diff --git a/docs/source/index.rst b/docs/source/index.rst
@@ -69,6 +69,7 @@ for details.
    missing
    histogram
    random
+   sfinae
    file_loading
    build-options
    pitfall

diff --git a/docs/source/sfinae.rst b/docs/source/sfinae.rst
@@ -0,0 +1,77 @@
+.. Copyright (c) 2016, Johan Mabille, Sylvain Corlay and Wolf Vollprecht
+
+   Distributed under the terms of the BSD 3-Clause License.
+
+   The full license is in the file LICENSE, distributed with this software.
+
+.. _histogram:
+
+SFINAE
+======
+
+Rank overload
+-------------
+
+All `xtensor`'s classes have a member ``rank`` that can be used
+to overload based on rank using *SFINAE*.
+Consider the following example:
+
+.. code-block:: cpp
+
+    template <class E, std::enable_if_t<!xt::has_rank_t<E, 2>::value, int> = 0>
+    inline E foo(E&& a)
+    {
+        ... // act on object of flexible rank, or fixed rank != 2
+    }
+
+    template <class E, std::enable_if_t<xt::has_rank_t<E, 2>::value, int> = 0>
+    inline E foo(E&& a)
+    {
+        ... // act on object of fixed rank == 2
+    }
+
+    TEST(sfinae, rank_basic)
+    {
+        xt::xarray<size_t> a = {{9, 9}, {9, 9}};
+        xt::xtensor<size_t, 1> b = {9, 9};
+        xt::xtensor<size_t, 2> c = {{9, 9}, {9, 9}};
+
+        foo(a); // flexible rank -> first overload
+        foo(b); // fixed rank == 2 -> first overload
+        foo(c); // fixed rank == 2 -> second overload
+    }
+
+.. note::
+
+    If one wants to test for more than a single value for ``rank``,
+    one can use the default value ``SIZE_MAX`` used for flexible rank objects.
+    For example, one could have the following overloads:
+
+    .. code-block:: cpp
+
+        // flexible rank
+        template <class E, std::enable_if_t<xt::has_rank_t<E, SIZE_MAX>::value, int> = 0>
+        inline E foo(E&& a);
+
+        // fixed rank == 1
+        template <class E, std::enable_if_t<xt::has_rank_t<E, 1>::value, int> = 0>
+        inline E foo(E&& a);
+
+        // fixed rank == 2
+        template <class E, std::enable_if_t<xt::has_rank_t<E, 2>::value, int> = 0>
+        inline E foo(E&& a);
+
+    Note that fixed ranks other than 1 and 2 will raise a compiler error.
+
+    Of course, if one wants a more limited scope, one could also do the following:
+
+    .. code-block:: cpp
+
+        // flexible rank
+        inline void foo(xt::xarray<double>& a);
+
+        // fixed rank == 1
+        inline void foo(xt::xtensor<double,1>& a);
+
+        // fixed rank == 2
+        inline void foo(xt::xtensor<double,2>& a);
diff --git a/include/xtensor/xarray.hpp b/include/xtensor/xarray.hpp
@@ -104,6 +104,7 @@ namespace xt
         using inner_backstrides_type = typename base_type::inner_backstrides_type;
         using temporary_type = typename semantic_base::temporary_type;
         using expression_tag = Tag;
+        constexpr static std::size_t rank = SIZE_MAX;
 
         xarray_container();
         explicit xarray_container(const shape_type& shape, layout_type l = L);

diff --git a/include/xtensor/xfixed.hpp b/include/xtensor/xfixed.hpp
@@ -54,7 +54,7 @@ namespace xt
     namespace detail
     {
         /**************************************************************************************
-           The following is something we can currently only dream about -- for when we drop 
+           The following is something we can currently only dream about -- for when we drop
            support for a lot of the old compilers (e.g. GCC 4.9, MSVC 2017 ;)
 
