pandas-select is a collection of DataFrame selectors that facilitates indexing and selecting data, fully compatible with pandas vanilla indexing.

The selector functions can choose variables based on their name, data type, arbitrary conditions, or any combination of these.

pandas-select is inspired by the excellent R library tidyselect.

pandas-select is a Python-only package hosted on PyPI. It can be installed via pip:

pip install pandas-select

• Fully compatible with the pandas.DataFrame [] operator and the pandas.DataFrame.loc accessor.
• Emphasise readability and conciseness by cutting boilerplate:

# pandas-select
df[AllNumeric()]
# vanilla
df.select_dtypes("number").columns

# pandas-select
df[StartsWith("Type") | "Legendary"]
# vanilla
df.loc[:, df.columns.str.startswith("Type") | (df.columns == "Legendary")]

• Ease the challenges of indexing with hierarchical index and offers an alternative to slicers when the labels cannot be listed manually.

# pandas-select
df_mi.loc[Contains("Jeff", axis="index", level="Name")]

# vanilla
df_mi.loc[df_mi.index.get_level_values("Name").str.contains("Jeff")]

• Play well with machine learning applications.

  • Respect the columns order.

  • Allow deferred selection when the DataFrame's columns are not known in advance, for example in automated machine learning applications.

  • Offer integration with sklearn.

    from pandas_select import AnyOf, AllBool, AllNominal, AllNumeric, ColumnSelector
    from sklearn.compose import make_column_transformer
    from sklearn.preprocessing import OneHotEncoder, StandardScaler
    
    ct = make_column_transformer(
       (StandardScaler(), ColumnSelector(AllNumeric() & ~AnyOf("Generation"))),
       (OneHotEncoder(), ColumnSelector(AllNominal() | AllBool() | "Generation")),
    )
    ct.fit_transform(df)

pandas-select is released under the BS3 license, its documentation lives at Read the Docs, the code on GitHub, and the latest release on PyPI. It is tested on Python 3.6+.