filter_names#

skrub.selectors.filter_names(predicate, *args, **kwargs)[source]#

Select columns based only on their name.

The predicate takes as input the column name and must return True if it should be selected and False otherwise.

Parameters:

predicatepython:callable(): A function that takes a column name (string) and optional extra arguments, returning True to select the column or False to exclude it. Signature: predicate(col_name, *args, **kwargs) -> bool
*argspython:tuple: Extra positional arguments passed to the predicate. Using explicit arguments (instead of closures) helps with pickling the selector.
**kwargspython:dict: Extra keyword arguments passed to the predicate. Using explicit arguments (instead of closures) helps with pickling the selector.

Returns:

Selector: A NameFilter selector that matches columns where predicate(name, *args, **kwargs) returns True.

See also

filter: Select columns based on their content or properties
glob: Select columns by wildcard pattern matching
regex: Select columns by regular expression pattern matching

Notes

filter_names passes the column NAME (a string), while filter passes the column OBJECT (Series/DataFrame)

For common patterns, prefer simpler selectors:

s.glob('*_mm') instead of s.filter_names(lambda n: n.endswith('_mm'))
s.regex(r'col_\d+') instead of s.filter_names(lambda n: re.match(...))

To pickle the selector, use importable functions as predicates rather than lambdas. Pass parameters via *args or **kwargs:

# Picklable (importable function + explicit args)
s.filter_names(str.endswith, '_mm')

# Picklable alternative (with functools.partial)
import functools
s.filter_names(functools.partial(str.endswith, '_mm'))

# NOT picklable (closure)
suffix = '_mm'
s.filter_names(lambda name: name.endswith(suffix))

Examples

>>> from skrub import selectors as s
>>> import pandas as pd
>>> df = pd.DataFrame(
...     {
...         "height_mm": [297.0, 420.0],
...         "width_mm": [210.0, 297.0],
...         "kind": ["A4", "A3"],
...         "ID": [4, 3],
...     }
... )
>>> df
   height_mm  width_mm kind  ID
0      297.0     210.0   A4   4
1      420.0     297.0   A3   3

Prefer using *args and **kwargs to pass extra arguments to the predicate, rather than defining a dynamic function which may cause pickling errors.

Select columns ending with a suffix (using lambda):

>>> selector = s.filter_names(lambda name: name.endswith('_mm'))
>>> s.select(df, selector)
   height_mm  width_mm
0      297.0     210.0
1      420.0     297.0

Use an importable function with explicit arguments (picklable):

>>> selector = s.filter_names(str.endswith, '_mm')
>>> selector
filter_names(str.endswith, '_mm')

>>> s.select(df, selector)
   height_mm  width_mm
0      297.0     210.0
1      420.0     297.0

>>> import pickle
>>> _ = pickle.dumps(selector)  # Pickling works!

Select columns with uppercase names:

>>> s.select(df, s.filter_names(str.isupper))
   ID
0   4
1   3

Combine with type selectors:

>>> s.select(df, s.numeric() & s.filter_names(str.islower))
   height_mm  width_mm
0      297.0     210.0
1      420.0     297.0