Add ability to build with clang

.github/composite-actions/download-libraries/action.yml

@@ -11,11 +11,26 @@ inputs:
     description: 'Download googletest'
     default: true
 
-  install-boost:
+  install-boost-gcc:
     type: boolean
-    description: 'Install boost'
+    description: 'Install boost built with GCC'
     default: true
 
+  install-boost-clang:
+    type: boolean
+    description: 'Install boost built with clang'
+    default: false
+
+  install-gcc:
+    type: boolean
+    description: 'Install GCC toolset (compiler and build tools)'
+    default: true
+
+  install-clang:
+    type: boolean
+    description: 'Install clang toolset (compiler and build tools)'
+    default: false
+
 runs:
   using: 'composite'
   steps:
@@ -25,6 +40,19 @@ runs:
         sudo apt-get update -y
         sudo apt-get install gcc-10 g++-10 cmake build-essential -y
       shell: bash
+      if: inputs.install-gcc != 'false'
+
+    - name: Install clang
+      # llvm.sh installs all needed libraries, no need in build-essential
+      # "all" option is needed to install libc++ and libc++abi
+      run: |
+        sudo apt-get update -y
+        sudo apt-get install cmake make -y
+        wget https://apt.llvm.org/llvm.sh
+        chmod +x llvm.sh
+        sudo ./llvm.sh 17 all
+      shell: bash
+      if: inputs.install-clang != 'false'
 
     - name: Make lib directory
       run: |
@@ -35,7 +63,7 @@ runs:
       uses: ./.github/composite-actions/download-library
       with:
         directory: googletest
-        download-command: git clone https://github.com/google/googletest/ --branch release-1.12.1 --depth 1
+        download-command: git clone https://github.com/google/googletest/ --branch release-1.14.0 --depth 1
       if: inputs.download-googletest != 'false'
 
     - name: Download easyloggingpp
@@ -72,13 +100,22 @@ runs:
         directory: boost
         download-command: wget -O boost_1_81_0.tar.gz https://sourceforge.net/projects/boost/files/boost/1.81.0/boost_1_81_0.tar.gz/download && tar xzvf boost_1_81_0.tar.gz && mv boost_1_81_0 boost
 
-    - name: Install Boost
+    - name: Install Boost built with GCC
       run: |
         cd lib/boost
         ./bootstrap.sh --prefix=/usr
         sudo ./b2 install --prefix=/usr
       shell: bash
-      if: inputs.install-boost != 'false'
+      if: inputs.install-boost-gcc != 'false'
+    - name: Install Boost built with clang
+      run: |
+        cd lib/boost
+        ./bootstrap.sh --prefix=/usr --with-toolset=clang
+        ./b2 clean
+        ./b2 toolset=clang cxxflags="-stdlib=libc++" linkflags="-stdlib=libc++"
+        sudo ./b2 install --prefix=/usr
+      shell: bash
+      if: inputs.install-boost-clang != 'false'
     - name: Download frozen
       uses: ./.github/composite-actions/download-library
       with:

.github/workflows/core-tests.yml

@@ -27,7 +27,7 @@ on:
       #- examples/**
   workflow_dispatch:
 jobs:
-  run_tests:
+  run_tests_gcc:
     runs-on: ubuntu-latest
     strategy:
       matrix:
@@ -44,6 +44,43 @@ jobs:
         uses: ./.github/composite-actions/download-datasets
       - name: Build
         run: |
+          export CC=gcc-10
+          export CXX=g++-10
+          if [[ "${{matrix.cfg.BUILD_TYPE}}" == "Debug" ]]; then
+            ./build.sh --debug --sanitizer=${{ matrix.cfg.SANITIZER }}
+          else
+            ./build.sh
+          fi
+      - name: Test
+        working-directory: ${{github.workspace}}/build/target
+        shell: bash
+        run: ./Desbordante_test --gtest_filter='*:-*HeavyDatasets*'
+  run_tests_clang:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        cfg:
+          - { BUILD_TYPE: Release }
+          - { BUILD_TYPE: Debug }
+          - { BUILD_TYPE: Debug, SANITIZER : ADDRESS }
+          - { BUILD_TYPE: Debug, SANITIZER : UB }
+    steps:
+      - uses: actions/checkout@v3
+      - name: Download libraries
+        uses: ./.github/composite-actions/download-libraries
+        with:
+          install-gcc: false
+          install-clang: true
+          install-boost-gcc: false
+          install-boost-clang: true
+      - name: Download datasets
+        uses: ./.github/composite-actions/download-datasets
+      - name: Build
+        run: |
+          export CC=clang-17
+          export CXX=clang++-17
+          export CXXFLAGS="-stdlib=libc++"
+          export LDFLAGS="-lc++abi"
           if [[ "${{matrix.cfg.BUILD_TYPE}}" == "Debug" ]]; then
             ./build.sh --debug --sanitizer=${{ matrix.cfg.SANITIZER }}
           else

