NdArrayCore.hpp

Go to the documentation of this file.

 
#pragma once
 
#include <array>
#include <cmath>
#include <deque>
#include <filesystem>
#include <forward_list>
#include <fstream>
#include <initializer_list>
#include <iostream>
#include <iterator>
#include <list>
#include <memory>
#include <numeric>
#include <set>
#include <string>
#include <type_traits>
#include <utility>
#include <vector>
 
#include "NumCpp/Core/Constants.hpp"
#include "NumCpp/Core/DtypeInfo.hpp"
#include "NumCpp/Core/Enums.hpp"
#include "NumCpp/Core/Internal/Endian.hpp"
#include "NumCpp/Core/Internal/Error.hpp"
#include "NumCpp/Core/Internal/StaticAsserts.hpp"
#include "NumCpp/Core/Internal/StdComplexOperators.hpp"
#include "NumCpp/Core/Internal/StlAlgorithms.hpp"
#include "NumCpp/Core/Internal/TypeTraits.hpp"
#include "NumCpp/Core/Shape.hpp"
#include "NumCpp/Core/Slice.hpp"
#include "NumCpp/Core/Types.hpp"
#include "NumCpp/NdArray/NdArrayIterators.hpp"
#include "NumCpp/Utils/essentiallyEqual.hpp"
#include "NumCpp/Utils/essentiallyEqualComplex.hpp"
#include "NumCpp/Utils/num2str.hpp"
#include "NumCpp/Utils/power.hpp"
#include "NumCpp/Utils/sqr.hpp"
#include "NumCpp/Utils/value2str.hpp"
 
namespace nc
{
    namespace type_traits
    {
        //============================================================================
        // Class Description:
        template<typename>
        struct is_ndarray_int : std::false_type
        {
        };
 
        //============================================================================
        // Class Description:
 
        template<typename dtype, typename Allocator>
        struct is_ndarray_int<NdArray<dtype, Allocator>>
        {
            static constexpr bool value = std::is_integral_v<dtype>;
        };
 
        //============================================================================
        // Class Description:
        template<typename T>
        constexpr bool is_ndarray_int_v = is_ndarray_int<T>::value;
 
        //============================================================================
        // Class Description:
        template<typename>
        struct is_ndarray_signed_int : std::false_type
        {
        };
 
        //============================================================================
        // Class Description:
 
        template<typename dtype, typename Allocator>
        struct is_ndarray_signed_int<NdArray<dtype, Allocator>>
        {
            static constexpr bool value = std::is_signed_v<dtype>;
        };
 
        //============================================================================
        // Class Description:
        template<typename T>
        constexpr bool is_ndarray_signed_int_v = is_ndarray_signed_int<T>::value;
 
        //============================================================================
        // Class Description:
        template<typename T>
        using ndarray_int_concept = std::enable_if_t<is_ndarray_int_v<T>, int>;
    } // namespace type_traits
 
    //================================================================================
    // Class Description:
    template<typename dtype, class Allocator = std::allocator<dtype>>
    class NdArray
    {
    private:
        STATIC_ASSERT_VALID_DTYPE(dtype);
        static_assert(std::is_same_v<dtype, typename Allocator::value_type>,
                      "value_type and Allocator::value_type must match");
 
        using AllocType   = typename std::allocator_traits<Allocator>::template rebind_alloc<dtype>;
        using AllocTraits = std::allocator_traits<AllocType>;
 
    public:
        using self_type       = NdArray<dtype, Allocator>;
        using value_type      = dtype;
        using allocator_type  = Allocator;
        using pointer         = typename AllocTraits::pointer;
        using const_pointer   = typename AllocTraits::const_pointer;
        using reference       = dtype&;
        using const_reference = const dtype&;
        using size_type       = uint32;
        using index_type      = int32;
        using difference_type = typename AllocTraits::difference_type;
 
        using iterator               = NdArrayIterator<dtype, pointer, difference_type>;
        using const_iterator         = NdArrayConstIterator<dtype, const_pointer, difference_type>;
        using reverse_iterator       = std::reverse_iterator<iterator>;
        using const_reverse_iterator = std::reverse_iterator<const_iterator>;
 
        using column_iterator         = NdArrayColumnIterator<dtype, size_type, pointer, difference_type>;
        using const_column_iterator   = NdArrayConstColumnIterator<dtype, size_type, const_pointer, difference_type>;
        using reverse_column_iterator = std::reverse_iterator<column_iterator>;
        using const_reverse_column_iterator = std::reverse_iterator<const_column_iterator>;
 
        //============================================================================
        // Method Description:
        NdArray() = default;
 
        //============================================================================
        // Method Description:
        explicit NdArray(size_type inSquareSize) :
            shape_{ inSquareSize, inSquareSize },
            size_{ inSquareSize * inSquareSize }
        {
            newArray();
        }
 
        //============================================================================
        // Method Description:
        NdArray(size_type inNumRows, size_type inNumCols) :
            shape_{ inNumRows, inNumCols },
            size_{ inNumRows * inNumCols }
        {
            newArray();
        }
 
        //============================================================================
        // Method Description:
        explicit NdArray(const Shape& inShape) :
            shape_{ inShape },
            size_{ shape_.size() }
        {
            newArray();
        }
 
        //============================================================================
        // Method Description:
        NdArray(std::initializer_list<dtype> inList) :
            shape_{ 1, static_cast<uint32>(inList.size()) },
            size_{ shape_.size() }
        {
            newArray();
            if (size_ > 0)
            {
                stl_algorithms::copy(inList.begin(), inList.end(), begin());
            }
        }
 
        //============================================================================
        // Method Description:
        NdArray(const std::initializer_list<std::initializer_list<dtype>>& inList) :
            shape_{ static_cast<uint32>(inList.size()), 0 }
        {
            for (const auto& list : inList)
            {
                if (shape_.cols == 0)
                {
                    shape_.cols = static_cast<uint32>(list.size());
                }
                else if (list.size() != shape_.cols)
                {
                    THROW_INVALID_ARGUMENT_ERROR(
                        "All rows of the initializer list needs to have the same number of elements");
                }
            }
 
            size_ = shape_.size();
            newArray();
            uint32 row = 0;
            for (const auto& list : inList)
            {
                const auto ptr = begin() += row * shape_.cols;
                stl_algorithms::copy(list.begin(), list.end(), ptr);
                ++row;
            }
        }
 
        //============================================================================
        // Method Description:
        template<size_t ArraySize, std::enable_if_t<is_valid_dtype_v<dtype>, int> = 0>
        NdArray(std::array<dtype, ArraySize>& inArray, PointerPolicy policy = PointerPolicy::COPY) :
            shape_{ 1, static_cast<uint32>(ArraySize) },
            size_{ shape_.size() }
        {
            switch (policy)
            {
                case PointerPolicy::COPY:
                {
                    newArray();
                    if (size_ > 0)
                    {
                        stl_algorithms::copy(inArray.begin(), inArray.end(), begin());
                    }
                    break;
                }
                case PointerPolicy::SHELL:
                {
                    array_   = inArray.data();
                    ownsPtr_ = false;
                    break;
                }
                default:
                {
                    THROW_RUNTIME_ERROR("Unimplemented PointerPolicy type");
                }
            }
        }
 
        //============================================================================
        // Method Description:
        template<size_t Dim0Size, size_t Dim1Size>
        NdArray(std::array<std::array<dtype, Dim1Size>, Dim0Size>& in2dArray,
                PointerPolicy                                      policy = PointerPolicy::COPY) :
            shape_{ static_cast<uint32>(Dim0Size), static_cast<uint32>(Dim1Size) },
            size_{ shape_.size() }
        {
            switch (policy)
            {
                case PointerPolicy::COPY:
                {
                    newArray();
                    if (size_ > 0)
                    {
                        const auto start = in2dArray.front().begin();
                        stl_algorithms::copy(start, start + size_, begin());
                    }
                    break;
                }
                case PointerPolicy::SHELL:
                {
                    array_   = in2dArray.front().data();
                    ownsPtr_ = false;
                    break;
                }
                default:
                {
                    THROW_RUNTIME_ERROR("Unimplemented PointerPolicy type");
                }
            }
        }
 
        //============================================================================
        // Method Description:
        template<std::enable_if_t<is_valid_dtype_v<dtype>, int> = 0>
        NdArray(std::vector<dtype>& inVector, PointerPolicy policy = PointerPolicy::COPY) :
            shape_{ 1, static_cast<uint32>(inVector.size()) },
            size_{ shape_.size() }
        {
            switch (policy)
            {
                case PointerPolicy::COPY:
                {
                    newArray();
                    if (size_ > 0)
                    {
                        stl_algorithms::copy(inVector.begin(), inVector.end(), begin());
                    }
                    break;
                }
                case PointerPolicy::SHELL:
                {
                    array_   = inVector.data();
                    ownsPtr_ = false;
                    break;
                }
                default:
                {
                    THROW_RUNTIME_ERROR("Unimplemented PointerPolicy type");
                }
            }
        }
 
        //============================================================================
        // Method Description:
        explicit NdArray(const std::vector<std::vector<dtype>>& in2dVector) :
            shape_{ static_cast<uint32>(in2dVector.size()), 0 }
        {
            for (const auto& row : in2dVector)
            {
                if (shape_.cols == 0)
                {
                    shape_.cols = static_cast<uint32>(row.size());
                }
                else if (row.size() != shape_.cols)
                {
                    THROW_INVALID_ARGUMENT_ERROR("All rows of the 2d vector need to have the same number of elements");
                }
            }
 
            size_ = shape_.size();
 
            newArray();
            auto currentPosition = begin();
            for (const auto& row : in2dVector)
            {
                stl_algorithms::copy(row.begin(), row.end(), currentPosition);
                currentPosition += shape_.cols;
            }
        }
 
        //============================================================================
        // Method Description:
        template<size_t Dim1Size>
        NdArray(std::vector<std::array<dtype, Dim1Size>>& in2dArray, PointerPolicy policy = PointerPolicy::COPY) :
            shape_{ static_cast<uint32>(in2dArray.size()), static_cast<uint32>(Dim1Size) },
            size_{ shape_.size() }
        {
            switch (policy)
            {
                case PointerPolicy::COPY:
                {
                    newArray();
                    if (size_ > 0)
                    {
                        const auto start = in2dArray.front().begin();
                        stl_algorithms::copy(start, start + size_, begin());
                    }
                    break;
                }
                case PointerPolicy::SHELL:
                {
                    array_   = in2dArray.front().data();
                    ownsPtr_ = false;
                    break;
                }
                default:
                {
                    THROW_RUNTIME_ERROR("Unimplemented PointerPolicy type");
                }
            }
        }
 
