ORCA: allow partition table use the append node

In ORCA, both Dynamic Scan and Append operators can serve as scan methods
for partitioned tables.

Historically, ORCA initially supported the Append operator before switching
to Dynamic Any Scan, with several iterations of changes in between.

In current CBDB versions, Dynamic Scan is the default partitioned table scan
operator. However, the vectorized executor in CBDB (closed-source) cannot
support Dynamic Scan due to its multi-threaded architecture, which prohibits
metadata reading during scanning.

This commit reintroduces the Append operator as an alternative. The GUC
`optimizer_disable_dynamic_table_scan` (default: off) now allows choosing
between Dynamic Scan and Append operators.

Important considerations when using Append operator:

1. **Row-level security unsupported on root partition table**: Append cannot
be used with row-level security on partitioned tables. ORCA applies row-level
security at the final stage, making it impossible to determine the root partition
after generating Append plans. Dynamic Scan preserves root partition information,
enabling runtime application of row-level security to child partitions.

2. **No index-only scan support**: This is a CBDB executor limitation. Only
Dynamic Scan supports index-only scans for partitioned tables.

Author: jiaqizho

src/backend/gpopt/config/CConfigParamMapping.cpp

@@ -301,15 +301,18 @@ CConfigParamMapping::SConfigMappingElem CConfigParamMapping::m_elements[] = {
 	{EopttraceEnableUseDistributionInDQA,
 	 &optimizer_enable_use_distribution_in_dqa,
 	 false,	 // m_negate_param
-	 GPOS_WSZ_LIT(
-		 "Enable use the distribution key in DQA")},
+	 GPOS_WSZ_LIT("Enable use the distribution key in DQA")},
 	{EopttraceDisableInnerHashJoin, &optimizer_enable_hashjoin,
 	 true,	// m_negate_param
 	 GPOS_WSZ_LIT("Explore hash join alternatives")},
 	{EopttraceDisableInnerNLJ, &optimizer_enable_nljoin,
 	 true,	// m_negate_param
 	 GPOS_WSZ_LIT("Enable nested loop join alternatives")},
 
+	{EopttraceDisableDynamicTableScan, &optimizer_disable_dynamic_table_scan,
+	 false,	 // m_negate_param
+	 GPOS_WSZ_LIT(
+		 "Disable the dynamic seq/bitmap/index scan in partition table")},
 };
 
 //---------------------------------------------------------------------------

src/backend/gpopt/translate/CTranslatorDXLToPlStmt.cpp

@@ -3921,6 +3921,31 @@ CTranslatorDXLToPlStmt::TranslateDXLAppend(
 	GPOS_ASSERT(EdxlappendIndexFirstChild < arity);
 	append->appendplans = NIL;
 
+	// translate table descriptor into a range table entry
+	CDXLPhysicalAppend *phy_append_dxlop =
+		CDXLPhysicalAppend::Cast(append_dxlnode->GetOperator());
+
+	// If this append was create from a DynamicTableScan node in ORCA, it will
+	// contain the table descriptor of the root partitioned table. Add that to
+	// the range table in the PlStmt.
+	if (phy_append_dxlop->GetScanId() != gpos::ulong_max)
+	{
+		GPOS_ASSERT(nullptr != phy_append_dxlop->GetDXLTableDesc());
+
+		// translation context for column mappings in the base relation
+		CDXLTranslateContextBaseTable base_table_context(m_mp);
+
+		(void) ProcessDXLTblDescr(phy_append_dxlop->GetDXLTableDesc(),
+								  &base_table_context);
+
+		OID oid_type =
+			CMDIdGPDB::CastMdid(m_md_accessor->PtMDType<IMDTypeInt4>()->MDId())
+				->Oid();
+		append->join_prune_paramids =
+			TranslateJoinPruneParamids(phy_append_dxlop->GetSelectorIds(),
+									   oid_type, m_dxl_to_plstmt_context);
+	}
+
 	// translate children
 	CDXLTranslateContext child_context(m_mp, false,
 									   output_context->GetColIdToParamIdMap());

