Should use other->numeric_feature_map_[i] in AddFeaturesFrom

[EdmondElephant]

Description

In this line: https://github.com/microsoft/LightGBM/blob/master/src/io/dataset.cpp#L1486 I think

int feat_ind = numeric_feature_map_[i];

should be

int feat_ind = other->numeric_feature_map_[i];

The loop condition i < (other->numeric_feature_map_).size() tells the same story.

Reproducible example

Code

import lightgbm as lgb
import numpy as np

a = np.zeros((100, 1), dtype=np.float32)
b = np.random.normal(size=(100, 5))

dataset_a = lgb.Dataset(a).construct()
dataset_b = lgb.Dataset(b).construct()

dataset_a.add_features_from(dataset_b)

Output

[LightGBM] [Warning] There are no meaningful features, as all feature values are constant.
[LightGBM] [Warning] Find the same feature name (Column_0) in Dataset::AddFeaturesFrom, change its name to (D2_Column_0)

Traceback (most recent call last):
  File "add.py", line 10, in <module>
    dataset_a.add_features_from(dataset_b)
  File "C:\test_lgb_add\lib\site-packages\lightgbm\basic.py", line 2437, in add_features_from
    _safe_call(_LIB.LGBM_DatasetAddFeaturesFrom(self.handle, other.handle))
OSError: exception: access violation reading 0x0000000000000000

Environment info

Command(s) you used to install LightGBM

pip install lightgbm==3.3.2

[jmoralez]

@guolinke @shiyu1994 can you please share your thoughts on this?

[guolinke]

thank you, I think it is a bug

[jmoralez]

Thanks! Do you agree with the proposed change?

[guolinke]

yeah, the fix looks good to me!

[jmoralez]

@EdmondElephant would you like to submit a PR with your proposed fix and your example as a test?

[EdmondElephant]

@jmoralez Submitted a PR: https://github.com/microsoft/LightGBM/pull/5434

[github-actions[bot]]

This issue 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.