        //============================================================================
        // Method Description:
        template<std::enable_if_t<is_valid_dtype_v<dtype>, int> = 0>
        explicit NdArray(const std::deque<dtype>& inDeque) :
            shape_{ 1, static_cast<uint32>(inDeque.size()) },
            size_{ shape_.size() }
        {
            newArray();
            if (size_ > 0)
            {
                stl_algorithms::copy(inDeque.begin(), inDeque.end(), begin());
            }
        }
 
        //============================================================================
        // Method Description:
        explicit NdArray(const std::deque<std::deque<dtype>>& in2dDeque) :
            shape_{ static_cast<uint32>(in2dDeque.size()), 0 }
        {
            for (const auto& row : in2dDeque)
            {
                if (shape_.cols == 0)
                {
                    shape_.cols = static_cast<uint32>(row.size());
                }
                else if (row.size() != shape_.cols)
                {
                    THROW_INVALID_ARGUMENT_ERROR("All rows of the 2d vector need to have the same number of elements");
                }
            }
 
            size_ = shape_.size();
 
            newArray();
            auto currentPosition = begin();
            for (const auto& row : in2dDeque)
            {
                stl_algorithms::copy(row.begin(), row.end(), currentPosition);
                currentPosition += shape_.cols;
            }
        }
 
        //============================================================================
        // Method Description:
        explicit NdArray(const std::list<dtype>& inList) :
            shape_{ 1, static_cast<uint32>(inList.size()) },
            size_{ shape_.size() }
        {
            newArray();
            if (size_ > 0)
            {
                stl_algorithms::copy(inList.begin(), inList.end(), begin());
            }
        }
 
        //============================================================================
        // Method Description:
        template<typename Iterator,
                 std::enable_if_t<std::is_same_v<typename std::iterator_traits<Iterator>::value_type, dtype>, int> = 0>
        NdArray(Iterator inFirst, Iterator inLast) :
            shape_{ 1, static_cast<uint32>(std::distance(inFirst, inLast)) },
            size_{ shape_.size() }
        {
            newArray();
            if (size_ > 0)
            {
                stl_algorithms::copy(inFirst, inLast, begin());
            }
        }
 
        //============================================================================
        // Method Description:
        template<typename UIntType,
                 std::enable_if_t<std::is_integral_v<UIntType> && !std::is_same_v<UIntType, bool>, int> = 0>
        NdArray(const_pointer inPtr, UIntType size) :
            NdArray(inPtr, 1, size)
        {
        }
 
        //============================================================================
        // Method Description:
        template<typename UIntType1,
                 typename UIntType2,
                 std::enable_if_t<std::is_integral_v<UIntType1> && !std::is_same_v<UIntType1, bool>, int> = 0,
                 std::enable_if_t<std::is_integral_v<UIntType2> && !std::is_same_v<UIntType2, bool>, int> = 0>
        NdArray(const_pointer inPtr, UIntType1 numRows, UIntType2 numCols) :
            shape_(numRows, numCols),
            size_{ shape_.size() }
        {
            newArray();
            if (inPtr != nullptr && size_ > 0)
            {
                stl_algorithms::copy(inPtr, inPtr + size_, begin());
            }
        }
 
        //============================================================================
        // Method Description:
        template<typename UIntType,
                 std::enable_if_t<std::is_integral_v<UIntType> && !std::is_same_v<UIntType, bool>, int> = 0>
        NdArray(pointer inPtr, UIntType size, PointerPolicy policy) :
            NdArray(inPtr, 1, size, policy)
        {
        }
 
        //============================================================================
        // Method Description:
        template<typename UIntType1,
                 typename UIntType2,
                 std::enable_if_t<std::is_integral_v<UIntType1> && !std::is_same_v<UIntType1, bool>, int> = 0,
                 std::enable_if_t<std::is_integral_v<UIntType2> && !std::is_same_v<UIntType2, bool>, int> = 0>
        NdArray(pointer inPtr, UIntType1 numRows, UIntType2 numCols, PointerPolicy policy) :
            shape_(numRows, numCols),
            size_{ shape_.size() }
        {
            switch (policy)
            {
                case PointerPolicy::COPY:
                {
                    newArray();
                    if (inPtr != nullptr && size_ > 0)
                    {
                        stl_algorithms::copy(inPtr, inPtr + size_, begin());
                    }
                    break;
                }
                case PointerPolicy::SHELL:
                {
                    array_   = inPtr;
                    ownsPtr_ = false;
                    break;
                }
                default:
                {
                    THROW_RUNTIME_ERROR("Unimplemented PointerPolicy type");
                }
            }
        }
 
        //============================================================================
        // Method Description:
        NdArray(const self_type& inOtherArray) :
            shape_{ inOtherArray.shape_ },
            size_{ inOtherArray.size_ },
            endianess_{ inOtherArray.endianess_ }
        {
            newArray();
            if (size_ > 0)
            {
                stl_algorithms::copy(inOtherArray.cbegin(), inOtherArray.cend(), begin());
            }
        }
 
        //============================================================================
        // Method Description:
        NdArray(self_type&& inOtherArray) noexcept :
            shape_{ inOtherArray.shape_ },
            size_{ inOtherArray.size_ },
            endianess_{ inOtherArray.endianess_ },
            array_{ inOtherArray.array_ },
            ownsPtr_{ inOtherArray.ownsPtr_ }
        {
            inOtherArray.shape_.rows = inOtherArray.shape_.cols = 0;
            inOtherArray.size_                                  = 0;
            inOtherArray.ownsPtr_                               = false;
            inOtherArray.array_                                 = nullptr;
        }
 
        //============================================================================
        // Method Description:
        ~NdArray() noexcept
        {
            deleteArray();
        }
 
        //============================================================================
        // Method Description:
        explicit operator bool() const noexcept
        {
            return isempty();
        }
 
        //============================================================================
        // Method Description:
        self_type& operator=(const self_type& rhs)
        {
            if (&rhs != this)
            {
                if (rhs.size_ > 0)
                {
                    newArray(rhs.shape_);
                    endianess_ = rhs.endianess_;
 
                    stl_algorithms::copy(rhs.cbegin(), rhs.cend(), begin());
                }
            }
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        self_type& operator=(value_type inValue) noexcept
        {
            if (array_ != nullptr)
            {
                stl_algorithms::fill(begin(), end(), inValue);
            }
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        self_type& operator=(self_type&& rhs) noexcept
        {
            if (&rhs != this)
            {
                deleteArray();
                shape_     = rhs.shape_;
                size_      = rhs.size_;
                endianess_ = rhs.endianess_;
                array_     = rhs.array_;
                ownsPtr_   = rhs.ownsPtr_;
 
                rhs.shape_.rows = rhs.shape_.cols = rhs.size_ = 0;
                rhs.array_                                    = nullptr;
                rhs.ownsPtr_                                  = false;
            }
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        reference operator[](index_type inIndex) noexcept
        {
            return const_cast<reference>(const_cast<const self_type*>(this)->operator[](inIndex));
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reference operator[](index_type inIndex) const noexcept
        {
            if (inIndex < 0)
            {
                inIndex += size_;
            }
 
            return array_[inIndex];
        }
 
        //============================================================================
        // Method Description:
        reference operator()(index_type inRowIndex, index_type inColIndex) noexcept
        {
            return const_cast<reference>(const_cast<const self_type*>(this)->operator()(inRowIndex, inColIndex));
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reference operator()(index_type inRowIndex, index_type inColIndex) const noexcept
        {
            if (inRowIndex < 0)
            {
                inRowIndex += shape_.rows;
            }
 
            if (inColIndex < 0)
            {
                inColIndex += shape_.cols;
            }
 
            return array_[inRowIndex * shape_.cols + inColIndex];
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type operator[](Slice inSlice) const
        {
            return operator[](toIndices(inSlice, Axis::NONE));
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type operator[](const NdArray<bool>& inMask) const
        {
            return operator[](inMask.flatnonzero());
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        [[nodiscard]] self_type operator[](const Indices& inIndices) const
        {
            auto      outArray = self_type(1, static_cast<size_type>(inIndices.size()));
            size_type i        = 0;
            for (auto& index : inIndices)
            {
                outArray[i++] = operator[](static_cast<index_type>(index));
            }
 
            return outArray;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type operator()(Slice inRowSlice, Slice inColSlice) const
        {
            return operator()(toIndices(inRowSlice, Axis::ROW), toIndices(inColSlice, Axis::COL));
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type operator()(Slice inRowSlice, index_type inColIndex) const
        {
            const NdArray<index_type> colIndices = { inColIndex };
            return operator()(toIndices(inRowSlice, Axis::ROW), colIndices);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type operator()(index_type inRowIndex, Slice inColSlice) const
        {
            const NdArray<index_type> rowIndices = { inRowIndex };
            return operator()(rowIndices, toIndices(inColSlice, Axis::COL));
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        [[nodiscard]] self_type operator()(const Indices& rowIndices, index_type colIndex) const
        {
            const NdArray<index_type> colIndices = { colIndex };
            return operator()(rowIndices, colIndices);
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        [[nodiscard]] self_type operator()(const Indices& rowIndices, Slice colSlice) const
        {
            return operator()(rowIndices, toIndices(colSlice, Axis::COL));
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        [[nodiscard]] self_type operator()(index_type rowIndex, const Indices& colIndices) const
        {
            const NdArray<index_type> rowIndices = { rowIndex };
            return operator()(rowIndices, colIndices);
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        [[nodiscard]] self_type operator()(Slice rowSlice, const Indices& colIndices) const
        {
            return operator()(toIndices(rowSlice, Axis::ROW), colIndices);
        }
 
        //============================================================================
        // Method Description:
        template<typename RowIndices,
                 typename ColIndices,
                 type_traits::ndarray_int_concept<RowIndices> = 0,
                 type_traits::ndarray_int_concept<ColIndices> = 0>
        [[nodiscard]] self_type operator()(const RowIndices& rowIndices, const ColIndices& colIndices) const
        {
            self_type returnArray(rowIndices.size(), colIndices.size());
 
            size_type rowCounter = 0;
            for (auto rowIter = rowIndices.begin(); rowIter != rowIndices.end(); ++rowIter)
            {
                size_type colCounter = 0;
                for (auto colIter = colIndices.begin(); colIter != colIndices.end(); ++colIter)
                {
                    returnArray(rowCounter, colCounter++) = operator()(*rowIter, *colIter);
                }
 
                ++rowCounter;
            }
 
            return returnArray;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] Slice cSlice(index_type inStartIdx = 0, size_type inStepSize = 1) const
        {
            return Slice(inStartIdx, shape_.cols, inStepSize); // NOLINT(modernize-return-braced-init-list)
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] Slice rSlice(index_type inStartIdx = 0, size_type inStepSize = 1) const
        {
            return Slice(inStartIdx, shape_.rows, inStepSize); // NOLINT(modernize-return-braced-init-list)
        }
 
