Remove hashagg_avoid_disk_plan GUC.

Note: This GUC was originally named enable_hashagg_disk when it appeared
in commit 1f39bce0, which added disk-based hash aggregation.  It was
subsequently renamed in commit 92c58fd9.

Author: Peter Geoghegan
Reviewed-By: Jeff Davis, Álvaro Herrera
Discussion: https://postgr.es/m/9d9d1e1252a52ea1bad84ea40dbebfd54e672a0f.camel%40j-davis.com
Backpatch: 13-, where disk-based hash aggregation was introduced.

doc/src/sgml/config.sgml

@@ -4840,23 +4840,6 @@ ANY <replaceable class="parameter">num_sync</replaceable> ( <replaceable class="
       </listitem>
      </varlistentry>
 
-     <varlistentry id="guc-hashagg-avoid-disk-plan" xreflabel="hashagg_avoid_disk_plan">
-      <term><varname>hashagg_avoid_disk_plan</varname> (<type>boolean</type>)
-      <indexterm>
-       <primary><varname>hashagg_avoid_disk_plan</varname> configuration parameter</primary>
-      </indexterm>
-      </term>
-      <listitem>
-       <para>
-        If set to <literal>on</literal>, causes the planner to avoid choosing
-        hashed aggregation plans that are expected to use the disk. If hashed
-        aggregation is chosen, it may still require the use of disk at
-        execution time, even if this parameter is enabled. The default is
-        <literal>off</literal>.
-       </para>
-      </listitem>
-     </varlistentry>
-
      </variablelist>
      </sect2>
      <sect2 id="runtime-config-query-constants">

src/backend/optimizer/path/costsize.c

@@ -130,7 +130,6 @@ bool		enable_tidscan = true;
 bool		enable_sort = true;
 bool		enable_incremental_sort = true;
 bool		enable_hashagg = true;
-bool		hashagg_avoid_disk_plan = true;
 bool		enable_nestloop = true;
 bool		enable_material = true;
 bool		enable_mergejoin = true;

src/backend/optimizer/plan/planner.c

@@ -4850,11 +4850,10 @@ create_distinct_paths(PlannerInfo *root,
 	 * Consider hash-based implementations of DISTINCT, if possible.
 	 *
 	 * If we were not able to make any other types of path, we *must* hash or
-	 * die trying.  If we do have other choices, there are several things that
+	 * die trying.  If we do have other choices, there are two things that
 	 * should prevent selection of hashing: if the query uses DISTINCT ON
 	 * (because it won't really have the expected behavior if we hash), or if
-	 * enable_hashagg is off, or if it looks like the hashtable will exceed
+	 * enable_hashagg is off.
-	 * work_mem.
 	 *
 	 * Note: grouping_is_hashable() is much more expensive to check than the
 	 * other gating conditions, so we want to do it last.
@@ -4864,12 +4863,7 @@ create_distinct_paths(PlannerInfo *root,
 	else if (parse->hasDistinctOn || !enable_hashagg)
 		allow_hash = false;		/* policy-based decision not to hash */
 	else
-	{
+		allow_hash = true;		/* default */
-		Size		hashentrysize = hash_agg_entry_size(0, cheapest_input_path->pathtarget->width, 0);
-
-		allow_hash = !hashagg_avoid_disk_plan ||
-			(hashentrysize * numDistinctRows <= work_mem * 1024L);
-	}
 
 	if (allow_hash && grouping_is_hashable(parse->distinctClause))
 	{
@@ -6749,8 +6743,6 @@ add_paths_to_grouping_rel(PlannerInfo *root, RelOptInfo *input_rel,
 
 	if (can_hash)
 	{
-		double		hashaggtablesize;
-
 		if (parse->groupingSets)
 		{
 			/*
@@ -6762,63 +6754,41 @@ add_paths_to_grouping_rel(PlannerInfo *root, RelOptInfo *input_rel,
 		}
 		else
 		{
-			hashaggtablesize = estimate_hashagg_tablesize(cheapest_path,
-														  agg_costs,
-														  dNumGroups);
-
 			/*
-			 * Provided that the estimated size of the hashtable does not
+			 * Generate a HashAgg Path.  We just need an Agg over the
-			 * exceed work_mem, we'll generate a HashAgg Path, although if we
+			 * cheapest-total input path, since input order won't matter.
-			 * were unable to sort above, then we'd better generate a Path, so
-			 * that we at least have one.
 			 */
-			if (!hashagg_avoid_disk_plan ||
+			add_path(grouped_rel, (Path *)
-				hashaggtablesize < work_mem * 1024L ||
+					 create_agg_path(root, grouped_rel,
-				grouped_rel->pathlist == NIL)
+									 cheapest_path,
-			{
+									 grouped_rel->reltarget,
-				/*
+									 AGG_HASHED,
-				 * We just need an Agg over the cheapest-total input path,
+									 AGGSPLIT_SIMPLE,
-				 * since input order won't matter.
+									 parse->groupClause,
-				 */
+									 havingQual,
-				add_path(grouped_rel, (Path *)
+									 agg_costs,
-						 create_agg_path(root, grouped_rel,
+									 dNumGroups));
-										 cheapest_path,
-										 grouped_rel->reltarget,
-										 AGG_HASHED,
-										 AGGSPLIT_SIMPLE,
-										 parse->groupClause,
-										 havingQual,
-										 agg_costs,
-										 dNumGroups));
-			}
 		}
 
 		/*
 		 * Generate a Finalize HashAgg Path atop of the cheapest partially
-		 * grouped path, assuming there is one. Once again, we'll only do this
+		 * grouped path, assuming there is one
-		 * if it looks as though the hash table won't exceed work_mem.
 		 */
 		if (partially_grouped_rel && partially_grouped_rel->pathlist)
 		{
 			Path	   *path = partially_grouped_rel->cheapest_total_path;
 
