Squashed 'third_party/GSL/' content from commit 0cebbd7
Change-Id: Iffb2e32f2f45297ac6d3e82168cd9df4ac5acc2f
git-subtree-dir: third_party/GSL
git-subtree-split: 0cebbd77bfc21acbf1cc05983ad626539eeeb8e0
diff --git a/tests/strided_span_tests.cpp b/tests/strided_span_tests.cpp
new file mode 100644
index 0000000..8719336
--- /dev/null
+++ b/tests/strided_span_tests.cpp
@@ -0,0 +1,757 @@
+///////////////////////////////////////////////////////////////////////////////
+//
+// Copyright (c) 2015 Microsoft Corporation. All rights reserved.
+//
+// This code is licensed under the MIT License (MIT).
+//
+// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+// THE SOFTWARE.
+//
+///////////////////////////////////////////////////////////////////////////////
+
+#include <catch/catch.hpp> // for AssertionHandler, StringRef, CHECK, CHECK...
+
+#include <gsl/gsl_byte> // for byte
+#include <gsl/gsl_util> // for narrow_cast
+#include <gsl/multi_span> // for strided_span, index, multi_span, strided_...
+
+#include <iostream> // for size_t
+#include <iterator> // for begin, end
+#include <numeric> // for iota
+#include <type_traits> // for integral_constant<>::value, is_convertible
+#include <vector> // for vector
+
+namespace gsl {
+struct fail_fast;
+} // namespace gsl
+
+using namespace std;
+using namespace gsl;
+
+namespace
+{
+struct BaseClass
+{
+};
+struct DerivedClass : BaseClass
+{
+};
+}
+
+TEST_CASE("span_section_test")
+{
+ int a[30][4][5];
+
+ const auto av = as_multi_span(a);
+ const auto sub = av.section({15, 0, 0}, gsl::multi_span_index<3>{2, 2, 2});
+ const auto subsub = sub.section({1, 0, 0}, gsl::multi_span_index<3>{1, 1, 1});
+ (void) subsub;
+}
+
+TEST_CASE("span_section")
+{
+ std::vector<int> data(5 * 10);
+ std::iota(begin(data), end(data), 0);
+ const multi_span<int, 5, 10> av = as_multi_span(multi_span<int>{data}, dim<5>(), dim<10>());
+
+ const strided_span<int, 2> av_section_1 = av.section({1, 2}, {3, 4});
+ CHECK((av_section_1[{0, 0}] == 12));
+ CHECK((av_section_1[{0, 1}] == 13));
+ CHECK((av_section_1[{1, 0}] == 22));
+ CHECK((av_section_1[{2, 3}] == 35));
+
+ const strided_span<int, 2> av_section_2 = av_section_1.section({1, 2}, {2, 2});
+ CHECK((av_section_2[{0, 0}] == 24));
+ CHECK((av_section_2[{0, 1}] == 25));
+ CHECK((av_section_2[{1, 0}] == 34));
+}
+
+TEST_CASE("strided_span_constructors")
+{
+ // Check stride constructor
+ {
+ int arr[] = {1, 2, 3, 4, 5, 6, 7, 8, 9};
+ const int carr[] = {1, 2, 3, 4, 5, 6, 7, 8, 9};
+
+ strided_span<int, 1> sav1{arr, {{9}, {1}}}; // T -> T
+ CHECK(sav1.bounds().index_bounds() == multi_span_index<1>{9});
+ CHECK(sav1.bounds().stride() == 1);
+ CHECK((sav1[0] == 1 && sav1[8] == 9));
+
+ strided_span<const int, 1> sav2{carr, {{4}, {2}}}; // const T -> const T
+ CHECK(sav2.bounds().index_bounds() == multi_span_index<1>{4});
+ CHECK(sav2.bounds().strides() == multi_span_index<1>{2});
+ CHECK((sav2[0] == 1 && sav2[3] == 7));
+
+ strided_span<int, 2> sav3{arr, {{2, 2}, {6, 2}}}; // T -> const T
+ CHECK((sav3.bounds().index_bounds() == multi_span_index<2>{2, 2}));
+ CHECK((sav3.bounds().strides() == multi_span_index<2>{6, 2}));
+ CHECK((sav3[{0, 0}] == 1 && sav3[{0, 1}] == 3 && sav3[{1, 0}] == 7));
+ }
+
+ // Check multi_span constructor
+ {
