#include "language/lexer/variable-parser.h"
#include "libpspp/array.h"
#include "libpspp/assertion.h"
+#include "libpspp/hash-functions.h"
#include "libpspp/hmap.h"
#include "libpspp/message.h"
#include "libpspp/string-array.h"
#undef S
};
+enum ctables_domain_type
+ {
+ /* Within a section, where stacked variables divide one section from
+ another. */
+ CTDT_TABLE, /* All layers of a whole section. */
+ CTDT_LAYER, /* One layer within a section. */
+ CTDT_LAYERROW, /* Row in one layer within a section. */
+ CTDT_LAYERCOL, /* Column in one layer within a section. */
+
+ /* Within a subtable, where a subtable pairs an innermost row variable with
+ an innermost column variable within a single layer. */
+ CTDT_SUBTABLE, /* Whole subtable. */
+ CTDT_ROW, /* Row within a subtable. */
+ CTDT_COL, /* Column within a subtable. */
+#define N_CTDTS 7
+ };
+
+struct ctables_domain
+ {
+ struct hmap_node node;
+
+ const struct ctables_freq *example;
+
+ double valid;
+ double missing;
+ };
+
+struct ctables_freq
+ {
+ /* In struct ctables's 'ft' hmap. Indexed by all the values in all the
+ axes (except the scalar variable, if any). */
+ struct hmap_node node;
+
+ /* The domains that contains this cell. */
+ struct ctables_domain *domains[N_CTDTS];
+
+ struct
+ {
+ size_t vaa_idx;
+ union value *values;
+ int leaf;
+ }
+ axes[PIVOT_N_AXES];
+
+ union ctables_summary *summaries;
+ };
+
struct ctables
{
struct pivot_table_look *look;
struct variable *base_weight; /* WEIGHT. */
int hide_threshold; /* HIDESMALLCOUNTS. */
- struct ctables_table *tables;
+ struct ctables_table **tables;
size_t n_tables;
};
CTLP_LAYER,
};
+struct var_array
+ {
+ struct variable **vars;
+ size_t n;
+ size_t scale_idx;
+ size_t *domains[N_CTDTS];
+ size_t n_domains[N_CTDTS];
+
+ struct ctables_summary_spec *summaries;
+ size_t n_summaries;
+ struct variable *summary_var;
+ };
+
+struct var_array2
+ {
+ struct var_array *vas;
+ size_t n;
+ };
+
struct ctables_table
{
struct ctables_axis *axes[PIVOT_N_AXES];
+ struct var_array2 vaas[PIVOT_N_AXES];
+ enum pivot_axis_type summary_axis;
+ struct hmap ft;
+ struct hmap domains[N_CTDTS];
enum pivot_axis_type slabels_position;
bool slabels_visible;
struct ctables_chisq *chisq;
struct ctables_pairwise *pairwise;
-
- struct ctables_freqtab **fts;
- size_t n_fts;
};
struct ctables_var
}
}
+static char *
+ctables_summary_default_label (enum ctables_summary_function function,
+ double percentile)
+{
+ static const char *default_labels[] = {
+#define S(ENUM, NAME, LABEL, FORMAT, AVAILABILITY) [ENUM] = LABEL,
+ SUMMARIES
+#undef S
+ };
+
+ return (function == CTSF_PTILE
+ ? xasprintf (_("Percentile %.2f"), percentile)
+ : xstrdup (gettext (default_labels[function])));
+}
+
static const char *
ctables_summary_function_name (enum ctables_summary_function function)
{
label = ss_xstrdup (lex_tokss (ctx->lexer));
lex_get (ctx->lexer);
}
- else if (function == CTSF_PTILE)
- label = xasprintf (_("Percentile %.2f"), percentile);
else
- {
- static const char *default_labels[] = {
-#define S(ENUM, NAME, LABEL, FORMAT, AVAILABILITY) [ENUM] = LABEL,
- SUMMARIES
-#undef S
- };
- label = xstrdup (gettext (default_labels[function]));
- }
+ label = ctables_summary_default_label (function, percentile);
/* Parse format. */
struct fmt_spec format;
}
static void
-ctables_table_uninit (struct ctables_table *t)
+ctables_table_destroy (struct ctables_table *t)
{
if (!t)
return;
free (t->title);
ctables_chisq_destroy (t->chisq);
ctables_pairwise_destroy (t->pairwise);
+ free (t);
}
static void
free (ct->missing);
free (ct->vlabels);
for (size_t i = 0; i < ct->n_tables; i++)
