Namespaces
Variants

    std::strided_slice

    From cppreference.com
     
    C++
     
    Containers library
     
    std::mdspan
     
    Defined in header <mdspan>
    template< class OffsetType, class ExtentType, class StrideType > struct strided_slice;
    		 (since C++26)
    See More

    An instance of every specialization of strided_slice is a slice specifier used in std::submdspan to select a subset of elements using a set of regularly spaced indices at a specified dimension in std::mdspan.

    Each strided_slice object s is characterized by three data members: the offset index s.offset, the extent s.extent, and the stride s.stride.

    Given that s.stride is greater than zero, the number of selected indices denoted by N is determined by 1 + (s.extent - 1) / s.stride if s.extent is nonzero, otherwise 0. The half-open interval from which indices will be selected is given by [s.offset, s.offset + s.extent). The sequence of selected indices is produced as follows: s.offset, ..., s.offset + (N - 1) * s.stride.

    This class template has no base classes or declared members other than those shown below.

    Template parameters

    OffsetType - the type of offset
    ExtentType - the type of extent
    StrideType - the type of stride
    Type requirements
    -
    All template parameters must be unsigned or signed integer types or must satisfy integral-constant-like

    The program is ill-formed if the type requirement is not satisfied.

    Member types

    Member type Definition
    offset_type OffsetType
    extent_type ExtentType
    stride_type StrideType

    Data members

    Member name Definition
    offset
    		 a starting index of type offset_type
    (public member object)
    extent
    		 a value of type extent_type added to offset used to define the upper bound of indices
    (public member object)
    stride
    		 an increment value of type stride_type which is equivalent to the distance between two indices
    (public member object)

    All these members are declared with [[no_unique_address]] attribute and have default member initializers where each data member is value initialized.

    Notes

    Every specialization of strided_slice is an aggregate class which allows aggregate initialization (including designated initialization) of data members (for instance, std::strided_slice{.offset = 2, .extent = 10, .stride = 3}).

    The slice specification of strided_slice takes advantage of the data member extent, as opposed to other slice specifications that use end to indicate the upper bound value. This is because it can directly generate a static extent for subview of std::mdspan if both extent and stride are of types that satisfy integral-constant-like. This enables efficient extraction of subviews with static extents by mixing compile time values with a runtime value of offset.

    Example

    Run this code
    #include <mdspan>
#include <print>

template <typename View, typename O = int, typename E = int, typename S = int>
    requires (View::extents_type::rank() == 1)
void print_sliced_view(View v, std::strided_slice<O, E, S> s)
{
    using index_type = View::index_type;
    auto subview = std::submdspan(v, s);
    const auto& submap = subview.mapping();

    std::print("[");
    bool uses_comma = false;
    for (index_type i = 0; i != subview.extent(0); ++i)
    {
        if (uses_comma)
            std::print(", ");
        std::print("{}", subview[i]);
        uses_comma = true;
    }
    uses_comma = false;
    std::print("] extracted from indices [");
    for (index_type i = 0; i != subview.extent(0); ++i)
    {
        if (uses_comma)
            std::print(", ");
        std::print("{}", submap(i) + s.offset);
        uses_comma = true;
    }
    std::println("]");
}

int main()
{
    static constexpr char letters[]
    {
        'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J',
        'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T',
        'U', 'V', 'W', 'X', 'Y', 'Z'
    };
    constexpr std::mdspan md(letters, 26);
    print_sliced_view(md, {.offset = 0, .extent = 10, .stride = 1});
    print_sliced_view(md, {.offset = 2, .extent = 10, .stride = 1});
    print_sliced_view(md, {.offset = 0, .extent = 5,  .stride = 1});
    print_sliced_view(md, {.offset = 2, .extent = 5,  .stride = 1});
    print_sliced_view(md, {.offset = 0, .extent = 10, .stride = 2});
    print_sliced_view(md, {.offset = 2, .extent = 10, .stride = 3});
    print_sliced_view(md, {.offset = 0, .extent = 15, .stride = 5});
    print_sliced_view(md, {.offset = 6, .extent = 15, .stride = 5});
}

    Output: 

    [A, B, C, D, E, F, G, H, I, J] extracted from indices [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
[C, D, E, F, G, H, I, J, K, L] extracted from indices [2, 3, 4, 5, 6, 7, 8, 9, 10, 11]
[A, B, C, D, E] extracted from indices [0, 1, 2, 3, 4]
[C, D, E, F, G] extracted from indices [2, 3, 4, 5, 6]
[A, C, E, G, I] extracted from indices [0, 2, 4, 6, 8]
[C, F, I, L] extracted from indices [2, 5, 8, 11]
[A, F, K] extracted from indices [0, 5, 10]
[G, L, Q] extracted from indices [6, 11, 16]

    See also

    		 BLAS-like slice of a valarray: starting index, length, stride
    (class) [edit]
    (C++26)
    		 returns a view of a subset of an existing mdspan
    (function template) [edit]