Support for Python3.12 and up, update dependency versions

[ogencoglu]

Is this project abandoned? Has not been any updates for years. Dependency versions are really old.

Is there a planned support for newer python versions?

[ehoogeveen-medweb]

Still seems to be working okay for me with newer dependencies, with one exception: scikit-learn 1.7 removes BaseEstimator._validate_data in favor of utils.validation.validate_data. To make it work:

1. Find and open gplearn/genetic.py
2. On line 24, add validate_data to the imports from sklearn.utils.validation
3. On lines 292 and 312, replace self._validate_data(...) with validate_data(self, ...) (keep the other arguments the same)

By the way, I'm trying out free-threaded CPython and it seems to be working great - way lower overhead than the multi-process options (which need to serialize and deserialize the data). To enable it:

1. Make sure you're using Python 3.13 with the experimental free-threaded binaries installed
2. Create a virtual environment with the free-threaded Python and activate it
   1. Linux: python3.13t -m venv venv && source venv/bin/activate
   2. Windows: py -3.13t -m venv venv && venv\Scripts\activate.bat
3. Install gplearn, numpy 2.3.0 and scikit-learn 1.6 or higher using pip
4. Add from joblib import parallel_config to your script
5. Add parallel_config(backend = 'threading') to your script
6. Run Python, e.g. python myscript.py
7. If you get a warning about the GIL being enabled, try running with -Xgil=0, e.g. python -Xgil=0 myscript.py (but if your non-thread safe dependencies use threading, it might cause problems)

[fabianostermann]

It would be great if there was a pull request and an update to PIPy for this package. I use this package for teaching and it still works great! Thx in advance for considering an update.

[trevorstephens]

thanks all for flagging and for the tips for external users as well @ehoogeveen-medweb !

I'm working on an update now to update dependencies and get tests to pass, as well as update packaging to more modern platforms in a first pass. will review the issue tracker for other easy items to add into a release.

[Greesb]

Still seems to be working okay for me with newer dependencies, with one exception: scikit-learn 1.7 removes BaseEstimator._validate_data in favor of utils.validation.validate_data. To make it work:

1. Find and open `gplearn/genetic.py`

2. On line 24, add `validate_data` to the imports from `sklearn.utils.validation`

3. On lines 292 and 312, replace `self._validate_data(...)` with `validate_data(self, ...)` (keep the other arguments the same)

By the way, I'm trying out free-threaded CPython and it seems to be working great - way lower overhead than the multi-process options (which need to serialize and deserialize the data). To enable it:

1. Make sure you're using Python 3.13 with the experimental free-threaded binaries installed

2. Create a virtual environment with the free-threaded Python and activate it
   
   1. Linux: `python3.13t -m venv venv && source venv/bin/activate`
   2. Windows: `py -3.13t -m venv venv && venv\Scripts\activate.bat`

3. Install `gplearn`, `numpy` 2.3.0 and `scikit-learn` 1.6 or higher using `pip`

4. Add `from joblib import parallel_config` to your script

5. Add `parallel_config(backend = 'threading')` to your script

6. Run Python, e.g. `python myscript.py`

7. If you get a warning about the GIL being enabled, try running with `-Xgil=0`, e.g. `python -Xgil=0 myscript.py` (but if your non-thread safe dependencies use threading, it might cause problems)

Thank you for finding this.
If you do not want to edit the gplearn code each time in a new venv, adding this piece of code in your script should also work:

    from gplearn.genetic import BaseSymbolic
    from sklearn.utils.validation import validate_data

    BaseSymbolic._validate_data = lambda self, *args, **kwargs: validate_data(
        self,
        *args,
        **kwargs,
    )