Series — Polars documentation (original) (raw)
This page gives an overview of all public Series methods.
class polars.Series(
name: str | ArrayLike | None = None,
values: ArrayLike | None = None,
dtype: PolarsDataType | None = None,
*,
strict: bool = True,
nan_to_null: bool = False,
A Series represents a single column in a Polars DataFrame.
Parameters:
namestr, default None
Name of the Series. Will be used as a column name when used in a DataFrame. When not specified, name is set to an empty string.
valuesArrayLike, default None
One-dimensional data in various forms. Supported are: Sequence, Series, pyarrow Array, and numpy ndarray.
dtypeDataType, default None
Data type of the resulting Series. If set to None
(default), the data type is inferred from the values
input. The strategy for data type inference depends on the strict
parameter:
- If
strict
is set to True (default), the inferred data type is equal to the first non-null value, orNull
if all values are null. - If
strict
is set to False, the inferred data type is the supertype of the values, orObject
if no supertype can be found. WARNING: A full pass over the values is required to determine the supertype. - If no values were passed, the resulting data type is
Null
.
strictbool, default True
Throw an error if any value does not exactly match the given or inferred data type. If set to False
, values that do not match the data type are cast to that data type or, if casting is not possible, set to null instead.
nan_to_nullbool, default False
In case a numpy array is used to create this Series, indicate how to deal with np.nan values. (This parameter is a no-op on non-numpy data).
Examples
Constructing a Series by specifying name and values positionally:
s = pl.Series("a", [1, 2, 3]) s shape: (3,) Series: 'a' [i64] [ 1 2 3 ]
Notice that the dtype is automatically inferred as a polars Int64:
Constructing a Series with a specific dtype:
s2 = pl.Series("a", [1, 2, 3], dtype=pl.Float32) s2 shape: (3,) Series: 'a' [f32] [ 1.0 2.0 3.0 ]
It is possible to construct a Series with values as the first positional argument. This syntax considered an anti-pattern, but it can be useful in certain scenarios. You must specify any other arguments through keywords.
s3 = pl.Series([1, 2, 3]) s3 shape: (3,) Series: '' [i64] [ 1 2 3 ]
Methods:
Attributes:
abs() → Series[source]
Compute absolute values.
Same as abs(series)
.
Examples
s = pl.Series([1, -2, -3]) s.abs() shape: (3,) Series: '' [i64] [ 1 2 3 ]
alias(name: str) → Series[source]
Rename the series.
Parameters:
name
The new name.
Examples
s = pl.Series("a", [1, 2, 3]) s.alias("b") shape: (3,) Series: 'b' [i64] [ 1 2 3 ]
all(*, ignore_nulls: bool = True) → bool | None[source]
Return whether all values in the column are True
.
Only works on columns of data type Boolean
.
Parameters:
ignore_nulls
Ignore null values (default).
If set to False
, Kleene logic is used to deal with nulls: if the column contains any null values and no False
values, the output is None
.
Returns:
bool or None
Examples
pl.Series([True, True]).all() True pl.Series([False, True]).all() False pl.Series([None, True]).all() True
Enable Kleene logic by setting ignore_nulls=False
.
pl.Series([None, True]).all(ignore_nulls=False) # Returns None
any(*, ignore_nulls: bool = True) → bool | None[source]
Return whether any of the values in the column are True
.
Only works on columns of data type Boolean
.
Parameters:
ignore_nulls
Ignore null values (default).
If set to False
, Kleene logic is used to deal with nulls: if the column contains any null values and no True
values, the output is None
.
Returns:
bool or None
Examples
pl.Series([True, False]).any() True pl.Series([False, False]).any() False pl.Series([None, False]).any() False
Enable Kleene logic by setting ignore_nulls=False
.
pl.Series([None, False]).any(ignore_nulls=False) # Returns None
append(other: Series) → Self[source]
Append a Series to this one.
The resulting series will consist of multiple chunks.
Parameters:
other
Series to append.
Warning
This method modifies the series in-place. The series is returned for convenience only.
Examples
a = pl.Series("a", [1, 2, 3]) b = pl.Series("b", [4, 5]) a.append(b) shape: (5,) Series: 'a' [i64] [ 1 2 3 4 5 ]
The resulting series will consist of multiple chunks.
approx_n_unique() → PythonLiteral | None[source]
Approximate count of unique values.
This is done using the HyperLogLog++ algorithm for cardinality estimation.
arccos() → Series[source]
Compute the element-wise value for the inverse cosine.
Examples
s = pl.Series("a", [1.0, 0.0, -1.0]) s.arccos() shape: (3,) Series: 'a' [f64] [ 0.0 1.570796 3.141593 ]
arccosh() → Series[source]
Compute the element-wise value for the inverse hyperbolic cosine.
Examples
s = pl.Series("a", [5.0, 1.0, 0.0, -1.0]) s.arccosh() shape: (4,) Series: 'a' [f64] [ 2.292432 0.0 NaN NaN ]
arcsin() → Series[source]
Compute the element-wise value for the inverse sine.
Examples
s = pl.Series("a", [1.0, 0.0, -1.0]) s.arcsin() shape: (3,) Series: 'a' [f64] [ 1.570796 0.0 -1.570796 ]
arcsinh() → Series[source]
Compute the element-wise value for the inverse hyperbolic sine.
Examples
s = pl.Series("a", [1.0, 0.0, -1.0]) s.arcsinh() shape: (3,) Series: 'a' [f64] [ 0.881374 0.0 -0.881374 ]
arctan() → Series[source]
Compute the element-wise value for the inverse tangent.
Examples
s = pl.Series("a", [1.0, 0.0, -1.0]) s.arctan() shape: (3,) Series: 'a' [f64] [ 0.785398 0.0 -0.785398 ]
arctanh() → Series[source]
Compute the element-wise value for the inverse hyperbolic tangent.