-			hashaggtablesize = estimate_hashagg_tablesize(path,
+			add_path(grouped_rel, (Path *)
-														  agg_final_costs,
+					 create_agg_path(root,
-														  dNumGroups);
+									 grouped_rel,
-
+									 path,
-			if (!hashagg_avoid_disk_plan ||
+									 grouped_rel->reltarget,
-				hashaggtablesize < work_mem * 1024L)
+									 AGG_HASHED,
-				add_path(grouped_rel, (Path *)
+									 AGGSPLIT_FINAL_DESERIAL,
-						 create_agg_path(root,
+									 parse->groupClause,
-										 grouped_rel,
+									 havingQual,
-										 path,
+									 agg_final_costs,
-										 grouped_rel->reltarget,
+									 dNumGroups));
-										 AGG_HASHED,
-										 AGGSPLIT_FINAL_DESERIAL,
-										 parse->groupClause,
-										 havingQual,
-										 agg_final_costs,
-										 dNumGroups));
 		}
 	}
 
@@ -7171,65 +7141,43 @@ create_partial_grouping_paths(PlannerInfo *root,
 		}
 	}
 
+	/*
+	 * Add a partially-grouped HashAgg Path where possible
+	 */
 	if (can_hash && cheapest_total_path != NULL)
 	{
-		double		hashaggtablesize;
-
 		/* Checked above */
 		Assert(parse->hasAggs || parse->groupClause);
 
-		hashaggtablesize =
+		add_path(partially_grouped_rel, (Path *)
-			estimate_hashagg_tablesize(cheapest_total_path,
+				 create_agg_path(root,
-									   agg_partial_costs,
+								 partially_grouped_rel,
-									   dNumPartialGroups);
+								 cheapest_total_path,
-
+								 partially_grouped_rel->reltarget,
-		/*
+								 AGG_HASHED,
-		 * Tentatively produce a partial HashAgg Path, depending on if it
+								 AGGSPLIT_INITIAL_SERIAL,
-		 * looks as if the hash table will fit in work_mem.
+								 parse->groupClause,
-		 */
+								 NIL,
-		if ((!hashagg_avoid_disk_plan || hashaggtablesize < work_mem * 1024L) &&
+								 agg_partial_costs,
-			cheapest_total_path != NULL)
+								 dNumPartialGroups));
-		{
-			add_path(partially_grouped_rel, (Path *)
-					 create_agg_path(root,
-									 partially_grouped_rel,
-									 cheapest_total_path,
-									 partially_grouped_rel->reltarget,
-									 AGG_HASHED,
-									 AGGSPLIT_INITIAL_SERIAL,
-									 parse->groupClause,
-									 NIL,
-									 agg_partial_costs,
-									 dNumPartialGroups));
-		}
 	}
 
+	/*
+	 * Now add a partially-grouped HashAgg partial Path where possible
+	 */
 	if (can_hash && cheapest_partial_path != NULL)
 	{
-		double		hashaggtablesize;
+		add_partial_path(partially_grouped_rel, (Path *)
-
+						 create_agg_path(root,
-		hashaggtablesize =
+										 partially_grouped_rel,
-			estimate_hashagg_tablesize(cheapest_partial_path,
+										 cheapest_partial_path,
-									   agg_partial_costs,
+										 partially_grouped_rel->reltarget,
-									   dNumPartialPartialGroups);
+										 AGG_HASHED,
-
+										 AGGSPLIT_INITIAL_SERIAL,
-		/* Do the same for partial paths. */
+										 parse->groupClause,
-		if ((!hashagg_avoid_disk_plan ||
+										 NIL,
-			 hashaggtablesize < work_mem * 1024L) &&
+										 agg_partial_costs,
-			cheapest_partial_path != NULL)
+										 dNumPartialPartialGroups));
-		{
-			add_partial_path(partially_grouped_rel, (Path *)
-							 create_agg_path(root,
-											 partially_grouped_rel,
-											 cheapest_partial_path,
-											 partially_grouped_rel->reltarget,
-											 AGG_HASHED,
-											 AGGSPLIT_INITIAL_SERIAL,
-											 parse->groupClause,
-											 NIL,
-											 agg_partial_costs,
-											 dNumPartialPartialGroups));
-		}
 	}
 
 	/*

src/backend/utils/misc/guc.c

@@ -1006,16 +1006,6 @@ static struct config_bool ConfigureNamesBool[] =
 		true,
 		NULL, NULL, NULL
 	},
-	{
-		{"hashagg_avoid_disk_plan", PGC_USERSET, QUERY_TUNING_METHOD,
-			gettext_noop("Causes the planner to avoid hashed aggregation plans that are expected to use the disk."),
-			NULL,
-			GUC_EXPLAIN
-		},
-		&hashagg_avoid_disk_plan,
-		false,
-		NULL, NULL, NULL
-	},
 	{
 		{"enable_material", PGC_USERSET, QUERY_TUNING_METHOD,
 			gettext_noop("Enables the planner's use of materialization."),

src/include/optimizer/cost.h

@@ -55,7 +55,6 @@ extern PGDLLIMPORT bool enable_tidscan;
 extern PGDLLIMPORT bool enable_sort;
 extern PGDLLIMPORT bool enable_incremental_sort;
 extern PGDLLIMPORT bool enable_hashagg;
-extern PGDLLIMPORT bool hashagg_avoid_disk_plan;
 extern PGDLLIMPORT bool enable_nestloop;
 extern PGDLLIMPORT bool enable_material;
 extern PGDLLIMPORT bool enable_mergejoin;