| 3651 | |
| 3652 | @pytest.mark.numpy |
| 3653 | def test_array_from_numpy_str_utf8(): |
| 3654 | # ARROW-3890 -- in Python 3, NPY_UNICODE arrays are produced, but in Python |
| 3655 | # 2 they are NPY_STRING (binary), so we must do UTF-8 validation |
| 3656 | vec = np.array(["toto", "tata"]) |
| 3657 | vec2 = np.array(["toto", "tata"], dtype=object) |
| 3658 | |
| 3659 | arr = pa.array(vec, pa.string()) |
| 3660 | arr2 = pa.array(vec2, pa.string()) |
| 3661 | expected = pa.array(["toto", "tata"]) |
| 3662 | assert arr.equals(expected) |
| 3663 | assert arr2.equals(expected) |
| 3664 | |
| 3665 | # with mask, separate code path |
| 3666 | mask = np.array([False, False], dtype=bool) |
| 3667 | arr = pa.array(vec, pa.string(), mask=mask) |
| 3668 | assert arr.equals(expected) |
| 3669 | |
| 3670 | # UTF8 validation failures |
| 3671 | vec = np.array([('mañana').encode('utf-16-le')]) |
| 3672 | with pytest.raises(ValueError): |
| 3673 | pa.array(vec, pa.string()) |
| 3674 | |
| 3675 | with pytest.raises(ValueError): |
| 3676 | pa.array(vec, pa.string(), mask=np.array([False])) |
| 3677 | |
| 3678 | |
| 3679 | @pytest.mark.numpy |