Simple constraint exclusion. For now, only child tables of inheritance

scans are candidates for exclusion; this should be fixed eventually.
Simon Riggs, with some help from Tom Lane.

Author: tglsfdc

doc/src/sgml/runtime.sgml

@@ -1,5 +1,5 @@
 <!--
-$PostgreSQL: pgsql/doc/src/sgml/runtime.sgml,v 1.338 2005/07/14 05:13:38 tgl Exp $
+$PostgreSQL: pgsql/doc/src/sgml/runtime.sgml,v 1.339 2005/07/23 21:05:45 tgl Exp $
 -->
 
 <chapter Id="runtime">
@@ -2278,6 +2278,56 @@ archive_command = 'copy "%p" /mnt/server/archivedir/"%f"'  # Windows
       </listitem>
      </varlistentry>
 
+     <varlistentry id="guc-enable-constraint-exclusion" xreflabel="enable_constraint_exclusion">
+      <term><varname>enable_constraint_exclusion</varname> (<type>boolean</type>)</term>
+      <indexterm>
+       <primary>constraint exclusion</primary>
+      </indexterm>
+      <indexterm>
+       <primary><varname>enable_constraint_exclusion</> configuration parameter</primary>
+      </indexterm>
+      <listitem>
+       <para>
+        Enables or disables the query planner's use of table constraints.
+        The default is <literal>off</>.
+       </para>
+
+       <para>
+        When this parameter is <literal>on</>, the planner compares query
+        conditions to table CHECK constraints, and omits scanning tables
+        for which the conditions contradict the constraints.  (Presently
+        this is done only for child tables of inheritance scans.)  For
+        example:
+
+<programlisting>
+CREATE TABLE parent(key integer, ...);
+CREATE TABLE child1000(check (key between 1000 and 1999)) INHERITS(parent);
+CREATE TABLE child2000(check (key between 2000 and 2999)) INHERITS(parent);
+...
+SELECT * FROM parent WHERE key = 2400;
+</programlisting>
+
+        With constraint exclusion enabled, this SELECT will not scan
+        <structname>child1000</> at all.  This can improve performance when
+        inheritance is used to build partitioned tables.
+       </para>
+
+       <para>
+        Currently, <varname>enable_constraint_exclusion</> defaults to
+        <literal>off</>, because it creates a risk of wrong answers when
+        query plans are cached: if a table constraint is changed or dropped,
+        the previously generated plan may now be wrong, and there is no
+        built-in mechanism to force re-planning.  (This deficiency will
+        probably be addressed in a future
+        <productname>PostgreSQL</productname> release.)  Another reason
+        for keeping it off is that the constraint checks are relatively
+        expensive to make, and in many circumstances will yield no savings.
+        It is recommended to turn this on only if you are actually using
+        partitioned tables designed to take advantage of the feature.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry id="guc-from-collapse-limit" xreflabel="from_collapse_limit">
       <term><varname>from_collapse_limit</varname> (<type>integer</type>)</term>
       <indexterm>

src/backend/optimizer/path/allpaths.c

@@ -8,13 +8,14 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/path/allpaths.c,v 1.134 2005/06/10 03:32:21 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/path/allpaths.c,v 1.135 2005/07/23 21:05:46 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
 
 #include "postgres.h"
 
+#include "nodes/makefuncs.h"
 #ifdef OPTIMIZER_DEBUG
 #include "nodes/print.h"
 #endif
@@ -25,6 +26,7 @@
 #include "optimizer/paths.h"
 #include "optimizer/plancat.h"
 #include "optimizer/planner.h"
+#include "optimizer/predtest.h"
 #include "optimizer/prep.h"
 #include "optimizer/var.h"
 #include "parser/parsetree.h"
@@ -34,6 +36,7 @@
 
 
 /* These parameters are set by GUC */
+bool		enable_constraint_exclusion = false;
 bool		enable_geqo = false;	/* just in case GUC doesn't set it */
 int			geqo_threshold;
 
@@ -311,7 +314,37 @@ set_inherited_rel_pathlist(PlannerInfo *root, RelOptInfo *rel,
 								   childOID);
 
