/* PSPP - a program for statistical analysis.
- Copyright (C) 1997-9, 2000, 2006, 2007 Free Software Foundation, Inc.
+ Copyright (C) 1997-9, 2000, 2006, 2007, 2009, 2010, 2011, 2012, 2013, 2014, 2015 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 <config.h>
-#include "dictionary.h"
+#include "data/dictionary.h"
+#include <stdint.h>
#include <stdlib.h>
#include <ctype.h>
-
-#include "case.h"
-#include "category.h"
-#include "settings.h"
-#include "value-labels.h"
-#include "vardict.h"
-#include "variable.h"
-#include "vector.h"
-#include <libpspp/alloc.h>
-#include <libpspp/array.h>
-#include <libpspp/compiler.h>
-#include <libpspp/hash.h>
-#include <libpspp/message.h>
-#include <libpspp/misc.h>
-#include <libpspp/pool.h>
-#include <libpspp/str.h>
-
-#include "minmax.h"
+#include <unistr.h>
+
+#include "data/attributes.h"
+#include "data/case.h"
+#include "data/identifier.h"
+#include "data/mrset.h"
+#include "data/settings.h"
+#include "data/value-labels.h"
+#include "data/vardict.h"
+#include "data/variable.h"
+#include "data/vector.h"
+#include "libpspp/array.h"
+#include "libpspp/assertion.h"
+#include "libpspp/compiler.h"
+#include "libpspp/hash-functions.h"
+#include "libpspp/hmap.h"
+#include "libpspp/i18n.h"
+#include "libpspp/message.h"
+#include "libpspp/misc.h"
+#include "libpspp/pool.h"
+#include "libpspp/str.h"
+#include "libpspp/string-array.h"
+
+#include "gl/intprops.h"
+#include "gl/minmax.h"
+#include "gl/xalloc.h"
+#include "gl/xmemdup0.h"
#include "gettext.h"
#define _(msgid) gettext (msgid)
/* A dictionary. */
struct dictionary
{
- struct variable **var; /* Variables. */
+ struct vardict_info *var; /* Variables. */
size_t var_cnt, var_cap; /* Number of variables, capacity. */
- struct hsh_table *name_tab; /* Variable index by name. */
+ struct caseproto *proto; /* Prototype for dictionary cases
+ (updated lazily). */
+ struct hmap name_map; /* Variable index by name. */
int next_value_idx; /* Index of next `union value' to allocate. */
const struct variable **split; /* SPLIT FILE vars. */
size_t split_cnt; /* SPLIT FILE count. */
struct variable *weight; /* WEIGHT variable. */
struct variable *filter; /* FILTER variable. */
- size_t case_limit; /* Current case limit (N command). */
+ casenumber case_limit; /* Current case limit (N command). */
char *label; /* File label. */
- struct string documents; /* Documents, as a string. */
+ struct string_array documents; /* Documents. */
struct vector **vector; /* Vectors of variables. */
size_t vector_cnt; /* Number of vectors. */
+ struct attrset attributes; /* Custom attributes. */
+ struct mrset **mrsets; /* Multiple response sets. */
+ size_t n_mrsets; /* Number of multiple response sets. */
+
+ char *encoding; /* Character encoding of string data */
+
const struct dict_callbacks *callbacks; /* Callbacks on dictionary
modification */
void *cb_data ; /* Data passed to callbacks */
+
+ void (*changed) (struct dictionary *, void *); /* Generic change callback */
+ void *changed_data;
};
+static void dict_unset_split_var (struct dictionary *, struct variable *);
+static void dict_unset_mrset_var (struct dictionary *, struct variable *);
+
+/* Returns the encoding for data in dictionary D. The return value is a
+ nonnull string that contains an IANA character set name. */
+const char *
+dict_get_encoding (const struct dictionary *d)
+{
+ return d->encoding ;
+}
+
+/* Returns true if UTF-8 string ID is an acceptable identifier in DICT's
+ encoding, false otherwise. If ISSUE_ERROR is true, issues an explanatory
+ error message on failure. */
+bool
+dict_id_is_valid (const struct dictionary *dict, const char *id,
+ bool issue_error)
+{
+ return id_is_valid (id, dict->encoding, issue_error);
+}
+
+void
+dict_set_change_callback (struct dictionary *d,
+ void (*changed) (struct dictionary *, void*),
+ void *data)
+{
+ d->changed = changed;
+ d->changed_data = data;
+}
+
+/* Discards dictionary D's caseproto. (It will be regenerated
+ lazily, on demand.) */
+static void
+invalidate_proto (struct dictionary *d)
+{
+ caseproto_unref (d->proto);
+ d->proto = NULL;
+}
+
/* Print a representation of dictionary D to stdout, for
debugging purposes. */
void
int i;
for (i = 0 ; i < d->var_cnt ; ++i )
{
- const struct variable *v =
- d->var[i];
- printf ("Name: %s;\tdict_idx: %d; case_idx: %d\n",
+ const struct variable *v = d->var[i].var;
+ printf ("Name: %s;\tdict_idx: %zu; case_idx: %zu\n",
var_get_name (v),
var_get_dict_index (v),
var_get_case_index (v));
dest->cb_data = src->cb_data;
}
-/* Creates and returns a new dictionary. */
+/* Creates and returns a new dictionary with the specified ENCODING. */
struct dictionary *
-dict_create (void)
+dict_create (const char *encoding)
{
struct dictionary *d = xzalloc (sizeof *d);
- d->name_tab = hsh_create (8, compare_vars_by_name, hash_var_by_name,
- NULL, NULL);
+ d->encoding = xstrdup (encoding);
+ hmap_init (&d->name_map);
+ attrset_init (&d->attributes);
+
return d;
}
/* Creates and returns a (deep) copy of an existing
- dictionary. */
+ dictionary.
+
+ The new dictionary's case indexes are copied from the old
+ dictionary. If the new dictionary won't be used to access
+ cases produced with the old dictionary, then the new
+ dictionary's case indexes should be compacted with
+ dict_compact_values to save space.
+
+ Callbacks are not cloned. */
struct dictionary *
dict_clone (const struct dictionary *s)
{
struct dictionary *d;
size_t i;
- assert (s != NULL);
-
- d = dict_create ();
+ d = dict_create (s->encoding);
for (i = 0; i < s->var_cnt; i++)
{
- struct variable *sv = s->var[i];
- struct variable *dv = dict_clone_var_assert (d, sv, var_get_name (sv));
+ struct variable *sv = s->var[i].var;
+ struct variable *dv = dict_clone_var_assert (d, sv);
size_t i;
for (i = 0; i < var_get_short_name_cnt (sv); i++)
var_set_short_name (dv, i, var_get_short_name (sv, i));
+
+ var_get_vardict (dv)->case_index = var_get_vardict (sv)->case_index;
}
d->next_value_idx = s->next_value_idx;
for (i = 0; i < s->vector_cnt; i++)
d->vector[i] = vector_clone (s->vector[i], s, d);
+ dict_set_attributes (d, dict_get_attributes (s));
+
+ for (i = 0; i < s->n_mrsets; i++)
+ {
+ const struct mrset *old = s->mrsets[i];
+ struct mrset *new;
+ size_t j;
+
+ /* Clone old mrset, then replace vars from D by vars from S. */
+ new = mrset_clone (old);
+ for (j = 0; j < new->n_vars; j++)
+ new->vars[j] = dict_lookup_var_assert (d, var_get_name (new->vars[j]));
+
+ dict_add_mrset (d, new);
+ }
+
return d;
}
{
/* FIXME? Should we really clear case_limit, label, documents?