- ctables_table_uninit (&ct->tables[i]);
+ ctables_table_destroy (ct->tables[i]);
free (ct->tables);
free (ct);
}
return true;
}
-struct var_array
- {
- const struct ctables_axis *summary;
- struct variable **vars;
- size_t n;
- };
-
static void
var_array_uninit (struct var_array *va)
{
free (va->vars);
}
-struct var_array2
- {
- struct var_array *vas;
- size_t n;
- };
-
static void
var_array2_uninit (struct var_array2 *vaa)
{
size_t allocate = a->n + b->n;
struct variable **vars = xnmalloc (allocate, sizeof *vars);
+ enum pivot_axis_type *axes = xnmalloc (allocate, sizeof *axes);
size_t n = 0;
for (size_t k = 0; k < a->n; k++)
vars[n++] = a->vars[k];
vars[n++] = b->vars[k];
assert (n == allocate);
- assert (!(a->summary && b->summary));
+ const struct var_array *summary_src;
+ if (!a->summary_var)
+ summary_src = b;
+ else if (!b->summary_var)
+ summary_src = a;
+ else
+ NOT_REACHED ();
vaa.vas[vaa.n++] = (struct var_array) {
- .summary = a->summary ? a->summary : b->summary,
.vars = vars,
- .n = n
+ .scale_idx = (a->scale_idx != SIZE_MAX ? a->scale_idx
+ : b->scale_idx != SIZE_MAX ? a->n + b->scale_idx
+ : SIZE_MAX),
+ .n = n,
+ .summaries = summary_src->summaries,
+ .n_summaries = summary_src->n_summaries,
+ .summary_var = summary_src->summary_var,
};
}
var_array2_uninit (&va0);
}
static struct var_array2
-enumerate_fts (const struct ctables_axis *a)
+enumerate_fts (enum pivot_axis_type axis_type, const struct ctables_axis *a)
{
if (!a)
return (struct var_array2) { .n = 0 };
{
case CTAO_VAR:
assert (!a->var.is_mrset);
+
+ struct variable **vars = xmalloc (sizeof *vars);
+ *vars = a->var.var;
+
struct var_array *va = xmalloc (sizeof *va);
- if (a->scale)
- *va = (struct var_array) { .n = 0 };
- else
+ *va = (struct var_array) {
+ .vars = vars,
+ .n = 1,
+ .scale_idx = a->scale ? 0 : SIZE_MAX,
+ };
+ if (a->n_summaries || a->scale)
{
- struct variable **v = xmalloc (sizeof *v);
- *v = a->var.var;
- *va = (struct var_array) { .vars = v, .n = 1 };
+ va->summaries = a->summaries;
+ va->n_summaries = a->n_summaries;
+ va->summary_var = a->var.var;
}
- va->summary = a->scale || a->n_summaries ? a : NULL;
return (struct var_array2) { .vas = va, .n = 1 };
case CTAO_STACK:
- return stack_fts (enumerate_fts (a->subs[0]),
- enumerate_fts (a->subs[1]));
+ return stack_fts (enumerate_fts (axis_type, a->subs[0]),
+ enumerate_fts (axis_type, a->subs[1]));
case CTAO_NEST:
- return nest_fts (enumerate_fts (a->subs[0]),
- enumerate_fts (a->subs[1]));
+ return nest_fts (enumerate_fts (axis_type, a->subs[0]),
+ enumerate_fts (axis_type, a->subs[1]));
}
NOT_REACHED ();
}
}
-static void
+static void UNUSED
ctables_summary_uninit (union ctables_summary *s,
const struct ctables_summary_spec *ss)
{
}
}
-
static double
-ctables_summary_value (union ctables_summary *s,
+ctables_summary_value (const struct ctables_freq *f,
+ union ctables_summary *s,
const struct ctables_summary_spec *ss)
{
switch (ss->function)
case CTSF_ECOUNT:
return s->valid;
+ case CTSF_SUBTABLEPCT_COUNT:
+ return f->domains[CTDT_SUBTABLE]->valid ? s->valid / f->domains[CTDT_SUBTABLE]->valid * 100 : SYSMIS;
+
case CTSF_ROWPCT_COUNT:
+ return f->domains[CTDT_ROW]->valid ? s->valid / f->domains[CTDT_ROW]->valid * 100 : SYSMIS;
+
case CTSF_COLPCT_COUNT:
+ return f->domains[CTDT_COL]->valid ? s->valid / f->domains[CTDT_COL]->valid * 100 : SYSMIS;
+
case CTSF_TABLEPCT_COUNT:
- case CTSF_SUBTABLEPCT_COUNT:
+ return f->domains[CTDT_TABLE]->valid ? s->valid / f->domains[CTDT_TABLE]->valid * 100 : SYSMIS;
+
case CTSF_LAYERPCT_COUNT:
+ return f->domains[CTDT_LAYER]->valid ? s->valid / f->domains[CTDT_LAYER]->valid * 100 : SYSMIS;
+
case CTSF_LAYERROWPCT_COUNT:
+ return f->domains[CTDT_LAYERROW]->valid ? s->valid / f->domains[CTDT_LAYERROW]->valid * 100 : SYSMIS;
+
case CTSF_LAYERCOLPCT_COUNT:
+ return f->domains[CTDT_LAYERCOL]->valid ? s->valid / f->domains[CTDT_LAYERCOL]->valid * 100 : SYSMIS;
+
case CTSF_ROWPCT_VALIDN:
case CTSF_COLPCT_VALIDN:
case CTSF_TABLEPCT_VALIDN:
NOT_REACHED ();
}
-struct ctables_freq
+struct ctables_freq_sort_aux
{
- struct hmap_node node; /* Element in hash table. */
- union ctables_summary *summaries;
- union value values[]; /* The value. */
+ const struct ctables_table *t;
+ enum pivot_axis_type a;
};
-struct ctables_freqtab
- {
- struct var_array vars;
- struct hmap data; /* Contains "struct ctables_freq"s. */
- const struct ctables_summary_spec *summaries;
- size_t n_summaries;
- const struct variable *summary_var;
- struct ctables_freq **sorted;
- };
+static int
+ctables_freq_compare_3way (const void *a_, const void *b_, const void *aux_)
+{
+ const struct ctables_freq_sort_aux *aux = aux_;
+ struct ctables_freq *const *ap = a_;
+ struct ctables_freq *const *bp = b_;
+ const struct ctables_freq *a = *ap;
+ const struct ctables_freq *b = *bp;
+
+ size_t a_idx = a->axes[aux->a].vaa_idx;
+ size_t b_idx = b->axes[aux->a].vaa_idx;
+ if (a_idx != b_idx)
+ return a_idx < b_idx ? -1 : 1;
+
+ const struct var_array *va = &aux->t->vaas[aux->a].vas[a_idx];
+ for (size_t i = 0; i < va->n; i++)
+ if (i != va->scale_idx)
+ {
+ int cmp = value_compare_3way (&a->axes[aux->a].values[i],
+ &b->axes[aux->a].values[i],
+ var_get_width (va->vars[i]));
+ if (cmp)
+ return cmp;
+ }
+ return 0;
+}
-static struct ctables_freq *
-ctables_freq_create (struct ctables_freqtab *ft)
+/* Algorithm:
+
+ For each row:
+ For each ctables_table:
+ For each combination of row vars:
+ For each combination of column vars:
+ For each combination of layer vars:
+ Add entry
+ Make a table of row values:
+ Sort entries by row values
+ Assign a 0-based index to each actual value
+ Construct a dimension
+ Make a table of column values
+ Make a table of layer values
+ For each entry:
+ Fill the table entry using the indexes from before.
+ */
+
+static struct ctables_domain *
+ctables_domain_insert (struct ctables_table *t, struct ctables_freq *f,
+ enum ctables_domain_type domain)
{
- struct ctables_freq *f = xmalloc (sizeof *f + ft->vars.n * sizeof *f->values);
- f->summaries = xmalloc (ft->n_summaries * sizeof *f->summaries);
- for (size_t i = 0; i < ft->n_summaries; i++)
- ctables_summary_init (&f->summaries[i], &ft->summaries[i]);
- return f;
+ size_t hash = 0;
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ {
+ size_t idx = f->axes[a].vaa_idx;
+ const struct var_array *va = &t->vaas[a].vas[idx];
+ hash = hash_int (idx, hash);
+ for (size_t i = 0; i < va->n_domains[domain]; i++)
+ {
+ size_t v_idx = va->domains[domain][i];
+ hash = value_hash (&f->axes[a].values[v_idx],
+ var_get_width (va->vars[v_idx]), hash);
+ }
+ }
+
+ struct ctables_domain *d;
+ HMAP_FOR_EACH_WITH_HASH (d, struct ctables_domain, node, hash, &t->domains[domain])
+ {
+ const struct ctables_freq *df = d->example;
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ {
+ size_t idx = f->axes[a].vaa_idx;
+ if (idx != df->axes[a].vaa_idx)
+ goto not_equal;
+
+ const struct var_array *va = &t->vaas[a].vas[idx];
+ for (size_t i = 0; i < va->n_domains[domain]; i++)
+ {
+ size_t v_idx = va->domains[domain][i];
+ if (!value_equal (&df->axes[a].values[v_idx],
+ &f->axes[a].values[v_idx],
+ var_get_width (va->vars[v_idx])))
+ goto not_equal;
+ }
+ }
