| from io import StringIO | |
| import numpy as np | |
| import pytest | |
| import pandas as pd | |
| import pandas._testing as tm | |
| from pandas.core.arrays import ExtensionArray | |
| class BaseParsingTests: | |
| def test_EA_types(self, engine, data, request): | |
| if isinstance(data.dtype, pd.CategoricalDtype): | |
| # in parsers.pyx _convert_with_dtype there is special-casing for | |
| # Categorical that pre-empts _from_sequence_of_strings | |
| pass | |
| elif isinstance(data.dtype, pd.core.dtypes.dtypes.NumpyEADtype): | |
| # These get unwrapped internally so are treated as numpy dtypes | |
| # in the parsers.pyx code | |
| pass | |
| elif ( | |
| type(data)._from_sequence_of_strings.__func__ | |
| is ExtensionArray._from_sequence_of_strings.__func__ | |
| ): | |
| # i.e. the EA hasn't overridden _from_sequence_of_strings | |
| mark = pytest.mark.xfail( | |
| reason="_from_sequence_of_strings not implemented", | |
| raises=NotImplementedError, | |
| ) | |
| request.node.add_marker(mark) | |
| df = pd.DataFrame({"with_dtype": pd.Series(data, dtype=str(data.dtype))}) | |
| csv_output = df.to_csv(index=False, na_rep=np.nan) | |
| result = pd.read_csv( | |
| StringIO(csv_output), dtype={"with_dtype": str(data.dtype)}, engine=engine | |
| ) | |
| expected = df | |
| tm.assert_frame_equal(result, expected) | |