X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Fdata%2Fdictionary.c;h=c13321c7827bb2cd877c598d834f5bbc2a341c6e;hb=24cbabb1547682037cda854b46c2e4bdf87a2c8a;hp=1c50ac7feb26976ea1e9b63d835961a5ae3db243;hpb=503f53bfdde87fc40466dadb77bc04cee0be2567;p=pspp diff --git a/src/data/dictionary.c b/src/data/dictionary.c index 1c50ac7feb..c13321c782 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 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 @@ -16,32 +17,39 @@ #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 + +#include "data/attributes.h" +#include "data/case.h" +#include "data/identifier.h" +#include "data/mrset.h" +#include "data/settings.h" +#include "data/value-labels.h" +#include "data/vardict.h" +#include "data/variable.h" +#include "data/vector.h" +#include "libpspp/array.h" +#include "libpspp/assertion.h" +#include "libpspp/compiler.h" +#include "libpspp/hash-functions.h" +#include "libpspp/hmap.h" +#include "libpspp/i18n.h" +#include "libpspp/message.h" +#include "libpspp/misc.h" +#include "libpspp/pool.h" +#include "libpspp/str.h" +#include "libpspp/string-array.h" +#include "libpspp/ll.h" + +#include "gl/intprops.h" +#include "gl/minmax.h" +#include "gl/xalloc.h" +#include "gl/xmemdup0.h" #include "gettext.h" #define _(msgid) gettext (msgid) @@ -49,20 +57,32 @@ /* A dictionary. */ struct dictionary { - struct variable **var; /* Variables. */ - size_t var_cnt, var_cap; /* Number of variables, capacity. */ - struct hsh_table *name_tab; /* Variable index by name. */ + int ref_cnt; + struct vardict_info *vars; /* Variables. */ + size_t n_vars; /* Number of variables. */ + size_t allocated_vars; /* Allocated space in 'vars'. */ + 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. */ + size_t n_splits; /* SPLIT FILE count. */ + enum split_type split_type; struct variable *weight; /* WEIGHT variable. */ struct variable *filter; /* FILTER variable. */ casenumber case_limit; /* Current case limit (N command). */ char *label; /* File label. */ - struct string documents; /* Documents, as a string. */ + struct string_array documents; /* Documents. */ struct vector **vector; /* Vectors of variables. */ - size_t vector_cnt; /* Number of vectors. */ + size_t n_vectors; /* Number of vectors. */ struct attrset attributes; /* Custom attributes. */ + struct mrset **mrsets; /* Multiple response sets. */ + size_t n_mrsets; /* Number of multiple response sets. */ + + /* Whether variable names must be valid identifiers. Normally, this is + true, but sometimes a dictionary is prepared for external use + (e.g. output to a CSV file) where names don't have to be valid. */ + bool names_must_be_ids; char *encoding; /* Character encoding of string data */ @@ -74,19 +94,101 @@ struct dictionary void *changed_data; }; +static void dict_unset_split_var (struct dictionary *, struct variable *, bool); +static void dict_unset_mrset_var (struct dictionary *, struct variable *); -void -dict_set_encoding (struct dictionary *d, const char *enc) +/* 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) { - d->encoding = strdup (enc); + 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->vars[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->vars[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 * dict_get_encoding (const struct dictionary *d) { return d->encoding ; } +/* Returns true if UTF-8 string ID is an acceptable identifier in DICT's + encoding, false otherwise. If ISSUE_ERROR is true, issues an explanatory + error message on failure. */ +bool +dict_id_is_valid (const struct dictionary *dict, const char *id, + bool issue_error) +{ + return (!dict->names_must_be_ids + || id_is_valid (id, dict->encoding, issue_error)); +} void dict_set_change_callback (struct dictionary *d, @@ -97,6 +199,14 @@ dict_set_change_callback (struct dictionary *d, 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. */ @@ -104,11 +214,10 @@ void dict_dump (const struct dictionary *d) { int i; - for (i = 0 ; i < d->var_cnt ; ++i ) + for (i = 0 ; i < d->n_vars ; ++i) { - const struct variable *v = - d->var[i]; - printf ("Name: %s;\tdict_idx: %d; case_idx: %d\n", + const struct variable *v = d->vars[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)); @@ -139,15 +248,21 @@ dict_copy_callbacks (struct dictionary *dest, dest->cb_data = src->cb_data; } -/* Creates and returns a new dictionary. */ +/* Creates and returns a new dictionary with the specified ENCODING. */ struct dictionary * -dict_create (void) +dict_create (const char *encoding) { - struct dictionary *d = xzalloc (sizeof *d); + struct dictionary *d = xmalloc (sizeof *d); + + *d = (struct dictionary) { + .encoding = xstrdup (encoding), + .names_must_be_ids = true, + .name_map = HMAP_INITIALIZER (d->name_map), + .attributes = ATTRSET_INITIALIZER (d->attributes), + .split_type = SPLIT_NONE, + .ref_cnt = 1, + }; - d->name_tab = hsh_create (8, compare_vars_by_name, hash_var_by_name, - NULL, NULL); - attrset_init (&d->attributes); return d; } @@ -158,43 +273,40 @@ dict_create (void) 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. */ + dict_compact_values to save space. + + Callbacks are not cloned. */ struct dictionary * dict_clone (const struct dictionary *s) { struct dictionary *d; size_t i; - assert (s != NULL); - - d = dict_create (); + d = dict_create (s->encoding); + dict_set_names_must_be_ids (d, dict_get_names_must_be_ids (s)); - for (i = 0; i < s->var_cnt; i++) + for (i = 0; i < s->n_vars; 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->vars[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++) + for (i = 0; i < var_get_n_short_names (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; - d->split_cnt = s->split_cnt; - if (d->split_cnt > 0) + d->n_splits = s->n_splits; + if (d->n_splits > 0) { - d->split = xnmalloc (d->split_cnt, sizeof *d->split); - for (i = 0; i < d->split_cnt; i++) - d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i])); + d->split = xnmalloc (d->n_splits, sizeof *d->split); + for (i = 0; i < d->n_splits; i++) + d->split[i] = dict_lookup_var_assert (d, var_get_name (s->split[i])); } + d->split_type = s->split_type; if (s->weight != NULL) dict_set_weight (d, dict_lookup_var_assert (d, var_get_name (s->weight))); @@ -206,175 +318,487 @@ dict_clone (const struct dictionary *s) dict_set_label (d, dict_get_label (s)); dict_set_documents (d, dict_get_documents (s)); - d->vector_cnt = s->vector_cnt; - d->vector = xnmalloc (d->vector_cnt, sizeof *d->vector); - for (i = 0; i < s->vector_cnt; i++) + d->n_vectors = s->n_vectors; + d->vector = xnmalloc (d->n_vectors, sizeof *d->vector); + for (i = 0; i < s->n_vectors; i++) d->vector[i] = vector_clone (s->vector[i], s, d); - d->encoding = strdup (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; } + + +/* Returns the SPLIT FILE vars (see cmd_split_file()). Call + dict_get_n_splits() 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_n_splits (const struct dictionary *d) +{ + return d->n_splits; +} + +/* 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, bool skip_callbacks) +{ + int orig_count; + + assert (dict_contains_var (d, v)); + + orig_count = d->n_splits; + d->n_splits = remove_equal (d->split, d->n_splits, sizeof *d->split, + &v, compare_var_ptrs, NULL); + if (orig_count != d->n_splits && !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); + } +} + + +/* Sets N split vars SPLIT in dictionary D. N is silently capped to a maximum + of MAX_SPLITS. */ +static void +dict_set_split_vars__ (struct dictionary *d, + struct variable *const *split, size_t n, + enum split_type type, bool skip_callbacks) +{ + if (n > MAX_SPLITS) + n = MAX_SPLITS; + assert (n == 0 || split != NULL); + + d->n_splits = n; + d->split_type = type == SPLIT_NONE ? SPLIT_LAYERED : type; + if (n > 0) + { + d->split = xnrealloc (d->split, n, sizeof *d->split) ; + memcpy (d->split, split, n * sizeof *d->split); + } + else + { + free (d->split); + d->split = NULL; + } + + 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 N split vars SPLIT in dictionary D. */ +void +dict_set_split_vars (struct dictionary *d, + struct variable *const *split, size_t n, + enum split_type type) +{ + dict_set_split_vars__ (d, split, n, type, false); +} + +void +dict_clear_split_vars (struct dictionary *d) +{ + dict_set_split_vars (d, NULL, 0, SPLIT_NONE); +} + + +/* Deletes variable V from dictionary D and frees V. + + This is a very bad idea if there might be any pointers to V + from outside D. In general, no variable in the active 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. + + 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) +{ + int dict_index = var_get_dict_index (v); + const int case_index = var_get_case_index (v); + + assert (dict_contains_var (d, v)); + + dict_unset_split_var (d, v, skip_callbacks); + dict_unset_mrset_var (d, v); + + if (d->weight == v) + dict_set_weight (d, NULL); + + if (d->filter == v) + dict_set_filter (d, NULL); + + dict_clear_vectors (d); + + /* Remove V from var array. */ + unindex_vars (d, dict_index, d->n_vars); + remove_element (d->vars, d->n_vars, sizeof *d->vars, dict_index); + d->n_vars--; + + /* Update dict_index for each affected variable. */ + reindex_vars (d, dict_index, d->n_vars, skip_callbacks); + + /* Free memory. */ + var_clear_vardict (v); + + 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); + } + + invalidate_proto (d); + var_unref (v); +} + +/* Deletes variable V from dictionary D and frees V. + + This is a very bad idea if there might be any pointers to V + from outside D. In general, no variable in the active 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. + + 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) +{ + dict_delete_var__ (d, v, false); +} + + +/* 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) +{ + /* 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++); +} + +/* 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->n_vars - IDX) * COUNT variable changed callbacks + plus COUNT variable delete callbacks. + This here produces d->n_vars - IDX variable changed callbacks + plus COUNT variable delete callbacks. */ +void +dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count) +{ + assert (idx + count <= d->n_vars); + + /* 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->vars[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); + } + + dict_clear_vectors (d); + + /* Remove variables from var array. */ + unindex_vars (d, idx, d->n_vars); + remove_range (d->vars, d->n_vars, sizeof *d->vars, idx, count); + d->n_vars -= count; + + /* Reindexing will result variable-changed callback */ + reindex_vars (d, idx, d->n_vars, false); + + invalidate_proto (d); + if (d->changed) d->changed (d, d->changed_data); + + /* 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); + } +} + +/* Deletes scratch variables from dictionary D. */ +void +dict_delete_scratch_vars (struct dictionary *d) +{ + int i; + + /* FIXME: this can be done in O(count) time, but this algorithm + is O(count**2). */ + for (i = 0; i < d->n_vars;) + if (var_get_dict_class (d->vars[i].var) == DC_SCRATCH) + dict_delete_var (d, d->vars[i].var); + else + i++; +} + + + /* Clears the contents from a dictionary without destroying the dictionary itself. */ -void -dict_clear (struct dictionary *d) +static void +dict_clear__ (struct dictionary *d, bool skip_callbacks) { /* 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 ) + while (d->n_vars > 0) { - dict_delete_var (d, d->var[d->var_cnt - 1]); + dict_delete_var__ (d, d->vars[d->n_vars - 1].var, skip_callbacks); } - free (d->var); - d->var = NULL; - d->var_cnt = d->var_cap = 0; - hsh_clear (d->name_tab); + free (d->vars); + d->vars = NULL; + d->n_vars = d->allocated_vars = 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); + dict_set_split_vars__ (d, NULL, 0, SPLIT_NONE, skip_callbacks); + + if (skip_callbacks) + { + d->weight = NULL; + d->filter = NULL; + } + else + { + dict_set_weight (d, NULL); + dict_set_filter (d, NULL); + } d->case_limit = 0; free (d->label); d->label = NULL; - ds_destroy (&d->documents); + string_array_clear (&d->documents); dict_clear_vectors (d); attrset_clear (&d->attributes); } -/* Destroys the aux data for every variable in D, by calling - var_clear_aux() for each variable. */ +/* Clears the contents from a dictionary without destroying the + dictionary itself. */ void -dict_clear_aux (struct dictionary *d) +dict_clear (struct dictionary *d) { - int i; + dict_clear__ (d, false); +} - assert (d != NULL); +/* Clears a dictionary and destroys it. */ +static void +_dict_destroy (struct dictionary *d) +{ + /* In general, we don't want callbacks occurring, if the dictionary + is being destroyed */ + d->callbacks = NULL ; - for (i = 0; i < d->var_cnt; i++) - var_clear_aux (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); +} + +struct dictionary * +dict_ref (struct dictionary *d) +{ + d->ref_cnt++; + return d; } -/* Clears a dictionary and destroys it. */ void -dict_destroy (struct dictionary *d) +dict_unref (struct dictionary *d) { - if (d != NULL) - { - /* In general, we don't want callbacks occuring, if the dictionary - is being destroyed */ - d->callbacks = NULL ; - - dict_clear (d); - hsh_destroy (d->name_tab); - attrset_destroy (&d->attributes); - free (d); - } + if (d == NULL) + return; + d->ref_cnt--; + assert (d->ref_cnt >= 0); + if (d->ref_cnt == 0) + _dict_destroy (d); } /* Returns the number of variables in D. */ size_t -dict_get_var_cnt (const struct dictionary *d) +dict_get_n_vars (const struct dictionary *d) { - assert (d != NULL); - - return d->var_cnt; + return d->n_vars; } /* Returns the variable in D with dictionary index IDX, which must be between 0 and the count returned by - dict_get_var_cnt(), exclusive. */ + dict_get_n_vars(), exclusive. */ struct variable * dict_get_var (const struct dictionary *d, size_t idx) { - assert (d != NULL); - assert (idx < d->var_cnt); + assert (idx < d->n_vars); - return d->var[idx]; + return d->vars[idx].var; } -/* Sets *VARS to an array of pointers to variables in D and *CNT +/* Sets *VARS to an array of pointers to variables in D and *N 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) + size_t *n, enum dict_class exclude) { - dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude); + dict_get_vars_mutable (d, (struct variable ***) vars, n, exclude); } -/* Sets *VARS to an array of pointers to variables in D and *CNT +/* Sets *VARS to an array of pointers to variables in D and *N 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 *n, enum dict_class exclude) { 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++) + for (i = 0; i < d->n_vars; i++) { - enum dict_class class = var_get_dict_class (d->var[i]); + enum dict_class class = var_get_dict_class (d->vars[i].var); if (!(class & exclude)) count++; } *vars = xnmalloc (count, sizeof **vars); - *cnt = 0; - for (i = 0; i < d->var_cnt; i++) + *n = 0; + for (i = 0; i < d->n_vars; i++) { - enum dict_class class = var_get_dict_class (d->var[i]); + enum dict_class class = var_get_dict_class (d->vars[i].var); if (!(class & exclude)) - (*vars)[(*cnt)++] = d->var[i]; + (*vars)[(*n)++] = d->vars[i].var; } - assert (*cnt == count); + assert (*n == count); } static struct variable * -add_var (struct dictionary *d, struct variable *v) +add_var_with_case_index (struct dictionary *d, struct variable *v, + int case_index) { - /* Add dictionary info to variable. */ - struct vardict_info vdi; - vdi.case_index = d->next_value_idx; - vdi.dict_index = d->var_cnt; - vdi.dict = d; - var_set_vardict (v, &vdi); + struct vardict_info *vardict; + + assert (case_index >= d->next_value_idx); /* Update dictionary. */ - if (d->var_cnt >= d->var_cap) + if (d->n_vars >= d->allocated_vars) { - d->var_cap = 8 + 2 * d->var_cap; - d->var = xnrealloc (d->var, d->var_cap, sizeof *d->var); + size_t i; + + d->vars = x2nrealloc (d->vars, &d->allocated_vars, sizeof *d->vars); + hmap_clear (&d->name_map); + for (i = 0; i < d->n_vars; i++) + { + var_set_vardict (d->vars[i].var, &d->vars[i]); + hmap_insert_fast (&d->name_map, &d->vars[i].name_node, + d->vars[i].name_node.hash); + } } - d->var[d->var_cnt++] = v; - hsh_force_insert (d->name_tab, v); - if ( d->changed ) d->changed (d, d->changed_data); - if ( d->callbacks && d->callbacks->var_added ) + vardict = &d->vars[d->n_vars++]; + vardict->dict = d; + vardict->var = v; + hmap_insert (&d->name_map, &vardict->name_node, + utf8_hash_case_string (var_get_name (v), 0)); + vardict->case_index = case_index; + var_set_vardict (v, vardict); + + if (d->changed) d->changed (d, d->changed_data); + if (d->callbacks && d->callbacks->var_added) d->callbacks->var_added (d, var_get_dict_index (v), d->cb_data); - d->next_value_idx += var_get_value_cnt (v); + invalidate_proto (d); + d->next_value_idx = case_index + 1; return v; } +static struct variable * +add_var (struct dictionary *d, struct variable *v) +{ + return add_var_with_case_index (d, v, d->next_value_idx); +} + /* Creates and returns a new variable in D with the given NAME and WIDTH. Returns a null pointer if the given NAME would duplicate that of an existing variable in the dictionary. */ @@ -396,27 +820,46 @@ 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); @@ -424,200 +867,50 @@ dict_clone_var_assert (struct dictionary *d, const struct variable *old_var, return add_var (d, new_var); } -/* Returns the variable named NAME in D, or a null pointer if no - variable has that name. */ struct variable * -dict_lookup_var (const struct dictionary *d, const char *name) +dict_clone_var_in_place_assert (struct dictionary *d, + const struct variable *old_var) { - 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); - - if ( result && var_has_vardict (result)) - { - const struct vardict_info *vdi = var_get_vardict (result); - assert (vdi->dict == d); - } - - return result; + assert (dict_lookup_var (d, var_get_name (old_var)) == NULL); + return add_var_with_case_index (d, var_clone (old_var), + var_get_case_index (old_var)); } -/* Returns the variable named NAME in D. Assert-fails if no +/* Returns the variable named NAME in D, or a null pointer if no variable has that name. */ struct variable * -dict_lookup_var_assert (const struct dictionary *d, const char *name) +dict_lookup_var (const struct dictionary *d, const char *name) { - struct variable *v = dict_lookup_var (d, name); - assert (v != NULL); - return v; -} + struct vardict_info *vardict; -/* Returns true if variable V is in dictionary D, - false otherwise. */ -bool -dict_contains_var (const struct dictionary *d, const struct variable *v) -{ - if (var_has_vardict (v)) + HMAP_FOR_EACH_WITH_HASH (vardict, struct vardict_info, name_node, + utf8_hash_case_string (name, 0), &d->name_map) { - 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); + struct variable *var = vardict->var; + if (!utf8_strcasecmp (var_get_name (var), name)) + return var; } - else - return 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_; - - 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) -{ - struct vardict_info vdi = *var_get_vardict (v); - struct dictionary *d = vdi.dict; - vdi.dict_index = dict_index; - var_set_vardict (v, &vdi); - - 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); -} - -/* 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); -} - -/* 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) -{ - size_t i; - - for (i = from; i < to; i++) - set_var_dict_index (d->var[i], i); -} - -/* Deletes variable V from dictionary D and frees V. - - This is a very bad idea if there might be any pointers to V - from outside D. In general, no variable in the active file's - 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. - - 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) -{ - int dict_index = var_get_dict_index (v); - const int case_index = var_get_case_index (v); - const int val_cnt = var_get_value_cnt (v); - - assert (dict_contains_var (d, v)); - - /* Delete aux data. */ - var_clear_aux (v); - - dict_unset_split_var (d, v); - - if (d->weight == v) - dict_set_weight (d, NULL); - - if (d->filter == v) - dict_set_filter (d, NULL); - - dict_clear_vectors (d); - - /* Remove V from var array. */ - 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); - if (d->callbacks && d->callbacks->var_deleted ) - d->callbacks->var_deleted (d, dict_index, case_index, val_cnt, d->cb_data); -} - -/* 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) -{ - /* 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) - dict_delete_var (d, *vars++); -} - -/* Deletes the COUNT variables in D starting at index IDX. This - is unsafe; see the comment on dict_delete_var() for - details. */ -void -dict_delete_consecutive_vars (struct dictionary *d, size_t idx, size_t count) -{ - /* FIXME: this can be done in O(count) time, but this algorithm - is O(count**2). */ - assert (idx + count <= d->var_cnt); - - while (count-- > 0) - dict_delete_var (d, d->var[idx]); + return NULL; } -/* Deletes scratch variables from dictionary D. */ -void -dict_delete_scratch_vars (struct dictionary *d) -{ - int i; - - /* FIXME: this can be done in O(count) time, but this algorithm - is O(count**2). */ - assert (d != NULL); +/* 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; +} - for (i = 0; i < d->var_cnt; ) - if (var_get_dict_class (d->var[i]) == DC_SCRATCH) - dict_delete_var (d, d->var[i]); - 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, @@ -628,9 +921,11 @@ 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); - 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); + assert (new_index < d->n_vars); + + unindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1); + move_element (d->vars, d->n_vars, sizeof *d->vars, old_index, new_index); + reindex_vars (d, MIN (old_index, new_index), MAX (old_index, new_index) + 1, false); } /* Reorders the variables in D, placing the COUNT variables @@ -641,46 +936,78 @@ 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); + assert (count <= d->n_vars); + + new_var = xnmalloc (d->allocated_vars, sizeof *new_var); - new_var = xnmalloc (d->var_cnt, sizeof *new_var); - memcpy (new_var, order, count * 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; } - 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++; - } - free (d->var); - d->var = new_var; + + /* Add remaining variables to new_var. */ + for (i = 0; i < d->n_vars; i++) + if (d->vars[i].dict != NULL) + new_var[count++] = d->vars[i]; + assert (count == d->n_vars); + + /* Replace old vardicts by new ones. */ + free (d->vars); + d->vars = new_var; + + hmap_clear (&d->name_map); + reindex_vars (d, 0, d->n_vars, false); } -/* Changes the name of variable V in dictionary D to NEW_NAME. */ +/* Changes the name of variable V that is currently in a dictionary to + NEW_NAME. */ static void -rename_var (struct dictionary *d, struct variable *v, const char *new_name) +rename_var (struct variable *v, const char *new_name) { - struct vardict_info vdi; - - assert (dict_contains_var (d, v)); - - vdi = *var_get_vardict (v); + struct vardict_info *vardict = var_get_vardict (v); var_clear_vardict (v); var_set_name (v, new_name); - var_set_vardict (v, &vdi); + vardict->name_node.hash = utf8_hash_case_string (new_name, 0); + var_set_vardict (v, vardict); +} + +/* Tries to changes the name of V in D to name NEW_NAME. Returns true if + successful, false if a variable (other than V) with the given name already + exists in D. */ +bool +dict_try_rename_var (struct dictionary *d, struct variable *v, + const char *new_name) +{ + struct variable *conflict = dict_lookup_var (d, new_name); + if (conflict && v != conflict) + return false; + + 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), false); + + 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), VAR_TRAIT_NAME, old, d->cb_data); + + var_unref (old); + + return true; } /* Changes the name of V in D to name NEW_NAME. Assert-fails if @@ -690,19 +1017,8 @@ void dict_rename_var (struct dictionary *d, struct variable *v, const char *new_name) { - 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); - - 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); + bool ok UNUSED = dict_try_rename_var (d, v, new_name); + assert (ok); } /* Renames COUNT variables specified in VARS to the names given @@ -733,34 +1049,37 @@ 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]), false); + } + + pool_destroy (pool); + return false; + } + reindex_var (d, var_get_vardict (vars[i]), false); + } /* Clear short names. */ if (settings_get_algorithm () == ENHANCED) @@ -783,63 +1102,78 @@ 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) { + size_t hint_len = strlen (hint); bool dropped = false; - char *cp; - - for (cp = name; *hint && cp < name + VAR_NAME_LEN; hint++) + char *root, *rp; + 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 + replaced by '_' too. */ + root = rp = xmalloc (hint_len + 1); + for (ofs = 0; ofs < hint_len; ofs += mblen) { - if (cp == name - ? lex_is_id1 (*hint) && *hint != '$' - : lex_is_idn (*hint)) + ucs4_t uc; + + mblen = u8_mbtouc (&uc, CHAR_CAST (const uint8_t *, hint + ofs), + hint_len - ofs); + if (rp == root + ? lex_uc_is_id1 (uc) && uc != '$' + : lex_uc_is_idn (uc)) { if (dropped) { - *cp++ = '_'; + *rp++ = '_'; dropped = false; } - if (cp < name + VAR_NAME_LEN) - *cp++ = *hint; + rp += u8_uctomb (CHAR_CAST (uint8_t *, rp), uc, 6); } - else if (cp > name) + else if (rp != root) dropped = true; } - *cp = '\0'; + *rp = '\0'; - if (name[0] != '\0') + if (root[0] != '\0') { - size_t len = strlen (name); unsigned long int i; - if (var_name_is_insertable (dict, name)) - return true; + if (var_name_is_insertable (dict, root)) + return root; for (i = 0; i < ULONG_MAX; i++) { char suffix[INT_BUFSIZE_BOUND (i) + 1]; - int ofs; + char *name; suffix[0] = '_'; - if (!str_format_26adic (i + 1, &suffix[1], sizeof suffix - 1)) + if (!str_format_26adic (i + 1, true, &suffix[1], sizeof suffix - 1)) NOT_REACHED (); - ofs = MIN (VAR_NAME_LEN - strlen (suffix), len); - strcpy (&name[ofs], suffix); - + name = utf8_encoding_concat (root, suffix, dict->encoding, 64); if (var_name_is_insertable (dict, name)) - return true; + { + free (root); + return name; + } + free (name); } } - return false; + free (root); + + 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; @@ -847,27 +1181,24 @@ 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 @@ -878,14 +1209,40 @@ 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) +{ + 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 whether variable names must be valid identifiers. Normally, this is + true, but sometimes a dictionary is