Merge branch 'master' into ubuntu-cmake

Author: apavlo

.travis.yml

@@ -133,7 +133,7 @@ script:
     # create build directory
     - mkdir build
     - cd build
-    # run cmake
+    # run cmake. NOTE: the PATH is made explicit to avoid automatic selection of the preinstalled llvm version in the Travis trusty image
     - PATH=/usr/lib/llvm-3.7/bin:/usr/bin:$PATH cmake -DCOVERALLS=$COVERALLS -DCMAKE_PREFIX_PATH=`llvm-config-3.7 --prefix` -DCMAKE_BUILD_TYPE=$PELOTON_BUILD_TYPE -DUSE_SANITIZER=Address ..
     # build
     - make -j4

Jenkinsfile

@@ -12,8 +12,8 @@ RUN mkdir /peloton/build && cd /peloton/build && cmake -DCMAKE_BUILD_TYPE=Debug
 RUN echo -n "Peloton Release build with "; g++ --version | head -1
 RUN rm -rf /peloton/build && mkdir /peloton/build && cd /peloton/build && cmake -DCMAKE_BUILD_TYPE=Release -DCOVERALLS=False .. && make -j4 && make install
 
-RUN echo -n "Peloton Debug build with "; clang++-3.7 --version | head -1
-RUN rm -rf /peloton/build && mkdir /peloton/build && cd /peloton/build && CC=clang-3.7 CXX=clang++-3.7 cmake -DCMAKE_BUILD_TYPE=Debug -DCOVERALLS=False .. && make -j4 && make install
+RUN echo -n "Peloton Debug build with "; clang++ --version | head -1
+RUN rm -rf /peloton/build && mkdir /peloton/build && cd /peloton/build && CC=clang CXX=clang++ cmake -DCMAKE_BUILD_TYPE=Debug -DCOVERALLS=False .. && make -j4 && make install
 
-RUN echo -n "Peloton Release build with "; clang++-3.7 --version | head -1
-RUN rm -rf /peloton/build && mkdir /peloton/build && cd /peloton/build && CC=clang-3.7 CXX=clang++-3.7 cmake -DCMAKE_BUILD_TYPE=Release -DCOVERALLS=False .. && make -j4 && make install
+RUN echo -n "Peloton Release build with "; clang++ --version | head -1
+RUN rm -rf /peloton/build && mkdir /peloton/build && cd /peloton/build && CC=clang CXX=clang++ cmake -DCMAKE_BUILD_TYPE=Release -DCOVERALLS=False .. && make -j4 && make install

script/docker/debian-stretch/Dockerfile-jenkins

@@ -77,6 +77,33 @@ if [ "$DISTRO" = "UBUNTU" ]; then
         libedit-dev \
         postgresql-client
 
+## ------------------------------------------------
+## DEBIAN
+## ------------------------------------------------
+elif [ "$DISTRO" = "DEBIAN OS" ]; then
+    sudo apt-get -qq --ignore-missing -y install \
+        git \
+        g++ \
+        clang \
+        cmake \
+        libgflags-dev \
+        libprotobuf-dev \
+        protobuf-compiler \
+        bison \
+        flex \
+        libevent-dev \
+        libboost-dev \
+        libboost-thread-dev \
+        libboost-filesystem-dev \
+        libjemalloc-dev \
+        libssl-dev \
+        valgrind \
+        lcov \
+        libpqxx-dev \
+        llvm-dev \
+        libedit-dev \
+        postgresql-client
+
 ## ------------------------------------------------
 ## FEDORA
 ## ------------------------------------------------
@@ -147,7 +174,7 @@ elif [[ "$DISTRO" == *"REDHAT"* ]] && [[ "${DISTRO_VER%.*}" == "7" ]]; then
         git \
         gcc-c++ \
         make \
-        cmake \
+        cmake3 \
         flex \
         bison \
         libevent-devel \

