Invent recursive_worktable_factor GUC to replace hard-wired constant.

Up to now, the planner estimated the size of a recursive query's
worktable as 10 times the size of the non-recursive term.  It's hard
to see how to do significantly better than that automatically, but
we can give users control over the multiplier to allow tuning for
specific use-cases.  The default behavior remains the same.

Simon Riggs

Discussion: https://postgr.es/m/CANbhV-EuaLm4H3g0+BSTYHEGxJj3Kht0R+rJ8vT57Dejnh=_nA@mail.gmail.com

src/backend/optimizer/path/costsize.c

@@ -123,6 +123,7 @@ double		cpu_index_tuple_cost = DEFAULT_CPU_INDEX_TUPLE_COST;
 double		cpu_operator_cost = DEFAULT_CPU_OPERATOR_COST;
 double		parallel_tuple_cost = DEFAULT_PARALLEL_TUPLE_COST;
 double		parallel_setup_cost = DEFAULT_PARALLEL_SETUP_COST;
+double		recursive_worktable_factor = DEFAULT_RECURSIVE_WORKTABLE_FACTOR;
 
 int			effective_cache_size = DEFAULT_EFFECTIVE_CACHE_SIZE;
 
@@ -5665,10 +5666,11 @@ set_cte_size_estimates(PlannerInfo *root, RelOptInfo *rel, double cte_rows)
 	if (rte->self_reference)
 	{
 		/*
-		 * In a self-reference, arbitrarily assume the average worktable size
-		 * is about 10 times the nonrecursive term's size.
+		 * In a self-reference, we assume the average worktable size is a
+		 * multiple of the nonrecursive term's size.  The best multiplier will
+		 * vary depending on query "fan-out", so make its value adjustable.
 		 */
-		rel->tuples = 10 * cte_rows;
+		rel->tuples = clamp_row_est(recursive_worktable_factor * cte_rows);
 	}
 	else
 	{

src/backend/utils/misc/guc.c

@@ -3740,6 +3740,18 @@ static struct config_real ConfigureNamesReal[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"recursive_worktable_factor", PGC_USERSET, QUERY_TUNING_OTHER,
+			gettext_noop("Sets the planner's estimate of the average size "
+						 "of a recursive query's working table."),
+			NULL,
+			GUC_EXPLAIN
+		},
+		&recursive_worktable_factor,
+		DEFAULT_RECURSIVE_WORKTABLE_FACTOR, 0.001, 1000000.0,
+		NULL, NULL, NULL
+	},
+
 	{
 		{"geqo_selection_bias", PGC_USERSET, QUERY_TUNING_GEQO,
 			gettext_noop("GEQO: selective pressure within the population."),

src/backend/utils/misc/postgresql.conf.sample

@@ -426,6 +426,7 @@
 					# JOIN clauses
 #plan_cache_mode = auto			# auto, force_generic_plan or
 					# force_custom_plan
+#recursive_worktable_factor = 10.0	# range 0.001-1000000
 
 
 #------------------------------------------------------------------------------

src/include/optimizer/cost.h

@@ -29,6 +29,8 @@
 #define DEFAULT_PARALLEL_TUPLE_COST 0.1
 #define DEFAULT_PARALLEL_SETUP_COST  1000.0
 
+/* defaults for non-Cost parameters */
+#define DEFAULT_RECURSIVE_WORKTABLE_FACTOR  10.0
 #define DEFAULT_EFFECTIVE_CACHE_SIZE  524288	/* measured in pages */
 
 typedef enum

src/include/optimizer/optimizer.h

@@ -91,6 +91,7 @@ extern PGDLLIMPORT double cpu_index_tuple_cost;
 extern PGDLLIMPORT double cpu_operator_cost;
 extern PGDLLIMPORT double parallel_tuple_cost;
 extern PGDLLIMPORT double parallel_setup_cost;
+extern PGDLLIMPORT double recursive_worktable_factor;
 extern PGDLLIMPORT int effective_cache_size;
 
 extern double clamp_row_est(double nrows);