Selectors#

Selectors allow for more intuitive selection of columns from DataFrame or LazyFrame objects based on their name, dtype or other properties. They unify and build on the related functionality that is available through the col() expression and can also broadcast expressions over the selected columns.

Importing#

Selectors are available as functions imported from polars.selectors

Typical/recommended usage is to import the module as cs and employ selectors from there.

import polars.selectors as cs
import polars as pl

df = pl.DataFrame(
    {
        "w": ["xx", "yy", "xx", "yy", "xx"],
        "x": [1, 2, 1, 4, -2],
        "y": [3.0, 4.5, 1.0, 2.5, -2.0],
        "z": ["a", "b", "a", "b", "b"],
    },
)
df.group_by(by=cs.string()).agg(cs.numeric().sum())

Set operations#

Selectors support the following set operations:

Operation	Expression
`UNION`	`A \| B`
`INTERSECTION`	`A & B`
`DIFFERENCE`	`A - B`
`SYMMETRIC DIFFERENCE`	`A ^ B`
`COMPLEMENT`	`~A`

Note that both individual selector results and selector set operations will always return matching columns in the same order as the underlying frame schema.

Examples#

import polars.selectors as cs
import polars as pl

# set up an empty dataframe with plenty of columns of various dtypes
df = pl.DataFrame(
    schema={
        "abc": pl.UInt16,
        "bbb": pl.UInt32,
        "cde": pl.Float64,
        "def": pl.Float32,
        "eee": pl.Boolean,
        "fgg": pl.Boolean,
        "ghi": pl.Time,
        "JJK": pl.Date,
        "Lmn": pl.Duration,
        "opp": pl.Datetime("ms"),
        "qqR": pl.String,
    },
)

# Select the UNION of temporal, strings and columns that start with "e"
assert df.select(cs.temporal() | cs.string() | cs.starts_with("e")).schema == {
    "eee": pl.Boolean,
    "ghi": pl.Time,
    "JJK": pl.Date,
    "Lmn": pl.Duration,
    "opp": pl.Datetime("ms"),
    "qqR": pl.String,
}

# Select the INTERSECTION of temporal and column names that match "opp" OR "JJK"
assert df.select(cs.temporal() & cs.matches("opp|JJK")).schema == {
    "JJK": pl.Date,
    "opp": pl.Datetime("ms"),
}

# Select the DIFFERENCE of temporal columns and columns that contain the name "opp" OR "JJK"
assert df.select(cs.temporal() - cs.matches("opp|JJK")).schema == {
    "ghi": pl.Time,
    "Lmn": pl.Duration,
}

# Select the SYMMETRIC DIFFERENCE of numeric columns and columns that contain an "e"
assert df.select(cs.contains("e") ^ cs.numeric()).schema == {
    "abc": UInt16,
    "bbb": UInt32,
    "eee": Boolean,
}

# Select the COMPLEMENT of all columns of dtypes Duration and Time
assert df.select(~cs.by_dtype([pl.Duration, pl.Time])).schema == {
    "abc": pl.UInt16,
    "bbb": pl.UInt32,
    "cde": pl.Float64,
    "def": pl.Float32,
    "eee": pl.Boolean,
    "fgg": pl.Boolean,
    "JJK": pl.Date,
    "opp": pl.Datetime("ms"),
    "qqR": pl.String,
}

Note

If you don’t want to use the set operations on the selectors, you can materialize them as expressions by calling as_expr. This ensures the operations OR, AND, etc are dispatched to the underlying expressions instead.

Functions#

Available selector functions:

`all`()	Select all columns.
`alpha`([ascii_only, ignore_spaces])	Select all columns with alphabetic names (eg: only letters).
`alphanumeric`([ascii_only, ignore_spaces])	Select all columns with alphanumeric names (eg: only letters and the digits 0-9).
`binary`()	Select all binary columns.
`boolean`()	Select all boolean columns.
`by_dtype`(*dtypes)	Select all columns matching the given dtypes.
`by_index`(*indices)	Select all columns matching the given indices (or range objects).
`by_name`(*names[, require_all])	Select all columns matching the given names.
`categorical`()	Select all categorical columns.
`contains`(*substring)	Select columns whose names contain the given literal substring(s).
`date`()	Select all date columns.
`datetime`([time_unit, time_zone])	Select all datetime columns, optionally filtering by time unit/zone.
`decimal`()	Select all decimal columns.
`digit`([ascii_only])	Select all columns having names consisting only of digits.
`duration`([time_unit])	Select all duration columns, optionally filtering by time unit.
`ends_with`(*suffix)	Select columns that end with the given substring(s).
`exclude`(columns, *more_columns)	Select all columns except those matching the given columns, datatypes, or selectors.
`expand_selector`(target, selector, *[, strict])	Expand selector to column names, with respect to a specific frame or target schema.
`first`()	Select the first column in the current scope.
`float`()	Select all float columns.
`integer`()	Select all integer columns.
`is_selector`(obj)	Indicate whether the given object/expression is a selector.
`last`()	Select the last column in the current scope.
`matches`(pattern)	Select all columns that match the given regex pattern.
`numeric`()	Select all numeric columns.
`signed_integer`()	Select all signed integer columns.
`starts_with`(*prefix)	Select columns that start with the given substring(s).
`string`(*[, include_categorical])	Select all String (and, optionally, Categorical) string columns.
`temporal`()	Select all temporal columns.
`time`()	Select all time columns.
`unsigned_integer`()	Select all unsigned integer columns.

polars.selectors.all() → SelectorType[source]#

Select all columns.

See also

first: Select the first column in the current scope.
last: Select the last column in the current scope.

Examples

>>> from datetime import date
>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "dt": [date(1999, 12, 31), date(2024, 1, 1)],
...         "value": [1_234_500, 5_000_555],
...     },
...     schema_overrides={"value": pl.Int32},
... )

Select all columns, casting them to string:

>>> df.select(cs.all().cast(pl.String))
shape: (2, 2)
┌────────────┬─────────┐
│ dt         ┆ value   │
│ ---        ┆ ---     │
│ str        ┆ str     │
╞════════════╪═════════╡
│ 1999-12-31 ┆ 1234500 │
│ 2024-01-01 ┆ 5000555 │
└────────────┴─────────┘

Select all columns except for those matching the given dtypes:

>>> df.select(cs.all() - cs.numeric())
shape: (2, 1)
┌────────────┐
│ dt         │
│ ---        │
│ date       │
╞════════════╡
│ 1999-12-31 │
│ 2024-01-01 │
└────────────┘

polars.selectors.alpha(ascii_only: bool = False, *, ignore_spaces: bool = False) → SelectorType[source]#

Select all columns with alphabetic names (eg: only letters).

Parameters:

ascii_only: Indicate whether to consider only ASCII alphabetic characters, or the full Unicode range of valid letters (accented, idiographic, etc).
ignore_spaces: Indicate whether to ignore the presence of spaces in column names; if so, only the other (non-space) characters are considered.

Notes

Matching column names cannot contain any non-alphabetic characters. Note that the definition of “alphabetic” consists of all valid Unicode alphabetic characters (p{Alphabetic}) by default; this can be changed by setting ascii_only=True.

