C++讀取保存為二進制的 numpy 數組

首先需要確定C++和Python中變量對應的精度類型，
https://docs.scipy.org/doc/numpy/user/basics.types.html#array-types-and-conversions-between-types

常用的，
C++int對應Pythonnp.intc
C++float對應Pythonnp.single
C++double對應Pythonnp.double

numpy數組保存為二進制文件

import numpy as np
a = np.array([[1.888,2.8888,3.88888],[4.666,5.6666,6.66666]])
a = a.astype(np.double)
a.astype('double').tofile("varr.data")

C++讀取二進制文件

#include <iostream>
#include <fstream>
using namespace std;

int main()
{
  double fnum[2][3] = {0};

  ifstream in("varr.data", ios::in | ios::binary);
  in.read((char *) &fnum, sizeof fnum);

  // see how many bytes have been read
  cout << in.gcount() << " bytes read\n";

  for(int j=0; j<3; j++) // show values read from file
  cout << fnum[0][j] << " ";

  in.close();

  return 0;
}

如果數據類型對應關系錯了，例如把a.astype('double').tofile("varr.data")改成a.astype('float).tofile("varr.data")`，C++中讀取的數組將不再是原來的數組。需要特別注意。