Add test to verify dropna behaviour on pd.SparseArray

verify that calling `dropna` on a pd.SparseArray does not inadvertently drop non-na records on both DataFrames and the SparseArray itself.

pandas/tests/arrays/sparse/test_array.py

@@ -1295,3 +1295,14 @@ def test_map_missing():
 
  result = arr.map({0: 10, 1: 11})
  tm.assert_sp_array_equal(result, expected)
+
+
+def test_dropna_sparse_column_doesnt_drop_nonna():
+ # GH-28287
+ arr = SparseArray([np.nan, 1])
+ exp = SparseArray([1.0])
+ tm.assert_sp_array_equal(arr.dropna(), exp)
+
+ df = pd.DataFrame({"a": [0, 1], "b": arr})
+ expected_df = pd.DataFrame({"a": [1], "b": exp}, index=pd.Int64Index([1]))
+ tm.assert_equal(df.dropna(), expected_df)