X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Fdata%2Fdictionary.c;h=467f347efd9f5a5e290435265cbd1077032bfb0b;hb=d8fdf0b4fa919e48397b438e9453d6b82215ff51;hp=02c6aa76bc5c991a2bc077ed9bc91a3e60972eea;hpb=38c2638eed90f33225870563c4be183097e4155a;p=pspp-builds.git diff --git a/src/data/dictionary.c b/src/data/dictionary.c index 02c6aa76..467f347e 100644 --- a/src/data/dictionary.c +++ b/src/data/dictionary.c @@ -1,5 +1,5 @@ /* 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 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 @@ -16,32 +16,34 @@ #include -#include "dictionary.h" +#include "data/dictionary.h" +#include #include #include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include "intprops.h" -#include "minmax.h" -#include "xalloc.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/message.h" +#include "libpspp/misc.h" +#include "libpspp/pool.h" +#include "libpspp/str.h" + +#include "gl/intprops.h" +#include "gl/minmax.h" +#include "gl/xalloc.h" #include "gettext.h" #define _(msgid) gettext (msgid) @@ -49,9 +51,11 @@ /* 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. */ @@ -63,11 +67,57 @@ struct dictionary 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 *); + +void +dict_set_encoding (struct dictionary *d, const char *enc) +{ + if (enc) + { + free (d->encoding); + d->encoding = xstrdup (enc); + } +} + +const char * +dict_get_encoding (const struct dictionary *d) +{ + return d->encoding ; +} + + +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 @@ -76,9 +126,8 @@ 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", + 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)); @@ -115,8 +164,7 @@ dict_create (void) { struct dictionary *d = xzalloc (sizeof *d); - d->name_tab = hsh_create (8, compare_vars_by_name, hash_var_by_name, - NULL, NULL); + hmap_init (&d->name_map); attrset_init (&d->attributes); return d; } @@ -135,25 +183,18 @@ dict_clone (const struct dictionary *s) struct dictionary *d; size_t i; - assert (s != NULL); - d = dict_create (); 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)); + 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)); - svdi = var_get_vardict (sv); - dvdi = *svdi; - dvdi.dict = d; - var_set_vardict (dv, &dvdi); + var_get_vardict (dv)->case_index = var_get_vardict (sv)->case_index; } d->next_value_idx = s->next_value_idx; @@ -181,8 +222,25 @@ dict_clone (const struct dictionary *s) for (i = 0; i < s->vector_cnt; i++) d->vector[i] = vector_clone (s->vector[i], s, d); + if ( s->encoding) + d->encoding = xstrdup (s->encoding); + 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; } @@ -193,17 +251,16 @@ dict_clear (struct dictionary *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 ) { - dict_delete_var (d, d->var[d->var_cnt - 1]); + 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); @@ -223,10 +280,8 @@ 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]); + var_clear_aux (d->var[i].var); } /* Clears a dictionary and destroys it. */ @@ -240,8 +295,10 @@ dict_destroy (struct dictionary *d) d->callbacks = NULL ; dict_clear (d); - hsh_destroy (d->name_tab); + hmap_destroy (&d->name_map); attrset_destroy (&d->attributes); + dict_clear_mrsets (d); + free (d->encoding); free (d); } } @@ -250,8 +307,6 @@ dict_destroy (struct dictionary *d) size_t dict_get_var_cnt (const struct dictionary *d) { - assert (d != NULL); - return d->var_cnt; } @@ -261,10 +316,9 @@ dict_get_var_cnt (const struct dictionary *d) 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; } /* Sets *VARS to an array of pointers to variables in D and *CNT @@ -291,15 +345,12 @@ dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars, size_t count; size_t i; - assert (d != NULL); - assert (vars != NULL); - assert (cnt != NULL); assert (exclude == (exclude & DC_ALL)); count = 0; for (i = 0; i < d->var_cnt; i++) { - enum dict_class class = var_get_dict_class (d->var[i]); + enum dict_class class = var_get_dict_class (d->var[i].var); if (!(class & exclude)) count++; } @@ -308,9 +359,9 @@ dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars, *cnt = 0; for (i = 0; i < d->var_cnt; i++) { - enum dict_class class = var_get_dict_class (d->var[i]); + enum dict_class class = var_get_dict_class (d->var[i].var); if (!