#include <math.h>
#include "data/casereader.h"
+#include "data/casewriter.h"
#include "data/dataset.h"
#include "data/dictionary.h"
#include "data/mrset.h"
+#include "data/subcase.h"
#include "data/value-labels.h"
#include "language/command.h"
#include "language/lexer/format-parser.h"
#include "libpspp/message.h"
#include "libpspp/string-array.h"
#include "math/moments.h"
+#include "math/percentiles.h"
+#include "math/sort.h"
#include "output/pivot-table.h"
#include "gl/minmax.h"
struct hmap_node node;
/* The domains that contain this cell. */
+ bool contributes_to_domains;
struct ctables_domain *domains[N_CTDTS];
bool hide;
const struct variable *clabels_example;
struct hmap clabels_values_map;
- union value *clabels_values;
+ struct ctables_value **clabels_values;
size_t n_clabels_values;
enum pivot_axis_type slabels_axis;
/* MEAN, SEMEAN, STDDEV, SUM, VARIANCE, *.SUM. */
struct moments1 *moments;
- /* XXX percentiles, median, mode, multiple response */
+ struct
+ {
+ struct casewriter *writer;
+ double mvalid;
+ double median;
+ };
+
+ /* XXX percentiles, mode, multiple response */
};
static void
case CTSF_LAYERPCT_TOTALN:
case CTSF_LAYERROWPCT_TOTALN:
case CTSF_LAYERCOLPCT_TOTALN:
+ case CTSF_MISSING:
case CSTF_TOTALN:
case CTSF_ETOTALN:
case CTSF_VALIDN:
break;
case CTSF_MEDIAN:
- case CTSF_MISSING:
+ {
+ struct caseproto *proto = caseproto_create ();
+ proto = caseproto_add_width (proto, 0);
+ proto = caseproto_add_width (proto, 0);
+
+ struct subcase ordering;
+ subcase_init (&ordering, 0, 0, SC_ASCEND);
+ s->writer = sort_create_writer (&ordering, proto);
+ subcase_uninit (&ordering);
+ caseproto_unref (proto);
+
+ s->mvalid = 0;
+ s->median = SYSMIS;
+ }
+ break;
+
case CTSF_MODE:
case CTSF_PTILE:
NOT_REACHED ();
case CTSF_LAYERPCT_TOTALN:
case CTSF_LAYERROWPCT_TOTALN:
case CTSF_LAYERCOLPCT_TOTALN:
+ case CTSF_MISSING:
case CSTF_TOTALN:
case CTSF_ETOTALN:
case CTSF_VALIDN:
break;
case CTSF_MEDIAN:
- case CTSF_MISSING:
+ casewriter_destroy (s->writer);
+ break;
+
case CTSF_MODE:
case CTSF_PTILE:
NOT_REACHED ();
case CTSF_LAYERPCT_TOTALN:
case CTSF_LAYERROWPCT_TOTALN:
case CTSF_LAYERCOLPCT_TOTALN:
+ case CTSF_MISSING:
case CSTF_TOTALN:
case CTSF_ETOTALN:
case CTSF_VALIDN:
case CTSF_LAYERPCT_SUM:
case CTSF_LAYERROWPCT_SUM:
case CTSF_LAYERCOLPCT_SUM:
- moments1_add (s->moments, value->f, weight);
+ if (!var_is_value_missing (var, value))
+ moments1_add (s->moments, value->f, weight);
break;
case CTSF_MEDIAN:
- case CTSF_MISSING:
+ if (var_is_value_missing (var, value))
+ {
+ s->mvalid += weight;
+
+ struct ccase *c = case_create (casewriter_get_proto (s->writer));
+ *case_num_rw_idx (c, 0) = value->f;
+ *case_num_rw_idx (c, 1) = weight;
+ casewriter_write (s->writer, c);
+ }
+ break;
+
case CTSF_MODE:
case CTSF_PTILE:
NOT_REACHED ();
case CTSF_LAYERCOLPCT_TOTALN:
NOT_REACHED ();
+ case CTSF_MISSING:
+ return s->missing;
+
case CSTF_TOTALN:
case CTSF_ETOTALN:
return s->valid + s->missing;
NOT_REACHED ();
case CTSF_MEDIAN:
- case CTSF_MISSING:
+ if (s->writer)
+ {
+ struct casereader *reader = casewriter_make_reader (s->writer);
+ s->writer = NULL;
+
+ struct percentile *median = percentile_create (0.5, s->mvalid);
+ struct order_stats *os = &median->parent;
+ order_stats_accumulate_idx (&os, 1, reader, 1, 0);
+ s->median = percentile_calculate (median, PC_HAVERAGE);
+ statistic_destroy (&median->parent.parent);
+ }
+ return s->median;
+
case CTSF_MODE:
case CTSF_PTILE:
NOT_REACHED ();
