Added support for semi/anti and outer joins to hash join algorithm. Reimplemented support for semi/anti joins inside the nested loop algorithm. Slightly changed implementation of full outer joins. Added transformation of IN/EXISTS subqueries into lateral semi-joins. Basic optimizer support for semi-joins. More efficient optimization for cross joins. Added some debug info (hash table statistics) for hash joins.

Author: dyemanov

src/jrd/RecordSourceNodes.cpp

@@ -52,6 +52,124 @@ static void genDeliverUnmapped(CompilerScratch* csb, const BoolExprNodeStack& pa
 static ValueExprNode* resolveUsingField(DsqlCompilerScratch* dsqlScratch, const MetaName& name,
 	ValueListNode* list, const FieldNode* flawedNode, const TEXT* side, dsql_ctx*& ctx);
 
+namespace
+{
+	// Search through the list of ANDed booleans to find comparisons
+	// referring streams of other select expressions.
+	// Extract those booleans and return them to the caller.
+
+	bool findDependentBooleans(CompilerScratch* csb,
+							   const StreamList& rseStreams,
+							   BoolExprNode** parentBoolean,
+							   BoolExprNodeStack& booleanStack)
+	{
+		const auto boolean = *parentBoolean;
+
+		const auto binaryNode = nodeAs<BinaryBoolNode>(boolean);
+		if (binaryNode && binaryNode->blrOp == blr_and)
+		{
+			const bool found1 = findDependentBooleans(csb, rseStreams,
+				binaryNode->arg1.getAddress(), booleanStack);
+			const bool found2 = findDependentBooleans(csb, rseStreams,
+				binaryNode->arg2.getAddress(), booleanStack);
+
+			if (!binaryNode->arg1 && !binaryNode->arg2)
+				*parentBoolean = nullptr;
+			else if (!binaryNode->arg1)
+				*parentBoolean = binaryNode->arg2;
+			else if (!binaryNode->arg2)
+				*parentBoolean = binaryNode->arg1;
+
+			return (found1 || found2);
+		}
+
+		if (const auto cmpNode = nodeAs<ComparativeBoolNode>(boolean))
+		{
+			SortedStreamList streams;
+			cmpNode->collectStreams(streams);
+
+			for (const auto stream : streams)
+			{
+				if (!rseStreams.exist(stream))
+				{
+					booleanStack.push(boolean);
+					*parentBoolean = nullptr;
+					return true;
+				}
+			}
+		}
+
+		return false;
+	}
+
+	// Search through the list of ANDed booleans to find correlated EXISTS/IN sub-queries.
+	// They are candidates to be converted into semi- or anti-joins.
+
+	bool findPossibleJoins(CompilerScratch* csb,
+						   BoolExprNode** parentBoolean,
+						   RecordSourceNodeStack& rseStack,
+						   BoolExprNodeStack& booleanStack)
+	{
+		auto boolNode = *parentBoolean;
+
+		const auto binaryNode = nodeAs<BinaryBoolNode>(boolNode);
+		if (binaryNode && binaryNode->blrOp == blr_and)
+		{
+			const bool found1 = findPossibleJoins(csb, binaryNode->arg1.getAddress(),
+				rseStack, booleanStack);
+			const bool found2 = findPossibleJoins(csb, binaryNode->arg2.getAddress(),
+				rseStack, booleanStack);
+
+			if (!binaryNode->arg1 && !binaryNode->arg2)
+				*parentBoolean = nullptr;
+			else if (!binaryNode->arg1)
+				*parentBoolean = binaryNode->arg2;
+			else if (!binaryNode->arg2)
+				*parentBoolean = binaryNode->arg1;
+
+			return (found1 || found2);
+		}
+
+		const auto rseNode = nodeAs<RseBoolNode>(boolNode);
+		// Both EXISTS (blr_any) and IN (blr_ansi_any) sub-queries are handled
+		if (rseNode && (rseNode->blrOp == blr_any || rseNode->blrOp == blr_ansi_any))
+		{
+			auto rse = rseNode->rse;
+			fb_assert(rse);
+
+			if (rse->rse_boolean)
+			{
+				StreamList streams;
+				rse->computeRseStreams(streams);
+
+				BoolExprNodeStack booleans;
+				if (findDependentBooleans(csb, streams,
+										  rse->rse_boolean.getAddress(),
+										  booleans))
+				{
+					fb_assert(booleans.hasData());
+					auto boolean = booleans.pop();
+					while (booleans.hasData())
+					{
+						const auto andNode = FB_NEW_POOL(csb->csb_pool)
+							BinaryBoolNode(csb->csb_pool, blr_and);
+						andNode->arg1 = boolean;
+						andNode->arg2 = booleans.pop();
+						boolean = andNode;
+					}
+
+					rse->flags |= RseNode::FLAG_SEMI_JOINED;
+					rseStack.push(rse);
+					booleanStack.push(boolean);
+					*parentBoolean = nullptr;
+					return true;
+				}
+			}
+		}
+
+		return false;
+	}
+}
 