Examples
s = pl.Series("a", [2.0, 1.0, 0.5, 0.0, -0.5, -1.0, -1.1]) s.arctanh() shape: (7,) Series: 'a' [f64] [ NaN inf 0.549306 0.0 -0.549306 -inf NaN ]
arg_max() → int | None[source]
Get the index of the maximal value.
Returns:
int
Examples
s = pl.Series("a", [3, 2, 1]) s.arg_max() 0
arg_min() → int | None[source]
Get the index of the minimal value.
Returns:
int
Examples
s = pl.Series("a", [3, 2, 1]) s.arg_min() 2
arg_sort(
*,
descending: bool = False,
nulls_last: bool = False,
) → Series[source]
Get the index values that would sort this Series.
Parameters:
descending
Sort in descending order.
nulls_last
Place null values last instead of first.
Examples
s = pl.Series("a", [5, 3, 4, 1, 2]) s.arg_sort() shape: (5,) Series: 'a' [u32] [ 3 4 1 2 0 ]
arg_true() → Series[source]
Get index values where Boolean Series evaluate True.
Returns:
Series
Series of data type UInt32
.
Examples
s = pl.Series("a", [1, 2, 3]) (s == 2).arg_true() shape: (1,) Series: 'a' [u32] [ 1 ]
arg_unique() → Series[source]
Get unique index as Series.
Returns:
Series
Examples
s = pl.Series("a", [1, 2, 2, 3]) s.arg_unique() shape: (3,) Series: 'a' [u32] [ 0 1 3 ]
backward_fill(limit: int | None = None) → Series[source]
Fill missing values with the next non-null value.
This is an alias of .fill_null(strategy="backward")
.
Parameters:
limit
The number of consecutive null values to backward fill.
bitwise_and() → PythonLiteral | None[source]
Perform an aggregation of bitwise ANDs.
bitwise_count_ones() → Self[source]
Evaluate the number of set bits.
bitwise_count_zeros() → Self[source]
Evaluate the number of unset Self.
bitwise_leading_ones() → Self[source]
Evaluate the number most-significant set bits before seeing an unset bit.
bitwise_leading_zeros() → Self[source]
Evaluate the number most-significant unset bits before seeing a set bit.
bitwise_or() → PythonLiteral | None[source]
Perform an aggregation of bitwise ORs.
bitwise_trailing_ones() → Self[source]
Evaluate the number least-significant set bits before seeing an unset bit.
bitwise_trailing_zeros() → Self[source]
Evaluate the number least-significant unset bits before seeing a set bit.
bitwise_xor() → PythonLiteral | None[source]
Perform an aggregation of bitwise XORs.
bottom_k(k: int = 5) → Series[source]
Return the k
smallest elements.
Non-null elements are always preferred over null elements. The output is not guaranteed to be in any particular order, call sort() after this function if you wish the output to be sorted.
This has time complexity:
\[O(n)\]
Parameters:
k
Number of elements to return.
Examples
s = pl.Series("a", [2, 5, 1, 4, 3]) s.bottom_k(3) shape: (3,) Series: 'a' [i64] [ 1 2 3 ]
cast(
dtype: type[int | float | str | bool] | PolarsDataType,
*,
strict: bool = True,
wrap_numerical: bool = False,
) → Self[source]
Cast between data types.
Parameters:
dtype
DataType to cast to.
strict
If True invalid casts generate exceptions instead of null
s.
wrap_numerical
If True numeric casts wrap overflowing values instead of marking the cast as invalid.
Examples
s = pl.Series("a", [True, False, True]) s shape: (3,) Series: 'a' [bool] [ true false true ]
s.cast(pl.UInt32) shape: (3,) Series: 'a' [u32] [ 1 0 1 ]
cbrt() → Series[source]
Compute the cube root of the elements.
Optimization for
pl.Series([1, 2]) ** (1.0 / 3) shape: (2,) Series: '' [f64] [ 1.0 1.259921 ]
Examples
s = pl.Series([1, 2, 3]) s.cbrt() shape: (3,) Series: '' [f64] [ 1.0 1.259921 1.44225 ]
ceil() → Series[source]
Rounds up to the nearest integer value.
Only works on floating point Series.
Examples
s = pl.Series("a", [1.12345, 2.56789, 3.901234]) s.ceil() shape: (3,) Series: 'a' [f64] [ 2.0 3.0 4.0 ]
chunk_lengths() → list[int][source]
Get the length of each individual chunk.
Examples
s = pl.Series("a", [1, 2, 3]) s2 = pl.Series("a", [4, 5, 6])
Concatenate Series with rechunk = True
pl.concat([s, s2], rechunk=True).chunk_lengths() [6]
Concatenate Series with rechunk = False
pl.concat([s, s2], rechunk=False).chunk_lengths() [3, 3]
clear(n: int = 0) → Series[source]
Create an empty copy of the current Series, with zero to ‘n’ elements.
The copy has an identical name/dtype, but no data.
Parameters:
n
Number of (empty) elements to return in the cleared frame.
See also
Cheap deepcopy/clone.
Examples
s = pl.Series("a", [None, True, False]) s.clear() shape: (0,) Series: 'a' [bool] [ ]
s.clear(n=2) shape: (2,) Series: 'a' [bool] [ null null ]
clip(
lower_bound: NumericLiteral | TemporalLiteral | IntoExprColumn | None = None,
upper_bound: NumericLiteral | TemporalLiteral | IntoExprColumn | None = None,
) → Series[source]
Set values outside the given boundaries to the boundary value.
Parameters:
lower_bound
Lower bound. Accepts expression input. Non-expression inputs are parsed as literals. If set to None
(default), no lower bound is applied.
upper_bound
Upper bound. Accepts expression input. Non-expression inputs are parsed as literals. If set to None
(default), no upper bound is applied.
Notes
This method only works for numeric and temporal columns. To clip other data types, consider writing a when-then-otherwise
expression. See when().
Examples
Specifying both a lower and upper bound:
s = pl.Series([-50, 5, 50, None]) s.clip(1, 10) shape: (4,) Series: '' [i64] [ 1 5 10 null ]
Specifying only a single bound:
s.clip(upper_bound=10) shape: (4,) Series: '' [i64] [ -50 5 10 null ]
clone() → Self[source]
Create a copy of this Series.
