#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 "language/lexer/lexer.h"
#include "libpspp/hmap.h"
#include "libpspp/message.h"
#include "libpspp/string-array.h"
+#include "math/mode.h"
#include "math/moments.h"
+#include "math/percentiles.h"
+#include "math/sort.h"
#include "output/pivot-table.h"
#include "gl/minmax.h"
const struct ctables_cell *example;
- double valid;
- double missing;
+ double d_valid; /* Dictionary weight. */
+ double d_missing;
+ double e_valid; /* Effective weight */
+ double e_missing;
};
enum ctables_summary_variant
struct hmap_node node;
/* The domains that contain this cell. */
+ bool contributes_to_domains;
struct ctables_domain *domains[N_CTDTS];
bool hide;
struct
{
- size_t stack_idx;
+ size_t nest_idx;
struct ctables_cell_value
{
const struct ctables_category *category;
bool mrsets_count_duplicates; /* MRSETS. */
bool smissing_listwise; /* SMISSING. */
- struct variable *base_weight; /* WEIGHT. */
+ struct variable *e_weight; /* WEIGHT. */
int hide_threshold; /* HIDESMALLCOUNTS. */
struct ctables_table **tables;
};
};
-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;
+ int leaf;
+ };
+
struct ctables_table
{
struct ctables_axis *axes[PIVOT_N_AXES];
struct ctables_stack stacks[PIVOT_N_AXES];
enum pivot_axis_type summary_axis;
+ struct ctables_summary_spec_set summary_specs;
struct hmap cells;
struct hmap domains[N_CTDTS];
- enum pivot_axis_type slabels_position;
+ const struct variable *clabels_example;
+ struct hmap clabels_values_map;
+ struct ctables_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;
}
}
+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
{
struct msg_location *loc = lex_ofs_location (ctx->lexer, start_ofs,
lex_ofs (ctx->lexer) - 1);
- printf ("add %s\n", ctables_summary_function_name (function));
add_summary_spec (sub, function, percentile, label, formatp, loc, sv);
free (label);
msg_location_destroy (loc);
/* MEAN, SEMEAN, STDDEV, SUM, VARIANCE, *.SUM. */
struct moments1 *moments;
- /* XXX percentiles, median, mode, multiple response */
+ /* MEDIAN, MODE, PTILE. */
+ struct
+ {
+ struct casewriter *writer;
+ double ovalid;
+ double ovalue;
+ };
+
+ /* XXX 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:
case CTSF_MODE:
case CTSF_PTILE:
- NOT_REACHED ();
+ {
+ 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->ovalid = 0;
+ s->ovalue = SYSMIS;
+ }
+ break;
case CTSF_RESPONSES:
case CTSF_ROWPCT_RESPONSES:
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:
case CTSF_MODE:
case CTSF_PTILE:
- NOT_REACHED ();
+ casewriter_destroy (s->writer);
+ break;
case CTSF_RESPONSES:
case CTSF_ROWPCT_RESPONSES:
ctables_summary_add (union ctables_summary *s,
const struct ctables_summary_spec *ss,
const struct variable *var, const union value *value,
- double weight)
+ double d_weight, double e_weight)
{
switch (ss->function)
{
case CTSF_COUNT:
+ case CSTF_TOTALN:
+ case CTSF_VALIDN:
+ if (var_is_value_missing (var, value))
+ s->missing += d_weight;
+ else
+ s->valid += d_weight;
+ break;
+
case CTSF_ECOUNT:
case CTSF_ROWPCT_COUNT:
case CTSF_COLPCT_COUNT:
case CTSF_LAYERPCT_TOTALN:
case CTSF_LAYERROWPCT_TOTALN:
case CTSF_LAYERCOLPCT_TOTALN:
- case CSTF_TOTALN:
+ case CTSF_MISSING:
case CTSF_ETOTALN:
- case CTSF_VALIDN:
case CTSF_EVALIDN:
if (var_is_value_missing (var, value))
- s->missing += weight;
+ s->missing += e_weight;
else
- s->valid += weight;
+ s->valid += e_weight;
break;
case CTSF_MAXIMUM:
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, e_weight);
break;
case CTSF_MEDIAN:
- case CTSF_MISSING:
case CTSF_MODE:
case CTSF_PTILE:
- NOT_REACHED ();
+ if (var_is_value_missing (var, value))
+ {
+ s->ovalid += e_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) = e_weight;
+ casewriter_write (s->writer, c);
+ }
+ break;
case CTSF_RESPONSES:
case CTSF_ROWPCT_RESPONSES:
}
}
+static enum ctables_domain_type
+ctables_function_domain (enum ctables_summary_function function)
+{
+ switch (function)
+ {
+ case CTSF_COUNT:
+ case CTSF_ECOUNT:
+ case CTSF_MISSING:
+ case CSTF_TOTALN:
+ case CTSF_ETOTALN:
+ case CTSF_VALIDN:
+ case CTSF_EVALIDN:
+ case CTSF_MAXIMUM:
+ case CTSF_MINIMUM:
+ case CTSF_RANGE:
+ case CTSF_MEAN:
+ case CTSF_SEMEAN:
+ case CTSF_STDDEV:
+ case CTSF_SUM:
+ case CTSF_VARIANCE:
+ case CTSF_MEDIAN:
+ case CTSF_PTILE:
+ case CTSF_MODE:
+ case CTSF_RESPONSES:
+ NOT_REACHED ();
+
+ case CTSF_COLPCT_COUNT:
+ case CTSF_COLPCT_COUNT_RESPONSES:
+ case CTSF_COLPCT_RESPONSES:
+ case CTSF_COLPCT_RESPONSES_COUNT:
+ case CTSF_COLPCT_SUM:
+ case CTSF_COLPCT_TOTALN:
+ case CTSF_COLPCT_VALIDN:
+ return CTDT_COL;
+
+ case CTSF_LAYERCOLPCT_COUNT:
+ case CTSF_LAYERCOLPCT_COUNT_RESPONSES:
+ case CTSF_LAYERCOLPCT_RESPONSES:
+ case CTSF_LAYERCOLPCT_RESPONSES_COUNT:
+ case CTSF_LAYERCOLPCT_SUM:
+ case CTSF_LAYERCOLPCT_TOTALN:
+ case CTSF_LAYERCOLPCT_VALIDN:
+ return CTDT_LAYERCOL;
+
+ case CTSF_LAYERPCT_COUNT:
+ case CTSF_LAYERPCT_COUNT_RESPONSES:
+ case CTSF_LAYERPCT_RESPONSES:
+ case CTSF_LAYERPCT_RESPONSES_COUNT:
+ case CTSF_LAYERPCT_SUM:
+ case CTSF_LAYERPCT_TOTALN:
+ case CTSF_LAYERPCT_VALIDN:
+ return CTDT_LAYER;
+
+ case CTSF_LAYERROWPCT_COUNT:
+ case CTSF_LAYERROWPCT_COUNT_RESPONSES:
+ case CTSF_LAYERROWPCT_RESPONSES:
+ case CTSF_LAYERROWPCT_RESPONSES_COUNT:
+ case CTSF_LAYERROWPCT_SUM:
+ case CTSF_LAYERROWPCT_TOTALN:
+ case CTSF_LAYERROWPCT_VALIDN:
+ return CTDT_LAYERROW;
+
+ case CTSF_ROWPCT_COUNT:
+ case CTSF_ROWPCT_COUNT_RESPONSES:
+ case CTSF_ROWPCT_RESPONSES:
+ case CTSF_ROWPCT_RESPONSES_COUNT:
+ case CTSF_ROWPCT_SUM:
+ case CTSF_ROWPCT_TOTALN:
+ case CTSF_ROWPCT_VALIDN:
+ return CTDT_ROW;
+
+ case CTSF_SUBTABLEPCT_COUNT:
+ case CTSF_SUBTABLEPCT_COUNT_RESPONSES:
+ case CTSF_SUBTABLEPCT_RESPONSES:
+ case CTSF_SUBTABLEPCT_RESPONSES_COUNT:
+ case CTSF_SUBTABLEPCT_SUM:
+ case CTSF_SUBTABLEPCT_TOTALN:
+ case CTSF_SUBTABLEPCT_VALIDN:
+ return CTDT_SUBTABLE;
+
+ case CTSF_TABLEPCT_COUNT:
+ case CTSF_TABLEPCT_COUNT_RESPONSES:
+ case CTSF_TABLEPCT_RESPONSES:
+ case CTSF_TABLEPCT_RESPONSES_COUNT:
+ case CTSF_TABLEPCT_SUM:
+ case CTSF_TABLEPCT_TOTALN:
+ case CTSF_TABLEPCT_VALIDN:
+ return CTDT_TABLE;
+ }
+
+ NOT_REACHED ();
+}
+
static double
ctables_summary_value (const struct ctables_cell *cell,
union ctables_summary *s,
case CTSF_ECOUNT:
return s->valid;
- case CTSF_SUBTABLEPCT_COUNT:
- return cell->domains[CTDT_SUBTABLE]->valid ? s->valid / cell->domains[CTDT_SUBTABLE]->valid * 100 : SYSMIS;
-
case CTSF_ROWPCT_COUNT:
- return cell->domains[CTDT_ROW]->valid ? s->valid / cell->domains[CTDT_ROW]->valid * 100 : SYSMIS;
-
case CTSF_COLPCT_COUNT:
- return cell->domains[CTDT_COL]->valid ? s->valid / cell->domains[CTDT_COL]->valid * 100 : SYSMIS;
-
case CTSF_TABLEPCT_COUNT:
- return cell->domains[CTDT_TABLE]->valid ? s->valid / cell->domains[CTDT_TABLE]->valid * 100 : SYSMIS;
-
+ case CTSF_SUBTABLEPCT_COUNT:
case CTSF_LAYERPCT_COUNT:
- return cell->domains[CTDT_LAYER]->valid ? s->valid / cell->domains[CTDT_LAYER]->valid * 100 : SYSMIS;
-
case CTSF_LAYERROWPCT_COUNT:
- return cell->domains[CTDT_LAYERROW]->valid ? s->valid / cell->domains[CTDT_LAYERROW]->valid * 100 : SYSMIS;
-
case CTSF_LAYERCOLPCT_COUNT:
- return cell->domains[CTDT_LAYERCOL]->valid ? s->valid / cell->domains[CTDT_LAYERCOL]->valid * 100 : SYSMIS;
+ {
+ enum ctables_domain_type d = ctables_function_domain (ss->function);
+ return (cell->domains[d]->e_valid
+ ? s->valid / cell->domains[d]->e_valid * 100
+ : SYSMIS);
+ }
case CTSF_ROWPCT_VALIDN:
case CTSF_COLPCT_VALIDN:
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:
- case CTSF_MODE:
case CTSF_PTILE:
- NOT_REACHED ();
+ if (s->writer)
+ {
+ struct casereader *reader = casewriter_make_reader (s->writer);
+ s->writer = NULL;
+
+ struct percentile *ptile = percentile_create (
+ ss->function == CTSF_PTILE ? ss->percentile : 0.5, s->ovalid);
+ struct order_stats *os = &ptile->parent;
+ order_stats_accumulate_idx (&os, 1, reader, 1, 0);
+ s->ovalue = percentile_calculate (ptile, PC_HAVERAGE);
+ statistic_destroy (&ptile->parent.parent);
+ }
+ return s->ovalue;
+
+ case CTSF_MODE:
+ if (s->writer)
+ {
+ struct casereader *reader = casewriter_make_reader (s->writer);
+ s->writer = NULL;
+
+ struct mode *mode = mode_create ();
+ struct order_stats *os = &mode->parent;
+ order_stats_accumulate_idx (&os, 1, reader, 1, 0);
+ s->ovalue = mode->mode;
+ statistic_destroy (&mode->parent.parent);
+ }
+ return s->ovalue;
case CTSF_RESPONSES:
case CTSF_ROWPCT_RESPONSES:
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;
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++)
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];
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
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]];
- 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);
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]));
}
ctables_cell_add__ (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)
+ double d_weight, double e_weight)
{
struct ctables_cell *cell = ctables_cell_insert__ (t, c, ix, cats);
const struct ctables_nest *ss = &t->stacks[t->summary_axis].nests[ix[t->summary_axis]];
const struct ctables_summary_spec_set *specs = &ss->specs[cell->sv];
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;
+ case_data (c, specs->var), d_weight, e_weight);
+ if (cell->contributes_to_domains)
+ {
+ for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++)
+ {
+ cell->domains[dt]->d_valid += d_weight;
+ cell->domains[dt]->e_valid += e_weight;
+ }
+ }
}
static void
recurse_totals (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,
+ double d_weight, double e_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_category *save = cats[a][i];
cats[a][i] = total;
- ctables_cell_add__ (t, c, ix, cats, weight);
- recurse_totals (t, c, ix, cats, weight, a, i + 1);
+ ctables_cell_add__ (t, c, ix, cats, d_weight, e_weight);
+ recurse_totals (t, c, ix, cats, d_weight, e_weight, a, i + 1);
+ cats[a][i] = save;
+ }
+ }
+ start_nest = 0;
+ }
+}
+
+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 d_weight, double e_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, d_weight, e_weight);
+ recurse_subtotals (t, c, ix, cats, d_weight, e_weight, a, i + 1);
cats[a][i] = save;
}
}
ctables_cell_insert (struct ctables_table *t,
const struct ccase *c,
size_t ir, size_t ic, size_t il,