+ return d;
+
+ not_equal: ;
+ }
+
+ d = xmalloc (sizeof *d);
+ *d = (struct ctables_domain) { .example = f };
+ hmap_insert (&t->domains[domain], &d->node, hash);
+ return d;
}
-static void
-ctables_freq_add (struct ctables_freqtab *ft, struct ctables_freq *f,
- const struct variable *var, const union value *value,
- double weight)
+static const struct ctables_cat_value *
+ctables_categories_match (const struct ctables_categories *cats,
+ const union value *v, const struct variable *var)
{
- for (size_t i = 0; i < ft->n_summaries; i++)
- ctables_summary_add (&f->summaries[i], &ft->summaries[i],
- var, value, weight);
+ const struct ctables_cat_value *othernm = NULL;
+ for (size_t i = cats->n_values; i-- > 0; )
+ {
+ const struct ctables_cat_value *cv = &cats->values[i];
+ switch (cv->type)
+ {
+ case CCVT_NUMBER:
+ if (cv->number == v->f)
+ return cv;
+ break;
+
+ case CCVT_STRING:
+ NOT_REACHED ();
+
+ case CCVT_RANGE:
+ if ((cv->range[0] == -DBL_MAX || v->f >= cv->range[0])
+ && (cv->range[1] == DBL_MAX || v->f <= cv->range[1]))
+ return cv;
+ break;
+
+ case CCVT_MISSING:
+ if (var_is_value_missing (var, v))
+ return cv;
+ break;
+
+ case CCVT_OTHERNM:
+ if (!othernm)
+ othernm = cv;
+ break;
+
+ case CCVT_SUBTOTAL:
+ case CCVT_HSUBTOTAL:
+ break;
+ }
+ }
+
+ return var_is_value_missing (var, v) ? NULL : othernm;
}
-static int
-ctables_freq_compare_3way (const void *a_, const void *b_, const void *vars_)
+static void
+ctables_freqtab_insert (struct ctables_table *t,
+ const struct ccase *c,
+ size_t ir, size_t ic, size_t il,
+ double weight)
{
- const struct var_array *vars = vars_;
- struct ctables_freq *const *a = a_;
- struct ctables_freq *const *b = b_;
+ size_t ix[PIVOT_N_AXES] = {
+ [PIVOT_AXIS_ROW] = ir,
+ [PIVOT_AXIS_COLUMN] = ic,
+ [PIVOT_AXIS_LAYER] = il,
+ };
+ const struct var_array *ss = &t->vaas[t->summary_axis].vas[ix[t->summary_axis]];
- for (size_t i = 0; i < vars->n; i++)
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
{
- int cmp = value_compare_3way (&(*a)->values[i], &(*b)->values[i],
- var_get_width (vars->vars[i]));
- if (cmp)
- return cmp;
+ const struct var_array *va = &t->vaas[a].vas[ix[a]];
+ for (size_t i = 0; i < va->n; i++)
+ {
+ if (i == va->scale_idx)
+ continue;
+
+ const struct ctables_categories *cats = t->categories[var_get_dict_index (va->vars[i])];
+ if (!cats || !cats->n_values)
+ continue;
+
+ if (!ctables_categories_match (cats, case_data (c, va->vars[i]), va->vars[i]))
+ return;
+ }
}
- return 0;
+
+ size_t hash = 0;
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ {
+ const struct var_array *va = &t->vaas[a].vas[ix[a]];
+ hash = hash_int (ix[a], hash);
+ for (size_t i = 0; i < va->n; i++)
+ if (i != va->scale_idx)
+ hash = value_hash (case_data (c, va->vars[i]),
+ var_get_width (va->vars[i]), hash);
+ }
+
+ struct ctables_freq *f;
+ HMAP_FOR_EACH_WITH_HASH (f, struct ctables_freq, node, hash, &t->ft)
+ {
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ {
+ const struct var_array *va = &t->vaas[a].vas[ix[a]];
+ if (f->axes[a].vaa_idx != ix[a])
+ goto not_equal;
+ for (size_t i = 0; i < va->n; i++)
+ if (i != va->scale_idx
+ && !value_equal (case_data (c, va->vars[i]),
+ &f->axes[a].values[i],
+ var_get_width (va->vars[i])))
+ goto not_equal;
+ }
+
+ goto summarize;
+
+ not_equal: ;
+ }
+
+ f = xmalloc (sizeof *f);
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ {
+ const struct var_array *va = &t->vaas[a].vas[ix[a]];
+ f->axes[a].vaa_idx = ix[a];
+ f->axes[a].values = (va->n
+ ? xnmalloc (va->n, sizeof *f->axes[a].values)
+ : NULL);
+ for (size_t i = 0; i < va->n; i++)
