variable *v, const struct fmt_spec *print)
{
}
/* Sets V's write format specification to WRITE, which must be a
- valid format specification for outputting a variable of V's
- width. */
+ valid format specification for a variable of V's width
+ (ordinarily an output format, but input formats are not
+ rejected). */
void
var_set_write_format (struct variable *v, const struct fmt_spec *write)
{
}
/* Sets V's print and write format specifications to FORMAT,
- which must be a valid format specification for outputting a
- variable of V's width. */
+ which must be a valid format specification for a variable of
+ V's width (ordinarily an output format, but input formats are
+ not rejected). */
void
var_set_both_formats (struct variable *v, const struct fmt_spec *format)
{
var_set_print_format (v, format);
var_set_write_format (v, format);
}
+
+/* Returns the default print and write format for a variable of
+ the given TYPE, as set by var_create. The return value can be
+ used to reset a variable's print and write formats to the
+ default. */
+struct fmt_spec
+var_default_formats (int width)
+{
+ return (width == 0
+ ? fmt_for_output (FMT_F, 8, 2)
+ : fmt_for_output (FMT_A, width, 0));
+}
\f
/* Return a string representing this variable, in the form most
appropriate from a human factors perspective, that is, its
ss_truncate (&s, 255);
if (!ss_is_empty (s))
v->label = ss_xstrdup (s);
- dict_var_changed (v);
}
+ dict_var_changed (v);
}
/* Removes any variable label from V. */
v->measure = measure;
dict_var_changed (v);
}
+
+/* Returns the default measurement level for a variable of the
+ given TYPE, as set by var_create. The return value can be
+ used to reset a variable's measurement level to the
+ default. */
+enum measure
+var_default_measure (enum val_type type)
+{
+ return type == VAL_NUMERIC ? MEASURE_SCALE : MEASURE_NOMINAL;
+}
\f
/* Returns V's display width, which applies only to GUIs. */
int
return v->display_width;
}
-
-
-
/* Sets V's display width to DISPLAY_WIDTH. */
void
-var_set_display_width (struct variable *v, int display_width)
+var_set_display_width (struct variable *v, int new_width)
{
- v->display_width = display_width;
+ int old_width = v->display_width;
+
+ v->display_width = new_width;
+
+ if ( old_width != new_width)
+ dict_var_display_width_changed (v);
+
dict_var_changed (v);
}
+
+/* Returns the default display width for a variable of the given
+ WIDTH, as set by var_create. The return value can be used to
+ reset a variable's display width to the default. */
+int
+var_default_display_width (int width)
+{
+ return width == 0 ? 8 : MIN (width, 32);
+}
\f
/* Returns true if A is a valid alignment,
false otherwise. */
v->alignment = alignment;
dict_var_changed (v);
}
+
+/* Returns the default display alignment for a variable of the
+ given TYPE, as set by var_create. The return value can be
+ used to reset a variable's display alignment to the default. */
+enum alignment
+var_default_alignment (enum val_type type)
+{
+ return type == VAL_NUMERIC ? ALIGN_RIGHT : ALIGN_LEFT;
+}
\f
/* Whether variables' values should be preserved from case to
case. */
/* Returns true if variable V's value should be left from case to
- case, instead of being reset to 0, system-missing, or blanks. */
+ case, instead of being reset to system-missing or blanks. */
bool
var_get_leave (const struct variable *v)
{
bool
var_must_leave (const struct variable *v)
{
- return dict_class_from_id (v->name) == DC_SCRATCH;
+ return var_get_dict_class (v) == DC_SCRATCH;
}
\f
-/* Returns V's short name, if it has one, or a null pointer
- otherwise.
+/* Returns the number of short names stored in VAR.