Others are necessarily cleared by deleting all the variables.*/
- assert (d != NULL);
-
while (d->var_cnt > 0 )
{
- struct variable *v = d->var[d->var_cnt - 1];
- int dict_index = var_get_dict_index (v);
- int case_index = var_get_case_index (v);
- int val_cnt = var_get_value_cnt (v);
-
- var_clear_vardict (v);
- var_destroy (v);
-
- d->var_cnt--;
-
- if (d->callbacks && d->callbacks->var_deleted )
- d->callbacks->var_deleted (d,
- dict_index, case_index, val_cnt,
- d->cb_data);
+ dict_delete_var (d, d->var[d->var_cnt - 1].var);
}
free (d->var);
d->var = NULL;
d->var_cnt = d->var_cap = 0;
- hsh_clear (d->name_tab);
+ invalidate_proto (d);
+ hmap_clear (&d->name_map);
d->next_value_idx = 0;
dict_set_split_vars (d, NULL, 0);
dict_set_weight (d, NULL);
d->case_limit = 0;
free (d->label);
d->label = NULL;
- ds_destroy (&d->documents);
+ string_array_clear (&d->documents);
dict_clear_vectors (d);
-}
-
-/* Destroys the aux data for every variable in D, by calling
- var_clear_aux() for each variable. */
-void
-dict_clear_aux (struct dictionary *d)
-{
- int i;
-
- assert (d != NULL);
-
- for (i = 0; i < d->var_cnt; i++)
- var_clear_aux (d->var[i]);
+ attrset_clear (&d->attributes);
}
/* Clears a dictionary and destroys it. */
{
if (d != NULL)
{
- /* In general, we don't want callbacks occuring, if the dictionary
+ /* In general, we don't want callbacks occurring, if the dictionary
is being destroyed */
d->callbacks = NULL ;
dict_clear (d);
- hsh_destroy (d->name_tab);
+ string_array_destroy (&d->documents);
+ hmap_destroy (&d->name_map);
+ attrset_destroy (&d->attributes);
+ dict_clear_mrsets (d);
+ free (d->encoding);
free (d);
}
}
size_t
dict_get_var_cnt (const struct dictionary *d)
{
- assert (d != NULL);
-
return d->var_cnt;
}
struct variable *
dict_get_var (const struct dictionary *d, size_t idx)
{
- assert (d != NULL);
assert (idx < d->var_cnt);
- return d->var[idx];
+ return d->var[idx].var;
}
-inline void
+/* Sets *VARS to an array of pointers to variables in D and *CNT
+ to the number of variables in *D. All variables are returned
+ except for those, if any, in the classes indicated by EXCLUDE.
+ (There is no point in putting DC_SYSTEM in EXCLUDE as
+ dictionaries never include system variables.) */
+void
dict_get_vars (const struct dictionary *d, const struct variable ***vars,
- size_t *cnt, unsigned exclude_classes)
+ size_t *cnt, enum dict_class exclude)
{
- dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude_classes);
+ dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude);
}
/* Sets *VARS to an array of pointers to variables in D and *CNT
to the number of variables in *D. All variables are returned
- if EXCLUDE_CLASSES is 0, or it may contain one or more of (1u
- << DC_ORDINARY), (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to
- exclude the corresponding type of variable. */
+ except for those, if any, in the classes indicated by EXCLUDE.
+ (There is no point in putting DC_SYSTEM in EXCLUDE as
+ dictionaries never include system variables.) */
void
dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars,
- size_t *cnt, unsigned exclude_classes)
+ size_t *cnt, enum dict_class exclude)
{
size_t count;
size_t i;
- assert (d != NULL);
- assert (vars != NULL);
- assert (cnt != NULL);
- assert ((exclude_classes & ~((1u << DC_ORDINARY)
- | (1u << DC_SYSTEM)
- | (1u << DC_SCRATCH))) == 0);
+ assert (exclude == (exclude & DC_ALL));
count = 0;
for (i = 0; i < d->var_cnt; i++)
{
- enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
- if (!(exclude_classes & (1u << class)))
+ enum dict_class class = var_get_dict_class (d->var[i].var);
+ if (!(class & exclude))
count++;
}
*cnt = 0;
for (i = 0; i < d->var_cnt; i++)
{
- enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
- if (!(exclude_classes & (1u << class)))
- (*vars)[(*cnt)++] = d->var[i];
+ enum dict_class class = var_get_dict_class (d->var[i].var);
+ if (!(class & exclude))
+ (*vars)[(*cnt)++] = d->var[i].var;
}
assert (*cnt == count);
}
static struct variable *
-add_var (struct dictionary *d, struct variable *v)
+add_var_with_case_index (struct dictionary *d, struct variable *v,
+ int case_index)
{
- /* Add dictionary info to variable. */
- struct vardict_info vdi;
- vdi.case_index = d->next_value_idx;
- vdi.dict_index = d->var_cnt;
- vdi.dict = d;
- var_set_vardict (v, &vdi);
+ struct vardict_info *vardict;
+
+ assert (case_index >= d->next_value_idx);
/* Update dictionary. */
if (d->var_cnt >= d->var_cap)
{
- d->var_cap = 8 + 2 * d->var_cap;
- d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var);
+ size_t i;
+
+ d->var = x2nrealloc (d->var, &d->var_cap, sizeof *d->var);
+ hmap_clear (&d->name_map);
+ for (i = 0; i < d->var_cnt; i++)
+ {
+ var_set_vardict (d->var[i].var, &d->var[i]);
+ hmap_insert_fast (&d->name_map, &d->var[i].name_node,
+ d->var[i].name_node.hash);
+ }
}
- d->var[d->var_cnt++] = v;
- hsh_force_insert (d->name_tab, v);
+ vardict = &d->var[d->var_cnt++];
+ vardict->dict = d;
+ vardict->var = v;
+ hmap_insert (&d->name_map, &vardict->name_node,
+ utf8_hash_case_string (var_get_name (v), 0));
+ vardict->case_index = case_index;
+ var_set_vardict (v, vardict);
+
+ if ( d->changed ) d->changed (d, d->changed_data);
if ( d->callbacks && d->callbacks->var_added )
d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data);
- d->next_value_idx += var_get_value_cnt (v);
+ invalidate_proto (d);
+ d->next_value_idx = case_index + 1;
return v;
}
+static struct variable *
+add_var (struct dictionary *d, struct variable *v)
+{
+ return add_var_with_case_index (d, v, d->next_value_idx);
+}
+
/* Creates and returns a new variable in D with the given NAME
and WIDTH. Returns a null pointer if the given NAME would
duplicate that of an existing variable in the dictionary. */
return add_var (d, var_create (name, width));
}
-/* Creates and returns a new variable in D with name NAME, as a
- copy of existing variable OLD_VAR, which need not be in D or
- in any dictionary. Returns a null pointer if the given NAME
- would duplicate that of an existing variable in the
+/* Creates and returns a new variable in D, as a copy of existing variable
+ OLD_VAR, which need not be in D or in any dictionary. Returns a null
+ pointer if OLD_VAR's name would duplicate that of an existing variable in
+ the dictionary. */
+struct variable *
+dict_clone_var (struct dictionary *d, const struct variable *old_var)
+{
+ return dict_clone_var_as (d, old_var, var_get_name (old_var));
+}
+
+/* Creates and returns a new variable in D, as a copy of existing variable
+ OLD_VAR, which need not be in D or in any dictionary. Assert-fails if
+ OLD_VAR's name would duplicate that of an existing variable in the
dictionary. */
struct variable *
-dict_clone_var (struct dictionary *d, const struct variable *old_var,
- const char *name)
+dict_clone_var_assert (struct dictionary *d, const struct variable *old_var)
+{
+ return dict_clone_var_as_assert (d, old_var, var_get_name (old_var));
+}
+
+/* Creates and returns a new variable in D with name NAME, as a copy of
+ existing variable OLD_VAR, which need not be in D or in any dictionary.
