X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Fdata%2Fdictionary.c;h=526bb280a0956b70a24f4db3e32943f98701a7ec;hb=da299bd9871b178336a440c6ac53aebc3cea672e;hp=7964ad55e94b83c869c2ea531f10efccf8116051;hpb=6e352378b703f57313fdd0f628b99b19ff25d055;p=pspp-builds.git diff --git a/src/data/dictionary.c b/src/data/dictionary.c index 7964ad55..526bb280 100644 --- a/src/data/dictionary.c +++ b/src/data/dictionary.c @@ -23,13 +23,14 @@ #include "case.h" #include "category.h" +#include "identifier.h" #include "settings.h" #include "value-labels.h" #include "vardict.h" #include "variable.h" #include "vector.h" -#include #include +#include #include #include #include @@ -37,7 +38,9 @@ #include #include +#include "intprops.h" #include "minmax.h" +#include "xalloc.h" #include "gettext.h" #define _(msgid) gettext (msgid) @@ -53,7 +56,7 @@ struct dictionary size_t split_cnt; /* SPLIT FILE count. */ struct variable *weight; /* WEIGHT variable. */ struct variable *filter; /* FILTER variable. */ - size_t case_limit; /* Current case limit (N command). */ + casenumber case_limit; /* Current case limit (N command). */ char *label; /* File label. */ struct string documents; /* Documents, as a string. */ struct vector **vector; /* Vectors of variables. */ @@ -63,6 +66,23 @@ struct dictionary void *cb_data ; /* Data passed to callbacks */ }; +/* Print a representation of dictionary D to stdout, for + debugging purposes. */ +void +dict_dump (const struct dictionary *d) +{ + int i; + for (i = 0 ; i < d->var_cnt ; ++i ) + { + const struct variable *v = + d->var[i]; + printf ("Name: %s;\tdict_idx: %d; case_idx: %d\n", + var_get_name (v), + var_get_dict_index (v), + var_get_case_index (v)); + + } +} /* Associate CALLBACKS with DICT. Callbacks will be invoked whenever the dictionary or any of the variables it contains are modified. @@ -99,7 +119,13 @@ dict_create (void) } /* Creates and returns a (deep) copy of an existing - dictionary. */ + dictionary. + + The new dictionary's case indexes are copied from the old + dictionary. If the new dictionary won't be used to access + cases produced with the old dictionary, then the new + dictionary's case indexes should be compacted with + dict_compact_values to save space. */ struct dictionary * dict_clone (const struct dictionary *s) { @@ -112,12 +138,19 @@ dict_clone (const struct dictionary *s) for (i = 0; i < s->var_cnt; i++) { + const struct vardict_info *svdi; + struct vardict_info dvdi; struct variable *sv = s->var[i]; struct variable *dv = dict_clone_var_assert (d, sv, var_get_name (sv)); size_t i; for (i = 0; i < var_get_short_name_cnt (sv); i++) var_set_short_name (dv, i, var_get_short_name (sv, i)); + + svdi = var_get_vardict (sv); + dvdi = *svdi; + dvdi.dict = d; + var_set_vardict (dv, &dvdi); } d->next_value_idx = s->next_value_idx; @@ -159,20 +192,7 @@ dict_clear (struct dictionary *d) while (d->var_cnt > 0 ) { - struct variable *v = d->var[d->var_cnt - 1]; - int dict_index = var_get_dict_index (v); - int case_index = var_get_case_index (v); - int val_cnt = var_get_value_cnt (v); - - var_clear_vardict (v); - var_destroy (v); - - d->var_cnt--; - - if (d->callbacks && d->callbacks->var_deleted ) - d->callbacks->var_deleted (d, - dict_index, case_index, val_cnt, - d->cb_data); + dict_delete_var (d, d->var[d->var_cnt - 1]); } free (d->var); @@ -240,21 +260,26 @@ dict_get_var (const struct dictionary *d, size_t idx) return d->var[idx]; } -inline void +/* Sets *VARS to an array of pointers to variables in D and *CNT + to the number of variables in *D. All