Short names are used only for system and portable file input
and output. They are upper-case only, not necessarily unique,
and limited to SHORT_NAME_LEN characters (plus a null
- terminator). Any variable may have no short name, indicated
- by returning a null pointer. */
+ terminator). Ordinarily a variable has at most one short
+ name, but very long string variables (longer than 255 bytes)
+ may have more. A variable might not have any short name at
+ all if it hasn't been saved to or read from a system or
+ portable file. */
+size_t
+var_get_short_name_cnt (const struct variable *var)
+{
+ return var->short_name_cnt;
+}
+
+/* Returns VAR's short name with the given IDX, if it has one
+ with that index, or a null pointer otherwise. Short names may
+ be sparse: even if IDX is less than the number of short names
+ in VAR, this function may return a null pointer. */
const char *
-var_get_short_name (const struct variable *v)
+var_get_short_name (const struct variable *var, size_t idx)
{
- return v->short_name[0] != '\0' ? v->short_name : NULL;
+ return idx < var->short_name_cnt ? var->short_names[idx] : NULL;
}
-/* Sets V's short_name to SHORT_NAME, truncating it to
- SHORT_NAME_LEN characters and converting it to uppercase in
- the process. Specifying a null pointer for SHORT_NAME clears
- the variable's short name. */
+/* Sets VAR's short name with the given IDX to SHORT_NAME,
+ truncating it to SHORT_NAME_LEN characters and converting it
+ to uppercase in the process. Specifying a null pointer for
+ SHORT_NAME clears the specified short name. */
void
-var_set_short_name (struct variable *v, const char *short_name)
+var_set_short_name (struct variable *var, size_t idx, const char *short_name)
{
- assert (v != NULL);
+ assert (var != NULL);
assert (short_name == NULL || var_is_plausible_name (short_name, false));
- if (short_name != NULL)
+ /* Clear old short name numbered IDX, if any. */
+ if (idx < var->short_name_cnt)
{
- str_copy_trunc (v->short_name, sizeof v->short_name, short_name);
- str_uppercase (v->short_name);
+ free (var->short_names[idx]);
+ var->short_names[idx] = NULL;
}
- else
- v->short_name[0] = '\0';
- dict_var_changed (v);
+
+ /* Install new short name for IDX. */
+ if (short_name != NULL)
+ {
+ if (idx >= var->short_name_cnt)
+ {
+ size_t old_cnt = var->short_name_cnt;
+ size_t i;
+ var->short_name_cnt = MAX (idx * 2, 1);
+ var->short_names = xnrealloc (var->short_names, var->short_name_cnt,
+ sizeof *var->short_names);
+ for (i = old_cnt; i < var->short_name_cnt; i++)
+ var->short_names[i] = NULL;
+ }
+ var->short_names[idx] = xstrndup (short_name, MAX_SHORT_STRING);
+ str_uppercase (var->short_names[idx]);
+ }
+
+ dict_var_changed (var);
}
-/* Clears V's short name. */
+/* Clears V's short names. */
void
-var_clear_short_name (struct variable *v)
+var_clear_short_names (struct variable *v)
{
- assert (v != NULL);
+ size_t i;
- v->short_name[0] = '\0';
+ for (i = 0; i < v->short_name_cnt; i++)
+ free (v->short_names[i]);
+ free (v->short_names);
+ v->short_names = NULL;
+ v->short_name_cnt = 0;
}
\f
/* Relationship with dictionary. */
return v->obs_vals;
}
-/* Sets V's observed categorical values to CAT_VALS. */
+/* Sets V's observed categorical values to CAT_VALS.
+ V becomes the owner of CAT_VALS. */
void
var_set_obs_vals (const struct variable *v_, struct cat_vals *cat_vals)
{
return v->obs_vals != NULL;
}
\f
-/* Returns the dictionary class corresponding to a variable named
- NAME. */
-enum dict_class
-dict_class_from_id (const char *name)
+/* Returns variable V's attribute set. The caller may examine or
+ modify the attribute set, but must not destroy it. Destroying
+ V, or calling var_set_attributes() on V, will also destroy its
+ attribute set. */
+struct attrset *
+var_get_attributes (const struct variable *v)
{
- switch (name[0])
- {
- default:
- return DC_ORDINARY;
- case '$':
- return DC_SYSTEM;
- case '#':
- return DC_SCRATCH;
- }
+ return (struct attrset *) &v->attributes;
}
-/* Returns the name of dictionary class DICT_CLASS. */
-const char *
-dict_class_to_name (enum dict_class dict_class)
+/* Replaces variable V's attributes set by a copy of ATTRS. */
+void
+var_set_attributes (struct variable *v, const struct attrset *attrs)
{
- switch (dict_class)
- {
- case DC_ORDINARY:
- return _("ordinary");
- case DC_SYSTEM:
- return _("system");
- case DC_SCRATCH:
- return _("scratch");
- default:
- NOT_REACHED ();
- }
+ attrset_destroy (&v->attributes);
+ attrset_clone (&v->attributes, attrs);
+}
+
+/* Returns true if V has any custom attributes, false if it has none. */
+bool
+var_has_attributes (const struct variable *v)
+{
+ return attrset_count (&v->attributes) > 0;
}
\f
/* Returns V's vardict structure. */