v7-0001-Allow-estimate_num_groups-to-pass-back-further-de.patch

application/octet-stream

Filename: v7-0001-Allow-estimate_num_groups-to-pass-back-further-de.patch
Type: application/octet-stream
Part: 0
Message: Re: Hybrid Hash/Nested Loop joins and caching results from subplans

Patch

Format: format-patch
Series: patch v7-0001
Subject: Allow estimate_num_groups() to pass back further details about the estimation
File+
contrib/postgres_fdw/postgres_fdw.c 1 1
src/backend/optimizer/path/costsize.c 2 1
src/backend/optimizer/path/indxpath.c 1 0
src/backend/optimizer/plan/planner.c 6 4
src/backend/optimizer/prep/prepunion.c 1 0
src/backend/optimizer/util/pathnode.c 1 0
src/backend/utils/adt/selfuncs.c 20 1
src/include/utils/selfuncs.h 10 1
From 5b7f411e829e7b902d807f09c7b6be4dcddc0fc5 Mon Sep 17 00:00:00 2001
From: "dgrowley@gmail.com" <dgrowley@gmail.com>
Date: Thu, 2 Jul 2020 16:06:36 +1200
Subject: [PATCH v7 1/3] Allow estimate_num_groups() to pass back further
 details about the estimation

Here we add a new output parameter to estimate_num_groups() to allow it to
set a flags variable with some bits to allow it to pass back additional
details to the caller which may be useful for decision making.

For now, the only new flag is one which indicates if the estimation
fell back on using the hard-coded constants in any part of the estimation.
Callers may like to change their behavior if this is set, and this gives
them the ability to do so. Callers may pass the flag pointer as NULL if
they have no interest in any of the flags.

We're not adding any actual usages of these flags here.  Some follow-up
commits will make use of this feature.
---
 contrib/postgres_fdw/postgres_fdw.c    |  2 +-
 src/backend/optimizer/path/costsize.c  |  3 ++-
 src/backend/optimizer/path/indxpath.c  |  1 +
 src/backend/optimizer/plan/planner.c   | 10 ++++++----
 src/backend/optimizer/prep/prepunion.c |  1 +
 src/backend/optimizer/util/pathnode.c  |  1 +
 src/backend/utils/adt/selfuncs.c       | 21 ++++++++++++++++++++-
 src/include/utils/selfuncs.h           | 11 ++++++++++-
 8 files changed, 42 insertions(+), 8 deletions(-)

diff --git a/contrib/postgres_fdw/postgres_fdw.c b/contrib/postgres_fdw/postgres_fdw.c
index 9fc53cad68..70f6fa2493 100644
--- a/contrib/postgres_fdw/postgres_fdw.c
+++ b/contrib/postgres_fdw/postgres_fdw.c
@@ -2960,7 +2960,7 @@ estimate_path_cost_size(PlannerInfo *root,
 			numGroups = estimate_num_groups(root,
 											get_sortgrouplist_exprs(root->parse->groupClause,
 																	fpinfo->grouped_tlist),
-											input_rows, NULL);
+											input_rows, NULL, NULL);
 
 			/*
 			 * Get the retrieved_rows and rows estimates.  If there are HAVING
diff --git a/src/backend/optimizer/path/costsize.c b/src/backend/optimizer/path/costsize.c
index fda4b2c6e8..5a7f5afb94 100644
--- a/src/backend/optimizer/path/costsize.c
+++ b/src/backend/optimizer/path/costsize.c
@@ -1864,7 +1864,8 @@ cost_incremental_sort(Path *path,
 