+ int arr[] = {1, 2};
+
+ // From non-cv-qualified source
+ {
+ const multi_span<int> src = arr;
+
+ strided_span<int, 1> sav{src, {2, 1}};
+ CHECK(sav.bounds().index_bounds() == multi_span_index<1>{2});
+ CHECK(sav.bounds().strides() == multi_span_index<1>{1});
+ CHECK(sav[1] == 2);
+
+#if _MSC_VER > 1800
+ // strided_span<const int, 1> sav_c{ {src}, {2, 1} };
+ strided_span<const int, 1> sav_c{multi_span<const int>{src},
+ strided_bounds<1>{2, 1}};
+#else
+ strided_span<const int, 1> sav_c{multi_span<const int>{src},
+ strided_bounds<1>{2, 1}};
+#endif
+ CHECK(sav_c.bounds().index_bounds() == multi_span_index<1>{2});
+ CHECK(sav_c.bounds().strides() == multi_span_index<1>{1});
+ CHECK(sav_c[1] == 2);
+
+#if _MSC_VER > 1800
+ strided_span<volatile int, 1> sav_v{src, {2, 1}};
+#else
+ strided_span<volatile int, 1> sav_v{multi_span<volatile int>{src},
+ strided_bounds<1>{2, 1}};
+#endif
+ CHECK(sav_v.bounds().index_bounds() == multi_span_index<1>{2});
+ CHECK(sav_v.bounds().strides() == multi_span_index<1>{1});
+ CHECK(sav_v[1] == 2);
+
+#if _MSC_VER > 1800
+ strided_span<const volatile int, 1> sav_cv{src, {2, 1}};
+#else
+ strided_span<const volatile int, 1> sav_cv{multi_span<const volatile int>{src},
+ strided_bounds<1>{2, 1}};
+#endif
+ CHECK(sav_cv.bounds().index_bounds() == multi_span_index<1>{2});
+ CHECK(sav_cv.bounds().strides() == multi_span_index<1>{1});
+ CHECK(sav_cv[1] == 2);
+ }
+
+ // From const-qualified source
+ {
+ const multi_span<const int> src{arr};
+
+ strided_span<const int, 1> sav_c{src, {2, 1}};
+ CHECK(sav_c.bounds().index_bounds() == multi_span_index<1>{2});
+ CHECK(sav_c.bounds().strides() == multi_span_index<1>{1});
+ CHECK(sav_c[1] == 2);
+
+#if _MSC_VER > 1800
+ strided_span<const volatile int, 1> sav_cv{src, {2, 1}};
+#else
+ strided_span<const volatile int, 1> sav_cv{multi_span<const volatile int>{src},
+ strided_bounds<1>{2, 1}};
+#endif
+
+ CHECK(sav_cv.bounds().index_bounds() == multi_span_index<1>{2});
+ CHECK(sav_cv.bounds().strides() == multi_span_index<1>{1});
+ CHECK(sav_cv[1] == 2);
+ }
+
+ // From volatile-qualified source
+ {
+ const multi_span<volatile int> src{arr};
+
+ strided_span<volatile int, 1> sav_v{src, {2, 1}};
+ CHECK(sav_v.bounds().index_bounds() == multi_span_index<1>{2});
+ CHECK(sav_v.bounds().strides() == multi_span_index<1>{1});
+ CHECK(sav_v[1] == 2);
+
+#if _MSC_VER > 1800
+ strided_span<const volatile int, 1> sav_cv{src, {2, 1}};
+#else
+ strided_span<const volatile int, 1> sav_cv{multi_span<const volatile int>{src},
+ strided_bounds<1>{2, 1}};
+#endif
+ CHECK(sav_cv.bounds().index_bounds() == multi_span_index<1>{2});
+ CHECK(sav_cv.bounds().strides() == multi_span_index<1>{1});
+ CHECK(sav_cv[1] == 2);
+ }
+
+ // From cv-qualified source
+ {
+ const multi_span<const volatile int> src{arr};
+
+ strided_span<const volatile int, 1> sav_cv{src, {2, 1}};
+ CHECK(sav_cv.bounds().index_bounds() == multi_span_index<1>{2});
+ CHECK(sav_cv.bounds().strides() == multi_span_index<1>{1});
+ CHECK(sav_cv[1] == 2);
+ }
+ }
+
+ // Check const-casting constructor
+ {
+ int arr[2] = {4, 5};
+
+ const multi_span<int, 2> av(arr, 2);
+ multi_span<const int, 2> av2{av};
+ CHECK(av2[1] == 5);
+
+ static_assert(
+ std::is_convertible<const multi_span<int, 2>, multi_span<const int, 2>>::value,