This is a cheap operation that does not copy data.
See also
Create an empty copy of the current Series, with identical schema but no data.
Examples
s = pl.Series("a", [1, 2, 3]) s.clone() shape: (3,) Series: 'a' [i64] [ 1 2 3 ]
cos() → Series[source]
Compute the element-wise value for the cosine.
Examples
import math s = pl.Series("a", [0.0, math.pi / 2.0, math.pi]) s.cos() shape: (3,) Series: 'a' [f64] [ 1.0 6.1232e-17 -1.0 ]
cosh() → Series[source]
Compute the element-wise value for the hyperbolic cosine.
Examples
s = pl.Series("a", [1.0, 0.0, -1.0]) s.cosh() shape: (3,) Series: 'a' [f64] [ 1.543081 1.0 1.543081 ]
cot() → Series[source]
Compute the element-wise value for the cotangent.
Examples
import math s = pl.Series("a", [0.0, math.pi / 2.0, math.pi]) s.cot() shape: (3,) Series: 'a' [f64] [ inf 6.1232e-17 -8.1656e15 ]
Return the number of non-null elements in the column.
Examples
s = pl.Series("a", [1, 2, None]) s.count() 2
cum_count(*, reverse: bool = False) → Self[source]
Return the cumulative count of the non-null values in the column.
Parameters:
reverse
Reverse the operation.
Examples
s = pl.Series(["x", "k", None, "d"]) s.cum_count() shape: (4,) Series: '' [u32] [ 1 2 2 3 ]
cum_max(*, reverse: bool = False) → Series[source]
Get an array with the cumulative max computed at every element.
Parameters:
reverse
reverse the operation.
Examples
s = pl.Series("s", [3, 5, 1]) s.cum_max() shape: (3,) Series: 's' [i64] [ 3 5 5 ]
cum_min(*, reverse: bool = False) → Series[source]
Get an array with the cumulative min computed at every element.
Parameters:
reverse
reverse the operation.
Examples
s = pl.Series("s", [1, 2, 3]) s.cum_min() shape: (3,) Series: 's' [i64] [ 1 1 1 ]
cum_prod(*, reverse: bool = False) → Series[source]
Get an array with the cumulative product computed at every element.
Parameters:
reverse
reverse the operation.
Notes
Dtypes in {Int8, UInt8, Int16, UInt16} are cast to Int64 before summing to prevent overflow issues.
Examples
s = pl.Series("a", [1, 2, 3]) s.cum_prod() shape: (3,) Series: 'a' [i64] [ 1 2 6 ]
cum_sum(*, reverse: bool = False) → Series[source]
Get an array with the cumulative sum computed at every element.
Parameters:
reverse
reverse the operation.
Notes
Dtypes in {Int8, UInt8, Int16, UInt16} are cast to Int64 before summing to prevent overflow issues.
Examples
s = pl.Series("a", [1, 2, 3]) s.cum_sum() shape: (3,) Series: 'a' [i64] [ 1 3 6 ]
cumulative_eval(
expr: Expr,
*,
min_samples: int = 1,
parallel: bool = False,
) → Series[source]
Run an expression over a sliding window that increases 1
slot every iteration.
Warning
This functionality is considered unstable. It may be changed at any point without it being considered a breaking change.
Parameters:
expr
Expression to evaluate
min_samples
Number of valid values there should be in the window before the expression is evaluated. valid values = length - null_count
parallel
Run in parallel. Don’t do this in a group by or another operation that already has much parallelization.
Warning
This can be really slow as it can have O(n^2)
complexity. Don’t use this for operations that visit all elements.
Examples
s = pl.Series("values", [1, 2, 3, 4, 5]) s.cumulative_eval(pl.element().first() - pl.element().last() ** 2) shape: (5,) Series: 'values' [i64] [ 0 -3 -8 -15 -24 ]
cut(
*,
labels: Sequence[str] | None = None,
left_closed: bool = False,
include_breaks: bool = False,
) → Series[source]
Bin continuous values into discrete categories.
Warning
This functionality is considered unstable. It may be changed at any point without it being considered a breaking change.
Parameters:
breaks
List of unique cut points.
labels
Names of the categories. The number of labels must be equal to the number of cut points plus one.
left_closed
Set the intervals to be left-closed instead of right-closed.
include_breaks
Include a column with the right endpoint of the bin each observation falls in. This will change the data type of the output from aCategorical
to a Struct
.
Returns:
Series
Series of data type Categorical
if include_breaks
is set toFalse
(default), otherwise a Series of data type Struct
.
Examples
Divide the column into three categories.
s = pl.Series("foo", [-2, -1, 0, 1, 2]) s.cut([-1, 1], labels=["a", "b", "c"]) shape: (5,) Series: 'foo' [cat] [ "a" "a" "b" "b" "c" ]
Create a DataFrame with the breakpoint and category for each value.
cut = s.cut([-1, 1], include_breaks=True).alias("cut") s.to_frame().with_columns(cut).unnest("cut") shape: (5, 3) ┌─────┬────────────┬────────────┐ │ foo ┆ breakpoint ┆ category │ │ --- ┆ --- ┆ --- │ │ i64 ┆ f64 ┆ cat │ ╞═════╪════════════╪════════════╡ │ -2 ┆ -1.0 ┆ (-inf, -1] │ │ -1 ┆ -1.0 ┆ (-inf, -1] │ │ 0 ┆ 1.0 ┆ (-1, 1] │ │ 1 ┆ 1.0 ┆ (-1, 1] │ │ 2 ┆ inf ┆ (1, inf] │ └─────┴────────────┴────────────┘
describe(
percentiles: Sequence[float] | float | None = (0.25, 0.5, 0.75),
interpolation: RollingInterpolationMethod = 'nearest',
) → DataFrame[source]
Quick summary statistics of a Series.
Series with mixed datatypes will return summary statistics for the datatype of the first value.