         template <class T>
@@ -281,7 +281,7 @@ namespace xt
      * with tensor semantic and fixed dimension
      *
      * @tparam ET The type of the elements.
-     * @tparam S The xshape template paramter of the container. 
+     * @tparam S The xshape template paramter of the container.
      * @tparam L The layout_type of the tensor.
      * @tparam SH Wether the tensor can be used as a shared expression.
      * @tparam Tag The expression tag.
@@ -313,6 +313,7 @@ namespace xt
         using expression_tag = Tag;
 
         constexpr static std::size_t N = std::tuple_size<shape_type>::value;
+        constexpr static std::size_t rank = N;
 
         xfixed_container() = default;
         xfixed_container(const value_type& v);
@@ -616,7 +617,7 @@ namespace xt
 
     /**
      * Allocates an xfixed_container with shape S with values from a C array.
-     * The type returned by get_init_type_t is raw C array ``value_type[X][Y][Z]`` for ``xt::xshape<X, Y, Z>``. 
+     * The type returned by get_init_type_t is raw C array ``value_type[X][Y][Z]`` for ``xt::xshape<X, Y, Z>``.
      * C arrays can be initialized with the initializer list syntax, but the size is checked at compile
      * time to prevent errors.
      * Note: for clang < 3.8 this is an initializer_list and the size is not checked at compile-or runtime.

diff --git a/include/xtensor/xtensor.hpp b/include/xtensor/xtensor.hpp
@@ -105,6 +105,7 @@ namespace xt
         using inner_strides_type = typename base_type::inner_strides_type;
         using temporary_type = typename semantic_base::temporary_type;
         using expression_tag = Tag;
+        constexpr static std::size_t rank = N;
 
         xtensor_container();
         xtensor_container(nested_initializer_list_t<value_type, N> t);
@@ -759,7 +760,7 @@ namespace xt
         std::fill(m_storage.begin(), m_storage.end(), e);
         return *this;
     }
-    
+
     template <class EC, std::size_t N, layout_type L, class Tag>
     inline auto xtensor_view<EC, N, L, Tag>::storage_impl() noexcept -> storage_type&
     {

diff --git a/include/xtensor/xutils.hpp b/include/xtensor/xutils.hpp
@@ -857,6 +857,19 @@ namespace xt
 
     template <class ST>
     using inner_reference_t = typename inner_reference<ST>::type;
+
+    /************
+     * has_rank *
+     ************/
+
+    template <class E, size_t N>
+    struct has_rank
+    {
+        using type = std::integral_constant<bool, std::decay_t<E>::rank == N>;
+    };
+
+    template <class E, size_t N>
+    using has_rank_t = typename has_rank<std::decay_t<E>, N>::type;
 }
 
 #endif
diff --git a/include/xtensor/xview.hpp b/include/xtensor/xview.hpp
@@ -432,6 +432,7 @@ namespace xt
 
         using container_iterator = pointer;
         using const_container_iterator = const_pointer;
+        constexpr static std::size_t rank = SIZE_MAX;
 
         // The FSL argument prevents the compiler from calling this constructor
         // instead of the copy constructor when sizeof...(SL) == 0.

diff --git a/test/CMakeLists.txt b/test/CMakeLists.txt
@@ -224,6 +224,7 @@ set(XTENSOR_TESTS
     test_extended_xhistogram.cpp
     test_extended_xsort.cpp
     test_xchunked_array.cpp
+    test_sfinae.cpp
 )
 
 if(nlohmann_json_FOUND)