- double weight)
+ double d_weight, double e_weight)
{
size_t ix[PIVOT_N_AXES] = {
[PIVOT_AXIS_ROW] = ir,
}
}
- ctables_cell_add__ (t, c, ix, cats, weight);
-
- recurse_totals (t, c, ix, cats, weight, 0, 0);
+ ctables_cell_add__ (t, c, ix, cats, d_weight, e_weight);
- 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_totals (t, c, ix, cats, d_weight, e_weight, 0, 0);
+ recurse_subtotals (t, c, ix, cats, d_weight, e_weight, 0, 0);
}
struct merge_item
{
- size_t tiebreaker;
const struct ctables_summary_spec_set *set;
size_t ofs;
};
return strcmp (as->label, bs->label);
}
-static bool
-ctables_execute (struct dataset *ds, struct ctables *ct)
+static struct pivot_value *
+ctables_category_create_label (const struct ctables_category *cat,
+ const struct variable *var,
+ const union value *value)
{
- 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])
- {
- t->stacks[a] = enumerate_fts (a, t->axes[a]);
+ 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));
+}
- 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++)
- {
- nest->domains[dt] = xmalloc (nest->n * sizeof *nest->domains[dt]);
- nest->n_domains[dt] = 0;
+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;
+}
- for (size_t k = 0; k < nest->n; k++)
- {
- if (k == nest->scale_idx)
- continue;
+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));
+}
- switch (dt)
- {
- case CTDT_TABLE:
- continue;
+static void
+ctables_table_output (struct ctables *ct, 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));
+
+ 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_("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));
+ }
- 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)
- continue;
- break;
+ 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 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->value));
+ }
+ }
- case CTDT_LAYERCOL:
- if (a == PIVOT_AXIS_ROW)
- continue;
- break;
- }
+ 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;
- 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 };
- }
+ assert (t->axes[a]);
- struct ctables_stack *stack = &t->stacks[t->summary_axis];
- for (size_t i = 0; i < stack->n; i++)
+ struct ctables_cell **sorted = xnmalloc (t->cells.count, sizeof *sorted);
+ size_t n_sorted = 0;
+
+ struct ctables_cell *cell;
+ HMAP_FOR_EACH (cell, struct ctables_cell, node, &t->cells)
+ if (!cell->hide)
+ sorted[n_sorted++] = cell;
+ assert (n_sorted <= t->cells.count);
+
+ struct ctables_cell_sort_aux aux = { .t = t, .a = a };
+ 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;
+
+ /* 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
{
- struct ctables_nest *nest = &stack->nests[i];
- if (!nest->specs[CSV_CELL].n)
+ enum ctables_level_type
{
- 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]);
+ CTL_VAR, /* Variable label for nest->vars[var_idx]. */
+ CTL_CATEGORY, /* Category for nest->vars[var_idx]. */
+ CTL_SUMMARY, /* Summary functions. */
}
- else if (!nest->specs[CSV_TOTAL].n)
- ctables_summary_spec_set_clone (&nest->specs[CSV_TOTAL],
- &nest->specs[CSV_CELL]);
- }
+ type;
- struct ctables_summary_spec_set merged = { .n = 0 };
- struct merge_item *items = xnmalloc (2 * stack->n, sizeof *items);
- size_t n_left = 0;
- for (size_t j = 0; j < stack->n; j++)
+ 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_sorted; j++)
{
- const struct ctables_nest *nest = &stack->nests[j];
- if (!nest->n)
- continue;
+ struct ctables_cell *cell = sorted[j];
