Fix size() and namespace lookups so that we can compile with C++17

.github/workflows/main.yml

@@ -1,4 +1,4 @@
-name: Windows RTools 3.5
+name: Windows RTools 4.0
 
 on:
   pull_request:
@@ -22,19 +22,17 @@ jobs:
       with:
         token: ${{ secrets.GITHUB_TOKEN }}
       if: "!startsWith(github.ref, 'refs/tags/') && github.ref != 'refs/heads/master' && github.ref != 'refs/heads/develop'"
-  
+
     - uses: actions/checkout@v2
     - uses: actions/setup-python@v2
       with:
         python-version: '2.x'
-    - name: Download RTools
-      run: Invoke-WebRequest -Uri https://cran.rstudio.com/bin/windows/Rtools/Rtools35.exe -OutFile ./R35.exe
-    - name: Install RTools
-      shell: powershell
-      run: Start-Process -FilePath ./R35.exe -ArgumentList /VERYSILENT -NoNewWindow -Wait
-    - name: PATH Setup
-      shell: powershell
-      run: echo "C:/Rtools/bin;C:/Rtools/mingw_64/bin" | Out-File -Append -FilePath $env:GITHUB_PATH -Encoding utf8
+    - uses: r-lib/actions/setup-r@v1
+      with:
+        r-version: 'release'
+    - name: Set path for RTools 4.0
+      if: runner.os == 'Windows'
+      run: echo "C:/rtools40/usr/bin;C:/rtools40/mingw64/bin" | Out-File -Append -FilePath $env:GITHUB_PATH -Encoding utf8
     - name: Print g++ & make version and path
       shell: powershell
       run: |
@@ -56,7 +54,7 @@ jobs:
         python.exe runTests.py -j2 test/unit/math/prim
         python.exe runTests.py -j2 test/unit/math/rev
         python.exe runTests.py -j2 test/unit/math/memory
-        
+
     - name: Upload gtest_output xml
       uses: actions/upload-artifact@v2
       if: failure()
@@ -72,14 +70,12 @@ jobs:
     - uses: actions/setup-python@v2
       with:
         python-version: '2.x'
-    - name: Download RTools
-      run: Invoke-WebRequest -Uri https://cran.rstudio.com/bin/windows/Rtools/Rtools35.exe -OutFile ./R35.exe
-    - name: Install RTools
-      shell: powershell
-      run: Start-Process -FilePath ./R35.exe -ArgumentList /VERYSILENT -NoNewWindow -Wait
-    - name: PATH Setup
-      shell: powershell
-      run: echo "C:/Rtools/bin;C:/Rtools/mingw_64/bin" | Out-File -Append -FilePath $env:GITHUB_PATH -Encoding utf8
+    - uses: r-lib/actions/setup-r@v1
+      with:
+        r-version: 'release'
+    - name: Set path for RTools 4.0
+      if: runner.os == 'Windows'
+      run: echo "C:/rtools40/usr/bin;C:/rtools40/mingw64/bin" | Out-File -Append -FilePath $env:GITHUB_PATH -Encoding utf8
     - name: Print g++ & make version and path
       shell: powershell
       run: |
@@ -93,9 +89,6 @@ jobs:
     - name: Add TBB to PATH
       shell: powershell
       run: echo "D:/a/math/math/lib/tbb" | Out-File -Append -FilePath $env:GITHUB_PATH -Encoding utf8
-    - name: Disable running fwd/mix tests
-      shell: powershell
-      run: echo "CXXFLAGS+= -DSTAN_MATH_TESTS_REV_ONLY" | Out-File -Append -FilePath make/local -Encoding utf8      
     - name: Run fwd unit tests and all the mix tests except those in mix/fun
       shell: powershell
       run: |
@@ -121,14 +114,12 @@ jobs:
     - uses: actions/setup-python@v2
       with:
         python-version: '2.x'
-    - name: Download RTools
-      run: Invoke-WebRequest -Uri https://cran.rstudio.com/bin/windows/Rtools/Rtools35.exe -OutFile ./R35.exe
-    - name: Install RTools
-      shell: powershell
-      run: Start-Process -FilePath ./R35.exe -ArgumentList /VERYSILENT -NoNewWindow -Wait
-    - name: PATH Setup
-      shell: powershell
-      run: echo "C:/Rtools/bin;C:/Rtools/mingw_64/bin" | Out-File -Append -FilePath $env:GITHUB_PATH -Encoding utf8
+    - uses: r-lib/actions/setup-r@v1
+      with:
+        r-version: 'release'
+    - name: Set path for RTools 4.0
+      if: runner.os == 'Windows'
+      run: echo "C:/rtools40/usr/bin;C:/rtools40/mingw64/bin" | Out-File -Append -FilePath $env:GITHUB_PATH -Encoding utf8
     - name: Print g++ & make version and path
       shell: powershell
       run: |
@@ -142,14 +133,11 @@ jobs:
     - name: Add TBB to PATH
       shell: powershell
       run: echo "D:/a/math/math/lib/tbb" | Out-File -Append -FilePath $env:GITHUB_PATH -Encoding utf8
-    - name: Disable running fwd/mix tests
-      shell: powershell
-      run: echo "CXXFLAGS+= -DSTAN_MATH_TESTS_REV_ONLY" | Out-File -Append -FilePath make/local -Encoding utf8      
     - name: Run mix/fun unit tests
       shell: powershell
       run: |
         python.exe runTests.py test/unit/math/mix/fun
-        
+
     - name: Upload gtest_output xml
       uses: actions/upload-artifact@v2
       if: failure()

