X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Fdata%2Fdictionary.c;h=4fbcdf2a1e16f237b2988b10d80800a3752c7343;hb=b4062aaae1b2686f7d4ece15420550519e9a9823;hp=40352b0f2ba9e97708fb21c8aec422c75dddb4b7;hpb=96994a54e60e9c95b8bba54c2281acf7059b1203;p=pspp diff --git a/src/data/dictionary.c b/src/data/dictionary.c index 40352b0f2b..4fbcdf2a1e 100644 --- a/src/data/dictionary.c +++ b/src/data/dictionary.c @@ -1,5 +1,6 @@ /* PSPP - a program for statistical analysis. - Copyright (C) 1997-9, 2000, 2006, 2007, 2009, 2010, 2011, 2012, 2013, 2014, 2015 Free Software Foundation, Inc. + Copyright (C) 1997-9, 2000, 2006, 2007, 2009, 2010, 2011, 2012, 2013, 2014, + 2015, 2020 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 @@ -91,9 +92,83 @@ struct dictionary void *changed_data; }; -static void dict_unset_split_var (struct dictionary *, struct variable *); +static void dict_unset_split_var (struct dictionary *, struct variable *, bool); static void dict_unset_mrset_var (struct dictionary *, struct variable *); +/* Compares two double pointers to variables, which should point + to elements of a struct dictionary's `var' member array. */ +static int +compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED) +{ + struct variable *const *a = a_; + struct variable *const *b = b_; + + return *a < *b ? -1 : *a > *b; +} + +static void +unindex_var (struct dictionary *d, struct vardict_info *vardict) +{ + 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, bool skip_callbacks) +{ + 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 (! skip_callbacks) + { + 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_unref (old); + } + } +} + +/* Sets the case_index in V's vardict to CASE_INDEX. */ +static void +set_var_case_index (struct variable *v, int case_index) +{ + 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 + indexes from FROM to TO (exclusive). */ +static void +reindex_vars (struct dictionary *d, size_t from, size_t to, bool skip_callbacks) +{ + size_t i; + + for (i = from; i < to; i++) + reindex_var (d, &d->var[i], skip_callbacks); +} + + + /* Returns the encoding for data in dictionary D. The return value is a nonnull string that contains an IANA character set name. */ const char * @@ -175,7 +250,7 @@ dict_copy_callbacks (struct dictionary *dest, struct dictionary * dict_create (const char *encoding) { - struct dictionary *d = xzalloc (sizeof *d); + struct dictionary *d = XZALLOC (struct dictionary); d->encoding = xstrdup (encoding); d->names_must_be_ids = true; @@ -186,13 +261,6 @@ dict_create (const char *encoding) return d; } -struct dictionary * -dict_ref (struct dictionary *s) -{ - s->ref_cnt++; - return s; -} - /* Creates and returns a (deep) copy of an existing dictionary. @@ -268,506 +336,564 @@ dict_clone (const struct dictionary *s) return d; } -/* Clears the contents from a dictionary without destroying the - dictionary itself. */ -void -dict_clear (struct dictionary *d) + + +/* 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 + SPLIT FILE vars. */ +const struct variable *const * +dict_get_split_vars (const struct dictionary *d) { - /* FIXME? Should we really clear case_limit, label, documents? - Others are necessarily cleared by deleting all the variables.*/ - while (d->var_cnt > 0) - { - dict_delete_var (d, d->var[d->var_cnt - 1].var); - } + return d->split; +} - free (d->var); - d->var = NULL; - d->var_cnt = d->var_cap = 0; - 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); - dict_set_filter (d, NULL); - d->case_limit = 0; - free (d->label); - d->label = NULL; - string_array_clear (&d->documents); - dict_clear_vectors (d); - attrset_clear (&d->attributes); +/* Returns the number of SPLIT FILE vars. */ +size_t +dict_get_split_cnt (const struct dictionary *d) +{ + return d->split_cnt; } -/* Clears a dictionary and destroys it. */ +/* Removes variable V, which must be in D, from D's set of split + variables. */ static void -_dict_destroy (struct dictionary *d) +dict_unset_split_var (struct dictionary *d, struct variable *v, bool skip_callbacks) { - /* In general, we don't want callbacks occurring, if the dictionary - is being destroyed */ - d->callbacks = NULL ; + int orig_count; - dict_clear (d); - string_array_destroy (&d->documents); - hmap_destroy (&d->name_map); - attrset_destroy (&d->attributes); - dict_clear_mrsets (d); - free (d->encoding); - free (d); -} + assert (dict_contains_var (d, v)); -void -dict_unref (struct dictionary *d) -{ - if (d == NULL) - return; - d->ref_cnt--; - assert (d->ref_cnt >= 0); - if (d->ref_cnt == 0) - _dict_destroy (d); + 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 && !skip_callbacks) + { + 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); + } } -/* Returns the number of variables in D. */ -size_t -dict_get_var_cnt (const struct dictionary *d) -{ - return d->var_cnt; -} -/* Returns the variable in D with dictionary index IDX, which - must be between 0 and the count returned by - dict_get_var_cnt(), exclusive. */ -struct variable * -dict_get_var (const struct dictionary *d, size_t idx) +/* Sets CNT split vars SPLIT in dictionary D. */ +static void +dict_set_split_vars__ (struct dictionary *d, + struct variable *const *split, size_t cnt, bool skip_callbacks) { - assert (idx < d->var_cnt); + assert (cnt == 0 || split != NULL); - return d->var[idx].var; -} + d->split_cnt = cnt; + 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; + } -/* 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, enum dict_class exclude) -{ - dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude); + if (!skip_callbacks) + { + if (d->changed) d->changed (d, d->changed_data); + if (d->callbacks && d->callbacks->split_changed) + d->callbacks->split_changed (d, d->cb_data); + } } -/* 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.) */ +/* Sets CNT split vars SPLIT in dictionary D. */ void -dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars, - size_t *cnt, enum dict_class exclude) +dict_set_split_vars (struct dictionary *d, + struct variable *const *split, size_t cnt) { - size_t count; - size_t i; + dict_set_split_vars__ (d, split, cnt, false); +} - 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].var); - if (!(class & exclude)) - count++; - } +/* Deletes variable V from dictionary D and frees V. - *vars = xnmalloc (count, sizeof **vars); - *cnt = 0; - for (i = 0; i < d->var_cnt; 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); -} + This is a very bad idea if there might be any pointers to V + 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 DELETE VARIABLES. -static struct variable * -add_var_with_case_index (struct dictionary *d, struct variable *v, - int case_index) + Pointers to V within D are not a problem, because + dict_delete_var() knows to remove V from split variables, + weights, filters, etc. */ +static void +dict_delete_var__ (struct dictionary *d, struct variable *v, bool skip_callbacks) { - struct vardict_info *vardict; + int dict_index = var_get_dict_index (v); + const int case_index = var_get_case_index (v); - assert (case_index >= d->next_value_idx); + assert (dict_contains_var (d, v)); - /* Update dictionary. */ - if (d->var_cnt >= d->var_cap) - { - size_t i; + dict_unset_split_var (d, v, skip_callbacks); + dict_unset_mrset_var (d, v); - 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); - } - } + if (d->weight == v) + dict_set_weight (d, NULL); - 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->filter == v) + dict_set_filter (d, NULL); - 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); + dict_clear_vectors (d); - invalidate_proto (d); - d->next_value_idx = case_index + 1; + /* 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--; - return v; -} + /* Update dict_index for each affected variable. */ + reindex_vars (d, dict_index, d->var_cnt, skip_callbacks); -static struct variable * -add_var (struct dictionary *d, struct variable *v) -{ - return add_var_with_case_index (d, v, d->next_value_idx); -} + /* Free memory. */ + var_clear_vardict (v); -/* 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. */ -struct variable * -dict_create_var (struct dictionary *d, const char *name, int width) -{ - return (dict_lookup_var (d, name) == NULL - ? dict_create_var_assert (d, name, width) - : NULL); -} + if (! skip_callbacks) + { + if (d->changed) d->changed (d, d->changed_data); + if (d->callbacks && d->callbacks->var_deleted) + d->callbacks->var_deleted (d, v, dict_index, case_index, d->cb_data); + } -/* Creates and returns a new variable in D with the given NAME - and WIDTH. Assert-fails if the given NAME would duplicate - that of an existing variable in the dictionary. */ -struct variable * -dict_create_var_assert (struct dictionary *d, const char *name, int width) -{ - assert (dict_lookup_var (d, name) == NULL); - return add_var (d, var_create (name, width)); + invalidate_proto (d); + var_unref (v); } -/* 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)); -} +/* Deletes variable V from dictionary D and frees V. -/* 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_assert (struct dictionary *d, const struct variable *old_var) -{ - return dict_clone_var_as_assert (d, old_var, var_get_name (old_var)); -} + This is a very bad idea if there might be any pointers to V + 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 DELETE VARIABLES. -/* 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) + Pointers to V within D are not a problem, because + dict_delete_var() knows to remove V from split variables, + weights, filters, etc. */ +void +dict_delete_var (struct dictionary *d, struct variable *v) { - return (dict_lookup_var (d, name) == NULL - ? dict_clone_var_as_assert (d, old_var, name) - : NULL); + dict_delete_var__ (d, v, false); } -/* 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_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); - var_set_name (new_var, name); - return add_var (d, new_var); -} -struct variable * -dict_clone_var_in_place_assert (struct dictionary *d, - const struct variable *old_var) +/* Deletes the COUNT variables listed in VARS from D. This is + unsafe; see the comment on dict_delete_var() for details. */ +void +dict_delete_vars (struct dictionary *d, + struct variable *const *vars, size_t count) { - 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)); + /* FIXME: this can be done in O(count) time, but this algorithm + is O(count**2). */ + assert (count == 0 || vars != NULL); + + while (count-- > 0) + dict_delete_var (d, *vars++); } -/* 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) +/* Deletes the COUNT variables in D starting at index IDX. This + is unsafe; see the comment on dict_delete_var() for + details. Deleting consecutive vars will result in less callbacks + compared to iterating over dict_delete_var. + A simple while loop over dict_delete_var will + produce (d->var_cnt - IDX) * COUNT variable changed callbacks + plus COUNT variable delete callbacks. + This here produces d->var_cnt - IDX variable changed callbacks + plus COUNT variable delete callbacks. */ +void +dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count) { - struct vardict_info *vardict; + assert (idx + count <= d->var_cnt); - HMAP_FOR_EACH_WITH_HASH (vardict, struct vardict_info, name_node, - utf8_hash_case_string (name, 0), &d->name_map) + /* We need to store the variable and the corresponding case_index + for the delete callbacks later. We store them in a linked list.*/ + struct delvar { + struct ll ll; + struct variable *var; + int case_index; + }; + struct ll_list list = LL_INITIALIZER (list); + + for (size_t i = idx; i < idx + count; i++) { - struct variable *var = vardict->var; - if (!utf8_strcasecmp (var_get_name (var), name)) - return var; + struct delvar *dv = xmalloc (sizeof (struct delvar)); + assert (dv); + struct variable *v = d->var[i].var; + + dict_unset_split_var (d, v, false); + dict_unset_mrset_var (d, v); + + if (d->weight == v) + dict_set_weight (d, NULL); + + if (d->filter == v) + dict_set_filter (d, NULL); + + dv->var = v; + dv->case_index = var_get_case_index (v); + ll_push_tail (&list, (struct ll *)dv); } - return NULL; -} + dict_clear_vectors (d); -/* Returns the variable named NAME in D. Assert-fails if no - variable has that name. */ -struct variable * -dict_lookup_var_assert (const struct dictionary *d, const char *name) -{ - struct variable *v = dict_lookup_var (d, name); - assert (v != NULL); - return v; -} + /* Remove variables from var array. */ + unindex_vars (d, idx, d->var_cnt); + remove_range (d->var, d->var_cnt, sizeof *d->var, idx, count); + d->var_cnt -= count; -/* Returns true if variable V is in dictionary D, - false otherwise. */ -bool -dict_contains_var (const struct dictionary *d, const struct variable *v) -{ - return (var_has_vardict (v) - && vardict_get_dictionary (var_get_vardict (v)) == d); -} + /* Reindexing will result variable-changed callback */ + reindex_vars (d, idx, d->var_cnt, false); -/* Compares two double pointers to variables, which should point - to elements of a struct dictionary's `var' member array. */ -static int -compare_var_ptrs (const void *a_, const void *b_, const void *aux UNUSED) -{ - struct variable *const *a = a_; - struct variable *const *b = b_; + invalidate_proto (d); + if (d->changed) d->changed (d, d->changed_data); - return *a < *b ? -1 : *a > *b; + /* Now issue the variable delete callbacks and delete + the variables. The vardict is not valid at this point + anymore. That is the reason why we stored the + caseindex before reindexing. */ + for (size_t vi = idx; vi < idx + count; vi++) + { + struct delvar *dv = (struct delvar *) ll_pop_head (&list); + var_clear_vardict (dv->var); + if (d->callbacks && d->callbacks->var_deleted) + d->callbacks->var_deleted (d, dv->var, vi, dv->case_index, d->cb_data); + var_unref (dv->var); + free (dv); + } } -static void -unindex_var (struct dictionary *d, struct vardict_info *vardict) +/* Deletes scratch variables from dictionary D. */ +void +dict_delete_scratch_vars (struct dictionary *d) { - hmap_delete (&d->name_map, &vardict->name_node); + int i; + + /* FIXME: this can be done in O(count) time, but this algorithm + is O(count**2). */ + for (i = 0; i < d->var_cnt;) + if (var_get_dict_class (d->var[i].var) == DC_SCRATCH) + dict_delete_var (d, d->var[i].var); + else + i++; } -/* 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). */ + + +/* Clears the contents from a dictionary without destroying the + dictionary itself. */ static void -reindex_var (struct dictionary *d, struct vardict_info *vardict) +dict_clear__ (struct dictionary *d, bool skip_callbacks) { - struct variable *old = (d->callbacks && d->callbacks->var_changed - ? var_clone (vardict->var) - : NULL); + /* FIXME? Should we really clear case_limit, label, documents? + Others are necessarily cleared by deleting all the variables.*/ + while (d->var_cnt > 0) + { + dict_delete_var__ (d, d->var[d->var_cnt - 1].var, skip_callbacks); + } - struct variable *var = vardict->var; - var_set_vardict (var, vardict); - hmap_insert_fast (&d->name_map, &vardict->name_node, - vardict->name_node.hash); + free (d->var); + d->var = NULL; + d->var_cnt = d->var_cap = 0; + invalidate_proto (d); + hmap_clear (&d->name_map); + d->next_value_idx = 0; + dict_set_split_vars__ (d, NULL, 0, skip_callbacks); - if (d->changed) d->changed (d, d->changed_data); - if (old) + if (skip_callbacks) + { + d->weight = NULL; + d->filter = NULL; + } + else { - d->callbacks->var_changed (d, var_get_dict_index (var), VAR_TRAIT_POSITION, old, d->cb_data); - var_destroy (old); + dict_set_weight (d, NULL); + dict_set_filter (d, NULL); } + d->case_limit = 0; + free (d->label); + d->label = NULL; + string_array_clear (&d->documents); + dict_clear_vectors (d); + attrset_clear (&d->attributes); } -/* Sets the case_index in V's vardict to CASE_INDEX. */ -static void -set_var_case_index (struct variable *v, int case_index) +/* Clears the contents from a dictionary without destroying the + dictionary itself. */ +void +dict_clear (struct dictionary *d) { - var_get_vardict (v)->case_index = case_index; + dict_clear__ (d, false); } -/* Removes the dictionary variables with indexes from FROM to TO (exclusive) - from name_map. */ +/* Clears a dictionary and destroys it. */ static void -unindex_vars (struct dictionary *d, size_t from, size_t to) +_dict_destroy (struct dictionary *d) { - size_t i; + /* In general, we don't want callbacks occurring, if the dictionary + is being destroyed */ + d->callbacks = NULL ; - for (i = from; i < to; i++) - unindex_var (d, &d->var[i]); + dict_clear__ (d, true); + string_array_destroy (&d->documents); + hmap_destroy (&d->name_map); + attrset_destroy (&d->attributes); + dict_clear_mrsets (d); + free (d->encoding); + free (d); } -/* Re-sets the dict_index in the dictionary variables with - indexes from FROM to TO (exclusive). */ -static void -reindex_vars (struct dictionary *d, size_t from, size_t to) +struct dictionary * +dict_ref (struct dictionary *d) { - size_t i; + d->ref_cnt++; + return d; +} - for (i = from; i < to; i++) - reindex_var (d, &d->var[i]); +void +dict_unref (struct dictionary *d) +{ + if (d == NULL) + return; + d->ref_cnt--; + assert (d->ref_cnt >= 0); + if (d->ref_cnt == 0) + _dict_destroy (d); } -/* Deletes variable V from dictionary D and frees V. +/* Returns the number of variables in D. */ +size_t +dict_get_var_cnt (const struct dictionary *d) +{ + return d->var_cnt; +} - This is a very bad idea if there might be any pointers to V - 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 DELETE VARIABLES. +/* Returns the variable in D with dictionary index IDX, which + must be between 0 and the count returned by + dict_get_var_cnt(), exclusive. */ +struct variable * +dict_get_var (const struct dictionary *d, size_t idx) +{ + assert (idx < d->var_cnt); - Pointers to V within D are not a problem, because - dict_delete_var() knows to remove V from split variables, - weights, filters, etc. */ + return d->var[idx].var; +} + +/* 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_delete_var (struct dictionary *d, struct variable *v) +dict_get_vars (const struct dictionary *d, const struct variable ***vars, + size_t *cnt, enum dict_class exclude) { - int dict_index = var_get_dict_index (v); - const int case_index = var_get_case_index (v); + dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude); +} - assert (dict_contains_var (d, v)); +/* 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_mutable (const struct dictionary *d, struct variable ***vars, + size_t *cnt, enum dict_class exclude) +{ + size_t count; + size_t i; - dict_unset_split_var (d, v); - dict_unset_mrset_var (d, v); + assert (exclude == (exclude & DC_ALL)); - if (d->weight == v) - dict_set_weight (d, NULL); + count = 0; + for (i = 0; i < d->var_cnt; i++) + { + enum dict_class class = var_get_dict_class (d->var[i].var); + if (!(class & exclude)) + count++; + } - if (d->filter == v) - dict_set_filter (d, NULL); + *vars = xnmalloc (count, sizeof **vars); + *cnt = 0; + for (i = 0; i < d->var_cnt; 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); +} - dict_clear_vectors (d); +static struct variable * +add_var_with_case_index (struct dictionary *d, struct variable *v, + int case_index) +{ + struct vardict_info *vardict; - /* 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--; + assert (case_index >= d->next_value_idx); - /* Update dict_index for each affected variable. */ - reindex_vars (d, dict_index, d->var_cnt); + /* Update dictionary. */ + if (d->var_cnt >= d->var_cap) + { + size_t i; - /* Free memory. */ - var_clear_vardict (v); + 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); + } + } + + 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); invalidate_proto (d); - if (d->callbacks && d->callbacks->var_deleted) - d->callbacks->var_deleted (d, v, dict_index, case_index, d->cb_data); + d->next_value_idx = case_index + 1; - var_destroy (v); + return v; } -/* Deletes the COUNT variables listed in VARS from D. This is - unsafe; see the comment on dict_delete_var() for details. */ -void -dict_delete_vars (struct dictionary *d, - struct variable *const *vars, size_t count) +static struct variable * +add_var (struct dictionary *d, struct variable *v) { - /* FIXME: this can be done in O(count) time, but this algorithm - is O(count**2). */ - assert (count == 0 || vars != NULL); - - while (count-- > 0) - dict_delete_var (d, *vars++); + return add_var_with_case_index (d, v, d->next_value_idx); } -/* Deletes the COUNT variables in D starting at index IDX. This - is unsafe; see the comment on dict_delete_var() for - details. Deleting consecutive vars will result in less callbacks - compared to iterating over dict_delete_var. - A simple while loop over dict_delete_var will - produce (d->var_cnt - IDX) * COUNT variable changed callbacks - plus COUNT variable delete callbacks. - This here produces d->var_cnt - IDX variable changed callbacks - plus COUNT variable delete callbacks. */ -void -dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count) +/* 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. */ +struct variable * +dict_create_var (struct dictionary *d, const char *name, int width) { - assert (idx + count <= d->var_cnt); - - /* We need to store the variable and the corresponding case_index - for the delete callbacks later. We store them in a linked list.