2

I'm using Chaquopy in my android project. I have a function in my python class which it returns a 2D array as a PyObject type. Now, I want to convert it to 2D array in my java class. How can I achieve this?

EDIT : Here is my Python code :

import numpy as np
from scipy.io import wavfile

def get_python_audio(file_path):
 fs, data_test = wavfile.read(file_path)
 print('data_test:', data_test.shape)
 data = data_test[:, 0]
 data = data[:, np.newaxis]
 print('data:', data.shape)


 return data
hdiz
  • 1,141
  • 2
  • 13
  • 27
  • What is the data type in the array? And what is the structure of the array itself, e.g. a NumPy array, a list of lists, or something else? – mhsmith Oct 26 '19 at 09:49
  • Actually, it's a .wav file with 2 dimension as i expanded it with numpy newaxis : fs, data_test = wavfile.read('file_path') data = data_test[:, 0] data = data[:, np.newaxis] – hdiz Oct 26 '19 at 10:02
  • shape of my array is (16000,1) – hdiz Oct 26 '19 at 10:17

1 Answers1

4

Chaquopy 8.x and newer:

NumPy arrays and Java arrays of the same type can now be converted directly. Assuming get_python_audio returns a 1-dimensional int16 array:

short[] data = yourModule.callAttr("get_python_audio", filePath).toJava(short[].class);

Chaquopy 7.x and older:

The fastest way to convert numerical arrays between Java and Python is to use a byte array.

We'll assume that wavfile.read returns a 1-dimensional int16 array, but this technique is easily adapted to other data types. Then in Python, you can do this:

def get_python_audio(file_path):
   fs, data_test = wavfile.read(file_path)
   return data_test.tobytes()

And in Java:

byte[] bytesData = yourModule.callAttr("get_python_audio", filePath).toJava(byte[].class);
short[] shortData = new short[bytesData.length / 2];
ByteBuffer.wrap(bytesData).order(ByteOrder.nativeOrder()).asShortBuffer().get(shortData);

If your audio has multiple channels, then you'll have to convert each channel separately.

mhsmith
  • 6,675
  • 3
  • 41
  • 58