variables are returned + except for those, if any, in the classes indicated by EXCLUDE. + (There is no point in putting DC_SYSTEM in EXCLUDE as + dictionaries never include system variables.) */ +void dict_get_vars (const struct dictionary *d, const struct variable ***vars, - size_t *cnt, unsigned exclude_classes) + size_t *cnt, enum dict_class exclude) { - dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude_classes); + dict_get_vars_mutable (d, (struct variable ***) vars, cnt, exclude); } /* Sets *VARS to an array of pointers to variables in D and *CNT to the number of variables in *D. All variables are returned - if EXCLUDE_CLASSES is 0, or it may contain one or more of (1u - << DC_ORDINARY), (1u << DC_SYSTEM), or (1u << DC_SCRATCH) to - exclude the corresponding type of variable. */ + except for those, if any, in the classes indicated by EXCLUDE. + (There is no point in putting DC_SYSTEM in EXCLUDE as + dictionaries never include system variables.) */ void dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars, - size_t *cnt, unsigned exclude_classes) + size_t *cnt, enum dict_class exclude) { size_t count; size_t i; @@ -262,15 +287,13 @@ dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars, assert (d != NULL); assert (vars != NULL); assert (cnt != NULL); - assert ((exclude_classes & ~((1u << DC_ORDINARY) - | (1u << DC_SYSTEM) - | (1u << DC_SCRATCH))) == 0); + assert (exclude == (exclude & DC_ALL)); count = 0; for (i = 0; i < d->var_cnt; i++) { - enum dict_class class = dict_class_from_id (var_get_name (d->var[i])); - if (!(exclude_classes & (1u << class))) + enum dict_class class = var_get_dict_class (d->var[i]); + if (!(class & exclude)) count++; } @@ -278,8 +301,8 @@ dict_get_vars_mutable (const struct dictionary *d, struct variable ***vars, *cnt = 0; for (i = 0; i < d->var_cnt; i++) { - enum dict_class class = dict_class_from_id (var_get_name (d->var[i])); - if (!(exclude_classes & (1u << class))) + enum dict_class class = var_get_dict_class (d->var[i]); + if (!(class & exclude)) (*vars)[(*cnt)++] = d->var[i]; } assert (*cnt == count); @@ -369,13 +392,19 @@ dict_lookup_var (const struct dictionary *d, const char *name) struct variable *target ; struct variable *result ; - if ( ! var_is_valid_name (name, false)) + if ( ! var_is_plausible_name (name, false)) return NULL; target = var_create (name, 0); result = hsh_find (d->name_tab, target); var_destroy (target); + if ( result && var_has_vardict (result)) + { + const struct vardict_info *vdi = var_get_vardict (result); + assert (vdi->dict == d); + } + return result; } @@ -457,7 +486,7 @@ reindex_vars (struct dictionary *d, size_t from, size_t to) active on the dictionary's dataset, because those transformations might reference the deleted variable. The safest time to delete a variable is just after a procedure has - been executed, as done by MODIFY VARS. + been executed, as done by DELETE VARIABLES. Pointers to V within D are not a problem, because dict_delete_var() knows to remove V from split variables, @@ -499,6 +528,7 @@ dict_delete_var (struct dictionary *d, struct variable *v) var_clear_vardict (v); var_destroy (v); + if (d->callbacks && d->callbacks->var_deleted ) d->callbacks->var_deleted (d, dict_index, case_index, val_cnt, d->cb_data); } @@ -543,7 +573,7 @@ dict_delete_scratch_vars (struct dictionary *d) assert (d != NULL); for (i = 0; i < d->var_cnt; ) - if (dict_class_from_id (var_get_name (d->var[i])) == DC_SCRATCH) + if (var_get_dict_class (d->var[i]) == DC_SCRATCH) dict_delete_var (d, d->var[i]); else