Python3 Version Upgrade

[mbrzus]

Description

Thank you for recently releasing this tool—it looks very useful, and I'd like to start using it in my projects.
I noticed the tool explicitly requires Python 3.8.0, which I acknowledge. However, Python 3.8 reached its end-of-life (EOL) in October 2024, meaning it no longer receives security updates, bug fixes, or support from the Python core team. This is concerning for a newly released tool, as it could expose users to vulnerabilities and limit compatibility with modern libraries, environments, or operating systems. Additionally, many third-party libraries have already dropped support for Python 3.8, making it more difficult to set up a compatible environment without relying on outdated or vulnerable dependencies.
To improve adoption, security, and long-term maintainability, I kindly request considering an upgrade to a currently supported Python version, such as 3.12 (coming in standard new OS releases such as Ubuntu 24.04).

Suggested Changes

• Update the setup.py, pyproject.toml, or equivalent to specify a range like python_requires='>=3.10'.
• Test the tool on recent Python versions (e.g., 3.10–3.12) and address any deprecations or incompatibilities (tools like pyupgrade can help automate this).
• If there's a specific reason for sticking with 3.8 (e.g., compatibility with certain dependencies), perhaps document it or provide guidance for users on newer setups.

I'm happy to help test a patched version on my environment or provide more feedback. Thanks for your consideration!

[perecanals]

Hi @mbrzus, I agree it would be best to try to update to a more updated, secure version of Python. Were you able to correctly install the conda environment as described without any incompatibilities? I have failed to do so following the instructions in the repo in two independent servers using Ubuntu 22.04 and was wondering if anybody else was running into the same issue.

Thanks!

[ezequieldlrosa]

Hi @mbrzus,

Thanks a lot for raising these important points and for your suggestions.

DeepISLES integrates three independent algorithms from different research groups, and finding a single Python version that satisfies all dependencies was quite challenging. Some of the libraries the challenge researchers rely on (e.g., SimpleITK-SimpleElastix) are deprecated and not maintained in newer Python releases, which is the main reason we had to pin the codebase to Python 3.8.

That said, we fully agree with your concern and we plan to:

1. Identify and replace libraries that no longer build under newer Python versions.
2. Stress-test again DeepISLES to ensure we don't degrade performance by introducing such changes.

This will take us some time, but we do aim to upgrade the supported Python versions (e.g., 3.10–3.12) in a future release. In the meantime, we’ll update the documentation addressing your 'Documentation' suggestions (points 1. and 3.)

If you prefer not to run Python 3.8 locally, please also consider using our Docker image, which isolates the environment and avoids dependency conflicts.

Thanks again for your constructive feedback. We'll be very happy if you help us test the newer release once we have it in place.

Best,
Ezequiel

[ezequieldlrosa]

Hi @perecanals,

Thanks for raising this. Could you share a bit more detail on the issues you ran into? For example, whether the errors appeared during a specific package installation.

We developed and tested the code on Ubuntu 24.04.2 across a few different machines, so it would be very helpful to know exactly where your installation is getting stuck.

Looking forward to your feedback,
Best,
Ezequiel

[perecanals]

Hi @ezequieldlrosa

Thanks for the explanation on the upgrade plan for DeepIsles. I will open up a new issue with my specific problem.