DSL Support for Layout and Buffer classes (#192)

* backup [skip ci]

* backup [skip ci]

* works, let's see what GCC hates...

* Committing clang-format changes

* remove ambiguity in string_type

---------

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

include/tensorwrapper/buffer/buffer_base.hpp

@@ -15,6 +15,7 @@
  */
 
 #pragma once
+#include <tensorwrapper/detail_/dsl_base.hpp>
 #include <tensorwrapper/detail_/polymorphic_base.hpp>
 #include <tensorwrapper/dsl/labeled.hpp>
 #include <tensorwrapper/layout/layout_base.hpp>
@@ -25,30 +26,33 @@ namespace tensorwrapper::buffer {
  *
  *  All classes which wrap existing tensor libraries derive from this class.
  */
-class BufferBase : public detail_::PolymorphicBase<BufferBase> {
+class BufferBase : public detail_::PolymorphicBase<BufferBase>,
+                   public detail_::DSLBase<BufferBase> {
 private:
     /// Type of *this
     using my_type = BufferBase;
 
+protected:
     /// Type *this inherits from
-    using my_base_type = detail_::PolymorphicBase<my_type>;
+    using polymorphic_base = detail_::PolymorphicBase<my_type>;
 
 public:
     /// Type all buffers inherit from
-    using buffer_base_type = typename my_base_type::base_type;
+    using buffer_base_type = typename polymorphic_base::base_type;
 
     /// Type of a mutable reference to a buffer_base_type object
-    using buffer_base_reference = typename my_base_type::base_reference;
+    using buffer_base_reference = typename polymorphic_base::base_reference;
 
     /// Type of a read-only reference to a buffer_base_type object
     using const_buffer_base_reference =
-      typename my_base_type::const_base_reference;
+      typename polymorphic_base::const_base_reference;
 
     /// Type of a pointer to an object of type buffer_base_type
-    using buffer_base_pointer = typename my_base_type::base_pointer;
+    using buffer_base_pointer = typename polymorphic_base::base_pointer;
 
     /// Type of a pointer to a read-only object of type buffer_base_type
-    using const_buffer_base_pointer = typename my_base_type::const_base_pointer;
+    using const_buffer_base_pointer =
+      typename polymorphic_base::const_base_pointer;
 
     /// Type of the class describing the physical layout of the buffer
     using layout_type = layout::LayoutBase;
@@ -59,6 +63,9 @@ class BufferBase : public detail_::PolymorphicBase<BufferBase> {
     /// Type of a pointer to the layout
     using layout_pointer = typename layout_type::layout_pointer;
 
+    /// Type used to represent the tensor's rank
+    using rank_type = typename layout_type::size_type;
+
     // -------------------------------------------------------------------------
     // -- Accessors
     // -------------------------------------------------------------------------
@@ -90,6 +97,10 @@ class BufferBase : public detail_::PolymorphicBase<BufferBase> {
         return *m_layout_;
     }
 
+    rank_type rank() const noexcept {
+        return has_layout() ? layout().rank() : 0;
+    }
+
     // -------------------------------------------------------------------------
     // -- Utility methods
     // -------------------------------------------------------------------------
@@ -191,6 +202,21 @@ class BufferBase : public detail_::PolymorphicBase<BufferBase> {
         return *this;
     }
 
+    dsl_reference addition_assignment_(label_type this_labels,
+                                       const_labeled_reference lhs,
+                                       const_labeled_reference rhs) override;
+
+    dsl_reference subtraction_assignment_(label_type this_labels,
+                                          const_labeled_reference lhs,
+                                          const_labeled_reference rhs) override;
+
+    dsl_reference multiplication_assignment_(
+      label_type this_labels, const_labeled_reference lhs,
+      const_labeled_reference rhs) override;
+
+    dsl_reference permute_assignment_(label_type this_labels,
+                                      const_labeled_reference rhs) override;
+
 private:
     /// Throws std::runtime_error when there is no layout
     void assert_layout_() const {

include/tensorwrapper/buffer/eigen.hpp

@@ -180,10 +180,36 @@ class Eigen : public Replicated {
         return my_base_type::are_equal_impl_<my_type>(rhs);
     }
 
+    /// Implements addition_assignment by calling addition_assignment on state
+    dsl_reference addition_assignment_(label_type this_labels,
+                                       const_labeled_reference lhs,
+                                       const_labeled_reference rhs) override;
+
+    /// Calls subtraction_assignment on each member
+    dsl_reference subtraction_assignment_(label_type this_labels,
+                                          const_labeled_reference lhs,
+                                          const_labeled_reference rhs) override;
+
+    /// Calls multiplication_assignment on each member
+    dsl_reference multiplication_assignment_(
+      label_type this_labels, const_labeled_reference lhs,
+      const_labeled_reference rhs) override;
+
+    /// Calls permute_assignment on each member
+    dsl_reference permute_assignment_(label_type this_labels,
+                                      const_labeled_reference rhs) override;
+
     /// Implements to_string
-    typename my_base_type::string_type to_string_() const override;
+    typename polymorphic_base::string_type to_string_() const override;
 