+ value_clone (&f->axes[a].values[i], case_data (c, va->vars[i]),
+ var_get_width (va->vars[i]));
+ }
+ f->summaries = xmalloc (ss->n_summaries * sizeof *f->summaries);
+ for (size_t i = 0; i < ss->n_summaries; i++)
+ ctables_summary_init (&f->summaries[i], &ss->summaries[i]);
+ for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++)
+ f->domains[dt] = ctables_domain_insert (t, f, dt);
+ hmap_insert (&t->ft, &f->node, hash);
+
+summarize:
+ for (size_t i = 0; i < ss->n_summaries; i++)
+ ctables_summary_add (&f->summaries[i], &ss->summaries[i], ss->summary_var,
+ case_data (c, ss->summary_var), weight);
+ for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++)
+ f->domains[dt]->valid += weight;
}
static bool
{
for (size_t i = 0; i < ct->n_tables; i++)
{
- size_t allocated_fts = 0;
+ struct ctables_table *t = ct->tables[i];
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ if (t->axes[a])
+ {
+ t->vaas[a] = enumerate_fts (a, t->axes[a]);
- struct ctables_table *t = &ct->tables[i];
- struct var_array2 vaa = enumerate_fts (t->axes[PIVOT_AXIS_ROW]);
- vaa = nest_fts (vaa, enumerate_fts (t->axes[PIVOT_AXIS_COLUMN]));
- vaa = nest_fts (vaa, enumerate_fts (t->axes[PIVOT_AXIS_LAYER]));
- for (size_t i = 0; i < vaa.n; i++)
- {
- for (size_t j = 0; j < vaa.vas[i].n; j++)
- {
- if (j)
- fputs (", ", stdout);
- fputs (var_get_name (vaa.vas[i].vars[j]), stdout);
- }
- putchar ('\n');
- }
+ for (size_t j = 0; j < t->vaas[a].n; j++)
+ {
+ struct var_array *va = &t->vaas[a].vas[j];
+ for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++)
+ {
+ va->domains[dt] = xmalloc (va->n * sizeof *va->domains[dt]);
+ va->n_domains[dt] = 0;
+
+ for (size_t k = 0; k < va->n; k++)
+ {
+ if (k == va->scale_idx)
+ continue;
+
+ switch (dt)
+ {
+ 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 == va->n - 1
+ || (va->scale_idx == va->n - 1
+ && k == va->n - 2))
+ continue;
+ }
+ break;
+
+ case CTDT_LAYERROW:
+ if (a == PIVOT_AXIS_COLUMN)
+ continue;
+ break;
+
+ case CTDT_LAYERCOL:
+ if (a == PIVOT_AXIS_ROW)
+ continue;
+ break;
+ }
+
+ va->domains[dt][va->n_domains[dt]++] = k;
+ }
+ }
+ }
+ }
+ else
+ {
+ struct var_array *va = xmalloc (sizeof *va);
+ *va = (struct var_array) { .n = 0 };
+ t->vaas[a] = (struct var_array2) { .vas = va, .n = 1 };
+ }
- for (size_t j = 0; j < vaa.n; j++)
+ for (size_t i = 0; i < t->vaas[t->summary_axis].n; i++)
{
- const struct var_array *va = &vaa.vas[j];
- const struct ctables_summary_spec *summaries;
- size_t n_summaries;
- const struct variable *summary_var;
- if (!va->summary)
+ struct var_array *va = &t->vaas[t->summary_axis].vas[i];
+ if (!va->n_summaries)
{
- static const struct ctables_summary_spec count = {
- .function = CTSF_COUNT,
- .label = (char *) N_("Count"),
- .format = { .type = FMT_F, .w = 40 },
+ va->summaries = xmalloc (sizeof *va->summaries);
+ va->n_summaries = 1;
+
+ enum ctables_summary_function function
+ = va->summary_var ? CTSF_MEAN : CTSF_COUNT;
+ struct ctables_var var = { .is_mrset = false, .var = va->summary_var };
+
+ *va->summaries = (struct ctables_summary_spec) {
+ .function = function,
+ .format = ctables_summary_default_format (function, &var),
+ .label = ctables_summary_default_label (function, 0),
};
- summaries = &count;
- n_summaries = 1;
- summary_var = va->vars[0];
- }
- else if (va->summary->n_summaries)
- {
- summaries = va->summary->summaries;
- n_summaries = va->summary->n_summaries;
- summary_var = va->summary->var.var;
+ if (!va->summary_var)
+ va->summary_var = va->vars[0];
}
- else
- {
- static const struct ctables_summary_spec mean = {
- .function = CTSF_MEAN,