prepared for external use (e.g. output + to a CSV file) where names don't have to be valid. */ +bool +dict_get_names_must_be_ids (const struct dictionary *d) +{ + return d->names_must_be_ids; +} + +/* Sets whether variable names must be valid identifiers. Normally, this is + true, but sometimes a dictionary is prepared for external use (e.g. output + to a CSV file) where names don't have to be valid. + + Changing this setting from false to true doesn't make the dictionary check + all the existing variable names, so it can cause an invariant violation. */ +void +dict_set_names_must_be_ids (struct dictionary *d, bool names_must_be_ids) { - return ((hint != NULL && make_hinted_name (dict, hint, name)) - || make_numeric_name (dict, num_start, name)); + d->names_must_be_ids = names_must_be_ids; } /* Returns the weighting variable in dictionary D, or a null @@ -893,7 +1250,6 @@ 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; @@ -909,7 +1265,6 @@ 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) @@ -917,34 +1272,33 @@ dict_get_case_weight (const struct dictionary *d, const struct ccase *c, else { double w = case_num (c, d->weight); - if (w < 0.0 || var_is_num_missing (d->weight, w, MV_ANY)) - w = 0.0; - if ( w == 0.0 && warn_on_invalid != NULL && *warn_on_invalid ) { - *warn_on_invalid = false; - msg (SW, _("At least one case in the data file had a weight value " - "that was user-missing, system-missing, zero, or " - "negative. These case(s) were ignored.")); - } - return w; + + return var_force_valid_weight (d->weight, w, warn_on_invalid); } } +/* Returns the format to use for weights. */ +const struct fmt_spec * +dict_get_weight_format (const struct dictionary *d) +{ + return d->weight ? var_get_print_format (d->weight) : &F_8_0; +} + /* Sets the weighting variable of D to V, or turning off weighting if V is a null pointer. */ 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 ) + 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()) @@ -952,7 +1306,6 @@ 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; @@ -963,17 +1316,16 @@ 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 ) + 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 @@ -981,8 +1333,6 @@ 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; } @@ -991,19 +1341,34 @@ 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->n_vars); + for (i = 0; i < d->n_vars; i++) + d->proto = caseproto_set_width (d->proto, + var_get_case_index (d->vars[i].var), + var_get_width (d->vars[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; } @@ -1012,8 +1377,6 @@ 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); } @@ -1025,40 +1388,14 @@ dict_compact_values (struct dictionary *d) size_t i; 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) + for (i = 0; i < d->n_vars; 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->vars[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), @@ -1072,237 +1409,211 @@ dict_pad_values (struct dictionary *d, int start, int padding) size_t dict_count_values (const struct dictionary *d, unsigned int exclude_classes) { - size_t i; - size_t cnt; - assert ((exclude_classes & ~((1u << DC_ORDINARY) | (1u << DC_SYSTEM) | (1u << DC_SCRATCH))) == 0); - cnt = 0; - for (i = 0; i < d->var_cnt; i++) + size_t n = 0; + for (size_t i = 0; i < d->n_vars; i++) { - enum dict_class class = var_get_dict_class (d->var[i]); + enum dict_class class = var_get_dict_class (d->vars[i].var); if (!(exclude_classes & (1u << class))) - cnt += var_get_value_cnt (d->var[i]); + n++; } - return cnt; -} - -/* 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) -{ - assert (d != NULL); - - return d->split; + return n; } -/* Returns the number of SPLIT FILE vars. */ -size_t -dict_get_split_cnt (const struct dictionary *d) -{ - assert (d != NULL); - - return d->split_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(). -/* 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) + 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) { - int orig_count; + struct caseproto *proto; + size_t i; - assert (dict_contains_var (d, v)); + assert ((exclude_classes & ~((1u << DC_ORDINARY) + | (1u << DC_SYSTEM) + | (1u << DC_SCRATCH))) == 0); - 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) + proto = caseproto_create (); + for (i = 0; i < d->n_vars; i++) { - 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); + struct variable *v = d->vars[i].var; + if (!(exclude_classes & (1u << var_get_dict_class (v)))) + proto = caseproto_add_width (proto, var_get_width (v)); } + return proto; } - -/* Sets CNT split vars SPLIT in dictionary D. */ -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; - 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 * dict_get_label (const struct dictionary *d) { - assert (d != NULL); - return d->label; } -/* Sets D's file label to LABEL, truncating it to a maximum of 60 - characters. */ +/* Sets D's file label to LABEL, truncating it to at most 60 bytes in D's + encoding. + + Removes D's label if LABEL is null or the empty string. */ void dict_set_label (struct dictionary *d, const char *label) { - assert (d != NULL); - free (d->label); - d->label = label != NULL ? xstrndup (label, 60) : NULL; + if (label == NULL || label[0] == '\0') + d->label = NULL; + else + d->label = utf8_encoding_trunc (label, d->encoding, 60); } -/* Returns the documents for D, or a null pointer if D has no - documents. If the return value is nonnull, then the string - will be an exact multiple of DOC_LINE_LENGTH bytes in length, - with each segment corresponding to one line. */ -const char * +/* Returns the documents for D, as an UTF-8 encoded string_array. The + return value is always nonnull; if there are no documents then the + string_arary is empty.