Calling Python Functions from Java (Without Jython, because it is too slow.) [closed]

There are a bunch of Python/Java bridges which allow you to do this:

• JEP embeds CPython in Java through JNI allowing Java to call down into Python. It's purely in-process.

• jpy is another in-process implementation. One of its key features is support for fast pass-by-value operations with arrays by use of pointer hand-off.

• jpype is another in-process implementation. Since it also uses internal C-based handoff it's highly performant.

• PyJNIus is another implementation which accesses Java classes using JNI.

I am a maintainer of another one: PJRmi. For this case, the PJRmi version of the code might look something like this:

import com.deshaw.pjrmi.PythonMinion;
import com.deshaw.pjrmi.PythonMinionProvider;
import java.util.Arrays;

public class PythonFunctionRunner
{
    public static void main(String... args)
        throws Exception
    {
        // Create the Python instance
        final PythonMinion python = PythonMinionProvider.spawn();

        // Pretend that we read in this Python code from a file on disk
        final String code =
            "import numpy\n" +
            "def my_sum(array):\n" +
            "    return numpy.sum(array)\n"
            ;

        // Push it into the interpreter
        python.exec(code);

        // Create an argument to give it
        final double[] array = new double[10];
        for (int i=0; i < array.length; i++) {
            array[i] = i;
        }

        // Call the method on the array, seeing how fast it goes
        final int count = 10000;
        final long start = System.nanoTime();
        double result = 0;
        for (int i=0; i < count; i++) {
            result = python.invoke("my_sum", double.class, array);
        }
        final long end = System.nanoTime();
        final long micros = (end - start) / 1000;
        System.out.println("my_sum(" + Arrays.toString(array) + ") = " + result);
        System.out.println("Handled " + (count * 1_000_000 / micros) + " calls per second");
    }
}

which would yield something like:

$ java PythonFunctionRunner
my_sum([0.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0]) = 45.0
Handled 109 calls per second

In my limited experience I have found the other bridges to be similar in nature. Which one is best for you likely depends on exactly what you want.