+ Returns a null pointer if the given NAME would duplicate that of an existing
+ variable in the dictionary. */
+struct variable *
+dict_clone_var_as (struct dictionary *d, const struct variable *old_var,
+ const char *name)
{
return (dict_lookup_var (d, name) == NULL
- ? dict_clone_var_assert (d, old_var, name)
+ ? dict_clone_var_as_assert (d, old_var, name)
: NULL);
}
-/* Creates and returns a new variable in D with name NAME, as a
- copy of existing variable OLD_VAR, which need not be in D or
- in any dictionary. Assert-fails if the given NAME would
- duplicate that of an existing variable in the dictionary. */
+/* Creates and returns a new variable in D with name NAME, as a copy of
+ existing variable OLD_VAR, which need not be in D or in any dictionary.
+ Assert-fails if the given NAME would duplicate that of an existing variable
+ in the dictionary. */
struct variable *
-dict_clone_var_assert (struct dictionary *d, const struct variable *old_var,
- const char *name)
+dict_clone_var_as_assert (struct dictionary *d, const struct variable *old_var,
+ const char *name)
{
struct variable *new_var = var_clone (old_var);
assert (dict_lookup_var (d, name) == NULL);
return add_var (d, new_var);
}
+struct variable *
+dict_clone_var_in_place_assert (struct dictionary *d,
+ const struct variable *old_var)
+{
+ assert (dict_lookup_var (d, var_get_name (old_var)) == NULL);
+ return add_var_with_case_index (d, var_clone (old_var),
+ var_get_case_index (old_var));
+}
+
/* Returns the variable named NAME in D, or a null pointer if no
variable has that name. */
struct variable *
dict_lookup_var (const struct dictionary *d, const char *name)
{
- struct variable *target ;
- struct variable *result ;
+ struct vardict_info *vardict;
- if ( ! var_is_plausible_name (name, false))
- return NULL;
-
- target = var_create (name, 0);
- result = hsh_find (d->name_tab, target);
- var_destroy (target);
+ HMAP_FOR_EACH_WITH_HASH (vardict, struct vardict_info, name_node,
+ utf8_hash_case_string (name, 0), &d->name_map)
+ {
+ struct variable *var = vardict->var;
+ if (!utf8_strcasecmp (var_get_name (var), name))
+ return var;
+ }
- return result;
+ return NULL;
}
/* Returns the variable named NAME in D. Assert-fails if no
bool
dict_contains_var (const struct dictionary *d, const struct variable *v)
{
- if (var_has_vardict (v))
- {
- const struct vardict_info *vdi = var_get_vardict (v);
- return (vdi->dict_index >= 0
- && vdi->dict_index < d->var_cnt
- && d->var[vdi->dict_index] == v);
- }
- else
- return false;
+ return (var_has_vardict (v)
+ && vardict_get_dictionary (var_get_vardict (v)) == d);
}
/* Compares two double pointers to variables, which should point
return *a < *b ? -1 : *a > *b;
}
-/* Sets the dict_index in V's vardict to DICT_INDEX. */
static void
-set_var_dict_index (struct variable *v, int dict_index)
+unindex_var (struct dictionary *d, struct vardict_info *vardict)
{
- struct vardict_info vdi = *var_get_vardict (v);
- struct dictionary *d = vdi.dict;
- vdi.dict_index = dict_index;
- var_set_vardict (v, &vdi);
+ hmap_delete (&d->name_map, &vardict->name_node);
+}
- if ( d->callbacks && d->callbacks->var_changed )
- d->callbacks->var_changed (d, dict_index, d->cb_data);
+/* This function assumes that vardict->name_node.hash is valid, that is, that
+ its name has not changed since it was hashed (rename_var() updates this
+ hash along with the name itself). */
+static void
+reindex_var (struct dictionary *d, struct vardict_info *vardict)
+{
+ struct variable *old = (d->callbacks && d->callbacks->var_changed
+ ? var_clone (vardict->var)
+ : NULL);
+
+ struct variable *var = vardict->var;
+ var_set_vardict (var, vardict);
+ hmap_insert_fast (&d->name_map, &vardict->name_node,
+ vardict->name_node.hash);
+
+ if ( d->changed ) d->changed (d, d->changed_data);
+ if (old)
+ {
+ d->callbacks->var_changed (d, var_get_dict_index (var), VAR_TRAIT_POSITION, old, d->cb_data);
+ var_destroy (old);
+ }
}
/* Sets the case_index in V's vardict to CASE_INDEX. */
static void
set_var_case_index (struct variable *v, int case_index)
{
- struct vardict_info vdi = *var_get_vardict (v);
- vdi.case_index = case_index;
- var_set_vardict (v, &vdi);
+ var_get_vardict (v)->case_index = case_index;
+}
+
+/* Removes the dictionary variables with indexes from FROM to TO (exclusive)
+ from name_map. */
+static void
+unindex_vars (struct dictionary *d, size_t from, size_t to)
+{
+ size_t i;
+
+ for (i = from; i < to; i++)
+ unindex_var (d, &d->var[i]);
}
/* Re-sets the dict_index in the dictionary variables with
size_t i;
for (i = from; i < to; i++)
- set_var_dict_index (d->var[i], i);
+ reindex_var (d, &d->var[i]);
}
/* Deletes variable V from dictionary D and frees V.
This is a very bad idea if there might be any pointers to V
- from outside D. In general, no variable in the active file's
+ from outside D. In general, no variable in the active dataset's
dictionary should be deleted when any transformations are
active on the dictionary's dataset, because those
transformations might reference the deleted variable. The
safest time to delete a variable is just after a procedure has
- been executed, as done by MODIFY VARS.
+ been executed, as done by DELETE VARIABLES.