 		/*
-		 * Now compute child access paths, and save the cheapest.
+		 * If we can prove we don't need to scan this child via constraint
+		 * exclusion, just ignore it.  (We have to have converted the
+		 * baserestrictinfo Vars before we can make the test.)
+		 */
+		if (enable_constraint_exclusion)
+		{
+			List   *constraint_pred;
+
+			constraint_pred = get_relation_constraints(childOID, childrel);
+			/*
+			 * We do not currently enforce that CHECK constraints contain
+			 * only immutable functions, so it's necessary to check here.
+			 * We daren't draw conclusions from plan-time evaluation of
+			 * non-immutable functions.
+			 */
+			if (!contain_mutable_functions((Node *) constraint_pred))
+			{
+				/*
+				 * The constraints are effectively ANDed together, so we can
+				 * just try to refute the entire collection at once.  This may
+				 * allow us to make proofs that would fail if we took them
+				 * individually.
+				 */
+				if (predicate_refuted_by(constraint_pred,
+										 childrel->baserestrictinfo))
+					continue;
+			}
+		}
+
+		/*
+		 * Compute the child's access paths, and save the cheapest.
 		 */
 		set_plain_rel_pathlist(root, childrel, childrte);
 
@@ -345,7 +378,8 @@ set_inherited_rel_pathlist(PlannerInfo *root, RelOptInfo *rel,
 
 	/*
 	 * Finally, build Append path and install it as the only access path
-	 * for the parent rel.
+	 * for the parent rel.  (Note: this is correct even if we have zero
+	 * or one live subpath due to constraint exclusion.)
 	 */
 	add_path(rel, (Path *) create_append_path(rel, subpaths));
 

src/backend/optimizer/plan/createplan.c

@@ -10,7 +10,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/plan/createplan.c,v 1.194 2005/07/15 22:02:51 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/plan/createplan.c,v 1.195 2005/07/23 21:05:46 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -40,7 +40,7 @@ static List *build_relation_tlist(RelOptInfo *rel);
 static bool use_physical_tlist(RelOptInfo *rel);
 static void disuse_physical_tlist(Plan *plan, Path *path);
 static Join *create_join_plan(PlannerInfo *root, JoinPath *best_path);
-static Append *create_append_plan(PlannerInfo *root, AppendPath *best_path);
+static Plan *create_append_plan(PlannerInfo *root, AppendPath *best_path);
 static Result *create_result_plan(PlannerInfo *root, ResultPath *best_path);
 static Material *create_material_plan(PlannerInfo *root, MaterialPath *best_path);
 static Plan *create_unique_plan(PlannerInfo *root, UniquePath *best_path);
@@ -435,14 +435,33 @@ create_join_plan(PlannerInfo *root, JoinPath *best_path)
  *
  *	  Returns a Plan node.
  */
-static Append *
+static Plan *
 create_append_plan(PlannerInfo *root, AppendPath *best_path)
 {
 	Append	   *plan;
 	List	   *tlist = build_relation_tlist(best_path->path.parent);
 	List	   *subplans = NIL;
 	ListCell   *subpaths;
 
+	/*
+	 * It is possible for the subplans list to contain only one entry,
+	 * or even no entries.  Handle these cases specially.
+	 *
+	 * XXX ideally, if there's just one entry, we'd not bother to generate
+	 * an Append node but just return the single child.  At the moment this
+	 * does not work because the varno of the child scan plan won't match
+	 * the parent-rel Vars it'll be asked to emit.
+	 */
+	if (best_path->subpaths == NIL)
+	{
+		/* Generate a Result plan with constant-FALSE gating qual */
+		return (Plan *) make_result(tlist,
+									(Node *) list_make1(makeBoolConst(false,
+																	  false)),
+									NULL);
+	}
+
+	/* Normal case with multiple subpaths */
 	foreach(subpaths, best_path->subpaths)
 	{
 		Path	   *subpath = (Path *) lfirst(subpaths);
@@ -452,7 +471,7 @@ create_append_plan(PlannerInfo *root, AppendPath *best_path)
 
 	plan = make_append(subplans, false, tlist);
 