Parameters:
percentiles
One or more percentiles to include in the summary statistics (if the Series has a numeric dtype). All values must be in the range [0, 1]
.
interpolation{‘nearest’, ‘higher’, ‘lower’, ‘midpoint’, ‘linear’}
Interpolation method used when calculating percentiles.
Returns:
DataFrame
Mapping with summary statistics of a Series.
Notes
The median is included by default as the 50% percentile.
Examples
s = pl.Series([1, 2, 3, 4, 5]) s.describe() shape: (9, 2) ┌────────────┬──────────┐ │ statistic ┆ value │ │ --- ┆ --- │ │ str ┆ f64 │ ╞════════════╪══════════╡ │ count ┆ 5.0 │ │ null_count ┆ 0.0 │ │ mean ┆ 3.0 │ │ std ┆ 1.581139 │ │ min ┆ 1.0 │ │ 25% ┆ 2.0 │ │ 50% ┆ 3.0 │ │ 75% ┆ 4.0 │ │ max ┆ 5.0 │ └────────────┴──────────┘
Non-numeric data types may not have all statistics available.
s = pl.Series(["aa", "aa", None, "bb", "cc"]) s.describe() shape: (4, 2) ┌────────────┬───────┐ │ statistic ┆ value │ │ --- ┆ --- │ │ str ┆ str │ ╞════════════╪═══════╡ │ count ┆ 4 │ │ null_count ┆ 1 │ │ min ┆ aa │ │ max ┆ cc │ └────────────┴───────┘
diff(n: int = 1, null_behavior: NullBehavior = 'ignore') → Series[source]
Calculate the first discrete difference between shifted items.
Parameters:
n
Number of slots to shift.
null_behavior{‘ignore’, ‘drop’}
How to handle null values.
Examples
s = pl.Series("s", values=[20, 10, 30, 25, 35], dtype=pl.Int8) s.diff() shape: (5,) Series: 's' [i8] [ null -10 20 -5 10 ]
s.diff(n=2) shape: (5,) Series: 's' [i8] [ null null 10 15 5 ]
s.diff(n=2, null_behavior="drop") shape: (3,) Series: 's' [i8] [ 10 15 5 ]
dot(other: Series | ArrayLike) → int | float | None[source]
Compute the dot/inner product between two Series.
Parameters:
other
Series (or array) to compute dot product with.
Examples
s = pl.Series("a", [1, 2, 3]) s2 = pl.Series("b", [4.0, 5.0, 6.0]) s.dot(s2) 32.0
drop_nans() → Series[source]
Drop all floating point NaN values.
The original order of the remaining elements is preserved.
Notes
A NaN value is not the same as a null value. To drop null values, use drop_nulls().
Examples
s = pl.Series([1.0, None, 3.0, float("nan")]) s.drop_nans() shape: (3,) Series: '' [f64] [ 1.0 null 3.0 ]
drop_nulls() → Series[source]
Drop all null values.
The original order of the remaining elements is preserved.
Notes
A null value is not the same as a NaN value. To drop NaN values, use drop_nans().
Examples
s = pl.Series([1.0, None, 3.0, float("nan")]) s.drop_nulls() shape: (3,) Series: '' [f64] [ 1.0 3.0 NaN ]
property dtype_: DataType_[source]
Get the data type of this Series.
Examples
s = pl.Series("a", [1, 2, 3]) s.dtype Int64
entropy(
base: float = 2.718281828459045,
*,
normalize: bool = True,
Computes the entropy.
Uses the formula -sum(pk * log(pk)
where pk
are discrete probabilities.
Parameters:
base
Given base, defaults to e
normalize
Normalize pk if it doesn’t sum to 1.
Examples
a = pl.Series([0.99, 0.005, 0.005]) a.entropy(normalize=True) 0.06293300616044681 b = pl.Series([0.65, 0.10, 0.25]) b.entropy(normalize=True) 0.8568409950394724
eq(other: Any) → Series | Expr[source]
Method equivalent of operator expression series == other
.
eq_missing(other: Any) → Series | Expr[source]
Method equivalent of equality operator series == other
where None == None
.
This differs from the standard eq
where null values are propagated.
Parameters:
other
A literal or expression value to compare with.
Examples
s1 = pl.Series("a", [333, 200, None]) s2 = pl.Series("a", [100, 200, None]) s1.eq(s2) shape: (3,) Series: 'a' [bool] [ false true null ] s1.eq_missing(s2) shape: (3,) Series: 'a' [bool] [ false true true ]
equals(
other: Series,
*,
check_dtypes: bool = False,
check_names: bool = False,
null_equal: bool = True,
Check whether the Series is equal to another Series.
Parameters:
other
Series to compare with.
check_dtypes
Require data types to match.
check_names
Require names to match.
null_equal
Consider null values as equal.
Examples
s1 = pl.Series("a", [1, 2, 3]) s2 = pl.Series("b", [4, 5, 6]) s1.equals(s1) True s1.equals(s2) False
estimated_size(unit: SizeUnit = 'b') → int | float[source]
Return an estimation of the total (heap) allocated size of the Series.
Estimated size is given in the specified unit (bytes by default).
This estimation is the sum of the size of its buffers, validity, including nested arrays. Multiple arrays may share buffers and bitmaps. Therefore, the size of 2 arrays is not the sum of the sizes computed from this function. In particular, [StructArray
]’s size is an upper bound.
When an array is sliced, its allocated size remains constant because the buffer unchanged. However, this function will yield a smaller number. This is because this function returns the visible size of the buffer, not its total capacity.
FFI buffers are included in this estimation.
Parameters:
unit{‘b’, ‘kb’, ‘mb’, ‘gb’, ‘tb’}
Scale the returned size to the given unit.
Notes
For data with Object dtype, the estimated size only reports the pointer size, which is a huge underestimation.
Examples
s = pl.Series("values", list(range(1_000_000)), dtype=pl.UInt32) s.estimated_size() 4000000 s.estimated_size("mb") 3.814697265625
ewm_mean(
*,
half_life: float | None = None,
adjust: bool = True,
min_samples: int = 1,
ignore_nulls: bool = False,
) → Series[source]
Compute exponentially-weighted moving average.