+ struct ctables_cell *prev = j > 0 ? sorted[j - 1] : NULL;
+ const struct ctables_nest *nest = &t->stacks[a].nests[cell->axes[a].nest_idx];
- for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++)
+ bool new_subtable = !prev || prev->axes[a].nest_idx != cell->axes[a].nest_idx;
+ if (new_subtable)
{
- items[n_left] = (struct merge_item) {
- .tiebreaker = n_left,
- .set = &nest->specs[sv]
- };
- n_left++;
- }
- }
+ n_levels = 0;
+ 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)
+ {
+ levels[n_levels++] = (struct ctables_level) {
+ .type = CTL_VAR,
+ .var_idx = k,
+ };
+ }
- 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 (nest->scale_idx != k
+ && (k != nest->n - 1 || t->label_axis[a] == a))
+ {
+ levels[n_levels++] = (struct ctables_level) {
+ .type = CTL_CATEGORY,
+ .var_idx = k,
+ };
+ }
+ }
- /* XXX Add to 'merged' */
- 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 (!summary_dimension && a == t->slabels_axis)
+ {
+ levels[n_levels++] = (struct ctables_level) {
+ .type = CTL_SUMMARY,
+ .var_idx = SIZE_MAX,
+ };
+ }
+ }
- for (size_t j = 0; j < n_left; )
+ size_t n_common = 0;
+ if (!new_subtable)
{
- if (merge_item_compare_3way (&items[j], &min) == 0)
+ for (; n_common < n_levels; n_common++)
{
- struct merge_item *item = &items[j];
- item->set->specs[item->ofs].axis_idx = merged.n - 1;
- if (++item->ofs >= item->set->n)
+ const struct ctables_level *level = &levels[n_common];
+ if (level->type == CTL_CATEGORY)
{
- items[j] = items[--n_left];
- continue;
+ size_t var_idx = level->var_idx;
+ 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;
}
}
- j++;
}
- }
- for (size_t j = 0; j < merged.n; j++)
- printf ("%s\n", ctables_summary_function_name (merged.specs[j].function));
+ 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;
+ }
+ }
+ 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 (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;
+ }
+ free (sorted);
+ free (groups);
}
- struct casereader *input = casereader_create_filter_weight (proc_open (ds),
- dataset_dict (ds),
- NULL, NULL);
- bool warn_on_invalid = true;
- double total_weight = 0;
- for (struct ccase *c = casereader_read (input); c;
- case_unref (c), c = casereader_read (input))
+ struct ctables_cell *cell;
+ HMAP_FOR_EACH (cell, struct ctables_cell, node, &t->cells)
{
- double weight = dict_get_case_weight (dataset_dict (ds), c,
- &warn_on_invalid);
- total_weight += weight;
+ if (cell->hide)
+ continue;
- for (size_t i = 0; i < ct->n_tables; i++)
+ 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++)
{
- struct ctables_table *t = ct->tables[i];
+ size_t dindexes[5];
+ size_t n_dindexes = 0;
- for (size_t ir = 0; ir < t->stacks[PIVOT_AXIS_ROW].n; ir++)
- 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);
+ 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));
+ assert (ctv != NULL);
+ dindexes[n_dindexes++] = ctv->leaf;
+ }
+
+ 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 && !summary_dimension)
+ leaf += j;
+ dindexes[n_dindexes++] = leaf;
+ }
+
+ double d = ctables_summary_value (cell, &cell->summaries[j], &specs->specs[j]);
+ struct pivot_value *value = pivot_value_new_number (d);
+ value->numeric.format = specs->specs[j].format;
+ pivot_table_put (pt, dindexes, n_dindexes, value);
}
}
- casereader_destroy (input);
- for (size_t i = 0; i < ct->n_tables; i++)
+ 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;
+