 private:
+    dsl_reference hadamard_(label_type this_labels, const_labeled_reference lhs,
+                            const_labeled_reference rhs);
+
+    dsl_reference contraction_(label_type this_labels,
+                               const_labeled_reference lhs,
+                               const_labeled_reference rhs);
+
     /// The actual Eigen tensor
     data_type m_tensor_;
 };

include/tensorwrapper/dsl/dummy_indices.hpp

@@ -163,6 +163,51 @@ class DummyIndices
         return true;
     }
 
+    /** @brief Is a thruple of DummyIndices consistent with a pure element-wise
+     *         product?
+     *
+     *  In generalized Einstein notation a pure element-wise (also commonly
+     *  termed Hadamard) product is denoted by *this, @p lhs, and @p rhs
+     *  having the same ordered set of dummy indices, up to permutation.
+     *  Additionally, the dummy indices associated with any given tensor may
+     *  not include a repeated index.
+     *
+     *  @param[in] lhs The dummy indices associated with the tensor to the
+     *                 left of the times operator.
+     *  @param[in] rhs The dummy indices associated with the tensor to the
+     *                 right of the times operator.
+     *
+     *  @return True If the dummy indices given by *this, @p lhs, and @p rhs
+     *          are consistent with a purely element-wise product of the tensors
+     *          that @p lhs and @p rhs label.
+     *
+     *  @throw None No throw guarantee.
+     */
+    bool is_hadamard_product(const DummyIndices& lhs,
+                             const DummyIndices& rhs) const noexcept;
+
+    /** @brief Does a thruple of DummyIndices indicate a product is a pure
+     *         contraction?
+     *
+     *  In generalized Einstein notation a pure contraction is an operation
+     *  where indices common to @p lhs and @p rhs are summed over and do NOT
+     *  appear in the result, i.e., *this. Additionally, we stipulate that
+     *  there must be at least one index summed over (if no index is summed over
+     *  the operation is a pure direct-product).
+     *
+     *  @param[in] lhs The dummy indices associated with the tensor to the
+     *                 left of the times operator.
+     *  @param[in] rhs The dummy indices associated with the tensor to the
+     *                 right of the times operator.
+     *
+     *  @return True if the indices associated with *this, @p lhs, and @p rhs
+     *          are consistent with a contraction and false otherwise.
+     *
+     *  @throw None No throw guarantee.
+     */
+    bool is_contraction(const DummyIndices& lhs,
+                        const DummyIndices& rhs) const noexcept;
+
     /** @brief Computes the permutation needed to convert *this into @p other.
      *
      *  Each DummyIndices object is viewed as an ordered set of objects. If
@@ -366,6 +411,31 @@ class DummyIndices
         return rv;
     }
 
+    /** @brief Returns the  set difference of *this and @p other.
+     *
+     *  The set difference of *this with @p other is the set of indices which
+     *  appear in *this, but not in @p other. This method will return the set
+     *  (indices which appear more than once in *this will only appear once
+     *  in the result) which results from the set difference of *this with
+     *  @p other.
+     *
+     *  @param[in] other The set to remove from *this.
+     *
+     *  @return The set difference of *this and @p rhs.
+     *
+     *  @throw std::bad_alloc if there is a problem allocating the return.
+     *                        Strong throw guarantee.
+     */
+    DummyIndices difference(const DummyIndices& other) const {
+        DummyIndices rv;
+        for(const auto& x : *this) {
+            if(other.count(x)) continue;
+            if(rv.count(x)) continue;
+            rv.m_dummy_indices_.push_back(x);
+        }
+        return rv;
+    }
+
 protected:
     /// Main ctor for setting the value, throws if any index is empty
     explicit DummyIndices(split_string_type split_dummy_indices) :
@@ -401,4 +471,28 @@ class DummyIndices
     split_string_type m_dummy_indices_;
 };
 
+template<typename StringType>
+bool DummyIndices<StringType>::is_hadamard_product(
+  const DummyIndices& lhs, const DummyIndices& rhs) const noexcept {
+    if(has_repeated_indices()) return false;
+    if(lhs.has_repeated_indices()) return false;
+    if(rhs.has_repeated_indices()) return false;
+    if(!is_permutation(lhs)) return false;
+    if(!is_permutation(rhs)) return false;
+    return true;
+}
+
+template<typename StringType>
+bool DummyIndices<StringType>::is_contraction(
+  const DummyIndices& lhs, const DummyIndices& rhs) const noexcept {
+    if(has_repeated_indices()) return false;
+    if(lhs.has_repeated_indices()) return false;
+    if(rhs.has_repeated_indices()) return false;
+    auto lhs_cap_rhs = lhs.intersection(rhs);
+    if(lhs_cap_rhs.empty()) return false; // No common indices
+    if(!intersection(lhs_cap_rhs).empty())
+        return false; // Common index not summed
+    return true;
+}
+
 } // namespace tensorwrapper::dsl