        //============================================================================
        // Method Description:
        reference at(index_type inIndex)
        {
            return const_cast<reference>(const_cast<const self_type*>(this)->at(inIndex));
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reference at(index_type inIndex) const
        {
            // this doesn't allow for calling the first element as -size_...
            // but why would you really want to do that anyway?
            if (std::abs(inIndex) > static_cast<int64>(size_ - 1))
            {
                std::string errStr = "Input index " + utils::num2str(inIndex);
                errStr += " is out of bounds for array of size " + utils::num2str(size_) + ".";
                THROW_INVALID_ARGUMENT_ERROR(errStr);
            }
 
            return operator[](inIndex); // cppcheck-suppress returnTempReference
        }
 
        //============================================================================
        // Method Description:
        reference at(index_type inRowIndex, index_type inColIndex)
        {
            return const_cast<reference>(const_cast<const self_type*>(this)->at(inRowIndex, inColIndex));
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reference at(index_type inRowIndex, index_type inColIndex) const
        {
            // this doesn't allow for calling the first element as -size_...
            // but why would you really want to do that anyway?
            if (std::abs(inRowIndex) > static_cast<index_type>(shape_.rows - 1))
            {
                std::string errStr = "Row index " + utils::num2str(inRowIndex);
                errStr += " is out of bounds for array of size " + utils::num2str(shape_.rows) + ".";
                THROW_INVALID_ARGUMENT_ERROR(errStr);
            }
 
            // this doesn't allow for calling the first element as -size_...
            // but why would you really want to do that anyway?
            if (std::abs(inColIndex) > static_cast<index_type>(shape_.cols - 1))
            {
                std::string errStr = "Column index " + utils::num2str(inColIndex);
                errStr += " is out of bounds for array of size " + utils::num2str(shape_.cols) + ".";
                THROW_INVALID_ARGUMENT_ERROR(errStr);
            }
 
            return operator()(inRowIndex, inColIndex); // cppcheck-suppress returnTempReference
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type at(const Slice& inSlice) const
        {
            return at(toIndices(inSlice, Axis::NONE));
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type at(const NdArray<bool>& inMask) const
        {
            if (inMask.shape() != shape_)
            {
                THROW_INVALID_ARGUMENT_ERROR("Input mask must have the same dimensions as array.");
            }
 
            return operator[](inMask);
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        [[nodiscard]] self_type at(const Indices& inIndices) const
        {
            stl_algorithms::for_each(inIndices.begin(),
                                     inIndices.end(),
                                     [this](auto index)
                                     {
                                         auto indexSigned = static_cast<index_type>(index);
                                         if (indexSigned < 0)
                                         {
                                             indexSigned += size_;
                                         }
 
                                         if (indexSigned < 0 || indexSigned > static_cast<index_type>(size_ - 1))
                                         {
                                             THROW_INVALID_ARGUMENT_ERROR("Index exceeds matrix dimensions");
                                         }
                                     });
 
            return operator[](inIndices);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type at(const Slice& inRowSlice, const Slice& inColSlice) const
        {
            return at(toIndices(inRowSlice, Axis::ROW), toIndices(inColSlice, Axis::COL));
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type at(const Slice& inRowSlice, index_type inColIndex) const
        {
            const NdArray<index_type> colIndices = { inColIndex };
            return at(toIndices(inRowSlice, Axis::ROW), colIndices);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type at(index_type inRowIndex, const Slice& inColSlice) const
        {
            const NdArray<index_type> rowIndices = { inRowIndex };
            return at(rowIndices, toIndices(inColSlice, Axis::COL));
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        [[nodiscard]] self_type at(const Indices& rowIndices, index_type colIndex) const
        {
            const NdArray<index_type> colIndices = { colIndex };
            return at(rowIndices, colIndices);
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        [[nodiscard]] self_type at(const Indices& rowIndices, Slice colSlice) const
        {
            return at(rowIndices, toIndices(colSlice, Axis::COL));
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        [[nodiscard]] self_type at(index_type rowIndex, const Indices& colIndices) const
        {
            const NdArray<index_type> rowIndices = { rowIndex };
            return at(rowIndices, colIndices);
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        [[nodiscard]] self_type at(Slice rowSlice, const Indices& colIndices) const
        {
            return at(toIndices(rowSlice, Axis::ROW), colIndices);
        }
 
        //============================================================================
        // Method Description:
        template<typename RowIndices,
                 typename ColIndices,
                 type_traits::ndarray_int_concept<RowIndices> = 0,
                 type_traits::ndarray_int_concept<ColIndices> = 0>
        [[nodiscard]] self_type at(const RowIndices& rowIndices, const ColIndices& colIndices) const
        {
            stl_algorithms::for_each(rowIndices.begin(),
                                     rowIndices.end(),
                                     [this](auto row)
                                     {
                                         auto rowSigned = static_cast<index_type>(row);
                                         if (rowSigned < 0)
                                         {
                                             rowSigned += shape_.rows;
                                         }
 
                                         if (rowSigned < 0 || rowSigned > static_cast<index_type>(shape_.rows - 1))
                                         {
                                             THROW_INVALID_ARGUMENT_ERROR("Row index exceeds matrix dimensions");
                                         }
                                     });
 
            stl_algorithms::for_each(colIndices.begin(),
                                     colIndices.end(),
                                     [this](auto col)
                                     {
                                         auto colSigned = static_cast<index_type>(col);
                                         if (colSigned < 0)
                                         {
                                             colSigned += shape_.cols;
                                         }
 
                                         if (colSigned < 0 || colSigned > static_cast<index_type>(shape_.cols - 1))
                                         {
                                             THROW_INVALID_ARGUMENT_ERROR("Column index exceeds matrix dimensions");
                                         }
                                     });
 
            return operator()(rowIndices, colIndices);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] iterator begin() noexcept
        {
            return iterator(array_);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] iterator begin(size_type inRow)
        {
            if (inRow >= shape_.rows)
            {
                THROW_INVALID_ARGUMENT_ERROR("input row is greater than the number of rows in the array.");
            }
 
            return begin() += (inRow * shape_.cols);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_iterator begin() const noexcept
        {
            return cbegin();
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_iterator begin(size_type inRow) const
        {
            return cbegin(inRow);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_iterator cbegin() const noexcept
        {
            return const_iterator(array_);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_iterator cbegin(size_type inRow) const
        {
            if (inRow >= shape_.rows)
            {
                THROW_INVALID_ARGUMENT_ERROR("input row is greater than the number of rows in the array.");
            }
 
            return cbegin() += (inRow * shape_.cols);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] column_iterator colbegin() noexcept
        {
            return column_iterator(array_, shape_.rows, shape_.cols);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] column_iterator colbegin(size_type inCol)
        {
            if (inCol >= shape_.cols)
            {
                THROW_INVALID_ARGUMENT_ERROR("input col is greater than the number of cols in the array.");
            }
 
            return colbegin() += (inCol * shape_.rows);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_column_iterator colbegin() const noexcept
        {
            return ccolbegin();
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_column_iterator colbegin(size_type inCol) const
        {
            return ccolbegin(inCol);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_column_iterator ccolbegin() const noexcept
        {
            return const_column_iterator(array_, shape_.rows, shape_.cols);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_column_iterator ccolbegin(size_type inCol) const
        {
            if (inCol >= shape_.cols)
            {
                THROW_INVALID_ARGUMENT_ERROR("input col is greater than the number of cols in the array.");
            }
 
            return ccolbegin() += (inCol * shape_.rows);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] reverse_iterator rbegin() noexcept
        {
            return reverse_iterator(end());
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] reverse_iterator rbegin(size_type inRow)
        {
            if (inRow >= shape_.rows)
            {
                THROW_INVALID_ARGUMENT_ERROR("input row is greater than the number of rows in the array.");
            }
 
            return rbegin() += (shape_.rows - inRow - 1) * shape_.cols;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_iterator rbegin() const noexcept
        {
            return crbegin();
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_iterator rbegin(size_type inRow) const
        {
            return crbegin(inRow);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_iterator crbegin() const noexcept
        {
            return const_reverse_iterator(cend());
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_iterator crbegin(size_type inRow) const
        {
            if (inRow >= shape_.rows)
            {
                THROW_INVALID_ARGUMENT_ERROR("input row is greater than the number of rows in the array.");
            }
 
            return crbegin() += (shape_.rows - inRow - 1) * shape_.cols;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] reverse_column_iterator rcolbegin() noexcept
        {
            return reverse_column_iterator(colend());
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] reverse_column_iterator rcolbegin(size_type inCol)
        {
            if (inCol >= shape_.cols)
            {
                THROW_INVALID_ARGUMENT_ERROR("input col is greater than the number of cols in the array.");
            }
 
            return rcolbegin() += (shape_.cols - inCol - 1) * shape_.rows;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_column_iterator rcolbegin() const noexcept
        {
            return crcolbegin();
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_column_iterator rcolbegin(size_type inCol) const
        {
            return crcolbegin(inCol);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_column_iterator crcolbegin() const noexcept
        {
            return const_reverse_column_iterator(ccolend());
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_column_iterator crcolbegin(size_type inCol) const
        {
            if (inCol >= shape_.cols)
            {
                THROW_INVALID_ARGUMENT_ERROR("input col is greater than the number of cols in the array.");
            }
 
            return crcolbegin() += (shape_.cols - inCol - 1) * shape_.rows;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] iterator end() noexcept
        {
            return begin() += size_;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] iterator end(size_type inRow)
        {
            if (inRow >= shape_.rows)
            {
                THROW_INVALID_ARGUMENT_ERROR("input row is greater than the number of rows in the array.");
            }
 
            return begin(inRow) += shape_.cols;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_iterator end() const noexcept
        {
            return cend();
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_iterator end(size_type inRow) const
        {
            return cend(inRow);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_iterator cend() const noexcept
        {
            return cbegin() += size_;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_iterator cend(size_type inRow) const
        {
            if (inRow >= shape_.rows)
            {
                THROW_INVALID_ARGUMENT_ERROR("input row is greater than the number of rows in the array.");
            }
 
            return cbegin(inRow) += shape_.cols;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] reverse_iterator rend() noexcept
        {
            return rbegin() += size_;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] reverse_iterator rend(size_type inRow)
        {
            if (inRow >= shape_.rows)
            {
                THROW_INVALID_ARGUMENT_ERROR("input row is greater than the number of rows in the array.");
            }
 
