Code review for commands/statscmds.c.

Fix machine-dependent sorting of column numbers.  (Odd behavior
would only materialize for column numbers above 255, but that's
certainly legal.)

Fix poor choice of SQLSTATE for some errors, and improve error message
wording.  (Notably, "is not a scalar type" is a totally misleading way
to explain "does not have a default btree opclass".)

Avoid taking AccessExclusiveLock on the associated relation during DROP
STATISTICS.  That's neither necessary nor desirable, and it could easily
have put us into situations where DROP fails (compare commit 68ea2b7f9).

Adjust/improve comments.

David Rowley and Tom Lane

Discussion: https://postgr.es/m/CAKJS1f-GmCfPvBbAEaM5xoVOaYdVgVN1gicALSoYQ77z-+vLbw@mail.gmail.com

src/backend/commands/statscmds.c

@@ -31,11 +31,15 @@
 #include "utils/typcache.h"
 
 
-/* used for sorting the attnums in CreateStatistics */
+/* qsort comparator for the attnums in CreateStatistics */
 static int
 compare_int16(const void *a, const void *b)
 {
-	return memcmp(a, b, sizeof(int16));
+	int			av = *(const int16 *) a;
+	int			bv = *(const int16 *) b;
+
+	/* this can't overflow if int is wider than int16 */
+	return (av - bv);
 }
 
 /*
@@ -44,8 +48,6 @@ compare_int16(const void *a, const void *b)
 ObjectAddress
 CreateStatistics(CreateStatsStmt *stmt)
 {
-	int			i;
-	ListCell   *l;
 	int16		attnums[STATS_MAX_DIMENSIONS];
 	int			numcols = 0;
 	ObjectAddress address = InvalidObjectAddress;
@@ -68,6 +70,8 @@ CreateStatistics(CreateStatsStmt *stmt)
 	bool		build_ndistinct;
 	bool		build_dependencies;
 	bool		requested_type = false;
+	int			i;
+	ListCell   *l;
 
 	Assert(IsA(stmt, CreateStatsStmt));
 
@@ -76,10 +80,10 @@ CreateStatistics(CreateStatsStmt *stmt)
 	namestrcpy(&stxname, namestr);
 
 	/*
-	 * If if_not_exists was given and the statistics already exists, bail out.
+	 * Deal with the possibility that the named statistics already exist.
 	 */
 	if (SearchSysCacheExists2(STATEXTNAMENSP,
-							  PointerGetDatum(&stxname),
+							  NameGetDatum(&stxname),
 							  ObjectIdGetDatum(namespaceId)))
 	{
 		if (stmt->if_not_exists)
@@ -97,10 +101,11 @@ CreateStatistics(CreateStatsStmt *stmt)
 	}
 
 	/*
-	 * CREATE STATISTICS will influence future execution plans but does
+	 * CREATE STATISTICS will influence future execution plans but does not
-	 * not interfere with currently executing plans so it is safe to
+	 * interfere with currently executing plans.  So it should be enough to
 	 * take only ShareUpdateExclusiveLock on relation, conflicting with
-	 * ANALYZE and other DDL that sets statistical information.
+	 * ANALYZE and other DDL that sets statistical information, but not with
+	 * normal queries.
 	 */
 	rel = relation_openrv(stmt->relation, ShareUpdateExclusiveLock);
 	relid = RelationGetRelid(rel);
@@ -137,25 +142,26 @@ CreateStatistics(CreateStatsStmt *stmt)
 		if (attForm->attnum < 0)
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("statistic creation on system columns is not supported")));
+					 errmsg("statistics creation on system columns is not supported")));
 
 		/* Disallow data types without a less-than operator */
 		type = lookup_type_cache(attForm->atttypid, TYPECACHE_LT_OPR);
 		if (type->lt_opr == InvalidOid)
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					 errmsg("only scalar types can be used in extended statistics")));
+					 errmsg("column \"%s\" cannot be used in statistics because its type has no default btree operator class",
+							attname)));
 
 		/* Make sure no more than STATS_MAX_DIMENSIONS columns are used */
 		if (numcols >= STATS_MAX_DIMENSIONS)
 			ereport(ERROR,
 					(errcode(ERRCODE_TOO_MANY_COLUMNS),
-					 errmsg("cannot have more than %d keys in statistics",
+					 errmsg("cannot have more than %d columns in statistics",
 							STATS_MAX_DIMENSIONS)));
 
 		attnums[numcols] = ((Form_pg_attribute) GETSTRUCT(atttuple))->attnum;
-		ReleaseSysCache(atttuple);
 		numcols++;
+		ReleaseSysCache(atttuple);
 	}
 
 	/*
@@ -164,26 +170,27 @@ CreateStatistics(CreateStatsStmt *stmt)
 	 */
 	if (numcols < 2)
 		ereport(ERROR,
-				(errcode(ERRCODE_TOO_MANY_COLUMNS),
+				(errcode(ERRCODE_INVALID_OBJECT_DEFINITION),
-				 errmsg("statistics require at least 2 columns")));
+				 errmsg("extended statistics require at least 2 columns")));
 