Pointers to V within D are not a problem, because
dict_delete_var() knows to remove V from split variables,
{
int dict_index = var_get_dict_index (v);
const int case_index = var_get_case_index (v);
- const int val_cnt = var_get_value_cnt (v);
assert (dict_contains_var (d, v));
- /* Delete aux data. */
- var_clear_aux (v);
-
dict_unset_split_var (d, v);
+ dict_unset_mrset_var (d, v);
if (d->weight == v)
dict_set_weight (d, NULL);
dict_clear_vectors (d);
/* Remove V from var array. */
+ unindex_vars (d, dict_index, d->var_cnt);
remove_element (d->var, d->var_cnt, sizeof *d->var, dict_index);
d->var_cnt--;
/* Update dict_index for each affected variable. */
reindex_vars (d, dict_index, d->var_cnt);
- /* Update name hash. */
- hsh_force_delete (d->name_tab, v);
-
-
/* Free memory. */
var_clear_vardict (v);
- var_destroy (v);
+ if ( d->changed ) d->changed (d, d->changed_data);
+
+ invalidate_proto (d);
if (d->callbacks && d->callbacks->var_deleted )
- d->callbacks->var_deleted (d, dict_index, case_index, val_cnt, d->cb_data);
+ d->callbacks->var_deleted (d, v, dict_index, case_index, d->cb_data);
+
+ var_destroy (v);
}
/* Deletes the COUNT variables listed in VARS from D. This is
{
/* FIXME: this can be done in O(count) time, but this algorithm
is O(count**2). */
- assert (d != NULL);
assert (count == 0 || vars != NULL);
while (count-- > 0)
assert (idx + count <= d->var_cnt);
while (count-- > 0)
- dict_delete_var (d, d->var[idx]);
+ dict_delete_var (d, d->var[idx].var);
}
/* Deletes scratch variables from dictionary D. */
/* FIXME: this can be done in O(count) time, but this algorithm
is O(count**2). */
- assert (d != NULL);
-
for (i = 0; i < d->var_cnt; )
- if (dict_class_from_id (var_get_name (d->var[i])) == DC_SCRATCH)
- dict_delete_var (d, d->var[i]);
+ if (var_get_dict_class (d->var[i].var) == DC_SCRATCH)
+ dict_delete_var (d, d->var[i].var);
else
i++;
}
size_t old_index = var_get_dict_index (v);
assert (new_index < d->var_cnt);
+
+ unindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
move_element (d->var, d->var_cnt, sizeof *d->var, old_index, new_index);
reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1);
}
dict_reorder_vars (struct dictionary *d,
struct variable *const *order, size_t count)
{
- struct variable **new_var;
+ struct vardict_info *new_var;
size_t i;
- assert (d != NULL);
assert (count == 0 || order != NULL);
assert (count <= d->var_cnt);
- new_var = xnmalloc (d->var_cnt, sizeof *new_var);
- memcpy (new_var, order, count * sizeof *new_var);
+ new_var = xnmalloc (d->var_cap, sizeof *new_var);
+
+ /* Add variables in ORDER to new_var. */
for (i = 0; i < count; i++)
{
- size_t index = var_get_dict_index (order[i]);
- assert (d->var[index] == order[i]);
- d->var[index] = NULL;
- set_var_dict_index (order[i], i);
+ struct vardict_info *old_var;
+
+ assert (dict_contains_var (d, order[i]));
+
+ old_var = var_get_vardict (order[i]);
+ new_var[i] = *old_var;
+ old_var->dict = NULL;
}
+
+ /* Add remaining variables to new_var. */
for (i = 0; i < d->var_cnt; i++)
- if (d->var[i] != NULL)
- {
- assert (count < d->var_cnt);
- new_var[count] = d->var[i];
- set_var_dict_index (new_var[count], count);
- count++;
- }
+ if (d->var[i].dict != NULL)
+ new_var[count++] = d->var[i];
+ assert (count == d->var_cnt);
+
+ /* Replace old vardicts by new ones. */
free (d->var);
d->var = new_var;
+
+ hmap_clear (&d->name_map);
+ reindex_vars (d, 0, d->var_cnt);
}
-/* Changes the name of variable V in dictionary D to NEW_NAME. */
+/* Changes the name of variable V that is currently in a dictionary to
+ NEW_NAME. */
static void
-rename_var (struct dictionary *d, struct variable *v, const char *new_name)
+rename_var (struct variable *v, const char *new_name)
{
- struct vardict_info vdi;
-
- assert (dict_contains_var (d, v));
-
- vdi = *var_get_vardict (v);
+ struct vardict_info *vardict = var_get_vardict (v);
var_clear_vardict (v);
var_set_name (v, new_name);
- var_set_vardict (v, &vdi);
+ vardict->name_node.hash = utf8_hash_case_string (new_name, 0);
+ var_set_vardict (v, vardict);
}
/* Changes the name of V in D to name NEW_NAME. Assert-fails if
dict_rename_var (struct dictionary *d, struct variable *v,
const char *new_name)
{
- assert (!strcasecmp (var_get_name (v), new_name)
+ struct variable *old = var_clone (v);
+ assert (!utf8_strcasecmp (var_get_name (v), new_name)
|| dict_lookup_var (d, new_name) == NULL);
- hsh_force_delete (d->name_tab, v);
- rename_var (d, v, new_name);
- hsh_force_insert (d->name_tab, v);
+ unindex_var (d, var_get_vardict (v));
+ rename_var (v, new_name);
+ reindex_var (d, var_get_vardict (v));
- if (get_algorithm () == ENHANCED)
+ if (settings_get_algorithm () == ENHANCED)
var_clear_short_names (v);
+ if ( d->changed ) d->changed (d, d->changed_data);
if ( d->callbacks && d->callbacks->var_changed )
- d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
+ d->callbacks->var_changed (d, var_get_dict_index (v), VAR_TRAIT_NAME, old, d->cb_data);
+
+ var_destroy (old);
}
/* Renames COUNT variables specified in VARS to the names given
and rename them. */
for (i = 0; i < count; i++)
{
- hsh_force_delete (d->name_tab, vars[i]);
- rename_var (d, vars[i], new_names[i]);
+ unindex_var (d, var_get_vardict (vars[i]));
+ rename_var (vars[i], new_names[i]);
}
/* Add the renamed variables back into the name hash,
checking for conflicts. */
for (i = 0; i < count; i++)
- if (hsh_insert (d->name_tab, vars[i]) != NULL)
- {
- /* There is a name conflict.
- Back out all the name changes that have already
- taken place, and indicate failure. */
- size_t fail_idx = i;
- if (err_name != NULL)
- *err_name = new_names[i];
-
- for (i = 0; i < fail_idx; i++)
- hsh_force_delete (d->name_tab, vars[i]);
-
- for (i = 0; i < count; i++)
- {
- rename_var (d, vars[i], old_names[i]);
- hsh_force_insert (d->name_tab, vars[i]);
- }
-
- pool_destroy (pool);
- return false;
- }
+ {
+ if (dict_lookup_var (d, var_get_name (vars[i])) != NULL)
+ {
+ /* There is a name conflict.