.github/workflows/wheel.yml

@@ -75,7 +75,7 @@ jobs:
         with:
           download-pybind: true
           download-googletest: false
-          install-boost: false
+          install-boost-gcc: false
 
       - name: Build wheels
         uses: pypa/[email protected]

CMakeLists.txt

@@ -66,6 +66,22 @@ else()
     if (ASAN)
         # Set DEBUG build options specific for build with ASAN
         set(ASAN_OPTS "-fsanitize=address")
+
+        find_program(APT_FOUND apt-get)
+        if (APT_FOUND AND CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
+            # alloc-dealloc-mismatch generates false positives on boost exceptions
+            # This applies only to Ubuntu package:
+            # https://github.com/llvm/llvm-project/issues/59432?ysclid=m4y0iqca2c577414782
+            # Disable this check on files listed in address_sanitizer_ignore_list.txt if compiler
+            # is Clang and apt-get is installed on system:
+            # FIXME(senichenkov): apt-get is not an ideal check -- maybe it wouldn't be so hard to
+            # ask apt-get if repository is ubuntu-...?
+            message("Running on Ubuntu")
+            message(WARNING "ASAN is broken in Ubuntu package, therefore alloc-dealloc-mismatch")
+            message(WARNING "check will be supressed. Consider using another distro for full ASAN coverage.")
+            string(JOIN ";" ASAN_OPTS "-fsanitize-ignorelist=${CMAKE_SOURCE_DIR}/address_sanitizer_ignore_list.txt")
+        endif()
+
         string(JOIN ";" DEBUG_BUILD_OPTS "${DEBUG_BUILD_OPTS}"
             "-O1"
             "-Wno-error" # Use of -Werror is discouraged with sanitizers
@@ -95,6 +111,16 @@ else()
         add_compile_options(-ggdb3)
     endif()
 
+    # Workaround clang-18 bug:
+    # https://github.com/llvm/llvm-project/issues/76515?ysclid=m406q4it5k674680045
+    if (CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
+        string(FIND "${CMAKE_CXX_COMPILER_VERSION}" "18" IDX)
+        if (IDX EQUAL 0)  # clang major version is 18
+            message(WARNING "C++ compiler is Clang++-18. Supressing deprecated declaration warnings. Consider using another version of Clang")
+            string(JOIN ";" DEBUG_BUILD_OPTS "${DEBUG_BUILD_OPTS}" "-Wno-deprecated-declarations")
+        endif()
+    endif()
+
     add_compile_options("$<$<CONFIG:Debug>:${DEBUG_BUILD_OPTS}>")
     add_link_options("$<$<CONFIG:Debug>:${DEBUG_LINK_OPTS}>")
 
@@ -122,7 +148,7 @@ include_directories(SYSTEM "lib/easyloggingpp/src" "lib/better-enums/" "lib/emha
 
 # adding submodules
 if (COMPILE_TESTS)
-    add_subdirectory("lib/googletest")
+    add_subdirectory("lib/googletest" SYSTEM)
 endif()
 
 set( CMAKE_BUILD_TYPE_COPY "${CMAKE_BUILD_TYPE}" )

README.md

@@ -245,15 +245,16 @@ The following instructions were tested on Ubuntu 20.04+ LTS and macOS Sonoma 14.
 ### Dependencies
 Prior to cloning the repository and attempting to build the project, ensure that you have the following software:
 
-- GNU GCC, version 10+
+- GNU GCC, version 10+ or Clang, version 16+
 - CMake, version 3.13+
-- Boost library built with GCC, version 1.81.0+
+- Boost library built with compiler you're going to use (GCC or Clang), version 1.81.0+
 