 	/*
-	 * Sort the attnums, which makes detecting duplicities somewhat easier, and
+	 * Sort the attnums, which makes detecting duplicates somewhat easier, and
 	 * it does not hurt (it does not affect the efficiency, unlike for
 	 * indexes, for example).
 	 */
 	qsort(attnums, numcols, sizeof(int16), compare_int16);
 
 	/*
-	 * Look for duplicities in the list of columns. The attnums are sorted so
+	 * Check for duplicates in the list of columns. The attnums are sorted so
 	 * just check consecutive elements.
 	 */
 	for (i = 1; i < numcols; i++)
 		if (attnums[i] == attnums[i - 1])
 			ereport(ERROR,
-					(errcode(ERRCODE_UNDEFINED_COLUMN),
+					(errcode(ERRCODE_DUPLICATE_COLUMN),
 				  errmsg("duplicate column name in statistics definition")));
 
+	/* Form an int2vector representation of the sorted column list */
 	stxkeys = buildint2vector(attnums, numcols);
 
 	/*
@@ -225,7 +232,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 		types[ntypes++] = CharGetDatum(STATS_EXT_NDISTINCT);
 	if (build_dependencies)
 		types[ntypes++] = CharGetDatum(STATS_EXT_DEPENDENCIES);
-	Assert(ntypes > 0);
+	Assert(ntypes > 0 && ntypes <= lengthof(types));
 	stxkind = construct_array(types, ntypes, CHAROID, 1, true, 'c');
 
 	/*
@@ -240,7 +247,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 	values[Anum_pg_statistic_ext_stxkeys - 1] = PointerGetDatum(stxkeys);
 	values[Anum_pg_statistic_ext_stxkind - 1] = PointerGetDatum(stxkind);
 
-	/* no statistics build yet */
+	/* no statistics built yet */
 	nulls[Anum_pg_statistic_ext_stxndistinct - 1] = true;
 	nulls[Anum_pg_statistic_ext_stxdependencies - 1] = true;
 
@@ -260,7 +267,7 @@ CreateStatistics(CreateStatsStmt *stmt)
 	relation_close(rel, NoLock);
 
 	/*
-	 * Add a dependency on a table, so that stats get dropped on DROP TABLE.
+	 * Add a dependency on the table, so that stats get dropped on DROP TABLE.
 	 */
 	ObjectAddressSet(parentobject, RelationRelationId, relid);
 	ObjectAddressSet(childobject, StatisticExtRelationId, statoid);
@@ -269,12 +276,15 @@ CreateStatistics(CreateStatsStmt *stmt)
 	/*
 	 * Also add dependency on the schema.  This is required to ensure that we
 	 * drop the statistics on DROP SCHEMA.  This is not handled automatically
-	 * by DROP TABLE because the statistics are not an object in the table's
+	 * by DROP TABLE because the statistics might be in a different schema
-	 * schema.
+	 * from the table itself.  (This definition is a bit bizarre for the
+	 * single-table case, but it will make more sense if/when we support
+	 * extended stats across multiple tables.)
 	 */
 	ObjectAddressSet(parentobject, NamespaceRelationId, namespaceId);
 	recordDependencyOn(&childobject, &parentobject, DEPENDENCY_AUTO);
 
+	/* Return stats object's address */
 	ObjectAddressSet(address, StatisticExtRelationId, statoid);
 
 	return address;
@@ -287,13 +297,13 @@ void
 RemoveStatisticsById(Oid statsOid)
 {
 	Relation	relation;
-	Oid			relid;
-	Relation	rel;
 	HeapTuple	tup;
 	Form_pg_statistic_ext statext;
+	Oid			relid;
 
 	/*
-	 * Delete the pg_statistic_ext tuple.
+	 * Delete the pg_statistic_ext tuple.  Also send out a cache inval on the
+	 * associated table, so that dependent plans will be rebuilt.
 	 */
 	relation = heap_open(StatisticExtRelationId, RowExclusiveLock);
 
@@ -305,14 +315,11 @@ RemoveStatisticsById(Oid statsOid)
 	statext = (Form_pg_statistic_ext) GETSTRUCT(tup);
 	relid = statext->stxrelid;
 
-	rel = heap_open(relid, AccessExclusiveLock);
+	CacheInvalidateRelcacheByRelid(relid);
 
 	simple_heap_delete(relation, &tup->t_self);
 
-	CacheInvalidateRelcache(rel);
-
 	ReleaseSysCache(tup);
 
 	heap_close(relation, RowExclusiveLock);
-	heap_close(rel, NoLock);
 }

src/test/regress/expected/stats_ext.out

@@ -163,7 +163,7 @@ CREATE STATISTICS s10 ON (unknown_column) FROM ndistinct;
 ERROR:  column "unknown_column" referenced in statistics does not exist
 -- single column
 CREATE STATISTICS s10 ON (a) FROM ndistinct;
-ERROR:  statistics require at least 2 columns
+ERROR:  extended statistics require at least 2 columns
 -- single column, duplicated
 CREATE STATISTICS s10 ON (a,a) FROM ndistinct;
 ERROR:  duplicate column name in statistics definition