Refactoring preliminaries for lazy operations (Part 1) (#1352)

This PR contains all the changes from the infrastructure for lazy operation evaluation (#1350)  that are simple and repetitive, but touch many files. In particular:

* Rename the `ResultTable` class to `Result` (a TODO suggested by @hannahbast some time ago).
* Add a new parameter `bool requestLaziness` to `Operation::computeResult`. This parameter is currently unused.

src/engine/Bind.cpp

@@ -81,10 +81,10 @@ std::vector<QueryExecutionTree*> Bind::getChildren() {
 }
 
 // _____________________________________________________________________________
-ResultTable Bind::computeResult() {
+Result Bind::computeResult([[maybe_unused]] bool requestLaziness) {
   using std::endl;
   LOG(DEBUG) << "Get input to BIND operation..." << endl;
-  shared_ptr<const ResultTable> subRes = _subtree->getResult();
+  std::shared_ptr<const Result> subRes = _subtree->getResult();
   LOG(DEBUG) << "Got input to Bind operation." << endl;
   IdTable idTable{getExecutionContext()->getAllocator()};
 
@@ -114,7 +114,7 @@ ResultTable Bind::computeResult() {
 template <size_t IN_WIDTH, size_t OUT_WIDTH>
 void Bind::computeExpressionBind(
     IdTable* outputIdTable, LocalVocab* outputLocalVocab,
-    const ResultTable& inputResultTable,
+    const Result& inputResultTable,
     sparqlExpression::SparqlExpression* expression) const {
   sparqlExpression::EvaluationContext evaluationContext(
       *getExecutionContext(), _subtree->getVariableColumns(),

src/engine/Bind.h

@@ -46,13 +46,13 @@ class Bind : public Operation {
   [[nodiscard]] vector<ColumnIndex> resultSortedOn() const override;
 
  private:
-  ResultTable computeResult() override;
+  Result computeResult([[maybe_unused]] bool requestLaziness) override;
 
   // Implementation for the binding of arbitrary expressions.
   template <size_t IN_WIDTH, size_t OUT_WIDTH>
   void computeExpressionBind(
       IdTable* outputIdTable, LocalVocab* outputLocalVocab,
-      const ResultTable& inputResultTable,
+      const Result& inputResultTable,
       sparqlExpression::SparqlExpression* expression) const;
 
   [[nodiscard]] VariableToColumnMap computeVariableToColumnMap() const override;

src/engine/CMakeLists.txt

@@ -2,7 +2,7 @@ add_subdirectory(sparqlExpressions)
 add_library(SortPerformanceEstimator SortPerformanceEstimator.cpp)
 qlever_target_link_libraries(SortPerformanceEstimator)
 add_library(engine
-        Engine.cpp QueryExecutionTree.cpp Operation.cpp ResultTable.cpp LocalVocab.cpp
+        Engine.cpp QueryExecutionTree.cpp Operation.cpp Result.cpp LocalVocab.cpp
         IndexScan.cpp Join.cpp Sort.cpp
         Distinct.cpp OrderBy.cpp Filter.cpp
         Server.cpp QueryPlanner.cpp QueryPlanningCostFactors.cpp

src/engine/CartesianProductJoin.cpp

@@ -132,10 +132,11 @@ void CartesianProductJoin::writeResultColumn(std::span<Id> targetColumn,
   }
 }
 // ____________________________________________________________________________
-ResultTable CartesianProductJoin::computeResult() {
+Result CartesianProductJoin::computeResult(
+    [[maybe_unused]] bool requestLaziness) {
   IdTable result{getExecutionContext()->getAllocator()};
   result.setNumColumns(getResultWidth());
-  std::vector<std::shared_ptr<const ResultTable>> subResults;
+  std::vector<std::shared_ptr<const Result>> subResults;
 