            return rbegin(inRow) += shape_.cols;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_iterator rend() const noexcept
        {
            return crend();
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_iterator rend(size_type inRow) const
        {
            return crend(inRow);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_iterator crend() const noexcept
        {
            return crbegin() += size_;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_iterator crend(size_type inRow) const
        {
            if (inRow >= shape_.rows)
            {
                THROW_INVALID_ARGUMENT_ERROR("input row is greater than the number of rows in the array.");
            }
 
            return crbegin(inRow) += shape_.cols;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] column_iterator colend() noexcept
        {
            return colbegin() += size_;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] column_iterator colend(size_type inCol)
        {
            if (inCol >= shape_.cols)
            {
                THROW_INVALID_ARGUMENT_ERROR("input col is greater than the number of cols in the array.");
            }
 
            return colbegin(inCol) += shape_.rows;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_column_iterator colend() const noexcept
        {
            return ccolend();
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_column_iterator colend(size_type inCol) const
        {
            return ccolend(inCol);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_column_iterator ccolend() const noexcept
        {
            return ccolbegin() += size_;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_column_iterator ccolend(size_type inCol) const
        {
            if (inCol >= shape_.cols)
            {
                THROW_INVALID_ARGUMENT_ERROR("input col is greater than the number of cols in the array.");
            }
 
            return ccolbegin(inCol) += shape_.rows;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] reverse_column_iterator rcolend() noexcept
        {
            return rcolbegin() += size_;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] reverse_column_iterator rcolend(size_type inCol)
        {
            if (inCol >= shape_.cols)
            {
                THROW_INVALID_ARGUMENT_ERROR("input col is greater than the number of cols in the array.");
            }
 
            return rcolbegin(inCol) += shape_.rows;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_column_iterator rcolend() const noexcept
        {
            return crcolend();
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_column_iterator rcolend(size_type inCol) const
        {
            return crcolend(inCol);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_column_iterator crcolend() const noexcept
        {
            return crcolbegin() += size_;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reverse_column_iterator crcolend(size_type inCol) const
        {
            if (inCol >= shape_.cols)
            {
                THROW_INVALID_ARGUMENT_ERROR("input col is greater than the number of cols in the array.");
            }
 
            return crcolbegin(inCol) += shape_.rows;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] NdArray<bool> all(Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            const auto function = [](dtype i) -> bool { return !utils::essentiallyEqual(i, dtype{ 0 }); };
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    NdArray<bool> returnArray = { stl_algorithms::all_of(cbegin(), cend(), function) };
                    return returnArray;
                }
                case Axis::COL:
                {
                    NdArray<bool> returnArray(1, shape_.rows);
                    for (uint32 row = 0; row < shape_.rows; ++row)
                    {
                        returnArray(0, row) = stl_algorithms::all_of(cbegin(row), cend(row), function);
                    }
 
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().all(Axis::COL);
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] NdArray<bool> any(Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            const auto function = [](dtype i) -> bool { return !utils::essentiallyEqual(i, dtype{ 0 }); };
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    NdArray<bool> returnArray = { stl_algorithms::any_of(cbegin(), cend(), function) };
                    return returnArray;
                }
                case Axis::COL:
                {
                    NdArray<bool> returnArray(1, shape_.rows);
                    for (uint32 row = 0; row < shape_.rows; ++row)
                    {
                        returnArray(0, row) = stl_algorithms::any_of(cbegin(row), cend(row), function);
                    }
 
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().any(Axis::COL);
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] NdArray<size_type> argmax(Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            const auto comparitor = [](dtype lhs, dtype rhs) noexcept -> bool { return lhs < rhs; };
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    NdArray<size_type> returnArray = { static_cast<size_type>(
                        stl_algorithms::max_element(cbegin(), cend(), comparitor) - cbegin()) };
                    return returnArray;
                }
                case Axis::COL:
                {
                    NdArray<size_type> returnArray(1, shape_.rows);
                    for (size_type row = 0; row < shape_.rows; ++row)
                    {
                        returnArray(0, row) = static_cast<size_type>(
                            stl_algorithms::max_element(cbegin(row), cend(row), comparitor) - cbegin(row));
                    }
 
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().argmax(Axis::COL);
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] NdArray<size_type> argmin(Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            const auto comparitor = [](dtype lhs, dtype rhs) noexcept -> bool { return lhs < rhs; };
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    NdArray<size_type> returnArray = { static_cast<size_type>(
                        stl_algorithms::min_element(cbegin(), cend(), comparitor) - cbegin()) };
                    return returnArray;
                }
                case Axis::COL:
                {
                    NdArray<size_type> returnArray(1, shape_.rows);
                    for (size_type row = 0; row < shape_.rows; ++row)
                    {
                        returnArray(0, row) = static_cast<size_type>(
                            stl_algorithms::min_element(cbegin(row), cend(row), comparitor) - cbegin(row));
                    }
 
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().argmin(Axis::COL);
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] NdArray<size_type> argsort(Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    std::vector<size_type> idx(size_);
                    std::iota(idx.begin(), idx.end(), 0);
 
                    const auto function = [this](size_type i1, size_type i2) noexcept -> bool
                    { return (*this)[i1] < (*this)[i2]; };
 
                    stl_algorithms::stable_sort(idx.begin(), idx.end(), function);
                    return NdArray<size_type>(idx); // NOLINT(modernize-return-braced-init-list)
                }
                case Axis::COL:
                {
                    NdArray<size_type>     returnArray(shape_);
                    std::vector<size_type> idx(shape_.cols);
 
                    for (index_type row = 0; row < static_cast<index_type>(shape_.rows); ++row)
                    {
                        std::iota(idx.begin(), idx.end(), 0);
 
                        const auto function = [this, row](size_type i1, size_type i2) noexcept -> bool
                        { return operator()(row, i1) < operator()(row, i2); };
 
                        stl_algorithms::stable_sort(idx.begin(), idx.end(), function);
 
                        for (index_type col = 0; col < static_cast<index_type>(shape_.cols); ++col)
                        {
                            returnArray(row, col) = idx[static_cast<size_type>(col)];
                        }
                    }
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().argsort(Axis::COL).transpose();
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        template<typename dtypeOut,
                 typename dtype_                                       = dtype,
                 std::enable_if_t<std::is_same_v<dtype_, dtype>, int>  = 0,
                 std::enable_if_t<std::is_arithmetic_v<dtype_>, int>   = 0,
                 std::enable_if_t<std::is_arithmetic_v<dtypeOut>, int> = 0>
        [[nodiscard]] NdArray<dtypeOut> astype() const
        {
            if constexpr (std::is_same_v<dtypeOut, dtype>)
            {
                return *this;
            }
            else
            {
                NdArray<dtypeOut> outArray(shape_);
                stl_algorithms::transform(cbegin(),
                                          cend(),
                                          outArray.begin(),
                                          [](dtype value) -> dtypeOut { return static_cast<dtypeOut>(value); });
 
                return outArray;
            }
        }
 
        //============================================================================
        // Method Description:
        template<typename dtypeOut,
                 typename dtype_                                      = dtype,
                 std::enable_if_t<std::is_same_v<dtype_, dtype>, int> = 0,
                 std::enable_if_t<std::is_arithmetic_v<dtype_>, int>  = 0,
                 std::enable_if_t<is_complex_v<dtypeOut>, int>        = 0>
        [[nodiscard]] NdArray<dtypeOut> astype() const
        {
            NdArray<dtypeOut> outArray(shape_);
 
            const auto function = [](const_reference value) -> dtypeOut
            { return std::complex<typename dtypeOut::value_type>(value); };
 
            stl_algorithms::transform(cbegin(), cend(), outArray.begin(), function);
 
            return outArray;
        }
 
        //============================================================================
        // Method Description:
        template<typename dtypeOut,
                 typename dtype_                                      = dtype,
                 std::enable_if_t<std::is_same_v<dtype_, dtype>, int> = 0,
                 std::enable_if_t<is_complex_v<dtype_>, int>          = 0,
                 std::enable_if_t<is_complex_v<dtypeOut>, int>        = 0>
        [[nodiscard]] NdArray<dtypeOut> astype() const
        {
            if constexpr (std::is_same_v<dtypeOut, dtype>)
            {
                return *this;
            }
            else
            {
                const auto function = [](const_reference value) noexcept -> dtypeOut
                { return complex_cast<typename dtypeOut::value_type>(value); };
 
                NdArray<dtypeOut> outArray(shape_);
                stl_algorithms::transform(cbegin(), cend(), outArray.begin(), function);
                return outArray;
            }
        }
 
        //============================================================================
        // Method Description:
        template<typename dtypeOut,
                 typename dtype_                                       = dtype,
                 std::enable_if_t<std::is_same_v<dtype_, dtype>, int>  = 0,
                 std::enable_if_t<is_complex_v<dtype_>, int>           = 0,
                 std::enable_if_t<std::is_arithmetic_v<dtypeOut>, int> = 0>
        [[nodiscard]] NdArray<dtypeOut> astype() const
        {
            NdArray<dtypeOut> outArray(shape_);
 
            const auto function = [](const_reference value) -> dtypeOut { return static_cast<dtypeOut>(value.real()); };
 
            stl_algorithms::transform(cbegin(), cend(), outArray.begin(), function);
 
            return outArray;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reference back() const noexcept
        {
            return *(cend() - 1);
        }
 
        //============================================================================
        // Method Description:
        reference back() noexcept
        {
            return *(end() - 1);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reference back(size_type row) const
        {
            return *(cend(row) - 1);
        }
 
        //============================================================================
        // Method Description:
        reference back(size_type row)
        {
            return *(end(row) - 1);
        }
 
        //============================================================================
        // Method Description:
        self_type& byteswap() noexcept
        {
            STATIC_ASSERT_INTEGER(dtype);
 
            stl_algorithms::for_each(begin(),
                                     end(),
                                     [](dtype& value) noexcept -> void { value = endian::byteSwap(value); });
 
            switch (endianess_)
            {
                case Endian::NATIVE:
                {
                    endianess_ = endian::isLittleEndian() ? Endian::BIG : Endian::LITTLE;
                    break;
                }
                case Endian::LITTLE:
                {
                    endianess_ = Endian::BIG;
                    break;
                }
                case Endian::BIG:
                {
                    endianess_ = Endian::LITTLE;
                    break;
                }
            }
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type clip(value_type inMin, value_type inMax) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            self_type outArray(shape_);
            stl_algorithms::transform(cbegin(),
                                      cend(),
                                      outArray.begin(),
                                      [inMin, inMax](dtype value) noexcept -> dtype
                                      {
#ifdef __cpp_lib_clamp
                                          const auto comparitor = [](dtype lhs, dtype rhs) noexcept -> bool
                                          { return lhs < rhs; };
 