Jenkinsfile

@@ -256,7 +256,8 @@ pipeline {
             }
             steps {
                 unstash 'MathSetup'
-	            sh "echo CXXFLAGS += -fsanitize=address >> make/local"
+                sh "echo CXX=${CLANG_CXX} > make/local"
+	              sh "echo CXXFLAGS += -fsanitize=address >> make/local"
                 script {
                     if (isUnix()) {
                         runTests("test/unit", false)

make/compiler_flags

@@ -116,7 +116,7 @@ CPPFLAGS_SUNDIALS ?= -DNO_FPRINTF_OUTPUT $(CPPFLAGS_OPTIM_SUNDIALS) $(CXXFLAGS_F
 
 
 ## setup compiler flags
-CXXFLAGS_LANG ?= -std=c++1y
+CXXFLAGS_LANG ?= -std=c++17
 #CXXFLAGS_BOOST ?=
 CXXFLAGS_SUNDIALS ?= -pipe $(CXXFLAGS_OPTIM_SUNDIALS) $(CPPFLAGS_FLTO_SUNDIALS)
 #CXXFLAGS_GTEST

stan/math/fwd/functor/operands_and_partials.hpp

@@ -111,10 +111,10 @@ class operands_and_partials<Op1, Op2, Op3, Op4, Op5, fvar<Dx>> {
 namespace internal {
 
 // Vectorized Univariate
-template <typename Dx>
-class ops_partials_edge<Dx, std::vector<fvar<Dx>>> {
+// std::vector<fvar<Dx>>
+template <typename Dx, typename Op>
+class ops_partials_edge<Dx, Op, require_std_vector_vt<is_fvar, Op>> {
  public:
-  using Op = std::vector<fvar<Dx>>;
   using partials_t = Eigen::Matrix<Dx, -1, 1>;
   partials_t partials_;                       // For univariate use-cases
   broadcast_array<partials_t> partials_vec_;  // For multivariate
@@ -137,11 +137,12 @@ class ops_partials_edge<Dx, std::vector<fvar<Dx>>> {
   }
 };
 
