Use ctypes to pass parameters (fixes #3398)

[btrotta]

I was able to reproduce the bug from #3398 consistently on my machine with the script below (Windows, Visual Studio, Python 2.7). It always crashes on seed=105.

I think the cause is that LGBM_DatasetGetFeatureNames is expecting a size_t input for buffer_len but we pass a Python integer. Apparently this can cause unpredictable behavior depending on the compiler, see https://stackoverflow.com/questions/51903911/calling-c-functions-via-python-ctypes-why-does-passing-uints-to-a-function-expe

import lightgbm as lgb
from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(*load_breast_cancer(return_X_y=True),
                                                                        test_size=0.1, random_state=1)
train_data = lgb.Dataset(X_train, y_train)

for seed in range(0, 200):
    print(seed)
    params = {
        "objective": "binary",
        "verbose": -1,
        "num_leaves": 3,
        "seed": seed
    }

    gbm1 = lgb.LGBMClassifier(n_estimators=3, num_leaves=3, silent=True)
    gbm1.fit(X_train, y_train)
    print(gbm1.booster_.feature_name()[27])

[github-actions]

This pull request has been automatically locked since there has not been any recent activity since it was closed. To start a new related discussion, open a new issue at https://github.com/microsoft/LightGBM/issues including a reference to this.