Contributing to FSDA Python Integration - Wrapper Package Ready for Review (Issue #212)

[qaseem97]

Hi everyone,

I’ve been following this initiative, and I’m very excited about the goal of making FSDA accessible to Python users. I have already started working on this by developing a robust wrapper architecture that addresses the challenges of bridging FSDA with the Python ecosystem.

have successfully compiled the FSDA toolbox into a standalone Python package (fsda_wrapper_pkg) and developed a wrapper that handles:
Seamless Integration: Instead of manual engine configuration, I’ve implemented an initialize() function that handles pathing and setup automatically.

Robust Data Handling: A dedicated fsda_wrapper.m interface that performs dynamic data type conversion (e.g., handling matlab.double casting and empty input arguments) which was a critical roadblock.

Cross-Functionality: I have already tested and integrated the following core FSDA functions:

• FSM (with and without interactive plots)
• FSR (Robust Regression)
• zscoreFS (including 3D/Sn dimension support)
• CorAna (Correspondence Analysis)

Developer Experience: Built a menu-driven Python interface to demonstrate how natural the workflow feels for Python users.

I have documented the architectural choices (Why I chose a compiled package over direct engine calls) and the error-handling strategies I used to overcome MATLAB-Python interface limitations.

would love to share my progress, contribute this code to the project and i am sharing my drive link where i have added my package + architecture + my testing script and live testing video

Prerequisites

Before you begin, ensure you have these installed:

-Python (Version 3.12.10 or less greater than this may create issues).
-MATLAB Software
-MATLAB Runtime (The version must match the one used to compile the package, e.g., 2025b).

How to Install and Run

Open your Project Folder:
Navigate to the folder named package where you see the file setup_project.bat.

Run the Installer:

Simply double-click the setup_project.bat file.

-The script will automatically create a Virtual Environment (venv) for you.
-It will install all necessary libraries (matlabengine and numpy) inside that environment.
-it will trigger pip install -e . using your setup.py file, which registers your fsda_wrapper_pkg so you can use it in your code.

Once finished, the environment will be ready, and your package will be linked to your Python setup.

i have share my code file in drive link you can test that code too.

link : https://drive.google.com/drive/folders/16JlIWzzOvxwUOU8Zq9SZjknCpfueWEx3?usp=sharing

@AldoCorbelliniUNIPR @MarcoRianiUNIPR @UniprJRC @Asadunipr @qaseem97

[UniprJRC]

Hi @qaseem97, nice work.
However, before moving further with coding, let us (me and prof. Riani) outline a few foundational points.
This project was designed to be a collaborative, public endeavor among you students. Before anyone begins writing code, we would like to setup a meeting with all interested students. During this session, we will clearly outline the project objectives, discuss the technological tools, define each participant’s role, and map out the overall timeline.
That said, we should point out that we have already extensively tested the solution you are proposing in the past. Unfortunately, both the results and the user reception were not as good as expected.
In your schematic workflow, the lower layer utilizes the MATLAB Runtime (MCR) to connect the FSDA core functions with the function wrappers. This approach is mandatory only if you plan to distribute your software to users who do not have a MathWorks license. However, if a license is available, it is much better to use the MATLAB Engine (which I see you are already using) and then completely ditch the MCR framework. The MCR is completely unnecessary here, massive in size (now around 6 GB), difficult to set up due to required environment variables, inflexible, and in our experience, rarely well-accepted by users.
Shortly, Dr. Domenico Perrotta from the European Union’s Joint Research Centre will send out invitations to all interested students for a kickoff meeting. We will jointly discuss the architectural details and outline potential development scenarios there, so please hold off on further coding until then.

[qaseem97]

@UniprJRC @MarcoRianiUNIPR @AldoCorbelliniUNIPR

Thank you for the feedback and clarification.

I understand the concerns regarding the MATLAB Runtime (MCR) dependency. My intention was not to make MCR a mandatory part of the architecture. The main focus of my work was the wrapper layer, including automatic initialization, data conversion, argument handling, and a generic gateway for calling FSDA functions from Python.

The current prototype already uses the MATLAB Engine for function execution, while I also experimented with packaging and deployment options, which is why MCR appeared in the documentation and architecture diagram.

I will hold off on further development until the kickoff meeting and I look forward to discussing the architecture and possible development directions with the team.

Thank you again for the guidance.