Parameters:
com
Specify decay in terms of center of mass, \(\gamma\), with
\[\alpha = \frac{1}{1 + \gamma} \; \forall \; \gamma \geq 0\]
span
Specify decay in terms of span, \(\theta\), with
\[\alpha = \frac{2}{\theta + 1} \; \forall \; \theta \geq 1\]
half_life
Specify decay in terms of half-life, \(\tau\), with
\[\alpha = 1 - \exp \left\{ \frac{ -\ln(2) }{ \tau } \right\} \; \forall \; \tau > 0\]
alpha
Specify smoothing factor alpha directly, \(0 < \alpha \leq 1\).
adjust
Divide by decaying adjustment factor in beginning periods to account for imbalance in relative weightings
- When
adjust=True
(the default) the EW function is calculated using weights \(w_i = (1 - \alpha)^i\)- When
adjust=False
the EW function is calculated recursively by
\[\begin{split}y_0 &= x_0 \\ y_t &= (1 - \alpha)y_{t - 1} + \alpha x_t\end{split}\]
min_samples
Minimum number of observations in window required to have a value (otherwise result is null).
ignore_nulls
Ignore missing values when calculating weights.
- When
ignore_nulls=False
(default), weights are based on absolute positions. For example, the weights of \(x_0\) and \(x_2\) used in calculating the final weighted average of [\(x_0\), None, \(x_2\)] are\((1-\alpha)^2\) and \(1\) ifadjust=True
, and\((1-\alpha)^2\) and \(\alpha\) ifadjust=False
.- When
ignore_nulls=True
, weights are based on relative positions. For example, the weights of\(x_0\) and \(x_2\) used in calculating the final weighted average of [\(x_0\), None, \(x_2\)] are\(1-\alpha\) and \(1\) ifadjust=True
, and \(1-\alpha\) and \(\alpha\) ifadjust=False
.
Examples
s = pl.Series([1, 2, 3]) s.ewm_mean(com=1, ignore_nulls=False) shape: (3,) Series: '' [f64] [ 1.0 1.666667 2.428571 ]
ewm_mean_by(by: IntoExpr, *, half_life: str | timedelta) → Series[source]
Compute time-based exponentially weighted moving average.
Given observations \(x_0, x_1, \ldots, x_{n-1}\) at times\(t_0, t_1, \ldots, t_{n-1}\), the EWMA is calculated as
\[ \begin{align}\begin{aligned}y_0 &= x_0\\\alpha_i &= 1 - \exp \left\{ \frac{ -\ln(2)(t_i-t_{i-1}) } { \tau } \right\}\\y_i &= \alpha_i x_i + (1 - \alpha_i) y_{i-1}; \quad i > 0\end{aligned}\end{align} \]
where \(\tau\) is the half_life
.
Parameters:
by
Times to calculate average by. Should be DateTime
, Date
, UInt64
,UInt32
, Int64
, or Int32
data type.
half_life
Unit over which observation decays to half its value.
Can be created either from a timedelta, or by using the following string language:
- 1ns (1 nanosecond)
- 1us (1 microsecond)
- 1ms (1 millisecond)
- 1s (1 second)
- 1m (1 minute)
- 1h (1 hour)
- 1d (1 day)
- 1w (1 week)
- 1i (1 index count)
Or combine them: “3d12h4m25s” # 3 days, 12 hours, 4 minutes, and 25 seconds
Note that half_life
is treated as a constant duration - calendar durations such as months (or even days in the time-zone-aware case) are not supported, please express your duration in an approximately equivalent number of hours (e.g. ‘370h’ instead of ‘1mo’).
Returns:
Expr
Float32 if input is Float32, otherwise Float64.
Examples
from datetime import date, timedelta df = pl.DataFrame( ... { ... "values": [0, 1, 2, None, 4], ... "times": [ ... date(2020, 1, 1), ... date(2020, 1, 3), ... date(2020, 1, 10), ... date(2020, 1, 15), ... date(2020, 1, 17), ... ], ... } ... ).sort("times") df["values"].ewm_mean_by(df["times"], half_life="4d") shape: (5,) Series: 'values' [f64] [ 0.0 0.292893 1.492474 null 3.254508 ]
ewm_std(
*,
half_life: float | None = None,
adjust: bool = True,
bias: bool = False,
min_samples: int = 1,
ignore_nulls: bool = False,
) → Series[source]
Compute exponentially-weighted moving standard deviation.
Parameters:
com
Specify decay in terms of center of mass, \(\gamma\), with
\[\alpha = \frac{1}{1 + \gamma} \; \forall \; \gamma \geq 0\]
span
Specify decay in terms of span, \(\theta\), with
\[\alpha = \frac{2}{\theta + 1} \; \forall \; \theta \geq 1\]
half_life
Specify decay in terms of half-life, \(\lambda\), with
\[\alpha = 1 - \exp \left\{ \frac{ -\ln(2) }{ \lambda } \right\} \; \forall \; \lambda > 0\]
alpha
Specify smoothing factor alpha directly, \(0 < \alpha \leq 1\).
adjust
Divide by decaying adjustment factor in beginning periods to account for imbalance in relative weightings
- When
adjust=True
(the default) the EW function is calculated using weights \(w_i = (1 - \alpha)^i\)- When
adjust=False
the EW function is calculated recursively by
\[\begin{split}y_0 &= x_0 \\ y_t &= (1 - \alpha)y_{t - 1} + \alpha x_t\end{split}\]
bias
When bias=False
, apply a correction to make the estimate statistically unbiased.
min_samples
Minimum number of observations in window required to have a value (otherwise result is null).
ignore_nulls
Ignore missing values when calculating weights.