*/ - struct delvar { - struct ll ll; - struct variable *var; - int case_index; - }; - struct ll_list list = LL_INITIALIZER (list); - - for (size_t i = idx; i < idx + count; i++) - { - struct delvar *dv = xmalloc (sizeof (struct delvar)); - assert (dv); - struct variable *v = d->var[i].var; - - dict_unset_split_var (d, v); - dict_unset_mrset_var (d, v); + return (dict_lookup_var (d, name) == NULL + ? dict_create_var_assert (d, name, width) + : NULL); +} - if (d->weight == v) - dict_set_weight (d, NULL); +/* Creates and returns a new variable in D with the given NAME + and WIDTH. Assert-fails if the given NAME would duplicate + that of an existing variable in the dictionary. */ +struct variable * +dict_create_var_assert (struct dictionary *d, const char *name, int width) +{ + assert (dict_lookup_var (d, name) == NULL); + return add_var (d, var_create (name, width)); +} - if (d->filter == v) - dict_set_filter (d, NULL); +/* 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)); +} - dv->var = v; - dv->case_index = var_get_case_index (v); - ll_push_tail (&list, (struct ll *)dv); - } +/* 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_assert (struct dictionary *d, const struct variable *old_var) +{ + return dict_clone_var_as_assert (d, old_var, var_get_name (old_var)); +} - dict_clear_vectors (d); +/* 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_as_assert (d, old_var, name) + : NULL); +} - /* Remove variables from var array. */ - unindex_vars (d, idx, d->var_cnt); - remove_range (d->var, d->var_cnt, sizeof *d->var, idx, count); - d->var_cnt -= count; +/* 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_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); + var_set_name (new_var, name); + return add_var (d, new_var); +} - /* Reindexing will result variable-changed callback */ - reindex_vars (d, idx, d->var_cnt); +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)); +} - invalidate_proto (d); - if (d->changed) d->changed (d, d->changed_data); +/* 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 vardict_info *vardict; - /* Now issue the variable delete callbacks and delete - the variables. The vardict is not valid at this point - anymore. That is the reason why we stored the - caseindex before reindexing. */ - for (size_t vi = idx; vi < idx + count; vi++) + HMAP_FOR_EACH_WITH_HASH (vardict, struct vardict_info, name_node, + utf8_hash_case_string (name, 0), &d->name_map) { - struct delvar *dv = (struct delvar *) ll_pop_head (&list); - var_clear_vardict (dv->var); - if (d->callbacks && d->callbacks->var_deleted) - d->callbacks->var_deleted (d, dv->var, vi, dv->case_index, d->cb_data); - var_destroy (dv->var); - free (dv); + struct variable *var = vardict->var; + if (!utf8_strcasecmp (var_get_name (var), name)) + return var; } + + return NULL; } -/* Deletes scratch variables from dictionary D. */ -void -dict_delete_scratch_vars (struct dictionary *d) +/* Returns the variable named NAME in D. Assert-fails if no + variable has that name. */ +struct variable * +dict_lookup_var_assert (const struct dictionary *d, const char *name) { - int i; + struct variable *v = dict_lookup_var (d, name); + assert (v != NULL); + return v; +} - /* FIXME: this can be done in O(count) time, but this algorithm - is O(count**2). */ - for (i = 0; i < d->var_cnt;) - if (var_get_dict_class (d->var[i].var) == DC_SCRATCH) - dict_delete_var (d, d->var[i].var); - else - i++; +/* Returns true if variable V is in dictionary D, + false otherwise. */ +bool +dict_contains_var (const struct dictionary *d, const struct variable *v) +{ + return (var_has_vardict (v) + && vardict_get_dictionary (var_get_vardict (v)) == d); } /* Moves V to 0-based position IDX in D. Other variables in D, @@ -782,7 +908,7 @@ dict_reorder_var (struct dictionary *d, struct variable *v, size_t new_index) 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); + reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1, false); } /* Reorders the variables in D, placing the COUNT variables @@ -824,7 +950,7 @@ dict_reorder_vars (struct dictionary *d, d->var = new_var; hmap_clear (&d->name_map); - reindex_vars (d, 0, d->var_cnt); + reindex_vars (d, 0, d->var_cnt, false); } /* Changes the name of variable V that is currently in a dictionary to @@ -853,7 +979,7 @@ dict_try_rename_var (struct dictionary *d, struct variable *v, struct variable *old = var_clone (v); unindex_var (d, var_get_vardict (v)); rename_var (v, new_name); - reindex_var (d, var_get_vardict (v)); + reindex_var (d, var_get_vardict (v), false); if (settings_get_algorithm () == ENHANCED) var_clear_short_names (v); @@ -862,7 +988,7 @@ dict_try_rename_var (struct dictionary *d, struct variable *v, if (d->callbacks && d->callbacks->var_changed) d->callbacks->var_changed (d, var_get_dict_index (v), VAR_TRAIT_NAME, old, d->cb_data); - var_destroy (old); + var_unref (old); return true; } @@ -929,13 +1055,13 @@ dict_rename_vars (struct dictionary *d, for (i = 0; i < count; i++) { rename_var (vars[i], old_names[i]); - reindex_var (d, var_get_vardict (vars[i])); + reindex_var (d, var_get_vardict (vars[i]), false); } pool_destroy (pool); return false; } - reindex_var (d, var_get_vardict (vars[i])); + reindex_var (d, var_get_vardict (vars[i]), false); } /* Clear short names. */ @@ -968,6 +1094,9 @@ make_hinted_name (const struct dictionary *dict, const char *hint) size_t ofs; int mblen; + if (hint_len > ID_MAX_LEN) + hint_len = ID_MAX_LEN; + /* 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 @@ -1074,6 +1203,7 @@ dict_make_unique_var_name (const struct dictionary *dict, const char *hint, if (hinted_name != NULL) return hinted_name; } + return make_numeric_name (dict, num_start); } @@ -1150,8 +1280,8 @@ dict_set_weight (struct dictionary *d, struct variable *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, - d->cb_data); + v ? var_get_dict_index (v) : -1, + d->cb_data); } /* Returns the filter variable in dictionary D (see cmd_filter()) @@ -1172,29 +1302,13 @@ dict_set_filter (struct dictionary *d, struct variable *v) assert (v == NULL || dict_contains_var (d, v)); assert (v == NULL || var_is_numeric (v)); - /* When a filter is set, we ref the dictionary. - This is because the GUI maintains a pointer - to the dict's variables, and the variables' - addresses change in the callback. */ - if (d->filter == NULL && v != NULL) - { - d = dict_ref (d); - } - - /* Deref the dict when a filter is removed. */ - if (d->filter != NULL && v == NULL) - { - assert (d->ref_cnt > 0); - dict_unref (d); - } - 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, - d->cb_data); + v ? var_get_dict_index (v) : -1, + d->cb_data); } /* Returns the case limit for dictionary D, or zero if the number @@ -1322,70 +1436,6 @@ dict_get_compacted_proto (const struct dictionary *d, } return proto; } - -/* 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 - SPLIT FILE vars. */ -const struct variable *const * -dict_get_split_vars (const struct dictionary *d) -{ - return d->split; -} - -/* Returns the number of SPLIT FILE vars. */ -size_t -dict_get_split_cnt (const struct dictionary *d) -{ - return d->split_cnt; -} - -/* 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) -{ - int orig_count; - - assert (dict_contains_var (d, v)); - - 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. */ -void -dict_set_split_vars (struct dictionary *d, - struct variable *const *split, size_t cnt) -{ - assert (cnt == 0 || split != NULL); - - d->split_cnt = cnt; - 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); -} - /* Returns the file label for D, or a null pointer if D is unlabeled (see cmd_file_label()). */ const char * @@ -1421,12 +1471,15 @@ dict_get_documents (const struct dictionary *d) void dict_set_documents (struct dictionary *d, const struct string_array *new_docs) { - size_t i; + /* Swap out the old documents, instead of destroying them immediately, to + allow the new documents to include pointers into the old ones. */ + struct string_array old_docs = STRING_ARRAY_INITIALIZER; + string_array_swap (&d->documents, &old_docs); - dict_clear_documents (d); - - for (i = 0; i < new_docs->n; i++) + for (size_t i = 0; i < new_docs->n; i++) dict_add_document_line (d, new_docs->strings[i], false); + + string_array_destroy (&old_docs); } /* Replaces the documents for D by UTF-8 encoded string NEW_DOCS, dividing it @@ -1749,11 +1802,14 @@ dict_var_changed (const struct variable *v, unsigned int what, struct variable * if (NULL == d) return; + if (what & (VAR_TRAIT_WIDTH | VAR_TRAIT_POSITION)) + invalidate_proto (d); + 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); + d->callbacks->var_changed (d, var_get_dict_index (v), what, oldvar, d->cb_data); } - var_destroy (oldvar); + var_unref (oldvar); }