object

skrub.selectors.object()

Select columns whose dtype is object (pandas) or pl.Object (polars).

Note that object columns may contain mixed types (e.g., strings and numbers) and are broader than string columns. Use this selector when you specifically need object-typed columns, and prefer more specific selectors like string() or categorical().

See also

string

Select string columns (preferred for text data). Use this instead of object() for text columns.

categorical

Select categorical columns.

has_dtype

Select columns whose dtype matches specific dtypes.

Notes

Warning

The behavior of string columns may change depending on the pandas version:

• Before pandas 3.0: String columns may have the object dtype

• From pandas 3.0 onwards: String columns have only the string dtype

This selector selects all object dtype columns regardless of content, including mixed-type columns. For text data, prefer string() which is more selective.

Examples

>>> from skrub import selectors as s
>>> import pandas as pd
>>> df = pd.DataFrame(
...     dict(
...         mixed=pd.Series(['A', 10]),
...         numeric=pd.Series([1, 2]),
...         string=pd.Series(['A', 'B']).convert_dtypes(),
...     )
... )
>>> df.dtypes
mixed       object
numeric      int64
string         ...
dtype: object

Select object dtype columns (note: can contain mixed types):

>>> s.select(df, s.object())
  mixed
0     A
1    10

Prefer string() for text columns:

>>> s.select(df, s.string())
  string
0      A
1      B