Allow setting statistics target for extended statistics

When building statistics, we need to decide how many rows to sample and
how accurate the resulting statistics should be. Until now, it was not
possible to explicitly define statistics target for extended statistics
objects, the value was always computed from the per-attribute targets
with a fallback to the system-wide default statistics target.

That's a bit inconvenient, as it ties together the statistics target set
for per-column and extended statistics. In some cases it may be useful
to require larger sample / higher accuracy for extended statics (or the
other way around), but with this approach that's not possible.

So this commit introduces a new command, allowing to specify statistics
target for individual extended statistics objects, overriding the value
derived from per-attribute targets (and the system default).

  ALTER STATISTICS stat_name SET STATISTICS target_value;

When determining statistics target for an extended statistics object we
first look at this explicitly set value. When this value is -1, we fall
back to the old formula, looking at the per-attribute targets first and
then the system default. This means the behavior is backwards compatible
with older PostgreSQL releases.

Author: Tomas Vondra
Discussion: https://postgr.es/m/20190618213357.vli3i23vpkset2xd@development
Reviewed-by: Kirk Jamison, Dean Rasheed

Author: tvondra

doc/src/sgml/ref/alter_statistics.sgml

@@ -26,6 +26,7 @@ PostgreSQL documentation
 ALTER STATISTICS <replaceable class="parameter">name</replaceable> OWNER TO { <replaceable class="parameter">new_owner</replaceable> | CURRENT_USER | SESSION_USER }
 ALTER STATISTICS <replaceable class="parameter">name</replaceable> RENAME TO <replaceable class="parameter">new_name</replaceable>
 ALTER STATISTICS <replaceable class="parameter">name</replaceable> SET SCHEMA <replaceable class="parameter">new_schema</replaceable>
+ALTER STATISTICS <replaceable class="parameter">name</replaceable> SET STATISTICS <replaceable class="parameter">new_target</replaceable>
 </synopsis>
  </refsynopsisdiv>
 
@@ -93,6 +94,22 @@ ALTER STATISTICS <replaceable class="parameter">name</replaceable> SET SCHEMA <r
       </listitem>
      </varlistentry>
 
+     <varlistentry>
+      <term><replaceable class="parameter">new_target</replaceable></term>
+      <listitem>
+       <para>
+        The statistic-gathering target for this statistics object for subsequent
+        <xref linkend="sql-analyze"/> operations.
+        The target can be set in the range 0 to 10000; alternatively, set it
+        to -1 to revert to using the system default statistics
+        target (<xref linkend="guc-default-statistics-target"/>).
+        For more information on the use of statistics by the
+        <productname>PostgreSQL</productname> query planner, refer to
+        <xref linkend="planner-stats"/>.
+       </para>
+      </listitem>
+     </varlistentry>
+
     </variablelist>
    </para>
   </refsect1>

src/backend/commands/analyze.c

@@ -307,7 +307,8 @@ do_analyze_rel(Relation onerel, VacuumParams *params,
 	VacAttrStats **vacattrstats;
 	AnlIndexData *indexdata;
 	int			targrows,
-				numrows;
+				numrows,
+				minrows;
 	double		totalrows,
 				totaldeadrows;
 	HeapTuple  *rows;
@@ -491,6 +492,16 @@ do_analyze_rel(Relation onerel, VacuumParams *params,
 		}
 	}
 
+	/*
+	 * Look at extended statistics objects too, as those may define custom
+	 * statistics target. So we may need to sample more rows and then build
+	 * the statistics with enough detail.
+	 */
+	minrows = ComputeExtStatisticsRows(onerel, attr_cnt, vacattrstats);
+
+	if (targrows < minrows)
+		targrows = minrows;
+
 	/*
 	 * Acquire the sample rows
 	 */

src/backend/commands/statscmds.c

@@ -14,13 +14,15 @@
  */
 #include "postgres.h"
 
+#include "access/heapam.h"
 #include "access/relation.h"
 #include "access/relscan.h"
 #include "access/table.h"
 #include "catalog/catalog.h"
 #include "catalog/dependency.h"
 #include "catalog/indexing.h"
 #include "catalog/namespace.h"
+#include "catalog/objectaccess.h"
 #include "catalog/pg_namespace.h"
 #include "catalog/pg_statistic_ext.h"
 #include "catalog/pg_statistic_ext_data.h"
@@ -29,6 +31,7 @@
 #include "miscadmin.h"
 #include "statistics/statistics.h"
 #include "utils/builtins.h"
+#include "utils/fmgroids.h"
 #include "utils/inval.h"
 #include "utils/memutils.h"
 #include "utils/rel.h"
@@ -336,6 +339,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 	values[Anum_pg_statistic_ext_stxrelid - 1] = ObjectIdGetDatum(relid);
 	values[Anum_pg_statistic_ext_stxname - 1] = NameGetDatum(&stxname);
 	values[Anum_pg_statistic_ext_stxnamespace - 1] = ObjectIdGetDatum(namespaceId);
+	values[Anum_pg_statistic_ext_stxstattarget - 1] = Int32GetDatum(-1);
 	values[Anum_pg_statistic_ext_stxowner - 1] = ObjectIdGetDatum(stxowner);
 	values[Anum_pg_statistic_ext_stxkeys - 1] = PointerGetDatum(stxkeys);
 	values[Anum_pg_statistic_ext_stxkind - 1] = PointerGetDatum(stxkind);
@@ -414,6 +418,110 @@ CreateStatistics(CreateStatsStmt *stmt)
 	return myself;
 }
 