*/ +const struct string_array * dict_get_documents (const struct dictionary *d) { - return ds_is_empty (&d->documents) ? NULL : ds_cstr (&d->documents); + return &d->documents; } -/* Sets the documents for D to DOCUMENTS, or removes D's - documents if DOCUMENT is a null pointer. If DOCUMENTS is - nonnull, then it should be an exact multiple of - DOC_LINE_LENGTH bytes in length, with each segment - corresponding to one line. */ +/* Replaces the documents for D by NEW_DOCS, a UTF-8 encoded string_array. */ void -dict_set_documents (struct dictionary *d, const char *documents) +dict_set_documents (struct dictionary *d, const struct string_array *new_docs) { - size_t remainder; + /* 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); + + 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); +} - ds_assign_cstr (&d->documents, documents != NULL ? documents : ""); +/* Replaces the documents for D by UTF-8 encoded string NEW_DOCS, dividing it + into individual lines at new-line characters. Each line is truncated to at + most DOC_LINE_LENGTH bytes in D's encoding. */ +void +dict_set_documents_string (struct dictionary *d, const char *new_docs) +{ + const char *s; - /* In case the caller didn't get it quite right, pad out the - final line with spaces. */ - remainder = ds_length (&d->documents) % DOC_LINE_LENGTH; - if (remainder != 0) - ds_put_char_multiple (&d->documents, ' ', DOC_LINE_LENGTH - remainder); + dict_clear_documents (d); + for (s = new_docs; *s != '\0';) + { + size_t len = strcspn (s, "\n"); + char *line = xmemdup0 (s, len); + dict_add_document_line (d, line, false); + free (line); + + s += len; + if (*s == '\n') + s++; + } } /* Drops the documents from dictionary D. */ void dict_clear_documents (struct dictionary *d) { - ds_clear (&d->documents); + string_array_clear (&d->documents); } -/* Appends LINE to the documents in D. LINE will be truncated or - padded on the right with spaces to make it exactly - DOC_LINE_LENGTH bytes long. */ -void -dict_add_document_line (struct dictionary *d, const char *line) +/* Appends the UTF-8 encoded LINE to the documents in D. LINE will be + truncated so that it is no more than 80 bytes in the dictionary's + encoding. If this causes some text to be lost, and ISSUE_WARNING is true, + then a warning will be issued. */ +bool +dict_add_document_line (struct dictionary *d, const char *line, + bool issue_warning) { - if (strlen (line) > DOC_LINE_LENGTH) + size_t trunc_len; + bool truncated; + + trunc_len = utf8_encoding_trunc_len (line, d->encoding, DOC_LINE_LENGTH); + truncated = line[trunc_len] != '\0'; + if (truncated && issue_warning) { - /* Note to translators: "bytes" is correct, not characters */ + /* TRANSLATORS: "bytes" is correct, not characters due to UTF encoding */ msg (SW, _("Truncating document line to %d bytes."), DOC_LINE_LENGTH); } - buf_copy_str_rpad (ds_put_uninit (&d->documents, DOC_LINE_LENGTH), - DOC_LINE_LENGTH, line); + + string_array_append_nocopy (&d->documents, xmemdup0 (line, trunc_len)); + + return !truncated; } /* Returns the number of document lines in dictionary D. */ size_t -dict_get_document_line_cnt (const struct dictionary *d) +dict_get_document_n_lines (const struct dictionary *d) { - return ds_length (&d->documents) / DOC_LINE_LENGTH; + return d->documents.n; } -/* Copies document line number IDX from dictionary D into - LINE, trimming off any trailing white space. */ -void -dict_get_document_line (const struct dictionary *d, - size_t idx, struct string *line) +/* Returns document line number IDX in dictionary D. The caller must not + modify or free the returned string. */ +const char * +dict_get_document_line (const struct dictionary *d, size_t idx) { - assert (idx < dict_get_document_line_cnt (d)); - ds_assign_substring (line, ds_substr (&d->documents, idx * DOC_LINE_LENGTH, - DOC_LINE_LENGTH)); - ds_rtrim (line, ss_cstr (CC_SPACES)); + assert (idx < d->documents.n); + return d->documents.strings[idx]; } -/* Creates in D a vector named NAME that contains the CNT +/* Creates in D a vector named NAME that contains the N variables in VAR. Returns true if successful, or false if a vector named NAME already exists in D. */ bool dict_create_vector (struct dictionary *d, const char *name, - struct variable **var, size_t cnt) + struct variable **var, size_t n) { - size_t i; - - assert (var != NULL); - assert (cnt > 0); - for (i = 0; i < cnt; i++) + assert (n > 0); + for (size_t i = 0; i < n; i++) assert (dict_contains_var (d, var[i])); if (dict_lookup_vector (d, name) == NULL) { - d->vector = xnrealloc (d->vector, d->vector_cnt + 1, sizeof *d->vector); - d->vector[d->vector_cnt++] = vector_create (name, var, cnt); + d->vector = xnrealloc (d->vector, d->n_vectors + 1, sizeof *d->vector); + d->vector[d->n_vectors++] = vector_create (name, var, n); return true; } else return false; } -/* Creates in D a vector named NAME that contains