- .label = (char *) N_("Mean"),
- .format = { .type = FMT_F, .w = 40, .d = 2}, /* XXX */
- };
- summaries = &mean;
- n_summaries = 1;
- summary_var = va->summary->var.var;
- };
-
- struct ctables_freqtab *ft = xmalloc (sizeof *ft);
- *ft = (struct ctables_freqtab) {
- .vars = *va,
- .summaries = summaries,
- .n_summaries = n_summaries,
- .summary_var = summary_var,
- .data = HMAP_INITIALIZER (ft->data),
- };
-
- if (t->n_fts >= allocated_fts)
- t->fts = x2nrealloc (t->fts, &allocated_fts, sizeof *t->fts);
- t->fts[t->n_fts++] = ft;
}
-
- free (vaa.vas);
}
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))
{
double weight = dict_get_case_weight (dataset_dict (ds), c,
&warn_on_invalid);
+ total_weight += weight;
for (size_t i = 0; i < ct->n_tables; i++)
{
- struct ctables_table *t = &ct->tables[i];
+ struct ctables_table *t = ct->tables[i];
- for (size_t j = 0; j < t->n_fts; j++)
- {
- struct ctables_freqtab *ft = t->fts[j];
-
- for (size_t k = 0; k < ft->vars.n; k++)
- {
- const struct variable *var = ft->vars.vars[k];
- switch (var_is_value_missing (var, case_data (c, var)))
- {
- case MV_SYSTEM:
- goto next_ft;
-
- case MV_USER:
- if (!t->categories[var_get_dict_index (var)]
- || !t->categories[var_get_dict_index (var)]->include_missing)
- goto next_ft;
- break;
- }
- }
- size_t hash = 0;
- for (size_t k = 0; k < ft->vars.n; k++)
- {
- const struct variable *var = ft->vars.vars[k];
- hash = value_hash (case_data (c, var), var_get_width (var), hash);
- }
-
- struct ctables_freq *f;
- HMAP_FOR_EACH_WITH_HASH (f, struct ctables_freq, node, hash, &ft->data)
- {
- for (size_t k = 0; k < ft->vars.n; k++)
- {
- const struct variable *var = ft->vars.vars[k];
- if (!value_equal (case_data (c, var), &f->values[k],
- var_get_width (var)))
- goto next_hash_node;
- }
- goto found;
-
- next_hash_node: ;
- }
-
- f = ctables_freq_create (ft);
- for (size_t k = 0; k < ft->vars.n; k++)
- {
- const struct variable *var = ft->vars.vars[k];
- value_clone (&f->values[k], case_data (c, var),
- var_get_width (var));
- }
- hmap_insert (&ft->data, &f->node, hash);
-
- found:
- ctables_freq_add (ft, f, ft->summary_var,
- case_data (c, ft->summary_var), weight);
-
- next_ft: ;
- }
+ for (size_t ir = 0; ir < t->vaas[PIVOT_AXIS_ROW].n; ir++)
+ for (size_t ic = 0; ic < t->vaas[PIVOT_AXIS_COLUMN].n; ic++)
+ for (size_t il = 0; il < t->vaas[PIVOT_AXIS_LAYER].n; il++)
+ ctables_freqtab_insert (t, c, ir, ic, il, weight);
}
}
casereader_destroy (input);
for (size_t i = 0; i < ct->n_tables; i++)
{
- struct ctables_table *t = &ct->tables[i];
-
- struct pivot_table *pt = pivot_table_create (N_("Custom Tables"));
- struct pivot_dimension *d = pivot_dimension_create (
- pt, PIVOT_AXIS_ROW, N_("Rows"));
- for (size_t j = 0; j < t->n_fts; j++)
+ struct ctables_table *t = ct->tables[i];
+
+ 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"))),
+ NULL);
+ 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++)
{
- struct ctables_freqtab *ft = t->fts[j];
- ft->sorted = xnmalloc (ft->data.count, sizeof *ft->sorted);
+ 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_freq **sorted = xnmalloc (t->ft.count, sizeof *sorted);
struct ctables_freq *f;
size_t n = 0;
- HMAP_FOR_EACH (f, struct ctables_freq, node, &ft->data)
- ft->sorted[n++] = f;
- assert (n == ft->data.count);
- sort (ft->sorted, n, sizeof *ft->sorted,
- ctables_freq_compare_3way, &ft->vars);
-
- struct pivot_category **groups = xnmalloc (ft->vars.n,
- sizeof *groups);
- for (size_t k = 0; k < n; k++)
+ HMAP_FOR_EACH (f, struct ctables_freq, node, &t->ft)
+ sorted[n++] = f;
+ assert (n == t->ft.count);