   // We don't need to fully materialize the child results if we have a LIMIT
   // specified and an OFFSET of 0.
@@ -154,21 +155,22 @@ ResultTable CartesianProductJoin::computeResult() {
     }
     subResults.push_back(child.getResult());
     // Early stopping: If one of the results is empty, we can stop early.
-    if (subResults.back()->size() == 0) {
+    if (subResults.back()->idTable().size() == 0) {
       break;
     }
     // Example for the following calculation: If we have a LIMIT of 1000 and
     // the first child already has a result of size 100, then the second child
     // needs to evaluate only its first 10 results. The +1 is because integer
     // divisions are rounded down by default.
     if (limitIfPresent.has_value()) {
-      limitIfPresent.value()._limit =
-          limitIfPresent.value()._limit.value() / subResults.back()->size() + 1;
+      limitIfPresent.value()._limit = limitIfPresent.value()._limit.value() /
+                                          subResults.back()->idTable().size() +
+                                      1;
     }
   }
 
   auto sizesView = std::views::transform(
-      subResults, [](const auto& child) { return child->size(); });
+      subResults, [](const auto& child) { return child->idTable().size(); });
   auto totalResultSize = std::accumulate(sizesView.begin(), sizesView.end(),
                                          1UL, std::multiplies{});
 
@@ -210,7 +212,7 @@ ResultTable CartesianProductJoin::computeResult() {
   auto subResultsDeref = std::views::transform(
       subResults, [](auto& x) -> decltype(auto) { return *x; });
   return {std::move(result), resultSortedOn(),
-          ResultTable::getMergedLocalVocab(subResultsDeref)};
+          Result::getMergedLocalVocab(subResultsDeref)};
 }
 
 // ____________________________________________________________________________

src/engine/CartesianProductJoin.h

@@ -79,7 +79,7 @@ class CartesianProductJoin : public Operation {
 
  private:
   //! Compute the result of the query-subtree rooted at this element..
-  ResultTable computeResult() override;
+  Result computeResult([[maybe_unused]] bool requestLaziness) override;
 
   // Copy each element from the `inputColumn` `groupSize` times to the
   // `targetColumn`. Repeat until the `targetColumn` is copletely filled. Skip

src/engine/CountAvailablePredicates.cpp

@@ -100,7 +100,8 @@ size_t CountAvailablePredicates::getCostEstimate() {
 }
 
 // _____________________________________________________________________________
-ResultTable CountAvailablePredicates::computeResult() {
+Result CountAvailablePredicates::computeResult(
+    [[maybe_unused]] bool requestLaziness) {
   LOG(DEBUG) << "CountAvailablePredicates result computation..." << std::endl;
   IdTable idTable{getExecutionContext()->getAllocator()};
   idTable.setNumColumns(2);
@@ -137,7 +138,7 @@ ResultTable CountAvailablePredicates::computeResult() {
                                                              patterns);
     return {std::move(idTable), resultSortedOn(), LocalVocab{}};
   } else {
-    std::shared_ptr<const ResultTable> subresult = subtree_->getResult();
+    std::shared_ptr<const Result> subresult = subtree_->getResult();
     LOG(DEBUG) << "CountAvailablePredicates subresult computation done."
                << std::endl;
 

src/engine/CountAvailablePredicates.h

@@ -16,8 +16,8 @@
 using std::string;
 using std::vector;
 
-// This Operation takes a ResultTable with at least one column containing ids,
-// and a column index referring to such a column. It then creates a ResultTable
+// This Operation takes a Result with at least one column containing ids,
+// and a column index referring to such a column. It then creates a Result
 // containing two columns, the first one filled with the ids of all predicates
 // for which there is an entry in the index with one of the entities in the
 // specified input column as its subject. The second output column contains a
@@ -103,6 +103,6 @@ class CountAvailablePredicates : public Operation {
   void computePatternTrickAllEntities(
       IdTable* result, const CompactVectorOfStrings<Id>& patterns) const;
 
-  ResultTable computeResult() override;
+  Result computeResult([[maybe_unused]] bool requestLaziness) override;
   [[nodiscard]] VariableToColumnMap computeVariableToColumnMap() const override;
 };

src/engine/Distinct.cpp

@@ -7,6 +7,7 @@
 #include <sstream>
 
 #include "engine/CallFixedSize.h"
+#include "engine/Engine.h"
 #include "engine/QueryExecutionTree.h"
 
 using std::endl;
@@ -37,10 +38,10 @@ VariableToColumnMap Distinct::computeVariableToColumnMap() const {
 }
 
 // _____________________________________________________________________________
-ResultTable Distinct::computeResult() {
+Result Distinct::computeResult([[maybe_unused]] bool requestLaziness) {
   IdTable idTable{getExecutionContext()->getAllocator()};
   LOG(DEBUG) << "Getting sub-result for distinct result computation..." << endl;
-  shared_ptr<const ResultTable> subRes = _subtree->getResult();
+  std::shared_ptr<const Result> subRes = _subtree->getResult();
 