+ "ctor is not implicit!");
+
+ const strided_span<int, 1> src{arr, {2, 1}};
+ strided_span<const int, 1> sav{src};
+ CHECK(sav.bounds().index_bounds() == multi_span_index<1>{2});
+ CHECK(sav.bounds().stride() == 1);
+ CHECK(sav[1] == 5);
+
+ static_assert(
+ std::is_convertible<const strided_span<int, 1>, strided_span<const int, 1>>::value,
+ "ctor is not implicit!");
+ }
+
+ // Check copy constructor
+ {
+ int arr1[2] = {3, 4};
+ const strided_span<int, 1> src1{arr1, {2, 1}};
+ strided_span<int, 1> sav1{src1};
+
+ CHECK(sav1.bounds().index_bounds() == multi_span_index<1>{2});
+ CHECK(sav1.bounds().stride() == 1);
+ CHECK(sav1[0] == 3);
+
+ int arr2[6] = {1, 2, 3, 4, 5, 6};
+ const strided_span<const int, 2> src2{arr2, {{3, 2}, {2, 1}}};
+ strided_span<const int, 2> sav2{src2};
+ CHECK((sav2.bounds().index_bounds() == multi_span_index<2>{3, 2}));
+ CHECK((sav2.bounds().strides() == multi_span_index<2>{2, 1}));
+ CHECK((sav2[{0, 0}] == 1 && sav2[{2, 0}] == 5));
+ }
+
+ // Check const-casting assignment operator
+ {
+ int arr1[2] = {1, 2};
+ int arr2[6] = {3, 4, 5, 6, 7, 8};
+
+ const strided_span<int, 1> src{arr1, {{2}, {1}}};
+ strided_span<const int, 1> sav{arr2, {{3}, {2}}};
+ strided_span<const int, 1>& sav_ref = (sav = src);
+ CHECK(sav.bounds().index_bounds() == multi_span_index<1>{2});
+ CHECK(sav.bounds().strides() == multi_span_index<1>{1});
+ CHECK(sav[0] == 1);
+ CHECK(&sav_ref == &sav);
+ }
+
+ // Check copy assignment operator
+ {
+ int arr1[2] = {3, 4};
+ int arr1b[1] = {0};
+ const strided_span<int, 1> src1{arr1, {2, 1}};
+ strided_span<int, 1> sav1{arr1b, {1, 1}};
+ strided_span<int, 1>& sav1_ref = (sav1 = src1);
+ CHECK(sav1.bounds().index_bounds() == multi_span_index<1>{2});
+ CHECK(sav1.bounds().strides() == multi_span_index<1>{1});
+ CHECK(sav1[0] == 3);
+ CHECK(&sav1_ref == &sav1);
+
+ const int arr2[6] = {1, 2, 3, 4, 5, 6};
+ const int arr2b[1] = {0};
+ const strided_span<const int, 2> src2{arr2, {{3, 2}, {2, 1}}};
+ strided_span<const int, 2> sav2{arr2b, {{1, 1}, {1, 1}}};
+ strided_span<const int, 2>& sav2_ref = (sav2 = src2);
+ CHECK((sav2.bounds().index_bounds() == multi_span_index<2>{3, 2}));
+ CHECK((sav2.bounds().strides() == multi_span_index<2>{2, 1}));
+ CHECK((sav2[{0, 0}] == 1 && sav2[{2, 0}] == 5));
+ CHECK(&sav2_ref == &sav2);
+ }
+}
+
+TEST_CASE("strided_span_slice")
+{
+ std::vector<int> data(5 * 10);
+ std::iota(begin(data), end(data), 0);
+ const multi_span<int, 5, 10> src =
+ as_multi_span(multi_span<int>{data}, dim<5>(), dim<10>());
+
+ const strided_span<int, 2> sav{src, {{5, 10}, {10, 1}}};
+#ifdef CONFIRM_COMPILATION_ERRORS
+ const strided_span<const int, 2> csav{{src}, {{5, 10}, {10, 1}}};
+#endif
+ const strided_span<const int, 2> csav{multi_span<const int, 5, 10>{src},
+ {{5, 10}, {10, 1}}};
+
+ strided_span<int, 1> sav_sl = sav[2];
+ CHECK(sav_sl[0] == 20);
+ CHECK(sav_sl[9] == 29);
+
+ strided_span<const int, 1> csav_sl = sav[3];
+ CHECK(csav_sl[0] == 30);
+ CHECK(csav_sl[9] == 39);
+
+ CHECK(sav[4][0] == 40);
+ CHECK(sav[4][9] == 49);
+}
+
+TEST_CASE("strided_span_column_major")
+{
+ // strided_span may be used to accommodate more peculiar
+ // use cases, such as column-major multidimensional array
+ // (aka. "FORTRAN" layout).