 //--------------------
 
@@ -2783,6 +2901,9 @@ RseNode* RseNode::pass1(thread_db* tdbb, CompilerScratch* csb)
 {
 	SET_TDBB(tdbb);
 
+	if (const auto newRse = processPossibleJoins(tdbb, csb))
+		return newRse->pass1(tdbb, csb);
+
 	// for scoping purposes, maintain a stack of RseNode's which are
 	// currently being parsed; if there are none on the stack as
 	// yet, mark the RseNode as variant to make sure that statement-
@@ -2888,6 +3009,12 @@ RseNode* RseNode::pass1(thread_db* tdbb, CompilerScratch* csb)
 void RseNode::pass1Source(thread_db* tdbb, CompilerScratch* csb, RseNode* rse,
 	BoolExprNode** boolean, RecordSourceNodeStack& stack)
 {
+	if (const auto newRse = processPossibleJoins(tdbb, csb))
+	{
+		newRse->pass1Source(tdbb, csb, rse, boolean, stack);
+		return;
+	}
+
 	if (rse_jointype != blr_inner)
 	{
 		// Check whether any of the upper level booleans (those belonging to the WHERE clause)
@@ -2941,15 +3068,15 @@ void RseNode::pass1Source(thread_db* tdbb, CompilerScratch* csb, RseNode* rse,
 		}
 	}
 
-	// in the case of an RseNode, it is possible that a new RseNode will be generated,
+	// In the case of an RseNode, it is possible that a new RseNode will be generated,
 	// so wait to process the source before we push it on the stack (bug 8039)
 
 	// The addition of the JOIN syntax for specifying inner joins causes an
 	// RseNode tree to be generated, which is undesirable in the simplest case
 	// where we are just trying to inner join more than 2 streams. If possible,
 	// try to flatten the tree out before we go any further.
 
-	if (!isLateral() &&
+	if (!isLateral() && !isSemiJoined() &&
 		rse->rse_jointype == blr_inner &&
 		rse_jointype == blr_inner &&
 		!rse_sorted && !rse_projection &&
@@ -3054,11 +3181,11 @@ RecordSource* RseNode::compile(thread_db* tdbb, Optimizer* opt, bool innerSubStr
 
 		StreamStateHolder stateHolder(csb, opt->getOuterStreams());
 
-		if (opt->isLeftJoin() || isLateral())
+		if (opt->isLeftJoin() || isLateral() || isSemiJoined())
 		{
 			stateHolder.activate();
 
-			if (opt->isLeftJoin())
+			if (opt->isLeftJoin() || isSemiJoined())
 			{
 				// Push all conjuncts except "missing" ones (e.g. IS NULL)
 				for (auto iter = opt->getConjuncts(false, true); iter.hasData(); ++iter)
@@ -3081,6 +3208,42 @@ RecordSource* RseNode::compile(thread_db* tdbb, Optimizer* opt, bool innerSubStr
 	return opt->compile(this, &conjunctStack);
 }
 
+RseNode* RseNode::processPossibleJoins(thread_db* tdbb, CompilerScratch* csb)
+{
+	if (rse_jointype != blr_inner || !rse_boolean)
+		return nullptr;
+
+	RecordSourceNodeStack rseStack;
+	BoolExprNodeStack booleanStack;
+
+	// Find possibly joinable sub-queries
+
+	if (!findPossibleJoins(csb, rse_boolean.getAddress(), rseStack, booleanStack))
+		return nullptr;
+
+	fb_assert(rseStack.hasData() && booleanStack.hasData());
+	fb_assert(rseStack.getCount() == booleanStack.getCount());
+
+	// Create joins between the original node and detected joinable nodes
+
+	auto rse = this;
+	while (rseStack.hasData())
+	{
+		const auto newRse = FB_NEW_POOL(*tdbb->getDefaultPool())
+			RseNode(*tdbb->getDefaultPool());
+
+		newRse->rse_relations.add(rse);
+		newRse->rse_relations.add(rseStack.pop());
+
+		newRse->rse_jointype = blr_inner;
+		newRse->rse_boolean = booleanStack.pop();
+
+		rse = newRse;
+	}
+
+	return rse;
+}
+
 // Check that all streams in the RseNode have a plan specified for them.
 // If they are not, there are streams in the RseNode which were not mentioned in the plan.
 void RseNode::planCheck(const CompilerScratch* csb) const