   LOG(DEBUG) << "Distinct result computation..." << endl;
   idTable.setNumColumns(subRes->idTable().numColumns());

src/engine/Distinct.h

@@ -55,7 +55,7 @@ class Distinct : public Operation {
   [[nodiscard]] string getCacheKeyImpl() const override;
 
  private:
-  virtual ResultTable computeResult() override;
+  virtual Result computeResult([[maybe_unused]] bool requestLaziness) override;
 
   VariableToColumnMap computeVariableToColumnMap() const override;
 };

src/engine/ExportQueryExecutionTrees.cpp

@@ -29,7 +29,7 @@ cppcoro::generator<QueryExecutionTree::StringTriple>
 ExportQueryExecutionTrees::constructQueryResultToTriples(
     const QueryExecutionTree& qet,
     const ad_utility::sparql_types::Triples& constructTriples,
-    LimitOffsetClause limitAndOffset, std::shared_ptr<const ResultTable> res,
+    LimitOffsetClause limitAndOffset, std::shared_ptr<const Result> res,
     CancellationHandle cancellationHandle) {
   for (size_t i : getRowIndices(limitAndOffset, res->idTable())) {
     ConstructQueryExportContext context{i, *res, qet.getVariableColumns(),
@@ -57,7 +57,7 @@ ad_utility::streams::stream_generator ExportQueryExecutionTrees::
         const QueryExecutionTree& qet,
         const ad_utility::sparql_types::Triples& constructTriples,
         LimitOffsetClause limitAndOffset,
-        std::shared_ptr<const ResultTable> resultTable,
+        std::shared_ptr<const Result> resultTable,
         CancellationHandle cancellationHandle) {
   resultTable->logResultSize();
   auto generator = ExportQueryExecutionTrees::constructQueryResultToTriples(
@@ -91,8 +91,7 @@ nlohmann::json
 ExportQueryExecutionTrees::constructQueryResultBindingsToQLeverJSON(
     const QueryExecutionTree& qet,
     const ad_utility::sparql_types::Triples& constructTriples,
-    const LimitOffsetClause& limitAndOffset,
-    std::shared_ptr<const ResultTable> res,
+    const LimitOffsetClause& limitAndOffset, std::shared_ptr<const Result> res,
     CancellationHandle cancellationHandle) {
   auto generator = constructQueryResultToTriples(qet, constructTriples,
                                                  limitAndOffset, std::move(res),
@@ -110,7 +109,7 @@ ExportQueryExecutionTrees::constructQueryResultBindingsToQLeverJSON(
 nlohmann::json ExportQueryExecutionTrees::idTableToQLeverJSONArray(
     const QueryExecutionTree& qet, const LimitOffsetClause& limitAndOffset,
     const QueryExecutionTree::ColumnIndicesAndTypes& columns,
-    std::shared_ptr<const ResultTable> resultTable,
+    std::shared_ptr<const Result> resultTable,
     CancellationHandle cancellationHandle) {
   AD_CORRECTNESS_CHECK(resultTable != nullptr);
   const IdTable& data = resultTable->idTable();
@@ -268,7 +267,7 @@ nlohmann::json ExportQueryExecutionTrees::selectQueryResultToSparqlJSON(
     const QueryExecutionTree& qet,
     const parsedQuery::SelectClause& selectClause,
     const LimitOffsetClause& limitAndOffset,
-    shared_ptr<const ResultTable> resultTable,
+    std::shared_ptr<const Result> resultTable,
     CancellationHandle cancellationHandle) {
   using nlohmann::json;
 
@@ -388,7 +387,7 @@ nlohmann::json ExportQueryExecutionTrees::selectQueryResultBindingsToQLeverJSON(
     const QueryExecutionTree& qet,
     const parsedQuery::SelectClause& selectClause,
     const LimitOffsetClause& limitAndOffset,
-    shared_ptr<const ResultTable> resultTable,
+    std::shared_ptr<const Result> resultTable,
     CancellationHandle cancellationHandle) {
   AD_CORRECTNESS_CHECK(resultTable != nullptr);
   LOG(DEBUG) << "Resolving strings for finished binary result...\n";
@@ -418,7 +417,7 @@ ExportQueryExecutionTrees::selectQueryResultToStream(
 