+/*
+ *		ALTER STATISTICS
+ */
+ObjectAddress
+AlterStatistics(AlterStatsStmt *stmt)
+{
+	Relation	rel;
+	Oid			stxoid;
+	HeapTuple	oldtup;
+	HeapTuple	newtup;
+	Datum		repl_val[Natts_pg_statistic_ext];
+	bool		repl_null[Natts_pg_statistic_ext];
+	bool		repl_repl[Natts_pg_statistic_ext];
+	ObjectAddress	address;
+	int			newtarget = stmt->stxstattarget;
+
+	/* Limit statistics target to a sane range */
+	if (newtarget < -1)
+	{
+		ereport(ERROR,
+				(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+				 errmsg("statistics target %d is too low",
+						newtarget)));
+	}
+	else if (newtarget > 10000)
+	{
+		newtarget = 10000;
+		ereport(WARNING,
+				(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+				 errmsg("lowering statistics target to %d",
+						newtarget)));
+	}
+
+	/* lookup OID of the statistics object */
+	stxoid = get_statistics_object_oid(stmt->defnames, stmt->missing_ok);
+
+	/*
+	 * If we got here and the OID is not valid, it means the statistics
+	 * does not exist, but the command specified IF EXISTS. So report
+	 * this as a simple NOTICE and we're done.
+	 */
+	if (!OidIsValid(stxoid))
+	{
+		char	   *schemaname;
+		char	   *statname;
+
+		Assert(stmt->missing_ok);
+
+		DeconstructQualifiedName(stmt->defnames, &schemaname, &statname);
+
+		if (schemaname)
+			ereport(NOTICE,
+					(errmsg("statistics object \"%s.%s\" does not exist, skipping",
+							schemaname, statname)));
+		else
+			ereport(NOTICE,
+					(errmsg("statistics object \"%s\" does not exist, skipping",
+							statname)));
+
+		return InvalidObjectAddress;
+	}
+
+	/* Search pg_statistic_ext */
+	rel = table_open(StatisticExtRelationId, RowExclusiveLock);
+
+	oldtup = SearchSysCache1(STATEXTOID, ObjectIdGetDatum(stxoid));
+
+	/* Must be owner of the existing statistics object */
+	if (!pg_statistics_object_ownercheck(stxoid, GetUserId()))
+		aclcheck_error(ACLCHECK_NOT_OWNER, OBJECT_STATISTIC_EXT,
+					   NameListToString(stmt->defnames));
+
+	/* Build new tuple. */
+	memset(repl_val, 0, sizeof(repl_val));
+	memset(repl_null, false, sizeof(repl_null));
+	memset(repl_repl, false, sizeof(repl_repl));
+
+	/* replace the stxstattarget column */
+	repl_repl[Anum_pg_statistic_ext_stxstattarget - 1] = true;
+	repl_val[Anum_pg_statistic_ext_stxstattarget - 1] = Int32GetDatum(newtarget);
+
+	newtup = heap_modify_tuple(oldtup, RelationGetDescr(rel),
+							   repl_val, repl_null, repl_repl);
+
+	/* Update system catalog. */
+	CatalogTupleUpdate(rel, &newtup->t_self, newtup);
+
+	InvokeObjectPostAlterHook(StatisticExtRelationId, stxoid, 0);
+
+	ObjectAddressSet(address, StatisticExtRelationId, stxoid);
+
+	/*
+	 * NOTE: because we only support altering the statistics target, not the
+	 * other fields, there is no need to update dependencies.
+	 */
+
+	heap_freetuple(newtup);
+	ReleaseSysCache(oldtup);
+
+	table_close(rel, RowExclusiveLock);
+
+	return address;
+}
+
 /*
  * Guts of statistics object deletion.
  */

src/backend/nodes/copyfuncs.c

@@ -3497,6 +3497,18 @@ _copyCreateStatsStmt(const CreateStatsStmt *from)
 	return newnode;
 }
 