src/backend/gpopt/translate/CTranslatorQueryToDXL.cpp

@@ -528,6 +528,11 @@ CTranslatorQueryToDXL::CheckRangeTable(Query *query)
 			GPOS_RAISE(gpdxl::ExmaDXL, gpdxl::ExmiQuery2DXLUnsupportedFeature,
 					   GPOS_WSZ_LIT("TABLESAMPLE in the FROM clause"));
 		}
+
+		if (rte->relkind == 'p' && query->hasRowSecurity && GPOS_FTRACE(EopttraceDisableDynamicTableScan)) {
+			GPOS_RAISE(gpdxl::ExmaDXL, gpdxl::ExmiQuery2DXLUnsupportedFeature,
+				GPOS_WSZ_LIT("ORCA not support row-level security if dynamic table scan is disabled."));
+		}
 	}
 }
 

src/backend/gpopt/translate/CTranslatorRelcacheToDXL.cpp

@@ -3044,7 +3044,7 @@ CTranslatorRelcacheToDXL::RetrieveIndexPartitions(CMemoryPool *mp, OID rel_oid)
 	{
 		OID oid = lfirst_oid(lc);
 		partition_oids->Append(GPOS_NEW(mp)
-								   CMDIdGPDB(IMDId::EmdidGeneral, oid));
+								   CMDIdGPDB(IMDId::EmdidInd, oid));
 	}
 
 	return partition_oids;

src/backend/gporca/libgpopt/include/gpopt/operators/CLogicalDynamicGetBase.h

@@ -70,11 +70,6 @@ class CLogicalDynamicGetBase : public CLogical
 	static ColRefToUlongMapArray *ConstructRootColMappingPerPart(
 		CMemoryPool *mp, CColRefArray *root_cols, IMdIdArray *partition_mdids);
 
-	using ColNameToIndexMap =
-		CHashMap<const CWStringConst, ULONG, CWStringConst::HashValue,
-				 CWStringConst::Equals, CleanupNULL<const CWStringConst>,
-				 CleanupDelete<ULONG>>;
-
 public:
 	// ctors
 	explicit CLogicalDynamicGetBase(CMemoryPool *mp);

src/backend/gporca/libgpopt/include/gpopt/operators/CLogicalDynamicIndexOnlyGet.h

@@ -87,9 +87,11 @@ class CLogicalDynamicIndexOnlyGet : public CLogicalDynamicIndexGet
 	{
 		CXformSet *xform_set = GPOS_NEW(mp) CXformSet(mp);
 
-		(void) xform_set->ExchangeSet(
-			CXform::ExfDynamicIndexOnlyGet2DynamicIndexOnlyScan);
-
+		// Append node have not implements the index only scan
+		if (!GPOS_FTRACE(EopttraceDisableDynamicTableScan)) {
+			(void) xform_set->ExchangeSet(
+				CXform::ExfDynamicIndexOnlyGet2DynamicIndexOnlyScan);
+		}
 		return xform_set;
 	}
 

src/backend/gporca/libgpopt/include/gpopt/translate/CTranslatorExprToDXL.h

@@ -313,6 +313,26 @@ class CTranslatorExprToDXL
 							CDistributionSpecArray *pdrgpdsBaseTables,
 							ULONG *pulNonGatherMotions, BOOL *pfDML);
 
+	// translate a dynamic seq/foreign scan to append
+	template <class PhysicalScanType>
+	CDXLNode *PdxlnDynamicScanToAppend(
+		CExpression *pexprDTS, CColRefArray *colref_array,
+		CDistributionSpecArray *pdrgpdsBaseTables, CExpression *pexprScalarCond,
+		CDXLPhysicalProperties *dxl_properties);
+	
+	// translate a dynamic index/index only scan to append
+	template <class PhysicalScanType>
+	CDXLNode *PdxlnDynamicIndexScanToAppend(CExpression *pexprDIS,
+		CColRefArray *colref_array,
+		CDXLPhysicalProperties *dxl_properties,
+		CReqdPropPlan *prpp);
+
+	// translate a dynamic bitmap scan to append
+	CDXLNode *PdxlnDynamicBitmapScanToAppend(
+		CExpression *pexprDTS, CColRefArray *colref_array,
+		CDistributionSpecArray *pdrgpdsBaseTables, CExpression *pexprScalarCond,
+		CDXLPhysicalProperties *dxl_properties);
+
 	// translate a dynamic table scan
 	CDXLNode *PdxlnDynamicTableScan(CExpression *pexprDTS,
 									CColRefArray *colref_array,
@@ -562,6 +582,16 @@ class CTranslatorExprToDXL
 									const CColRefArray *root_colrefs,
 									CExpression *pred);
 