 To use test datasets you will need:
 - Git Large File Storage, version 3.0.2+
 
 #### Ubuntu dependencies installation
 
+##### GCC
 Run the following commands:
 ```sh 
 sudo apt install gcc g++ cmake libboost-all-dev git-lfs
@@ -263,6 +264,45 @@ export CXX=g++
 The last 2 lines set gcc as CMake compiler in your terminal session.
 You can also add them to the end of `~/.profile` to set this by default in all sessions.
 
+##### Clang
+Firstly, you'll need to build Boost with Clang, as packaged versions, distributed by package managers, are built with GCC and have different ABI.
+Instructions below are given for Boost-1.81.0. If you want to use another version, you'll need to change `tar ...` command on step 3.
+For further details on Boost installation, please consult [Boost documentation](https://www.boost.org/doc/libs/1_81_0/more/getting_started/unix-variants.html).
+1) It's recommended to install Boost into `/usr/local`. You can use any other location, but you'll need to adapt instructions for it.
+```sh
+cd /usr/local
+```
+2) Download an official Boost distribuition from [SourceForge](https://sourceforge.net/projects/boost/files/boost/1.81.0/)
+3) Unpack downloaded archive:
+```sh
+tar --bzip2 -xf boost_1_81_0.tar.bz2
+```
+4) Compile Boost:
+```sh
+./bootstrap --with-toolset=clang
+./b2 clean
+./b2 toolset=clang cxxflags="-stdlib=libc++" linkflags="-stdlib=libc++"
+```
+5) Install Boost:
+```sh
+sudo ./b2 install
+```
+
+Run the following commands:
+```sh
+sudo apt install cmake git-lfs
+bash -c "$(wget -O - https://apt.llvm.org/llvm.sh)"
+export CC=clang
+export CXX=clang++
+export CXXFLAGS="-stdlib=libc++"
+# libc++ is fully compatible with GCC's ABI, so you can omit the next line if you want to use libstdc++ ABI:
+export LDFLAGS="-lc++abi"
+export BOOST_ROOT="/usr/local/"
+```
+Second command installs the latest version of LLVM (which includes Clang). For other installation options, see [LLVM packages page](https://apt.llvm.org/).
+The last 5 lines set Clang as CMake compiler in your terminal session and set directory where Boost libraries are located.
+You can also add them to the end of `~/.profile` to set this by default in all sessions.
+
 #### MacOS dependencies installation
 
 Install Xcode Command Line Tools if you don't have them. Run:

address_sanitizer_ignore_list.txt

@@ -0,0 +1,4 @@
+# Disable alloc_dealloc_mismatch ASAN check
+[alloc_dealloc_mismatch]
+# in file:
+src:typed_column_data.h

build.sh

@@ -1,5 +1,8 @@
 #!/bin/bash
 
+# Stop on error:
+set -e
+
 function print_help() {
 cat << EOF
 Usage: ./build.sh [options]
@@ -85,7 +88,7 @@ if [[ $NO_TESTS == true ]]; then
   PREFIX="$PREFIX -D COMPILE_TESTS=OFF"
 else
   if [[ ! -d "googletest" ]] ; then
-    git clone https://github.com/google/googletest/ --branch v1.13.0 --depth 1
+    git clone https://github.com/google/googletest/ --branch v1.14.0 --depth 1
   fi
 fi
 
@@ -116,5 +119,5 @@ fi
 cd ..
 mkdir -p build
 cd build
-rm CMakeCache.txt
+rm -f CMakeCache.txt
 cmake $PREFIX .. && make $JOBS_OPTION

src/core/algorithms/fd/fdep/fd_tree_element.cpp

@@ -1,6 +1,7 @@
 #include "fd_tree_element.h"
 
 #include "boost/dynamic_bitset.hpp"
+#include "util/bitset_extensions.h"
 
 FDTreeElement::FDTreeElement(size_t max_attribute_number)
     : max_attribute_number_(max_attribute_number) {
@@ -45,7 +46,7 @@ bool FDTreeElement::ContainsGeneralization(std::bitset<kMaxAttrNum> const& lhs,
         return true;
     }
 
-    size_t next_set_attr = lhs._Find_next(current_attr);
+    size_t next_set_attr = util::FindNext(lhs, current_attr);
     if (next_set_attr == kMaxAttrNum) {
