#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/str.h>
#include "minmax.h"
+#include "xalloc.h"
#include "gettext.h"
#define _(msgid) gettext (msgid)
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 vector **vector; /* Vectors of variables. */
void *cb_data ; /* Data passed to callbacks */
};
+/* Print a representation of dictionary D to stdout, for
+ debugging purposes. */
+void
+dict_dump (const struct dictionary *d)
+{
+ 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",
+ var_get_name (v),
+ var_get_dict_index (v),
+ var_get_case_index (v));
+
+ }
+}
/* Associate CALLBACKS with DICT. Callbacks will be invoked whenever
the dictionary or any of the variables it contains are modified.
}
/* 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. */
struct dictionary *
dict_clone (const struct dictionary *s)
{
for (i = 0; i < s->var_cnt; i++)
{
+ const struct vardict_info *svdi;
+ struct vardict_info dvdi;
struct variable *sv = s->var[i];
struct variable *dv = dict_clone_var_assert (d, sv, var_get_name (sv));
- var_set_short_name (dv, var_get_short_name (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));
+
+ svdi = var_get_vardict (sv);
+ dvdi = *svdi;
+ dvdi.dict = d;
+ var_set_vardict (dv, &dvdi);
}
d->next_value_idx = s->next_value_idx;
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]);
}
free (d->var);
return d->var[idx];
}
-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]);
+ 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)))
+ enum dict_class class = var_get_dict_class (d->var[i]);
+ if (!(class & exclude))
(*vars)[(*cnt)++] = d->var[i];
}
assert (*cnt == count);
struct variable *target ;
struct variable *result ;
- if ( ! var_is_valid_name (name, false))
+ if ( ! var_is_plausible_name (name, false))
return NULL;
target = var_create (name, 0);
result = hsh_find (d->name_tab, target);
var_destroy (target);
+ if ( result && var_has_vardict (result))
+ {
+ const struct vardict_info *vdi = var_get_vardict (result);
+ assert (vdi->dict == d);
+ }
+
return result;
}
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,
var_clear_vardict (v);
var_destroy (v);
+
if (d->callbacks && d->callbacks->var_deleted )
d->callbacks->var_deleted (d, dict_index, case_index, val_cnt, d->cb_data);
}
assert (d != NULL);
for (i = 0; i < d->var_cnt; )
- if (dict_class_from_id (var_get_name (d->var[i])) == DC_SCRATCH)
+ if (var_get_dict_class (d->var[i]) == DC_SCRATCH)
dict_delete_var (d, d->var[i]);
else
i++;
rename_var (d, v, new_name);
hsh_force_insert (d->name_tab, v);
- if (get_algorithm () == ENHANCED)
- var_clear_short_name (v);
+ if (settings_get_algorithm () == ENHANCED)
+ var_clear_short_names (v);
if ( d->callbacks && d->callbacks->var_changed )
d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
}
/* Clear short names. */
- if (get_algorithm () == ENHANCED)
+ if (settings_get_algorithm () == ENHANCED)
for (i = 0; i < count; i++)
- var_clear_short_name (vars[i]);
+ var_clear_short_names (vars[i]);
pool_destroy (pool);
return true;
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 (v == NULL || dict_contains_var (d, v));
+ assert (v == NULL || var_is_numeric (v));
d->filter = v;
/* 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);
/* 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);
return sizeof (union value) * dict_get_next_value_idx (d);
}
-/* Deletes scratch variables in dictionary D and reassigns values
- so that fragmentation is eliminated. */
+/* Reassigns values in dictionary D so that fragmentation is
+ eliminated. */
void
dict_compact_values (struct dictionary *d)
{
size_t i;
d->next_value_idx = 0;
- for (i = 0; i < d->var_cnt; )
+ for (i = 0; i < d->var_cnt; i++)
{
struct variable *v = d->var[i];
-
- if (dict_class_from_id (var_get_name (v)) != DC_SCRATCH)
- {
- set_var_case_index (v, d->next_value_idx);
- d->next_value_idx += var_get_value_cnt (v);
- i++;
- }
- else
- dict_delete_var (d, v);
+ 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.
}
-/* Returns the number of values that would be used by a case if
- dict_compact_values() were called. */
+/* 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),
+ (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to exclude the
+ corresponding type of variable.