-template <typename Dx, int R, int C>
-class ops_partials_edge<Dx, Eigen::Matrix<fvar<Dx>, R, C>> {
+// Eigen::Matrix<fvar<Dx>, R, C>
+template <typename Dx, typename Op>
+class ops_partials_edge<Dx, Op, require_eigen_vt<is_fvar, Op>> {
  public:
-  using partials_t = Eigen::Matrix<Dx, R, C>;
-  using Op = Eigen::Matrix<fvar<Dx>, R, C>;
+  using partials_t
+      = Eigen::Matrix<Dx, Op::RowsAtCompileTime, Op::ColsAtCompileTime>;
   partials_t partials_;                       // For univariate use-cases
   broadcast_array<partials_t> partials_vec_;  // For multivariate
   explicit ops_partials_edge(const Op& ops)
@@ -164,11 +165,15 @@ class ops_partials_edge<Dx, Eigen::Matrix<fvar<Dx>, R, C>> {
 };
 
 // Multivariate; vectors of eigen types
-template <typename Dx, int R, int C>
-class ops_partials_edge<Dx, std::vector<Eigen::Matrix<fvar<Dx>, R, C>>> {
+// std::vector<Eigen::Matrix<fvar<Dx>, R, C>>
+template <typename Dx, typename Op>
+class ops_partials_edge<
+    Dx, Op,
+    require_all_t<is_std_vector<Op>, is_eigen<value_type_t<Op>>,
+                  is_fvar<value_type_t<value_type_t<Op>>>>> {
  public:
-  using Op = std::vector<Eigen::Matrix<fvar<Dx>, R, C>>;
-  using partial_t = Eigen::Matrix<Dx, R, C>;
+  using partial_t = Eigen::Matrix<Dx, value_type_t<Op>::RowsAtCompileTime,
+                                  value_type_t<Op>::ColsAtCompileTime>;
   std::vector<partial_t> partials_vec_;
   explicit ops_partials_edge(const Op& ops)
       : partials_vec_(ops.size()), operands_(ops) {
@@ -193,10 +198,12 @@ class ops_partials_edge<Dx, std::vector<Eigen::Matrix<fvar<Dx>, R, C>>> {
   }
 };
 
-template <typename Dx>
-class ops_partials_edge<Dx, std::vector<std::vector<fvar<Dx>>>> {
+template <typename Dx, typename Op>
+class ops_partials_edge<
+    Dx, Op,
+    require_all_t<is_std_vector<Op>, is_std_vector<value_type_t<Op>>,
+                  is_fvar<value_type_t<value_type_t<Op>>>>> {
  public:
-  using Op = std::vector<std::vector<fvar<Dx>>>;
   using partial_t = std::vector<Dx>;
   std::vector<partial_t> partials_vec_;
   explicit ops_partials_edge(const Op& ops)

stan/math/opencl/kernel_generator/operation_cl.hpp

@@ -372,7 +372,9 @@ class operation_cl : public operation_cl_base {
    * expression.
    * @return number of elements
    */
-  inline int size() const { return derived().rows() * derived().cols(); }
+  inline Eigen::Index size() const {
+    return derived().rows() * derived().cols();
+  }
 
   /**
    * Number of rows threads need to be launched for. For most expressions this

stan/math/opencl/matrix_cl.hpp

@@ -64,7 +64,7 @@ class matrix_cl : public matrix_cl_base {
 
   int cols() const { return cols_; }
 
-  int size() const { return rows_ * cols_; }
+  Eigen::Index size() const { return rows_ * cols_; }
 