   // This call triggers the possibly expensive computation of the query result
   // unless the result is already cached.
-  shared_ptr<const ResultTable> resultTable = qet.getResult();
+  std::shared_ptr<const Result> resultTable = qet.getResult();
   resultTable->logResultSize();
   LOG(DEBUG) << "Converting result IDs to their corresponding strings ..."
              << std::endl;
@@ -563,7 +562,7 @@ ad_utility::streams::stream_generator ExportQueryExecutionTrees::
       selectClause.getSelectedVariablesAsStrings();
   // This call triggers the possibly expensive computation of the query result
   // unless the result is already cached.
-  shared_ptr<const ResultTable> resultTable = qet.getResult();
+  std::shared_ptr<const Result> resultTable = qet.getResult();
 
   // In the XML format, the variables don't include the question mark.
   auto varsWithoutQuestionMark = std::views::transform(
@@ -605,8 +604,7 @@ ad_utility::streams::stream_generator
 ExportQueryExecutionTrees::constructQueryResultToStream(
     const QueryExecutionTree& qet,
     const ad_utility::sparql_types::Triples& constructTriples,
-    LimitOffsetClause limitAndOffset,
-    std::shared_ptr<const ResultTable> resultTable,
+    LimitOffsetClause limitAndOffset, std::shared_ptr<const Result> resultTable,
     CancellationHandle cancellationHandle) {
   static_assert(format == MediaType::octetStream || format == MediaType::csv ||
                 format == MediaType::tsv || format == MediaType::sparqlXml);
@@ -638,11 +636,11 @@ nlohmann::json ExportQueryExecutionTrees::computeQueryResultAsQLeverJSON(
     const ParsedQuery& query, const QueryExecutionTree& qet,
     const ad_utility::Timer& requestTimer, uint64_t maxSend,
     CancellationHandle cancellationHandle) {
-  shared_ptr<const ResultTable> resultTable = qet.getResult();
+  std::shared_ptr<const Result> resultTable = qet.getResult();
   resultTable->logResultSize();
   auto timeResultComputation = requestTimer.msecs();
 
-  size_t resultSize = resultTable->size();
+  size_t resultSize = resultTable->idTable().size();
 
   nlohmann::json j;
 
@@ -725,7 +723,7 @@ nlohmann::json ExportQueryExecutionTrees::computeSelectQueryResultAsSparqlJSON(
     AD_THROW(
         "SPARQL-compliant JSON format is only supported for SELECT queries");
   }
-  shared_ptr<const ResultTable> resultTable = qet.getResult();
+  std::shared_ptr<const Result> resultTable = qet.getResult();
   resultTable->logResultSize();
   nlohmann::json j;
   auto limitAndOffset = query._limitOffset;

src/engine/ExportQueryExecutionTrees.h

@@ -111,7 +111,7 @@ class ExportQueryExecutionTrees {
       const QueryExecutionTree& qet,
       const parsedQuery::SelectClause& selectClause,
       const LimitOffsetClause& limitAndOffset,
-      shared_ptr<const ResultTable> resultTable,
+      std::shared_ptr<const Result> resultTable,
       CancellationHandle cancellationHandle);
 
   /**
@@ -132,31 +132,30 @@ class ExportQueryExecutionTrees {
   static nlohmann::json idTableToQLeverJSONArray(
       const QueryExecutionTree& qet, const LimitOffsetClause& limitAndOffset,
       const QueryExecutionTree::ColumnIndicesAndTypes& columns,
-      std::shared_ptr<const ResultTable> resultTable,
+      std::shared_ptr<const Result> resultTable,
       CancellationHandle cancellationHandle);
 
   // ___________________________________________________________________________
   static nlohmann::json constructQueryResultBindingsToQLeverJSON(
       const QueryExecutionTree& qet,
       const ad_utility::sparql_types::Triples& constructTriples,
       const LimitOffsetClause& limitAndOffset,
-      std::shared_ptr<const ResultTable> res,
-      CancellationHandle cancellationHandle);
+      std::shared_ptr<const Result> res, CancellationHandle cancellationHandle);
 