script/installation/packages.sh

@@ -82,21 +82,16 @@ bool BinderContext::GetColumnPosTuple(
     const std::string &col_name,
     std::shared_ptr<catalog::TableCatalogObject> table_obj,
     std::tuple<oid_t, oid_t, oid_t> &col_pos_tuple, type::TypeId &value_type) {
-  try {
-    auto column_object = table_obj->GetColumnObject(col_name);
-    if (column_object == nullptr) {
-      return false;
-    }
-
-    oid_t col_pos = column_object->GetColumnId();
-    col_pos_tuple = std::make_tuple(table_obj->GetDatabaseOid(),
-                                    table_obj->GetTableOid(), col_pos);
-    value_type = column_object->GetColumnType();
-    return true;
-  } catch (CatalogException &e) {
-    LOG_TRACE("Can't find table %d! Return false", std::get<1>(table_id_tuple));
+  auto column_object = table_obj->GetColumnObject(col_name);
+  if (column_object == nullptr) {
     return false;
   }
+
+  oid_t col_pos = column_object->GetColumnId();
+  col_pos_tuple = std::make_tuple(table_obj->GetDatabaseOid(),
+                                  table_obj->GetTableOid(), col_pos);
+  value_type = column_object->GetColumnType();
+  return true;
 }
 
 bool BinderContext::GetColumnPosTuple(

src/binder/binder_context.cpp

@@ -94,7 +94,7 @@ bool CreateExecutor::CreateDatabase(const planner::CreatePlan &node) {
       database_name, txn);
   txn->SetResult(result);
   LOG_TRACE("Result is: %s",
-            ResultTypeToString(current_txn->GetResult()).c_str());
+            ResultTypeToString(txn->GetResult()).c_str());
   return (true);
 }
 

src/executor/create_executor.cpp

@@ -138,6 +138,13 @@ class TupleValueExpression : public AbstractExpression {
 
   std::string GetColumnName() const { return col_name_; }
 
+  std::string GetColFullName() const {
+    if (!table_name_.empty()) {
+      return table_name_ + "." + col_name_;
+    }
+    return col_name_;
+  }
+
   void SetTableName(std::string table_alias) { table_name_ = table_alias; }
 
   void SetColName(std::string col_name) { col_name_ = col_name; }

src/include/expression/tuple_value_expression.h

@@ -0,0 +1,53 @@
+//===----------------------------------------------------------------------===//
+//
+//                         Peloton
+//
+// cost_and_stats_calculator.h
+//
+// Identification: src/include/optimizer/stats_calculator.h
+//
+// Copyright (c) 2015-16, Carnegie Mellon University Database Group
+//
+//===----------------------------------------------------------------------===//
+
+#pragma once
+
+#include "optimizer/operator_visitor.h"
+
+namespace peloton {
+
+namespace expression {
+class AbstractExpression;
+}
+namespace optimizer {
+
+class Memo;
+
+// Derive child stats that has not yet been calculated for a logical group
+// expression
+class ChildStatsDeriver : public OperatorVisitor {
+ public:
+  std::vector<ExprSet> DeriveInputStats(
+      GroupExpression *gexpr,
+      ExprSet required_cols, Memo *memo);
+
+  void Visit(const LogicalQueryDerivedGet *) override;
+  void Visit(const LogicalInnerJoin *) override;
+  void Visit(const LogicalLeftJoin *) override;
+  void Visit(const LogicalRightJoin *) override;
+  void Visit(const LogicalOuterJoin *) override;
+  void Visit(const LogicalSemiJoin *) override;
+  void Visit(const LogicalAggregateAndGroupBy *) override;
+
+ private:
+  void PassDownRequiredCols();
+  void PassDownColumn(expression::AbstractExpression* col);
+  ExprSet required_cols_;
+  GroupExpression *gexpr_;
+  Memo *memo_;
+
+  std::vector<ExprSet> output_;
+};
+
+}  // namespace optimizer
+}  // namespace peloton

src/include/optimizer/child_stats_deriver.h

@@ -17,12 +17,11 @@
 namespace peloton {
 namespace optimizer {
 
+class Memo;
 // Derive cost for a physical group expressionh
 class CostCalculator : public OperatorVisitor {
  public:
-  double CalculatorCost(
-      GroupExpression* gexpr,
-      const PropertySet *output_properties);
+  double CalculateCost(GroupExpression *gexpr, Memo *memo);
 
   void Visit(const DummyScan *) override;
   void Visit(const PhysicalSeqScan *) override;
@@ -48,10 +47,12 @@ class CostCalculator : public OperatorVisitor {
   void Visit(const PhysicalAggregate *) override;
 
  private:
-  // We cannot use reference here because otherwise we have to initialize them
-  // when constructing the class
-  GroupExpression* gexpr_;
-  const PropertySet *output_properties_;
+  double HashCost();
+  double SortCost();
+  double GroupByCost();
+
+  GroupExpression *gexpr_;
+  Memo *memo_;
   double output_cost_ = 0;
 };