- When
ignore_nulls=False
(default), weights are based on absolute positions. For example, the weights of \(x_0\) and \(x_2\) used in calculating the final weighted average of [\(x_0\), None, \(x_2\)] are\((1-\alpha)^2\) and \(1\) ifadjust=True
, and\((1-\alpha)^2\) and \(\alpha\) ifadjust=False
.- When
ignore_nulls=True
, weights are based on relative positions. For example, the weights of\(x_0\) and \(x_2\) used in calculating the final weighted average of [\(x_0\), None, \(x_2\)] are\(1-\alpha\) and \(1\) ifadjust=True
, and \(1-\alpha\) and \(\alpha\) ifadjust=False
.
Examples
s = pl.Series("a", [1, 2, 3]) s.ewm_std(com=1, ignore_nulls=False) shape: (3,) Series: 'a' [f64] [ 0.0 0.707107 0.963624 ]
ewm_var(
*,
half_life: float | None = None,
adjust: bool = True,
bias: bool = False,
min_samples: int = 1,
ignore_nulls: bool = False,
) → Series[source]
Compute exponentially-weighted moving variance.
Parameters:
com
Specify decay in terms of center of mass, \(\gamma\), with
\[\alpha = \frac{1}{1 + \gamma} \; \forall \; \gamma \geq 0\]
span
Specify decay in terms of span, \(\theta\), with
\[\alpha = \frac{2}{\theta + 1} \; \forall \; \theta \geq 1\]
half_life
Specify decay in terms of half-life, \(\lambda\), with
\[\alpha = 1 - \exp \left\{ \frac{ -\ln(2) }{ \lambda } \right\} \; \forall \; \lambda > 0\]
alpha
Specify smoothing factor alpha directly, \(0 < \alpha \leq 1\).
adjust
Divide by decaying adjustment factor in beginning periods to account for imbalance in relative weightings
- When
adjust=True
(the default) the EW function is calculated using weights \(w_i = (1 - \alpha)^i\)- When
adjust=False
the EW function is calculated recursively by
\[\begin{split}y_0 &= x_0 \\ y_t &= (1 - \alpha)y_{t - 1} + \alpha x_t\end{split}\]
bias
When bias=False
, apply a correction to make the estimate statistically unbiased.
min_samples
Minimum number of observations in window required to have a value (otherwise result is null).
ignore_nulls
Ignore missing values when calculating weights.
- When
ignore_nulls=False
(default), weights are based on absolute positions. For example, the weights of \(x_0\) and \(x_2\) used in calculating the final weighted average of [\(x_0\), None, \(x_2\)] are\((1-\alpha)^2\) and \(1\) ifadjust=True
, and\((1-\alpha)^2\) and \(\alpha\) ifadjust=False
.- When
ignore_nulls=True
, weights are based on relative positions. For example, the weights of\(x_0\) and \(x_2\) used in calculating the final weighted average of [\(x_0\), None, \(x_2\)] are\(1-\alpha\) and \(1\) ifadjust=True
, and \(1-\alpha\) and \(\alpha\) ifadjust=False
.
Examples
s = pl.Series("a", [1, 2, 3]) s.ewm_var(com=1, ignore_nulls=False) shape: (3,) Series: 'a' [f64] [ 0.0 0.5 0.928571 ]
exp() → Series[source]
Compute the exponential, element-wise.
Examples
s = pl.Series([1, 2, 3]) s.exp() shape: (3,) Series: '' [f64] [ 2.718282 7.389056 20.085537 ]
explode() → Series[source]
Explode a list Series.
This means that every item is expanded to a new row.
Returns:
Series
Series with the data type of the list elements.
Examples
s = pl.Series("a", [[1, 2, 3], [4, 5, 6]]) s shape: (2,) Series: 'a' [list[i64]] [ [1, 2, 3] [4, 5, 6] ] s.explode() shape: (6,) Series: 'a' [i64] [ 1 2 3 4 5 6 ]
extend(other: Series) → Self[source]
Extend the memory backed by this Series with the values from another.
Different from append
, which adds the chunks from other
to the chunks of this series, extend
appends the data from other
to the underlying memory locations and thus may cause a reallocation (which is expensive).
If this does not
cause a reallocation, the resulting data structure will not have any extra chunks and thus will yield faster queries.
Prefer extend
over append
when you want to do a query after a single append. For instance, during online operations where you add n
rows and rerun a query.
Prefer append
over extend
when you want to append many times before doing a query. For instance, when you read in multiple files and want to store them in a single Series
. In the latter case, finish the sequence of append
operations with a rechunk
.
Parameters:
other
Series to extend the series with.
Warning
This method modifies the series in-place. The series is returned for convenience only.
Examples
a = pl.Series("a", [1, 2, 3]) b = pl.Series("b", [4, 5]) a.extend(b) shape: (5,) Series: 'a' [i64] [ 1 2 3 4 5 ]
The resulting series will consist of a single chunk.
extend_constant(value: IntoExpr, n: int | IntoExprColumn) → Series[source]
Extremely fast method for extending the Series with ‘n’ copies of a value.
Parameters:
value
A constant literal value or a unit expression with which to extend the expression result Series; can pass None to extend with nulls.
n
The number of additional values that will be added.
Examples
s = pl.Series([1, 2, 3]) s.extend_constant(99, n=2) shape: (5,) Series: '' [i64] [ 1 2 3 99 99 ]
fill_nan(value: int | float | Expr | None) → Series[source]
Fill floating point NaN value with a fill value.
Parameters:
value
Value used to fill NaN values.
Warning
Note that floating point NaNs (Not a Number) are not missing values. To replace missing values, use fill_null().
Examples
s = pl.Series("a", [1.0, 2.0, 3.0, float("nan")]) s.fill_nan(0) shape: (4,) Series: 'a' [f64] [ 1.0 2.0 3.0 0.0 ]
fill_null(
value: Any | Expr | None = None,
strategy: FillNullStrategy | None = None,
) → Series[source]
Fill null values using the specified value or strategy.
Parameters:
value
Value used to fill null values.
strategy{None, ‘forward’, ‘backward’, ‘min’, ‘max’, ‘mean’, ‘zero’, ‘one’}
Strategy used to fill null values.
limit
Number of consecutive null values to fill when using the ‘forward’ or ‘backward’ strategy.
