jclay / cppyy-knearestneighbors-example

This example demostrates using the cppyy CMake tools to automatically create Python bindings from C++.

Readme

This is an example project showing how to integrate a C++ kNearestNeighbors implementation with Cppyy to enable calling from Python.

There are two C++ classes defined - Point and NearestNeighbor. From C++ we can use it as follows:

#include "nearest_neighbors.h"
#include <iostream>

using namespace std;
int main(void) {
  NearestNeighbors knn;

  knn.points.push_back(Point(2, 0));
  knn.points.push_back(Point(1, 0));
  knn.points.push_back(Point(0, 10));
  knn.points.push_back(Point(5, 5));
  knn.points.push_back(Point(2, 5));

  auto nearest = knn.nearest(Point(0, 0), 3);
  for (const auto n : nearest) {
    cout << n.x << ", " << n.y << endl;
  }
}

And from the Python side:

from cppyy.gbl import std
import knn_example
from knn_example import NearestNeighbors, Point

# TODO put this in a pythonization that gets loaded in automatically
# Tell python how to print our C++ Class
Point.__repr__ = lambda self: repr(str(self.x) + ", " + str(self.y))

knn = NearestNeighbors()
points = [Point(2,0), Point(1,0), Point(0,10), Point(5,5), Point(2,5)]
knn.points = std.vector[Point](points)
result = knn.nearest(Point(2.1, 5.3), 4)
print(list(result))

Setup

Note: I'm using Python 2

This is because even the latest clang-python in Debian (for clang 6) provides Python 2 bindings only. It seems that Python 3 support should work but, but for now it will likely require building Clang from source.

I have only tested this setup using Debian (running in Windows Subsystem for Linux) with the latest testing repository (buster).

The Cppyy docs are a great resource and I highly recommend you read them before trying the project.

Install dependencies

sudo apt-get install python2 python-pip
sudo apt-get install clang-6.0
sudo apt-get install libclang-6.0-dev
sudo apt-get install python-clang-6.0

I encountered the following bug which is addressed by explicitly installing with the following:

pip install cppyy-cling
pip install cppyy-backend
pip install CpyCppyy
pip install cppyy

We can verify that things are working with:

$ python2
>>>  from clang.cindex import Config
>>>  Config().lib
>>>  <CDLL 'libclang-6.0.so', handle 7fffe4d35490 at 7f496430ee90>

Building

mkdir build && cd build
cmake -G Ninja ..
cmake --build .

You can test the C++ driver with ./knn to ensure the program runs correctly.

Using with Python

From the directory containing the knn_examples folder (build in this case):

from cppyy.gbl import std
import knn_example
from knn_example import NearestNeighbors, Point

# TODO put this in a pythonization that gets loaded in automatically
# Tell python how to print our C++ Class
Point.__repr__ = lambda self: repr(str(self.x) + ", " + str(self.y))

knn = NearestNeighbors()
points = [Point(2,0), Point(1,0), Point(0,10), Point(5,5), Point(2,5)]
knn.points = std.vector[Point](points)
result = knn.nearest(Point(2.1, 5.3), 4)
print(list(result))