+	CDXLNode *PdxlnBitmapIndexProbeForChildPart(
+		const ColRefToUlongMap *root_col_mapping,
+		const CColRefArray *part_colrefs, const CColRefArray *root_colrefs,
+		const IMDRelation *part, CExpression *pexprBitmapIndexProbe);
+								
+	CDXLNode *PdxlnBitmapIndexPathForChildPart(
+		const ColRefToUlongMap *root_col_mapping,
+		const CColRefArray *part_colrefs, const CColRefArray *root_colrefs,
+		const IMDRelation *part, CExpression *pexprBitmapIndexPath);
+
 	// translate a project list expression into a DXL proj list node
 	// according to the order specified in the dynamic array
 	CDXLNode *PdxlnProjList(const CExpression *pexprProjList,

src/backend/gporca/libgpopt/src/operators/CLogicalDynamicGetBase.cpp

@@ -236,15 +236,6 @@ CLogicalDynamicGetBase::ConstructRootColMappingPerPart(
 	CMDAccessor *mda = COptCtxt::PoctxtFromTLS()->Pmda();
 
 	ColRefToUlongMapArray *part_maps = GPOS_NEW(mp) ColRefToUlongMapArray(mp);
-
-	// Build hashmap of colname to the index
-	ColNameToIndexMap *root_mapping = GPOS_NEW(mp) ColNameToIndexMap(mp);
-	for (ULONG i = 0; i < root_cols->Size(); ++i)
-	{
-		CColRef *root_colref = (*root_cols)[i];
-		root_mapping->Insert(root_colref->Name().Pstr(), GPOS_NEW(mp) ULONG(i));
-	}
-
 	for (ULONG ul = 0; ul < partition_mdids->Size(); ++ul)
 	{
 		IMDId *part_mdid = (*partition_mdids)[ul];
@@ -253,28 +244,35 @@ CLogicalDynamicGetBase::ConstructRootColMappingPerPart(
 		GPOS_ASSERT(nullptr != partrel);
 
 		ColRefToUlongMap *mapping = GPOS_NEW(mp) ColRefToUlongMap(mp);
-		// The root mapping cannot contain dropped columns, but may be
-		// in a different order than the child cols.Iterate through each of the child
-		// cols, and retrieve the corresponding index in the parent table
-		for (ULONG j = 0; j < partrel->ColumnCount(); ++j)
+
+		for (ULONG i = 0; i < root_cols->Size(); ++i)
 		{
-			const IMDColumn *coldesc = partrel->GetMdCol(j);
-			const CWStringConst *colname = coldesc->Mdname().GetMDName();
+			CColRef *root_colref = (*root_cols)[i];
 
-			if (coldesc->IsDropped())
+			BOOL found_mapping = false;
+			for (ULONG j = 0, idx = 0; j < partrel->ColumnCount(); ++j, ++idx)
 			{
-				continue;
+				const IMDColumn *coldesc = partrel->GetMdCol(j);
+				const CWStringConst *colname = coldesc->Mdname().GetMDName();
+
+				if (coldesc->IsDropped())
+				{
+					--idx;
+					continue;
+				}
+
+				if (colname->Equals(root_colref->Name().Pstr()))
+				{
+					// Found the corresponding column in the child partition
+					// Save the index in the mapping
+					mapping->Insert(root_colref, GPOS_NEW(mp) ULONG(idx));
+					found_mapping = true;
+					break;
+				}
 			}
 
-			ULONG *root_idx = root_mapping->Find(colname);
-			if (nullptr != root_idx)
-			{
-				mapping->Insert((*root_cols)[*root_idx],
-								GPOS_NEW(mp) ULONG(*root_idx));
-			}
-			else
+			if (!found_mapping)
 			{
-				root_mapping->Release();
 				GPOS_RAISE(
 					CException::ExmaInvalid, CException::ExmiInvalid,
 					GPOS_WSZ_LIT(
@@ -283,6 +281,5 @@ CLogicalDynamicGetBase::ConstructRootColMappingPerPart(
 		}
 		part_maps->Append(mapping);
 	}
-	root_mapping->Release();
 	return part_maps;
 }