the CNT +/* Creates in D a vector named NAME that contains the N variables in VAR. A vector named NAME must not already exist in D. */ void dict_create_vector_assert (struct dictionary *d, const char *name, - struct variable **var, size_t cnt) + struct variable **var, size_t n) { assert (dict_lookup_vector (d, name) == NULL); - dict_create_vector (d, name, var, cnt); + dict_create_vector (d, name, var, n); } /* Returns the vector in D with index IDX, which must be less - than dict_get_vector_cnt (D). */ + than dict_get_n_vectors (D). */ const struct vector * dict_get_vector (const struct dictionary *d, size_t idx) { - assert (d != NULL); - assert (idx < d->vector_cnt); + assert (idx < d->n_vectors); return d->vector[idx]; } /* Returns the number of vectors in D. */ size_t -dict_get_vector_cnt (const struct dictionary *d) +dict_get_n_vectors (const struct dictionary *d) { - assert (d != NULL); - - return d->vector_cnt; + return d->n_vectors; } /* Looks up and returns the vector within D with the given @@ -1311,8 +1622,8 @@ const struct vector * dict_lookup_vector (const struct dictionary *d, const char *name) { size_t i; - for (i = 0; i < d->vector_cnt; i++) - if (!strcasecmp (vector_get_name (d->vector[i]), name)) + for (i = 0; i < d->n_vectors; i++) + if (!utf8_strcasecmp (vector_get_name (d->vector[i]), name)) return d->vector[i]; return NULL; } @@ -1323,22 +1634,153 @@ dict_clear_vectors (struct dictionary *d) { size_t i; - for (i = 0; i < d->vector_cnt; i++) + for (i = 0; i < d->n_vectors; i++) vector_destroy (d->vector[i]); free (d->vector); d->vector = NULL; - d->vector_cnt = 0; + d->n_vectors = 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 (!utf8_strcasecmp (name, dict->mrsets[i]->name)) + return i; + + return SIZE_MAX; +} + +/* Looks for a multiple response set named NAME in DICT. If it finds one, + returns it; otherwise, returns NULL. */ +const struct mrset * +dict_lookup_mrset (const struct dictionary *dict, const char *name) +{ + size_t idx = dict_lookup_mrset_idx (dict, name); + return idx != SIZE_MAX ? dict->mrsets[idx] : NULL; } +/* Adds MRSET to DICT, replacing any existing set with the same name. Returns + true if a set was replaced, false if none existed with the specified name. + + Ownership of MRSET is transferred to DICT. */ +bool +dict_add_mrset (struct dictionary *dict, struct mrset *mrset) +{ + size_t idx; + + assert (mrset_ok (mrset, dict)); + + idx = dict_lookup_mrset_idx (dict, mrset->name); + if (idx == SIZE_MAX) + { + 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 attribute set. */ struct attrset * -dict_get_attributes (const struct dictionary *d) +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. */ @@ -1352,66 +1794,89 @@ dict_set_attributes (struct dictionary *d, const struct attrset *attrs) /* Returns true if D has at least one attribute in its attribute set, false if D's attribute set is empty. */ bool -dict_has_attributes (const struct dictionary *d) +dict_has_attributes (const struct dictionary *d) { return attrset_count (&d->attributes) > 0; } -/* Called from variable.c to notify the dictionary that some property of - the variable has changed */ +/* Called from variable.c to notify the dictionary that some property (indicated + by WHAT) of the variable has changed. OLDVAR is a copy of V as it existed + prior to the change. OLDVAR is destroyed by this function. +*/ void -dict_var_changed (const struct variable *v) +dict_var_changed (const struct variable *v, unsigned int what, struct variable *oldvar) { - if ( var_has_vardict (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) + 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); + 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); } + var_unref (oldvar); } -/* 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) + +/* 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 ( var_has_vardict (v)) - { - const struct vardict_info *vdi = var_get_vardict (v); - struct dictionary *d; + if (internal_dict == NULL) + internal_dict = dict_create ("UTF-8"); - d = vdi->dict; + for (;;) + { + static int counter = INT_MAX / 2; + struct variable *var; + char name[64]; - dict_pad_values (d, var_get_case_index(v) + 1, delta); + if (++counter == INT_MAX) + counter = INT_MAX / 2; - if (d->changed) d->changed (d, d->changed_data); - if ( d->callbacks && d->callbacks->var_resized ) - d->callbacks->var_resized (d, var_get_dict_index (v), delta, d->cb_data); + 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; + } } } -/* Called from variable.c to notify the dictionary that the variable's display width - has changed */ +/* Destroys VAR, which must have been created with + dict_create_internal_var(). */ void -dict_var_display_width_changed (const struct variable *v) +dict_destroy_internal_var (struct variable *var) { - if ( var_has_vardict (v)) + if (var != NULL) { - const struct vardict_info *vdi = var_get_vardict (v); - struct dictionary *d; - - d = vdi->dict; + dict_delete_var (internal_dict, var); - 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); + /* 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_n_vars (internal_dict) == 0) + { + dict_unref (internal_dict); + internal_dict = NULL; + } } } - + +int +vardict_get_dict_index (const struct vardict_info *vardict) +{ + return vardict - vardict->dict->vars; +}