   // Generate an RDF graph for a CONSTRUCT query.
   static cppcoro::generator<QueryExecutionTree::StringTriple>
   constructQueryResultToTriples(
       const QueryExecutionTree& qet,
       const ad_utility::sparql_types::Triples& constructTriples,
-      LimitOffsetClause limitAndOffset, std::shared_ptr<const ResultTable> res,
+      LimitOffsetClause limitAndOffset, std::shared_ptr<const Result> res,
       CancellationHandle cancellationHandle);
 
   // ___________________________________________________________________________
   static nlohmann::json selectQueryResultToSparqlJSON(
       const QueryExecutionTree& qet,
       const parsedQuery::SelectClause& selectClause,
       const LimitOffsetClause& limitAndOffset,
-      shared_ptr<const ResultTable> resultTable,
+      std::shared_ptr<const Result> resultTable,
       CancellationHandle cancellationHandle);
 
   // ___________________________________________________________________________
@@ -165,7 +164,7 @@ class ExportQueryExecutionTrees {
       const QueryExecutionTree& qet,
       const ad_utility::sparql_types::Triples& constructTriples,
       LimitOffsetClause limitAndOffset,
-      std::shared_ptr<const ResultTable> resultTable,
+      std::shared_ptr<const Result> resultTable,
       CancellationHandle cancellationHandle);
 
   // _____________________________________________________________________________

src/engine/Filter.cpp

@@ -43,9 +43,9 @@ string Filter::getDescriptor() const {
 }
 
 // _____________________________________________________________________________
-ResultTable Filter::computeResult() {
+Result Filter::computeResult([[maybe_unused]] bool requestLaziness) {
   LOG(DEBUG) << "Getting sub-result for Filter result computation..." << endl;
-  shared_ptr<const ResultTable> subRes = _subtree->getResult();
+  std::shared_ptr<const Result> subRes = _subtree->getResult();
   LOG(DEBUG) << "Filter result computation..." << endl;
   checkCancellation();
 
@@ -63,7 +63,7 @@ ResultTable Filter::computeResult() {
 // _____________________________________________________________________________
 template <size_t WIDTH>
 void Filter::computeFilterImpl(IdTable* outputIdTable,
-                               const ResultTable& inputResultTable) {
+                               const Result& inputResultTable) {
   sparqlExpression::EvaluationContext evaluationContext(
       *getExecutionContext(), _subtree->getVariableColumns(),
       inputResultTable.idTable(), getExecutionContext()->getAllocator(),

src/engine/Filter.h

@@ -58,9 +58,9 @@ class Filter : public Operation {
     return _subtree->getVariableColumns();
   }
 
-  ResultTable computeResult() override;
+  Result computeResult([[maybe_unused]] bool requestLaziness) override;
 
   template <size_t WIDTH>
   void computeFilterImpl(IdTable* outputIdTable,
-                         const ResultTable& inputResultTable);
+                         const Result& inputResultTable);
 };

src/engine/GroupBy.cpp

@@ -222,9 +222,9 @@ void GroupBy::processGroup(
  * @param blockEnd Where the group ends.
  * @param input The input Table.
  * @param result
- * @param inTable The input ResultTable, which is required for its local
+ * @param inTable The input Result, which is required for its local
  *                vocabulary
- * @param outTable The output ResultTable, the vocabulary of which needs to be
+ * @param outTable The output Result, the vocabulary of which needs to be
  *                 expanded for GROUP_CONCAT aggregates
  * @param distinctHashSet An empty hash set. This is only passed in as an
  *                        argument to allow for efficient reusage of its
@@ -309,7 +309,7 @@ void GroupBy::doGroupBy(const IdTable& dynInput,
   *dynResult = std::move(result).toDynamic();
 }
 
-ResultTable GroupBy::computeResult() {
+Result GroupBy::computeResult([[maybe_unused]] bool requestLaziness) {
   LOG(DEBUG) << "GroupBy result computation..." << std::endl;
 
   IdTable idTable{getExecutionContext()->getAllocator()};
@@ -335,7 +335,7 @@ ResultTable GroupBy::computeResult() {
   auto hashMapOptimizationParams =
       checkIfHashMapOptimizationPossible(aggregates);
 
-  std::shared_ptr<const ResultTable> subresult;
+  std::shared_ptr<const Result> subresult;
   if (hashMapOptimizationParams.has_value()) {
     const auto* child = _subtree->getRootOperation()->getChildren().at(0);
     // Skip sorting