check sorted indices in Subset (#2510)

* Update sparse_bin.hpp

* check sorted in c_api

* fix python package

* fix tests

* fix test

* std::is_sorted

* Update basic.py

python-package/lightgbm/basic.py

@@ -1095,7 +1095,7 @@ class Dataset(object):
                       free_raw_data=self.free_raw_data)
         ret._predictor = self._predictor
         ret.pandas_categorical = self.pandas_categorical
-        ret.used_indices = used_indices
+        ret.used_indices = sorted(used_indices)
         return ret
 
     def save_binary(self, filename):

python-package/lightgbm/engine.py

@@ -339,8 +339,8 @@ def _make_n_folds(full_data, folds, nfold, params, seed, fpreproc=None, stratifi
 
     ret = _CVBooster()
     for train_idx, test_idx in folds:
-        train_set = full_data.subset(train_idx)
-        valid_set = full_data.subset(test_idx)
+        train_set = full_data.subset(sorted(train_idx))
+        valid_set = full_data.subset(sorted(test_idx))
         # run preprocessing on the data set if needed
         if fpreproc is not None:
             train_set, valid_set, tparam = fpreproc(train_set, valid_set, params.copy())

src/c_api.cpp

@@ -921,6 +921,9 @@ int LGBM_DatasetGetSubset(
   const int32_t lower = 0;
   const int32_t upper = full_dataset->num_data() - 1;
   Common::CheckElementsIntervalClosed(used_row_indices, lower, upper, num_used_row_indices, "Used indices of subset");
+  if (!std::is_sorted(used_row_indices, used_row_indices + num_used_row_indices)) {
+    Log::Fatal("used_row_indices should be sorted in Subset");
+  }
   auto ret = std::unique_ptr<Dataset>(new Dataset(num_used_row_indices));
   ret->CopyFeatureMapperFrom(full_dataset);
   ret->CopySubset(full_dataset, used_row_indices, num_used_row_indices, true);

tests/python_package_test/test_engine.py

@@ -820,9 +820,9 @@ class TestEngine(unittest.TestCase):
         data = np.random.random((500, 2))
         y = [1] * 250 + [0] * 250
         lgb_train = lgb.Dataset(data, y, free_raw_data=False)
-        subset_index_1 = np.random.choice(np.arange(500), 300, replace=False)
+        subset_index_1 = sorted(np.random.choice(np.arange(500), 300, replace=False))
         subset_data_1 = lgb_train.subset(subset_index_1)
-        subset_index_2 = np.random.choice(np.arange(500), 200, replace=False)
+        subset_index_2 = sorted(np.random.choice(np.arange(500), 200, replace=False))
         subset_data_2 = lgb_train.subset(subset_index_2)
         params = {
             'objective': 'binary',
@@ -1601,8 +1601,8 @@ class TestEngine(unittest.TestCase):
         iter_min = min([iter_min_l1, iter_min_l2])
         iter_min_valid1 = min([iter_valid1_l1, iter_valid1_l2])
 
-        iter_cv_l1 = 3
-        iter_cv_l2 = 17
+        iter_cv_l1 = 4
+        iter_cv_l2 = 12
         self.assertEqual(len(set([iter_cv_l1, iter_cv_l2])), 2)
         iter_cv_min = min([iter_cv_l1, iter_cv_l2])