如何从2d Boost.MultiArray获取子阵列?

How to take subarray from 2d Boost.MultiArray?

本文关键字：获取 阵列 MultiArray Boost 2d      更新时间：2023-10-16

我正在开发一个需要使用2d Boost.MultiArray的程序。我设法初始化它并用数据填充它。但是我不明白如何取大小为i的子阵列，j如果多数组的大小为m，n。i<=m和j<=n的地方.谁能帮我？
法典：

matrix_type matrix(boost::extents[width][height]);
read_matrix_from_file(file_content, matrix);
for (int rank = 1; rank < workers; rank++) {
auto subarray_size = (rest > 0) ? lines_per_worker + 1 : lines_per_worker;
rest--;
typedef boost::multi_array_types::index_range range;

size_t finish_line = subarray_size + bias - 1;
finish_line = (finish_line==bias)? finish_line+1:finish_line;

matrix_type::array_view<2>::type
current_process_batch = matrix[boost::indices[range(bias, subarray_size + bias - 1)][range(0, width)]];
}

文档 https://www.boost.org/doc/libs/1_73_0/libs/multi_array/doc/user.html#sec_generators

Boost.MultiArray 提供了用于创建 已存在的阵列组件。它允许您创建子视图 保留与原始数组相同的维数，或 一个尺寸也比原版小。

子视图 通过调用运算符 []，向其传递 index_gen类型。index_gen由传递index_range填充 对象到其运算符 []。index_range和index_gen类型为 在 multi_array_types 命名空间中定义并作为 的嵌套成员 每种数组类型。与 boost：：extent 类似，默认情况下是库 构造对象 boost：：index。您可以通过以下方式抑制此对象 在包含库之前定义BOOST_MULTI_ARRAY_NO_GENERATORS 页眉。下面是一个简单的子视图创建示例。

样本：

住在科里鲁 

#include <boost/multi_array.hpp>
#include <iostream>
template <typename Ma>
auto dump(Ma const& r) -> std::enable_if_t<1 == Ma::dimensionality> {
for (auto c: r)
std::cout << " " << c;
std::cout << "n";
}
template <typename Ma>
auto dump(Ma const& ma) -> std::enable_if_t<2 == Ma::dimensionality> {
for (auto const& r: ma)
dump(r);
}
int main() {using range = boost::multi_array_types::index_range;
using boost::extents;
using boost::indices;
auto mn = extents[7][4];
boost::multi_array<int, 2> ma(mn);
// fill with numbers from 10..37
std::iota(ma.data(), ma.data() + ma.num_elements(), 10);
dump(ma);
std::cout << "Slice [3..5][1..3]n";
dump(ma[ indices[range(3,5)][range(1,3)] ]);
std::cout << "Slice [2..4][2..3]n";
dump(ma[ indices[range(2,4)][range(2,3)] ]);
std::cout << "Slice [1,3,5][2..3]n";
dump(ma[ indices[range(1,7,2)][range(2,3)] ]);
std::cout << "Degenerate views (reduced dimensionality):n";
std::cout << "Slice [1,3,5][2]n";
dump(ma[ indices[range(1,7,2)][2] ]);
std::cout << "Slice [0][1,3]n";
dump(ma[ indices[0][range(1,4,2)] ]);
}

指纹

10 11 12 13
14 15 16 17
18 19 20 21
22 23 24 25
26 27 28 29
30 31 32 33
34 35 36 37
Slice [3..5][1..3]
23 24
27 28
Slice [2..4][2..3]
20
24
Slice [1,3,5][2..3]
16
24
32
Degenerate views (reduced dimensionality):
Slice [1,3,5][2]
16 24 32
Slice [0][1,3]
11 13

更新

评论：

如果要避免退化维度，方法如下：

住在科里鲁

std::cout << "Degenerate views (reduced dimensionality):n";
std::cout << "Slice [1,3,5][2]n";
dump(ma[ indices[range(1,7,2)][2] ]);
std::cout << "Slice [0][1,3]n";
dump(ma[ indices[0][range(1,4,2)] ]);
// RE: Comment
std::cout << "NON-Degenerate views (reduced dimensionality):n";
std::cout << "Slice [1,3,5][2]n";
dump(ma[ indices[range(1,7,2)][range(2,3)] ]);
std::cout << "Slice [0][1,3]n";
dump(ma[ indices[range(0,1)][range(1,4,2)] ]);

指纹

Degenerate views (reduced dimensionality):
Slice [1,3,5][2]
16 24 32
Slice [0][1,3]
11 13
NON-Degenerate views (reduced dimensionality):
Slice [1,3,5][2]
16
24
32
Slice [0][1,3]
11 13

基本上，不要使用文字[i][j]而是[range(i,i+1)][range(j,j+1)]

相关文章：