i++; @@ -626,7 +656,7 @@ dict_rename_var (struct dictionary *d, struct variable *v, rename_var (d, v, new_name); hsh_force_insert (d->name_tab, v); - if (get_algorithm () == ENHANCED) + if (settings_get_algorithm () == ENHANCED) var_clear_short_names (v); if ( d->callbacks && d->callbacks->var_changed ) @@ -691,7 +721,7 @@ dict_rename_vars (struct dictionary *d, } /* Clear short names. */ - if (get_algorithm () == ENHANCED) + if (settings_get_algorithm () == ENHANCED) for (i = 0; i < count; i++) var_clear_short_names (vars[i]); @@ -699,6 +729,123 @@ dict_rename_vars (struct dictionary *d, return true; } +/* Returns true if a variable named NAME may be inserted in DICT; + that is, if there is not already a variable with that name in + DICT and if NAME is not a reserved word. (The caller's checks + have already verified that NAME is otherwise acceptable as a + variable name.) */ +static bool +var_name_is_insertable (const struct dictionary *dict, const char *name) +{ + return (dict_lookup_var (dict, name) == NULL + && lex_id_to_token (ss_cstr (name)) == T_ID); +} + +static bool +make_hinted_name (const struct dictionary *dict, const char *hint, + char name[VAR_NAME_LEN + 1]) +{ + bool dropped = false; + char *cp; + + for (cp = name; *hint && cp < name + VAR_NAME_LEN; hint++) + { + if (cp == name + ? lex_is_id1 (*hint) && *hint != '$' + : lex_is_idn (*hint)) + { + if (dropped) + { + *cp++ = '_'; + dropped = false; + } + if (cp < name + VAR_NAME_LEN) + *cp++ = *hint; + } + else if (cp > name) + dropped = true; + } + *cp = '\0'; + + if (name[0] != '\0') + { + size_t len = strlen (name); + unsigned long int i; + + if (var_name_is_insertable (dict, name)) + return true; + + for (i = 0; i < ULONG_MAX; i++) + { + char suffix[INT_BUFSIZE_BOUND (i) + 1]; + int ofs; + + suffix[0] = '_'; + if (!str_format_26adic (i + 1, &suffix[1], sizeof suffix - 1)) + NOT_REACHED (); + + ofs = MIN (VAR_NAME_LEN - strlen (suffix), len); + strcpy (&name[ofs], suffix); + + if (var_name_is_insertable (dict, name)) + return true; + } + } + + return false; +} + +static bool +make_numeric_name (const struct dictionary *dict, unsigned long int *num_start, + char name[VAR_NAME_LEN + 1]) +{ + unsigned long int number; + + for (number = num_start != NULL ? MAX (*num_start, 1) : 1; + number < ULONG_MAX; + number++) + { + sprintf (name, "VAR%03lu", number); + if (dict_lookup_var (dict, name) == NULL) + { + if (num_start != NULL) + *num_start = number + 1; + return true; + } + } + + if (num_start != NULL) + *num_start = ULONG_MAX; + return false; +} + + +/* 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.) + + HINT, if it is non-null, is used as a suggestion that will be + modified for suitability as a variable name and for + uniqueness. + + If HINT is null or entirely unsuitable, a name in the form + "VAR%03d" will be generated, where the smallest unused integer + value is used. If NUM_START is non-null, then its value is + used as the minimum numeric value to check, and it is updated + to the next value to be checked. + */ +bool +dict_make_unique_var_name (const struct dictionary *dict, const char *hint, + unsigned long int *num_start, + char name[VAR_NAME_LEN + 1]) +{ + return ((hint != NULL && make_hinted_name (dict, hint, name)) + || make_numeric_name (dict, num_start, name)); +} + /* Returns the weighting variable in dictionary D, or a null pointer if the dictionary is unweighted. */ struct variable * @@ -710,11 +857,12 @@ dict_get_weight (const struct dictionary *d) return d->weight; } -/* Returns