+ Back out all the name changes that have already
+ taken place, and indicate failure. */
+ size_t fail_idx = i;
+ if (err_name != NULL)
+ *err_name = new_names[i];
+
+ for (i = 0; i < fail_idx; i++)
+ unindex_var (d, var_get_vardict (vars[i]));
+
+ for (i = 0; i < count; i++)
+ {
+ rename_var (vars[i], old_names[i]);
+ reindex_var (d, var_get_vardict (vars[i]));
+ }
+
+ pool_destroy (pool);
+ return false;
+ }
+ reindex_var (d, var_get_vardict (vars[i]));
+ }
/* Clear short names. */
- if (get_algorithm () == ENHANCED)
+ if (settings_get_algorithm () == ENHANCED)
for (i = 0; i < count; i++)
var_clear_short_names (vars[i]);
return true;
}
+/* Returns true if a variable named NAME may be inserted in DICT;
+ that is, if there is not already a variable with that name in
+ DICT and if NAME is not a reserved word. (The caller's checks
+ have already verified that NAME is otherwise acceptable as a
+ variable name.) */
+static bool
+var_name_is_insertable (const struct dictionary *dict, const char *name)
+{
+ return (dict_lookup_var (dict, name) == NULL
+ && lex_id_to_token (ss_cstr (name)) == T_ID);
+}
+
+static char *
+make_hinted_name (const struct dictionary *dict, const char *hint)
+{
+ size_t hint_len = strlen (hint);
+ bool dropped = false;
+ char *root, *rp;
+ size_t ofs;
+ int mblen;
+
+ /* The allocation size here is OK: characters that are copied directly fit
+ OK, and characters that are not copied directly are replaced by a single
+ '_' byte. If u8_mbtouc() replaces bad input by 0xfffd, then that will get
+ replaced by '_' too. */
+ root = rp = xmalloc (hint_len + 1);
+ for (ofs = 0; ofs < hint_len; ofs += mblen)
+ {
+ ucs4_t uc;
+
+ mblen = u8_mbtouc (&uc, CHAR_CAST (const uint8_t *, hint + ofs),
+ hint_len - ofs);
+ if (rp == root
+ ? lex_uc_is_id1 (uc) && uc != '$'
+ : lex_uc_is_idn (uc))
+ {
+ if (dropped)
+ {
+ *rp++ = '_';
+ dropped = false;
+ }
+ rp += u8_uctomb (CHAR_CAST (uint8_t *, rp), uc, 6);
+ }
+ else if (rp != root)
+ dropped = true;
+ }
+ *rp = '\0';
+
+ if (root[0] != '\0')
+ {
+ unsigned long int i;
+
+ if (var_name_is_insertable (dict, root))
+ return root;
+
+ for (i = 0; i < ULONG_MAX; i++)
+ {
+ char suffix[INT_BUFSIZE_BOUND (i) + 1];
+ char *name;
+
+ suffix[0] = '_';
+ if (!str_format_26adic (i + 1, true, &suffix[1], sizeof suffix - 1))
+ NOT_REACHED ();
+
+ name = utf8_encoding_concat (root, suffix, dict->encoding, 64);
+ if (var_name_is_insertable (dict, name))
+ {
+ free (root);
+ return name;
+ }
+ free (name);
+ }
+ }
+
+ free (root);
+
+ return NULL;
+}
+
+static char *
+make_numeric_name (const struct dictionary *dict, unsigned long int *num_start)
+{
+ unsigned long int number;
+
+ for (number = num_start != NULL ? MAX (*num_start, 1) : 1;
+ number < ULONG_MAX;
+ number++)
+ {
+ char name[3 + INT_STRLEN_BOUND (number) + 1];
+
+ sprintf (name, "VAR%03lu", number);
+ if (dict_lookup_var (dict, name) == NULL)
+ {
+ if (num_start != NULL)
+ *num_start = number + 1;
+ return xstrdup (name);
+ }
+ }
+
+ NOT_REACHED ();
+}
+
+
+/* Devises and returns a variable name unique within DICT. The variable name
+ is owned by the caller, which must free it with free() when it is no longer
+ needed.
+
+ HINT, if it is non-null, is used as a suggestion that will be
+ modified for suitability as a variable name and for
+ uniqueness.
+
+ If HINT is null or entirely unsuitable, a name in the form
+ "VAR%03d" will be generated, where the smallest unused integer
+ value is used. If NUM_START is non-null, then its value is
+ used as the minimum numeric value to check, and it is updated
+ to the next value to be checked.
+*/
+char *
+dict_make_unique_var_name (const struct dictionary *dict, const char *hint,
+ unsigned long int *num_start)
+{
+ if (hint != NULL)
+ {
+ char *hinted_name = make_hinted_name (dict, hint);
+ if (hinted_name != NULL)
+ return hinted_name;
+ }
+ return make_numeric_name (dict, num_start);
+}
+
/* Returns the weighting variable in dictionary D, or a null
pointer if the dictionary is unweighted. */
struct variable *
dict_get_weight (const struct dictionary *d)
{
- assert (d != NULL);
assert (d->weight == NULL || dict_contains_var (d, d->weight));
return d->weight;
}
-/* Returns the value of D's weighting variable in case C, except that a
- negative weight is returned as 0. Returns 1 if the dictionary is
- unweighted. Will warn about missing, negative, or zero values if
- warn_on_invalid is true. The function will set warn_on_invalid to false
- if an invalid weight is found. */
+/* Returns the value of D's weighting variable in case C, except
+ that a negative weight is returned as 0. Returns 1 if the
+ dictionary is unweighted. Will warn about missing, negative,
+ or zero values if *WARN_ON_INVALID is true. The function will
+ set *WARN_ON_INVALID to false if an invalid weight is
+ found. */
double
dict_get_case_weight (const struct dictionary *d, const struct ccase *c,
bool *warn_on_invalid)
{
- assert (d != NULL);
assert (c != NULL);
if (d->weight == NULL)
else
{
double w = case_num (c, d->weight);
- if (w < 0.0 || var_is_num_missing (d->weight, w, MV_ANY))
- w = 0.0;
- if ( w == 0.0 && warn_on_invalid != NULL && *warn_on_invalid ) {
- *warn_on_invalid = false;
- msg (SW, _("At least one case in the data file had a weight value "
- "that was user-missing, system-missing, zero, or "
- "negative. These case(s) were ignored."));
- }
- return w;
+
+ return var_force_valid_weight (d->weight, w, warn_on_invalid);
}
}
void
dict_set_weight (struct dictionary *d, struct variable *v)
{
- assert (d != NULL);
assert (v == NULL || dict_contains_var (d, v));
assert (v == NULL || var_is_numeric (v));
d->weight = v;
+ if (d->changed) d->changed (d, d->changed_data);
if ( d->callbacks && d->callbacks->weight_changed )
d->callbacks->weight_changed (d,
v ? var_get_dict_index (v) : -1,
struct variable *
dict_get_filter (const struct dictionary *d)
{
- assert (d != NULL);
assert (d->filter == NULL || dict_contains_var (d, d->filter));
return d->filter;
void
dict_set_filter (struct dictionary *d, struct variable *v)
{
- assert (d != NULL);
assert (v == NULL || dict_contains_var (d, v));
+ assert (v == NULL || var_is_numeric (v));
d->filter = v;
+ if (d->changed) d->changed (d, d->changed_data);
if ( d->callbacks && d->callbacks->filter_changed )
d->callbacks->filter_changed (d,
v ? var_get_dict_index (v) : -1,
/* Returns the case limit for dictionary D, or zero if the number
of cases is unlimited. */
-size_t
+casenumber
dict_get_case_limit (const struct dictionary *d)
{
- assert (d != NULL);
-
return d->case_limit;
}
/* Sets CASE_LIMIT as the case limit for dictionary D. Use
0 for CASE_LIMIT to indicate no limit. */
void
-dict_set_case_limit (struct dictionary *d, size_t case_limit)
+dict_set_case_limit (struct dictionary *d, casenumber case_limit)
{
- assert (d != NULL);
-
d->case_limit = case_limit;
}
+/* Returns the prototype used for cases created by dictionary D. */
+const struct caseproto *
+dict_get_proto (const struct dictionary *d_)
+{
+ struct dictionary *d = CONST_CAST (struct dictionary *, d_);
+ if (d->proto == NULL)
+ {
+ size_t i;
+
+ d->proto = caseproto_create ();
+ d->proto = caseproto_reserve (d->proto, d->var_cnt);
+ for (i = 0; i < d->var_cnt; i++)
+ d->proto = caseproto_set_width (d->proto,
+ var_get_case_index (d->var[i].var),
+ var_get_width (d->var[i].var));
+ }
+ return d->proto;
+}
+
/* Returns the case index of the next value to be added to D.
