struct var_array
{
- const struct ctables_axis *summary;
struct variable **vars;
size_t n;
+
+ struct ctables_summary_spec *summaries;
+ size_t n_summaries;
+ struct variable *summary_var;
};
struct var_array2
{
struct ctables_axis *axes[PIVOT_N_AXES];
struct var_array2 vaas[PIVOT_N_AXES];
+ enum pivot_axis_type summary_axis;
struct hmap ft;
enum pivot_axis_type slabels_position;
}
}
+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;
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
+ .n = n,
+ .summaries = summary_src->summaries,
+ .n_summaries = summary_src->n_summaries,
+ .summary_var = summary_src->summary_var,
};
}
var_array2_uninit (&va0);
*axes = axis_type;
*va = (struct var_array) { .vars = vars, .n = 1 };
}
- va->summary = a->scale || a->n_summaries ? a : NULL;
+ if (a->n_summaries || a->scale)
+ {
+ va->summaries = a->summaries;
+ va->n_summaries = a->n_summaries;
+ va->summary_var = a->var.var;
+ }
return (struct var_array2) { .vas = va, .n = 1 };
case CTAO_STACK:
/* XXX percentiles, median, mode, multiple response */
};
-#if 0
static void
ctables_summary_init (union ctables_summary *s,
const struct ctables_summary_spec *ss)
}
}
-static void
+static void UNUSED
ctables_summary_uninit (union ctables_summary *s,
const struct ctables_summary_spec *ss)
{
NOT_REACHED ();
}
-#endif
struct ctables_freq
{
{
size_t vaa_idx;
union value *values;
+ int leaf;
}
axes[PIVOT_N_AXES];
- //union ctables_summary *summaries;
- double count;
+ union ctables_summary *summaries;
};
#if 0
[PIVOT_AXIS_COLUMN] = ic,
[PIVOT_AXIS_LAYER] = il,
};
+ const struct var_array *ss = &t->vaas[t->summary_axis].vas[ix[t->summary_axis]];
size_t hash = 0;
for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
goto not_equal;
}
- f->count += weight;
- return;
+ goto summarize;
not_equal: ;
}
value_clone (&f->axes[a].values[i], case_data (c, va->vars[i]),
var_get_width (va->vars[i]));
}
- f->count = weight;
+ 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]);
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);
}
static bool
*va = (struct var_array) { .n = 0 };
t->vaas[a] = (struct var_array2) { .vas = va, .n = 1 };
}
+
+ for (size_t i = 0; i < t->vaas[t->summary_axis].n; i++)
+ {
+ struct var_array *va = &t->vaas[t->summary_axis].vas[i];
+ if (!va->n_summaries)
+ {
+ 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),
+ };
+ if (!va->summary_var)
+ va->summary_var = va->vars[0];
+ }
+ }
}
struct casereader *input = casereader_create_filter_weight (proc_open (ds),
[PIVOT_AXIS_COLUMN] = N_("Columns"),
[PIVOT_AXIS_LAYER] = N_("Layers"),
};
- d[a] = (t->axes[a]
+ 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;
assert (n == t->ft.count);
struct ctables_freq_sort_aux aux = { .t = t, .a = a };
- n = sort_unique (sorted, n, sizeof *sorted, ctables_freq_compare_3way, &aux);
+ 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++)
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 *f = sorted[j];
}
else
new_subtable = true;
+
if (new_subtable)
- top = pivot_category_create_group__ (
- d[a]->root, pivot_value_new_variable (va->vars[0]));
- printf ("n_common=%zu\n", n_common);
+ {
+ enum ctables_vlabel vlabel = ct->vlabels[var_get_dict_index (va->vars[0])];
+ top = d[a]->root;
+ if (vlabel != CTVL_NONE)
+ 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;
+ }
for (size_t k = n_common; k < va->n; k++)
{
struct pivot_category *parent = k > 0 ? groups[k - 1] : top;
+ struct pivot_value *label = pivot_value_new_var_value (
+ va->vars[k], &f->axes[a].values[k]);
+
if (k == va->n - 1)
{
- pivot_category_create_leaf (
- parent,
- pivot_value_new_var_value (va->vars[va->n - 1],
- &f->axes[a].values[va->n - 1]));
+ if (a == t->summary_axis)
+ {
+ 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
+ prev_leaf = pivot_category_create_leaf (parent, label);
break;
}
- parent = pivot_category_create_group__ (
- parent,
- pivot_value_new_var_value (va->vars[k], &f->axes[a].values[k]));
+ parent = pivot_category_create_group__ (parent, label);
- parent = pivot_category_create_group__ (
- parent, pivot_value_new_variable (va->vars[k]));
+ 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;
-
-#if 0
- for (size_t p = 0; p < ft->n_summaries; p++)
- {
- if (a == t->slabels_position)
- pivot_category_create_leaf (
- c, pivot_value_new_text (ft->summaries[p].label));
- //pivot_table_put1 (pt, leaf, pivot_value_new_number (value));
- }
-#endif
- }
}
+
+ 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 value = ctables_summary_value (&f->summaries[j], &ss->summaries[j]);
+ pivot_table_put (pt, dindexes, n_dindexes,
+ pivot_value_new_number (value));
+ }
+ }
+
pivot_table_submit (pt);
}
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))