Moved re-ordering logic into assignment

Author: spectre-ns

include/xtensor/xassign.hpp

@@ -439,16 +439,70 @@ namespace xt
         using requested_value_type = detail::conditional_promote_to_complex_t<e1_value_type, e2_requested_value_type>;
     };
 
+    /**********************************
+     * Expression Order Optimizations *
+     **********************************/
+
+    class optimize_expression
+    {
+    private:
+
+        template <class E1, class E2>
+        struct equal_rank
+        {
+            static constexpr bool value = get_rank<E1>::value == get_rank<E2>::value;
+        };
+
+        template <class E1, class... E>
+        struct all_equal_rank
+        {
+            static constexpr bool value = xtl::conjunction<equal_rank<E1, E>...>::value
+                                          && (get_rank<E1>::value != SIZE_MAX);
+        };
+
+        template <class F, class... CT, class... S, size_t... I, size_t... J>
+        inline auto
+        impl_reorder_function(const xfunction<F, CT...>& e, std::tuple<S...> slices, std::index_sequence<I...>, std::index_sequence<J...>)
+        {
+            return make_lambda_xfunction(F(), view(std::get<I>(e.arguments()), std::get<J>(slices)...)...);
+        }
+
+    public:
+
+        // when we have a view of a function where the closures of the functions are of equal rank (i.e no
+        // broadcasting) we can flip the order of the function and the view such that we have a function of
+        // views of containers which can be linearly assigned unlike the inverse.
+        template <class F, class... CT, class... S, class = std::enable_if_t<all_equal_rank<std::decay_t<CT>...>::value>>
+        inline auto reorder(const xview<xfunction<F, CT...>, S...>& e)
+        {
+            return impl_reorder_function(
+                e.expression(),
+                e.slices(),
+                std::make_index_sequence<sizeof...(CT)>(),
+                std::make_index_sequence<sizeof...(S)>()
+            );
+        }
+
+        // base case no applicable optimization
+        template <class E>
+        inline auto& reorder(E&& e)
+        {
+            return std::forward<E>(e);
+        }
+    };
+
     template <class E1, class E2>
     inline void xexpression_assigner_base<xtensor_expression_tag>::assign_data(
         xexpression<E1>& e1,
         const xexpression<E2>& e2,
         bool trivial
     )
     {
-        E1& de1 = e1.derived_cast();
-        const E2& de2 = e2.derived_cast();
-        using traits = xassign_traits<E1, E2>;
+        auto& de1 = e1.derived_cast();
+        const auto& de2 = optimize_expression().reorder(e2.derived_cast());
+        using dst_type = typename std::decay_t<decltype(de1)>;
+        using src_type = typename std::decay_t<decltype(de2)>;
+        using traits = xassign_traits<dst_type, src_type>;
 
         bool linear_assign = traits::linear_assign(de1, de2, trivial);
         constexpr bool simd_assign = traits::simd_assign();