+
+ struct ctables_freq_sort_aux aux = { .t = t, .a = a };
+ sort (sorted, n, sizeof *sorted, ctables_freq_compare_3way, &aux);
+
+ size_t max_depth = 0;
+ for (size_t j = 0; j < t->vaas[a].n; j++)
+ if (t->vaas[a].vas[j].n > max_depth)
+ max_depth = t->vaas[a].vas[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_freq *prev = k > 0 ? ft->sorted[k - 1] : NULL;
- struct ctables_freq *f = ft->sorted[k];
+ struct ctables_freq *f = sorted[j];
+ const struct var_array *va = &t->vaas[a].vas[f->axes[a].vaa_idx];
size_t n_common = 0;
- if (prev)
- for (; n_common + 1 < ft->vars.n; n_common++)
- if (!value_equal (&prev->values[n_common],
- &f->values[n_common],
- var_get_type (ft->vars.vars[n_common])))
- break;
-
- for (size_t m = n_common; m < ft->vars.n; m++)
+ bool new_subtable = false;
+ if (j > 0)
{
- struct pivot_category *parent = m > 0 ? groups[m - 1] : d->root;
- const struct variable *var = ft->vars.vars[m];
- enum ctables_vlabel vlabel = ct->vlabels[var_get_dict_index (var)];
+ struct ctables_freq *prev = sorted[j - 1];
+ if (prev->axes[a].vaa_idx == f->axes[a].vaa_idx)
+ {
+ for (; n_common < va->n; n_common++)
+ if (n_common != va->scale_idx
+ && !value_equal (&prev->axes[a].values[n_common],
+ &f->axes[a].values[n_common],
+ var_get_type (va->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 (va->vars[0])];
+ top = d[a]->root;
if (vlabel != CTVL_NONE)
- parent = pivot_category_create_group__ (
- parent, pivot_value_new_variable (ft->vars.vars[m]));
-
- if (m + 1 < ft->vars.n)
- parent = pivot_category_create_group__ (
- parent,
- pivot_value_new_var_value (ft->vars.vars[m], &f->values[m]));
- groups[m] = parent;
+ top = pivot_category_create_group__ (
+ top, pivot_value_new_variable (va->vars[0]));
+ }
+ if (n_common == va->n)
+ {
+ f->axes[a].leaf = prev_leaf;
+ continue;
+ }
- if (m == ft->vars.n - 1)
+ for (size_t k = n_common; k < va->n; k++)
+ {
+ struct pivot_category *parent = k > 0 ? groups[k - 1] : top;
+
+ struct pivot_value *label
+ = (k != va->scale_idx
+ ? pivot_value_new_var_value (va->vars[k],
+ &f->axes[a].values[k])
+ : NULL);
+ if (k == va->n - 1)
{
- struct pivot_category *c = pivot_category_create_group__ (
- parent,
- pivot_value_new_var_value (ft->vars.vars[ft->vars.n - 1],
- &f->values[ft->vars.n - 1]));
- for (size_t p = 0; p < ft->n_summaries; p++)
+ if (a == t->summary_axis)
+ {
+ if (label)
+ parent = pivot_category_create_group__ (parent, label);
+ for (size_t m = 0; m < va->n_summaries; m++)
+ {
+ int leaf = pivot_category_create_leaf (
+ parent, pivot_value_new_text (va->summaries[m].label));
+ if (m == 0)
+ prev_leaf = leaf;
+ }
+ }
+ else
{
- double value = ctables_summary_value (
- &f->summaries[p], &ft->summaries[p]);
- int leaf = pivot_category_create_leaf (
- c, pivot_value_new_text (ft->summaries[p].label));
- pivot_table_put1 (pt, leaf, pivot_value_new_number (value));
+ /* 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);
}
+ break;
}
+
+ if (label)
+ parent = pivot_category_create_group__ (parent, label);
+
+ enum ctables_vlabel vlabel = ct->vlabels[var_get_dict_index (va->vars[k + 1])];
+ if (vlabel != CTVL_NONE)
+ parent = pivot_category_create_group__ (
+ parent, pivot_value_new_variable (va->vars[k + 1]));
+ groups[k] = parent;
}
+
+ f->axes[a].leaf = prev_leaf;
}
+ free (sorted);
free (groups);
}
+ struct ctables_freq *f;
+ HMAP_FOR_EACH (f, struct ctables_freq, node, &t->ft)
+ {
+ const struct var_array *ss = &t->vaas[t->summary_axis].vas[f->axes[t->summary_axis].vaa_idx];
+ for (size_t j = 0; j < ss->n_summaries; j++)
+ {
+ size_t dindexes[3];
+ size_t n_dindexes = 0;