This value is the number of `union value's that need to be
allocated to store a case for dictionary D. */
int
dict_get_next_value_idx (const struct dictionary *d)
{
- assert (d != NULL);
-
return d->next_value_idx;
}
size_t
dict_get_case_size (const struct dictionary *d)
{
- assert (d != NULL);
-
return sizeof (union value) * dict_get_next_value_idx (d);
}
d->next_value_idx = 0;
for (i = 0; i < d->var_cnt; i++)
{
- struct variable *v = d->var[i];
- set_var_case_index (v, d->next_value_idx);
- d->next_value_idx += var_get_value_cnt (v);
- }
-}
-
-/*
- Reassigns case indices for D, increasing each index above START by
- the value PADDING.
-*/
-static void
-dict_pad_values (struct dictionary *d, int start, int padding)
-{
- size_t i;
-
- if ( padding <= 0 )
- return;
-
- for (i = 0; i < d->var_cnt; ++i)
- {
- struct variable *v = d->var[i];
-
- int index = var_get_case_index (v);
-
- if ( index >= start)
- set_var_case_index (v, index + padding);
+ struct variable *v = d->var[i].var;
+ set_var_case_index (v, d->next_value_idx++);
}
-
- d->next_value_idx += padding;
+ invalidate_proto (d);
}
-
/* Returns the number of values occupied by the variables in
dictionary D. All variables are considered if EXCLUDE_CLASSES
is 0, or it may contain one or more of (1u << DC_ORDINARY),
cnt = 0;
for (i = 0; i < d->var_cnt; i++)
{
- enum dict_class class = dict_class_from_id (var_get_name (d->var[i]));
+ enum dict_class class = var_get_dict_class (d->var[i].var);
if (!(exclude_classes & (1u << class)))
- cnt += var_get_value_cnt (d->var[i]);
+ cnt++;
}
return cnt;
}
+
+/* Returns the case prototype that would result after deleting
+ all variables from D that are not in one of the
+ EXCLUDE_CLASSES and compacting the dictionary with
+ dict_compact().
+
+ The caller must unref the returned caseproto when it is no
+ longer needed. */
+struct caseproto *
+dict_get_compacted_proto (const struct dictionary *d,
+ unsigned int exclude_classes)
+{
+ struct caseproto *proto;
+ size_t i;
+
+ assert ((exclude_classes & ~((1u << DC_ORDINARY)
+ | (1u << DC_SYSTEM)
+ | (1u << DC_SCRATCH))) == 0);
+
+ proto = caseproto_create ();
+ for (i = 0; i < d->var_cnt; i++)
+ {
+ struct variable *v = d->var[i].var;
+ if (!(exclude_classes & (1u << var_get_dict_class (v))))
+ proto = caseproto_add_width (proto, var_get_width (v));
+ }
+ return proto;
+}
\f
/* Returns the SPLIT FILE vars (see cmd_split_file()). Call
dict_get_split_cnt() to determine how many SPLIT FILE vars
const struct variable *const *
dict_get_split_vars (const struct dictionary *d)
{
- assert (d != NULL);
-
return d->split;
}
size_t
dict_get_split_cnt (const struct dictionary *d)
{
- assert (d != NULL);
-
return d->split_cnt;
}
-/* Removes variable V from the set of split variables in dictionary D */
-void
-dict_unset_split_var (struct dictionary *d,
- struct variable *v)
+/* Removes variable V, which must be in D, from D's set of split
+ variables. */
+static void
+dict_unset_split_var (struct dictionary *d, struct variable *v)
{
- const int count = d->split_cnt;
- d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
- &v, compare_var_ptrs, NULL);
+ int orig_count;
- if ( count == d->split_cnt)
- return;
+ assert (dict_contains_var (d, v));
- if ( d->callbacks && d->callbacks->split_changed )
- d->callbacks->split_changed (d, d->cb_data);
+ orig_count = d->split_cnt;
+ d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split,
+ &v, compare_var_ptrs, NULL);
+ if (orig_count != d->split_cnt)
+ {
+ if (d->changed) d->changed (d, d->changed_data);
+ /* We changed the set of split variables so invoke the
+ callback. */
+ if (d->callbacks && d->callbacks->split_changed)
+ d->callbacks->split_changed (d, d->cb_data);
+ }
}
/* Sets CNT split vars SPLIT in dictionary D. */
dict_set_split_vars (struct dictionary *d,
struct variable *const *split, size_t cnt)
{
- assert (d != NULL);
assert (cnt == 0 || split != NULL);
d->split_cnt = cnt;
- d->split = cnt > 0 ? xnrealloc (d->split, cnt, sizeof *d->split) : NULL;
- memcpy (d->split, split, cnt * sizeof *d->split);
+ if ( cnt > 0 )
+ {
+ d->split = xnrealloc (d->split, cnt, sizeof *d->split) ;
+ memcpy (d->split, split, cnt * sizeof *d->split);
+ }
+ else
+ {
+ free (d->split);
+ d->split = NULL;
+ }
+ if (d->changed) d->changed (d, d->changed_data);
if ( d->callbacks && d->callbacks->split_changed )
d->callbacks->split_changed (d, d->cb_data);
}
const char *
dict_get_label (const struct dictionary *d)
{
- assert (d != NULL);
-
return d->label;
}
-/* Sets D's file label to LABEL, truncating it to a maximum of 60
- characters. */
+/* Sets D's file label to LABEL, truncating it to at most 60 bytes in D's
+ encoding.