+
+ int cm_array[3 * 5] = {1, 4, 7, 10, 13, 2, 5, 8, 11, 14, 3, 6, 9, 12, 15};
+ strided_span<int, 2> cm_sav{cm_array, {{5, 3}, {1, 5}}};
+
+ // Accessing elements
+ CHECK((cm_sav[{0, 0}] == 1));
+ CHECK((cm_sav[{0, 1}] == 2));
+ CHECK((cm_sav[{1, 0}] == 4));
+ CHECK((cm_sav[{4, 2}] == 15));
+
+ // Slice
+ strided_span<int, 1> cm_sl = cm_sav[3];
+
+ CHECK(cm_sl[0] == 10);
+ CHECK(cm_sl[1] == 11);
+ CHECK(cm_sl[2] == 12);
+
+ // Section
+ strided_span<int, 2> cm_sec = cm_sav.section({2, 1}, {3, 2});
+
+ CHECK((cm_sec.bounds().index_bounds() == multi_span_index<2>{3, 2}));
+ CHECK((cm_sec[{0, 0}] == 8));
+ CHECK((cm_sec[{0, 1}] == 9));
+ CHECK((cm_sec[{1, 0}] == 11));
+ CHECK((cm_sec[{2, 1}] == 15));
+}
+
+TEST_CASE("strided_span_bounds")
+{
+ int arr[] = {0, 1, 2, 3};
+ multi_span<int> av(arr);
+
+ {
+ // incorrect sections
+
+ CHECK_THROWS_AS(av.section(0, 0)[0], fail_fast);
+ CHECK_THROWS_AS(av.section(1, 0)[0], fail_fast);
+ CHECK_THROWS_AS(av.section(1, 1)[1], fail_fast);
+
+ CHECK_THROWS_AS(av.section(2, 5), fail_fast);
+ CHECK_THROWS_AS(av.section(5, 2), fail_fast);
+ CHECK_THROWS_AS(av.section(5, 0), fail_fast);
+ CHECK_THROWS_AS(av.section(0, 5), fail_fast);
+ CHECK_THROWS_AS(av.section(5, 5), fail_fast);
+ }
+
+ {
+ // zero stride
+ strided_span<int, 1> sav{av, {{4}, {}}};
+ CHECK(sav[0] == 0);
+ CHECK(sav[3] == 0);
+ CHECK_THROWS_AS(sav[4], fail_fast);
+ }
+
+ {
+ // zero extent
+ strided_span<int, 1> sav{av, {{}, {1}}};
+ CHECK_THROWS_AS(sav[0], fail_fast);
+ }
+
+ {
+ // zero extent and stride
+ strided_span<int, 1> sav{av, {{}, {}}};
+ CHECK_THROWS_AS(sav[0], fail_fast);
+ }
+
+ {
+ // strided array ctor with matching strided bounds
+ strided_span<int, 1> sav{arr, {4, 1}};
+ CHECK(sav.bounds().index_bounds() == multi_span_index<1>{4});
+ CHECK(sav[3] == 3);
+ CHECK_THROWS_AS(sav[4], fail_fast);
+ }
+
+ {
+ // strided array ctor with smaller strided bounds
+ strided_span<int, 1> sav{arr, {2, 1}};
+ CHECK(sav.bounds().index_bounds() == multi_span_index<1>{2});
+ CHECK(sav[1] == 1);
+ CHECK_THROWS_AS(sav[2], fail_fast);
+ }
+
+ {
+ // strided array ctor with fitting irregular bounds
+ strided_span<int, 1> sav{arr, {2, 3}};
+ CHECK(sav.bounds().index_bounds() == multi_span_index<1>{2});
+ CHECK(sav[0] == 0);
+ CHECK(sav[1] == 3);
+ CHECK_THROWS_AS(sav[2], fail_fast);
+ }
+
+ {
+ // bounds cross data boundaries - from static arrays
+ CHECK_THROWS_AS((strided_span<int, 1>{arr, {3, 2}}), fail_fast);
+ CHECK_THROWS_AS((strided_span<int, 1>{arr, {3, 3}}), fail_fast);
+ CHECK_THROWS_AS((strided_span<int, 1>{arr, {4, 5}}), fail_fast);
+ CHECK_THROWS_AS((strided_span<int, 1>{arr, {5, 1}}), fail_fast);