the value of D's weighting variable in case C, except that a - negative weight is returned as 0. Returns 1 if the dictionary is - unweighted. Will warn about missing, negative, or zero values if - warn_on_invalid is true. The function will set warn_on_invalid to false - if an invalid weight is found. */ +/* Returns the value of D's weighting variable in case C, except + that a negative weight is returned as 0. Returns 1 if the + dictionary is unweighted. Will warn about missing, negative, + or zero values if *WARN_ON_INVALID is true. The function will + set *WARN_ON_INVALID to false if an invalid weight is + found. */ double dict_get_case_weight (const struct dictionary *d, const struct ccase *c, bool *warn_on_invalid) @@ -774,6 +922,7 @@ 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; @@ -785,7 +934,7 @@ dict_set_filter (struct dictionary *d, struct variable *v) /* Returns the case limit for dictionary D, or zero if the number of cases is unlimited. */ -size_t +casenumber dict_get_case_limit (const struct dictionary *d) { assert (d != NULL); @@ -796,7 +945,7 @@ dict_get_case_limit (const struct dictionary *d) /* Sets CASE_LIMIT as the case limit for dictionary D. Use 0 for CASE_LIMIT to indicate no limit. */ void -dict_set_case_limit (struct dictionary *d, size_t case_limit) +dict_set_case_limit (struct dictionary *d, casenumber case_limit) { assert (d != NULL); @@ -889,112 +1038,13 @@ dict_count_values (const struct dictionary *d, unsigned int exclude_classes) cnt = 0; for (i = 0; i < d->var_cnt; i++) { - enum dict_class class = dict_class_from_id (var_get_name (d->var[i])); + enum dict_class class = var_get_dict_class (d->var[i]); if (!(exclude_classes & (1u << class))) cnt += var_get_value_cnt (d->var[i]); } return cnt; } -/* How to copy a contiguous range of values between cases. */ -struct copy_map - { - size_t src_idx; /* Starting value index in source case. */ - size_t dst_idx; /* Starting value index in target case. */ - size_t cnt; /* Number of values. */ - }; - -/* How to compact a case. */ -struct dict_compactor - { - struct copy_map *maps; /* Array of mappings. */ - size_t map_cnt; /* Number of mappings. */ - }; - -/* Creates and returns a dict_compactor that can be used to - compact cases for dictionary D. - - Compacting a case eliminates "holes" between values and after - the last value. (Holes are created by deleting variables.) - - All variables are compacted if EXCLUDE_CLASSES is 0, or it may - contain one or more of (1u << DC_ORDINARY), (1u << DC_SYSTEM), - or (1u << DC_SCRATCH) to cause the corresponding type of - variable to be deleted during compaction. */ -struct dict_compactor * -dict_make_compactor (const struct dictionary *d, unsigned int exclude_classes) -{ - struct dict_compactor *compactor; - struct copy_map *map; - size_t map_allocated; - size_t value_idx; - size_t i; - - assert ((exclude_classes & ~((1u << DC_ORDINARY) - | (1u << DC_SYSTEM) - | (1u << DC_SCRATCH))) == 0); - - compactor = xmalloc (sizeof *compactor); - compactor->maps = NULL; - compactor->map_cnt = 0; - map_allocated = 0; - - value_idx = 0; - map = NULL; - for (i = 0; i < d->var_cnt; i++) - { - struct variable *v = d->var[i]; - enum dict_class class = dict_class_from_id (var_get_name (v)); - if (exclude_classes & (1u << class)) - continue; - - if (map != NULL && map->src_idx + map->cnt == var_get_case_index (v)) - map->cnt += var_get_value_cnt (v); - else - { - if (compactor->map_cnt == map_allocated) - compactor->maps = x2nrealloc (compactor->maps, &map_allocated, - sizeof *compactor->maps); - map = &compactor->maps[compactor->map_cnt++]; - map->src_idx = var_get_case_index (v); - map->dst_idx = value_idx; - map->cnt = var_get_value_cnt (v); - } - value_idx += var_get_value_cnt (v); - } - - return compactor; -} - -/* Compacts SRC by copying it to DST according to the scheme in - COMPACTOR. - - Compacting a case eliminates "holes" between values and after - the last value. (Holes are created by deleting variables.) */ -void -dict_compactor_compact (const struct dict_compactor *compactor, - struct ccase *dst, const struct ccase *src) -{ - size_t i; - - for (i = 0; i < compactor->map_cnt; i++) - { - const struct copy_map *map = &compactor->maps[i]; - case_copy (dst, map->dst_idx, src, map->src_idx, map->cnt); - } -} - -/* Destroys COMPACTOR. */ -void -dict_compactor_destroy (struct dict_compactor *compactor) -{ - if (compactor != NULL) - { - free (compactor->maps); - free (compactor); - } -} - /* Returns the SPLIT FILE vars (see cmd_split_file()). Call dict_get_split_cnt() to determine how many SPLIT FILE vars there are. Returns a null pointer if and only if there are no @@ -1016,20 +1066,25 @@ dict_get_split_cnt (const struct dictionary *d) return d->split_cnt; } -/* Removes variable V from the set of split variables in dictionary D */ +/* Removes variable V, which must be in D, from D's set of split + variables. */ void -dict_unset_split_var (struct dictionary *d, - struct variable *v) +dict_unset_split_var (struct dictionary *d, struct variable *v) { - const int count = d->split_cnt; - d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split, - &v, compare_var_ptrs, NULL); + int orig_count; - if ( count == d->split_cnt) - return; + assert (dict_contains_var (d, v)); - if ( d->callbacks && d->callbacks->split_changed ) - d->callbacks->split_changed (d, d->cb_data); + orig_count = d->split_cnt; + d->split_cnt = remove_equal (d->split, d->split_cnt, sizeof *d->split, + &v, compare_var_ptrs, NULL); + if (orig_count != d->split_cnt) + { + /* We changed the set of split variables so invoke the + callback. */ + if (d->callbacks && d->callbacks->split_changed) + d->callbacks->split_changed (d, d->cb_data); + } } /* Sets CNT split vars SPLIT in dictionary D. */ @@ -1041,8 +1096,16 @@ dict_set_split_vars (struct dictionary *d, assert (cnt == 0 || split != NULL); d->split_cnt = cnt; - d->split = cnt > 0 ? xnrealloc (d->split, cnt, sizeof *d->split) : NULL; - memcpy (d->split, split, cnt * sizeof *d->split); + if ( cnt > 0 ) + { + d->split = xnrealloc (d->split, cnt, sizeof *d->split) ; + memcpy (d->split, split, cnt * sizeof *d->split); + } + else + { + free (d->split); + d->split = NULL; + } if ( d->callbacks && d->callbacks->split_changed ) d->callbacks->split_changed (d, d->cb_data); @@ -1066,16 +1129,7 @@ dict_set_label (struct dictionary *d, const char *label) assert (d != NULL); free (d->label); - if (label == NULL) - d->label = NULL; - else if (strlen (label) < 60) - d->label = xstrdup (label); - else - { - d->label = xmalloc (61); - memcpy (d->label, label, 60); - d->label[60] = '\0'; - } + d->label = label != NULL ? xstrndup (label, 60) : NULL; } /* Returns the documents for D, or a null pointer if D has no @@ -1239,9 +1293,7 @@ dict_var_changed (const struct variable *v) if ( var_has_vardict (v)) { const struct vardict_info *vdi = var_get_vardict (v); - struct dictionary *d; - - d = vdi->dict; + struct dictionary *d = vdi->dict; if ( d->callbacks && d->callbacks->var_changed ) d->callbacks->var_changed (d, var_get_dict_index (v), d->cb_data);