CPPYY Part II

动态特性

我们来实现一个简单的KNN算法

"""
新建一个文件 nearest_neighbors.h
------------------
#pragma once

#include <iostream>
#include <vector>

class Point {
public:
  Point() {};
  Point(double x, double y) : x(x), y(y) {};
  double x, y;
};

class NearestNeighbors {
 public:
  NearestNeighbors() {};
  std::vector<Point> points;
  std::vector<Point> nearest(Point, int k);
};
------------------
"""
# 将头文件include 进来
cppyy.include("nearest_neighbors.h")

# 现在已经可以使用 Point 和 NearestNeighbors 类了
Point = cppyy.gbl.Point
KNN = cppyy.gbl.NearestNeighbors

# 构造一个 Point
p = Point(1.0, 2.0)
print(repr(p))

<cppyy.gbl.Point object at 0x000001E120E918B0>

# 我们希望print(p) 可以看到具体的信息
Point.__repr__ = lambda self: "<Point object at 0x{:04X}, x: {}, y: {}>".format(id(self), self.x, self.y)

# 再次print(p)
print(repr(p))  # >> p   # in ipython

<Point object at 0x1E122C5CB40, x: 1.0, y: 2.0>

# 现在初始化 KNN，需要一系列的Points
from random import random
points = [Point(random()*10, random()*10) for _ in range(10)]
knn = KNN()
knn.points = Vector[Point](points)

list(knn.points)

[<Point object at 0x1E122EC5CA8, x: 8.423026486604515, y: 4.205179067516682>,
 <Point object at 0x1E1217F8438, x: 1.38495460424919, y: 8.902270221641716>,
 <Point object at 0x1E122FB5870, x: 5.4137809335642615, y: 5.081185223495259>,
 <Point object at 0x1E122FB56C0, x: 4.43062339130408, y: 0.3717755170621895>,
 <Point object at 0x1E122FB5990, x: 7.143169297375893, y: 7.112832543272915>,
 <Point object at 0x1E122FB5B88, x: 2.619781100077131, y: 2.0009995637839486>,
 <Point object at 0x1E122FB55E8, x: 9.678854624172509, y: 7.42921725146636>,
 <Point object at 0x1E122FB53F0, x: 9.058931286852678, y: 6.025951815850337>,
 <Point object at 0x1E122FB5FC0, x: 1.2802522942068906, y: 3.7686758142413557>,
 <Point object at 0x1E122FB5BD0, x: 1.5544746354777161, y: 1.2622207703902388>]

p = Point(2.4, 3.5)

# 现在可以直接调用下面的knn.nearest(p, 2)吗？
# 答案是不行的，因为只include了头文件，knn.nearest函数是未定义的，如果调用将会报错
knn.nearest(p, 2)
-------------------------------------------------
IncrementalExecutor::executeFunction: symbol '?nearest@NearestNeighbors@@QEAA?AV?$vector@VPoint@@V?$allocator@VPoint@@@std@@@std@@VPoint@@H@Z' unresolved while linking symbol '__cf_17'!
You are probably missing the definition of public: class std::vector<class Point,class std::allocator<class Point> > __cdecl NearestNeighbors::nearest(class Point,int) __ptr64
Maybe you need to load the corresponding shared library?
    ValueError                                Traceback (most recent call last)
    <ipython-input-55-44717562f51d> in <module>
          1 # 现在可以直接调用下面的knn.nearest(p, 2)吗？
          2 # 答案是不行的，因为只include了头文件，knn.nearest函数是未定义的，如果调用将会报错
    ----> 3 knn.nearest(p, 2)
   
    ValueError: vector<Point> NearestNeighbors::nearest(Point, int k) =>
    ValueError: nullptr result where temporary expected

# 需要将包含该函数定义的动态库文件导入进来
# knn_lib.dll 为用MSVC编译的，可能是符号不一致，还是会报错，在linux下导入gcc编译的.so文件以后，是确定可行的
cppyy.load_library("knn_lib.dll")

# 此时应该输出正确的结果
knn.nearest(p, 2)

总结

• cppyy生成的python 类可以很方便的修改函数实现等
• 使用cppyy.include 与 cppyy.load_libraries来实现导入c++模块