+ CHECK_THROWS_AS((strided_span<int, 1>{arr, {5, 5}}), fail_fast);
+ }
+
+ {
+ // bounds cross data boundaries - from array view
+ CHECK_THROWS_AS((strided_span<int, 1>{av, {3, 2}}), fail_fast);
+ CHECK_THROWS_AS((strided_span<int, 1>{av, {3, 3}}), fail_fast);
+ CHECK_THROWS_AS((strided_span<int, 1>{av, {4, 5}}), fail_fast);
+ CHECK_THROWS_AS((strided_span<int, 1>{av, {5, 1}}), fail_fast);
+ CHECK_THROWS_AS((strided_span<int, 1>{av, {5, 5}}), fail_fast);
+ }
+
+ {
+ // bounds cross data boundaries - from dynamic arrays
+ CHECK_THROWS_AS((strided_span<int, 1>{av.data(), 4, {3, 2}}), fail_fast);
+ CHECK_THROWS_AS((strided_span<int, 1>{av.data(), 4, {3, 3}}), fail_fast);
+ CHECK_THROWS_AS((strided_span<int, 1>{av.data(), 4, {4, 5}}), fail_fast);
+ CHECK_THROWS_AS((strided_span<int, 1>{av.data(), 4, {5, 1}}), fail_fast);
+ CHECK_THROWS_AS((strided_span<int, 1>{av.data(), 4, {5, 5}}), fail_fast);
+ CHECK_THROWS_AS((strided_span<int, 1>{av.data(), 2, {2, 2}}), fail_fast);
+ }
+
+#ifdef CONFIRM_COMPILATION_ERRORS
+ {
+ strided_span<int, 1> sav0{av.data(), {3, 2}};
+ strided_span<int, 1> sav1{arr, {1}};
+ strided_span<int, 1> sav2{arr, {1, 1, 1}};
+ strided_span<int, 1> sav3{av, {1}};
+ strided_span<int, 1> sav4{av, {1, 1, 1}};
+ strided_span<int, 2> sav5{av.as_multi_span(dim<2>(), dim<2>()), {1}};
+ strided_span<int, 2> sav6{av.as_multi_span(dim<2>(), dim<2>()), {1, 1, 1}};
+ strided_span<int, 2> sav7{av.as_multi_span(dim<2>(), dim<2>()),
+ {{1, 1}, {1, 1}, {1, 1}}};
+
+ multi_span_index<1> index{0, 1};
+ strided_span<int, 1> sav8{arr, {1, {1, 1}}};
+ strided_span<int, 1> sav9{arr, {{1, 1}, {1, 1}}};
+ strided_span<int, 1> sav10{av, {1, {1, 1}}};
+ strided_span<int, 1> sav11{av, {{1, 1}, {1, 1}}};
+ strided_span<int, 2> sav12{av.as_multi_span(dim<2>(), dim<2>()), {{1}, {1}}};
+ strided_span<int, 2> sav13{av.as_multi_span(dim<2>(), dim<2>()), {{1}, {1, 1, 1}}};
+ strided_span<int, 2> sav14{av.as_multi_span(dim<2>(), dim<2>()), {{1, 1, 1}, {1}}};
+ }
+#endif
+}
+
+TEST_CASE("strided_span_type_conversion")
+{
+ int arr[] = {0, 1, 2, 3};
+ multi_span<int> av(arr);
+
+ {
+ strided_span<int, 1> sav{av.data(), av.size(), {av.size() / 2, 2}};
+#ifdef CONFIRM_COMPILATION_ERRORS
+ strided_span<long, 1> lsav1 = sav.as_strided_span<long, 1>();
+#endif
+ }
+ {
+ strided_span<int, 1> sav{av, {av.size() / 2, 2}};
+#ifdef CONFIRM_COMPILATION_ERRORS
+ strided_span<long, 1> lsav1 = sav.as_strided_span<long, 1>();
+#endif
+ }
+
+ multi_span<const byte, dynamic_range> bytes = as_bytes(av);
+
+ // retype strided array with regular strides - from raw data
+ {
+ strided_bounds<2> bounds{{2, bytes.size() / 4}, {bytes.size() / 2, 1}};
+ strided_span<const byte, 2> sav2{bytes.data(), bytes.size(), bounds};
