X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Fmath%2Fcategoricals.c;h=c088ac00732bd23dc7a9db178f058c434f1579b4;hb=12140b59f70771295af36404296219cba9f3f6bc;hp=d4d43243703cd30dd2a848f7f56bf148b8ffc5b8;hpb=04c7908ba9dbc11eca3b20f43e9adfc2840b3679;p=pspp diff --git a/src/math/categoricals.c b/src/math/categoricals.c index d4d4324370..c088ac0073 100644 --- a/src/math/categoricals.c +++ b/src/math/categoricals.c @@ -34,8 +34,6 @@ #define CATEGORICALS_DEBUG 0 -#define EFFECTS_CODING 1 - struct value_node { struct hmap_node node; /* Node in hash map. */ @@ -96,10 +94,10 @@ lookup_variable (const struct hmap *map, const struct variable *var, unsigned in struct interact_params { - /* A map indexed by a interaction_value */ + /* A map of cases indexed by a interaction_value */ struct hmap ivmap; - const struct interaction *iact; + struct interaction *iact; int base_subscript_short; int base_subscript_long; @@ -165,12 +163,17 @@ struct categoricals struct pool *pool; - /* Missing values to be excluded */ - enum mv_class exclude; + /* Missing values in the dependent varirable to be excluded */ + enum mv_class dep_excl; + + /* Missing values in the factor variables to be excluded */ + enum mv_class fctr_excl; const void *aux1; void *aux2; + bool sane; + const struct payload *payload; }; @@ -248,16 +251,15 @@ categoricals_destroy (struct categoricals *cat) /* Interate over each interaction value, and unref any cases that we reffed */ HMAP_FOR_EACH (iv, struct interaction_value, node, &cat->iap[i].ivmap) { -#if 0 - if (cat->payload) - cat->payload->destroy (cat->aux1, iv->user_data); -#endif + if (cat->payload && cat->payload->destroy) + cat->payload->destroy (cat->aux1, cat->aux2, iv->user_data); case_unref (iv->ccase); } free (cat->iap[i].enc_sum); free (cat->iap[i].df_prod); hmap_destroy (&cat->iap[i].ivmap); + interaction_destroy (cat->iap[i].iact); } /* Interate over each variable and delete its value map */ @@ -292,10 +294,15 @@ lookup_case (const struct hmap *map, const struct interaction *iact, const struc return iv; } +bool +categoricals_sane (const struct categoricals *cat) +{ + return cat->sane; +} struct categoricals * categoricals_create (struct interaction *const*inter, size_t n_inter, - const struct variable *wv, enum mv_class exclude) + const struct variable *wv, enum mv_class dep_excl, enum mv_class fctr_excl) { size_t i; struct categoricals *cat = xmalloc (sizeof *cat); @@ -307,9 +314,11 @@ categoricals_create (struct interaction *const*inter, size_t n_inter, cat->reverse_variable_map_short = NULL; cat->reverse_variable_map_long = NULL; cat->pool = pool_create (); - cat->exclude = exclude; + cat->dep_excl = dep_excl; + cat->fctr_excl = fctr_excl; cat->payload = NULL; cat->aux2 = NULL; + cat->sane = false; cat->iap = pool_calloc (cat->pool, cat->n_iap, sizeof *cat->iap); @@ -347,7 +356,12 @@ categoricals_update (struct categoricals *cat, const struct ccase *c) { int i; struct variable_node *vn = NULL; - const double weight = cat->wv ? case_data (c, cat->wv)->f : 1.0; + double weight; + + if (NULL == cat) + return; + + weight = cat->wv ? case_data (c, cat->wv)->f : 1.0; assert (NULL == cat->reverse_variable_map_short); assert (NULL == cat->reverse_variable_map_long); @@ -378,7 +392,7 @@ categoricals_update (struct categoricals *cat, const struct ccase *c) size_t hash; struct interaction_value *node; - if ( interaction_case_is_missing (iact, c, cat->exclude)) + if ( interaction_case_is_missing (iact, c, cat->fctr_excl)) continue; hash = interaction_case_hash (iact, c, 0); @@ -442,6 +456,9 @@ categoricals_n_total (const struct categoricals *cat) size_t categoricals_df_total (const struct categoricals *cat) { + if (NULL == cat) + return 0; + return cat->df_sum; } @@ -454,7 +471,7 @@ categoricals_is_complete (const struct categoricals *cat) /* This function must be called *before* any call to categoricals_get_*_by subscript and *after* all calls to categoricals_update */ -bool +void categoricals_done (const struct categoricals *cat_) { /* Implementation Note: Whilst this function is O(n) in