Examples

>>> import polars as pl
>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "no1": [100, 200, 300],
...         "café": ["espresso", "latte", "mocha"],
...         "t or f": [True, False, None],
...         "hmm": ["aaa", "bbb", "ccc"],
...         "都市": ["東京", "大阪", "京都"],
...     }
... )

Select columns with alphabetic names; note that accented characters and kanji are recognised as alphabetic here:

>>> df.select(cs.alpha())
shape: (3, 3)
┌──────────┬─────┬──────┐
│ café     ┆ hmm ┆ 都市 │
│ ---      ┆ --- ┆ ---  │
│ str      ┆ str ┆ str  │
╞══════════╪═════╪══════╡
│ espresso ┆ aaa ┆ 東京 │
│ latte    ┆ bbb ┆ 大阪 │
│ mocha    ┆ ccc ┆ 京都 │
└──────────┴─────┴──────┘

Constrain the definition of “alphabetic” to ASCII characters only:

>>> df.select(cs.alpha(ascii_only=True))
shape: (3, 1)
┌─────┐
│ hmm │
│ --- │
│ str │
╞═════╡
│ aaa │
│ bbb │
│ ccc │
└─────┘

>>> df.select(cs.alpha(ascii_only=True, ignore_spaces=True))
shape: (3, 2)
┌────────┬─────┐
│ t or f ┆ hmm │
│ ---    ┆ --- │
│ bool   ┆ str │
╞════════╪═════╡
│ true   ┆ aaa │
│ false  ┆ bbb │
│ null   ┆ ccc │
└────────┴─────┘

Select all columns except for those with alphabetic names:

>>> df.select(~cs.alpha())
shape: (3, 2)
┌─────┬────────┐
│ no1 ┆ t or f │
│ --- ┆ ---    │
│ i64 ┆ bool   │
╞═════╪════════╡
│ 100 ┆ true   │
│ 200 ┆ false  │
│ 300 ┆ null   │
└─────┴────────┘

>>> df.select(~cs.alpha(ignore_spaces=True))
shape: (3, 1)
┌─────┐
│ no1 │
│ --- │
│ i64 │
╞═════╡
│ 100 │
│ 200 │
│ 300 │
└─────┘

polars.selectors.alphanumeric(ascii_only: bool = False, *, ignore_spaces: bool = False) → SelectorType[source]#

Select all columns with alphanumeric names (eg: only letters and the digits 0-9).

Parameters:

ascii_only: Indicate whether to consider only ASCII alphabetic characters, or the full Unicode range of valid letters (accented, idiographic, etc).
ignore_spaces: Indicate whether to ignore the presence of spaces in column names; if so, only the other (non-space) characters are considered.

Notes

Matching column names cannot contain any non-alphabetic or integer characters. Note that the definition of “alphabetic” consists of all valid Unicode alphabetic characters (p{Alphabetic}) and digit characters (d) by default; this can be changed by setting ascii_only=True.

Examples

>>> import polars as pl
>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "1st_col": [100, 200, 300],
...         "flagged": [True, False, True],
...         "00prefix": ["01:aa", "02:bb", "03:cc"],
...         "last col": ["x", "y", "z"],
...     }
... )

Select columns with alphanumeric names:

>>> df.select(cs.alphanumeric())
shape: (3, 2)
┌─────────┬──────────┐
│ flagged ┆ 00prefix │
│ ---     ┆ ---      │
│ bool    ┆ str      │
╞═════════╪══════════╡
│ true    ┆ 01:aa    │
│ false   ┆ 02:bb    │
│ true    ┆ 03:cc    │
└─────────┴──────────┘

>>> df.select(cs.alphanumeric(ignore_spaces=True))
shape: (3, 3)
┌─────────┬──────────┬──────────┐
│ flagged ┆ 00prefix ┆ last col │
│ ---     ┆ ---      ┆ ---      │
│ bool    ┆ str      ┆ str      │
╞═════════╪══════════╪══════════╡
│ true    ┆ 01:aa    ┆ x        │
│ false   ┆ 02:bb    ┆ y        │
│ true    ┆ 03:cc    ┆ z        │
└─────────┴──────────┴──────────┘

Select all columns except for those with alphanumeric names:

>>> df.select(~cs.alphanumeric())
shape: (3, 2)
┌─────────┬──────────┐
│ 1st_col ┆ last col │
│ ---     ┆ ---      │
│ i64     ┆ str      │
╞═════════╪══════════╡
│ 100     ┆ x        │
│ 200     ┆ y        │
│ 300     ┆ z        │
└─────────┴──────────┘

>>> df.select(~cs.alphanumeric(ignore_spaces=True))
shape: (3, 1)
┌─────────┐
│ 1st_col │
│ ---     │
│ i64     │
╞═════════╡
│ 100     │
│ 200     │
│ 300     │
└─────────┘

polars.selectors.binary() → SelectorType[source]#

Select all binary columns.

See also

by_dtype: Select all columns matching the given dtype(s).
string: Select all string columns (optionally including categoricals).

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame({"a": [b"hello"], "b": ["world"], "c": [b"!"], "d": [":)"]})
>>> df
shape: (1, 4)
┌──────────┬───────┬────────┬─────┐
│ a        ┆ b     ┆ c      ┆ d   │
│ ---      ┆ ---   ┆ ---    ┆ --- │
│ binary   ┆ str   ┆ binary ┆ str │
╞══════════╪═══════╪════════╪═════╡
│ b"hello" ┆ world ┆ b"!"   ┆ :)  │
└──────────┴───────┴────────┴─────┘

Select binary columns and export as a dict:

>>> df.select(cs.binary()).to_dict(as_series=False)
{'a': [b'hello'], 'c': [b'!']}

Select all columns except for those that are binary:

>>> df.select(~cs.binary()).to_dict(as_series=False)
{'b': ['world'], 'd': [':)']}

polars.selectors.boolean() → SelectorType[source]#

Select all boolean columns.

See also

by_dtype: Select all columns matching the given dtype(s).

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame({"n": range(1, 5)}).with_columns(n_even=pl.col("n") % 2 == 0)
>>> df
shape: (4, 2)
┌─────┬────────┐
│ n   ┆ n_even │
│ --- ┆ ---    │
│ i64 ┆ bool   │
╞═════╪════════╡
│ 1   ┆ false  │
│ 2   ┆ true   │
│ 3   ┆ false  │
│ 4   ┆ true   │
└─────┴────────┘

Select and invert boolean columns:

>>> df.with_columns(is_odd=cs.boolean().not_())
shape: (4, 3)
┌─────┬────────┬────────┐
│ n   ┆ n_even ┆ is_odd │
│ --- ┆ ---    ┆ ---    │
│ i64 ┆ bool   ┆ bool   │
╞═════╪════════╪════════╡
│ 1   ┆ false  ┆ true   │
│ 2   ┆ true   ┆ false  │
│ 3   ┆ false  ┆ true   │
│ 4   ┆ true   ┆ false  │
└─────┴────────┴────────┘

Select all columns except for those that are boolean:

>>> df.select(~cs.boolean())
shape: (4, 1)
┌─────┐
│ n   │
│ --- │
│ i64 │
╞═════╡
│ 1   │
│ 2   │
│ 3   │
│ 4   │
└─────┘

polars.selectors.by_dtype(*dtypes: PolarsDataType | Collection[PolarsDataType]) → SelectorType[source]#

Select all columns matching the given dtypes.

See also

by_name: Select all columns matching the given names.
by_index: Select all columns matching the given indices.

