clean up

src/include/planner/subplans_table.h

@@ -13,7 +13,7 @@ namespace kuzu {
 namespace planner {
 
 const uint64_t MAX_LEVEL_TO_PLAN_EXACTLY = 7;
-const uint64_t MAX_NUM_SUBGRAPHS_PER_LEVEL = 100;
+const uint64_t MAX_NUM_SUBGRAPHS_PER_LEVEL = 50;
 const uint64_t MAX_NUM_PLANS_PER_SUBGRAPH = 50;
 
 class SubPlansTable {

src/planner/join_order_enumerator.cpp

@@ -79,7 +79,6 @@ std::vector<std::unique_ptr<LogicalPlan>> JoinOrderEnumerator::enumerate(
     while (context->currentLevel < context->maxLevel) {
         planLevel(context->currentLevel++);
     }
-    auto s = context->getPlans(context->getFullyMatchedSubqueryGraph()).size();
     return std::move(context->getPlans(context->getFullyMatchedSubqueryGraph()));
 }
 

src/planner/subplans_table.cpp

@@ -49,7 +49,7 @@ void SubPlansTable::clear() {
 }
 
 void SubPlansTable::PlanSet::addPlan(std::unique_ptr<LogicalPlan> plan) {
-    if (plans.size() > MAX_NUM_PLANS_PER_SUBGRAPH) {
+    if (plans.size() >= MAX_NUM_PLANS_PER_SUBGRAPH) {
         return;
     }
     auto schema = plan->getSchema();
@@ -58,6 +58,7 @@ void SubPlansTable::PlanSet::addPlan(std::unique_ptr<LogicalPlan> plan) {
         plans.push_back(std::move(plan));
     } else { // swap plan for lower cost under the same factorization schema
         auto idx = schemaToPlanIdx.at(schema);
+        assert(idx < MAX_NUM_PLANS_PER_SUBGRAPH);
         auto currentPlan = plans[idx].get();
         if (currentPlan->getCost() > plan->getCost()) {
             plans[idx] = std::move(plan);

src/processor/mapper/map_intersect.cpp

@@ -11,6 +11,7 @@ namespace processor {
 std::unique_ptr<PhysicalOperator> PlanMapper::mapLogicalIntersectToPhysical(
     LogicalOperator* logicalOperator) {
     auto logicalIntersect = (LogicalIntersect*)logicalOperator;
+    auto intersectNodeID = logicalIntersect->getIntersectNodeID();
     auto outSchema = logicalIntersect->getSchema();
     std::vector<std::unique_ptr<PhysicalOperator>> children;
     children.resize(logicalOperator->getNumChildren());
@@ -22,17 +23,19 @@ std::unique_ptr<PhysicalOperator> PlanMapper::mapLogicalIntersectToPhysical(
         auto buildSchema = logicalIntersect->getChild(i)->getSchema();
         auto buildSidePrevOperator = mapLogicalOperatorToPhysical(logicalIntersect->getChild(i));
         std::vector<DataPos> payloadsDataPos;
-        auto buildDataInfo =
-            generateBuildDataInfo(*buildSchema, {keyNodeID}, buildSchema->getExpressionsInScope());
-        for (auto& [dataPos, _] : buildDataInfo.payloadsPosAndType) {
-            auto expression = buildSchema->getGroup(dataPos.dataChunkPos)
-                                  ->getExpressions()[dataPos.valueVectorPos];
-            if (expression->getUniqueName() ==
-                logicalIntersect->getIntersectNodeID()->getUniqueName()) {
+        binder::expression_vector expressionsToMaterialize;
+        expressionsToMaterialize.push_back(keyNodeID);
+        expressionsToMaterialize.push_back(intersectNodeID);
+        for (auto& expression : buildSchema->getExpressionsInScope()) {
+            if (expression->getUniqueName() == keyNodeID->getUniqueName() ||
+                expression->getUniqueName() == intersectNodeID->getUniqueName()) {
                 continue;
             }
+            expressionsToMaterialize.push_back(expression);
             payloadsDataPos.emplace_back(outSchema->getExpressionPos(*expression));
         }
+        auto buildDataInfo =
+            generateBuildDataInfo(*buildSchema, {keyNodeID}, expressionsToMaterialize);
         auto sharedState = std::make_shared<IntersectSharedState>();
         sharedStates.push_back(sharedState);
         children[i] = make_unique<IntersectBuild>(

