Show estimated number of groups for Incremental Sort in EXPLAIN

Incremental Sort's costs heavily depend on the estimated number of input
groups with equal presorted key values.  Overestimations can cause the
planner to choose Incremental Sort over Sort when Sort would have
been a better choice.

Here, we add the planner's estimate to EXPLAIN to allow easier
understanding of why Incremental Sort has been chosen.

Author: Ilia Evdokimov <ilya.evdokimov@tantorlabs.com>
Reviewed-by: Enrique Sánchez <enriqueesanchz@gmail.com>
Reviewed-by: David Rowley <dgrowleyml@gmail.com>
Reviewed-by: solai v <solai.cdac@gmail.com>
Discussion: https://postgr.es/m/10682fef-3748-43f5-a932-7adcdd9bd2b8%40tantorlabs.com

doc/src/sgml/perform.sgml

@@ -325,6 +325,7 @@ EXPLAIN SELECT * FROM tenk1 ORDER BY hundred, ten LIMIT 100;
    ->  Incremental Sort  (cost=19.35..2033.39 rows=10000 width=244)
          Sort Key: hundred, ten
          Presorted Key: hundred
+         Estimated Groups: 100
          ->  Index Scan using tenk1_hundred on tenk1  (cost=0.29..1574.20 rows=10000 width=244)
 </screen>
 

src/backend/commands/explain.c

@@ -3306,11 +3306,16 @@ static void
 show_incremental_sort_info(IncrementalSortState *incrsortstate,
 						   ExplainState *es)
 {
+	IncrementalSort *plan = (IncrementalSort *) incrsortstate->ss.ps.plan;
 	IncrementalSortGroupInfo *fullsortGroupInfo;
 	IncrementalSortGroupInfo *prefixsortGroupInfo;
 
 	fullsortGroupInfo = &incrsortstate->incsort_info.fullsortGroupInfo;
 
+	if (es->costs)
+		ExplainPropertyFloat("Estimated Groups", NULL,
+							 plan->numGroups, 0, es);
+
 	if (!es->analyze)
 		return;
 

src/backend/optimizer/path/costsize.c

@@ -2049,6 +2049,9 @@ cost_tuplesort(Cost *startup_cost, Cost *run_cost,
  * We estimate the number of groups into which the relation is divided by the
  * leading pathkeys, and then calculate the cost of sorting a single group
  * with tuplesort using cost_tuplesort().
+ *
+ * If num_groups is not NULL, *num_groups gets set to the estimated number of
+ * sort groups.
  */
 void
 cost_incremental_sort(Path *path,
@@ -2056,7 +2059,8 @@ cost_incremental_sort(Path *path,
 					  int input_disabled_nodes,
 					  Cost input_startup_cost, Cost input_total_cost,
 					  double input_tuples, int width, Cost comparison_cost, int sort_mem,
-					  double limit_tuples)
+					  double limit_tuples,
+					  Cardinality *num_groups)
 {
 	Cost		startup_cost,
 				run_cost,
@@ -2184,6 +2188,10 @@ cost_incremental_sort(Path *path,
 
 	path->startup_cost = startup_cost;
 	path->total_cost = startup_cost + run_cost;
+
+	/* set output parameter values */
+	if (num_groups)
+		*num_groups = input_groups;
 }
 
 /*
@@ -2407,7 +2415,8 @@ cost_append(AppendPath *apath, PlannerInfo *root)
 											  subpath->pathtarget->width,
 											  0.0,
 											  work_mem,
-											  apath->limit_tuples);
+											  apath->limit_tuples,
+											  NULL);
 					}
 					else
 					{
@@ -3850,7 +3859,8 @@ initial_cost_mergejoin(PlannerInfo *root, JoinCostWorkspace *workspace,
 								  outer_path->pathtarget->width,
 								  0.0,
 								  work_mem,
-								  -1.0);
+								  -1.0,
+								  NULL);
 		}
 		else
 		{

src/backend/optimizer/plan/createplan.c

@@ -2074,6 +2074,7 @@ create_incrementalsort_plan(PlannerInfo *root, IncrementalSortPath *best_path,
 											  best_path->nPresortedCols);
 
 	copy_generic_path_info(&plan->sort.plan, (Path *) best_path);
+	plan->numGroups = best_path->numGroups;
 
 	return plan;
 }
@@ -5469,7 +5470,8 @@ label_incrementalsort_with_costsize(PlannerInfo *root, IncrementalSort *plan,
 						  lefttree->plan_width,
 						  0.0,
 						  work_mem,
-						  limit_tuples);
+						  limit_tuples,
+						  &plan->numGroups);
 	plan->sort.plan.startup_cost = sort_path.startup_cost;
 	plan->sort.plan.total_cost = sort_path.total_cost;
 	plan->sort.plan.plan_rows = lefttree->plan_rows;

src/backend/optimizer/util/pathnode.c

@@ -1605,7 +1605,8 @@ create_merge_append_path(PlannerInfo *root,
 									  subpath->pathtarget->width,
 									  0.0,
 									  work_mem,
-									  pathnode->limit_tuples);
+									  pathnode->limit_tuples,
+									  NULL);
 			}
 			else
 			{
@@ -2883,7 +2884,8 @@ create_incremental_sort_path(PlannerInfo *root,
 						  subpath->rows,
 						  subpath->pathtarget->width,
 						  0.0,	/* XXX comparison_cost shouldn't be 0? */
-						  work_mem, limit_tuples);
+						  work_mem, limit_tuples,
+						  &sort->numGroups);
 
 	sort->nPresortedCols = presorted_keys;
 

src/include/nodes/pathnodes.h

@@ -2548,6 +2548,7 @@ typedef struct IncrementalSortPath
 {
 	SortPath	spath;
 	int			nPresortedCols; /* number of presorted columns */
+	Cardinality numGroups;		/* estimated number of groups in input */
 } IncrementalSortPath;
 
 /*

src/include/nodes/plannodes.h

@@ -1171,6 +1171,8 @@ typedef struct IncrementalSort
 	Sort		sort;
 	/* number of presorted columns */
 	int			nPresortedCols;
+	/* estimated number of groups in input */
+	Cardinality numGroups;
 } IncrementalSort;
 
 /* ---------------

src/include/optimizer/cost.h

@@ -118,7 +118,8 @@ extern void cost_incremental_sort(Path *path,
 								  int input_disabled_nodes,
 								  Cost input_startup_cost, Cost input_total_cost,
 								  double input_tuples, int width, Cost comparison_cost, int sort_mem,
-								  double limit_tuples);
+								  double limit_tuples,
+								  Cardinality *num_groups);
 extern void cost_append(AppendPath *apath, PlannerInfo *root);
 extern void cost_merge_append(Path *path, PlannerInfo *root,
 							  List *pathkeys, int n_streams,