Examples

>>> from datetime import date
>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "dt": [date(1999, 12, 31), date(2024, 1, 1), date(2010, 7, 5)],
...         "value": [1_234_500, 5_000_555, -4_500_000],
...         "other": ["foo", "bar", "foo"],
...     }
... )

Select all columns with date or string dtypes:

>>> df.select(cs.by_dtype(pl.Date, pl.String))
shape: (3, 2)
┌────────────┬───────┐
│ dt         ┆ other │
│ ---        ┆ ---   │
│ date       ┆ str   │
╞════════════╪═══════╡
│ 1999-12-31 ┆ foo   │
│ 2024-01-01 ┆ bar   │
│ 2010-07-05 ┆ foo   │
└────────────┴───────┘

Select all columns that are not of date or string dtype:

>>> df.select(~cs.by_dtype(pl.Date, pl.String))
shape: (3, 1)
┌──────────┐
│ value    │
│ ---      │
│ i64      │
╞══════════╡
│ 1234500  │
│ 5000555  │
│ -4500000 │
└──────────┘

Group by string columns and sum the numeric columns:

>>> df.group_by(cs.string()).agg(cs.numeric().sum()).sort(by="other")
shape: (2, 2)
┌───────┬──────────┐
│ other ┆ value    │
│ ---   ┆ ---      │
│ str   ┆ i64      │
╞═══════╪══════════╡
│ bar   ┆ 5000555  │
│ foo   ┆ -3265500 │
└───────┴──────────┘

polars.selectors.by_index(*indices: int | range | Sequence[int | range]) → SelectorType[source]#

Select all columns matching the given indices (or range objects).

Parameters:

*indices: One or more column indices (or range objects). Negative indexing is supported.

See also

by_dtype: Select all columns matching the given dtypes.
by_name: Select all columns matching the given names.

Notes

Matching columns are returned in the order in which their indexes appear in the selector, not the underlying schema order.

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "key": ["abc"],
...         **{f"c{i:02}": [0.5 * i] for i in range(100)},
...     },
... )
>>> print(df)
shape: (1, 101)
┌─────┬─────┬─────┬─────┬───┬──────┬──────┬──────┬──────┐
│ key ┆ c00 ┆ c01 ┆ c02 ┆ … ┆ c96  ┆ c97  ┆ c98  ┆ c99  │
│ --- ┆ --- ┆ --- ┆ --- ┆   ┆ ---  ┆ ---  ┆ ---  ┆ ---  │
│ str ┆ f64 ┆ f64 ┆ f64 ┆   ┆ f64  ┆ f64  ┆ f64  ┆ f64  │
╞═════╪═════╪═════╪═════╪═══╪══════╪══════╪══════╪══════╡
│ abc ┆ 0.0 ┆ 0.5 ┆ 1.0 ┆ … ┆ 48.0 ┆ 48.5 ┆ 49.0 ┆ 49.5 │
└─────┴─────┴─────┴─────┴───┴──────┴──────┴──────┴──────┘

Select columns by index (“key” column and the two first/last columns):

>>> df.select(cs.by_index(0, 1, 2, -2, -1))
shape: (1, 5)
┌─────┬─────┬─────┬──────┬──────┐
│ key ┆ c00 ┆ c01 ┆ c98  ┆ c99  │
│ --- ┆ --- ┆ --- ┆ ---  ┆ ---  │
│ str ┆ f64 ┆ f64 ┆ f64  ┆ f64  │
╞═════╪═════╪═════╪══════╪══════╡
│ abc ┆ 0.0 ┆ 0.5 ┆ 49.0 ┆ 49.5 │
└─────┴─────┴─────┴──────┴──────┘

Select the “key” column and use a range object to select various columns. Note that you can freely mix and match integer indices and range objects:

>>> df.select(cs.by_index(0, range(1, 101, 20)))
shape: (1, 6)
┌─────┬─────┬──────┬──────┬──────┬──────┐
│ key ┆ c00 ┆ c20  ┆ c40  ┆ c60  ┆ c80  │
│ --- ┆ --- ┆ ---  ┆ ---  ┆ ---  ┆ ---  │
│ str ┆ f64 ┆ f64  ┆ f64  ┆ f64  ┆ f64  │
╞═════╪═════╪══════╪══════╪══════╪══════╡
│ abc ┆ 0.0 ┆ 10.0 ┆ 20.0 ┆ 30.0 ┆ 40.0 │
└─────┴─────┴──────┴──────┴──────┴──────┘

>>> df.select(cs.by_index(0, range(101, 0, -25)))
shape: (1, 5)
┌─────┬──────┬──────┬──────┬─────┐
│ key ┆ c75  ┆ c50  ┆ c25  ┆ c00 │
│ --- ┆ ---  ┆ ---  ┆ ---  ┆ --- │
│ str ┆ f64  ┆ f64  ┆ f64  ┆ f64 │
╞═════╪══════╪══════╪══════╪═════╡
│ abc ┆ 37.5 ┆ 25.0 ┆ 12.5 ┆ 0.0 │
└─────┴──────┴──────┴──────┴─────┘

Select all columns except for the even-indexed ones:

>>> df.select(~cs.by_index(range(1, 100, 2)))
shape: (1, 51)
┌─────┬─────┬─────┬─────┬───┬──────┬──────┬──────┬──────┐
│ key ┆ c01 ┆ c03 ┆ c05 ┆ … ┆ c93  ┆ c95  ┆ c97  ┆ c99  │
│ --- ┆ --- ┆ --- ┆ --- ┆   ┆ ---  ┆ ---  ┆ ---  ┆ ---  │
│ str ┆ f64 ┆ f64 ┆ f64 ┆   ┆ f64  ┆ f64  ┆ f64  ┆ f64  │
╞═════╪═════╪═════╪═════╪═══╪══════╪══════╪══════╪══════╡
│ abc ┆ 0.5 ┆ 1.5 ┆ 2.5 ┆ … ┆ 46.5 ┆ 47.5 ┆ 48.5 ┆ 49.5 │
└─────┴─────┴─────┴─────┴───┴──────┴──────┴──────┴──────┘

polars.selectors.by_name(*names: str | Collection[str], require_all: bool = True) → SelectorType[source]#

Select all columns matching the given names.

Added in version 0.20.27: The require_all parameter was added.

Parameters:

*names: One or more names of columns to select.
require_all: Whether to match all names (the default) or any of the names.

See also

by_dtype: Select all columns matching the given dtypes.
by_index: Select all columns matching the given indices.

Notes

Matching columns are returned in the order in which they are declared in the selector, not the underlying schema order.

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "foo": ["x", "y"],
...         "bar": [123, 456],
...         "baz": [2.0, 5.5],
...         "zap": [False, True],
...     }
... )

Select columns by name:

>>> df.select(cs.by_name("foo", "bar"))
shape: (2, 2)
┌─────┬─────┐
│ foo ┆ bar │
│ --- ┆ --- │
│ str ┆ i64 │
╞═════╪═════╡
│ x   ┆ 123 │
│ y   ┆ 456 │
└─────┴─────┘

Match any of the given columns by name:

>>> df.select(cs.by_name("baz", "moose", "foo", "bear", require_all=False))
shape: (2, 2)
┌─────┬─────┐
│ foo ┆ baz │
│ --- ┆ --- │
│ str ┆ f64 │
╞═════╪═════╡
│ x   ┆ 2.0 │
│ y   ┆ 5.5 │
└─────┴─────┘

Match all columns except for those given:

>>> df.select(~cs.by_name("foo", "bar"))
shape: (2, 2)
┌─────┬───────┐
│ baz ┆ zap   │
│ --- ┆ ---   │
│ f64 ┆ bool  │
╞═════╪═══════╡
│ 2.0 ┆ false │
│ 5.5 ┆ true  │
└─────┴───────┘

polars.selectors.categorical() → SelectorType[source]#

Select all categorical columns.