 	/* Estimate number of groups with equal presorted keys. */
 	if (!unknown_varno)
-		input_groups = estimate_num_groups(root, presortedExprs, input_tuples, NULL);
+		input_groups = estimate_num_groups(root, presortedExprs, input_tuples,
+										   NULL, NULL);
 
 	group_tuples = input_tuples / input_groups;
 	group_input_run_cost = input_run_cost / input_groups;
diff --git a/src/backend/optimizer/path/indxpath.c b/src/backend/optimizer/path/indxpath.c
index bcb1bc6097..4f6ab5d635 100644
--- a/src/backend/optimizer/path/indxpath.c
+++ b/src/backend/optimizer/path/indxpath.c
@@ -1986,6 +1986,7 @@ adjust_rowcount_for_semijoins(PlannerInfo *root,
 			nunique = estimate_num_groups(root,
 										  sjinfo->semi_rhs_exprs,
 										  nraw,
+										  NULL,
 										  NULL);
 			if (rowcount > nunique)
 				rowcount = nunique;
diff --git a/src/backend/optimizer/plan/planner.c b/src/backend/optimizer/plan/planner.c
index b40a112c25..64d8cfb89f 100644
--- a/src/backend/optimizer/plan/planner.c
+++ b/src/backend/optimizer/plan/planner.c
@@ -3719,7 +3719,8 @@ get_number_of_groups(PlannerInfo *root,
 					double		numGroups = estimate_num_groups(root,
 																groupExprs,
 																path_rows,
-																&gset);
+																&gset,
+																NULL);
 
 					gs->numGroups = numGroups;
 					rollup->numGroups += numGroups;
@@ -3744,7 +3745,8 @@ get_number_of_groups(PlannerInfo *root,
 					double		numGroups = estimate_num_groups(root,
 																groupExprs,
 																path_rows,
-																&gset);
+																&gset,
+																NULL);
 
 					gs->numGroups = numGroups;
 					gd->dNumHashGroups += numGroups;
@@ -3760,7 +3762,7 @@ get_number_of_groups(PlannerInfo *root,
 												 target_list);
 
 			dNumGroups = estimate_num_groups(root, groupExprs, path_rows,
-											 NULL);
+											 NULL, NULL);
 		}
 	}
 	else if (parse->groupingSets)
@@ -4778,7 +4780,7 @@ create_distinct_paths(PlannerInfo *root,
 												parse->targetList);
 		numDistinctRows = estimate_num_groups(root, distinctExprs,
 											  cheapest_input_path->rows,
-											  NULL);
+											  NULL, NULL);
 	}
 
 	/*
diff --git a/src/backend/optimizer/prep/prepunion.c b/src/backend/optimizer/prep/prepunion.c
index 2ebd4ea332..20b2025272 100644
--- a/src/backend/optimizer/prep/prepunion.c
+++ b/src/backend/optimizer/prep/prepunion.c
@@ -338,6 +338,7 @@ recurse_set_operations(Node *setOp, PlannerInfo *root,
 				*pNumGroups = estimate_num_groups(subroot,
 												  get_tlist_exprs(subquery->targetList, false),
 												  subpath->rows,
+												  NULL,
 												  NULL);
 		}
 	}
diff --git a/src/backend/optimizer/util/pathnode.c b/src/backend/optimizer/util/pathnode.c
index c1fc866cbf..e528e05459 100644
--- a/src/backend/optimizer/util/pathnode.c
+++ b/src/backend/optimizer/util/pathnode.c
@@ -1688,6 +1688,7 @@ create_unique_path(PlannerInfo *root, RelOptInfo *rel, Path *subpath,
 	pathnode->path.rows = estimate_num_groups(root,
 											  sjinfo->semi_rhs_exprs,
 											  rel->rows,
+											  NULL,
 											  NULL);
 	numCols = list_length(sjinfo->semi_rhs_exprs);
 
diff --git a/src/backend/utils/adt/selfuncs.c b/src/backend/utils/adt/selfuncs.c
index 00c7afc66f..2f1c1b8ec4 100644
--- a/src/backend/utils/adt/selfuncs.c
+++ b/src/backend/utils/adt/selfuncs.c
@@ -3241,6 +3241,7 @@ typedef struct
 	Node	   *var;			/* might be an expression, not just a Var */
 	RelOptInfo *rel;			/* relation it belongs to */
 	double		ndistinct;		/* # distinct values */
+	bool		isdefault;		/* true if DEFAULT_NUM_DISTINCT was used */
 } GroupVarInfo;
 