cell = xmalloc (sizeof *cell);
cell->hide = false;
cell->sv = sv;
+ cell->contributes_to_domains = true;
for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
{
const struct ctables_nest *nest = &t->stacks[a].nests[ix[a]];
: NULL);
for (size_t i = 0; i < nest->n; i++)
{
+ const struct ctables_category *cat = cats[a][i];
+
if (i != nest->scale_idx)
{
- const struct ctables_category *subtotal = cats[a][i]->subtotal;
+ const struct ctables_category *subtotal = cat->subtotal;
if (subtotal && subtotal->type == CCT_HSUBTOTAL)
cell->hide = true;
+
+ if (cat->type == CCT_TOTAL || cat->type == CCT_SUBTOTAL || cat->type == CCT_HSUBTOTAL)
+ cell->contributes_to_domains = false;
}
- cell->axes[a].cvs[i].category = cats[a][i];
+ cell->axes[a].cvs[i].category = cat;
value_clone (&cell->axes[a].cvs[i].value, case_data (c, nest->vars[i]),
var_get_width (nest->vars[i]));
}
for (size_t i = 0; i < specs->n; i++)
ctables_summary_add (&cell->summaries[i], &specs->specs[i], specs->var,
case_data (c, specs->var), weight);
- for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++)
- cell->domains[dt]->valid += weight;
+ if (cell->contributes_to_domains)
+ for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++)
+ cell->domains[dt]->valid += weight;
}
static void
}
}
+static void
+recurse_subtotals (struct ctables_table *t, const struct ccase *c,
+ size_t ix[PIVOT_N_AXES],
+ const struct ctables_category *cats[PIVOT_N_AXES][10],
+ double weight,
+ enum pivot_axis_type start_axis, size_t start_nest)
+{
+ for (enum pivot_axis_type a = start_axis; a < PIVOT_N_AXES; a++)
+ {
+ const struct ctables_nest *nest = &t->stacks[a].nests[ix[a]];
+ for (size_t i = start_nest; i < nest->n; i++)
+ {
+ if (i == nest->scale_idx)
+ continue;
+
+ const struct ctables_category *save = cats[a][i];
+ if (save->subtotal)
+ {
+ cats[a][i] = save->subtotal;
+ ctables_cell_add__ (t, c, ix, cats, weight);
+ recurse_subtotals (t, c, ix, cats, weight, a, i + 1);
+ cats[a][i] = save;
+ }
+ }
+ start_nest = 0;
+ }
+}
+
static void
ctables_cell_insert (struct ctables_table *t,
const struct ccase *c,
ctables_cell_add__ (t, c, ix, cats, weight);
recurse_totals (t, c, ix, cats, weight, 0, 0);
-
- for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
- {
- const struct ctables_nest *nest = &t->stacks[a].nests[ix[a]];
- for (size_t i = 0; i < nest->n; i++)
- {
- if (i == nest->scale_idx)
- continue;
-
- const struct ctables_category *save = cats[a][i];
- if (save->subtotal)
- {
- cats[a][i] = save->subtotal;
- ctables_cell_add__ (t, c, ix, cats, weight);
- cats[a][i] = save;
- }
- }
- }
+ recurse_subtotals (t, c, ix, cats, weight, 0, 0);
}
struct merge_item
pivot_table_set_caption (
pt, pivot_value_new_user_text (t->corner, SIZE_MAX));
- bool summary_dimension = t->summary_axis != t->slabels_axis;
+ bool summary_dimension = (t->summary_axis != t->slabels_axis
+ || (!t->slabels_visible
+ && t->summary_specs.n > 1));
if (summary_dimension)
{
struct pivot_dimension *d = pivot_dimension_create (
- pt, t->slabels_axis, N_("Summaries"));
+ pt, t->slabels_axis, N_("Statistics"));
const struct ctables_summary_spec_set *specs = &t->summary_specs;
+ if (!t->slabels_visible)
+ d->hide_all_labels = true;
for (size_t i = 0; i < specs->n; i++)
pivot_category_create_leaf (
d->root, pivot_value_new_text (specs->specs[i].label));
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;
- }
+ const struct ctables_value *value = t->clabels_values[i];