See also

by_dtype: Select all columns matching the given dtype(s).
string: Select all string columns (optionally including categoricals).

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "foo": ["xx", "yy"],
...         "bar": [123, 456],
...         "baz": [2.0, 5.5],
...     },
...     schema_overrides={"foo": pl.Categorical},
... )

Select all categorical columns:

>>> df.select(cs.categorical())
shape: (2, 1)
┌─────┐
│ foo │
│ --- │
│ cat │
╞═════╡
│ xx  │
│ yy  │
└─────┘

Select all columns except for those that are categorical:

>>> df.select(~cs.categorical())
shape: (2, 2)
┌─────┬─────┐
│ bar ┆ baz │
│ --- ┆ --- │
│ i64 ┆ f64 │
╞═════╪═════╡
│ 123 ┆ 2.0 │
│ 456 ┆ 5.5 │
└─────┴─────┘

polars.selectors.contains(*substring: str) → SelectorType[source]#

Select columns whose names contain the given literal substring(s).

Parameters:

substring: Substring(s) that matching column names should contain.

See also

matches: Select all columns that match the given regex pattern.
ends_with: Select columns that end with the given substring(s).
starts_with: Select columns that start with the given substring(s).

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "foo": ["x", "y"],
...         "bar": [123, 456],
...         "baz": [2.0, 5.5],
...         "zap": [False, True],
...     }
... )

Select columns that contain the substring ‘ba’:

>>> df.select(cs.contains("ba"))
shape: (2, 2)
┌─────┬─────┐
│ bar ┆ baz │
│ --- ┆ --- │
│ i64 ┆ f64 │
╞═════╪═════╡
│ 123 ┆ 2.0 │
│ 456 ┆ 5.5 │
└─────┴─────┘

Select columns that contain the substring ‘ba’ or the letter ‘z’:

>>> df.select(cs.contains("ba", "z"))
shape: (2, 3)
┌─────┬─────┬───────┐
│ bar ┆ baz ┆ zap   │
│ --- ┆ --- ┆ ---   │
│ i64 ┆ f64 ┆ bool  │
╞═════╪═════╪═══════╡
│ 123 ┆ 2.0 ┆ false │
│ 456 ┆ 5.5 ┆ true  │
└─────┴─────┴───────┘

Select all columns except for those that contain the substring ‘ba’:

>>> df.select(~cs.contains("ba"))
shape: (2, 2)
┌─────┬───────┐
│ foo ┆ zap   │
│ --- ┆ ---   │
│ str ┆ bool  │
╞═════╪═══════╡
│ x   ┆ false │
│ y   ┆ true  │
└─────┴───────┘

polars.selectors.date() → SelectorType[source]#

Select all date columns.

See also

datetime: Select all datetime columns, optionally filtering by time unit/zone.
duration: Select all duration columns, optionally filtering by time unit.
temporal: Select all temporal columns.
time: Select all time columns.

Examples

>>> from datetime import date, datetime, time
>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "dtm": [datetime(2001, 5, 7, 10, 25), datetime(2031, 12, 31, 0, 30)],
...         "dt": [date(1999, 12, 31), date(2024, 8, 9)],
...         "tm": [time(0, 0, 0), time(23, 59, 59)],
...     },
... )

Select all date columns:

>>> df.select(cs.date())
shape: (2, 1)
┌────────────┐
│ dt         │
│ ---        │
│ date       │
╞════════════╡
│ 1999-12-31 │
│ 2024-08-09 │
└────────────┘

Select all columns except for those that are dates:

>>> df.select(~cs.date())
shape: (2, 2)
┌─────────────────────┬──────────┐
│ dtm                 ┆ tm       │
│ ---                 ┆ ---      │
│ datetime[μs]        ┆ time     │
╞═════════════════════╪══════════╡
│ 2001-05-07 10:25:00 ┆ 00:00:00 │
│ 2031-12-31 00:30:00 ┆ 23:59:59 │
└─────────────────────┴──────────┘

Select all datetime columns, optionally filtering by time unit/zone.

Parameters:

time_unit

One (or more) of the allowed timeunit precision strings, “ms”, “us”, and “ns”. Omit to select columns with any valid timeunit.

time_zone

One or more timezone strings, as defined in zoneinfo (to see valid options run import zoneinfo; zoneinfo.available_timezones() for a full list).
Set None to select Datetime columns that do not have a timezone.
Set “*” to select Datetime columns that have any timezone.

See also

date: Select all date columns.
duration: Select all duration columns, optionally filtering by time unit.
temporal: Select all temporal columns.
time: Select all time columns.

Examples

>>> from datetime import datetime, date, timezone
>>> import polars.selectors as cs
>>> from zoneinfo import ZoneInfo
>>> tokyo_tz = ZoneInfo("Asia/Tokyo")
>>> utc_tz = timezone.utc
>>> df = pl.DataFrame(
...     {
...         "tstamp_tokyo": [
...             datetime(1999, 7, 21, 5, 20, 16, 987654, tzinfo=tokyo_tz),
...             datetime(2000, 5, 16, 6, 21, 21, 123465, tzinfo=tokyo_tz),
...         ],
...         "tstamp_utc": [
...             datetime(2023, 4, 10, 12, 14, 16, 999000, tzinfo=utc_tz),
...             datetime(2025, 8, 25, 14, 18, 22, 666000, tzinfo=utc_tz),
...         ],
...         "tstamp": [
...             datetime(2000, 11, 20, 18, 12, 16, 600000),
...             datetime(2020, 10, 30, 10, 20, 25, 123000),
...         ],
...         "dt": [date(1999, 12, 31), date(2010, 7, 5)],
...     },
...     schema_overrides={
...         "tstamp_tokyo": pl.Datetime("ns", "Asia/Tokyo"),
...         "tstamp_utc": pl.Datetime("us", "UTC"),
...     },
... )

Select all datetime columns:

>>> df.select(cs.datetime())
shape: (2, 3)
┌────────────────────────────────┬─────────────────────────────┬─────────────────────────┐
│ tstamp_tokyo                   ┆ tstamp_utc                  ┆ tstamp                  │
│ ---                            ┆ ---                         ┆ ---                     │
│ datetime[ns, Asia/Tokyo]       ┆ datetime[μs, UTC]           ┆ datetime[μs]            │
╞════════════════════════════════╪═════════════════════════════╪═════════════════════════╡
│ 1999-07-21 05:20:16.987654 JST ┆ 2023-04-10 12:14:16.999 UTC ┆ 2000-11-20 18:12:16.600 │
│ 2000-05-16 06:21:21.123465 JST ┆ 2025-08-25 14:18:22.666 UTC ┆ 2020-10-30 10:20:25.123 │
└────────────────────────────────┴─────────────────────────────┴─────────────────────────┘

Select all datetime columns that have ‘us’ precision:

>>> df.select(cs.datetime("us"))
shape: (2, 2)
┌─────────────────────────────┬─────────────────────────┐
│ tstamp_utc                  ┆ tstamp                  │
│ ---                         ┆ ---                     │
│ datetime[μs, UTC]           ┆ datetime[μs]            │
╞═════════════════════════════╪═════════════════════════╡
│ 2023-04-10 12:14:16.999 UTC ┆ 2000-11-20 18:12:16.600 │
│ 2025-08-25 14:18:22.666 UTC ┆ 2020-10-30 10:20:25.123 │
└─────────────────────────────┴─────────────────────────┘