+
+ The return value may be less than the number of values in one
+ of dictionary D's cases (as returned by
+ dict_get_next_value_idx) even if E is 0, because there may be
+ gaps in D's cases due to deleted variables. */
size_t
-dict_get_compacted_value_cnt (const struct dictionary *d)
+dict_count_values (const struct dictionary *d, unsigned int exclude_classes)
{
size_t i;
size_t cnt;
- cnt = 0;
- for (i = 0; i < d->var_cnt; i++)
- if (dict_class_from_id (var_get_name (d->var[i])) != DC_SCRATCH)
- cnt += var_get_value_cnt (d->var[i]);
- return cnt;
-}
-
-/* Creates and returns an array mapping from a dictionary index
- to the case index that the corresponding variable will have
- after calling dict_compact_values(). Scratch variables
- receive -1 for case index because dict_compact_values() will
- delete them. */
-int *
-dict_get_compacted_dict_index_to_case_index (const struct dictionary *d)
-{
- size_t i;
- size_t next_value_idx;
- int *map;
+ assert ((exclude_classes & ~((1u << DC_ORDINARY)
+ | (1u << DC_SYSTEM)
+ | (1u << DC_SCRATCH))) == 0);
- map = xnmalloc (d->var_cnt, sizeof *map);
- next_value_idx = 0;
+ cnt = 0;
for (i = 0; i < d->var_cnt; i++)
{
- struct variable *v = d->var[i];
-
- if (dict_class_from_id (var_get_name (v)) != DC_SCRATCH)
- {
- map[i] = next_value_idx;
- next_value_idx += var_get_value_cnt (v);
- }
- else
- map[i] = -1;
- }
- return map;
-}
-
-/* Returns true if a case for dictionary D would be smaller after
- compacting, false otherwise. Compacting a case eliminates
- "holes" between values and after the last value. Holes are
- created by deleting variables (or by scratch variables).
-
- The return value may differ from whether compacting a case
- from dictionary D would *change* the case: compacting could
- rearrange values even if it didn't reduce space
- requirements. */
-bool
-dict_compacting_would_shrink (const struct dictionary *d)
-{
- return dict_get_compacted_value_cnt (d) < dict_get_next_value_idx (d);
-}
-
-/* Returns true if a case for dictionary D would change after
- compacting, false otherwise. Compacting a case eliminates
- "holes" between values and after the last value. Holes are
- created by deleting variables (or by scratch variables).
-
- The return value may differ from whether compacting a case
- from dictionary D would *shrink* the case: compacting could
- rearrange values without reducing space requirements. */
-bool
-dict_compacting_would_change (const struct dictionary *d)
-{
- size_t case_idx;
- size_t i;
-
- case_idx = 0;
- for (i = 0; i < dict_get_var_cnt (d); i++)
- {
- struct variable *v = dict_get_var (d, i);
- if (var_get_case_index (v) != case_idx)
- return true;
- case_idx += var_get_value_cnt (v);
+ enum dict_class class = var_get_dict_class (d->var[i]);
+ if (!(exclude_classes & (1u << class)))
+ cnt += var_get_value_cnt (d->var[i]);
}
- return false;
+ return cnt;
}
\f
-/* How to copy a contiguous range of values between cases. */
-struct copy_map
- {
- size_t src_idx; /* Starting value index in source case. */
- size_t dst_idx; /* Starting value index in target case. */
- size_t cnt; /* Number of values. */
- };
-
-/* How to compact a case. */
-struct dict_compactor
- {
- struct copy_map *maps; /* Array of mappings. */
- size_t map_cnt; /* Number of mappings. */
- };
-
-/* Creates and returns a dict_compactor that can be used to
- compact cases for dictionary D.
-
- Compacting a case eliminates "holes" between values and after
- the last value. Holes are created by deleting variables (or
- by scratch variables). */
-struct dict_compactor *
-dict_make_compactor (const struct dictionary *d)
-{
- struct dict_compactor *compactor;
- struct copy_map *map;
- size_t map_allocated;
- size_t value_idx;
- size_t i;
-
- compactor = xmalloc (sizeof *compactor);
- compactor->maps = NULL;
- compactor->map_cnt = 0;
- map_allocated = 0;
-
- value_idx = 0;
- map = NULL;
- for (i = 0; i < d->var_cnt; i++)
- {
- struct variable *v = d->var[i];
-
- if (dict_class_from_id (var_get_name (v)) == DC_SCRATCH)
- continue;
- if (map != NULL && map->src_idx + map->cnt == var_get_case_index (v))
- map->cnt += var_get_value_cnt (v);
- else
- {
- if (compactor->map_cnt == map_allocated)
- compactor->maps = x2nrealloc (compactor->maps, &map_allocated,
- sizeof *compactor->maps);
- map = &compactor->maps[compactor->map_cnt++];
- map->src_idx = var_get_case_index (v);
- map->dst_idx = value_idx;
- map->cnt = var_get_value_cnt (v);
- }
- value_idx += var_get_value_cnt (v);
- }
-
- return compactor;
-}
-
-/* Compacts SRC by copying it to DST according to the scheme in
- COMPACTOR.