   const matrix_cl_view& view() const { return view_; }
 
@@ -283,7 +283,7 @@ class matrix_cl : public matrix_cl_base {
   matrix_cl(const int rows, const int cols,
             matrix_cl_view partial_view = matrix_cl_view::Entire)
       : rows_(rows), cols_(cols), view_(partial_view) {
-    if (size() == 0) {
+    if (this->size() == 0) {
       return;
     }
     cl::Context& ctx = opencl_context.context();
@@ -321,7 +321,7 @@ class matrix_cl : public matrix_cl_base {
                      matrix_cl_view partial_view = matrix_cl_view::Entire)
       : rows_(A.rows()), cols_(A.cols()), view_(partial_view) {
     using Mat_type = std::decay_t<ref_type_for_opencl_t<Mat>>;
-    if (size() == 0) {
+    if (this->size() == 0) {
       return;
     }
     initialize_buffer_no_heap_if<
@@ -457,7 +457,7 @@ class matrix_cl : public matrix_cl_base {
       return *this;
     }
     this->wait_for_read_write_events();
-    if (size() != a.size()) {
+    if (this->size() != a.size()) {
       buffer_cl_ = cl::Buffer(opencl_context.context(), CL_MEM_READ_WRITE,
                               sizeof(T) * a.size());
     }
@@ -518,7 +518,7 @@ class matrix_cl : public matrix_cl_base {
   template <bool in_order = false>
   cl::Event initialize_buffer(const T* A) {
     cl::Event transfer_event;
-    if (size() == 0) {
+    if (this->size() == 0) {
       return transfer_event;
     }
     cl::Context& ctx = opencl_context.context();
@@ -538,7 +538,7 @@ class matrix_cl : public matrix_cl_base {
   template <bool in_order = false>
   cl::Event initialize_buffer(T* A) {
     cl::Event transfer_event;
-    if (size() == 0) {
+    if (this->size() == 0) {
       return transfer_event;
     }
     cl::Context& ctx = opencl_context.context();
@@ -577,7 +577,7 @@ class matrix_cl : public matrix_cl_base {
    */
   template <bool No_heap, typename U, std::enable_if_t<No_heap>* = nullptr>
   void initialize_buffer_no_heap_if(U&& obj) {
-    if (size() == 0) {
+    if (this->size() == 0) {
       return;
     }
     initialize_buffer(obj.data());
@@ -587,7 +587,7 @@ class matrix_cl : public matrix_cl_base {
   template <bool No_heap, typename U, std::enable_if_t<!No_heap>* = nullptr>
   void initialize_buffer_no_heap_if(U&& obj) {
     using U_val = std::decay_t<ref_type_for_opencl_t<U>>;
-    if (size() == 0) {
+    if (this->size() == 0) {
       return;
     }
     auto* obj_heap = new U_val(std::move(obj));

stan/math/opencl/prim/bernoulli_cdf.hpp

@@ -37,7 +37,7 @@ return_type_t<T_prob_cl> bernoulli_cdf(const T_n_cl& n,
 
   check_consistent_sizes(function, "Random variable", n,
                          "Probability parameter", theta);
-  const size_t N = is_n_vector ? size(n) : size(theta);
+  const size_t N = is_n_vector ? math::size(n) : math::size(theta);
   if (N == 0) {
     return 1.0;
   }

stan/math/opencl/prim/bernoulli_lccdf.hpp

@@ -38,7 +38,7 @@ return_type_t<T_prob_cl> bernoulli_lccdf(const T_n_cl& n,
 
   check_consistent_sizes(function, "Random variable", n,
                          "Probability parameter", theta);
-  const size_t N = is_n_vector ? size(n) : size(theta);
+  const size_t N = is_n_vector ? math::size(n) : math::size(theta);
   if (N == 0) {
     return 0.0;
   }

stan/math/opencl/prim/bernoulli_lcdf.hpp

@@ -38,7 +38,7 @@ return_type_t<T_prob_cl> bernoulli_lcdf(const T_n_cl& n,
 
   check_consistent_sizes(function, "Random variable", n,
                          "Probability parameter", theta);
-  const size_t N = is_n_vector ? size(n) : size(theta);
+  const size_t N = is_n_vector ? math::size(n) : math::size(theta);
   if (N == 0) {
     return 0.0;
   }

stan/math/opencl/prim/bernoulli_logit_glm_lpmf.hpp

@@ -66,13 +66,14 @@ return_type_t<T_x_cl, T_alpha_cl, T_beta_cl> bernoulli_logit_glm_lpmf(
   const size_t M = x.cols();
 
