Make attstattarget nullable

This changes the pg_attribute field attstattarget into a nullable
field in the variable-length part of the row.  If no value is set by
the user for attstattarget, it is now null instead of previously -1.
This saves space in pg_attribute and tuple descriptors for most
practical scenarios.  (ATTRIBUTE_FIXED_PART_SIZE is reduced from 108
to 104.)  Also, null is the semantically more correct value.

The ANALYZE code internally continues to represent the default
statistics target by -1, so that that code can avoid having to deal
with null values.  But that is now contained to the ANALYZE code.
Only the DDL code deals with attstattarget possibly null.

For system columns, the field is now always null.  The ANALYZE code
skips system columns anyway.

To set a column's statistics target to the default value, the new
command form ALTER TABLE ... SET STATISTICS DEFAULT can be used.  (SET
STATISTICS -1 still works.)

Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
Discussion: https://www.postgresql.org/message-id/flat/4da8d211-d54d-44b9-9847-f2a9f1184c76@eisentraut.org

src/backend/access/common/tupdesc.c

@@ -453,8 +453,6 @@ equalTupleDescs(TupleDesc tupdesc1, TupleDesc tupdesc2)
 			return false;
 		if (attr1->atttypid != attr2->atttypid)
 			return false;
-		if (attr1->attstattarget != attr2->attstattarget)
-			return false;
 		if (attr1->attlen != attr2->attlen)
 			return false;
 		if (attr1->attndims != attr2->attndims)
@@ -639,7 +637,6 @@ TupleDescInitEntry(TupleDesc desc,
 	else if (attributeName != NameStr(att->attname))
 		namestrcpy(&(att->attname), attributeName);
 
-	att->attstattarget = -1;
 	att->attcacheoff = -1;
 	att->atttypmod = typmod;
 
@@ -702,7 +699,6 @@ TupleDescInitBuiltinEntry(TupleDesc desc,
 	Assert(attributeName != NULL);
 	namestrcpy(&(att->attname), attributeName);
 
-	att->attstattarget = -1;
 	att->attcacheoff = -1;
 	att->atttypmod = typmod;
 

src/backend/bootstrap/bootstrap.c

@@ -552,7 +552,6 @@ DefineAttr(char *name, char *type, int attnum, int nullness)
 	if (OidIsValid(attrtypes[attnum]->attcollation))
 		attrtypes[attnum]->attcollation = C_COLLATION_OID;
 
-	attrtypes[attnum]->attstattarget = -1;
 	attrtypes[attnum]->attcacheoff = -1;
 	attrtypes[attnum]->atttypmod = -1;
 	attrtypes[attnum]->attislocal = true;

src/backend/catalog/genbki.pl

@@ -840,7 +840,6 @@ sub gen_pg_attribute
 				my %row;
 				$row{attnum} = $attnum;
 				$row{attrelid} = $table->{relation_oid};
-				$row{attstattarget} = '0';
 
 				morph_row_for_pgattr(\%row, $schema, $attr, 1);
 				print_bki_insert(\%row, $schema);

src/backend/catalog/heap.c

@@ -749,14 +749,16 @@ InsertPgAttributeTuples(Relation pg_attribute_rel,
 		slot[slotCount]->tts_values[Anum_pg_attribute_attisdropped - 1] = BoolGetDatum(attrs->attisdropped);
 		slot[slotCount]->tts_values[Anum_pg_attribute_attislocal - 1] = BoolGetDatum(attrs->attislocal);
 		slot[slotCount]->tts_values[Anum_pg_attribute_attinhcount - 1] = Int16GetDatum(attrs->attinhcount);
-		slot[slotCount]->tts_values[Anum_pg_attribute_attstattarget - 1] = Int16GetDatum(attrs->attstattarget);
 		slot[slotCount]->tts_values[Anum_pg_attribute_attcollation - 1] = ObjectIdGetDatum(attrs->attcollation);
 		if (attoptions && attoptions[natts] != (Datum) 0)
 			slot[slotCount]->tts_values[Anum_pg_attribute_attoptions - 1] = attoptions[natts];
 		else
 			slot[slotCount]->tts_isnull[Anum_pg_attribute_attoptions - 1] = true;
 
-		/* start out with empty permissions and empty options */
+		/*
+		 * The remaining fields are not set for new columns.
+		 */
+		slot[slotCount]->tts_isnull[Anum_pg_attribute_attstattarget - 1] = true;
 		slot[slotCount]->tts_isnull[Anum_pg_attribute_attacl - 1] = true;
 		slot[slotCount]->tts_isnull[Anum_pg_attribute_attfdwoptions - 1] = true;
 		slot[slotCount]->tts_isnull[Anum_pg_attribute_attmissingval - 1] = true;
@@ -818,9 +820,6 @@ AddNewAttributeTuples(Oid new_rel_oid,
 
 	indstate = CatalogOpenIndexes(rel);
 
