Reduce "Var IS [NOT] NULL" quals during constant folding

In commit b262ad440, we introduced an optimization that reduces an IS
[NOT] NULL qual on a NOT NULL column to constant true or constant
false, provided we can prove that the input expression of the NullTest
is not nullable by any outer joins or grouping sets.  This deduction
happens quite late in the planner, during the distribution of quals to
rels in query_planner.  However, this approach has some drawbacks: we
can't perform any further folding with the constant, and it turns out
to be prone to bugs.

Ideally, this deduction should happen during constant folding.
However, the per-relation information about which columns are defined
as NOT NULL is not available at that point.  This information is
currently collected from catalogs when building RelOptInfos for base
or "other" relations.

This patch moves the collection of NOT NULL attribute information for
relations before pull_up_sublinks, storing it in a hash table keyed by
relation OID.  It then uses this information to perform the NullTest
deduction for Vars during constant folding.  This also makes it
possible to leverage this information to pull up NOT IN subqueries.

Note that this patch does not get rid of restriction_is_always_true
and restriction_is_always_false.  Removing them would prevent us from
reducing some IS [NOT] NULL quals that we were previously able to
reduce, because (a) the self-join elimination may introduce new IS NOT
NULL quals after constant folding, and (b) if some outer joins are
converted to inner joins, previously irreducible NullTest quals may
become reducible.

Author: Richard Guo <guofenglinux@gmail.com>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CAMbWs4-bFJ1At4btk5wqbezdu8PLtQ3zv-aiaY3ry9Ymm=jgFQ@mail.gmail.com

3 files changed, 22 insertions, 34 deletions

diff --git a/src/backend/optimizer/plan/initsplan.c b/src/backend/optimizer/plan/initsplan.c
index 01804b085b3..3e3fec89252 100644
--- a/src/backend/optimizer/plan/initsplan.c
+++ b/src/backend/optimizer/plan/initsplan.c
@@ -3048,36 +3048,16 @@ add_base_clause_to_rel(PlannerInfo *root, Index relid,
  * expr_is_nonnullable
  *	  Check to see if the Expr cannot be NULL
  *
- * If the Expr is a simple Var that is defined NOT NULL and meanwhile is not
- * nulled by any outer joins, then we can know that it cannot be NULL.
+ * Currently we only support simple Vars.
  */
 static bool
 expr_is_nonnullable(PlannerInfo *root, Expr *expr)
 {
-	RelOptInfo *rel;
-	Var		   *var;
-
 	/* For now only check simple Vars */
 	if (!IsA(expr, Var))
 		return false;
 
-	var = (Var *) expr;
-
-	/* could the Var be nulled by any outer joins? */
-	if (!bms_is_empty(var->varnullingrels))
-		return false;
-
-	/* system columns cannot be NULL */
-	if (var->varattno < 0)
-		return true;
-
-	/* is the column defined NOT NULL? */
-	rel = find_base_rel(root, var->varno);
-	if (var->varattno > 0 &&
-		bms_is_member(var->varattno, rel->notnullattnums))
-		return true;
-
-	return false;
+	return var_is_nonnullable(root, (Var *) expr, true);
 }
 
 /*
diff --git a/src/backend/optimizer/plan/planner.c b/src/backend/optimizer/plan/planner.c
index fc13d921d0c..c989e72cac5 100644
--- a/src/backend/optimizer/plan/planner.c
+++ b/src/backend/optimizer/plan/planner.c
@@ -342,6 +342,7 @@ standard_planner(Query *parse, const char *query_string, int cursorOptions,
 	glob->transientPlan = false;
 	glob->dependsOnRole = false;
 	glob->partition_directory = NULL;
+	glob->rel_notnullatts_hash = NULL;
 
 	/*
 	 * Assess whether it's feasible to use parallel mode for this query. We
@@ -723,11 +724,12 @@ subquery_planner(PlannerGlobal *glob, Query *parse, PlannerInfo *parent_root,
 	/*
 	 * Scan the rangetable for relation RTEs and retrieve the necessary
 	 * catalog information for each relation.  Using this information, clear
-	 * the inh flag for any relation that has no children, and expand virtual
-	 * generated columns for any relation that contains them.  Note that this
-	 * step does not descend into sublinks and subqueries; if we pull up any
-	 * sublinks or subqueries below, their relation RTEs are processed just
-	 * before pulling them up.
+	 * the inh flag for any relation that has no children, collect not-null
+	 * attribute numbers for any relation that has column not-null
+	 * constraints, and expand virtual generated columns for any relation that
+	 * contains them.  Note that this step does not descend into sublinks and
+	 * subqueries; if we pull up any sublinks or subqueries below, their
+	 * relation RTEs are processed just before pulling them up.
 	 */
 	parse = root->parse = preprocess_relation_rtes(root);
 
diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c
index 4bdca59df64..d71ed958e31 100644
--- a/src/backend/optimizer/plan/subselect.c
+++ b/src/backend/optimizer/plan/subselect.c
@@ -1519,8 +1519,10 @@ convert_EXISTS_sublink_to_join(PlannerInfo *root, SubLink *sublink,
 	/*
 	 * Scan the rangetable for relation RTEs and retrieve the necessary
 	 * catalog information for each relation.  Using this information, clear
-	 * the inh flag for any relation that has no children, and expand virtual
-	 * generated columns for any relation that contains them.
+	 * the inh flag for any relation that has no children, collect not-null
+	 * attribute numbers for any relation that has column not-null
+	 * constraints, and expand virtual generated columns for any relation that
+	 * contains them.
 	 *
 	 * Note: we construct up an entirely dummy PlannerInfo for use here.  This
 	 * is fine because only the "glob" and "parse" links will be used in this
@@ -1760,6 +1762,7 @@ convert_EXISTS_to_ANY(PlannerInfo *root, Query *subselect,
 					  Node **testexpr, List **paramIds)
 {
 	Node	   *whereClause;
+	PlannerInfo subroot;
 	List	   *leftargs,
 			   *rightargs,
 			   *opids,
@@ -1819,12 +1822,15 @@ convert_EXISTS_to_ANY(PlannerInfo *root, Query *subselect,
 	 * parent aliases were flattened already, and we're not going to pull any
 	 * child Vars (of any description) into the parent.
 	 *
-	 * Note: passing the parent's root to eval_const_expressions is
-	 * technically wrong, but we can get away with it since only the
-	 * boundParams (if any) are used, and those would be the same in a
-	 * subroot.
+	 * Note: we construct up an entirely dummy PlannerInfo to pass to
+	 * eval_const_expressions.  This is fine because only the "glob" and
+	 * "parse" links are used by eval_const_expressions.
 	 */
-	whereClause = eval_const_expressions(root, whereClause);
+	MemSet(&subroot, 0, sizeof(subroot));
+	subroot.type = T_PlannerInfo;
+	subroot.glob = root->glob;
+	subroot.parse = subselect;
+	whereClause = eval_const_expressions(&subroot, whereClause);
 	whereClause = (Node *) canonicalize_qual((Expr *) whereClause, false);
 	whereClause = (Node *) make_ands_implicit((Expr *) whereClause);