                                          return std::clamp(value, inMin, inMax, comparitor);
#else
                    if (value < inMin)
                    {
                        return inMin;
                    }
                    else if (value > inMax)
                    {
                        return inMax;
                    }
 
                    return value;
#endif
                                      });
 
            return outArray;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type column(size_type inColumn) const
        {
            return operator()(rSlice(), inColumn);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type columns(const NdArray<size_type>& inCols) const
        {
            auto       returnArray = self_type(shape_.rows, inCols.size());
            const auto rSlice      = returnArray.rSlice();
 
            for (size_type i = 0; i < inCols.size(); ++i)
            {
                returnArray.put(rSlice, i, column(inCols[i]));
            }
 
            return returnArray;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] NdArray<bool> contains(value_type inValue, Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    NdArray<bool> returnArray = { stl_algorithms::find(cbegin(), cend(), inValue) != cend() };
                    return returnArray;
                }
                case Axis::COL:
                {
                    NdArray<bool> returnArray(1, shape_.rows);
                    for (size_type row = 0; row < shape_.rows; ++row)
                    {
                        returnArray(0, row) = stl_algorithms::find(cbegin(row), cend(row), inValue) != cend(row);
                    }
 
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().contains(inValue, Axis::COL);
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type copy() const
        {
            return self_type(*this);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type cumprod(Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    self_type returnArray(1, size_);
                    returnArray[0] = front();
                    for (size_type i = 1; i < size_; ++i)
                    {
                        returnArray[i] = returnArray[i - 1] * array_[i];
                    }
 
                    return returnArray;
                }
                case Axis::COL:
                {
                    self_type returnArray(shape_);
                    for (uint32 row = 0; row < shape_.rows; ++row)
                    {
                        returnArray(row, 0) = operator()(row, 0);
                        for (uint32 col = 1; col < shape_.cols; ++col)
                        {
                            returnArray(row, col) = returnArray(row, col - 1) * operator()(row, col);
                        }
                    }
 
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().cumprod(Axis::COL).transpose();
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type cumsum(Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    self_type returnArray(1, size_);
                    returnArray[0] = front();
                    for (size_type i = 1; i < size_; ++i)
                    {
                        returnArray[i] = returnArray[i - 1] + array_[i];
                    }
 
                    return returnArray;
                }
                case Axis::COL:
                {
                    self_type returnArray(shape_);
                    for (uint32 row = 0; row < shape_.rows; ++row)
                    {
                        returnArray(row, 0) = operator()(row, 0);
                        for (uint32 col = 1; col < shape_.cols; ++col)
                        {
                            returnArray(row, col) = returnArray(row, col - 1) + operator()(row, col);
                        }
                    }
 
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().cumsum(Axis::COL).transpose();
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] pointer data() noexcept
        {
            return array_;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_pointer data() const noexcept
        {
            return array_;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] pointer dataRelease() noexcept
        {
            ownsPtr_ = false;
            return data();
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type diagonal(index_type inOffset = 0, Axis inAxis = Axis::ROW) const
        {
            switch (inAxis)
            {
                case Axis::COL:
                {
                    std::vector<dtype> diagnolValues;
                    size_type          col = 0;
                    for (index_type row = inOffset; row < static_cast<index_type>(shape_.rows); ++row)
                    {
                        if (row < 0)
                        {
                            ++col;
                            continue;
                        }
                        if (col >= shape_.cols)
                        {
                            break;
                        }
 
                        diagnolValues.push_back(operator()(static_cast<size_type>(row), col));
                        ++col;
                    }
 
                    return self_type(diagnolValues);
                }
                case Axis::ROW:
                {
                    return transpose().diagonal(inOffset, Axis::COL);
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] size_type dimSize(Axis inAxis) const noexcept
        {
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    return size();
                }
                case Axis::ROW:
                {
                    return numRows();
                }
                case Axis::COL:
                {
                    return numCols();
                }
                default:
                {
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type dot(const self_type& inOtherArray) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            if (shape_ == inOtherArray.shape_ && (shape_.rows == 1 || shape_.cols == 1))
            {
                dtype     dotProduct  = std::inner_product(cbegin(), cend(), inOtherArray.cbegin(), dtype{ 0 });
                self_type returnArray = { dotProduct };
                return returnArray;
            }
            if (shape_.cols == inOtherArray.shape_.rows)
            {
                // 2D array, use matrix multiplication
                self_type returnArray(shape_.rows, inOtherArray.shape_.cols);
                auto      otherArrayT = inOtherArray.transpose();
 
                for (uint32 i = 0; i < shape_.rows; ++i)
                {
                    for (uint32 j = 0; j < otherArrayT.shape_.rows; ++j)
                    {
                        returnArray(i, j) =
                            std::inner_product(otherArrayT.cbegin(j), otherArrayT.cend(j), cbegin(i), dtype{ 0 });
                    }
                }
 
                return returnArray;
            }
 
            std::string errStr = "shapes of [" + utils::num2str(shape_.rows) + ", " + utils::num2str(shape_.cols) + "]";
            errStr += " and [" + utils::num2str(inOtherArray.shape_.rows) + ", " +
                      utils::num2str(inOtherArray.shape_.cols) + "]";
            errStr += " are not consistent.";
            THROW_INVALID_ARGUMENT_ERROR(errStr);
 
            return self_type(); // get rid of compiler warning
        }
 
        //============================================================================
        // Method Description:
        void dump(const std::string& inFilename) const
        {
            std::filesystem::path f(inFilename);
            if (!f.has_extension())
            {
                f.replace_extension("bin");
            }
 
            std::ofstream ofile(f.c_str(), std::ios::binary);
            if (!ofile.good())
            {
                THROW_RUNTIME_ERROR("Unable to open the input file:\n\t" + inFilename);
            }
 
            if (array_ != nullptr)
            {
                ofile.write(reinterpret_cast<const char*>(array_), size_ * sizeof(dtype));
            }
            ofile.close();
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] Endian endianess() const noexcept
        {
            STATIC_ASSERT_ARITHMETIC(dtype);
 
            return endianess_;
        }
 
        //============================================================================
        // Method Description:
        self_type& fill(value_type inFillValue) noexcept
        {
            stl_algorithms::fill(begin(), end(), inFillValue);
            return *this;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] NdArray<size_type> flatnonzero() const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            std::vector<size_type> indices;
            size_type              idx = 0;
            for (auto value : *this)
            {
                if (!utils::essentiallyEqual(value, dtype{ 0 }))
                {
                    indices.push_back(idx);
                }
                ++idx;
            }
 
            return NdArray<size_type>(indices); // NOLINT(modernize-return-braced-init-list)
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type flatten() const
        {
            self_type outArray(1, size_);
            stl_algorithms::copy(cbegin(), cend(), outArray.begin());
            return outArray;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reference front() const noexcept
        {
            return *cbegin();
        }
 
        //============================================================================
        // Method Description:
        reference front() noexcept
        {
            return *begin();
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const_reference front(size_type row) const
        {
            return *cbegin(row);
        }
 
        //============================================================================
        // Method Description:
        reference front(size_type row)
        {
            return *begin(row);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type getByIndices(const NdArray<size_type>& inIndices) const
        {
            return operator[](inIndices);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type getByMask(const NdArray<bool>& inMask) const
        {
            return operator[](inMask);
        }
 
        //============================================================================
        // Method Description:
        // NOLINTNEXTLINE(modernize-use-nodiscard)
        bool isempty() const noexcept
        {
            return size_ == 0;
        }
 
        //============================================================================
        // Method Description:
        // NOLINTNEXTLINE(modernize-use-nodiscard)
        bool isflat() const noexcept
        {
            return !isscalar() && (shape_.rows == 1 || shape_.cols == 1);
        }
 
        //============================================================================
        // Method Description:
        // NOLINTNEXTLINE(modernize-use-nodiscard)
        bool isscalar() const noexcept
        {
            return size_ == 1;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] NdArray<bool> issorted(Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            const auto comparitor = [](dtype lhs, dtype rhs) noexcept -> bool { return lhs < rhs; };
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    return { stl_algorithms::is_sorted(cbegin(), cend(), comparitor) };
                }
                case Axis::COL:
                {
                    NdArray<bool> returnArray(1, shape_.rows);
                    for (uint32 row = 0; row < shape_.rows; ++row)
                    {
                        returnArray(0, row) = stl_algorithms::is_sorted(cbegin(row), cend(row), comparitor);
                    }
 
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().issorted(Axis::COL);
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        // NOLINTNEXTLINE(modernize-use-nodiscard)
        bool issquare() const noexcept
        {
            return shape_.issquare();
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] value_type item() const
        {
            if (!isscalar())
            {
                THROW_INVALID_ARGUMENT_ERROR("Can only convert an array of size 1 to a C++ scalar");
            }
 
            return front();
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type max(Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            const auto comparitor = [](dtype lhs, dtype rhs) noexcept -> bool { return lhs < rhs; };
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    self_type returnArray = { *stl_algorithms::max_element(cbegin(), cend(), comparitor) };
                    return returnArray;
                }
                case Axis::COL:
                {
                    self_type returnArray(1, shape_.rows);
                    for (uint32 row = 0; row < shape_.rows; ++row)
                    {
                        returnArray(0, row) = *stl_algorithms::max_element(cbegin(row), cend(row), comparitor);
                    }
 
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().max(Axis::COL);
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type min(Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            const auto comparitor = [](dtype lhs, dtype rhs) noexcept -> bool { return lhs < rhs; };
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    self_type returnArray = { *stl_algorithms::min_element(cbegin(), cend(), comparitor) };
                    return returnArray;
                }
                case Axis::COL:
                {
                    self_type returnArray(1, shape_.rows);
                    for (uint32 row = 0; row < shape_.rows; ++row)
                    {
                        returnArray(0, row) = *stl_algorithms::min_element(cbegin(row), cend(row), comparitor);
                    }
 
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().min(Axis::COL);
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type median(Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            const auto comparitor = [](dtype lhs, dtype rhs) noexcept -> bool { return lhs < rhs; };
 
            if (size_ == 0)
            {
                THROW_RUNTIME_ERROR("Median is undefined for an array of size = 0.");
            }
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    self_type copyArray(*this);
 
                    const size_type middleIdx = size_ / 2; // integer division
                    stl_algorithms::nth_element(copyArray.begin(),
                                                copyArray.begin() + middleIdx,
                                                copyArray.end(),
                                                comparitor);
 
                    dtype medianValue = copyArray.array_[middleIdx];
                    if (size_ % 2 == 0)
                    {
                        const size_type lhsIndex = middleIdx - 1;
                        stl_algorithms::nth_element(copyArray.begin(),
                                                    copyArray.begin() + lhsIndex,
                                                    copyArray.end(),
                                                    comparitor);
                        medianValue =
                            (medianValue + copyArray.array_[lhsIndex]) / dtype{ 2 }; // potentially integer division, ok
                    }
 
                    return { medianValue };
                }
                case Axis::COL:
                {
                    self_type copyArray(*this);
                    self_type returnArray(1, shape_.rows);
 
                    const bool isEven = shape_.cols % 2 == 0;
                    for (uint32 row = 0; row < shape_.rows; ++row)
                    {
                        const uint32 middleIdx = shape_.cols / 2; // integer division
                        stl_algorithms::nth_element(copyArray.begin(row),
                                                    copyArray.begin(row) + middleIdx,
                                                    copyArray.end(row),
                                                    comparitor);
 
