How to correcty install and use JyNI with numpy support on Jython? about jyni HOT 2 OPEN

It seems that the previous problem was connected with too late version of Numpy: it was 1.16. As you recommend, I switched to Numpy 1.13.3. Jython currently cannot install numpy, but I found workaround: installed Python 2.7.9, installed their Numpy 1.13.3 and copied the folders numpy and numpy-1.13.3.dist-info from Python's directory site-packages into the same directory of Jython.

Now my test SimpleJythonNumpyTest.py consists of only 1 line:
import numpy as np
I call it via the following command:
java -cp C:\jython2.7.2\jython.jar;C:\JyNI\JyNI.jar org.python.util.jython SimpleJythonNumpyTest.py

Java version: "17.0.2" 2022-01-18 LTS
Java(TM) SE Runtime Environment (build 17.0.2+8-LTS-86)
Java HotSpot(TM) 64-Bit Server VM (build 17.0.2+8-LTS-86, mixed mode, sharing)

Numpy is successully imported, but the result is the system crash: see hs_err_pid8120.log
hs_err_pid8120.log

from jyni.

Hello Daniel, thank you for your interest in JyNI. However, if your goal is to do data-science beyond some numpy-numerics, Jython+JyNI is probably not sufficient. Already SciPy does not work and NumPy only works for a narrow set of versions and somewhat limited functionality. That JyNI is experimental is indicated by the "alpha" in the version name.

copied the folders numpy and numpy-1.13.3.dist-info from Python's directory site-packages into the same directory of Jython.

Yes, that is much how I used to do it. Using CPython 2.7 pip to install NumPy and dependencies is my recommended way. Then simply add the path to Jython's library path (sys.path.append(...)). No need to copy things around. E.g. from my setup:

sys.path.append('C:/Program Files/Python/Python2.7.13/Lib/site-packages')
sys.path.append('C:/Program Files/Python/Python2.7.13/DLLs')

Alternatively, one can build NumPy from source (in-place build!) (easy on Linux, difficult on Windows) and put that folder on the library path.

Anyway. There are plenty of reasons for broken config. Last but not least I never tested on recent Java (also because Jython itself used to have issues with recent Java). Easiest way to check out JyNI with NumPy is a stand-alone image I do not distribute publicly. Write to me at [email protected].

from jyni.