+ strided_span<const int, 2> sav3 = sav2.as_strided_span<const int>();
+ CHECK(sav3[0][0] == 0);
+ CHECK(sav3[1][0] == 2);
+ CHECK_THROWS_AS(sav3[1][1], fail_fast);
+ CHECK_THROWS_AS(sav3[0][1], fail_fast);
+ }
+
+ // retype strided array with regular strides - from multi_span
+ {
+ strided_bounds<2> bounds{{2, bytes.size() / 4}, {bytes.size() / 2, 1}};
+ multi_span<const byte, 2, dynamic_range> bytes2 =
+ as_multi_span(bytes, dim<2>(), dim(bytes.size() / 2));
+ strided_span<const byte, 2> sav2{bytes2, bounds};
+ strided_span<int, 2> sav3 = sav2.as_strided_span<int>();
+ CHECK(sav3[0][0] == 0);
+ CHECK(sav3[1][0] == 2);
+ CHECK_THROWS_AS(sav3[1][1], fail_fast);
+ CHECK_THROWS_AS(sav3[0][1], fail_fast);
+ }
+
+ // retype strided array with not enough elements - last dimension of the array is too small
+ {
+ strided_bounds<2> bounds{{4, 2}, {4, 1}};
+ multi_span<const byte, 2, dynamic_range> bytes2 =
+ as_multi_span(bytes, dim<2>(), dim(bytes.size() / 2));
+ strided_span<const byte, 2> sav2{bytes2, bounds};
+ CHECK_THROWS_AS(sav2.as_strided_span<int>(), fail_fast);
+ }
+
+ // retype strided array with not enough elements - strides are too small
+ {
+ strided_bounds<2> bounds{{4, 2}, {2, 1}};
+ multi_span<const byte, 2, dynamic_range> bytes2 =
+ as_multi_span(bytes, dim<2>(), dim(bytes.size() / 2));
+ strided_span<const byte, 2> sav2{bytes2, bounds};
+ CHECK_THROWS_AS(sav2.as_strided_span<int>(), fail_fast);
+ }
+
+ // retype strided array with not enough elements - last dimension does not divide by the new
+ // typesize
+ {
+ strided_bounds<2> bounds{{2, 6}, {4, 1}};
+ multi_span<const byte, 2, dynamic_range> bytes2 =
+ as_multi_span(bytes, dim<2>(), dim(bytes.size() / 2));
+ strided_span<const byte, 2> sav2{bytes2, bounds};
+ CHECK_THROWS_AS(sav2.as_strided_span<int>(), fail_fast);
+ }
+
+ // retype strided array with not enough elements - strides does not divide by the new
+ // typesize
+ {
+ strided_bounds<2> bounds{{2, 1}, {6, 1}};
+ multi_span<const byte, 2, dynamic_range> bytes2 =
+ as_multi_span(bytes, dim<2>(), dim(bytes.size() / 2));
+ strided_span<const byte, 2> sav2{bytes2, bounds};
+ CHECK_THROWS_AS(sav2.as_strided_span<int>(), fail_fast);
+ }
+
+ // retype strided array with irregular strides - from raw data
+ {
+ strided_bounds<1> bounds{bytes.size() / 2, 2};
+ strided_span<const byte, 1> sav2{bytes.data(), bytes.size(), bounds};
+ CHECK_THROWS_AS(sav2.as_strided_span<int>(), fail_fast);
+ }
+
+ // retype strided array with irregular strides - from multi_span
+ {
+ strided_bounds<1> bounds{bytes.size() / 2, 2};
+ strided_span<const byte, 1> sav2{bytes, bounds};
+ CHECK_THROWS_AS(sav2.as_strided_span<int>(), fail_fast);