-	/* set stats detail level to a sane default */
-	for (int i = 0; i < natts; i++)
-		tupdesc->attrs[i].attstattarget = -1;
 	InsertPgAttributeTuples(rel, tupdesc, new_rel_oid, NULL, indstate);
 
 	/* add dependencies on their datatypes and collations */
@@ -1685,9 +1684,6 @@ RemoveAttributeById(Oid relid, AttrNumber attnum)
 	/* Remove any not-null constraint the column may have */
 	attStruct->attnotnull = false;
 
-	/* We don't want to keep stats for it anymore */
-	attStruct->attstattarget = 0;
-
 	/* Unset this so no one tries to look up the generation expression */
 	attStruct->attgenerated = '\0';
 
@@ -1704,9 +1700,11 @@ RemoveAttributeById(Oid relid, AttrNumber attnum)
 	replacesAtt[Anum_pg_attribute_attmissingval - 1] = true;
 
 	/*
-	 * Clear the other variable-length fields.  This saves some space in
-	 * pg_attribute and removes no longer useful information.
+	 * Clear the other nullable fields.  This saves some space in pg_attribute
+	 * and removes no longer useful information.
 	 */
+	nullsAtt[Anum_pg_attribute_attstattarget - 1] = true;
+	replacesAtt[Anum_pg_attribute_attstattarget - 1] = true;
 	nullsAtt[Anum_pg_attribute_attacl - 1] = true;
 	replacesAtt[Anum_pg_attribute_attacl - 1] = true;
 	nullsAtt[Anum_pg_attribute_attoptions - 1] = true;

src/backend/catalog/index.c

@@ -325,7 +325,6 @@ ConstructTupleDescriptor(Relation heapRelation,
 
 		MemSet(to, 0, ATTRIBUTE_FIXED_PART_SIZE);
 		to->attnum = i + 1;
-		to->attstattarget = -1;
 		to->attcacheoff = -1;
 		to->attislocal = true;
 		to->attcollation = (i < numkeyatts) ? collationIds[i] : InvalidOid;
@@ -1780,10 +1779,12 @@ index_concurrently_swap(Oid newIndexId, Oid oldIndexId, const char *oldName)
 		while (HeapTupleIsValid((attrTuple = systable_getnext(scan))))
 		{
 			Form_pg_attribute att = (Form_pg_attribute) GETSTRUCT(attrTuple);
+			HeapTuple	tp;
+			Datum		dat;
+			bool		isnull;
 			Datum		repl_val[Natts_pg_attribute];
 			bool		repl_null[Natts_pg_attribute];
 			bool		repl_repl[Natts_pg_attribute];
-			int			attstattarget;
 			HeapTuple	newTuple;
 
 			/* Ignore dropped columns */
@@ -1793,10 +1794,18 @@ index_concurrently_swap(Oid newIndexId, Oid oldIndexId, const char *oldName)
 			/*
 			 * Get attstattarget from the old index and refresh the new value.
 			 */
-			attstattarget = get_attstattarget(oldIndexId, att->attnum);
+			tp = SearchSysCache2(ATTNUM, ObjectIdGetDatum(oldIndexId), Int16GetDatum(att->attnum));
+			if (!HeapTupleIsValid(tp))
+				elog(ERROR, "cache lookup failed for attribute %d of relation %u",
+					 att->attnum, oldIndexId);
+			dat = SysCacheGetAttr(ATTNUM, tp, Anum_pg_attribute_attstattarget, &isnull);
+			ReleaseSysCache(tp);
 
-			/* no need for a refresh if both match */
-			if (attstattarget == att->attstattarget)
+			/*
+			 * No need for a refresh if old index value is null.  (All new
+			 * index values are null at this point.)
+			 */
+			if (isnull)
 				continue;
 
 			memset(repl_val, 0, sizeof(repl_val));
@@ -1804,7 +1813,7 @@ index_concurrently_swap(Oid newIndexId, Oid oldIndexId, const char *oldName)
 			memset(repl_repl, false, sizeof(repl_repl));
 
 			repl_repl[Anum_pg_attribute_attstattarget - 1] = true;
-			repl_val[Anum_pg_attribute_attstattarget - 1] = Int16GetDatum(attstattarget);
+			repl_val[Anum_pg_attribute_attstattarget - 1] = dat;
 
 			newTuple = heap_modify_tuple(attrTuple,
 										 RelationGetDescr(pg_attribute),