Examples
s = pl.Series("a", [1, 2, 3, None]) s.fill_null(strategy="forward") shape: (4,) Series: 'a' [i64] [ 1 2 3 3 ] s.fill_null(strategy="min") shape: (4,) Series: 'a' [i64] [ 1 2 3 1 ] s = pl.Series("b", ["x", None, "z"]) s.fill_null(pl.lit("")) shape: (3,) Series: 'b' [str] [ "x" "" "z" ]
filter(predicate: Series | Iterable[bool]) → Self[source]
Filter elements by a boolean mask.
The original order of the remaining elements is preserved.
Elements where the filter does not evaluate to True are discarded, including nulls.
Parameters:
predicate
Boolean mask.
Examples
s = pl.Series("a", [1, 2, 3]) mask = pl.Series("", [True, False, True]) s.filter(mask) shape: (2,) Series: 'a' [i64] [ 1 3 ]
first() → PythonLiteral | None[source]
Get the first element of the Series.
Returns None
if the Series is empty.
property flags_: dict[str, bool]_[source]
Get flags that are set on the Series.
Examples
s = pl.Series("a", [1, 2, 3]) s.flags {'SORTED_ASC': False, 'SORTED_DESC': False}
floor() → Series[source]
Rounds down to the nearest integer value.
Only works on floating point Series.
Examples
s = pl.Series("a", [1.12345, 2.56789, 3.901234]) s.floor() shape: (3,) Series: 'a' [f64] [ 1.0 2.0 3.0 ]
forward_fill(limit: int | None = None) → Series[source]
Fill missing values with the last non-null value.
This is an alias of .fill_null(strategy="forward")
.
Parameters:
limit
The number of consecutive null values to forward fill.
gather(indices: int | list[int] | Expr | Series | np.ndarray[Any, Any]) → Series[source]
Take values by index.
Parameters:
indices
Index location used for selection.
Examples
s = pl.Series("a", [1, 2, 3, 4]) s.gather([1, 3]) shape: (2,) Series: 'a' [i64] [ 2 4 ]
gather_every(n: int, offset: int = 0) → Series[source]
Take every nth value in the Series and return as new Series.
Parameters:
n
Gather every _n_-th row.
offset
Start the row index at this offset.
Examples
s = pl.Series("a", [1, 2, 3, 4]) s.gather_every(2) shape: (2,) Series: 'a' [i64] [ 1 3 ] s.gather_every(2, offset=1) shape: (2,) Series: 'a' [i64] [ 2 4 ]
ge(other: Any) → Series | Expr[source]
Method equivalent of operator expression series >= other
.
get_chunks() → list[Series][source]
Get the chunks of this Series as a list of Series.
Examples
s1 = pl.Series("a", [1, 2, 3]) s2 = pl.Series("a", [4, 5, 6]) s = pl.concat([s1, s2], rechunk=False) s.get_chunks() [shape: (3,) Series: 'a' [i64] [ 1 2 3 ], shape: (3,) Series: 'a' [i64] [ 4 5 6 ]]
gt(other: Any) → Series | Expr[source]
Method equivalent of operator expression series > other
.
Check whether the Series contains one or more null values.
Examples
s = pl.Series([1, 2, None]) s.has_nulls() True s[:2].has_nulls() False
Check whether the Series contains one or more null values.
Deprecated since version 0.20.30: Use has_nulls() instead.
hash(
seed: int = 0,
) → Series[source]
Hash the Series.
The hash value is of type UInt64
.
Parameters:
seed
Random seed parameter. Defaults to 0.
seed_1
Random seed parameter. Defaults to seed
if not set.
seed_2
Random seed parameter. Defaults to seed
if not set.
seed_3
Random seed parameter. Defaults to seed
if not set.
Notes
This implementation of hash
does not guarantee stable results across different Polars versions. Its stability is only guaranteed within a single version.
Examples
s = pl.Series("a", [1, 2, 3]) s.hash(seed=42)
shape: (3,) Series: 'a' [u64] [ 10734580197236529959 3022416320763508302 13756996518000038261 ]
head(n: int = 10) → Series[source]
Get the first n
elements.
Parameters:
n
Number of elements to return. If a negative value is passed, return all elements except the last abs(n)
.
Examples
s = pl.Series("a", [1, 2, 3, 4, 5]) s.head(3) shape: (3,) Series: 'a' [i64] [ 1 2 3 ]
Pass a negative value to get all rows except
the last abs(n)
.
s.head(-3) shape: (2,) Series: 'a' [i64] [ 1 2 ]
hist(
bins: list[float] | None = None,
*,
include_category: bool = True,
include_breakpoint: bool = True,
) → DataFrame[source]
Bin values into buckets and count their occurrences.
Warning
This functionality is considered unstable. It may be changed at any point without it being considered a breaking change.
Parameters:
bins
Bin edges. If None given, we determine the edges based on the data.
bin_count
If bins
is not provided, bin_count
uniform bins are created that fully encompass the data.
include_breakpoint
Include a column that indicates the upper breakpoint.
include_category
Include a column that shows the intervals as categories.
Returns:
DataFrame
Examples
a = pl.Series("a", [1, 3, 8, 8, 2, 1, 3]) a.hist(bin_count=4) shape: (4, 3) ┌────────────┬─────────────┬───────┐ │ breakpoint ┆ category ┆ count │ │ --- ┆ --- ┆ --- │ │ f64 ┆ cat ┆ u32 │ ╞════════════╪═════════════╪═══════╡ │ 2.75 ┆ [1.0, 2.75] ┆ 3 │ │ 4.5 ┆ (2.75, 4.5] ┆ 2 │ │ 6.25 ┆ (4.5, 6.25] ┆ 0 │ │ 8.0 ┆ (6.25, 8.0] ┆ 2 │ └────────────┴─────────────┴───────┘
implode() → Self[source]
Aggregate values into a list.