   if (is_y_vector) {
-    check_size_match(function, "Rows of ", "x", N, "rows of ", "y", size(y));
+    check_size_match(function, "Rows of ", "x", N, "rows of ", "y",
+                     math::size(y));
   }
   check_size_match(function, "Columns of ", "x_cl", M, "size of ", "beta",
-                   size(beta));
+                   math::size(beta));
   if (is_alpha_vector) {
     check_size_match(function, "Rows of ", "x", N, "size of ", "alpha",
-                     size(alpha));
+                     math::size(alpha));
   }
 
   if (N == 0) {

stan/math/opencl/prim/bernoulli_logit_lpmf.hpp

@@ -36,7 +36,7 @@ return_type_t<T_prob_cl> bernoulli_logit_lpmf(const T_n_cl& n,
 
   check_consistent_sizes(function, "Random variable", n,
                          "Probability parameter", theta);
-  const size_t N = is_n_vector ? size(n) : size(theta);
+  const size_t N = is_n_vector ? math::size(n) : math::size(theta);
   if (N == 0) {
     return 0.0;
   }

stan/math/opencl/prim/bernoulli_lpmf.hpp

@@ -38,7 +38,7 @@ return_type_t<T_prob_cl> bernoulli_lpmf(const T_n_cl& n,
 
   check_consistent_sizes(function, "Random variable", n,
                          "Probability parameter", theta);
-  const size_t N = is_n_vector ? size(n) : size(theta);
+  const size_t N = is_n_vector ? math::size(n) : math::size(theta);
   if (N == 0) {
     return 0.0;
   }

stan/math/opencl/prim/categorical_logit_glm_lpmf.hpp

@@ -61,10 +61,10 @@ return_type_t<T_x, T_alpha, T_beta> categorical_logit_glm_lpmf(
   static const char* function = "categorical_logit_glm_lpmf";
   if (is_y_vector) {
     check_size_match(function, "Rows of ", "x", N_instances, "size of ", "y",
-                     size(y));
+                     math::size(y));
   }
   check_size_match(function, "Columns of ", "beta", N_classes, "size of ",
-                   "alpha", size(alpha));
+                   "alpha", math::size(alpha));
   check_size_match(function, "Columns of ", "x", N_attributes, "Rows of",
                    "beta", beta.rows());
 

stan/math/opencl/prim/cumulative_sum.hpp

@@ -38,8 +38,8 @@ inline auto cumulative_sum(T_vec&& v) {
       = opencl_kernels::cumulative_sum<T_scal>::kernel1.get_option(
           "LOCAL_SIZE_");
   const int work_groups = std::min(
-      (v.size() + local_size - 1) / local_size,
-      static_cast<int>(
+      static_cast<Eigen::Index>((v.size() + local_size - 1) / local_size),
+      static_cast<Eigen::Index>(
           opencl_context.device()[0].getInfo<CL_DEVICE_MAX_COMPUTE_UNITS>())
           * 16);
   const int local_size2

stan/math/opencl/prim/neg_binomial_2_log_glm_lpmf.hpp

@@ -82,17 +82,18 @@ neg_binomial_2_log_glm_lpmf(const T_y_cl& y, const T_x_cl& x,
   const size_t M = x.cols();
 
   if (is_y_vector) {
-    check_size_match(function, "Rows of ", "x", N, "rows of ", "y", size(y));
+    check_size_match(function, "Rows of ", "x", N, "rows of ", "y",
+                     math::size(y));
   }
   check_size_match(function, "Columns of ", "x", M, "size of ", "beta",
-                   size(beta));
+                   math::size(beta));
   if (is_phi_vector) {
     check_size_match(function, "Rows of ", "x", N, "size of ", "phi",
-                     size(phi));
+                     math::size(phi));
   }
   if (is_alpha_vector) {
     check_size_match(function, "Rows of ", "x", N, "size of ", "alpha",
-                     size(alpha));
+                     math::size(alpha));
   }
   if (N == 0) {
     return 0;