Proof of concept for SPARQL UPDATE

.gitignore

@@ -1,7 +1,11 @@
-# Build directory
+# Build directories
 build/
+debug/
 cmake-build*
 
+# Debugger history file
+.gdb_history
+
 # End-to-End data
 e2e_data/*
 # Compiled Object files

src/engine/Server.cpp

@@ -1,4 +1,4 @@
-// Copyright 2011 - 2022, University of Freiburg
+// Copyright 2011 - 2023, University of Freiburg
 // Chair of Algorithms and Data Structures
 // Authors: Björn Buchhold <[email protected]>
 //          Johannes Kalmbach <[email protected]>
@@ -13,6 +13,7 @@
 
 #include "engine/ExportQueryExecutionTrees.h"
 #include "engine/QueryPlanner.h"
+#include "parser/TurtleParser.h"
 #include "util/BoostHelpers/AsyncWaitForFuture.h"
 #include "util/OnDestructionDontThrowDuringStackUnwinding.h"
 
@@ -310,11 +311,63 @@ Awaitable<void> Server::process(
     logCommand(cmd, "clear cache completely (including unpinned elements)");
     cache_.clearAll();
     response = createJsonResponse(composeCacheStatsJson(), request);
+  } else if (auto cmd = checkParameter("cmd", "clear-delta-triples")) {
+    logCommand(cmd, "clear delta triples");
+    index_.deltaTriples().clear();
+    response = createJsonResponse(composeStatsJson(), request);
   } else if (auto cmd = checkParameter("cmd", "get-settings")) {
     logCommand(cmd, "get server settings");
     response = createJsonResponse(RuntimeParameters().toMap(), request);
   }
 
+  // Insert or delete triples.
+  //
+  // TODO: This is a preliminary interface for testing. Eventually, this should
+  // be included in our SPARQL grammer (where the line `updateUnit : update;` at
+  // the beginning is currently commented out).
+  //
+  // TODO: For testing purposes, allow insertions and deletions without access
+  // token. Eventually, this should be restricted, of course, which can be
+  // easily done by adding the argument `accessTokenOk` to each of the calls for
+  // `checkParameter`.
+  {
+    bool insertDetected = false;
+    bool deleteDetected = false;
+    std::optional<std::string> parameterValue;
+    if ((parameterValue = checkParameter("insert", std::nullopt))) {
+      LOG(INFO) << "INSERT: " << parameterValue.value() << std::endl;
+      insertDetected = true;
+    } else if ((parameterValue = checkParameter("delete", std::nullopt))) {
+      LOG(INFO) << "DELETE: " << parameterValue.value() << std::endl;
+      deleteDetected = true;
+    }
+    if (insertDetected || deleteDetected) {
+      AD_CORRECTNESS_CHECK(parameterValue.has_value());
+      const std::string& input = parameterValue.value();
+      TurtleStringParser<Tokenizer> parser;
+      parser.parseUtf8String(input);
+      if (parser.getTriples().size() == 0) {
+        throw std::runtime_error("Triple could not be parsed");
+      } else if (parser.getTriples().size() > 1) {
+        throw std::runtime_error("Only one triple per call please");
+      }
+      TurtleTriple turtleTriple = parser.getTriples()[0];
+      if (insertDetected) {
+        index_.deltaTriples().insertTriple(std::move(turtleTriple));
+        response =
+            createOkResponse(absl::StrCat("INSERT operation for triple \"",
+                                          input, "\" processed\n"),
+                             request, ad_utility::MediaType::textPlain);
+      } else {
+        index_.deltaTriples().deleteTriple(std::move(turtleTriple));
+        response =
+            createOkResponse(absl::StrCat("DELETE operation for triple \"",
+                                          input, "\" processed\n"),
+                             request, ad_utility::MediaType::textPlain);
+      }
+    }
+  }
+
   // Ping with or without messsage.
   if (urlPathAndParameters._path == "/ping") {
     if (auto msg = checkParameter("msg", std::nullopt)) {
@@ -455,6 +508,8 @@ json Server::composeStatsJson() const {
   result["num-text-records"] = index_.getNofTextRecords();
   result["num-word-occurrences"] = index_.getNofWordPostings();
   result["num-entity-occurrences"] = index_.getNofEntityPostings();
+  result["num-delta-triples-inserted"] = index_.deltaTriples().numInserted();
+  result["num-delta-triples-deleted"] = index_.deltaTriples().numDeleted();
   return result;
 }
 

src/engine/Server.h

@@ -13,6 +13,7 @@
 #include "engine/QueryExecutionContext.h"
 #include "engine/QueryExecutionTree.h"
 #include "engine/SortPerformanceEstimator.h"
+#include "index/DeltaTriples.h"
 #include "index/Index.h"
 #include "nlohmann/json.hpp"
 #include "parser/ParseException.h"

src/global/Id.h

@@ -8,9 +8,9 @@
 #include <cstdint>
 #include <limits>
 
-#include "../util/Exception.h"
-#include "./IndexTypes.h"
-#include "./ValueId.h"
+#include "global/IndexTypes.h"
+#include "global/ValueId.h"
+#include "util/Exception.h"
 
 using Id = ValueId;
 typedef uint16_t Score;

src/global/IdTriple.h

@@ -0,0 +1,18 @@
+// Copyright 2023, University of Freiburg
+// Chair of Algorithms and Data Structures
+// Authors: Hannah Bast <[email protected]>
+
+#pragma once
+
+#include <array>
+
+#include "global/Id.h"
+
+// Should we have an own class for this? We need this at several places.
+using IdTriple = std::array<Id, 3>;
+
+// Hash value for such triple.
+template <typename H>
+H AbslHashValue(H h, const IdTriple& triple) {
+  return H::combine(std::move(h), triple[0], triple[1], triple[2]);
+}

src/global/ValueId.h

@@ -282,10 +282,10 @@ class ValueId {
   /// This operator is only for debugging and testing. It returns a
   /// human-readable representation.
   friend std::ostream& operator<<(std::ostream& ostr, const ValueId& id) {
-    ostr << toString(id.getDatatype()) << ':';
+    ostr << toString(id.getDatatype())[0] << ':';
     auto visitor = [&ostr]<typename T>(T&& value) {
       if constexpr (ad_utility::isSimilar<T, ValueId::UndefinedType>) {
-        ostr << "Undefined";
+        ostr << "xx";
       } else if constexpr (ad_utility::isSimilar<T, double> ||
                            ad_utility::isSimilar<T, int64_t>) {
         ostr << std::to_string(value);

src/index/CMakeLists.txt

@@ -8,6 +8,8 @@ add_library(index
         VocabularyOnDisk.h VocabularyOnDisk.cpp
         IndexMetaData.h IndexMetaDataImpl.h
         MetaDataHandler.h
+        LocatedTriples.h LocatedTriples.cpp
+        DeltaTriples.h DeltaTriples.cpp
         StxxlSortFunctors.h
         TextMetaData.cpp TextMetaData.h
         DocsDB.cpp DocsDB.h