X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Flanguage%2Fstats%2Fctables.c;h=fb8bb3a113008c182828b1612af925af08673e1a;hb=bba95a8b863616cc849a7b89ec231efff5707b97;hp=7a1cb3b2ca08bc7149a0315e28c673dfe5e0a2d2;hpb=60f7b79a9bc29d1610b0ee59dcba1f6a8489c6e9;p=pspp diff --git a/src/language/stats/ctables.c b/src/language/stats/ctables.c index 7a1cb3b2ca..fb8bb3a113 100644 --- a/src/language/stats/ctables.c +++ b/src/language/stats/ctables.c @@ -22,6 +22,7 @@ #include "data/dataset.h" #include "data/dictionary.h" #include "data/mrset.h" +#include "data/value-labels.h" #include "language/command.h" #include "language/lexer/format-parser.h" #include "language/lexer/lexer.h" @@ -192,7 +193,7 @@ struct ctables_cell struct { - size_t stack_idx; + size_t nest_idx; struct ctables_cell_value { const struct ctables_category *category; @@ -277,13 +278,6 @@ struct ctables_postcompute_expr }; }; -enum ctables_label_position - { - CTLP_NORMAL, - CTLP_OPPOSITE, - CTLP_LAYER, - }; - struct ctables_summary_spec_set { struct ctables_summary_spec *specs; @@ -316,6 +310,13 @@ struct ctables_stack size_t n; }; +struct ctables_value + { + struct hmap_node node; + union value value; + int leaf; + }; + struct ctables_table { struct ctables_axis *axes[PIVOT_N_AXES]; @@ -325,11 +326,25 @@ struct ctables_table struct hmap cells; struct hmap domains[N_CTDTS]; + const struct variable *clabels_example; + struct hmap clabels_values_map; + union value *clabels_values; + size_t n_clabels_values; + enum pivot_axis_type slabels_axis; bool slabels_visible; - enum ctables_label_position row_labels; - enum ctables_label_position col_labels; + /* The innermost category labels for axis 'a' appear on axis label_axis[a]. + + Most commonly, label_axis[a] == a, and in particular we always have + label_axis{PIVOT_AXIS_LAYER] == PIVOT_AXIS_LAYER. + + If ROWLABELS or COLLABELS is specified, then one of + label_axis[PIVOT_AXIS_ROW] or label_axis[PIVOT_AXIS_COLUMN] can be the + opposite axis or PIVOT_AXIS_LAYER. Only one of them will differ. + */ + enum pivot_axis_type label_axis[PIVOT_N_AXES]; + enum pivot_axis_type clabels_from_axis; /* Indexed by variable dictionary index. */ struct ctables_categories **categories; @@ -451,6 +466,46 @@ ctables_category_uninit (struct ctables_category *cat) } } +static bool +ctables_category_equal (const struct ctables_category *a, + const struct ctables_category *b) +{ + if (a->type != b->type) + return false; + + switch (a->type) + { + case CCT_NUMBER: + return a->number == b->number; + + case CCT_STRING: + return strcmp (a->string, b->string); + + case CCT_RANGE: + return a->range[0] == b->range[0] && a->range[1] == b->range[1]; + + case CCT_MISSING: + case CCT_OTHERNM: + return true; + + case CCT_SUBTOTAL: + case CCT_HSUBTOTAL: + case CCT_TOTAL: + return !strcmp (a->total_label, b->total_label); + + case CCT_VALUE: + case CCT_LABEL: + case CCT_FUNCTION: + return (a->include_missing == b->include_missing + && a->sort_ascending == b->sort_ascending + && a->sort_function == b->sort_function + && a->sort_var == b->sort_var + && a->percentile == b->percentile); + } + + NOT_REACHED (); +} + static void ctables_categories_unref (struct ctables_categories *c) { @@ -467,6 +522,20 @@ ctables_categories_unref (struct ctables_categories *c) free (c); } +static bool +ctables_categories_equal (const struct ctables_categories *a, + const struct ctables_categories *b) +{ + if (a->n_cats != b->n_cats || a->show_empty != b->show_empty) + return false; + + for (size_t i = 0; i < a->n_cats; i++) + if (!ctables_category_equal (&a->cats[i], &b->cats[i])) + return false; + + return true; +} + /* Chi-square test (SIGTEST). */ struct ctables_chisq { @@ -2049,8 +2118,8 @@ ctables_cell_compare_3way (const void *a_, const void *b_, const void *aux_) const struct ctables_cell *a = *ap; const struct ctables_cell *b = *bp; - size_t a_idx = a->axes[aux->a].stack_idx; - size_t b_idx = b->axes[aux->a].stack_idx; + size_t a_idx = a->axes[aux->a].nest_idx; + size_t b_idx = b->axes[aux->a].nest_idx; if (a_idx != b_idx) return a_idx < b_idx ? -1 : 1; @@ -2139,7 +2208,7 @@ ctables_domain_insert (struct ctables_table *t, struct ctables_cell *cell, size_t hash = 0; for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++) { - size_t idx = cell->axes[a].stack_idx; + size_t idx = cell->axes[a].nest_idx; const struct ctables_nest *nest = &t->stacks[a].nests[idx]; hash = hash_int (idx, hash); for (size_t i = 0; i < nest->n_domains[domain]; i++) @@ -2156,8 +2225,8 @@ ctables_domain_insert (struct ctables_table *t, struct ctables_cell *cell, const struct ctables_cell *df = d->example; for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++) { - size_t idx = cell->axes[a].stack_idx; - if (idx != df->axes[a].stack_idx) + size_t idx = cell->axes[a].nest_idx; + if (idx != df->axes[a].nest_idx) goto not_equal; const struct ctables_nest *nest = &t->stacks[a].nests[idx]; @@ -2274,7 +2343,7 @@ ctables_cell_insert__ (struct ctables_table *t, const struct ccase *c, for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++) { const struct ctables_nest *nest = &t->stacks[a].nests[ix[a]]; - if (cell->axes[a].stack_idx != ix[a]) + if (cell->axes[a].nest_idx != ix[a]) goto not_equal; for (size_t i = 0; i < nest->n; i++) if (i != nest->scale_idx @@ -2299,7 +2368,7 @@ ctables_cell_insert__ (struct ctables_table *t, const struct ccase *c, for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++) { const struct ctables_nest *nest = &t->stacks[a].nests[ix[a]]; - cell->axes[a].stack_idx = ix[a]; + cell->axes[a].nest_idx = ix[a]; cell->axes[a].cvs = (nest->n ? xnmalloc (nest->n, sizeof *cell->axes[a].cvs) : NULL); @@ -2452,8 +2521,38 @@ merge_item_compare_3way (const struct merge_item *a, const struct merge_item *b) return strcmp (as->label, bs->label); } +static struct pivot_value * +ctables_category_create_label (const struct ctables_category *cat, + const struct variable *var, + const union value *value) +{ + return (cat->type == CCT_TOTAL || cat->type == CCT_SUBTOTAL || cat->type == CCT_HSUBTOTAL + ? pivot_value_new_user_text (cat->total_label, SIZE_MAX) + : pivot_value_new_var_value (var, value)); +} + +static struct ctables_value * +ctables_value_find__ (struct ctables_table *t, const union value *value, + int width, unsigned int hash) +{ + struct ctables_value *clv; + HMAP_FOR_EACH_WITH_HASH (clv, struct ctables_value, node, + hash, &t->clabels_values_map) + if (value_equal (value, &clv->value, width)) + return clv; + return NULL; +} + +static struct ctables_value * +ctables_value_find (struct ctables_table *t, + const union value *value, int width) +{ + return ctables_value_find__ (t, value, width, + value_hash (value, width, 0)); +} + static void -ctables_table_output_same_axis (struct ctables *ct, struct ctables_table *t) +ctables_table_output (struct ctables *ct, struct ctables_table *t) { struct pivot_table *pt = pivot_table_create__ ( (t->title @@ -2467,6 +2566,41 @@ ctables_table_output_same_axis (struct ctables *ct, struct ctables_table *t) pivot_table_set_caption ( pt, pivot_value_new_user_text (t->corner, SIZE_MAX)); + bool summary_dimension = t->summary_axis != t->slabels_axis; + if (summary_dimension) + { + struct pivot_dimension *d = pivot_dimension_create ( + pt, t->slabels_axis, N_("Summaries")); + const struct ctables_summary_spec_set *specs = &t->summary_specs; + for (size_t i = 0; i < specs->n; i++) + pivot_category_create_leaf ( + d->root, pivot_value_new_text (specs->specs[i].label)); + } + + bool categories_dimension = t->clabels_example != NULL; + if (categories_dimension) + { + struct pivot_dimension *d = pivot_dimension_create ( + pt, t->label_axis[t->clabels_from_axis], + t->clabels_from_axis == PIVOT_AXIS_ROW + ? N_("Row Categories") + : N_("Column Categories")); + const struct variable *var = t->clabels_example; + const struct ctables_categories *c = t->categories[var_get_dict_index (var)]; + for (size_t i = 0; i < t->n_clabels_values; i++) + { + const union value *value = &t->clabels_values[i]; + const struct ctables_category *cat = ctables_categories_match (c, value, var); + if (!cat) + { + pivot_category_create_leaf (d->root, pivot_value_new_integer (value->f)); /* XXX */ + continue; + } + pivot_category_create_leaf (d->root, ctables_category_create_label ( + cat, t->clabels_example, value)); + } + } + pivot_table_set_look (pt, ct->look); struct pivot_dimension *d[PIVOT_N_AXES]; for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++) @@ -2485,113 +2619,163 @@ ctables_table_output_same_axis (struct ctables *ct, struct ctables_table *t) assert (t->axes[a]); struct ctables_cell **sorted = xnmalloc (t->cells.count, sizeof *sorted); + size_t n_sorted = 0; struct ctables_cell *cell; - size_t n = 0; HMAP_FOR_EACH (cell, struct ctables_cell, node, &t->cells) if (!cell->hide) - sorted[n++] = cell; - assert (n <= t->cells.count); + sorted[n_sorted++] = cell; + assert (n_sorted <= t->cells.count); struct ctables_cell_sort_aux aux = { .t = t, .a = a }; - sort (sorted, n, sizeof *sorted, ctables_cell_compare_3way, &aux); + sort (sorted, n_sorted, sizeof *sorted, ctables_cell_compare_3way, &aux); size_t max_depth = 0; for (size_t j = 0; j < t->stacks[a].n; j++) if (t->stacks[a].nests[j].n > max_depth) max_depth = t->stacks[a].nests[j].n; - struct pivot_category **groups = xnmalloc (max_depth, sizeof *groups); - struct pivot_category *top = NULL; + /* Pivot categories: + + - variable label for nest->vars[0], if vlabel != CTVL_NONE + - category for nest->vars[0], if nest->scale_idx != 0 + - variable label for nest->vars[1], if vlabel != CTVL_NONE + - category for nest->vars[1], if nest->scale_idx != 1 + ... + - variable label for nest->vars[n - 1], if vlabel != CTVL_NONE + - category for nest->vars[n - 1], if t->label_axis[a] == a && nest->scale_idx != n - 1. + - summary function, if 'a == t->slabels_axis && a == + t->summary_axis'. + + Additional dimensions: + + - If 'a == t->slabels_axis && a != t->summary_axis', add a summary + dimension. + - If 't->label_axis[b] == a' for some 'b != a', add a category + dimension to 'a'. + */ + + struct ctables_level + { + enum ctables_level_type + { + CTL_VAR, /* Variable label for nest->vars[var_idx]. */ + CTL_CATEGORY, /* Category for nest->vars[var_idx]. */ + CTL_SUMMARY, /* Summary functions. */ + } + type; + + size_t var_idx; + }; + struct ctables_level *levels = xnmalloc (1 + 2 * max_depth, sizeof *levels); + size_t n_levels = 0; + + struct pivot_category **groups = xnmalloc (1 + 2 * max_depth, sizeof *groups); int prev_leaf = 0; - for (size_t j = 0; j < n; j++) + for (size_t j = 0; j < n_sorted; j++) { struct ctables_cell *cell = sorted[j]; - const struct ctables_nest *nest = &t->stacks[a].nests[cell->axes[a].stack_idx]; + struct ctables_cell *prev = j > 0 ? sorted[j - 1] : NULL; + const struct ctables_nest *nest = &t->stacks[a].nests[cell->axes[a].nest_idx]; - size_t n_common = 0; - bool new_subtable = false; - if (j > 0) + bool new_subtable = !prev || prev->axes[a].nest_idx != cell->axes[a].nest_idx; + if (new_subtable) { - struct ctables_cell *prev = sorted[j - 1]; - if (prev->axes[a].stack_idx == cell->axes[a].stack_idx) + n_levels = 0; + printf ("%s levels:", pivot_axis_type_to_string (a)); + for (size_t k = 0; k < nest->n; k++) { - for (; n_common < nest->n; n_common++) - if (n_common != nest->scale_idx - && (prev->axes[a].cvs[n_common].category - != cell->axes[a].cvs[n_common].category - || !value_equal (&prev->axes[a].cvs[n_common].value, - &cell->axes[a].cvs[n_common].value, - var_get_type (nest->vars[n_common])))) - break; + enum ctables_vlabel vlabel = ct->vlabels[var_get_dict_index (nest->vars[k])]; + if (vlabel != CTVL_NONE) + { + printf (" var(%s)", var_get_name (nest->vars[k])); + levels[n_levels++] = (struct ctables_level) { + .type = CTL_VAR, + .var_idx = k, + }; + } + + if (nest->scale_idx != k + && (k != nest->n - 1 || t->label_axis[a] == a)) + { + printf (" category(%s)", var_get_name (nest->vars[k])); + levels[n_levels++] = (struct ctables_level) { + .type = CTL_CATEGORY, + .var_idx = k, + }; + } } - else - new_subtable = true; - } - else - new_subtable = true; - if (new_subtable) - { - enum ctables_vlabel vlabel = ct->vlabels[var_get_dict_index (nest->vars[0])]; - top = d[a]->root; - if (vlabel != CTVL_NONE) - top = pivot_category_create_group__ ( - top, pivot_value_new_variable (nest->vars[0])); - } - if (n_common == nest->n) - { - cell->axes[a].leaf = prev_leaf; - continue; + if (a == t->slabels_axis && a == t->summary_axis) + { + printf (" summary"); + levels[n_levels++] = (struct ctables_level) { + .type = CTL_SUMMARY, + .var_idx = SIZE_MAX, + }; + } + printf ("\n"); } - for (size_t k = n_common; k < nest->n; k++) + size_t n_common = 0; + if (!new_subtable) { - struct pivot_category *parent = k > 0 ? groups[k - 1] : top; - - struct pivot_value *label - = (k == nest->scale_idx ? NULL - : (cell->axes[a].cvs[k].category->type == CCT_TOTAL - || cell->axes[a].cvs[k].category->type == CCT_SUBTOTAL - || cell->axes[a].cvs[k].category->type == CCT_HSUBTOTAL) - ? pivot_value_new_user_text (cell->axes[a].cvs[k].category->total_label, - SIZE_MAX) - : pivot_value_new_var_value (nest->vars[k], - &cell->axes[a].cvs[k].value)); - if (k == nest->n - 1) + for (; n_common < n_levels; n_common++) { - if (a == t->summary_axis) + const struct ctables_level *level = &levels[n_common]; + if (level->type == CTL_CATEGORY) { - if (label) - parent = pivot_category_create_group__ (parent, label); - const struct ctables_summary_spec_set *specs = &nest->specs[cell->sv]; - for (size_t m = 0; m < specs->n; m++) + size_t var_idx = level->var_idx; + if (prev->axes[a].cvs[var_idx].category + != cell->axes[a].cvs[var_idx].category) { - int leaf = pivot_category_create_leaf ( - parent, pivot_value_new_text (specs->specs[m].label)); - if (m == 0) - prev_leaf = leaf; + break; + } + else if (!value_equal (&prev->axes[a].cvs[var_idx].value, + &cell->axes[a].cvs[var_idx].value, + var_get_type (nest->vars[var_idx]))) + { + break; } } - else - { - /* This assertion is true as long as the summary axis - is the axis where the summaries are displayed. */ - assert (label); + } + } - prev_leaf = pivot_category_create_leaf (parent, label); + for (size_t k = n_common; k < n_levels; k++) + { + const struct ctables_level *level = &levels[k]; + struct pivot_category *parent = k ? groups[k - 1] : d[a]->root; + if (level->type == CTL_SUMMARY) + { + const struct ctables_summary_spec_set *specs = &t->summary_specs; + for (size_t m = 0; m < specs->n; m++) + { + int leaf = pivot_category_create_leaf ( + parent, pivot_value_new_text (specs->specs[m].label)); + if (!m) + prev_leaf = leaf; } - break; } + else + { + const struct variable *var = nest->vars[level->var_idx]; + struct pivot_value *label; + if (level->type == CTL_VAR) + label = pivot_value_new_variable (var); + else if (level->type == CTL_CATEGORY) + { + const struct ctables_cell_value *cv = &cell->axes[a].cvs[level->var_idx]; + label = ctables_category_create_label (cv->category, + var, &cv->value); + } + else + NOT_REACHED (); - if (label) - parent = pivot_category_create_group__ (parent, label); - - enum ctables_vlabel vlabel = ct->vlabels[var_get_dict_index (nest->vars[k + 1])]; - if (vlabel != CTVL_NONE) - parent = pivot_category_create_group__ ( - parent, pivot_value_new_variable (nest->vars[k + 1])); - groups[k] = parent; + if (k == n_levels - 1) + prev_leaf = pivot_category_create_leaf (parent, label); + else + groups[k] = pivot_category_create_group__ (parent, label); + } } cell->axes[a].leaf = prev_leaf; @@ -2599,24 +2783,40 @@ ctables_table_output_same_axis (struct ctables *ct, struct ctables_table *t) free (sorted); free (groups); } + struct ctables_cell *cell; HMAP_FOR_EACH (cell, struct ctables_cell, node, &t->cells) { if (cell->hide) continue; - const struct ctables_nest *nest = &t->stacks[t->summary_axis].nests[cell->axes[t->summary_axis].stack_idx]; - const struct ctables_summary_spec_set *specs = &nest->specs[cell->sv]; + const struct ctables_nest *specs_nest = &t->stacks[t->summary_axis].nests[cell->axes[t->summary_axis].nest_idx]; + const struct ctables_summary_spec_set *specs = &specs_nest->specs[cell->sv]; for (size_t j = 0; j < specs->n; j++) { - size_t dindexes[3]; + size_t dindexes[5]; size_t n_dindexes = 0; + if (summary_dimension) + dindexes[n_dindexes++] = specs->specs[j].axis_idx; + + if (categories_dimension) + { + const struct ctables_nest *clabels_nest = &t->stacks[t->clabels_from_axis].nests[cell->axes[t->clabels_from_axis].nest_idx]; + const struct variable *var = clabels_nest->vars[clabels_nest->n - 1]; + const union value *value = &cell->axes[t->clabels_from_axis].cvs[clabels_nest->n - 1].value; + const struct ctables_value *ctv = ctables_value_find (t, value, var_get_width (var)); + + printf ("leaf=%d\n", ctv ? ctv->leaf : 0); + dindexes[n_dindexes++] = ctv ? ctv->leaf : 0; /* XXX */ + //dindexes[n_dindexes++] = 0; + } + for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++) if (d[a]) { int leaf = cell->axes[a].leaf; - if (a == t->summary_axis) + if (a == t->summary_axis && !summary_dimension) leaf += j; dindexes[n_dindexes++] = leaf; } @@ -2633,311 +2833,219 @@ ctables_table_output_same_axis (struct ctables *ct, struct ctables_table *t) static void -ctables_table_output_different_axis (struct ctables *ct, struct ctables_table *t) +ctables_prepare_table (struct ctables_table *t) { - struct pivot_table *pt = pivot_table_create__ ( - (t->title - ? pivot_value_new_user_text (t->title, SIZE_MAX) - : pivot_value_new_text (N_("Custom Tables"))), - "Custom Tables"); - if (t->caption) - pivot_table_set_caption ( - pt, pivot_value_new_user_text (t->caption, SIZE_MAX)); - if (t->corner) - pivot_table_set_caption ( - pt, pivot_value_new_user_text (t->corner, SIZE_MAX)); - - pivot_table_set_look (pt, ct->look); - struct pivot_dimension *d[PIVOT_N_AXES]; for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++) - { - static const char *names[] = { - [PIVOT_AXIS_ROW] = N_("Rows"), - [PIVOT_AXIS_COLUMN] = N_("Columns"), - [PIVOT_AXIS_LAYER] = N_("Layers"), - }; - d[a] = (t->axes[a] || a == t->summary_axis - ? pivot_dimension_create (pt, a, names[a]) - : NULL); - if (!d[a]) - continue; - - assert (t->axes[a]); - - struct ctables_cell **sorted = xnmalloc (t->cells.count, sizeof *sorted); - - struct ctables_cell *cell; - size_t n = 0; - HMAP_FOR_EACH (cell, struct ctables_cell, node, &t->cells) - if (!cell->hide) - sorted[n++] = cell; - assert (n <= t->cells.count); - - struct ctables_cell_sort_aux aux = { .t = t, .a = a }; - sort (sorted, n, sizeof *sorted, ctables_cell_compare_3way, &aux); - - size_t max_depth = 0; - for (size_t j = 0; j < t->stacks[a].n; j++) - if (t->stacks[a].nests[j].n > max_depth) - max_depth = t->stacks[a].nests[j].n; - - struct pivot_category **groups = xnmalloc (max_depth, sizeof *groups); - struct pivot_category *top = NULL; - int prev_leaf = 0; - for (size_t j = 0; j < n; j++) - { - struct ctables_cell *cell = sorted[j]; - const struct ctables_nest *nest = &t->stacks[a].nests[cell->axes[a].stack_idx]; - - size_t n_common = 0; - bool new_subtable = false; - if (j > 0) - { - struct ctables_cell *prev = sorted[j - 1]; - if (prev->axes[a].stack_idx == cell->axes[a].stack_idx) - { - for (; n_common < nest->n; n_common++) - if (n_common != nest->scale_idx - && (prev->axes[a].cvs[n_common].category - != cell->axes[a].cvs[n_common].category - || !value_equal (&prev->axes[a].cvs[n_common].value, - &cell->axes[a].cvs[n_common].value, - var_get_type (nest->vars[n_common])))) - break; - } - else - new_subtable = true; - } - else - new_subtable = true; - - if (new_subtable) - { - enum ctables_vlabel vlabel = ct->vlabels[var_get_dict_index (nest->vars[0])]; - top = d[a]->root; - if (vlabel != CTVL_NONE) - top = pivot_category_create_group__ ( - top, pivot_value_new_variable (nest->vars[0])); - } - if (n_common == nest->n) - { - cell->axes[a].leaf = prev_leaf; - continue; - } - - for (size_t k = n_common; k < nest->n; k++) - { - struct pivot_category *parent = k > 0 ? groups[k - 1] : top; - - struct pivot_value *label - = (k == nest->scale_idx ? NULL - : (cell->axes[a].cvs[k].category->type == CCT_TOTAL - || cell->axes[a].cvs[k].category->type == CCT_SUBTOTAL - || cell->axes[a].cvs[k].category->type == CCT_HSUBTOTAL) - ? pivot_value_new_user_text (cell->axes[a].cvs[k].category->total_label, - SIZE_MAX) - : pivot_value_new_var_value (nest->vars[k], - &cell->axes[a].cvs[k].value)); - if (k == nest->n - 1) - { - if (a == t->slabels_axis) - { - if (label) - parent = pivot_category_create_group__ (parent, label); - const struct ctables_summary_spec_set *specs = &t->summary_specs; - for (size_t m = 0; m < specs->n; m++) - { - int leaf = pivot_category_create_leaf ( - parent, pivot_value_new_text (specs->specs[m].label)); - if (m == 0) - prev_leaf = leaf; - } - } - else - { - prev_leaf = pivot_category_create_leaf (parent, label ? label : pivot_value_new_user_text ("text", SIZE_MAX)); - } - break; - } - - if (label) - parent = pivot_category_create_group__ (parent, label); - - enum ctables_vlabel vlabel = ct->vlabels[var_get_dict_index (nest->vars[k + 1])]; - if (vlabel != CTVL_NONE) - parent = pivot_category_create_group__ ( - parent, pivot_value_new_variable (nest->vars[k + 1])); - groups[k] = parent; - } - - cell->axes[a].leaf = prev_leaf; - } - free (sorted); - free (groups); - } - pivot_table_submit (pt); -} - + if (t->axes[a]) + { + t->stacks[a] = enumerate_fts (a, t->axes[a]); -static bool -ctables_execute (struct dataset *ds, struct ctables *ct) -{ - for (size_t i = 0; i < ct->n_tables; i++) - { - struct ctables_table *t = ct->tables[i]; - for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++) - if (t->axes[a]) + for (size_t j = 0; j < t->stacks[a].n; j++) { - t->stacks[a] = enumerate_fts (a, t->axes[a]); - - for (size_t j = 0; j < t->stacks[a].n; j++) + struct ctables_nest *nest = &t->stacks[a].nests[j]; + for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++) { - struct ctables_nest *nest = &t->stacks[a].nests[j]; - for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++) + nest->domains[dt] = xmalloc (nest->n * sizeof *nest->domains[dt]); + nest->n_domains[dt] = 0; + + for (size_t k = 0; k < nest->n; k++) { - nest->domains[dt] = xmalloc (nest->n * sizeof *nest->domains[dt]); - nest->n_domains[dt] = 0; + if (k == nest->scale_idx) + continue; - for (size_t k = 0; k < nest->n; k++) + switch (dt) { - if (k == nest->scale_idx) - continue; + case CTDT_TABLE: + continue; - switch (dt) + case CTDT_LAYER: + if (a != PIVOT_AXIS_LAYER) + continue; + break; + + case CTDT_SUBTABLE: + case CTDT_ROW: + case CTDT_COL: + if (dt == CTDT_SUBTABLE ? a != PIVOT_AXIS_LAYER + : dt == CTDT_ROW ? a == PIVOT_AXIS_COLUMN + : a == PIVOT_AXIS_ROW) { - case