+ 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)
{
- struct ctables_table *t = ct->tables[i];
+ 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;
+ }
- 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++)
+ 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)
{
- 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;
+ 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++)
+ if (t->axes[a])
+ {
+ t->stacks[a] = enumerate_fts (a, t->axes[a]);
- assert (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++)
+ {
+ nest->domains[dt] = xmalloc (nest->n * sizeof *nest->domains[dt]);
+ nest->n_domains[dt] = 0;
- struct ctables_cell **sorted = xnmalloc (t->cells.count, sizeof *sorted);
+ for (size_t k = 0; k < nest->n; k++)
+ {
+ if (k == nest->scale_idx)
+ continue;
- 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);
+ switch (dt)
+ {
+ case CTDT_TABLE:
+ continue;
- struct ctables_cell_sort_aux aux = { .t = t, .a = a };
- sort (sorted, n, sizeof *sorted, ctables_cell_compare_3way, &aux);
+ 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;
- 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;
+ case CTDT_LAYERROW:
+ if (a == PIVOT_AXIS_COLUMN)
+ continue;
+ break;
- 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];
+ case CTDT_LAYERCOL:
+ if (a == PIVOT_AXIS_ROW)
+ continue;
+ break;
+ }
- 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;
+ 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 };
+ }
- 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;
- }
+ 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_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];
- 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->summary_axis)
- {
- 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++)
- {
- int leaf = pivot_category_create_leaf (
- parent, pivot_value_new_text (specs->specs[m].label));
- if (m == 0)
- prev_leaf = leaf;
- }
- }
- else
- {
- /* This assertion is true as long as the summary axis
- is the axis where the summaries are displayed. */
- assert (label);
+ 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]);
+ }
- prev_leaf = pivot_category_create_leaf (parent, label);
- }
- break;
- }
+ 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] };
+ }
- if (label)
- parent = pivot_category_create_group__ (parent, label);
+ 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];
- 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 (merged->n >= merged->allocated)
+ merged->specs = x2nrealloc (merged->specs, &merged->allocated,
+ sizeof *merged->specs);
+ merged->specs[merged->n++] = min.set->specs[min.ofs];
- cell->axes[a].leaf = prev_leaf;
+ for (size_t j = 0; j < n_left; )
+ {
+ 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)
+ {
+ items[j] = items[--n_left];
+ continue;
+ }
}
- free (sorted);
- free (groups);
+ j++;
}
- struct ctables_cell *cell;
- HMAP_FOR_EACH (cell, struct ctables_cell, node, &t->cells)
+ }
+
+#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 < stack->n; j++)
+ {
+ const struct ctables_nest *nest = &stack->nests[j];
+ for (enum ctables_summary_variant sv = 0; sv < N_CSVS; sv++)
{
- if (cell->hide)
- continue;
+ 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
- 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];
- for (size_t j = 0; j < specs->n; j++)
- {
- size_t dindexes[3];
- size_t n_dindexes = 0;
+ return (ctables_check_label_position (t, PIVOT_AXIS_ROW)