-	return plan;
+	return (Plan *) plan;
 }
 
 /*

src/backend/optimizer/plan/planagg.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/plan/planagg.c,v 1.5 2005/06/05 22:32:56 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/plan/planagg.c,v 1.6 2005/07/23 21:05:46 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -184,7 +184,6 @@ optimize_minmax_aggregates(PlannerInfo *root, List *tlist, Path *best_path)
 	 */
 	if (IsA(best_path, ResultPath))
 	{
-		Assert(((ResultPath *) best_path)->subpath != NULL);
 		constant_quals = ((ResultPath *) best_path)->constantqual;
 		/* no need to do this more than once: */
 		constant_quals = order_qual_clauses(root, constant_quals);

src/backend/optimizer/util/plancat.c

@@ -9,7 +9,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/util/plancat.c,v 1.112 2005/06/13 23:14:48 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/util/plancat.c,v 1.113 2005/07/23 21:05:47 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -25,6 +25,7 @@
 #include "nodes/makefuncs.h"
 #include "optimizer/clauses.h"
 #include "optimizer/plancat.h"
+#include "optimizer/prep.h"
 #include "optimizer/tlist.h"
 #include "parser/parsetree.h"
 #include "parser/parse_expr.h"
@@ -359,6 +360,85 @@ estimate_rel_size(Relation rel, int32 *attr_widths,
 	}
 }
 
+
+/*
+ * get_relation_constraints
+ *
+ * Retrieve the CHECK constraint expressions of the given relation.
+ *
+ * Returns a List (possibly empty) of constraint expressions.  Each one
+ * has been canonicalized, and its Vars are changed to have the varno
+ * indicated by rel->relid.  This allows the expressions to be easily
+ * compared to expressions taken from WHERE.
+ *
+ * Note: at present this is invoked at most once per relation per planner
+ * run, and in many cases it won't be invoked at all, so there seems no
+ * point in caching the data in RelOptInfo.
+ */
+List *
+get_relation_constraints(Oid relationObjectId, RelOptInfo *rel)
+{
+	List	   *result = NIL;
+	Index		varno = rel->relid;
+	Relation	relation;
+	TupleConstr *constr;
+
+	/*
+	 * We assume the relation has already been safely locked.
+	 */
+	relation = heap_open(relationObjectId, NoLock);
+
+	constr = relation->rd_att->constr;
+	if (constr != NULL)
+	{
+		int		num_check = constr->num_check;
+		int		i;
+
+		for (i = 0; i < num_check; i++)
+		{
+			Node	*cexpr;
+
+			cexpr = stringToNode(constr->check[i].ccbin);
+
+			/*
+			 * Run each expression through const-simplification and
+			 * canonicalization.  This is not just an optimization, but is
+			 * necessary, because we will be comparing it to
+			 * similarly-processed qual clauses, and may fail to detect valid
+			 * matches without this.  This must match the processing done to
+			 * qual clauses in preprocess_expression()!  (We can skip the
+			 * stuff involving subqueries, however, since we don't allow any
+			 * in check constraints.)
+			 */
+			cexpr = eval_const_expressions(cexpr);
+
+			cexpr = (Node *) canonicalize_qual((Expr *) cexpr);
+
+			/*
+			 * Also mark any coercion format fields as "don't care", so that
+			 * we can match to both explicit and implicit coercions.
+			 */
+			set_coercionform_dontcare(cexpr);
+
+			/* Fix Vars to have the desired varno */
+			if (varno != 1)
+				ChangeVarNodes(cexpr, 1, varno, 0);
+
+			/*
+			 * Finally, convert to implicit-AND format (that is, a List)
+			 * and append the resulting item(s) to our output list.
+			 */
+			result = list_concat(result,
+								 make_ands_implicit((Expr *) cexpr));
+		}
+	}
+
+	heap_close(relation, NoLock);
+
+	return result;
+}
+
+
 /*
  * build_physical_tlist
  *