-
- Compacting a case eliminates "holes" between values and after
- the last value. Holes are created by deleting variables (or
- by scratch variables). */
-void
-dict_compactor_compact (const struct dict_compactor *compactor,
- struct ccase *dst, const struct ccase *src)
-{
- size_t i;
-
- for (i = 0; i < compactor->map_cnt; i++)
- {
- const struct copy_map *map = &compactor->maps[i];
- case_copy (dst, map->dst_idx, src, map->src_idx, map->cnt);
- }
-}
-
-/* Destroys COMPACTOR. */
-void
-dict_compactor_destroy (struct dict_compactor *compactor)
-{
- if (compactor != NULL)
- {
- free (compactor->maps);
- free (compactor);
- }
-}
-
/* Returns the SPLIT FILE vars (see cmd_split_file()). Call
dict_get_split_cnt() to determine how many SPLIT FILE vars
there are. Returns a null pointer if and only if there are no
return d->split_cnt;
}
-/* Removes variable V from the set of split variables in dictionary D */
+/* Removes variable V, which must be in D, from D's set of split
+ variables. */
void
-dict_unset_split_var (struct dictionary *d,
- struct variable *v)
+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)
+ {
+ /* 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. */
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->callbacks && d->callbacks->split_changed )
d->callbacks->split_changed (d, d->cb_data);
assert (d != NULL);
free (d->label);
- if (label == NULL)
- 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 = label != NULL ? xstrndup (label, 60) : NULL;
}
/* Returns the documents for D, or a null pointer if D has no
d->vector_cnt = 0;
}
-/* Compares two strings. */
-static int
-compare_strings (const void *a, const void *b, const void *aux UNUSED)
-{
- return strcmp (a, b);
-}
-
-/* Hashes a string. */
-static unsigned
-hash_string (const void *s, const void *aux UNUSED)
-{
- return hsh_hash_string (s);
-}
-
-
-/* Sets V's short name to BASE, followed by a suffix of the form
- _A, _B, _C, ..., _AA, _AB, etc. according to the value of
- SUFFIX_NUMBER. Truncates BASE as necessary to fit. */
-static void
-set_var_short_name_suffix (struct variable *v, const char *base,
- int suffix_number)
-{
- char suffix[SHORT_NAME_LEN + 1];
- char short_name[SHORT_NAME_LEN + 1];
- char *start, *end;
- int len, ofs;
-
- assert (v != NULL);
- assert (suffix_number >= 0);
-
- /* Set base name. */
- var_set_short_name (v, base);
-
- /* Compose suffix. */
- start = end = suffix + sizeof suffix - 1;
- *end = '\0';
- do
- {
- *--start = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"[suffix_number % 26];
- if (start <= suffix + 1)
- msg (SE, _("Variable suffix too large."));
- suffix_number /= 26;
- }
- while (suffix_number > 0);
- *--start = '_';
-
- /* Append suffix to V's short name. */
- str_copy_trunc (short_name, sizeof short_name, base);
- len = end - start;
- if (len + strlen (short_name) > SHORT_NAME_LEN)
- ofs = SHORT_NAME_LEN - len;
- else
- ofs = strlen (short_name);
- strcpy (short_name + ofs, start);
-
- /* Set name. */
- var_set_short_name (v, short_name);
-}
-
-/* Assigns a valid, unique short_name[] to each variable in D.
- Each variable whose actual name is short has highest priority
- for that short name. Otherwise, variables with an existing
- short_name[] have the next highest priority for a given short
- name; if it is already taken, then the variable is treated as
- if short_name[] had been empty. Otherwise, long names are
- truncated to form short names. If that causes conflicts,
- variables are renamed as PREFIX_A, PREFIX_B, and so on. */
-void
-dict_assign_short_names (struct dictionary *d)
-{
- struct hsh_table *short_names;
- size_t i;
-
- /* Give variables whose names are short the corresponding short
- names, and clear short_names[] that conflict with a variable
- name. */
- for (i = 0; i < d->var_cnt; i++)
- {
- struct variable *v = d->var[i];
- const char *short_name = var_get_short_name (v);
- if (strlen (var_get_name (v)) <= SHORT_NAME_LEN)
- var_set_short_name (v, var_get_name (v));
- else if (short_name != NULL && dict_lookup_var (d, short_name) != NULL)
- var_clear_short_name (v);
- }
-
- /* Each variable with an assigned short_name[] now gets it
- unless there is a conflict. */
- short_names = hsh_create (d->var_cnt, compare_strings, hash_string,
- NULL, NULL);
- for (i = 0; i < d->var_cnt; i++)
- {
- struct variable *v = d->var[i];
- const char *name = var_get_short_name (v);
- if (name != NULL && hsh_insert (short_names, (char *) name) != NULL)
- var_clear_short_name (v);
- }
-
- /* Now assign short names to remaining variables. */
- for (i = 0; i < d->var_cnt; i++)
- {
- struct variable *v = d->var[i];
- const char *name = var_get_short_name (v);
- if (name == NULL)
- {
- /* Form initial short_name from the variable name, then
- try _A, _B, ... _AA, _AB, etc., if needed.*/
- int trial = 0;
- do
- {
- if (trial == 0)
- var_set_short_name (v, var_get_name (v));
- else
- set_var_short_name_suffix (v, var_get_name (v), trial - 1);
-
- trial++;
- }
- while (hsh_insert (short_names, (char *) var_get_short_name (v))
- != NULL);
- }
- }
-
- /* Get rid of hash table. */
- hsh_destroy (short_names);
-}
-
-
/* Called from variable.c to notify the dictionary that some property of
the variable has changed */
void
if ( var_has_vardict (v))
{
const struct vardict_info *vdi = var_get_vardict (v);
- struct dictionary *d;
-
- d = vdi->dict;
+ struct dictionary *d = vdi->dict;
if ( d->callbacks && d->callbacks->var_changed )
d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);
projects / pspp-builds.git / blobdiff