+ && ctables_check_label_position (t, PIVOT_AXIS_COLUMN));
+}
- 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)
- leaf += j;
- dindexes[n_dindexes++] = leaf;
- }
+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 *var = nest->vars[nest->n - 1];
+ int width = var_get_width (var);
+ const union value *value = case_data (c, var);
- double d = ctables_summary_value (cell, &cell->summaries[j], &specs->specs[j]);
- struct pivot_value *value = pivot_value_new_number (d);
- value->numeric.format = specs->specs[j].format;
- pivot_table_put (pt, dindexes, n_dindexes, value);
- }
+ 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);
+ 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 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->value, &b->value, *width);
+}
+
+static void
+ctables_sort_clabels_values (struct ctables_table *t)
+{
+ int width = var_get_width (t->clabels_example);
- pivot_table_submit (pt);
+ 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)
+ 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
+ctables_execute (struct dataset *ds, struct ctables *ct)
+{
+ struct casereader *input = proc_open (ds);
+ bool warn_on_invalid = true;
+ for (struct ccase *c = casereader_read (input); c;
+ case_unref (c), c = casereader_read (input))
+ {
+ double d_weight = dict_get_case_weight (dataset_dict (ds), c,
+ &warn_on_invalid);
+ double e_weight = (ct->e_weight
+ ? var_force_valid_weight (ct->e_weight,
+ case_num (c, ct->e_weight),
+ &warn_on_invalid)
+ : d_weight);
+
+ for (size_t i = 0; i < ct->n_tables; i++)
+ {
+ struct ctables_table *t = ct->tables[i];
+
+ for (size_t ir = 0; ir < t->stacks[PIVOT_AXIS_ROW].n; ir++)
+ 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, d_weight, e_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);
+ ctables_table_output (ct, ct->tables[i]);
+ }
return proc_commit (ds);
}
if (!lex_force_match_id (lexer, "VARIABLE"))
goto error;
lex_match (lexer, T_EQUALS);
- ct->base_weight = parse_variable (lexer, dataset_dict (ds));
- if (!ct->base_weight)
+ ct->e_weight = parse_variable (lexer, dataset_dict (ds));
+ if (!ct->e_weight)
goto error;
}
else if (lex_match_id (lexer, "HIDESMALLCOUNTS"))
struct ctables_table *t = xmalloc (sizeof *t);
*t = (struct ctables_table) {
.cells = HMAP_INITIALIZER (t->cells),
- .slabels_position = PIVOT_AXIS_COLUMN,
+ .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,
}
if (lex_token (lexer) == T_ENDCMD)
- break;
+ {
+ if (!ctables_prepare_table (t))
+ goto error;
+ break;
+ }
if (!lex_force_match (lexer, T_SLASH))
break;
{
if (lex_match_id (lexer, "SLABELS"))
{
- while (lex_token (lexer) != T_SLASH)
+ while (lex_token (lexer) != T_SLASH && lex_token (lexer) != T_ENDCMD)
{
if (lex_match_id (lexer, "POSITION"))
{
lex_match (lexer, T_EQUALS);
if (lex_match_id (lexer, "COLUMN"))
- t->slabels_position = PIVOT_AXIS_COLUMN;
+ t->slabels_axis = PIVOT_AXIS_COLUMN;
else if (lex_match_id (lexer, "ROW"))
- t->slabels_position = PIVOT_AXIS_ROW;
+ t->slabels_axis = PIVOT_AXIS_ROW;
else if (lex_match_id (lexer, "LAYER"))
- t->slabels_position = PIVOT_AXIS_LAYER;
+ t->slabels_axis = PIVOT_AXIS_LAYER;
else
{
lex_error_expecting (lexer, "COLUMN", "ROW", "LAYER");
}
else if (lex_match_id (lexer, "CLABELS"))
{
- while (lex_token (lexer) != T_SLASH)
+ 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;
}
+ if (!ctables_prepare_table (t))
+ goto error;
}
while (lex_token (lexer) != T_ENDCMD);