astropy-pandas

Heavily inspired by pint-pandas, we are trying to bring astropy units into pandas data frames. As of this writing this is just a proof-of-concept, a lot of work is needed to make this work properly.

Minimal example:

>>> import astropy_pandas
>>> import pandas as pd
>>> df = pd.DataFrame({
...      'x': pd.Series([13.], dtype="astropy[parsec]"),
...      'y': pd.Series([37.], dtype="astropy[angstrom]")})
>>> (df['x'] * df['y']).astropy.to('km2')
0    1484.210916697348
dtype: astropy[km2]

Roadmap to 1.0

• What can we learn from physipandas (MIT license)
• Make underlying data types explicit, astropy[parsec][float64] probably as in https://github.com/hgrecco/pint-pandas/pull/247/files
• Easy conversion to a unit-free data frame
• Implement (or test that default works) everything mentioned in https://pandas.pydata.org/docs/reference/api/pandas.api.extensions.ExtensionArray.html
• Show benchmarks (compare to data frames made from astropy quantities, compare to pint-pandas, compare to physipandas)
• Decide on a license. Probably we should use GPL3+ again. We'll heavily pull from pint-pandas which is BSD-3-clause but there should be no problem as long as we retain their copyright notice * [ ] Announce at https://stackoverflow.com/questions/39419178/how-can-i-manage-units-in-pandas-data
• Mention at https://pint.readthedocs.io/en/0.10.1/pint-pandas.html
• Mention at https://pandas.pydata.org/community/ecosystem.html somewhere. However, maybe it's better to stick to BSD-3-Clause so pint-pandas can pull things back from us. Also pandas and astropy are both BSD-3-clause.

Potential Limitations

• Plotly doesn't like these data frames: hgrecco/pint-pandas#249
• set_index might not work: hgrecco/pint-pandas#231 and hgrecco/pint-pandas#44
• Problems when comparing data frames with different units
• allclose might not work, see hgrecco/pint-pandas#108
• apply might not work on DataFrame, see hgrecco/pint-pandas#55