diff --git a/test/test_sfinae.cpp b/test/test_sfinae.cpp
@@ -0,0 +1,90 @@
+/***************************************************************************
+* Copyright (c) Johan Mabille, Sylvain Corlay and Wolf Vollprecht          *
+* Copyright (c) QuantStack                                                 *
+*                                                                          *
+* Distributed under the terms of the BSD 3-Clause License.                 *
+*                                                                          *
+* The full license is in the file LICENSE, distributed with this software. *
+****************************************************************************/
+
+#include <complex>
+#include <limits>
+
+#include "gtest/gtest.h"
+#include "xtensor/xtensor.hpp"
+#include "xtensor/xarray.hpp"
+// #include "xtensor/xfixed.hpp"
+#include "xtensor/xview.hpp"
+
+namespace xt
+{
+    template <class E, std::enable_if_t<!xt::has_rank_t<E, 2>::value, int> = 0>
+    inline E sfinae_rank_basic_func(E&& a)
+    {
+        E b = a;
+        b.fill(0);
+        return b;
+    }
+
+    template <class E, std::enable_if_t<xt::has_rank_t<E, 2>::value, int> = 0>
+    inline E sfinae_rank_basic_func(E&& a)
+    {
+        E b = a;
+        b.fill(2);
+        return b;
+    }
+
+    TEST(sfinae, rank_basic)
+    {
+        xt::xarray<size_t> a = {{9, 9, 9}, {9, 9, 9}};
+        xt::xtensor<size_t, 1> b = {9, 9};
+        xt::xtensor<size_t, 2> c = {{9, 9}, {9, 9}};
+        // xt::xtensor_fixed<size_t, xt::xshape<2, 2>> d = {{9, 9}, {9, 9}};
+        auto v = xt::view(c, 0, xt::all());
+
+        EXPECT_TRUE(xt::all(xt::equal(sfinae_rank_basic_func(a), 0ul)));
+        EXPECT_TRUE(xt::all(xt::equal(sfinae_rank_basic_func(b), 0ul)));
+        EXPECT_TRUE(xt::all(xt::equal(sfinae_rank_basic_func(c), 2ul)));
+        // EXPECT_TRUE(xt::all(xt::equal(sfinae_rank_basic_func(d), 2ul)));
+        EXPECT_TRUE(xt::all(xt::equal(sfinae_rank_basic_func(v), 0ul)));
+    }
+
+    template <class E, std::enable_if_t<xt::has_rank_t<E, SIZE_MAX>::value, int> = 0>
+    inline E sfinae_rank_func(E&& a)
+    {
+        E b = a;
+        b.fill(0);
+        return b;
+    }
+
+    template <class E, std::enable_if_t<xt::has_rank_t<E, 1>::value, int> = 0>
+    inline E sfinae_rank_func(E&& a)
+    {
+        E b = a;
+        b.fill(1);
+        return b;
+    }
+
+    template <class E, std::enable_if_t<xt::has_rank_t<E, 2>::value, int> = 0>
+    inline E sfinae_rank_func(E&& a)
+    {
+        E b = a;
+        b.fill(2);
+        return b;
+    }
+
+    TEST(sfinae, rank)
+    {
+        xt::xarray<size_t> a = {{9, 9, 9}, {9, 9, 9}};
+        xt::xtensor<size_t, 1> b = {9, 9};
+        xt::xtensor<size_t, 2> c = {{9, 9}, {9, 9}};
+        // xt::xtensor_fixed<size_t, xt::xshape<2, 2>> d = {{9, 9}, {9, 9}};
+        auto v = xt::view(c, 0, xt::all());
+
+        EXPECT_TRUE(xt::all(xt::equal(sfinae_rank_func(a), 0ul)));
+        EXPECT_TRUE(xt::all(xt::equal(sfinae_rank_func(b), 1ul)));
+        EXPECT_TRUE(xt::all(xt::equal(sfinae_rank_func(c), 2ul)));
+        // EXPECT_TRUE(xt::all(xt::equal(sfinae_rank_func(d), 2ul)));
+        EXPECT_TRUE(xt::all(xt::equal(sfinae_rank_func(v), 0ul)));
+    }
+}
-Original file line number
+Diff line change
@@ Expand Up / @@ -69,6 +69,7 @@ for details. @@
        missing
        histogram
        random
+       sfinae
        file_loading
        build-options
        pitfall
@@ Expand Down @@