                        dtype medianValue = copyArray(row, middleIdx);
                        if (isEven)
                        {
                            const size_type lhsIndex = middleIdx - 1;
                            stl_algorithms::nth_element(copyArray.begin(row),
                                                        copyArray.begin(row) + lhsIndex,
                                                        copyArray.end(row),
                                                        comparitor);
                            medianValue = (medianValue + copyArray(row, lhsIndex)) /
                                          dtype{ 2 }; // potentially integer division, ok
                        }
 
                        returnArray(0, row) = medianValue;
                    }
 
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().median(Axis::COL);
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        self_type& nans() noexcept
 
        {
            STATIC_ASSERT_FLOAT(dtype);
 
            fill(constants::nan);
            return *this;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] uint64 nbytes() const noexcept
        {
            return static_cast<uint64>(sizeof(dtype) * size_);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type newbyteorder(Endian inEndianess) const
        {
            STATIC_ASSERT_INTEGER(dtype);
 
            const bool nativeIsLittle = endian::isLittleEndian();
 
            switch (endianess_)
            {
                case Endian::NATIVE:
                {
                    switch (inEndianess)
                    {
                        case Endian::NATIVE:
                        {
                            return NdArray(*this);
                        }
                        case Endian::BIG:
                        {
                            if (nativeIsLittle)
                            {
                                self_type outArray(shape_);
 
                                stl_algorithms::transform(cbegin(), end(), outArray.begin(), endian::byteSwap<dtype>);
 
                                outArray.endianess_ = Endian::BIG;
                                return outArray;
                            }
                            else
                            {
                                auto outArray       = NdArray(*this);
                                outArray.endianess_ = Endian::BIG;
                                return outArray;
                            }
                        }
                        case Endian::LITTLE:
                        {
                            if (nativeIsLittle)
                            {
                                auto outArray       = NdArray(*this);
                                outArray.endianess_ = Endian::LITTLE;
                                return outArray;
                            }
                            else
                            {
                                self_type outArray(shape_);
 
                                stl_algorithms::transform(cbegin(), end(), outArray.begin(), endian::byteSwap<dtype>);
 
                                outArray.endianess_ = Endian::LITTLE;
                                return outArray;
                            }
                        }
                        default:
                        {
                            THROW_INVALID_ARGUMENT_ERROR("Unimplemented endian type.");
                            return {}; // get rid of compiler warning
                        }
                    }
                    break;
                }
                case Endian::BIG:
                {
                    switch (inEndianess)
                    {
                        case Endian::NATIVE:
                        {
                            if (nativeIsLittle)
                            {
                                self_type outArray(shape_);
 
                                stl_algorithms::transform(cbegin(), end(), outArray.begin(), endian::byteSwap<dtype>);
 
                                outArray.endianess_ = Endian::NATIVE;
                                return outArray;
                            }
                            else
                            {
                                auto outArray       = NdArray(*this);
                                outArray.endianess_ = Endian::NATIVE;
                                return outArray;
                            }
                        }
                        case Endian::BIG:
                        {
                            return NdArray(*this);
                        }
                        case Endian::LITTLE:
                        {
                            self_type outArray(shape_);
 
                            stl_algorithms::transform(cbegin(), end(), outArray.begin(), endian::byteSwap<dtype>);
 
                            outArray.endianess_ = Endian::LITTLE;
                            return outArray;
                        }
                        default:
                        {
                            THROW_INVALID_ARGUMENT_ERROR("Unimplemented endian type.");
                            return {}; // get rid of compiler warning
                        }
                    }
                    break;
                }
                case Endian::LITTLE:
                {
                    switch (inEndianess)
                    {
                        case Endian::NATIVE:
                        {
                            if (nativeIsLittle)
                            {
                                auto outArray       = NdArray(*this);
                                outArray.endianess_ = Endian::NATIVE;
                                return outArray;
                            }
                            else
                            {
                                self_type outArray(shape_);
 
                                stl_algorithms::transform(cbegin(), end(), outArray.begin(), endian::byteSwap<dtype>);
 
                                outArray.endianess_ = Endian::NATIVE;
                                return outArray;
                            }
                        }
                        case Endian::BIG:
                        {
                            self_type outArray(shape_);
 
                            stl_algorithms::transform(cbegin(), end(), outArray.begin(), endian::byteSwap<dtype>);
 
                            outArray.endianess_ = Endian::BIG;
                            return outArray;
                        }
                        case Endian::LITTLE:
                        {
                            return NdArray(*this);
                        }
                        default:
                        {
                            THROW_INVALID_ARGUMENT_ERROR("Unimplemented endian type.");
                            return {}; // get rid of compiler warning
                        }
                    }
                    break;
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented endian type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] NdArray<bool> none(Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            const auto function = [](dtype i) -> bool { return !utils::essentiallyEqual(i, dtype{ 0 }); };
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    NdArray<bool> returnArray = { stl_algorithms::none_of(cbegin(), cend(), function) };
                    return returnArray;
                }
                case Axis::COL:
                {
                    NdArray<bool> returnArray(1, shape_.rows);
                    for (uint32 row = 0; row < shape_.rows; ++row)
                    {
                        returnArray(0, row) = stl_algorithms::none_of(cbegin(row), cend(row), function);
                    }
 
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().none(Axis::COL);
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] std::pair<NdArray<size_type>, NdArray<size_type>> nonzero() const;
 
        //============================================================================
        // Method Description:
        [[nodiscard]] size_type numCols() const noexcept
        {
            return shape_.cols;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] size_type numRows() const noexcept
        {
            return shape_.rows;
        }
 
        //============================================================================
        // Method Description:
        self_type& ones() noexcept
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            fill(dtype{ 1 });
            return *this;
        }
 
        //============================================================================
        // Method Description:
        bool ownsInternalData() noexcept
        {
            return ownsPtr_;
        }
 
        //============================================================================
        // Method Description:
        self_type& partition(size_type inKth, Axis inAxis = Axis::NONE)
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            const auto comparitor = [](dtype lhs, dtype rhs) noexcept -> bool
            { return lhs < rhs; }; // cppcheck-suppress returnTempReference
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    if (inKth >= size_)
                    {
                        std::string errStr = "kth(=" + utils::num2str(inKth);
                        errStr += ") out of bounds (" + utils::num2str(size_) + ")";
                        THROW_INVALID_ARGUMENT_ERROR(errStr);
                    }
 
                    stl_algorithms::nth_element(begin(), begin() + inKth, end(), comparitor);
                    break;
                }
                case Axis::COL:
                {
                    if (inKth >= shape_.cols)
                    {
                        std::string errStr = "kth(=" + utils::num2str(inKth);
                        errStr += ") out of bounds (" + utils::num2str(shape_.cols) + ")";
                        THROW_INVALID_ARGUMENT_ERROR(errStr);
                    }
 
                    for (uint32 row = 0; row < shape_.rows; ++row)
                    {
                        stl_algorithms::nth_element(begin(row), begin(row) + inKth, end(row), comparitor);
                    }
                    break;
                }
                case Axis::ROW:
                {
                    if (inKth >= shape_.rows)
                    {
                        std::string errStr = "kth(=" + utils::num2str(inKth);
                        errStr += ") out of bounds (" + utils::num2str(shape_.rows) + ")";
                        THROW_INVALID_ARGUMENT_ERROR(errStr);
                    }
 
                    self_type transposedArray = transpose();
                    for (uint32 row = 0; row < transposedArray.shape_.rows; ++row)
                    {
                        stl_algorithms::nth_element(transposedArray.begin(row),
                                                    transposedArray.begin(row) + inKth,
                                                    transposedArray.end(row),
                                                    comparitor);
                    }
                    *this = transposedArray.transpose();
                    break;
                }
            }
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        void print() const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            std::cout << *this;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type prod(Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    dtype     product     = std::accumulate(cbegin(), cend(), dtype{ 1 }, std::multiplies<dtype>());
                    self_type returnArray = { product };
                    return returnArray;
                }
                case Axis::COL:
                {
                    self_type returnArray(1, shape_.rows);
                    for (uint32 row = 0; row < shape_.rows; ++row)
                    {
                        returnArray(0, row) =
                            std::accumulate(cbegin(row), cend(row), dtype{ 1 }, std::multiplies<dtype>());
                    }
 
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().prod(Axis::COL);
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type ptp(Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            const auto comparitor = [](dtype lhs, dtype rhs) noexcept -> bool { return lhs < rhs; };
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    const auto result      = stl_algorithms::minmax_element(cbegin(), cend(), comparitor);
                    self_type  returnArray = { *result.second - *result.first };
                    return returnArray;
                }
                case Axis::COL:
                {
                    self_type returnArray(1, shape_.rows);
                    for (uint32 row = 0; row < shape_.rows; ++row)
                    {
                        const auto result   = stl_algorithms::minmax_element(cbegin(row), cend(row), comparitor);
                        returnArray(0, row) = *result.second - *result.first;
                    }
 
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().ptp(Axis::COL);
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        self_type& put(index_type inIndex, const value_type& inValue)
        {
            at(inIndex) = inValue;
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        self_type& put(index_type inRow, index_type inCol, const value_type& inValue)
        {
            at(inRow, inCol) = inValue;
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        self_type& put(const Indices& inIndices, const value_type& inValue)
        {
            for (auto index : inIndices)
            {
                put(index, inValue);
            }
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        self_type& put(const Indices& inIndices, const self_type& inValues)
        {
            if (inValues.isscalar())
            {
                return put(inIndices, inValues.item());
            }
            else if (inIndices.size() != inValues.size())
            {
                THROW_INVALID_ARGUMENT_ERROR("Input indices do not match values dimensions.");
            }
 
            size_type counter = 0;
            for (auto index : inIndices)
            {
                put(index, inValues[counter++]);
            }
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        self_type& put(const Slice& inSlice, const value_type& inValue)
        {
            return put(toIndices(inSlice, Axis::NONE), inValue);
        }
 