+ }
+}
+
+TEST_CASE("empty_strided_spans")
+{
+ {
+ multi_span<int, 0> empty_av(nullptr);
+ strided_span<int, 1> empty_sav{empty_av, {0, 1}};
+
+ CHECK(empty_sav.bounds().index_bounds() == multi_span_index<1>{0});
+ CHECK_THROWS_AS(empty_sav[0], fail_fast);
+ CHECK_THROWS_AS(empty_sav.begin()[0], fail_fast);
+ CHECK_THROWS_AS(empty_sav.cbegin()[0], fail_fast);
+
+ for (const auto& v : empty_sav) {
+ (void) v;
+ CHECK(false);
+ }
+ }
+
+ {
+ strided_span<int, 1> empty_sav{nullptr, 0, {0, 1}};
+
+ CHECK(empty_sav.bounds().index_bounds() == multi_span_index<1>{0});
+ CHECK_THROWS_AS(empty_sav[0], fail_fast);
+ CHECK_THROWS_AS(empty_sav.begin()[0], fail_fast);
+ CHECK_THROWS_AS(empty_sav.cbegin()[0], fail_fast);
+
+ for (const auto& v : empty_sav) {
+ (void) v;
+ CHECK(false);
+ }
+ }
+}
+
+void iterate_every_other_element(multi_span<int, dynamic_range> av)
+{
+ // pick every other element
+
+ auto length = av.size() / 2;
+#if _MSC_VER > 1800
+ auto bounds = strided_bounds<1>({length}, {2});
+#else
+ auto bounds = strided_bounds<1>(multi_span_index<1>{length}, multi_span_index<1>{2});
+#endif
+ strided_span<int, 1> strided(&av.data()[1], av.size() - 1, bounds);
+
+ CHECK(strided.size() == length);
+ CHECK(strided.bounds().index_bounds()[0] == length);
+ for (auto i = 0; i < strided.size(); ++i) {
+ CHECK(strided[i] == av[2 * i + 1]);
+ }
+
+ int idx = 0;
+ for (auto num : strided) {
+ CHECK(num == av[2 * idx + 1]);
+ idx++;
+ }
+}
+
+TEST_CASE("strided_span_section_iteration")
+{
+ int arr[8] = {4, 0, 5, 1, 6, 2, 7, 3};
+
+ // static bounds
+ {
+ multi_span<int, 8> av(arr, 8);
+ iterate_every_other_element(av);
+ }
+
+ // dynamic bounds
+ {
+ multi_span<int, dynamic_range> av(arr, 8);
+ iterate_every_other_element(av);
+ }
+}
+
+TEST_CASE("dynamic_strided_span_section_iteration")
+{
+ auto arr = new int[8];
+ for (int i = 0; i < 4; ++i) {
+ arr[2 * i] = 4 + i;
+ arr[2 * i + 1] = i;
+ }
+
+ auto av = as_multi_span(arr, 8);
+ iterate_every_other_element(av);
+
+ delete[] arr;
+}
+
+void iterate_second_slice(multi_span<int, dynamic_range, dynamic_range, dynamic_range> av)
+{
+ const int expected[6] = {2, 3, 10, 11, 18, 19};
+ auto section = av.section({0, 1, 0}, {3, 1, 2});
+
+ for (auto i = 0; i < section.extent<0>(); ++i) {
+ for (auto j = 0; j < section.extent<1>(); ++j)
+ for (auto k = 0; k < section.extent<2>(); ++k) {
+ auto idx = multi_span_index<3>{i, j, k}; // avoid braces in the CHECK macro
+ CHECK(section[idx] == expected[2 * i + 2 * j + k]);
+ }
+ }
+
+ for (auto i = 0; i < section.extent<0>(); ++i) {
+ for (auto j = 0; j < section.extent<1>(); ++j)
+ for (auto k = 0; k < section.extent<2>(); ++k)