src/backend/commands/analyze.c

@@ -1004,6 +1004,10 @@ static VacAttrStats *
 examine_attribute(Relation onerel, int attnum, Node *index_expr)
 {
 	Form_pg_attribute attr = TupleDescAttr(onerel->rd_att, attnum - 1);
+	int			attstattarget;
+	HeapTuple	atttuple;
+	Datum		dat;
+	bool		isnull;
 	HeapTuple	typtuple;
 	VacAttrStats *stats;
 	int			i;
@@ -1013,15 +1017,28 @@ examine_attribute(Relation onerel, int attnum, Node *index_expr)
 	if (attr->attisdropped)
 		return NULL;
 
+	/*
+	 * Get attstattarget value.  Set to -1 if null.  (Analyze functions expect
+	 * -1 to mean use default_statistics_target; see for example
+	 * std_typanalyze.)
+	 */
+	atttuple = SearchSysCache2(ATTNUM, ObjectIdGetDatum(RelationGetRelid(onerel)), Int16GetDatum(attnum));
+	if (!HeapTupleIsValid(atttuple))
+		elog(ERROR, "cache lookup failed for attribute %d of relation %u",
+			 attnum, RelationGetRelid(onerel));
+	dat = SysCacheGetAttr(ATTNUM, atttuple, Anum_pg_attribute_attstattarget, &isnull);
+	attstattarget = isnull ? -1 : DatumGetInt16(dat);
+	ReleaseSysCache(atttuple);
+
 	/* Don't analyze column if user has specified not to */
-	if (attr->attstattarget == 0)
+	if (attstattarget == 0)
 		return NULL;
 
 	/*
 	 * Create the VacAttrStats struct.
 	 */
 	stats = (VacAttrStats *) palloc0(sizeof(VacAttrStats));
-	stats->attstattarget = attr->attstattarget;
+	stats->attstattarget = attstattarget;
 
 	/*
 	 * When analyzing an expression index, believe the expression tree's type

src/backend/commands/tablecmds.c

@@ -8543,10 +8543,14 @@ ATExecSetStatistics(Relation rel, const char *colName, int16 colNum, Node *newVa
 {
 	int			newtarget;
 	Relation	attrelation;
-	HeapTuple	tuple;
+	HeapTuple	tuple,
+				newtuple;
 	Form_pg_attribute attrtuple;
 	AttrNumber	attnum;
 	ObjectAddress address;
+	Datum		repl_val[Natts_pg_attribute];
+	bool		repl_null[Natts_pg_attribute];
+	bool		repl_repl[Natts_pg_attribute];
 
 	/*
 	 * We allow referencing columns by numbers only for indexes, since table
@@ -8559,8 +8563,18 @@ ATExecSetStatistics(Relation rel, const char *colName, int16 colNum, Node *newVa
 				(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
 				 errmsg("cannot refer to non-index column by number")));
 
-	Assert(IsA(newValue, Integer));
-	newtarget = intVal(newValue);
+	if (newValue)
+	{
+		Assert(IsA(newValue, Integer));
+		newtarget = intVal(newValue);
+	}
+	else
+	{
+		/*
+		 * -1 was used in previous versions to represent the default setting
+		 */
+		newtarget = -1;
+	}
 
 	/*
 	 * Limit target to a sane range
@@ -8585,7 +8599,7 @@ ATExecSetStatistics(Relation rel, const char *colName, int16 colNum, Node *newVa
 
 	if (colName)
 	{
-		tuple = SearchSysCacheCopyAttName(RelationGetRelid(rel), colName);
+		tuple = SearchSysCacheAttName(RelationGetRelid(rel), colName);
 
 		if (!HeapTupleIsValid(tuple))
 			ereport(ERROR,
@@ -8595,7 +8609,7 @@ ATExecSetStatistics(Relation rel, const char *colName, int16 colNum, Node *newVa
 	}
 	else
 	{
-		tuple = SearchSysCacheCopyAttNum(RelationGetRelid(rel), colNum);
+		tuple = SearchSysCacheAttNum(RelationGetRelid(rel), colNum);
 
 		if (!HeapTupleIsValid(tuple))
 			ereport(ERROR,
@@ -8629,16 +8643,27 @@ ATExecSetStatistics(Relation rel, const char *colName, int16 colNum, Node *newVa
 					 errhint("Alter statistics on table column instead.")));
 	}
 