        //============================================================================
        // Method Description:
        self_type& put(const Slice& inSlice, const self_type& inValues)
        {
            return put(toIndices(inSlice, Axis::NONE), inValues);
        }
 
        //============================================================================
        // Method Description:
        template<typename RowIndices,
                 typename ColIndices,
                 type_traits::ndarray_int_concept<RowIndices> = 0,
                 type_traits::ndarray_int_concept<ColIndices> = 0>
        self_type& put(const RowIndices& inRowIndices, const ColIndices& inColIndices, const value_type& inValue)
        {
            stl_algorithms::for_each(inRowIndices.begin(),
                                     inRowIndices.end(),
                                     [this, &inColIndices, &inValue](const auto row)
                                     {
                                         stl_algorithms::for_each(inColIndices.begin(),
                                                                  inColIndices.end(),
                                                                  [this, row, &inValue](const auto col)
                                                                  { this->put(row, col, inValue); });
                                     });
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        template<typename RowIndices, type_traits::ndarray_int_concept<RowIndices> = 0>
        self_type& put(const RowIndices& inRowIndices, const Slice& inColSlice, const value_type& inValue)
        {
            return put(inRowIndices, toIndices(inColSlice, Axis::COL), inValue);
        }
 
        //============================================================================
        // Method Description:
        template<typename ColIndices, type_traits::ndarray_int_concept<ColIndices> = 0>
        self_type& put(const Slice& inRowSlice, const ColIndices& inColIndices, const value_type& inValue)
        {
            return put(toIndices(inRowSlice, Axis::ROW), inColIndices, inValue);
        }
 
        //============================================================================
        // Method Description:
        self_type& put(const Slice& inRowSlice, const Slice& inColSlice, const value_type& inValue)
        {
            return put(toIndices(inRowSlice, Axis::ROW), toIndices(inColSlice, Axis::COL), inValue);
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        self_type& put(const Indices& inRowIndices, index_type inColIndex, const value_type& inValue)
        {
            const NdArray<index_type> colIndices = { inColIndex };
            return put(inRowIndices, colIndices, inValue);
        }
 
        //============================================================================
        // Method Description:
        self_type& put(const Slice& inRowSlice, index_type inColIndex, const value_type& inValue)
        {
            const NdArray<index_type> colIndices = { inColIndex };
            return put(toIndices(inRowSlice, Axis::ROW), colIndices, inValue);
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        self_type& put(index_type inRowIndex, const Indices& inColIndices, const value_type& inValue)
        {
            const NdArray<index_type> rowIndices = { inRowIndex };
            return put(rowIndices, inColIndices, inValue);
        }
 
        //============================================================================
        // Method Description:
        self_type& put(index_type inRowIndex, const Slice& inColSlice, const value_type& inValue)
        {
            const NdArray<index_type> rowIndices = { inRowIndex };
            return put(rowIndices, toIndices(inColSlice, Axis::COL), inValue);
        }
 
        //============================================================================
        // Method Description:
        template<typename RowIndices,
                 typename ColIndices,
                 type_traits::ndarray_int_concept<RowIndices> = 0,
                 type_traits::ndarray_int_concept<ColIndices> = 0>
        self_type& put(const RowIndices& inRowIndices, const ColIndices& inColIndices, const self_type& inValues)
        {
            std::vector<size_type> indices;
            indices.reserve(inRowIndices.size() * inColIndices.size());
            std::for_each(inRowIndices.begin(),
                          inRowIndices.end(),
                          [this, &inColIndices, &indices](auto row)
                          {
                              if constexpr (std::is_signed_v<decltype(row)>)
                              {
                                  if (row < 0)
                                  {
                                      row += shape_.rows;
                                  }
                                  // still
                                  if (row < 0)
                                  {
                                      THROW_INVALID_ARGUMENT_ERROR("row index exceeds matrix dimensions");
                                  }
                              }
                              std::for_each(inColIndices.begin(),
                                            inColIndices.end(),
                                            [this, row, &indices](auto col)
                                            {
                                                if constexpr (std::is_signed_v<decltype(col)>)
                                                {
                                                    if (col < 0)
                                                    {
                                                        col += shape_.cols;
                                                    }
                                                    // still
                                                    if (col < 0)
                                                    {
                                                        THROW_INVALID_ARGUMENT_ERROR(
                                                            "col index exceeds matrix dimensions");
                                                    }
                                                }
                                                indices.push_back(row * shape_.cols + col);
                                            });
                          });
 
            return put(NdArray<size_type>(indices.data(), indices.size(), PointerPolicy::SHELL), inValues);
        }
 
        //============================================================================
        // Method Description:
        template<typename RowIndices, type_traits::ndarray_int_concept<RowIndices> = 0>
        self_type& put(const RowIndices& inRowIndices, Slice inColSlice, const self_type& inValues)
        {
            return put(inRowIndices, toIndices(inColSlice, Axis::COL), inValues);
        }
 
        //============================================================================
        // Method Description:
        template<typename ColIndices, type_traits::ndarray_int_concept<ColIndices> = 0>
        self_type& put(Slice inRowSlice, const ColIndices& inColIndices, const self_type& inValues)
        {
            return put(toIndices(inRowSlice, Axis::ROW), inColIndices, inValues);
        }
 
        //============================================================================
        // Method Description:
        self_type& put(Slice inRowSlice, Slice inColSlice, const self_type& inValues)
        {
            return put(toIndices(inRowSlice, Axis::ROW), toIndices(inColSlice, Axis::COL), inValues);
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        self_type& put(const Indices& inRowIndices, index_type inColIndex, const self_type& inValues)
        {
            const NdArray<index_type> colIndices = { inColIndex };
            return put(inRowIndices, colIndices, inValues);
        }
 
        //============================================================================
        // Method Description:
        self_type& put(const Slice& inRowSlice, index_type inColIndex, const self_type& inValues)
        {
            const NdArray<index_type> colIndices = { inColIndex };
            return put(toIndices(inRowSlice, Axis::ROW), colIndices, inValues);
        }
 
        //============================================================================
        // Method Description:
        template<typename Indices, type_traits::ndarray_int_concept<Indices> = 0>
        self_type& put(index_type inRowIndex, const Indices& inColIndices, const self_type& inValues)
        {
            const NdArray<index_type> rowIndices = { inRowIndex };
            return put(rowIndices, inColIndices, inValues);
        }
 
        //============================================================================
        // Method Description:
        self_type& put(index_type inRowIndex, const Slice& inColSlice, const self_type& inValues)
        {
            const NdArray<index_type> rowIndices = { inRowIndex };
            return put(rowIndices, toIndices(inColSlice, Axis::COL), inValues);
        }
 
        //============================================================================
        // Method Description:
        self_type& putMask(const NdArray<bool>& inMask, const value_type& inValue)
        {
            if (inMask.shape() != shape_)
            {
                THROW_INVALID_ARGUMENT_ERROR("input inMask must be the same shape as the array it is masking.");
            }
 
            return put(inMask.flatnonzero(), inValue);
        }
 
        //============================================================================
        // Method Description:
        self_type& putMask(const NdArray<bool>& inMask, const self_type& inValues)
        {
            if (inMask.shape() != shape_)
            {
                THROW_INVALID_ARGUMENT_ERROR("input inMask must be the same shape as the array it is masking.");
            }
 
            if (inValues.isscalar())
            {
                put(inMask.flatnonzero(), inValues.item());
            }
            else
            {
                put(inMask.flatnonzero(), inValues);
            }
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        self_type& ravel()
        {
            reshape(size_);
            return *this;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type repeat(size_type inNumRows, size_type inNumCols) const
        {
            self_type returnArray(shape_.rows * inNumRows, shape_.cols * inNumCols);
 
            for (size_type row = 0; row < inNumRows; ++row)
            {
                for (size_type col = 0; col < inNumCols; ++col)
                {
                    std::vector<size_type> indices(shape_.size());
 
                    const size_type rowStart = row * shape_.rows;
                    const size_type colStart = col * shape_.cols;
 
                    const size_type rowEnd = (row + 1) * shape_.rows;
                    const size_type colEnd = (col + 1) * shape_.cols;
 
                    size_type counter = 0;
                    for (size_type rowIdx = rowStart; rowIdx < rowEnd; ++rowIdx)
                    {
                        for (size_type colIdx = colStart; colIdx < colEnd; ++colIdx)
                        {
                            indices[counter++] = rowIdx * returnArray.shape_.cols + colIdx;
                        }
                    }
 
                    returnArray.put(NdArray<size_type>(indices), *this);
                }
            }
 
            return returnArray;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type repeat(const Shape& inRepeatShape) const
        {
            return repeat(inRepeatShape.rows, inRepeatShape.cols);
        }
 
        //============================================================================
        // Method Description:
        self_type& replace(value_type oldValue, value_type newValue)
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            stl_algorithms::replace(begin(), end(), oldValue, newValue);
            return *this;
        }
 
        //============================================================================
        // Method Description:
        self_type& reshape(size_type inSize)
        {
            if (inSize != size_)
            {
                std::string errStr = "Cannot reshape array of size " + utils::num2str(size_) + " into shape ";
                errStr += "[" + utils::num2str(1) + ", " + utils::num2str(inSize) + "]";
                THROW_RUNTIME_ERROR(errStr);
            }
 
            shape_.rows = 1;
            shape_.cols = inSize;
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        self_type& reshape(index_type inNumRows, index_type inNumCols)
        {
            if (inNumRows < 0)
            {
                if (size_ % inNumCols == 0)
                {
                    return reshape(size_ / inNumCols, inNumCols);
                }
 
                std::string errStr = "Cannot reshape array of size " + utils::num2str(size_) + " into a shape ";
                errStr += "with " + utils::num2str(inNumCols) + " columns";
                THROW_INVALID_ARGUMENT_ERROR(errStr);
            }
 
            if (inNumCols < 0)
            {
                if (size_ % inNumRows == 0)
                {
                    return reshape(inNumRows, size_ / inNumRows);
                }
 
                std::string errStr = "Cannot reshape array of size " + utils::num2str(size_) + " into a shape ";
                errStr += "with " + utils::num2str(inNumRows) + " rows";
                THROW_INVALID_ARGUMENT_ERROR(errStr);
            }
 
            if (static_cast<size_type>(inNumRows * inNumCols) != size_)
            {
                std::string errStr = "Cannot reshape array of size " + utils::num2str(size_) + " into shape ";
                errStr += "[" + utils::num2str(inNumRows) + ", " + utils::num2str(inNumCols) + "]";
                THROW_INVALID_ARGUMENT_ERROR(errStr);
            }
 
            shape_.rows = static_cast<size_type>(inNumRows);
            shape_.cols = static_cast<size_type>(inNumCols);
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        self_type& reshape(const Shape& inShape)
        {
            return reshape(inShape.rows, inShape.cols);
        }
 