Select all datetime columns that have any timezone:

>>> df.select(cs.datetime(time_zone="*"))
shape: (2, 2)
┌────────────────────────────────┬─────────────────────────────┐
│ tstamp_tokyo                   ┆ tstamp_utc                  │
│ ---                            ┆ ---                         │
│ datetime[ns, Asia/Tokyo]       ┆ datetime[μs, UTC]           │
╞════════════════════════════════╪═════════════════════════════╡
│ 1999-07-21 05:20:16.987654 JST ┆ 2023-04-10 12:14:16.999 UTC │
│ 2000-05-16 06:21:21.123465 JST ┆ 2025-08-25 14:18:22.666 UTC │
└────────────────────────────────┴─────────────────────────────┘

Select all datetime columns that have a specific timezone:

>>> df.select(cs.datetime(time_zone="UTC"))
shape: (2, 1)
┌─────────────────────────────┐
│ tstamp_utc                  │
│ ---                         │
│ datetime[μs, UTC]           │
╞═════════════════════════════╡
│ 2023-04-10 12:14:16.999 UTC │
│ 2025-08-25 14:18:22.666 UTC │
└─────────────────────────────┘

Select all datetime columns that have NO timezone:

>>> df.select(cs.datetime(time_zone=None))
shape: (2, 1)
┌─────────────────────────┐
│ tstamp                  │
│ ---                     │
│ datetime[μs]            │
╞═════════════════════════╡
│ 2000-11-20 18:12:16.600 │
│ 2020-10-30 10:20:25.123 │
└─────────────────────────┘

Select all columns except for datetime columns:

>>> df.select(~cs.datetime())
shape: (2, 1)
┌────────────┐
│ dt         │
│ ---        │
│ date       │
╞════════════╡
│ 1999-12-31 │
│ 2010-07-05 │
└────────────┘

polars.selectors.decimal() → SelectorType[source]#

Select all decimal columns.

See also

float: Select all float columns.
integer: Select all integer columns.
numeric: Select all numeric columns.

Examples

>>> from decimal import Decimal as D
>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "foo": ["x", "y"],
...         "bar": [D(123), D(456)],
...         "baz": [D("2.0005"), D("-50.5555")],
...     },
...     schema_overrides={"baz": pl.Decimal(scale=5, precision=10)},
... )

Select all decimal columns:

>>> df.select(cs.decimal())
shape: (2, 2)
┌──────────────┬───────────────┐
│ bar          ┆ baz           │
│ ---          ┆ ---           │
│ decimal[*,0] ┆ decimal[10,5] │
╞══════════════╪═══════════════╡
│ 123          ┆ 2.00050       │
│ 456          ┆ -50.55550     │
└──────────────┴───────────────┘

Select all columns except the decimal ones:

>>> df.select(~cs.decimal())
shape: (2, 1)
┌─────┐
│ foo │
│ --- │
│ str │
╞═════╡
│ x   │
│ y   │
└─────┘

polars.selectors.digit(ascii_only: bool = False) → SelectorType[source]#

Select all columns having names consisting only of digits.

Notes

Matching column names cannot contain any non-digit characters. Note that the definition of “digit” consists of all valid Unicode digit characters (d) by default; this can be changed by setting ascii_only=True.

Examples

>>> import polars as pl
>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "key": ["aaa", "bbb", "aaa", "bbb", "bbb"],
...         "year": [2001, 2001, 2025, 2025, 2001],
...         "value": [-25, 100, 75, -15, -5],
...     }
... ).pivot(
...     values="value",
...     index="key",
...     on="year",
...     aggregate_function="sum",
... )
>>> print(df)
shape: (2, 3)
┌─────┬──────┬──────┐
│ key ┆ 2001 ┆ 2025 │
│ --- ┆ ---  ┆ ---  │
│ str ┆ i64  ┆ i64  │
╞═════╪══════╪══════╡
│ aaa ┆ -25  ┆ 75   │
│ bbb ┆ 95   ┆ -15  │
└─────┴──────┴──────┘

Select columns with digit names:

>>> df.select(cs.digit())
shape: (2, 2)
┌──────┬──────┐
│ 2001 ┆ 2025 │
│ ---  ┆ ---  │
│ i64  ┆ i64  │
╞══════╪══════╡
│ -25  ┆ 75   │
│ 95   ┆ -15  │
└──────┴──────┘

Select all columns except for those with digit names:

>>> df.select(~cs.digit())
shape: (2, 1)
┌─────┐
│ key │
│ --- │
│ str │
╞═════╡
│ aaa │
│ bbb │
└─────┘

Demonstrate use of ascii_only flag (by default all valid unicode digits are considered, but this can be constrained to ascii 0-9):

>>> df = pl.DataFrame({"१९९९": [1999], "२०७७": [2077], "3000": [3000]})
>>> df.select(cs.digit())
shape: (1, 3)
┌──────┬──────┬──────┐
│ १९९९ ┆ २०७७ ┆ 3000 │
│ ---  ┆ ---  ┆ ---  │
│ i64  ┆ i64  ┆ i64  │
╞══════╪══════╪══════╡
│ 1999 ┆ 2077 ┆ 3000 │
└──────┴──────┴──────┘

>>> df.select(cs.digit(ascii_only=True))
shape: (1, 1)
┌──────┐
│ 3000 │
│ ---  │
│ i64  │
╞══════╡
│ 3000 │
└──────┘

polars.selectors.duration(time_unit: TimeUnit | Collection[TimeUnit] | None = None) → SelectorType[source]#

Select all duration columns, optionally filtering by time unit.

Parameters:

time_unit: One (or more) of the allowed timeunit precision strings, “ms”, “us”, and “ns”. Omit to select columns with any valid timeunit.

See also

date: Select all date columns.
datetime: Select all datetime columns, optionally filtering by time unit/zone.
temporal: Select all temporal columns.
time: Select all time columns.

Examples

>>> from datetime import date, timedelta
>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "dt": [date(2022, 1, 31), date(2025, 7, 5)],
...         "td1": [
...             timedelta(days=1, milliseconds=123456),
...             timedelta(days=1, hours=23, microseconds=987000),
...         ],
...         "td2": [
...             timedelta(days=7, microseconds=456789),
...             timedelta(days=14, minutes=999, seconds=59),
...         ],
...         "td3": [
...             timedelta(weeks=4, days=-10, microseconds=999999),
...             timedelta(weeks=3, milliseconds=123456, microseconds=1),
...         ],
...     },
...     schema_overrides={
...         "td1": pl.Duration("ms"),
...         "td2": pl.Duration("us"),
...         "td3": pl.Duration("ns"),
...     },
... )

Select all duration columns:

>>> df.select(cs.duration())
shape: (2, 3)
┌────────────────┬─────────────────┬────────────────────┐
│ td1            ┆ td2             ┆ td3                │
│ ---            ┆ ---             ┆ ---                │
│ duration[ms]   ┆ duration[μs]    ┆ duration[ns]       │
╞════════════════╪═════════════════╪════════════════════╡
│ 1d 2m 3s 456ms ┆ 7d 456789µs     ┆ 18d 999999µs       │
│ 1d 23h 987ms   ┆ 14d 16h 39m 59s ┆ 21d 2m 3s 456001µs │
└────────────────┴─────────────────┴────────────────────┘

Select all duration columns that have ‘ms’ precision:

>>> df.select(cs.duration("ms"))
shape: (2, 1)
┌────────────────┐
│ td1            │
│ ---            │
│ duration[ms]   │
╞════════════════╡
│ 1d 2m 3s 456ms │
│ 1d 23h 987ms   │
└────────────────┘

Select all duration columns that have ‘ms’ OR ‘ns’ precision:

>>> df.select(cs.duration(["ms", "ns"]))
shape: (2, 2)
┌────────────────┬────────────────────┐
│ td1            ┆ td3                │
│ ---            ┆ ---                │
│ duration[ms]   ┆ duration[ns]       │
╞════════════════╪════════════════════╡
│ 1d 2m 3s 456ms ┆ 18d 999999µs       │
│ 1d 23h 987ms   ┆ 21d 2m 3s 456001µs │
└────────────────┴────────────────────┘

Select all columns except for duration columns:

>>> df.select(~cs.duration())
shape: (2, 1)
┌────────────┐
│ dt         │
│ ---        │
│ date       │
╞════════════╡
│ 2022-01-31 │
│ 2025-07-05 │
└────────────┘

polars.selectors.ends_with(*suffix: str) → SelectorType[source]#

Select columns that end with the given substring(s).

Parameters:

suffix: Substring(s) that matching column names should end with.

See also

contains: Select columns that contain the given literal substring(s).
matches: Select all columns that match the given regex pattern.
starts_with: Select columns that start with the given substring(s).

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "foo": ["x", "y"],
...         "bar": [123, 456],
...         "baz": [2.0, 5.5],
...         "zap": [False, True],
...     }
... )

Select columns that end with the substring ‘z’:

>>> df.select(cs.ends_with("z"))
shape: (2, 1)
┌─────┐
│ baz │
│ --- │
│ f64 │
╞═════╡
│ 2.0 │
│ 5.5 │
└─────┘

Select columns that end with either the letter ‘z’ or ‘r’:

>>> df.select(cs.ends_with("z", "r"))
shape: (2, 2)
┌─────┬─────┐
│ bar ┆ baz │
│ --- ┆ --- │
│ i64 ┆ f64 │
╞═════╪═════╡
│ 123 ┆ 2.0 │
│ 456 ┆ 5.5 │
└─────┴─────┘

Select all columns except for those that end with the substring ‘z’:

>>> df.select(~cs.ends_with("z"))
shape: (2, 3)
┌─────┬─────┬───────┐
│ foo ┆ bar ┆ zap   │
│ --- ┆ --- ┆ ---   │
│ str ┆ i64 ┆ bool  │
╞═════╪═════╪═══════╡
│ x   ┆ 123 ┆ false │
│ y   ┆ 456 ┆ true  │
└─────┴─────┴───────┘

Select all columns except those matching the given columns, datatypes, or selectors.

Parameters:

columns: One or more columns (col or name), datatypes, columns, or selectors representing the columns to exclude.
*more_columns: Additional columns, datatypes, or selectors to exclude, specified as positional arguments.

Notes

If excluding a single selector it is simpler to write as ~selector instead.

Examples

Exclude by column name(s):

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "aa": [1, 2, 3],
...         "ba": ["a", "b", None],
...         "cc": [None, 2.5, 1.5],
...     }
... )
>>> df.select(cs.exclude("ba", "xx"))
shape: (3, 2)
┌─────┬──────┐
│ aa  ┆ cc   │
│ --- ┆ ---  │
│ i64 ┆ f64  │
╞═════╪══════╡
│ 1   ┆ null │
│ 2   ┆ 2.5  │
│ 3   ┆ 1.5  │
└─────┴──────┘

Exclude using a column name, a selector, and a dtype:

>>> df.select(cs.exclude("aa", cs.string(), pl.UInt32))
shape: (3, 1)
┌──────┐
│ cc   │
│ ---  │
│ f64  │
╞══════╡
│ null │
│ 2.5  │
│ 1.5  │
└──────┘

polars.selectors.expand_selector( target: DataFrame | LazyFrame | Mapping[str, PolarsDataType], selector: SelectorType | Expr, *, strict: bool = True, ) → tuple[str, ...][source]#

Expand selector to column names, with respect to a specific frame or target schema.

Added in version 0.20.30: The strict parameter was added.

Parameters:

target: A polars DataFrame, LazyFrame or schema.
selector: An arbitrary polars selector (or compound selector).
strict: Setting False will additionally allow for a broader range of column selection expressions (such as bare columns or use of .exclude()) to be expanded, not just the dedicated selectors.

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "colx": ["a", "b", "c"],
...         "coly": [123, 456, 789],
...         "colz": [2.0, 5.5, 8.0],
...     }
... )

Expand selector with respect to an existing DataFrame:

>>> cs.expand_selector(df, cs.numeric())
('coly', 'colz')
>>> cs.expand_selector(df, cs.first() | cs.last())
('colx', 'colz')

This also works with LazyFrame:

>>> cs.expand_selector(df.lazy(), ~(cs.first() | cs.last()))
('coly',)

Expand selector with respect to a standalone schema:

>>> schema = {
...     "id": pl.Int64,
...     "desc": pl.String,
...     "count": pl.UInt32,
...     "value": pl.Float64,
... }
>>> cs.expand_selector(schema, cs.string() | cs.float())
('desc', 'value')

Allow for non-strict selection expressions (such as those including use of an .exclude() constraint) to be expanded:

>>> cs.expand_selector(schema, cs.numeric().exclude("id"), strict=False)
('count', 'value')

polars.selectors.first() → SelectorType[source]#

Select the first column in the current scope.

See also

all: Select all columns.
last: Select the last column in the current scope.

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "foo": ["x", "y"],
...         "bar": [123, 456],
...         "baz": [2.0, 5.5],
...         "zap": [0, 1],
...     }
... )

Select the first column:

>>> df.select(cs.first())
shape: (2, 1)
┌─────┐
│ foo │
│ --- │
│ str │
╞═════╡
│ x   │
│ y   │
└─────┘

Select everything except for the first column:

>>> df.select(~cs.first())
shape: (2, 3)
┌─────┬─────┬─────┐
│ bar ┆ baz ┆ zap │
│ --- ┆ --- ┆ --- │
│ i64 ┆ f64 ┆ i64 │
╞═════╪═════╪═════╡
│ 123 ┆ 2.0 ┆ 0   │
│ 456 ┆ 5.5 ┆ 1   │
└─────┴─────┴─────┘

polars.selectors.float() → SelectorType[source]#

Select all float columns.

See also

integer: Select all integer columns.
numeric: Select all numeric columns.
signed_integer: Select all signed integer columns.
unsigned_integer: Select all unsigned integer columns.

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "foo": ["x", "y"],
...         "bar": [123, 456],
...         "baz": [2.0, 5.5],
...         "zap": [0.0, 1.0],
...     },
...     schema_overrides={"baz": pl.Float32, "zap": pl.Float64},
... )

Select all float columns:

>>> df.select(cs.float())
shape: (2, 2)
┌─────┬─────┐
│ baz ┆ zap │
│ --- ┆ --- │
│ f32 ┆ f64 │
╞═════╪═════╡
│ 2.0 ┆ 0.0 │
│ 5.5 ┆ 1.0 │
└─────┴─────┘

Select all columns except for those that are float:

>>> df.select(~cs.float())
shape: (2, 2)
┌─────┬─────┐
│ foo ┆ bar │
│ --- ┆ --- │
│ str ┆ i64 │
╞═════╪═════╡
│ x   ┆ 123 │
│ y   ┆ 456 │
└─────┴─────┘

polars.selectors.integer() → SelectorType[source]#

Select all integer columns.

