#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"
};
};
-enum ctables_label_position
- {
- CTLP_NORMAL,
- CTLP_OPPOSITE,
- CTLP_LAYER,
- };
-
struct ctables_summary_spec_set
{
struct ctables_summary_spec *specs;
size_t n;
};
+struct ctables_value
+ {
+ struct hmap_node node;
+ union value value;
+ };
+
struct ctables_table
{
struct ctables_axis *axes[PIVOT_N_AXES];
struct hmap cells;
struct hmap domains[N_CTDTS];
+ const struct variable *clabels_example;
+ struct hmap clabels_values_map;
+ union value *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];
/* Indexed by variable dictionary index. */
struct ctables_categories **categories;
}
}
+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)
{
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
{
pivot_table_submit (pt);
}
-
static void
ctables_table_output_different_axis (struct ctables *ct, struct ctables_table *t)
{
struct ctables_cell *cell = sorted[j];
const struct ctables_nest *nest = &t->stacks[a].nests[cell->axes[a].stack_idx];
+ /* Pivot categories:
+
+ - variable label for nest->vars[0], if vlabel != CTVL_NONE
+ - category for nest->vars[0]
+ - variable label for nest->vars[1], if vlabel != CTVL_NONE
+ - category for nest->vars[1]
+ ...
+ - variable label for nest->vars[nest->n - 1], if vlabel != CTVL_NONE
+ - category for nest->vars[nest->n - 1], unless t->label_axis[a] != a.
+ - 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'.
+ */
+
size_t n_common = 0;
bool new_subtable = false;
if (j > 0)
#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;
+ HMAP_FOR_EACH_WITH_HASH (clv, struct ctables_value, node, hash,
+ &t->clabels_values_map)
+ if (value_equal (value, &clv->value, width))
+ goto next_stack;
+
+ clv = xmalloc (sizeof *clv);
+ value_clone (&clv->value, value, width);
+ hmap_insert (&t->clabels_values_map, &clv->node, hash);
+
+ next_stack: ;
+ }
+}
+
+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);
+
+ const struct ctables_value *clv;
+ size_t i = 0;
+ HMAP_FOR_EACH (clv, struct ctables_value, node, &t->clabels_values_map)
+ t->clabels_values[i++] = clv->value;
+ 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)
{
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);
for (size_t i = 0; i < ct->n_tables; i++)
{
struct ctables_table *t = ct->tables[i];
+
+ if (t->clabels_example)
+ ctables_sort_clabels_values (t);
+
if (t->summary_axis == t->slabels_axis)
ctables_table_output_same_axis (ct, ct->tables[i]);
else
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;
+
+ 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;
+ }
+
+ 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 (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)
{
.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,
+ },
.categories = categories,
.n_categories = n_vars,
.cilevel = 95,
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");
{
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");
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);