 static List *
@@ -3287,6 +3288,7 @@ add_unique_group_var(PlannerInfo *root, List *varinfos,
 	varinfo->var = var;
 	varinfo->rel = vardata->rel;
 	varinfo->ndistinct = ndistinct;
+	varinfo->isdefault = isdefault;
 	varinfos = lappend(varinfos, varinfo);
 	return varinfos;
 }
@@ -3311,6 +3313,11 @@ add_unique_group_var(PlannerInfo *root, List *varinfos,
  *	pgset - NULL, or a List** pointing to a grouping set to filter the
  *		groupExprs against
  *
+ * Outputs:
+ *	flags - When passed as non-NULL, the function sets bits in this
+ *		parameter to provide further details to callers about some
+ *		assumptions which were made when performing the estimation.
+ *
  * Given the lack of any cross-correlation statistics in the system, it's
  * impossible to do anything really trustworthy with GROUP BY conditions
  * involving multiple Vars.  We should however avoid assuming the worst
@@ -3358,7 +3365,7 @@ add_unique_group_var(PlannerInfo *root, List *varinfos,
  */
 double
 estimate_num_groups(PlannerInfo *root, List *groupExprs, double input_rows,
-					List **pgset)
+					List **pgset, int *flags)
 {
 	List	   *varinfos = NIL;
 	double		srf_multiplier = 1.0;
@@ -3366,6 +3373,10 @@ estimate_num_groups(PlannerInfo *root, List *groupExprs, double input_rows,
 	ListCell   *l;
 	int			i;
 
+	/* Zero the flags output parameter, if set */
+	if (flags != NULL)
+		*flags = 0;
+
 	/*
 	 * We don't ever want to return an estimate of zero groups, as that tends
 	 * to lead to division-by-zero and other unpleasantness.  The input_rows
@@ -3569,6 +3580,14 @@ estimate_num_groups(PlannerInfo *root, List *groupExprs, double input_rows,
 					if (relmaxndistinct < varinfo2->ndistinct)
 						relmaxndistinct = varinfo2->ndistinct;
 					relvarcount++;
+
+					/*
+					 * When varinfo2's isdefault is set then we'd better mark
+					 * that fact in the selectivity flags variable.
+					 */
+					if (flags != NULL && varinfo2->isdefault)
+						*flags |= SELFLAG_USED_DEFAULT;
+
 				}
 
 				/* we're done with this relation */
diff --git a/src/include/utils/selfuncs.h b/src/include/utils/selfuncs.h
index 7ac4a06391..455e1343ee 100644
--- a/src/include/utils/selfuncs.h
+++ b/src/include/utils/selfuncs.h
@@ -65,6 +65,14 @@
 			p = 1.0; \
 	} while (0)
 
+/*
+ * A set of flags which some selectivity estimation functions can pass back to
+ * callers to provide further details about some assumptions which were made
+ * during the estimation.
+ */
+#define SELFLAG_USED_DEFAULT		(1 << 0) /* Estimation fell back on one of
+											  * the DEFAULTs as defined above.
+											  */
 
 /* Return data from examine_variable and friends */
 typedef struct VariableStatData
@@ -194,7 +202,8 @@ extern void mergejoinscansel(PlannerInfo *root, Node *clause,
 							 Selectivity *rightstart, Selectivity *rightend);
 
 extern double estimate_num_groups(PlannerInfo *root, List *groupExprs,
-								  double input_rows, List **pgset);
+								  double input_rows, List **pgset,
+								  int *flags);
 
 extern void estimate_hash_bucket_stats(PlannerInfo *root,
 									   Node *hashkey, double nbuckets,
-- 
2.25.1