cat->n_cats_total, in most @@ -468,6 +485,10 @@ categoricals_done (const struct categoricals *cat_) int i; int idx_short = 0; int idx_long = 0; + + if (NULL == cat) + return; + cat->df_sum = 0; cat->n_cats_total = 0; @@ -488,7 +509,10 @@ categoricals_done (const struct categoricals *cat_) struct variable_node *vn = lookup_variable (&cat->varmap, var, hash_pointer (var, 0)); if (hmap_count (&vn->valmap) == 0) - return false; + { + cat->sane = false; + return; + } cat->iap[i].df_prod[v] = df * (hmap_count (&vn->valmap) - 1); df = cat->iap[i].df_prod[v]; @@ -528,7 +552,6 @@ categoricals_done (const struct categoricals *cat_) HMAP_FOR_EACH (ivn, struct interaction_value, node, &iap->ivmap) { iap->reverse_interaction_value_map[x++] = ivn; - } assert (x <= iap->n_cats); @@ -572,13 +595,15 @@ categoricals_done (const struct categoricals *cat_) struct interaction_value *iv = iap->reverse_interaction_value_map[y]; for (x = iap->base_subscript_short; x < iap->base_subscript_short + df ;++x) { - const double bin = categoricals_get_code_for_case (cat, x, iv->ccase); \ + const double bin = categoricals_get_effects_code_for_case (cat, x, iv->ccase); iap->enc_sum [x - iap->base_subscript_short] += bin * iv->cc; } + if (cat->payload && cat->payload->calculate) + cat->payload->calculate (cat->aux1, cat->aux2, iv->user_data); } } - return true; + cat->sane = true; } @@ -630,11 +655,13 @@ categoricals_get_sum_by_subscript (const struct categoricals *cat, int subscript return vp->enc_sum[subscript - vp->base_subscript_short]; } + /* Returns unity if the value in case C at SUBSCRIPT is equal to the category for that subscript */ -double +static double categoricals_get_code_for_case (const struct categoricals *cat, int subscript, - const struct ccase *c) + const struct ccase *c, + bool effects_coding) { const struct interaction *iact = categoricals_get_interaction_by_subscript (cat, subscript); @@ -667,13 +694,10 @@ categoricals_get_code_for_case (const struct categoricals *cat, int subscript, const int index = ((subscript - base_index) % iap->df_prod[v] ) / dfp; dfp = iap->df_prod [v]; -#if EFFECTS_CODING - if ( valn->index == df ) + if (effects_coding && valn->index == df ) bin = -1.0; - else -#endif - if ( valn->index != index ) - bin = 0; + else if ( valn->index != index ) + bin = 0; result *= bin; } @@ -682,6 +706,28 @@ categoricals_get_code_for_case (const struct categoricals *cat, int subscript, } +/* Returns unity if the value in case C at SUBSCRIPT is equal to the category + for that subscript */ +double +categoricals_get_dummy_code_for_case (const struct categoricals *cat, int subscript, + const struct ccase *c) +{ + return categoricals_get_code_for_case (cat, subscript, c, false); +} + +/* Returns unity if the value in case C at SUBSCRIPT is equal to the category + for that subscript. + Else if it is the last category, return -1. + Otherwise return 0. + */ +double +categoricals_get_effects_code_for_case (const struct categoricals *cat, int subscript, + const struct ccase *c) +{ + return categoricals_get_code_for_case (cat, subscript, c, true); +} + + size_t categoricals_get_n_variables (const struct categoricals *cat) { @@ -695,8 +741,14 @@ categoricals_get_n_variables (const struct categoricals *cat) const struct ccase * categoricals_get_case_by_category_real (const struct categoricals *cat, int iact, int n) { + const struct interaction_value *vn; + const struct interact_params *vp = &cat->iap[iact]; - const struct interaction_value *vn = vp->reverse_interaction_value_map [n]; + + if ( n >= hmap_count (&vp->ivmap)) + return NULL; + + vn = vp->reverse_interaction_value_map [n]; return vn->ccase; } @@ -706,7 +758,12 @@ void * categoricals_get_user_data_by_category_real (const struct categoricals *cat, int iact, int n) { const struct interact_params *vp = &cat->iap[iact]; - const struct interaction_value *iv = vp->reverse_interaction_value_map [n]; + const struct interaction_value *iv ; + + if ( n >= hmap_count (&vp->ivmap)) + return NULL; + + iv = vp->reverse_interaction_value_map [n]; return iv->user_data; }