        //============================================================================
        // Method Description:
        self_type& resizeFast(size_type inNumRows, size_type inNumCols)
        {
            newArray(Shape(inNumRows, inNumCols));
            return *this;
        }
 
        //============================================================================
        // Method Description:
        self_type& resizeFast(const Shape& inShape)
        {
            return resizeFast(inShape.rows, inShape.cols);
        }
 
        //============================================================================
        // Method Description:
        self_type& resizeSlow(size_type inNumRows, size_type inNumCols)
        {
            std::vector<dtype> oldData(size_);
            stl_algorithms::copy(begin(), end(), oldData.begin());
 
            const Shape inShape(inNumRows, inNumCols);
            const Shape oldShape = shape_;
 
            newArray(inShape);
 
            for (uint32 row = 0; row < inShape.rows; ++row)
            {
                for (uint32 col = 0; col < inShape.cols; ++col)
                {
                    if (row >= oldShape.rows || col >= oldShape.cols)
                    {
                        operator()(row, col) = dtype{ 0 }; // zero fill
                    }
                    else
                    {
                        operator()(row, col) = oldData[row * oldShape.cols + col];
                    }
                }
            }
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        self_type& resizeSlow(const Shape& inShape)
        {
            return resizeSlow(inShape.rows, inShape.cols);
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type round(uint8 inNumDecimals = 0) const
        {
            STATIC_ASSERT_FLOAT(dtype);
 
            self_type    returnArray(shape_);
            const double multFactor = utils::power(10., inNumDecimals);
            const auto   function   = [multFactor](dtype value) noexcept -> dtype
            { return static_cast<dtype>(std::nearbyint(static_cast<double>(value) * multFactor) / multFactor); };
 
            stl_algorithms::transform(cbegin(), cend(), returnArray.begin(), function);
 
            return returnArray;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type row(size_type inRow) const
        {
            return self_type(cbegin(inRow), cend(inRow));
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type rows(const NdArray<size_type>& inRows) const
        {
            auto       returnArray = self_type(inRows.size(), shape_.cols);
            const auto cSlice      = returnArray.cSlice();
 
            for (size_type i = 0; i < inRows.size(); ++i)
            {
                returnArray.put(i, cSlice, row(inRows[i]));
            }
 
            return returnArray;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] const Shape& shape() const noexcept
        {
            return shape_;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] size_type size() const noexcept
        {
            return size_;
        }
 
        //============================================================================
        // Method Description:
        self_type& sort(Axis inAxis = Axis::NONE)
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            const auto comparitor = [](dtype lhs, dtype rhs) noexcept -> bool
            { return lhs < rhs; }; // cppcheck-suppress returnTempReference
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    stl_algorithms::sort(begin(), end(), comparitor);
                    break;
                }
                case Axis::COL:
                {
                    for (uint32 row = 0; row < shape_.rows; ++row)
                    {
                        stl_algorithms::sort(begin(row), end(row), comparitor);
                    }
                    break;
                }
                case Axis::ROW:
                {
                    self_type transposedArray = transpose();
                    for (uint32 row = 0; row < transposedArray.shape_.rows; ++row)
                    {
                        stl_algorithms::sort(transposedArray.begin(row), transposedArray.end(row), comparitor);
                    }
 
                    *this = transposedArray.transpose();
                    break;
                }
            }
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] std::string str() const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            std::string out;
            out += "[";
            for (uint32 row = 0; row < shape_.rows; ++row)
            {
                out += "[";
                for (uint32 col = 0; col < shape_.cols; ++col)
                {
                    out += utils::value2str(operator()(row, col)) + ", ";
                }
 
                if (row == shape_.rows - 1)
                {
                    out += "]";
                }
                else
                {
                    out += "]\n";
                }
            }
            out += "]\n";
            return out;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type sum(Axis inAxis = Axis::NONE) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    self_type returnArray = { std::accumulate(cbegin(), cend(), dtype{ 0 }) };
                    return returnArray;
                }
                case Axis::COL:
                {
                    self_type returnArray(1, shape_.rows);
                    for (uint32 row = 0; row < shape_.rows; ++row)
                    {
                        returnArray(0, row) = std::accumulate(cbegin(row), cend(row), dtype{ 0 });
                    }
 
                    return returnArray;
                }
                case Axis::ROW:
                {
                    return transpose().sum(Axis::COL);
                }
                default:
                {
                    THROW_INVALID_ARGUMENT_ERROR("Unimplemented axis type.");
                    return {}; // get rid of compiler warning
                }
            }
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type swapaxes() const
        {
            return transpose();
        }
 
        //============================================================================
        // Method Description:
        self_type& swapCols(index_type colIdx1, index_type colIdx2) noexcept
        {
            for (index_type row = 0; row < static_cast<index_type>(shape_.rows); ++row)
            {
                std::swap(operator()(row, colIdx1), operator()(row, colIdx2));
            }
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        self_type& swapRows(index_type rowIdx1, index_type rowIdx2) noexcept
        {
            for (index_type col = 0; col < static_cast<index_type>(shape_.cols); ++col)
            {
                std::swap(operator()(rowIdx1, col), operator()(rowIdx2, col));
            }
 
            return *this;
        }
 
        //============================================================================
        // Method Description:
        void tofile(const std::string& inFilename) const
        {
            dump(inFilename);
        }
 
        //============================================================================
        // Method Description:
        void tofile(const std::string& inFilename, const char inSep) const
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            std::filesystem::path f(inFilename);
            if (!f.has_extension())
            {
                f.replace_extension("txt");
            }
 
            std::ofstream ofile(f.c_str());
            if (!ofile.good())
            {
                THROW_RUNTIME_ERROR("Input file could not be opened:\n\t" + inFilename);
            }
 
            size_type counter = 0;
            for (auto value : *this)
            {
                ofile << value;
                if (counter++ != size_ - 1)
                {
                    ofile << inSep;
                }
            }
            ofile.close();
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] NdArray<size_type> toIndices(Slice inSlice, Axis inAxis = Axis::NONE) const
        {
            size_type numElements = 0;
            switch (inAxis)
            {
                case Axis::NONE:
                {
                    numElements = inSlice.numElements(size_);
                    break;
                }
                case Axis::ROW:
                {
                    numElements = inSlice.numElements(shape_.rows);
                    break;
                }
                case Axis::COL:
                {
                    numElements = inSlice.numElements(shape_.cols);
                    break;
                }
                default:
                {
                    // not actually possible, getting rid of compiler warning
                    THROW_INVALID_ARGUMENT_ERROR("Invalid 'inAxis' option");
                }
            }
 
            if (numElements == 0)
            {
                return {};
            }
 
            NdArray<size_type> indices(1, numElements);
            indices[0] = static_cast<size_type>(inSlice.start);
            for (size_type i = 1; i < indices.size(); ++i)
            {
                indices[static_cast<index_type>(i)] = static_cast<size_type>(
                    indices[static_cast<index_type>(i - size_type{ 1 })] + static_cast<size_type>(inSlice.step));
            }
 
            return indices;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] std::vector<dtype> toStlVector() const
        {
            return std::vector<dtype>(cbegin(), cend());
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] value_type trace(size_type inOffset = 0, Axis inAxis = Axis::ROW) const noexcept
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            size_type rowStart = 0;
            size_type colStart = 0;
            switch (inAxis)
            {
                case Axis::ROW:
                {
                    rowStart += inOffset;
                    break;
                }
                case Axis::COL:
                {
                    colStart += inOffset;
                    break;
                }
                default:
                {
                    // if the user input NONE, override back to ROW
                    inAxis = Axis::ROW;
                    break;
                }
            }
 
            if (rowStart >= shape_.rows || colStart >= shape_.cols)
            {
                return dtype{ 0 };
            }
 
            size_type col = colStart;
            dtype     sum = 0;
            for (size_type row = rowStart; row < shape_.rows; ++row)
            {
                if (col >= shape_.cols)
                {
                    break;
                }
                sum += operator()(row, col++);
            }
 
            return sum;
        }
 
        //============================================================================
        // Method Description:
        [[nodiscard]] self_type transpose() const
        {
            self_type transArray(shape_.cols, shape_.rows);
            for (uint32 row = 0; row < shape_.rows; ++row)
            {
                for (uint32 col = 0; col < shape_.cols; ++col)
                {
                    transArray(col, row) = operator()(row, col);
                }
            }
            return transArray;
        }
 
        //============================================================================
        // Method Description:
        self_type& zeros() noexcept
        {
            STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
            fill(dtype{ 0 });
            return *this;
        }
 
    private:
        //====================================Attributes==============================
        allocator_type allocator_{};
        Shape          shape_{ 0, 0 };
        size_type      size_{ 0 };
        Endian         endianess_{ Endian::NATIVE };
        pointer        array_{ nullptr };
        bool           ownsPtr_{ false };
 
        //============================================================================
        // Method Description:
        void deleteArray() noexcept
        {
            if (ownsPtr_ && array_ != nullptr)
            {
                allocator_.deallocate(array_, size_);
            }
 
            array_      = nullptr;
            shape_.rows = shape_.cols = 0;
            size_                     = 0;
            ownsPtr_                  = false;
            endianess_                = Endian::NATIVE;
        }
 
        //============================================================================
        // Method Description:
        void newArray()
        {
            if (size_ > 0)
            {
                array_   = allocator_.allocate(size_);
                ownsPtr_ = true;
            }
        }
 
        //============================================================================
        // Method Description:
        void newArray(const Shape& inShape)
        {
            deleteArray();
 
            shape_ = inShape;
            size_  = inShape.size();
            newArray();
        }
    };
 
    // NOTE: this needs to be defined outside of the class to get rid of a compiler
    // error in Visual Studio
    template<typename dtype, class Alloc_>
    [[nodiscard]] std::pair<NdArray<uint32>, NdArray<uint32>> NdArray<dtype, Alloc_>::nonzero() const
    {
        STATIC_ASSERT_ARITHMETIC_OR_COMPLEX(dtype);
 
        std::vector<size_type> rowIndices;
        std::vector<size_type> colIndices;
 
        for (uint32 row = 0; row < shape_.rows; ++row)
        {
            for (uint32 col = 0; col < shape_.cols; ++col)
            {
                if (!utils::essentiallyEqual(operator()(row, col), dtype{ 0 }))
                {
                    rowIndices.push_back(row);
                    colIndices.push_back(col);
                }
            }
        }
 
        return std::make_pair(NdArray<size_type>(rowIndices), NdArray<size_type>(colIndices));
    }
} // namespace nc