Start work on slabels != summary_axis.
authorBen Pfaff <blp@cs.stanford.edu>
Sat, 15 Jan 2022 19:03:21 +0000 (11:03 -0800)
committerBen Pfaff <blp@cs.stanford.edu>
Sat, 2 Apr 2022 01:48:55 +0000 (18:48 -0700)
src/language/stats/ctables.c

index 08f7a74180a67bab097d1fad9b51d7f9cd76339a..7247651c7ed4cc8422e99f44ad0ddede78bdaf27 100644 (file)
@@ -321,6 +321,7 @@ 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];
 
@@ -2453,6 +2454,366 @@ merge_item_compare_3way (const struct merge_item *a, const struct merge_item *b)
   return strcmp (as->label, bs->label);
 }
 
+static void
+ctables_table_output_same_axis (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));
+
+  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;
+
+      assert (t->axes[a]);
+
+      struct ctables_cell **sorted = xnmalloc (t->cells.count, sizeof *sorted);
+
+      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);
+
+      struct ctables_cell_sort_aux aux = { .t = t, .a = a };
+      sort (sorted, n, 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;
+
+      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];
+
+          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;
+
+          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;
+            }
+
+          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);
+
+                      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 (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;
+            }
+
+          cell->axes[a].leaf = prev_leaf;
+        }
+      free (sorted);
+      free (groups);
+    }
+  struct ctables_cell *cell;
+  HMAP_FOR_EACH (cell, struct ctables_cell, node, &t->cells)
+    {
+      if (cell->hide)
+        continue;
+
+      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;
+
+          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;
+              }
+
+          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);
+        }
+    }
+
+  pivot_table_submit (pt);
+}
+
+
+static void
+ctables_table_output_different_axis (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));
+
+  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;
+
+      assert (t->axes[a]);
+
+      struct ctables_cell **sorted = xnmalloc (t->cells.count, sizeof *sorted);
+
+      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);
+
+      struct ctables_cell_sort_aux aux = { .t = t, .a = a };
+      sort (sorted, n, 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;
+
+      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];
+
+          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;
+
+          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;
+            }
+
+          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);
+
+                      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 (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;
+            }
+
+          cell->axes[a].leaf = prev_leaf;
+        }
+      free (sorted);
+      free (groups);
+    }
+  struct ctables_cell *cell;
+  HMAP_FOR_EACH (cell, struct ctables_cell, node, &t->cells)
+    {
+      if (cell->hide)
+        continue;
+
+      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;
+
+          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;
+              }
+
+          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);
+        }
+    }
+
+  pivot_table_submit (pt);
+}
+
+
 static bool
 ctables_execute (struct dataset *ds, struct ctables *ct)
 {
@@ -2554,7 +2915,7 @@ ctables_execute (struct dataset *ds, struct ctables *ct)
                                             &nest->specs[CSV_CELL]);
         }
 
-      struct ctables_summary_spec_set merged = { .n = 0 };
+      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++)
@@ -2581,17 +2942,17 @@ ctables_execute (struct dataset *ds, struct ctables *ct)
               min = items[j];
 
           /* 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 (merged->n >= merged->allocated)
+            merged->specs = x2nrealloc (merged->specs, &merged->allocated,
+                                        sizeof *merged->specs);
+          merged->specs[merged->n++] = min.set->specs[min.ofs];
 
           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;
+                  item->set->specs[item->ofs].axis_idx = merged->n - 1;
                   if (++item->ofs >= item->set->n)
                     {
                       items[j] = items[--n_left];
@@ -2602,8 +2963,8 @@ ctables_execute (struct dataset *ds, struct ctables *ct)
             }
         }
 
-      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 < merged->n; j++)
+        printf ("%s\n", ctables_summary_function_name (merged->specs[j].function));
 
       for (size_t j = 0; j < stack->n; j++)
         {
@@ -2646,183 +3007,11 @@ ctables_execute (struct dataset *ds, struct ctables *ct)
   for (size_t i = 0; i < ct->n_tables; i++)
     {
       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"))),
-        "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++)
-        {
-          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_cell **sorted = xnmalloc (t->cells.count, sizeof *sorted);
-
-          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);
-
-          struct ctables_cell_sort_aux aux = { .t = t, .a = a };
-          sort (sorted, n, 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;
-
-          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];
-
-              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;
-
-              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;
-                }
-
-              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);
-
-                          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 (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;
-                }
-
-              cell->axes[a].leaf = prev_leaf;
-            }
-          free (sorted);
-          free (groups);
-        }
-      struct ctables_cell *cell;
-      HMAP_FOR_EACH (cell, struct ctables_cell, node, &t->cells)
-        {
-          if (cell->hide)
-            continue;
-
-          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;
-
-              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;
-                  }
-
-              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);
-            }
-        }
-
-      pivot_table_submit (pt);
+      if (t->summary_axis == t->slabels_position)
+        ctables_table_output_same_axis (ct, ct->tables[i]);
+      else
+        ctables_table_output_different_axis (ct, ct->tables[i]);
     }
-
   return proc_commit (ds);
 }