         return false;
     }
@@ -71,7 +72,7 @@ bool FDTreeElement::GetGeneralizationAndDelete(std::bitset<kMaxAttrNum> const& l
         return true;
     }
 
-    size_t next_set_attr = lhs._Find_next(current_attr);
+    size_t next_set_attr = util::FindNext(lhs, current_attr);
     if (next_set_attr == kMaxAttrNum) {
         return false;
     }
@@ -104,7 +105,7 @@ bool FDTreeElement::GetSpecialization(std::bitset<kMaxAttrNum> const& lhs, size_
 
     bool found = false;
     size_t attr = (current_attr > 1 ? current_attr : 1);
-    size_t next_set_attr = lhs._Find_next(current_attr);
+    size_t next_set_attr = util::FindNext(lhs, current_attr);
 
     if (next_set_attr == kMaxAttrNum) {
         while (!found && attr <= this->max_attribute_number_) {
@@ -153,7 +154,8 @@ void FDTreeElement::AddFunctionalDependency(std::bitset<kMaxAttrNum> const& lhs,
     FDTreeElement* current_node = this;
     this->AddRhsAttribute(attr_num);
 
-    for (size_t i = lhs._Find_first(); i != kMaxAttrNum; i = lhs._Find_next(i)) {
+    auto iter = util::MakeBitsetIterator(lhs);
+    for (size_t i = iter->Pos(); i != kMaxAttrNum; iter->Next(), i = iter->Pos()) {
         if (current_node->children_[i - 1] == nullptr) {
             current_node->children_[i - 1] =
                     std::make_unique<FDTreeElement>(this->max_attribute_number_);
@@ -215,8 +217,8 @@ void FDTreeElement::PrintDependencies(std::bitset<kMaxAttrNum>& active_path, std
         if (this->is_fd_[attr - 1]) {
             out = "{";
 
-            for (size_t i = active_path._Find_first(); i != kMaxAttrNum;
-                 i = active_path._Find_next(i)) {
+            auto iter = util::MakeBitsetIterator(active_path);
+            for (size_t i = iter->Pos(); i != kMaxAttrNum; iter->Next(), i = iter->Pos()) {
                 if (!column_id.empty())
                     out += column_id + std::to_string(std::stoi(column_names[i - 1]) + 1) + ",";
                 else
@@ -257,11 +259,8 @@ void FDTreeElement::TransformTreeFdCollection(std::bitset<kMaxAttrNum>& active_p
 
     for (size_t attr = 1; attr <= this->max_attribute_number_; ++attr) {
         if (this->is_fd_[attr - 1]) {
-            boost::dynamic_bitset<> lhs_bitset(this->max_attribute_number_);
-            for (size_t i = active_path._Find_first(); i != kMaxAttrNum;
-                 i = active_path._Find_next(i)) {
-                lhs_bitset.set(i - 1);
-            }
+            auto lhs_bitset =
+                    util::CreateShiftedDynamicBitset(active_path, this->max_attribute_number_);
             Vertical lhs(scheme.get(), lhs_bitset);
             Column rhs(scheme.get(), scheme->GetColumn(attr - 1)->GetName(), attr - 1);
             fd_collection.emplace_back(FD{lhs, rhs, scheme});

src/core/algorithms/fd/fdep/fdep.cpp

@@ -5,6 +5,7 @@
 #include "config/equal_nulls/option.h"
 #include "config/tabular_data/input_table/option.h"
 #include "model/table/column_layout_relation_data.h"
+#include "util/bitset_extensions.h"
 
 // #ifndef PRINT_FDS
 // #define PRINT_FDS
@@ -96,8 +97,9 @@ void FDep::AddViolatedFDs(std::vector<size_t> const& t1, std::vector<size_t> con
     }
 
     equal_attr &= (~diff_attr);
-    for (size_t attr = diff_attr._Find_first(); attr != FDTreeElement::kMaxAttrNum;
-         attr = diff_attr._Find_next(attr)) {
+    auto iter = util::MakeBitsetIterator(diff_attr);
+    for (size_t attr = iter->Pos(); attr != FDTreeElement::kMaxAttrNum;
+         iter->Next(), attr = iter->Pos()) {
         this->neg_cover_tree_->AddFunctionalDependency(equal_attr, attr);
     }
 }