+
+ Removes D's label if LABEL is null or the empty string. */
void
dict_set_label (struct dictionary *d, const char *label)
{
- assert (d != NULL);
-
free (d->label);
- if (label == NULL)
+ if (label == NULL || label[0] == '\0')
d->label = NULL;
- else if (strlen (label) < 60)
- d->label = xstrdup (label);
else
- {
- d->label = xmalloc (61);
- memcpy (d->label, label, 60);
- d->label[60] = '\0';
- }
+ d->label = utf8_encoding_trunc (label, d->encoding, 60);
}
-/* Returns the documents for D, or a null pointer if D has no
- documents. If the return value is nonnull, then the string
- will be an exact multiple of DOC_LINE_LENGTH bytes in length,
- with each segment corresponding to one line. */
-const char *
+/* Returns the documents for D, as an UTF-8 encoded string_array. The
+ return value is always nonnull; if there are no documents then the
+ string_arary is empty.*/
+const struct string_array *
dict_get_documents (const struct dictionary *d)
{
- return ds_is_empty (&d->documents) ? NULL : ds_cstr (&d->documents);
+ return &d->documents;
}
-/* Sets the documents for D to DOCUMENTS, or removes D's
- documents if DOCUMENT is a null pointer. If DOCUMENTS is
- nonnull, then it should be an exact multiple of
- DOC_LINE_LENGTH bytes in length, with each segment
- corresponding to one line. */
+/* Replaces the documents for D by NEW_DOCS, a UTF-8 encoded string_array. */
void
-dict_set_documents (struct dictionary *d, const char *documents)
+dict_set_documents (struct dictionary *d, const struct string_array *new_docs)
{
- size_t remainder;
+ size_t i;
- ds_assign_cstr (&d->documents, documents != NULL ? documents : "");
+ dict_clear_documents (d);
- /* In case the caller didn't get it quite right, pad out the
- final line with spaces. */
- remainder = ds_length (&d->documents) % DOC_LINE_LENGTH;
- if (remainder != 0)
- ds_put_char_multiple (&d->documents, ' ', DOC_LINE_LENGTH - remainder);
+ for (i = 0; i < new_docs->n; i++)
+ dict_add_document_line (d, new_docs->strings[i], false);
+}
+
+/* Replaces the documents for D by UTF-8 encoded string NEW_DOCS, dividing it
+ into individual lines at new-line characters. Each line is truncated to at
+ most DOC_LINE_LENGTH bytes in D's encoding. */
+void
+dict_set_documents_string (struct dictionary *d, const char *new_docs)
+{
+ const char *s;
+
+ dict_clear_documents (d);
+ for (s = new_docs; *s != '\0'; )
+ {
+ size_t len = strcspn (s, "\n");
+ char *line = xmemdup0 (s, len);
+ dict_add_document_line (d, line, false);
+ free (line);
+
+ s += len;
+ if (*s == '\n')
+ s++;
+ }
}
/* Drops the documents from dictionary D. */
void
dict_clear_documents (struct dictionary *d)
{
- ds_clear (&d->documents);
+ string_array_clear (&d->documents);
}
-/* Appends LINE to the documents in D. LINE will be truncated or
- padded on the right with spaces to make it exactly
- DOC_LINE_LENGTH bytes long. */
-void
-dict_add_document_line (struct dictionary *d, const char *line)
+/* Appends the UTF-8 encoded LINE to the documents in D. LINE will be
+ truncated so that it is no more than 80 bytes in the dictionary's
+ encoding. If this causes some text to be lost, and ISSUE_WARNING is true,
+ then a warning will be issued. */
+bool
+dict_add_document_line (struct dictionary *d, const char *line,
+ bool issue_warning)
{
- if (strlen (line) > DOC_LINE_LENGTH)
+ size_t trunc_len;
+ bool truncated;
+
+ trunc_len = utf8_encoding_trunc_len (line, d->encoding, DOC_LINE_LENGTH);
+ truncated = line[trunc_len] != '\0';
+ if (truncated && issue_warning)
{
/* Note to translators: "bytes" is correct, not characters */
msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH);
}
- buf_copy_str_rpad (ds_put_uninit (&d->documents, DOC_LINE_LENGTH),
- DOC_LINE_LENGTH, line);
+
+ string_array_append_nocopy (&d->documents, xmemdup0 (line, trunc_len));
+
+ return !truncated;
}
/* Returns the number of document lines in dictionary D. */
size_t
dict_get_document_line_cnt (const struct dictionary *d)
{
- return ds_length (&d->documents) / DOC_LINE_LENGTH;
+ return d->documents.n;
}
-/* Copies document line number IDX from dictionary D into
- LINE, trimming off any trailing white space. */
-void
-dict_get_document_line (const struct dictionary *d,
- size_t idx, struct string *line)
+/* Returns document line number IDX in dictionary D. The caller must not
+ modify or free the returned string. */
+const char *
+dict_get_document_line (const struct dictionary *d, size_t idx)
{
- assert (idx < dict_get_document_line_cnt (d));
- ds_assign_substring (line, ds_substr (&d->documents, idx * DOC_LINE_LENGTH,
- DOC_LINE_LENGTH));
- ds_rtrim (line, ss_cstr (CC_SPACES));
+ assert (idx < d->documents.n);
+ return d->documents.strings[idx];
}
/* Creates in D a vector named NAME that contains the CNT
{
size_t i;
- assert (var != NULL);
assert (cnt > 0);
for (i = 0; i < cnt; i++)
assert (dict_contains_var (d, var[i]));
const struct vector *
dict_get_vector (const struct dictionary *d, size_t idx)
{
- assert (d != NULL);
assert (idx < d->vector_cnt);
return d->vector[idx];
size_t
dict_get_vector_cnt (const struct dictionary *d)
{
- assert (d != NULL);
-
return d->vector_cnt;
}
{
size_t i;
for (i = 0; i < d->vector_cnt; i++)
- if (!strcasecmp (vector_get_name (d->vector[i]), name))
+ if (!utf8_strcasecmp (vector_get_name (d->vector[i]), name))
return d->vector[i];
return NULL;
}
d->vector = NULL;
d->vector_cnt = 0;
}
+\f
+/* Multiple response sets. */
-/* Called from variable.c to notify the dictionary that some property of
- the variable has changed */
-void
-dict_var_changed (const struct variable *v)
+/* Returns the multiple response set in DICT with index IDX, which must be
+ between 0 and the count returned by dict_get_n_mrsets(), exclusive. */
+const struct mrset *
+dict_get_mrset (const struct dictionary *dict, size_t idx)
{
- if ( var_has_vardict (v))
+ assert (idx < dict->n_mrsets);
+ return dict->mrsets[idx];
+}
+
+/* Returns the number of multiple response sets in DICT. */
+size_t
+dict_get_n_mrsets (const struct dictionary *dict)
+{
+ return dict->n_mrsets;
+}
+
+/* Looks for a multiple response set named NAME in DICT. If it finds one,
+ returns its index; otherwise, returns SIZE_MAX. */
+static size_t
+dict_lookup_mrset_idx (const struct dictionary *dict, const char *name)
+{
+ size_t i;
+
+ for (i = 0; i < dict->n_mrsets; i++)
+ if (!utf8_strcasecmp (name, dict->mrsets[i]->name))
+ return i;
+
+ return SIZE_MAX;
+}
+
+/* Looks for a multiple response set named NAME in DICT. If it finds one,
+ returns it; otherwise, returns NULL. */
+const struct mrset *
+dict_lookup_mrset (const struct dictionary *dict, const char *name)
+{
+ size_t idx = dict_lookup_mrset_idx (dict, name);
+ return idx != SIZE_MAX ? dict->mrsets[idx] : NULL;
+}
+
+/* Adds MRSET to DICT, replacing any existing set with the same name. Returns
+ true if a set was replaced, false if none existed with the specified name.
