This is an automated email from the ASF dual-hosted git repository. oppenheimer01 pushed a commit to branch cbdb-postgres-merge in repository https://gitbox.apache.org/repos/asf/cloudberry.git
commit 4087e15b601cacbbdaaa627fc8eb0ca4438d2e1c Author: Jianghua.yjh <[email protected]> AuthorDate: Mon Apr 20 21:43:32 2026 -0700 ORCA: add optimizer_use_streaming_hashagg GUC (#1681) ORCA unconditionally sets stream_safe=true for all local HashAggs in FLocalHashAggStreamSafe, so the existing gp_use_streaming_hashagg GUC (which is only read by the Postgres planner path in cdbgroupingpaths.c) has no effect when optimizer=on. There was no way to disable streaming hash agg for ORCA plans. Introduce optimizer_use_streaming_hashagg (default on) and wire it through the standard CConfigParamMapping path: map it (negated) to a new EopttraceDisableStreamingHashAgg traceflag, and check that traceflag in FLocalHashAggStreamSafe. When the GUC is off, ORCA emits a non-streaming Partial HashAggregate that spills to disk and fully deduplicates. --- src/backend/gpopt/config/CConfigParamMapping.cpp | 7 ++++++- .../libgpopt/src/translate/CTranslatorExprToDXLUtils.cpp | 3 ++- .../libnaucrates/include/naucrates/traceflags/traceflags.h | 3 +++ src/backend/utils/misc/guc_gp.c | 11 +++++++++++ src/include/utils/guc.h | 1 + src/include/utils/unsync_guc_name.h | 1 + 6 files changed, 24 insertions(+), 2 deletions(-) diff --git a/src/backend/gpopt/config/CConfigParamMapping.cpp b/src/backend/gpopt/config/CConfigParamMapping.cpp index 603855c50ec..62255364ba8 100644 --- a/src/backend/gpopt/config/CConfigParamMapping.cpp +++ b/src/backend/gpopt/config/CConfigParamMapping.cpp @@ -331,7 +331,12 @@ CConfigParamMapping::SConfigMappingElem CConfigParamMapping::m_elements[] = { false, // m_negate_param GPOS_WSZ_LIT( "Enable create window hash agg")}, - + + {EopttraceDisableStreamingHashAgg, &optimizer_use_streaming_hashagg, + true, // m_negate_param + GPOS_WSZ_LIT( + "Disable streaming hash agg in ORCA-generated local partial aggregations.")}, + }; //--------------------------------------------------------------------------- diff --git a/src/backend/gporca/libgpopt/src/translate/CTranslatorExprToDXLUtils.cpp b/src/backend/gporca/libgpopt/src/translate/CTranslatorExprToDXLUtils.cpp index 27f5cb688fe..cf497223553 100644 --- a/src/backend/gporca/libgpopt/src/translate/CTranslatorExprToDXLUtils.cpp +++ b/src/backend/gporca/libgpopt/src/translate/CTranslatorExprToDXLUtils.cpp @@ -1296,7 +1296,8 @@ CTranslatorExprToDXLUtils::FLocalHashAggStreamSafe(CExpression *pexprAgg) // is a local hash aggregate and it generates duplicates (therefore safe to stream) return (COperator::EgbaggtypeLocal == popAgg->Egbaggtype()) && - popAgg->FGeneratesDuplicates(); + popAgg->FGeneratesDuplicates() && + !GPOS_FTRACE(EopttraceDisableStreamingHashAgg); } //--------------------------------------------------------------------------- diff --git a/src/backend/gporca/libnaucrates/include/naucrates/traceflags/traceflags.h b/src/backend/gporca/libnaucrates/include/naucrates/traceflags/traceflags.h index 2e489f214e5..8a18ace986a 100644 --- a/src/backend/gporca/libnaucrates/include/naucrates/traceflags/traceflags.h +++ b/src/backend/gporca/libnaucrates/include/naucrates/traceflags/traceflags.h @@ -250,6 +250,9 @@ enum EOptTraceFlag // Use the all key exclude the non-fixed key in AGG pds EopttraceAggRRSExcludeNonFixedKey = 103053, + // Disable streaming hash agg in ORCA-generated local partial aggregations + EopttraceDisableStreamingHashAgg = 103054, + /////////////////////////////////////////////////////// ///////////////////// statistics flags //////////////// ////////////////////////////////////////////////////// diff --git a/src/backend/utils/misc/guc_gp.c b/src/backend/utils/misc/guc_gp.c index 50ff4194684..4312c86ad15 100644 --- a/src/backend/utils/misc/guc_gp.c +++ b/src/backend/utils/misc/guc_gp.c @@ -154,6 +154,7 @@ bool enable_parallel_dedup_semi_join = true; bool enable_parallel_dedup_semi_reverse_join = true; bool parallel_query_use_streaming_hashagg = false; bool gp_use_streaming_hashagg = true; +bool optimizer_use_streaming_hashagg = true; int gp_appendonly_insert_files = 0; int gp_appendonly_insert_files_tuples_range = 0; int gp_random_insert_segments = 0; @@ -1909,6 +1910,16 @@ struct config_bool ConfigureNamesBool_gp[] = true, NULL, NULL }, + { + {"optimizer_use_streaming_hashagg", PGC_USERSET, DEVELOPER_OPTIONS, + gettext_noop("Use streaming hash agg in ORCA-generated local partial hash aggregations."), + NULL, + GUC_NO_SHOW_ALL | GUC_NOT_IN_SAMPLE + }, + &optimizer_use_streaming_hashagg, + true, NULL, NULL + }, + { {"gp_force_random_redistribution", PGC_USERSET, CUSTOM_OPTIONS, gettext_noop("Force redistribution of insert for randomly-distributed."), diff --git a/src/include/utils/guc.h b/src/include/utils/guc.h index 354bca248f5..4f6401fb43f 100644 --- a/src/include/utils/guc.h +++ b/src/include/utils/guc.h @@ -301,6 +301,7 @@ extern bool enable_parallel_dedup_semi_join; extern bool enable_parallel_dedup_semi_reverse_join; extern bool parallel_query_use_streaming_hashagg; extern bool gp_use_streaming_hashagg; +extern bool optimizer_use_streaming_hashagg; extern int gp_appendonly_insert_files; extern int gp_appendonly_insert_files_tuples_range; extern int gp_random_insert_segments; diff --git a/src/include/utils/unsync_guc_name.h b/src/include/utils/unsync_guc_name.h index e74955d0360..e9d5c7eb05b 100644 --- a/src/include/utils/unsync_guc_name.h +++ b/src/include/utils/unsync_guc_name.h @@ -511,6 +511,7 @@ "optimizer_skew_factor", "optimizer_use_external_constant_expression_evaluation_for_ints", "optimizer_use_gpdb_allocators", + "optimizer_use_streaming_hashagg", "optimizer_xform_bind_threshold", "parallel_leader_participation", "parallel_query_use_streaming_hashagg", --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