See also

by_dtype: Select columns by dtype.
float: Select all float columns.
numeric: Select all numeric columns.
signed_integer: Select all signed integer columns.
unsigned_integer: Select all unsigned integer columns.

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "foo": ["x", "y"],
...         "bar": [123, 456],
...         "baz": [2.0, 5.5],
...         "zap": [0, 1],
...     }
... )

Select all integer columns:

>>> df.select(cs.integer())
shape: (2, 2)
┌─────┬─────┐
│ bar ┆ zap │
│ --- ┆ --- │
│ i64 ┆ i64 │
╞═════╪═════╡
│ 123 ┆ 0   │
│ 456 ┆ 1   │
└─────┴─────┘

Select all columns except for those that are integer :

>>> df.select(~cs.integer())
shape: (2, 2)
┌─────┬─────┐
│ foo ┆ baz │
│ --- ┆ --- │
│ str ┆ f64 │
╞═════╪═════╡
│ x   ┆ 2.0 │
│ y   ┆ 5.5 │
└─────┴─────┘

polars.selectors.is_selector(obj: Any) → bool[source]#

Indicate whether the given object/expression is a selector.

Examples

>>> from polars.selectors import is_selector
>>> import polars.selectors as cs
>>> is_selector(pl.col("colx"))
False
>>> is_selector(cs.first() | cs.last())
True

polars.selectors.last() → SelectorType[source]#

Select the last column in the current scope.

See also

all: Select all columns.
first: Select the first column in the current scope.

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "foo": ["x", "y"],
...         "bar": [123, 456],
...         "baz": [2.0, 5.5],
...         "zap": [0, 1],
...     }
... )

Select the last column:

>>> df.select(cs.last())
shape: (2, 1)
┌─────┐
│ zap │
│ --- │
│ i64 │
╞═════╡
│ 0   │
│ 1   │
└─────┘

Select everything except for the last column:

>>> df.select(~cs.last())
shape: (2, 3)
┌─────┬─────┬─────┐
│ foo ┆ bar ┆ baz │
│ --- ┆ --- ┆ --- │
│ str ┆ i64 ┆ f64 │
╞═════╪═════╪═════╡
│ x   ┆ 123 ┆ 2.0 │
│ y   ┆ 456 ┆ 5.5 │
└─────┴─────┴─────┘

polars.selectors.matches(pattern: str) → SelectorType[source]#

Select all columns that match the given regex pattern.

Parameters:

pattern: A valid regular expression pattern, compatible with the regex crate.

See also

contains: Select all columns that contain the given substring.
ends_with: Select all columns that end with the given substring(s).
starts_with: Select all columns that start with the given substring(s).

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "foo": ["x", "y"],
...         "bar": [123, 456],
...         "baz": [2.0, 5.5],
...         "zap": [0, 1],
...     }
... )

Match column names containing an ‘a’, preceded by a character that is not ‘z’:

>>> df.select(cs.matches("[^z]a"))
shape: (2, 2)
┌─────┬─────┐
│ bar ┆ baz │
│ --- ┆ --- │
│ i64 ┆ f64 │
╞═════╪═════╡
│ 123 ┆ 2.0 │
│ 456 ┆ 5.5 │
└─────┴─────┘

Do not match column names ending in ‘R’ or ‘z’ (case-insensitively):

>>> df.select(~cs.matches(r"(?i)R|z$"))
shape: (2, 2)
┌─────┬─────┐
│ foo ┆ zap │
│ --- ┆ --- │
│ str ┆ i64 │
╞═════╪═════╡
│ x   ┆ 0   │
│ y   ┆ 1   │
└─────┴─────┘

polars.selectors.numeric() → SelectorType[source]#

Select all numeric columns.

See also

by_dtype: Select columns by dtype.
float: Select all float columns.
integer: Select all integer columns.
signed_integer: Select all signed integer columns.
unsigned_integer: Select all unsigned integer columns.

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "foo": ["x", "y"],
...         "bar": [123, 456],
...         "baz": [2.0, 5.5],
...         "zap": [0, 0],
...     },
...     schema_overrides={"bar": pl.Int16, "baz": pl.Float32, "zap": pl.UInt8},
... )

Match all numeric columns:

>>> df.select(cs.numeric())
shape: (2, 3)
┌─────┬─────┬─────┐
│ bar ┆ baz ┆ zap │
│ --- ┆ --- ┆ --- │
│ i16 ┆ f32 ┆ u8  │
╞═════╪═════╪═════╡
│ 123 ┆ 2.0 ┆ 0   │
│ 456 ┆ 5.5 ┆ 0   │
└─────┴─────┴─────┘

Match all columns except for those that are numeric:

>>> df.select(~cs.numeric())
shape: (2, 1)
┌─────┐
│ foo │
│ --- │
│ str │
╞═════╡
│ x   │
│ y   │
└─────┘

polars.selectors.signed_integer() → SelectorType[source]#

Select all signed integer columns.

See also

by_dtype: Select columns by dtype.
float: Select all float columns.
integer: Select all integer columns.
numeric: Select all numeric columns.
unsigned_integer: Select all unsigned integer columns.

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "foo": [-123, -456],
...         "bar": [3456, 6789],
...         "baz": [7654, 4321],
...         "zap": ["ab", "cd"],
...     },
...     schema_overrides={"bar": pl.UInt32, "baz": pl.UInt64},
... )

Select all signed integer columns:

>>> df.select(cs.signed_integer())
shape: (2, 1)
┌──────┐
│ foo  │
│ ---  │
│ i64  │
╞══════╡
│ -123 │
│ -456 │
└──────┘

>>> df.select(~cs.signed_integer())
shape: (2, 3)
┌──────┬──────┬─────┐
│ bar  ┆ baz  ┆ zap │
│ ---  ┆ ---  ┆ --- │
│ u32  ┆ u64  ┆ str │
╞══════╪══════╪═════╡
│ 3456 ┆ 7654 ┆ ab  │
│ 6789 ┆ 4321 ┆ cd  │
└──────┴──────┴─────┘

Select all integer columns (both signed and unsigned):

>>> df.select(cs.integer())
shape: (2, 3)
┌──────┬──────┬──────┐
│ foo  ┆ bar  ┆ baz  │
│ ---  ┆ ---  ┆ ---  │
│ i64  ┆ u32  ┆ u64  │
╞══════╪══════╪══════╡
│ -123 ┆ 3456 ┆ 7654 │
│ -456 ┆ 6789 ┆ 4321 │
└──────┴──────┴──────┘

polars.selectors.starts_with(*prefix: str) → SelectorType[source]#

Select columns that start with the given substring(s).

Parameters:

prefix: Substring(s) that matching column names should start with.

See also

contains: Select all columns that contain the given substring.
ends_with: Select all columns that end with the given substring(s).
matches: Select all columns that match the given regex pattern.

