add overflow tests for Delta filter

Author: ehgus

numcodecs/tests/test_delta.py

@@ -64,3 +64,44 @@ def test_errors():
         Delta(dtype=object)
     with pytest.raises(ValueError):
         Delta(dtype='i8', astype=object)
+
+
+error_proned_integer_type_pairs = [
+    ('i2', 'i1'),
+    ('i4', 'i2'),
+    ('i8', 'i4'),
+    ('u2', 'u1'),
+    ('u4', 'u2'),
+    ('u8', 'u4'),
+]
+
+
+# # TODO: Keep it until Numpy bug is fixed
+# # Numpy does not generate overflow bug properly for integer values
+# # Minimal working example:
+# # import numpy as np
+# # arr = np.zeros((1,), dtype = 'i1')
+# # arr[0:] = np.array([128], dtype='i2') # This should raise overflow error or warning, but it does not
+# # arr[0] = 128 # This raise overflow error
+# def test_error_proned_integer_encode():
+#     for dtype, astype in error_proned_integer_type_pairs:
+#         codec = Delta(dtype=dtype, astype=astype)
+#         arr = np.array([0, np.iinfo(astype).max + 1], dtype=dtype)
+#         with pytest.raises(RuntimeWarning) as runtime_warning:
+#             codec.encode(arr)
+#         assert "overflow" in str(runtime_warning.value)
+
+
+error_proned_float_type_pairs = [
+    ('f4', 'f2'),
+    ('f8', 'f4'),
+]
+
+
+def test_error_proned_float_encode():
+    for dtype, astype in error_proned_float_type_pairs:
+        codec = Delta(dtype=dtype, astype=astype)
+        arr = np.array([0, np.astype(np.finfo(astype).max, dtype) * 2], dtype=dtype)
+        with pytest.raises(RuntimeWarning) as runtime_warning:
+            codec.encode(arr)
+        assert "overflow" in str(runtime_warning.value)