+
+ Ownership of MRSET is transferred to DICT. */
+bool
+dict_add_mrset (struct dictionary *dict, struct mrset *mrset)
+{
+ size_t idx;
+
+ assert (mrset_ok (mrset, dict));
+
+ idx = dict_lookup_mrset_idx (dict, mrset->name);
+ if (idx == SIZE_MAX)
{
- const struct vardict_info *vdi = var_get_vardict (v);
- struct dictionary *d;
+ dict->mrsets = xrealloc (dict->mrsets,
+ (dict->n_mrsets + 1) * sizeof *dict->mrsets);
+ dict->mrsets[dict->n_mrsets++] = mrset;
+ return true;
+ }
+ else
+ {
+ mrset_destroy (dict->mrsets[idx]);
+ dict->mrsets[idx] = mrset;
+ return false;
+ }
+}
- d = vdi->dict;
+/* Looks for a multiple response set in DICT named NAME. If found, removes it
+ from DICT and returns true. If none is found, returns false without
+ modifying DICT.
- if ( d->callbacks && d->callbacks->var_changed )
- d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
+ Deleting one multiple response set causes the indexes of other sets within
+ DICT to change. */
+bool
+dict_delete_mrset (struct dictionary *dict, const char *name)
+{
+ size_t idx = dict_lookup_mrset_idx (dict, name);
+ if (idx != SIZE_MAX)
+ {
+ mrset_destroy (dict->mrsets[idx]);
+ dict->mrsets[idx] = dict->mrsets[--dict->n_mrsets];
+ return true;
+ }
+ else
+ return false;
+}
+
+/* Deletes all multiple response sets from DICT. */
+void
+dict_clear_mrsets (struct dictionary *dict)
+{
+ size_t i;
+
+ for (i = 0; i < dict->n_mrsets; i++)
+ mrset_destroy (dict->mrsets[i]);
+ free (dict->mrsets);
+ dict->mrsets = NULL;
+ dict->n_mrsets = 0;
+}
+
+/* Removes VAR, which must be in DICT, from DICT's multiple response sets. */
+static void
+dict_unset_mrset_var (struct dictionary *dict, struct variable *var)
+{
+ size_t i;
+
+ assert (dict_contains_var (dict, var));
+
+ for (i = 0; i < dict->n_mrsets; )
+ {
+ struct mrset *mrset = dict->mrsets[i];
+ size_t j;
+
+ for (j = 0; j < mrset->n_vars; )
+ if (mrset->vars[j] == var)
+ remove_element (mrset->vars, mrset->n_vars--,
+ sizeof *mrset->vars, j);
+ else
+ j++;
+
+ if (mrset->n_vars < 2)
+ {
+ mrset_destroy (mrset);
+ dict->mrsets[i] = dict->mrsets[--dict->n_mrsets];
+ }
+ else
+ i++;
}
}
+\f
+/* Returns D's attribute set. The caller may examine or modify
+ the attribute set, but must not destroy it. Destroying D or
+ calling dict_set_attributes for D will also destroy D's
+ attribute set. */
+struct attrset *
+dict_get_attributes (const struct dictionary *d)
+{
+ return CONST_CAST (struct attrset *, &d->attributes);
+}
+/* Replaces D's attributes set by a copy of ATTRS. */
+void
+dict_set_attributes (struct dictionary *d, const struct attrset *attrs)
+{
+ attrset_destroy (&d->attributes);
+ attrset_clone (&d->attributes, attrs);
+}
+
+/* Returns true if D has at least one attribute in its attribute
+ set, false if D's attribute set is empty. */
+bool
+dict_has_attributes (const struct dictionary *d)
+{
+ return attrset_count (&d->attributes) > 0;
+}
-/* Called from variable.c to notify the dictionary that the variable's width
- has changed */
+/* Called from variable.c to notify the dictionary that some property (indicated
+ by WHAT) of the variable has changed. OLDVAR is a copy of V as it existed
+ prior to the change. OLDVAR is destroyed by this function.
+*/
void
-dict_var_resized (const struct variable *v, int delta)
+dict_var_changed (const struct variable *v, unsigned int what, struct variable *oldvar)
{
if ( var_has_vardict (v))
{
- const struct vardict_info *vdi = var_get_vardict (v);
- struct dictionary *d;
+ const struct vardict_info *vardict = var_get_vardict (v);
+ struct dictionary *d = vardict->dict;
- d = vdi->dict;
+ if ( NULL == d)
+ return;
- dict_pad_values (d, var_get_case_index(v) + 1, delta);
+ if (d->changed ) d->changed (d, d->changed_data);
+ if ( d->callbacks && d->callbacks->var_changed )
+ d->callbacks->var_changed (d, var_get_dict_index (v), what, oldvar, d->cb_data);
+ }
+ var_destroy (oldvar);
+}
+
+
+\f
+/* Dictionary used to contain "internal variables". */
+static struct dictionary *internal_dict;
- if ( d->callbacks && d->callbacks->var_resized )
- d->callbacks->var_resized (d, var_get_dict_index (v), delta, d->cb_data);
+/* Create a variable of the specified WIDTH to be used for internal
+ calculations only. The variable is assigned case index CASE_IDX. */
+struct variable *
+dict_create_internal_var (int case_idx, int width)
+{
+ if (internal_dict == NULL)
+ internal_dict = dict_create ("UTF-8");
+
+ for (;;)
+ {
+ static int counter = INT_MAX / 2;
+ struct variable *var;
+ char name[64];
+
+ if (++counter == INT_MAX)
+ counter = INT_MAX / 2;
+
+ sprintf (name, "$internal%d", counter);
+ var = dict_create_var (internal_dict, name, width);
+ if (var != NULL)
+ {
+ set_var_case_index (var, case_idx);
+ return var;
+ }
}
}
+
+/* Destroys VAR, which must have been created with
+ dict_create_internal_var(). */
+void
+dict_destroy_internal_var (struct variable *var)
+{
+ if (var != NULL)
+ {
+ dict_delete_var (internal_dict, var);
+
+ /* Destroy internal_dict if it has no variables left, just so that
+ valgrind --leak-check --show-reachable won't show internal_dict. */
+ if (dict_get_var_cnt (internal_dict) == 0)
+ {
+ dict_destroy (internal_dict);
+ internal_dict = NULL;
+ }
+ }
+}
+\f
+int
+vardict_get_dict_index (const struct vardict_info *vardict)
+{
+ return vardict - vardict->dict->var;
+}
projects / pspp / blobdiff