-	attrtuple->attstattarget = newtarget;
-
-	CatalogTupleUpdate(attrelation, &tuple->t_self, tuple);
+	/* Build new tuple. */
+	memset(repl_null, false, sizeof(repl_null));
+	memset(repl_repl, false, sizeof(repl_repl));
+	if (newtarget != -1)
+		repl_val[Anum_pg_attribute_attstattarget - 1] = newtarget;
+	else
+		repl_null[Anum_pg_attribute_attstattarget - 1] = true;
+	repl_repl[Anum_pg_attribute_attstattarget - 1] = true;
+	newtuple = heap_modify_tuple(tuple, RelationGetDescr(attrelation),
+								 repl_val, repl_null, repl_repl);
+	CatalogTupleUpdate(attrelation, &tuple->t_self, newtuple);
 
 	InvokeObjectPostAlterHook(RelationRelationId,
 							  RelationGetRelid(rel),
 							  attrtuple->attnum);
 	ObjectAddressSubSet(address, RelationRelationId,
 						RelationGetRelid(rel), attnum);
-	heap_freetuple(tuple);
+
+	heap_freetuple(newtuple);
+
+	ReleaseSysCache(tuple);
 
 	table_close(attrelation, RowExclusiveLock);
 

src/backend/parser/gram.y

@@ -337,6 +337,7 @@ static Node *makeRecursiveViewSelect(char *relname, List *aliases, Node *query);
 %type <list>	alter_table_cmds alter_type_cmds
 %type <list>    alter_identity_column_option_list
 %type <defelt>  alter_identity_column_option
+%type <node>	set_statistics_value
 
 %type <list>	createdb_opt_list createdb_opt_items copy_opt_list
 				transaction_mode_list
@@ -2446,18 +2447,18 @@ alter_table_cmd:
 					n->missing_ok = true;
 					$$ = (Node *) n;
 				}
-			/* ALTER TABLE <name> ALTER [COLUMN] <colname> SET STATISTICS <SignedIconst> */
-			| ALTER opt_column ColId SET STATISTICS SignedIconst
+			/* ALTER TABLE <name> ALTER [COLUMN] <colname> SET STATISTICS */
+			| ALTER opt_column ColId SET STATISTICS set_statistics_value
 				{
 					AlterTableCmd *n = makeNode(AlterTableCmd);
 
 					n->subtype = AT_SetStatistics;
 					n->name = $3;
-					n->def = (Node *) makeInteger($6);
+					n->def = $6;
 					$$ = (Node *) n;
 				}
-			/* ALTER TABLE <name> ALTER [COLUMN] <colnum> SET STATISTICS <SignedIconst> */
-			| ALTER opt_column Iconst SET STATISTICS SignedIconst
+			/* ALTER TABLE <name> ALTER [COLUMN] <colnum> SET STATISTICS */
+			| ALTER opt_column Iconst SET STATISTICS set_statistics_value
 				{
 					AlterTableCmd *n = makeNode(AlterTableCmd);
 
@@ -2469,7 +2470,7 @@ alter_table_cmd:
 
 					n->subtype = AT_SetStatistics;
 					n->num = (int16) $3;
-					n->def = (Node *) makeInteger($6);
+					n->def = $6;
 					$$ = (Node *) n;
 				}
 			/* ALTER TABLE <name> ALTER [COLUMN] <colname> SET ( column_parameter = value [, ... ] ) */
@@ -3070,6 +3071,11 @@ alter_identity_column_option:
 				}
 		;
 
+set_statistics_value:
+			SignedIconst					{ $$ = (Node *) makeInteger($1); }
+			| DEFAULT						{ $$ = NULL; }
+		;
+
 PartitionBoundSpec:
 			/* a HASH partition */
 			FOR VALUES WITH '(' hash_partbound ')'

src/backend/utils/cache/lsyscache.c

@@ -872,33 +872,6 @@ get_attnum(Oid relid, const char *attname)
 		return InvalidAttrNumber;
 }
 
-/*
- * get_attstattarget
- *
- *		Given the relation id and the attribute number,
- *		return the "attstattarget" field from the attribute relation.
- *
- *		Errors if not found.
- */
-int
-get_attstattarget(Oid relid, AttrNumber attnum)
-{
-	HeapTuple	tp;
-	Form_pg_attribute att_tup;
-	int			result;
-
-	tp = SearchSysCache2(ATTNUM,
-						 ObjectIdGetDatum(relid),
-						 Int16GetDatum(attnum));
-	if (!HeapTupleIsValid(tp))
-		elog(ERROR, "cache lookup failed for attribute %d of relation %u",
-			 attnum, relid);
-	att_tup = (Form_pg_attribute) GETSTRUCT(tp);
-	result = att_tup->attstattarget;
-	ReleaseSysCache(tp);
-	return result;
-}
-
 /*
  * get_attgenerated
  *