CTDT_TABLE: - continue; - - case CTDT_LAYER: - if (a != PIVOT_AXIS_LAYER) - continue; - break; - - case CTDT_SUBTABLE: - case CTDT_ROW: - case CTDT_COL: - if (dt == CTDT_SUBTABLE ? a != PIVOT_AXIS_LAYER - : dt == CTDT_ROW ? a == PIVOT_AXIS_COLUMN - : a == PIVOT_AXIS_ROW) - { - if (k == nest->n - 1 - || (nest->scale_idx == nest->n - 1 - && k == nest->n - 2)) - continue; - } - break; - - case CTDT_LAYERROW: - if (a == PIVOT_AXIS_COLUMN) + if (k == nest->n - 1 + || (nest->scale_idx == nest->n - 1 + && k == nest->n - 2)) continue; - break; - - case CTDT_LAYERCOL: - if (a == PIVOT_AXIS_ROW) - continue; - break; } + break; + + case CTDT_LAYERROW: + if (a == PIVOT_AXIS_COLUMN) + continue; + break; - nest->domains[dt][nest->n_domains[dt]++] = k; + case CTDT_LAYERCOL: + if (a == PIVOT_AXIS_ROW) + continue; + break; } + + nest->domains[dt][nest->n_domains[dt]++] = k; } } } - else - { - struct ctables_nest *nest = xmalloc (sizeof *nest); - *nest = (struct ctables_nest) { .n = 0 }; - t->stacks[a] = (struct ctables_stack) { .nests = nest, .n = 1 }; - } + } + else + { + struct ctables_nest *nest = xmalloc (sizeof *nest); + *nest = (struct ctables_nest) { .n = 0 }; + t->stacks[a] = (struct ctables_stack) { .nests = nest, .n = 1 }; + } - struct ctables_stack *stack = &t->stacks[t->summary_axis]; - for (size_t i = 0; i < stack->n; i++) + struct ctables_stack *stack = &t->stacks[t->summary_axis]; + for (size_t i = 0; i < stack->n; i++) + { + struct ctables_nest *nest = &stack->nests[i]; + if (!nest->specs[CSV_CELL].n) { - struct ctables_nest *nest = &stack->nests[i]; - if (!nest->specs[CSV_CELL].n) - { - struct ctables_summary_spec_set *specs = &nest->specs[CSV_CELL]; - specs->specs = xmalloc (sizeof *specs->specs); - specs->n = 1; - - enum ctables_summary_function function - = specs->var ? CTSF_MEAN : CTSF_COUNT; - struct ctables_var var = { .is_mrset = false, .var = specs->var }; - - *specs->specs = (struct ctables_summary_spec) { - .function = function, - .format = ctables_summary_default_format (function, &var), - .label = ctables_summary_default_label (function, 0), - }; - if (!specs->var) - specs->var = nest->vars[0]; + struct ctables_summary_spec_set *specs = &nest->specs[CSV_CELL]; + specs->specs = xmalloc (sizeof *specs->specs); + specs->n = 1; + + enum ctables_summary_function function + = specs->var ? CTSF_MEAN : CTSF_COUNT; + struct ctables_var var = { .is_mrset = false, .var = specs->var }; + + *specs->specs = (struct ctables_summary_spec) { + .function = function, + .format = ctables_summary_default_format (function, &var), + .label = ctables_summary_default_label (function, 0), + }; + if (!specs->var) + specs->var = nest->vars[0]; - ctables_summary_spec_set_clone (&nest->specs[CSV_TOTAL], - &nest->specs[CSV_CELL]); - } - else if (!nest->specs[CSV_TOTAL].n) - ctables_summary_spec_set_clone (&nest->specs[CSV_TOTAL], - &nest->specs[CSV_CELL]); + ctables_summary_spec_set_clone (&nest->specs[CSV_TOTAL], + &nest->specs[CSV_CELL]); } + else if (!nest->specs[CSV_TOTAL].n) + ctables_summary_spec_set_clone (&nest->specs[CSV_TOTAL], + &nest->specs[CSV_CELL]); + } - struct ctables_summary_spec_set *merged = &t->summary_specs; - struct merge_item *items = xnmalloc (2 * stack->n, sizeof *items); - size_t n_left = 0; - for (size_t j = 0; j < stack->n; j++) - { - const struct ctables_nest *nest = &stack->nests[j]; - if (nest->n) - for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++) - items[n_left++] = (struct merge_item) { .set = &nest->specs[sv] }; - } + struct ctables_summary_spec_set *merged = &t->summary_specs; + struct merge_item *items = xnmalloc (2 * stack->n, sizeof *items); + size_t n_left = 0; + for (size_t j = 0; j < stack->n; j++) + { + const struct ctables_nest *nest = &stack->nests[j]; + if (nest->n) + for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++) + items[n_left++] = (struct merge_item) { .set = &nest->specs[sv] }; + } - while (n_left > 0) - { - struct merge_item min = items[0]; - for (size_t j = 1; j < n_left; j++) - if (merge_item_compare_3way (&items[j], &min) < 0) - min = items[j]; + while (n_left > 0) + { + struct merge_item min = items[0]; + for (size_t j = 1; j < n_left; j++) + if (merge_item_compare_3way (&items[j], &min) < 0) + min = items[j]; - if (merged->n >= merged->allocated) - merged->specs = x2nrealloc (merged->specs, &merged->allocated, - sizeof *merged->specs); - merged->specs[merged->n++] = min.set->specs[min.ofs]; + if (merged->n >= merged->allocated) + merged->specs = x2nrealloc (merged->specs, &merged->allocated, + sizeof *merged->specs); + merged->specs[merged->n++] = min.set->specs[min.ofs]; - for (size_t j = 0; j < n_left; ) + for (size_t j = 0; j < n_left; ) + { + if (merge_item_compare_3way (&items[j], &min) == 0) { - if (merge_item_compare_3way (&items[j], &min) == 0) + struct merge_item *item = &items[j]; + item->set->specs[item->ofs].axis_idx = merged->n - 1; + if (++item->ofs >= item->set->n) { - struct merge_item *item = &items[j]; - item->set->specs[item->ofs].axis_idx = merged->n - 1; - if (++item->ofs >= item->set->n) - { - items[j] = items[--n_left]; - continue; - } + items[j] = items[--n_left]; + continue; } - j++; } + j++; } + } #if 0 - for (size_t j = 0; j < merged->n; j++) - printf ("%s\n", ctables_summary_function_name (merged->specs[j].function)); + for (size_t j = 0; j < merged->n; j++) + printf ("%s\n", ctables_summary_function_name (merged->specs[j].function)); - for (size_t j = 0; j < stack->n; j++) + for (size_t j = 0; j < stack->n; j++) + { + const struct ctables_nest *nest = &stack->nests[j]; + for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++) { - const struct ctables_nest *nest = &stack->nests[j]; - for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++) - { - const struct ctables_summary_spec_set *specs = &nest->specs[sv]; - for (size_t k = 0; k < specs->n; k++) - printf ("(%s, %zu) ", ctables_summary_function_name (specs->specs[k].function), - specs->specs[k].axis_idx); - printf ("\n"); - } + const struct ctables_summary_spec_set *specs = &nest->specs[sv]; + for (size_t k = 0; k < specs->n; k++) + printf ("(%s, %zu) ", ctables_summary_function_name (specs->specs[k].function), + specs->specs[k].axis_idx); + printf ("\n"); } + } #endif +} + +static void +ctables_insert_clabels_values (struct ctables_table *t, const struct ccase *c, + enum pivot_axis_type a) +{ + struct ctables_stack *stack = &t->stacks[a]; + for (size_t i = 0; i < stack->n; i++) + { + const struct ctables_nest *nest = &stack->nests[i]; + const struct variable *v = nest->vars[nest->n - 1]; + int width = var_get_width (v); + const union value *value = case_data (c, v); + unsigned int hash = value_hash (value, width, 0); + + struct ctables_value *clv = ctables_value_find__ (t, value, width, hash); + if (!clv) + { + clv = xmalloc (sizeof *clv); + value_clone (&clv->value, value, width); + hmap_insert (&t->clabels_values_map, &clv->node, hash); + } } +} + +static int +compare_clabels_values_3way (const void *a_, const void *b_, const void *width_) +{ + const union value *a = a_; + const union value *b = b_; + const int *width = width_; + return value_compare_3way (a, b, *width); +} + +static void +ctables_sort_clabels_values (struct ctables_table *t) +{ + int width = var_get_width (t->clabels_example); + + size_t n = hmap_count (&t->clabels_values_map); + t->clabels_values = xnmalloc (n, sizeof *t->clabels_values); + + struct ctables_value *clv; + size_t i = 0; + HMAP_FOR_EACH (clv, struct ctables_value, node, &t->clabels_values_map) + { + clv->leaf = i; + t->clabels_values[i] = clv->value; + i++; + } + t->n_clabels_values = n; + assert (i == n); + sort (t->clabels_values, n, sizeof *t->clabels_values, + compare_clabels_values_3way, &width); +} + +static bool +ctables_execute (struct dataset *ds, struct ctables *ct) +{ struct casereader *input = casereader_create_filter_weight (proc_open (ds), dataset_dict (ds), NULL, NULL); @@ -2958,6 +3066,10 @@ ctables_execute (struct dataset *ds, struct ctables *ct) for (size_t ic = 0; ic < t->stacks[PIVOT_AXIS_COLUMN].n; ic++) for (size_t il = 0; il < t->stacks[PIVOT_AXIS_LAYER].n; il++) ctables_cell_insert (t, c, ir, ic, il, weight); + + for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++) + if (t->label_axis[a] != a) + ctables_insert_clabels_values (t, c, a); } } casereader_destroy (input); @@ -2965,14 +3077,102 @@ ctables_execute (struct dataset *ds, struct ctables *ct) for (size_t i = 0; i < ct->n_tables; i++) { struct ctables_table *t = ct->tables[i]; - if (t->summary_axis == t->slabels_axis) - ctables_table_output_same_axis (ct, ct->tables[i]); - else - ctables_table_output_different_axis (ct, ct->tables[i]); + + if (t->clabels_example) + ctables_sort_clabels_values (t); + + ctables_table_output (ct, ct->tables[i]); } return proc_commit (ds); } +static bool +ctables_check_label_position (struct ctables_table *t, enum pivot_axis_type a) +{ + enum pivot_axis_type label_pos = t->label_axis[a]; + if (label_pos == a) + return true; + + t->clabels_from_axis = a; + + const char *subcommand_name = a == PIVOT_AXIS_ROW ? "ROWLABELS" : "COLLABELS"; + const char *pos_name = label_pos == PIVOT_AXIS_LAYER ? "LAYER" : "OPPOSITE"; + + const struct ctables_stack *stack = &t->stacks[a]; + if (!stack->n) + return true; + + const struct ctables_nest *n0 = &stack->nests[0]; + assert (n0->n > 0); + const struct variable *v0 = n0->vars[n0->n - 1]; + struct ctables_categories *c0 = t->categories[var_get_dict_index (v0)]; + t->clabels_example = v0; + + for (size_t i = 0; i < c0->n_cats; i++) + if (c0->cats[i].type == CCT_FUNCTION) + { + msg (SE, _("%s=%s is not allowed with sorting based " + "on a summary function."), + subcommand_name, pos_name); + return false; + } + if (n0->n - 1 == n0->scale_idx) + { + msg (SE, _("%s=%s requires the variables to be moved to be categorical, " + "but %s is a scale variable."), + subcommand_name, pos_name, var_get_name (v0)); + return false; + } + + for (size_t i = 1; i < stack->n; i++) + { + const struct ctables_nest *ni = &stack->nests[i]; + assert (ni->n > 0); + const struct variable *vi = ni->vars[ni->n - 1]; + struct ctables_categories *ci = t->categories[var_get_dict_index (vi)]; + + if (ni->n - 1 == ni->scale_idx) + { + msg (SE, _("%s=%s requires the variables to be moved to be " + "categorical, but %s is a scale variable."), + subcommand_name, pos_name, var_get_name (vi)); + return false; + } + if (var_get_width (v0) != var_get_width (vi)) + { + msg (SE, _("%s=%s requires the variables to be " + "moved to have the same width, but %s has " + "width %d and %s has width %d."), + subcommand_name, pos_name, + var_get_name (v0), var_get_width (v0), + var_get_name (vi), var_get_width (vi)); + return false; + } + if (!val_labs_equal (var_get_value_labels (v0), + var_get_value_labels (vi))) + { + msg (SE, _("%s=%s requires the variables to be " + "moved to have the same value labels, but %s " + "and %s have different value labels."), + subcommand_name, pos_name, + var_get_name (v0), var_get_name (vi)); + return false; + } + if (!ctables_categories_equal (c0, ci)) + { + msg (SE, _("%s=%s requires the variables to be " + "moved to have the same category " + "specifications, but %s and %s have different " + "category specifications."), + subcommand_name, pos_name, + var_get_name (v0), var_get_name (vi)); + return false; + } + } + + return true; +} + int cmd_ctables (struct lexer *lexer, struct dataset *ds) { @@ -3210,8 +3410,13 @@ cmd_ctables (struct lexer *lexer, struct dataset *ds) .cells = HMAP_INITIALIZER (t->cells), .slabels_axis = PIVOT_AXIS_COLUMN, .slabels_visible = true, - .row_labels = CTLP_NORMAL, - .col_labels = CTLP_NORMAL, + .clabels_values_map = HMAP_INITIALIZER (t->clabels_values_map), + .label_axis = { + [PIVOT_AXIS_ROW] = PIVOT_AXIS_ROW, + [PIVOT_AXIS_COLUMN] = PIVOT_AXIS_COLUMN, + [PIVOT_AXIS_LAYER] = PIVOT_AXIS_LAYER, + }, + .clabels_from_axis = PIVOT_AXIS_LAYER, .categories = categories, .n_categories = n_vars, .cilevel = 95, @@ -3342,14 +3547,17 @@ cmd_ctables (struct lexer *lexer, struct dataset *ds) while (lex_token (lexer) != T_SLASH && lex_token (lexer) != T_ENDCMD) { if (lex_match_id (lexer, "AUTO")) - t->row_labels = t->col_labels = CTLP_NORMAL; + { + t->label_axis[PIVOT_AXIS_ROW] = PIVOT_AXIS_ROW; + t->label_axis[PIVOT_AXIS_COLUMN] = PIVOT_AXIS_COLUMN; + } else if (lex_match_id (lexer, "ROWLABELS")) { lex_match (lexer, T_EQUALS); if (lex_match_id (lexer, "OPPOSITE")) - t->row_labels = CTLP_OPPOSITE; + t->label_axis[PIVOT_AXIS_ROW] = PIVOT_AXIS_COLUMN; else if (lex_match_id (lexer, "LAYER")) - t->row_labels = CTLP_LAYER; + t->label_axis[PIVOT_AXIS_ROW] = PIVOT_AXIS_LAYER; else { lex_error_expecting (lexer, "OPPOSITE", "LAYER"); @@ -3360,9 +3568,9 @@ cmd_ctables (struct lexer *lexer, struct dataset *ds) { lex_match (lexer, T_EQUALS); if (lex_match_id (lexer, "OPPOSITE")) - t->col_labels = CTLP_OPPOSITE; + t->label_axis[PIVOT_AXIS_COLUMN] = PIVOT_AXIS_ROW; else if (lex_match_id (lexer, "LAYER")) - t->col_labels = CTLP_LAYER; + t->label_axis[PIVOT_AXIS_COLUMN] = PIVOT_AXIS_LAYER; else { lex_error_expecting (lexer, "OPPOSITE", "LAYER"); @@ -3637,12 +3845,17 @@ cmd_ctables (struct lexer *lexer, struct dataset *ds) break; } - if (t->row_labels != CTLP_NORMAL && t->col_labels != CTLP_NORMAL) + if (t->label_axis[PIVOT_AXIS_ROW] != PIVOT_AXIS_ROW + && t->label_axis[PIVOT_AXIS_COLUMN] != PIVOT_AXIS_COLUMN) { msg (SE, _("ROWLABELS and COLLABELS may not both be specified.")); goto error; } + ctables_prepare_table (t); + + ctables_check_label_position (t, PIVOT_AXIS_ROW); + ctables_check_label_position (t, PIVOT_AXIS_COLUMN); } while (lex_token (lexer) != T_ENDCMD);