+
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ if (d[a])
+ {
+ int leaf = f->axes[a].leaf;
+ if (a == t->summary_axis)
+ leaf += j;
+ dindexes[n_dindexes++] = leaf;
+ }
+
+ double d = ctables_summary_value (f, &f->summaries[j], &ss->summaries[j]);
+ struct pivot_value *value = pivot_value_new_number (d);
+ value->numeric.format = ss->summaries[j].format;
+ pivot_table_put (pt, dindexes, n_dindexes, value);
+ }
+ }
+
pivot_table_submit (pt);
}
+#if 0
for (size_t i = 0; i < ct->n_tables; i++)
{
- struct ctables_table *t = &ct->tables[i];
+ struct ctables_table *t = ct->tables[i];
for (size_t j = 0; j < t->n_fts; j++)
{
free (f);
}
hmap_destroy (&ft->data);
- free (ft->sorted);
var_array_uninit (&ft->vars);
free (ft);
}
free (t->fts);
}
-
+#endif
+
return proc_commit (ds);
}
.vlabels = vlabels,
.hide_threshold = 5,
};
+ ct->look->omit_empty = false;
if (!lex_force_match (lexer, T_SLASH))
goto error;
ct->tables = x2nrealloc (ct->tables, &allocated_tables,
sizeof *ct->tables);
- struct ctables_table *t = &ct->tables[ct->n_tables++];
+ struct ctables_table *t = xmalloc (sizeof *t);
*t = (struct ctables_table) {
+ .ft = HMAP_INITIALIZER (t->ft),
.slabels_position = PIVOT_AXIS_COLUMN,
.slabels_visible = true,
.row_labels = CTLP_NORMAL,
.n_categories = dict_get_n_vars (dataset_dict (ds)),
.cilevel = 95,
};
+ for (enum ctables_domain_type dt = 0; dt < N_CTDTS; dt++)
+ hmap_init (&t->domains[dt]);
+ ct->tables[ct->n_tables++] = t;
lex_match (lexer, T_EQUALS);
if (!ctables_axis_parse (lexer, dataset_dict (ds), ct, t, PIVOT_AXIS_ROW))
const struct ctables_axis *scales[PIVOT_N_AXES];
size_t n_scales = 0;
- for (size_t i = 0; i < 3; i++)
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
{
- scales[i] = find_scale (t->axes[i]);
- if (scales[i])
+ scales[a] = find_scale (t->axes[a]);
+ if (scales[a])
n_scales++;
}
if (n_scales > 1)
{
- msg (SE, _("Scale variables may appear only on one dimension."));
+ msg (SE, _("Scale variables may appear only on one axis."));
if (scales[PIVOT_AXIS_ROW])
msg_at (SN, scales[PIVOT_AXIS_ROW]->loc,
- _("This scale variable appears in the rows dimension."));
+ _("This scale variable appears on the rows axis."));
if (scales[PIVOT_AXIS_COLUMN])
msg_at (SN, scales[PIVOT_AXIS_COLUMN]->loc,
- _("This scale variable appears in the columns dimension."));
+ _("This scale variable appears on the columns axis."));
if (scales[PIVOT_AXIS_LAYER])
msg_at (SN, scales[PIVOT_AXIS_LAYER]->loc,
- _("This scale variable appears in the layer dimension."));
+ _("This scale variable appears on the layer axis."));
goto error;
}
+ const struct ctables_axis *summaries[PIVOT_N_AXES];
+ size_t n_summaries = 0;
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ {
+ summaries[a] = (scales[a]
+ ? scales[a]
+ : find_categorical_summary_spec (t->axes[a]));
+ if (summaries[a])
+ n_summaries++;
+ }
+ if (n_summaries > 1)
+ {
+ msg (SE, _("Summaries may appear only on one axis."));
+ if (summaries[PIVOT_AXIS_ROW])
+ msg_at (SN, summaries[PIVOT_AXIS_ROW]->loc,
+ _("This variable on the rows axis has a summary."));
+ if (summaries[PIVOT_AXIS_COLUMN])
+ msg_at (SN, summaries[PIVOT_AXIS_COLUMN]->loc,
+ _("This variable on the columns axis has a summary."));
+ if (summaries[PIVOT_AXIS_LAYER])
+ msg_at (SN, summaries[PIVOT_AXIS_LAYER]->loc,
+ _("This variable on the layers axis has a summary."));
+ goto error;
+ }
+ for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+ if (n_summaries ? summaries[a] : t->axes[a])
+ {
+ t->summary_axis = a;
+ break;
+ }
+
if (lex_token (lexer) == T_ENDCMD)
break;
if (!lex_force_match (lexer, T_SLASH))