src/jrd/RecordSourceNodes.h

@@ -719,15 +719,16 @@ class RseNode final : public TypedNode<RecordSourceNode, RecordSourceNode::TYPE_
 public:
 	enum : USHORT
 	{
-		FLAG_VARIANT			= 0x01,	// variant (not invariant?)
-		FLAG_SINGULAR			= 0x02,	// singleton select
-		FLAG_WRITELOCK			= 0x04,	// locked for write
-		FLAG_SCROLLABLE			= 0x08,	// scrollable cursor
-		FLAG_DSQL_COMPARATIVE	= 0x10,	// transformed from DSQL ComparativeBoolNode
-		FLAG_OPT_FIRST_ROWS		= 0x20,	// optimize retrieval for first rows
-		FLAG_LATERAL			= 0x40,	// lateral derived table
-		FLAG_SKIP_LOCKED		= 0x80,	// skip locked
-		FLAG_SUB_QUERY			= 0x100	// sub-query
+		FLAG_VARIANT			= 0x01,		// variant (not invariant?)
+		FLAG_SINGULAR			= 0x02,		// singleton select
+		FLAG_WRITELOCK			= 0x04,		// locked for write
+		FLAG_SCROLLABLE			= 0x08,		// scrollable cursor
+		FLAG_DSQL_COMPARATIVE	= 0x10,		// transformed from DSQL ComparativeBoolNode
+		FLAG_OPT_FIRST_ROWS		= 0x20,		// optimize retrieval for first rows
+		FLAG_LATERAL			= 0x40,		// lateral derived table
+		FLAG_SKIP_LOCKED		= 0x80,		// skip locked
+		FLAG_SUB_QUERY			= 0x100,	// sub-query
+		FLAG_SEMI_JOINED		= 0x200		// participates in semi-join
 	};
 
 	bool isInvariant() const
@@ -755,6 +756,11 @@ class RseNode final : public TypedNode<RecordSourceNode, RecordSourceNode::TYPE_
 		return (flags & FLAG_SUB_QUERY) != 0;
 	}
 
+	bool isSemiJoined() const
+	{
+		return (flags & FLAG_SEMI_JOINED) != 0;
+	}
+
 	bool hasWriteLock() const
 	{
 		return (flags & FLAG_WRITELOCK) != 0;
@@ -875,6 +881,7 @@ class RseNode final : public TypedNode<RecordSourceNode, RecordSourceNode::TYPE_
 private:
 	void planCheck(const CompilerScratch* csb) const;
 	static void planSet(CompilerScratch* csb, PlanNode* plan);
+	RseNode* processPossibleJoins(thread_db* tdbb, CompilerScratch* csb);
 
 public:
 	NestConst<ValueExprNode> dsqlFirst;

src/jrd/optimizer/InnerJoin.cpp

@@ -108,6 +108,8 @@ void InnerJoin::calculateStreamInfo()
 		innerStream->baseIndexes = candidate->indexes;
 		innerStream->baseUnique = candidate->unique;
 		innerStream->baseNavigated = candidate->navigated;
+		innerStream->baseMatches = candidate->matches;
+		innerStream->baseDependentFromStreams = candidate->dependentFromStreams;
 
 		csb->csb_rpt[innerStream->number].deactivate();
 	}
@@ -573,13 +575,36 @@ River* InnerJoin::formRiver()
 
 			// Create a hash join
 			rsb = FB_NEW_POOL(getPool())
-				HashJoin(tdbb, csb, 2, hashJoinRsbs, keys.begin(), stream.selectivity);
+				HashJoin(tdbb, csb, INNER_JOIN, 2, hashJoinRsbs, keys.begin(), stream.selectivity);
 
 			// Clear priorly processed rsb's, as they're already incorporated into a hash join
 			rsbs.clear();
 		}
 		else
+		{
+			StreamList depStreams;
+
+			if (optimizer->isSemiJoined() && rsbs.isEmpty())
+			{
+				const auto baseStream = getStreamInfo(stream.number);
+				for (const auto match : baseStream->baseMatches)
+				{
+					if (optimizer->checkEquiJoin(match))
+					{
+						for (const auto depStream : baseStream->baseDependentFromStreams)
+						{
+							if (match->containsStream(depStream))
+								depStreams.add(depStream);
+						}
+					}
+				}
+			}
+
+			StreamStateHolder stateHolder(csb, depStreams);
+			stateHolder.deactivate();
+
 			rsb = optimizer->generateRetrieval(stream.number, sortPtr, false, false);
+		}
 
 		rsbs.add(rsb);
 		streams.add(stream.number);