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "foo": [1.0, 2.0],
...         "bar": [3.0, 4.0],
...         "baz": [5, 6],
...         "zap": [7, 8],
...     }
... )

Match columns starting with a ‘b’:

>>> df.select(cs.starts_with("b"))
shape: (2, 2)
┌─────┬─────┐
│ bar ┆ baz │
│ --- ┆ --- │
│ f64 ┆ i64 │
╞═════╪═════╡
│ 3.0 ┆ 5   │
│ 4.0 ┆ 6   │
└─────┴─────┘

Match columns starting with either the letter ‘b’ or ‘z’:

>>> df.select(cs.starts_with("b", "z"))
shape: (2, 3)
┌─────┬─────┬─────┐
│ bar ┆ baz ┆ zap │
│ --- ┆ --- ┆ --- │
│ f64 ┆ i64 ┆ i64 │
╞═════╪═════╪═════╡
│ 3.0 ┆ 5   ┆ 7   │
│ 4.0 ┆ 6   ┆ 8   │
└─────┴─────┴─────┘

Match all columns except for those starting with ‘b’:

>>> df.select(~cs.starts_with("b"))
shape: (2, 2)
┌─────┬─────┐
│ foo ┆ zap │
│ --- ┆ --- │
│ f64 ┆ i64 │
╞═════╪═════╡
│ 1.0 ┆ 7   │
│ 2.0 ┆ 8   │
└─────┴─────┘

polars.selectors.string(*, include_categorical: bool = False) → SelectorType[source]#

Select all String (and, optionally, Categorical) string columns.

See also

binary: Select all binary columns.
by_dtype: Select all columns matching the given dtype(s).
categorical: Select all categorical columns.

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "w": ["xx", "yy", "xx", "yy", "xx"],
...         "x": [1, 2, 1, 4, -2],
...         "y": [3.0, 4.5, 1.0, 2.5, -2.0],
...         "z": ["a", "b", "a", "b", "b"],
...     },
... ).with_columns(
...     z=pl.col("z").cast(pl.Categorical("lexical")),
... )

Group by all string columns, sum the numeric columns, then sort by the string cols:

>>> df.group_by(cs.string()).agg(cs.numeric().sum()).sort(by=cs.string())
shape: (2, 3)
┌─────┬─────┬─────┐
│ w   ┆ x   ┆ y   │
│ --- ┆ --- ┆ --- │
│ str ┆ i64 ┆ f64 │
╞═════╪═════╪═════╡
│ xx  ┆ 0   ┆ 2.0 │
│ yy  ┆ 6   ┆ 7.0 │
└─────┴─────┴─────┘

Group by all string and categorical columns:

>>> df.group_by(cs.string(include_categorical=True)).agg(cs.numeric().sum()).sort(
...     by=cs.string(include_categorical=True)
... )
shape: (3, 4)
┌─────┬─────┬─────┬──────┐
│ w   ┆ z   ┆ x   ┆ y    │
│ --- ┆ --- ┆ --- ┆ ---  │
│ str ┆ cat ┆ i64 ┆ f64  │
╞═════╪═════╪═════╪══════╡
│ xx  ┆ a   ┆ 2   ┆ 4.0  │
│ xx  ┆ b   ┆ -2  ┆ -2.0 │
│ yy  ┆ b   ┆ 6   ┆ 7.0  │
└─────┴─────┴─────┴──────┘

polars.selectors.temporal() → SelectorType[source]#

Select all temporal columns.

See also

by_dtype: Select all columns matching the given dtype(s).
date: Select all date columns.
datetime: Select all datetime columns, optionally filtering by time unit/zone.
duration: Select all duration columns, optionally filtering by time unit.
time: Select all time columns.

Examples

>>> from datetime import date, time
>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "dt": [date(2021, 1, 1), date(2021, 1, 2)],
...         "tm": [time(12, 0, 0), time(20, 30, 45)],
...         "value": [1.2345, 2.3456],
...     }
... )

Match all temporal columns:

>>> df.select(cs.temporal())
shape: (2, 2)
┌────────────┬──────────┐
│ dt         ┆ tm       │
│ ---        ┆ ---      │
│ date       ┆ time     │
╞════════════╪══════════╡
│ 2021-01-01 ┆ 12:00:00 │
│ 2021-01-02 ┆ 20:30:45 │
└────────────┴──────────┘

Match all temporal columns except for time columns:

>>> df.select(cs.temporal() - cs.time())
shape: (2, 1)
┌────────────┐
│ dt         │
│ ---        │
│ date       │
╞════════════╡
│ 2021-01-01 │
│ 2021-01-02 │
└────────────┘

Match all columns except for temporal columns:

>>> df.select(~cs.temporal())
shape: (2, 1)
┌────────┐
│ value  │
│ ---    │
│ f64    │
╞════════╡
│ 1.2345 │
│ 2.3456 │
└────────┘

polars.selectors.time() → SelectorType[source]#

Select all time columns.

See also

date: Select all date columns.
datetime: Select all datetime columns, optionally filtering by time unit/zone.
duration: Select all duration columns, optionally filtering by time unit.
temporal: Select all temporal columns.

Examples

>>> from datetime import date, datetime, time
>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "dtm": [datetime(2001, 5, 7, 10, 25), datetime(2031, 12, 31, 0, 30)],
...         "dt": [date(1999, 12, 31), date(2024, 8, 9)],
...         "tm": [time(0, 0, 0), time(23, 59, 59)],
...     },
... )

Select all time columns:

>>> df.select(cs.time())
shape: (2, 1)
┌──────────┐
│ tm       │
│ ---      │
│ time     │
╞══════════╡
│ 00:00:00 │
│ 23:59:59 │
└──────────┘

Select all columns except for those that are times:

>>> df.select(~cs.time())
shape: (2, 2)
┌─────────────────────┬────────────┐
│ dtm                 ┆ dt         │
│ ---                 ┆ ---        │
│ datetime[μs]        ┆ date       │
╞═════════════════════╪════════════╡
│ 2001-05-07 10:25:00 ┆ 1999-12-31 │
│ 2031-12-31 00:30:00 ┆ 2024-08-09 │
└─────────────────────┴────────────┘

polars.selectors.unsigned_integer() → SelectorType[source]#

Select all unsigned integer columns.

See also

by_dtype: Select columns by dtype.
float: Select all float columns.
integer: Select all integer columns.
numeric: Select all numeric columns.
signed_integer: Select all signed integer columns.

Examples

>>> import polars.selectors as cs
>>> df = pl.DataFrame(
...     {
...         "foo": [-123, -456],
...         "bar": [3456, 6789],
...         "baz": [7654, 4321],
...         "zap": ["ab", "cd"],
...     },
...     schema_overrides={"bar": pl.UInt32, "baz": pl.UInt64},
... )

Select all unsigned integer columns:

>>> df.select(cs.unsigned_integer())
shape: (2, 2)
┌──────┬──────┐
│ bar  ┆ baz  │
│ ---  ┆ ---  │
│ u32  ┆ u64  │
╞══════╪══════╡
│ 3456 ┆ 7654 │
│ 6789 ┆ 4321 │
└──────┴──────┘

Select all columns except for those that are unsigned integers:

>>> df.select(~cs.unsigned_integer())
shape: (2, 2)
┌──────┬─────┐
│ foo  ┆ zap │
│ ---  ┆ --- │
│ i64  ┆ str │
╞══════╪═════╡
│ -123 ┆ ab  │
│ -456 ┆ cd  │
└──────┴─────┘

Select all integer columns (both signed and unsigned):

>>> df.select(cs.integer())
shape: (2, 3)
┌──────┬──────┬──────┐
│ foo  ┆ bar  ┆ baz  │
│ ---  ┆ ---  ┆ ---  │
│ i64  ┆ u32  ┆ u64  │
╞══════╪══════╪══════╡
│ -123 ┆ 3456 ┆ 7654 │
│ -456 ┆ 6789 ┆ 4321 │
└──────┴──────┴──────┘