+ CHECK(section[i][j][k] == expected[2 * i + 2 * j + k]);
+ }
+
+ int i = 0;
+ for (const auto num : section) {
+ CHECK(num == expected[i]);
+ i++;
+ }
+}
+
+TEST_CASE("strided_span_section_iteration_3d")
+{
+ int arr[3][4][2]{};
+ for (auto i = 0; i < 3; ++i) {
+ for (auto j = 0; j < 4; ++j)
+ for (auto k = 0; k < 2; ++k) arr[i][j][k] = 8 * i + 2 * j + k;
+ }
+
+ {
+ multi_span<int, 3, 4, 2> av = arr;
+ iterate_second_slice(av);
+ }
+}
+
+TEST_CASE("dynamic_strided_span_section_iteration_3d")
+{
+ const auto height = 12, width = 2;
+ const auto size = height * width;
+
+ auto arr = new int[static_cast<std::size_t>(size)];
+ for (auto i = 0; i < size; ++i) {
+ arr[i] = i;
+ }
+
+ {
+ auto av = as_multi_span(as_multi_span(arr, 24), dim<3>(), dim<4>(), dim<2>());
+ iterate_second_slice(av);
+ }
+
+ {
+ auto av = as_multi_span(as_multi_span(arr, 24), dim(3), dim<4>(), dim<2>());
+ iterate_second_slice(av);
+ }
+
+ {
+ auto av = as_multi_span(as_multi_span(arr, 24), dim<3>(), dim(4), dim<2>());
+ iterate_second_slice(av);
+ }
+
+ {
+ auto av = as_multi_span(as_multi_span(arr, 24), dim<3>(), dim<4>(), dim(2));
+ iterate_second_slice(av);
+ }
+ delete[] arr;
+}
+
+TEST_CASE("strided_span_conversion")
+{
+ // get an multi_span of 'c' values from the list of X's
+
+ struct X
+ {
+ int a;
+ int b;
+ int c;
+ };
+
+ X arr[4] = {{0, 1, 2}, {3, 4, 5}, {6, 7, 8}, {9, 10, 11}};
+
+ int s = sizeof(int) / sizeof(byte);
+ auto d2 = 3 * s;
+ auto d1 = narrow_cast<int>(sizeof(int)) * 12 / d2;
+
+ // convert to 4x12 array of bytes
+ auto av = as_multi_span(as_bytes(as_multi_span(arr, 4)), dim(d1), dim(d2));
+
+ CHECK(av.bounds().index_bounds()[0] == 4);
+ CHECK(av.bounds().index_bounds()[1] == 12);
+
+ // get the last 4 columns
+ auto section = av.section({0, 2 * s}, {4, s}); // { { arr[0].c[0], arr[0].c[1], arr[0].c[2],
+ // arr[0].c[3] } , { arr[1].c[0], ... } , ...
+ // }
+
+ // convert to array 4x1 array of integers
+ auto cs = section.as_strided_span<int>(); // { { arr[0].c }, {arr[1].c } , ... }
+
+ CHECK(cs.bounds().index_bounds()[0] == 4);
+ CHECK(cs.bounds().index_bounds()[1] == 1);
+
+ // transpose to 1x4 array
+ strided_bounds<2> reverse_bounds{
+ {cs.bounds().index_bounds()[1], cs.bounds().index_bounds()[0]},
+ {cs.bounds().strides()[1], cs.bounds().strides()[0]}};
+
+ strided_span<int, 2> transposed{cs.data(), cs.bounds().total_size(), reverse_bounds};
+
+ // slice to get a one-dimensional array of c's
+ strided_span<int, 1> result = transposed[0];
+
+ CHECK(result.bounds().index_bounds()[0] == 4);
+ CHECK_THROWS_AS(result.bounds().index_bounds()[1], fail_fast);
+
+ int i = 0;
+ for (auto& num : result) {
+ CHECK(num == arr[i].c);
+ i++;
+ }
+}