+ const struct ctables_category *cat = ctables_categories_match (c, &value->value, var);
+ assert (cat != NULL);
pivot_category_create_leaf (d->root, ctables_category_create_label (
- cat, t->clabels_example, value));
+ cat, t->clabels_example, &value->value));
}
}
if (new_subtable)
{
n_levels = 0;
- printf ("%s levels:", pivot_axis_type_to_string (a));
for (size_t k = 0; k < nest->n; k++)
{
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,
}
}
- if (a == t->slabels_axis && a == t->summary_axis)
+ if (!summary_dimension && a == t->slabels_axis)
{
- printf (" summary");
levels[n_levels++] = (struct ctables_level) {
.type = CTL_SUMMARY,
.var_idx = SIZE_MAX,
};
}
- printf ("\n");
}
size_t n_common = 0;
if (level->type == CTL_CATEGORY)
{
size_t var_idx = level->var_idx;
- if (prev->axes[a].cvs[var_idx].category
- != cell->axes[a].cvs[var_idx].category)
- {
- 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;
- }
+ const struct ctables_category *c = cell->axes[a].cvs[var_idx].category;
+ if (prev->axes[a].cvs[var_idx].category != c)
+ break;
+ else if (c->type != CCT_SUBTOTAL
+ && c->type != CCT_HSUBTOTAL
+ && c->type != CCT_TOTAL
+ && !value_equal (&prev->axes[a].cvs[var_idx].value,
+ &cell->axes[a].cvs[var_idx].value,
+ var_get_type (nest->vars[var_idx])))
+ break;
}
}
}
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;
+ assert (ctv != NULL);
+ dindexes[n_dindexes++] = ctv->leaf;
}
for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
pivot_table_submit (pt);
}
+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;
-static void
+ 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;
+}
+
+static bool
ctables_prepare_table (struct ctables_table *t)
{
for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
}
}
#endif
+
+ return (ctables_check_label_position (t, PIVOT_AXIS_ROW)
+ && ctables_check_label_position (t, PIVOT_AXIS_COLUMN));
}
static void
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);
+ const struct variable *var = nest->vars[nest->n - 1];
+ int width = var_get_width (var);
+ const union value *value = case_data (c, var);
+
+ if (var_is_numeric (var) && value->f == SYSMIS)
+ continue;
+
+ if (!ctables_categories_match (t->categories [var_get_dict_index (var)],
+ value, var))
+ continue;
+
unsigned int hash = value_hash (value, width, 0);
struct ctables_value *clv = ctables_value_find__ (t, value, width, 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 struct ctables_value *const *ap = a_;
+ const struct ctables_value *const *bp = b_;
+ const struct ctables_value *a = *ap;
+ const struct ctables_value *b = *bp;
const int *width = width_;
- return value_compare_3way (a, b, *width);
+ return value_compare_3way (&a->value, &b->value, *width);
}
static void
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->clabels_values[i++] = clv;
t->n_clabels_values = n;
assert (i == n);
sort (t->clabels_values, n, sizeof *t->clabels_values,
compare_clabels_values_3way, &width);
+
+ for (size_t i = 0; i < n; i++)
+ t->clabels_values[i]->leaf = i;
}
static bool
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)
{
}
if (lex_token (lexer) == T_ENDCMD)
- break;
+ {
+ if (!ctables_prepare_table (t))
+ goto error;
+ break;
+ }
if (!lex_force_match (lexer, T_SLASH))
break;
goto error;
}
- ctables_prepare_table (t);
-
- ctables_check_label_position (t, PIVOT_AXIS_ROW);
- ctables_check_label_position (t, PIVOT_AXIS_COLUMN);
+ if (!ctables_prepare_table (t))
+ goto error;
}
while (lex_token (lexer) != T_ENDCMD);