src/processor/operator/hash_join/join_hash_table.cpp

@@ -42,9 +42,12 @@ void JoinHashTable::append(const std::vector<ValueVector*>& vectorsToAppend) {
     // TODO(Guodong): use compiling information to remove the for loop.
     auto numTuplesToAppend = 1;
     for (auto i = 0u; i < numKeyColumns; i++) {
-        numTuplesToAppend *= (vectorsToAppend[i]->state->isFlat() ?
-                                  1 :
-                                  vectorsToAppend[i]->state->selVector->selectedSize);
+        // At most one unFlat key data chunk. If there are multiple unFlat key vectors, they must
+        // share the same state.
+        if (!vectorsToAppend[i]->state->isFlat()) {
+            numTuplesToAppend = vectorsToAppend[i]->state->selVector->selectedSize;
+            break;
+        }
     }
     auto appendInfos = factorizedTable->allocateFlatTupleBlocks(numTuplesToAppend);
     for (auto i = 0u; i < vectorsToAppend.size(); i++) {

src/processor/operator/intersect/intersect.cpp

@@ -17,9 +17,7 @@ void Intersect::initLocalStateInternal(ResultSet* resultSet, ExecutionContext* c
         for (auto i = 0u; i < dataInfo.payloadsDataPos.size(); i++) {
             auto vector = resultSet->getValueVector(dataInfo.payloadsDataPos[i]);
             // Always skip the first two columns in the fTable: build key and intersect key.
-            // TODO(Guodong): this is a potential bug because you cannot guarantee intersect key is
-            // the second column. Once this is solved, go back and refactor projection push down for
-            // intersect.
+            // TODO(Guodong): this is a bad practice. This assumption requires
             columnIdxesToScanFrom.push_back(i + 2);
             vectorsToReadInto.push_back(vector.get());
         }
@@ -147,7 +145,7 @@ void Intersect::populatePayloads(
     const std::vector<uint8_t*>& tuples, const std::vector<uint32_t>& listIdxes) {
     for (auto i = 0u; i < listIdxes.size(); i++) {
         auto listIdx = listIdxes[i];
-        sharedHTs[i]->getHashTable()->getFactorizedTable()->lookup(
+        sharedHTs[listIdx]->getHashTable()->getFactorizedTable()->lookup(
             payloadVectorsToScanInto[listIdx], intersectSelVectors[i].get(),
             payloadColumnIdxesToScanFrom[listIdx], tuples[listIdx]);
     }

test/test_files/tinysnb/cyclic/multi_label.test

@@ -3,10 +3,8 @@
 ---- 1
 3
 
-# TODO(Guodong): generic extend sames to give correct output but not getting correct final result
-# Also remember to add one with edge filter
-#-NAME MultiLabelTriangleTest
-#-QUERY MATCH (a:person)-[:knows]->(b:person)-[:studyAt|:mixed]->(c:organisation), (a)-[:studyAt]->(c) RETURN COUNT(*)
-#-ENUMERATE
-#---- 1
-#2
+-NAME MultiLabelTriangleTest
+-QUERY MATCH (a:person)-[:knows]->(b:person)-[:studyAt|:meets]->(c:organisation:person), (a)-[:studyAt|:meets]->(c) RETURN COUNT(*)
+-ENUMERATE
+---- 1
+4

test/test_files/tinysnb/cyclic/single_label.test

@@ -58,6 +58,13 @@ Dan|1950-05-14|Bob|1950-05-14|Carol|2000-01-01
 -ENUMERATE
 ---- 0
 
+-NAME TriangleTest6
+-QUERY MATCH (a:person)-[e1:knows]->(b:person)-[e2:knows]->(c:person), (a)-[e3:knows]->(c) WHERE a.fName='Alice' AND e1.meetTime=timestamp('1986-10-21 21:08:31.521') RETURN b.fName, c.fName
+-ENUMERATE
+---- 2
+Bob|Carol
+Bob|Dan
+
 -NAME TriangleWithProjectionTest
 -QUERY MATCH (a:person)-[e1:knows]->(b:person)-[e2:studyAt]->(c:organisation), (a)-[e3:studyAt]->(c) RETURN a.fName, e1.date, b.fName, e2.year, c.name, e3.year
 -ENUMERATE