+static AlterStatsStmt *
+_copyAlterStatsStmt(const AlterStatsStmt *from)
+{
+	AlterStatsStmt *newnode = makeNode(AlterStatsStmt);
+
+	COPY_NODE_FIELD(defnames);
+	COPY_SCALAR_FIELD(stxstattarget);
+	COPY_SCALAR_FIELD(missing_ok);
+
+	return newnode;
+}
+
 static CreateFunctionStmt *
 _copyCreateFunctionStmt(const CreateFunctionStmt *from)
 {
@@ -5211,6 +5223,9 @@ copyObjectImpl(const void *from)
 		case T_CreateStatsStmt:
 			retval = _copyCreateStatsStmt(from);
 			break;
+		case T_AlterStatsStmt:
+			retval = _copyAlterStatsStmt(from);
+			break;
 		case T_CreateFunctionStmt:
 			retval = _copyCreateFunctionStmt(from);
 			break;

src/backend/nodes/equalfuncs.c

@@ -1365,6 +1365,16 @@ _equalCreateStatsStmt(const CreateStatsStmt *a, const CreateStatsStmt *b)
 	return true;
 }
 
+static bool
+_equalAlterStatsStmt(const AlterStatsStmt *a, const AlterStatsStmt *b)
+{
+	COMPARE_NODE_FIELD(defnames);
+	COMPARE_SCALAR_FIELD(stxstattarget);
+	COMPARE_SCALAR_FIELD(missing_ok);
+
+	return true;
+}
+
 static bool
 _equalCreateFunctionStmt(const CreateFunctionStmt *a, const CreateFunctionStmt *b)
 {
@@ -3309,6 +3319,9 @@ equal(const void *a, const void *b)
 		case T_CreateStatsStmt:
 			retval = _equalCreateStatsStmt(a, b);
 			break;
+		case T_AlterStatsStmt:
+			retval = _equalAlterStatsStmt(a, b);
+			break;
 		case T_CreateFunctionStmt:
 			retval = _equalCreateFunctionStmt(a, b);
 			break;

src/backend/nodes/outfuncs.c

@@ -2668,6 +2668,16 @@ _outCreateStatsStmt(StringInfo str, const CreateStatsStmt *node)
 	WRITE_BOOL_FIELD(if_not_exists);
 }
 
+static void
+_outAlterStatsStmt(StringInfo str, const AlterStatsStmt *node)
+{
+	WRITE_NODE_TYPE("ALTERSTATSSTMT");
+
+	WRITE_NODE_FIELD(defnames);
+	WRITE_INT_FIELD(stxstattarget);
+	WRITE_BOOL_FIELD(missing_ok);
+}
+
 static void
 _outNotifyStmt(StringInfo str, const NotifyStmt *node)
 {
@@ -4130,6 +4140,9 @@ outNode(StringInfo str, const void *obj)
 			case T_CreateStatsStmt:
 				_outCreateStatsStmt(str, obj);
 				break;
+			case T_AlterStatsStmt:
+				_outAlterStatsStmt(str, obj);
+				break;
 			case T_NotifyStmt:
 				_outNotifyStmt(str, obj);
 				break;

src/backend/parser/gram.y

@@ -252,7 +252,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 		AlterOperatorStmt AlterSeqStmt AlterSystemStmt AlterTableStmt
 		AlterTblSpcStmt AlterExtensionStmt AlterExtensionContentsStmt AlterForeignTableStmt
 		AlterCompositeTypeStmt AlterUserMappingStmt
-		AlterRoleStmt AlterRoleSetStmt AlterPolicyStmt
+		AlterRoleStmt AlterRoleSetStmt AlterPolicyStmt AlterStatsStmt
 		AlterDefaultPrivilegesStmt DefACLAction
 		AnalyzeStmt CallStmt ClosePortalStmt ClusterStmt CommentStmt
 		ConstraintsSetStmt CopyStmt CreateAsStmt CreateCastStmt
@@ -852,6 +852,7 @@ stmt :
 			| AlterRoleSetStmt
 			| AlterRoleStmt
 			| AlterSubscriptionStmt
+			| AlterStatsStmt
 			| AlterTSConfigurationStmt
 			| AlterTSDictionaryStmt
 			| AlterUserMappingStmt
@@ -3984,6 +3985,34 @@ CreateStatsStmt:
 				}
 			;
 
+
+/*****************************************************************************
+ *
+ *		QUERY :
+ *				ALTER STATISTICS [IF EXISTS] stats_name
+ *					SET STATISTICS  <SignedIconst>
+ *
+ *****************************************************************************/
+
+AlterStatsStmt:
+			ALTER STATISTICS any_name SET STATISTICS SignedIconst
+				{
+					AlterStatsStmt *n = makeNode(AlterStatsStmt);
+					n->defnames = $3;
+					n->missing_ok = false;
+					n->stxstattarget = $6;
+					$$ = (Node *)n;
+				}
+			| ALTER STATISTICS IF_P EXISTS any_name SET STATISTICS SignedIconst
+				{
+					AlterStatsStmt *n = makeNode(AlterStatsStmt);
+					n->defnames = $5;
+					n->missing_ok = true;
+					n->stxstattarget = $8;
+					$$ = (Node *)n;
+				}
+			;
+
 /*****************************************************************************
  *
  *		QUERY :