Received: from malur.postgresql.org ([217.196.149.56]) by arkaria.postgresql.org with esmtps (TLS1.3:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.92) (envelope-from ) id 1oTqsR-00018C-2W for pgsql-hackers@arkaria.postgresql.org; Thu, 01 Sep 2022 20:33:27 +0000 Received: from localhost ([127.0.0.1] helo=malur.postgresql.org) by malur.postgresql.org with esmtp (Exim 4.92) (envelope-from ) id 1oTqsP-0007XP-Ry for pgsql-hackers@arkaria.postgresql.org; Thu, 01 Sep 2022 20:33:25 +0000 Received: from makus.postgresql.org ([2001:4800:3e1:1::229]) by malur.postgresql.org with esmtps (TLS1.3:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.92) (envelope-from ) id 1oTqsP-0007XF-Cx for pgsql-hackers@lists.postgresql.org; Thu, 01 Sep 2022 20:33:25 +0000 Received: from sss.pgh.pa.us ([66.207.139.130]) by makus.postgresql.org with esmtps (TLS1.3:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.92) (envelope-from ) id 1oTqsK-0003Qc-2d for pgsql-hackers@postgresql.org; Thu, 01 Sep 2022 20:33:21 +0000 Received: from sss1.sss.pgh.pa.us (localhost [127.0.0.1]) by sss.pgh.pa.us (8.15.2/8.15.2) with ESMTP id 281KXILh635920; Thu, 1 Sep 2022 16:33:19 -0400 From: Tom Lane To: Quan Zongliang cc: pgsql-hackers Subject: Re: Bug: When user-defined AM is used, the index path cannot be selected correctly In-reply-to: <7f498182-6736-8fee-d2fe-452e336de795@yeah.net> References: <1606486.1660701833@sss.pgh.pa.us> <7f498182-6736-8fee-d2fe-452e336de795@yeah.net> Comments: In-reply-to Quan Zongliang message dated "Wed, 17 Aug 2022 17:11:43 +0800" MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="----- =_aaaaaaaaaa0" Content-ID: <635869.1662064380.0@sss.pgh.pa.us> Date: Thu, 01 Sep 2022 16:33:18 -0400 Message-ID: <635919.1662064398@sss.pgh.pa.us> List-Id: List-Help: List-Subscribe: List-Post: List-Owner: List-Archive: Archived-At: Precedence: bulk ------- =_aaaaaaaaaa0 Content-Type: text/plain; charset="us-ascii" Content-ID: <635869.1662064380.1@sss.pgh.pa.us> Quan Zongliang writes: > New patch attached. > It seems that partitions do not use AM other than btree and hash. > Rewrite only indxpath.c and check if it is a custom AM. This seems drastically underdocumented, and the test you're using for extension opclasses is wrong. What we need to know before applying match_boolean_index_clause is that a clause using BooleanEqualOperator will be valid for the index. That's got next door to nothing to do with whether the opclass is default for the index AM. A non-default opclass might support that operator, and conversely a default one might not (although I concede it's not that easy to imagine what other set of operators-on-boolean an extension opclass might be interested in). I think we need something more like the attached. regards, tom lane ------- =_aaaaaaaaaa0 Content-Type: text/x-diff; name="recognize-extension-boolean-opclasses-2.patch"; charset="us-ascii" Content-ID: <635869.1662064380.2@sss.pgh.pa.us> Content-Description: recognize-extension-boolean-opclasses-2.patch Content-Transfer-Encoding: quoted-printable diff --git a/src/backend/optimizer/path/indxpath.c b/src/backend/optimizer= /path/indxpath.c index 045ff2e487..63a8eef45c 100644 --- a/src/backend/optimizer/path/indxpath.c +++ b/src/backend/optimizer/path/indxpath.c @@ -153,6 +153,7 @@ static IndexClause *match_clause_to_indexcol(PlannerIn= fo *root, RestrictInfo *rinfo, int indexcol, IndexOptInfo *index); +static bool IsBooleanOpfamily(Oid opfamily); static IndexClause *match_boolean_index_clause(PlannerInfo *root, RestrictInfo *rinfo, int indexcol, IndexOptInfo *index); @@ -2342,6 +2343,23 @@ match_clause_to_indexcol(PlannerInfo *root, return NULL; } = +/* + * IsBooleanOpfamily + * Detect whether an opfamily supports boolean equality as an operator. + * + * If the opfamily OID is in the range of built-in objects, we can rely + * on hard-wired knowledge of which built-in opfamilies support this. + * For extension opfamilies, there's no choice but to do a catcache looku= p. + */ +static bool +IsBooleanOpfamily(Oid opfamily) +{ + if (opfamily < FirstNormalObjectId) + return IsBuiltinBooleanOpfamily(opfamily); + else + return op_in_opfamily(BooleanEqualOperator, opfamily); +} + /* * match_boolean_index_clause * Recognize restriction clauses that can be matched to a boolean index= . diff --git a/src/backend/optimizer/path/pathkeys.c b/src/backend/optimizer= /path/pathkeys.c index 1fa7fc99b5..b9c1ce0a64 100644 --- a/src/backend/optimizer/path/pathkeys.c +++ b/src/backend/optimizer/path/pathkeys.c @@ -1191,8 +1191,13 @@ partkey_is_bool_constant_for_query(RelOptInfo *part= rel, int partkeycol) PartitionScheme partscheme =3D partrel->part_scheme; ListCell *lc; = - /* If the partkey isn't boolean, we can't possibly get a match */ - if (!IsBooleanOpfamily(partscheme->partopfamily[partkeycol])) + /* + * If the partkey isn't boolean, we can't possibly get a match. + * + * Partitioning currently can only use built-in AMs, so checking for + * built-in boolean opclasses is good enough. + */ + if (!IsBuiltinBooleanOpfamily(partscheme->partopfamily[partkeycol])) return false; = /* Check each restriction clause for the partitioned rel */ diff --git a/src/backend/partitioning/partprune.c b/src/backend/partitioni= ng/partprune.c index bf9fe5b7aa..b5403149d7 100644 --- a/src/backend/partitioning/partprune.c +++ b/src/backend/partitioning/partprune.c @@ -3596,7 +3596,11 @@ match_boolean_partition_clause(Oid partopfamily, Ex= pr *clause, Expr *partkey, = *outconst =3D NULL; = - if (!IsBooleanOpfamily(partopfamily)) + /* + * Partitioning currently can only use built-in AMs, so checking for + * built-in boolean opclasses is good enough. + */ + if (!IsBuiltinBooleanOpfamily(partopfamily)) return PARTCLAUSE_UNSUPPORTED; = if (IsA(clause, BooleanTest)) diff --git a/src/include/catalog/pg_opfamily.h b/src/include/catalog/pg_op= family.h index 8dc9ce01bb..b56d714322 100644 --- a/src/include/catalog/pg_opfamily.h +++ b/src/include/catalog/pg_opfamily.h @@ -55,7 +55,8 @@ DECLARE_UNIQUE_INDEX_PKEY(pg_opfamily_oid_index, 2755, O= pfamilyOidIndexId, on pg = #ifdef EXPOSE_TO_CLIENT_CODE = -#define IsBooleanOpfamily(opfamily) \ +/* This does not account for non-core opclasses that might accept boolean= */ +#define IsBuiltinBooleanOpfamily(opfamily) \ ((opfamily) =3D=3D BOOL_BTREE_FAM_OID || (opfamily) =3D=3D BOOL_HASH_FAM= _OID) = #endif /* EXPOSE_TO_CLIENT_CODE */ ------- =_aaaaaaaaaa0--