(class & exclude)) - (*vars)[(*cnt)++] = d->var[i]; + (*vars)[(*cnt)++] = d->var[i].var; } assert (*cnt == count); } @@ -318,26 +369,37 @@ dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars, static struct variable * add_var (struct dictionary *d, struct variable *v) { - /* 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; /* 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, + hash_case_string (var_get_name (v), 0)); + vardict->case_index = d->next_value_idx; + 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); + d->next_value_idx++; + invalidate_proto (d); return v; } @@ -363,27 +425,46 @@ dict_create_var_assert (struct dictionary *d, const char *name, int width) 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); @@ -396,23 +477,17 @@ dict_clone_var_assert (struct dictionary *d, const struct variable *old_var, struct variable * dict_lookup_var (const struct dictionary *d, const char *name) { - struct variable *target ; - struct variable *result ; - - if ( ! var_is_plausible_name (name, false)) - return NULL; - - target = var_create (name, 0); - result = hsh_find (d->name_tab, target); - var_destroy (target); + struct vardict_info *vardict; - if ( result && var_has_vardict (result)) - { - const struct vardict_info *vdi = var_get_vardict (result); - assert (vdi->dict == d); - } + HMAP_FOR_EACH_WITH_HASH (vardict, struct vardict_info, name_node, + hash_case_string (name, 0), &d->name_map) + { + struct variable *var = vardict->var; + if (!strcasecmp (var_get_name (var), name)) + return var; + } - return result; + return NULL; } /* Returns the variable named NAME in D. Assert-fails if no @@ -430,15 +505,8 @@ dict_lookup_var_assert (const struct dictionary *d, const char *name) 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 @@ -452,26 +520,45 @@ compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED) 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); +} +/* 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 *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 ( d->callbacks && d->callbacks->var_changed ) - d->callbacks->var_changed (d, dict_index, d->cb_data); + d->callbacks->var_changed (d, var_get_dict_index (var), d->cb_data); } /* 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 @@ -482,7 +569,7 @@ reindex_vars (struct dictionary *d, size_t from, size_t to) 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. @@ -503,7 +590,7 @@ dict_delete_var (struct dictionary *d, struct variable *v) { 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); + const int width = var_get_width (v); assert (dict_contains_var (d, v)); @@ -511,6 +598,7 @@ dict_delete_var (struct dictionary *d, struct variable *v) 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); @@ -521,23 +609,22 @@ dict_delete_var (struct dictionary *d, struct variable *v) 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, dict_index, case_index, width, d->cb_data); } /* Deletes the COUNT variables listed in VARS from D. This is @@ -548,7 +635,6 @@ dict_delete_vars (struct dictionary *d, { /* 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) @@ -566,7 +652,7 @@ dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count) 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. */ @@ -577,11 +663,9 @@ dict_delete_scratch_vars (struct 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 (var_get_dict_class (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++; } @@ -595,6 +679,8 @@ dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index) 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); } @@ -607,46 +693,50 @@ void 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 = 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 @@ -659,13 +749,14 @@ dict_rename_var (struct dictionary *d, struct variable *v, assert (!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 (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); } @@ -698,34 +789,37 @@ dict_rename_vars (struct dictionary *d, 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 (settings_get_algorithm () == ENHANCED) @@ -748,10 +842,10 @@ var_name_is_insertable (const struct dictionary *dict, const char *name) && lex_id_to_token (ss_cstr (name)) == T_ID); } -static bool -make_hinted_name (const struct dictionary *dict, const char *hint, - char name[VAR_NAME_LEN + 1]) +static char * +make_hinted_name (const struct dictionary *dict, const char *hint) { + char name[VAR_NAME_LEN + 1]; bool dropped = false; char *cp; @@ -780,7 +874,7 @@ make_hinted_name (const struct dictionary *dict, const char *hint, unsigned long int i; if (var_name_is_insertable (dict, name)) - return true; + return xstrdup (name); for (i = 0; i < ULONG_MAX; i++) { @@ -795,16 +889,15 @@ make_hinted_name (const struct dictionary *dict, const char *hint, strcpy (&name[ofs], suffix); if (var_name_is_insertable (dict, name)) - return true; + return xstrdup (name); } } - return false; + return NULL; } -static bool -make_numeric_name (const struct dictionary *dict, unsigned long int *num_start, - char name[VAR_NAME_LEN + 1]) +static char * +make_numeric_name (const struct dictionary *dict, unsigned long int *num_start) { unsigned long int number; @@ -812,27 +905,24 @@ make_numeric_name (const struct dictionary *dict, unsigned long int *num_start, 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 true; + return xstrdup (name); } } - if (num_start != NULL) - *num_start = ULONG_MAX; - return false; + NOT_REACHED (); } -/* Attempts to devise a variable name unique within DICT. - Returns true if successful, in which case the new variable - name is stored into NAME. Returns false if all names that can - be generated have already been taken. (Returning false is - quite unlikely: at least ULONG_MAX unique names can be - generated.) +/* 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 @@ -843,14 +933,18 @@ make_numeric_name (const struct dictionary *dict, unsigned long int *num_start, 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. - */ -bool +*/ +char * dict_make_unique_var_name (const struct dictionary *dict, const char *hint, - unsigned long int *num_start, - char name[VAR_NAME_LEN + 1]) + unsigned long int *num_start) { - return ((hint != NULL && make_hinted_name (dict, hint, name)) - || make_numeric_name (dict, num_start, name)); + 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 @@ -858,7 +952,6 @@ dict_make_unique_var_name (const struct dictionary *dict, const char *hint, 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; @@ -874,7 +967,6 @@ 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) @@ -899,12 +991,12 @@ dict_get_case_weight (const struct dictionary *d, const struct ccase *c, 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, @@ -916,7 +1008,6 @@ dict_set_weight (struct dictionary *d, struct variable *v) 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; @@ -927,12 +1018,12 @@ dict_get_filter (const struct dictionary *d) 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, @@ -944,8 +1035,6 @@ dict_set_filter (struct dictionary *d, struct variable *v) casenumber dict_get_case_limit (const struct dictionary *d) { - assert (d != NULL); - return d->case_limit; } @@ -954,19 +1043,34 @@ dict_get_case_limit (const struct dictionary *d) void 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; } @@ -975,8 +1079,6 @@ dict_get_next_value_idx (const struct dictionary *d) size_t dict_get_case_size (const struct dictionary *d) { - assert (d != NULL); - return sizeof (union value) * dict_get_next_value_idx (d); } @@ -990,38 +1092,12 @@ dict_compact_values (struct dictionary *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), @@ -1045,12 +1121,40 @@ dict_count_values (const struct dictionary *d, unsigned int exclude_classes) cnt = 0; for (i = 0; i < d->var_cnt; i++) { - enum dict_class class = var_get_dict_class (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; +} /* Returns the SPLIT FILE vars (see cmd_split_file()). Call dict_get_split_cnt() to determine how many SPLIT FILE vars @@ -1059,8 +1163,6 @@ dict_count_values (const struct dictionary *d, unsigned int exclude_classes) const struct variable *const * dict_get_split_vars (const struct dictionary *d) { - assert (d != NULL); - return d->split; } @@ -1068,14 +1170,12 @@ dict_get_split_vars (const struct dictionary *d) size_t dict_get_split_cnt (const struct dictionary *d) { - assert (d != NULL); - return d->split_cnt; } /* Removes variable V, which must be in D, from D's set of split variables. */ -void +static void dict_unset_split_var (struct dictionary *d, struct variable *v) { int orig_count; @@ -1087,6 +1187,7 @@ dict_unset_split_var (struct dictionary *d, struct variable *v) &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) @@ -1099,7 +1200,6 @@ void 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; @@ -1114,6 +1214,7 @@ dict_set_split_vars (struct dictionary *d, 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); } @@ -1123,20 +1224,18 @@ dict_set_split_vars (struct dictionary *d, 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. */ + characters. + + 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); - d->label = label != NULL ? xstrndup (label, 60) : NULL; + d->label = label != NULL && label[0] != '\0' ? xstrndup (label, 60) : NULL; } /* Returns the documents for D, or a null pointer if D has no @@ -1165,7 +1264,7 @@ dict_set_documents (struct dictionary *d, const char *documents) 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); + ds_put_byte_multiple (&d->documents, ' ', DOC_LINE_LENGTH - remainder); } /* Drops the documents from dictionary D. */ @@ -1187,7 +1286,7 @@ dict_add_document_line (struct dictionary *d, const char *line) 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); + DOC_LINE_LENGTH, line, ' '); } /* Returns the number of document lines in dictionary D. */ @@ -1219,7 +1318,6 @@ dict_create_vector (struct dictionary *d, { size_t i; - assert (var != NULL); assert (cnt > 0); for (i = 0; i < cnt; i++) assert (dict_contains_var (d, var[i])); @@ -1251,7 +1349,6 @@ dict_create_vector_assert (struct dictionary *d, 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]; @@ -1261,8 +1358,6 @@ dict_get_vector (const struct dictionary *d, size_t idx) size_t dict_get_vector_cnt (const struct dictionary *d) { - assert (d != NULL); - return d->vector_cnt; } @@ -1291,7 +1386,138 @@ dict_clear_vectors (struct dictionary *d) d->vector = NULL; d->vector_cnt = 0; } + +/* Multiple response sets. */ + +/* 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) +{ + 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 (!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) + { + 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; + } +} + +/* 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. + + 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++; + } +} + /* 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 @@ -1299,7 +1525,7 @@ dict_clear_vectors (struct dictionary *d) struct attrset * dict_get_attributes (const struct dictionary *d) { - return (struct attrset *) &d->attributes; + return CONST_CAST (struct attrset *, &d->attributes); } /* Replaces D's attributes set by a copy of ATTRS. */ @@ -1325,9 +1551,13 @@ dict_var_changed (const struct variable *v) { if ( var_has_vardict (v)) { - const struct vardict_info *vdi = var_get_vardict (v); - struct dictionary *d = vdi->dict; + const struct vardict_info *vardict = var_get_vardict (v); + struct dictionary *d = vardict->dict; + + if ( NULL == d) + return; + 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); } @@ -1337,19 +1567,21 @@ dict_var_changed (const struct variable *v) /* Called from variable.c to notify the dictionary that the variable's width has changed */ void -dict_var_resized (const struct variable *v, int delta) +dict_var_resized (const struct variable *v, int old_width) { if ( var_has_vardict (v)) { - const struct vardict_info *vdi = var_get_vardict (v); + const struct vardict_info *vardict = var_get_vardict (v); struct dictionary *d; - d = vdi->dict; + d = vardict->dict; - dict_pad_values (d, var_get_case_index(v) + 1, delta); + if (d->changed) d->changed (d, d->changed_data); + invalidate_proto (d); if ( d->callbacks && d->callbacks->var_resized ) - d->callbacks->var_resized (d, var_get_dict_index (v), delta, d->cb_data); + d->callbacks->var_resized (d, var_get_dict_index (v), old_width, + d->cb_data); } } @@ -1360,13 +1592,68 @@ dict_var_display_width_changed (const struct variable *v) { if ( var_has_vardict (v)) { - const struct vardict_info *vdi = var_get_vardict (v); + const struct vardict_info *vardict = var_get_vardict (v); struct dictionary *d; - d = vdi->dict; + d = vardict->dict; + if (d->changed) d->changed (d, d->changed_data); if ( d->callbacks && d->callbacks->var_display_width_changed ) d->callbacks->var_display_width_changed (d, var_get_dict_index (v), d->cb_data); } } + +/* Dictionary used to contain "internal variables". */ +static struct dictionary *internal_dict; +/* 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 (); + + 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; + } + } +} + +int +vardict_get_dict_index (const struct vardict_info *vardict) +{ + return vardict - vardict->dict->var; +}