Examples
s = pl.Series("a", [1, 2, 3]) s.implode() shape: (1,) Series: 'a' [list[i64]] [ [1, 2, 3] ]
index_of(element: IntoExpr) → int | None[source]
Get the index of the first occurrence of a value, or None
if it’s not found.
Parameters:
element
Value to find.
Examples
s = pl.Series("a", [1, None, 17]) s.index_of(17) 2 s.index_of(None) # search for a null 1 s.index_of(55) is None True
interpolate(method: InterpolationMethod = 'linear') → Series[source]
Interpolate intermediate values.
Nulls at the beginning and end of the series remain null.
Parameters:
method{‘linear’, ‘nearest’}
Interpolation method.
Examples
s = pl.Series("a", [1, 2, None, None, 5]) s.interpolate() shape: (5,) Series: 'a' [f64] [ 1.0 2.0 3.0 4.0 5.0 ]
interpolate_by(by: IntoExpr) → Series[source]
Interpolate intermediate values with x-coordinate based on another column.
Nulls at the beginning and end of the series remain null.
Parameters:
by
Column to interpolate values based on.
Examples
Fill null values using linear interpolation.
s = pl.Series([1, None, None, 3]) by = pl.Series([1, 2, 7, 8]) s.interpolate_by(by) shape: (4,) Series: '' [f64] [ 1.0 1.285714 2.714286 3.0 ]
is_between(
lower_bound: IntoExpr,
upper_bound: IntoExpr,
closed: ClosedInterval = 'both',
) → Series[source]
Get a boolean mask of the values that are between the given lower/upper bounds.
Parameters:
lower_bound
Lower bound value. Accepts expression input. Non-expression inputs (including strings) are parsed as literals.
upper_bound
Upper bound value. Accepts expression input. Non-expression inputs (including strings) are parsed as literals.
closed{‘both’, ‘left’, ‘right’, ‘none’}
Define which sides of the interval are closed (inclusive).
Notes
If the value of the lower_bound
is greater than that of the upper_bound
then the result will be False, as no value can satisfy the condition.
Examples
s = pl.Series("num", [1, 2, 3, 4, 5]) s.is_between(2, 4) shape: (5,) Series: 'num' [bool] [ false true true true false ]
Use the closed
argument to include or exclude the values at the bounds:
s.is_between(2, 4, closed="left") shape: (5,) Series: 'num' [bool] [ false true true false false ]
You can also use strings as well as numeric/temporal values:
s = pl.Series("s", ["a", "b", "c", "d", "e"]) s.is_between("b", "d", closed="both") shape: (5,) Series: 's' [bool] [ false true true true false ]
is_duplicated() → Series[source]
Get mask of all duplicated values.
Returns:
Series
Series of data type Boolean
.
Examples
s = pl.Series("a", [1, 2, 2, 3]) s.is_duplicated() shape: (4,) Series: 'a' [bool] [ false true true false ]
Check if the Series is empty.
Examples
s = pl.Series("a", [], dtype=pl.Float32) s.is_empty() True
is_finite() → Series[source]
Returns a boolean Series indicating which values are finite.
Returns:
Series
Series of data type Boolean
.
Examples
import numpy as np s = pl.Series("a", [1.0, 2.0, np.inf]) s.is_finite() shape: (3,) Series: 'a' [bool] [ true true false ]
is_first_distinct() → Series[source]
Return a boolean mask indicating the first occurrence of each distinct value.
Returns:
Series
Series of data type Boolean
.
Examples
s = pl.Series([1, 1, 2, 3, 2]) s.is_first_distinct() shape: (5,) Series: '' [bool] [ true false true true false ]
is_in(other: Series | Collection[Any], *, nulls_equal: bool = False) → Series[source]
Check if elements of this Series are in the other Series.
Parameters:
nulls_equalbool, default False
If True, treat null as a distinct value. Null values will not propagate.
Returns:
Series
Series of data type Boolean
.
Examples
s = pl.Series("a", [1, 2, 3]) s2 = pl.Series("b", [2, 4, None]) s2.is_in(s) shape: (3,) Series: 'b' [bool] [ true false null ]
when nulls_equal=True, None is treated as a distinct value
s2.is_in(s, nulls_equal=True) shape: (3,) Series: 'b' [bool] [ true false false ]
check if some values are a member of sublists
sets = pl.Series("sets", [[1, 2, 3], [1, 2], [9, 10]]) optional_members = pl.Series("optional_members", [1, 2, 3]) print(sets) shape: (3,) Series: 'sets' [list[i64]] [ [1, 2, 3] [1, 2] [9, 10] ] print(optional_members) shape: (3,) Series: 'optional_members' [i64] [ 1 2 3 ] optional_members.is_in(sets) shape: (3,) Series: 'optional_members' [bool] [ true true false ]
is_infinite() → Series[source]
Returns a boolean Series indicating which values are infinite.
Returns:
Series
Series of data type Boolean
.
Examples
import numpy as np s = pl.Series("a", [1.0, 2.0, np.inf]) s.is_infinite() shape: (3,) Series: 'a' [bool] [ false false true ]
is_last_distinct() → Series[source]
Return a boolean mask indicating the last occurrence of each distinct value.
Returns:
Series
Series of data type Boolean
.
Examples
s = pl.Series([1, 1, 2, 3, 2]) s.is_last_distinct() shape: (5,) Series: '' [bool] [ false true false true true ]
is_nan() → Series[source]
Returns a boolean Series indicating which values are NaN.
Returns:
Series
Series of data type Boolean
.
Examples
import numpy as np s = pl.Series("a", [1.0, 2.0, 3.0, np.nan]) s.is_nan() shape: (4,) Series: 'a' [bool] [ false false false true ]
is_not_nan() → Series[source]
Returns a boolean Series indicating which values are not NaN.
Returns:
Series
Series of data type Boolean
.
Examples
import numpy as np s = pl.Series("a", [1.0, 2.0, 3.0, np.nan]) s.is_not_nan() shape: (4,) Series: 'a' [bool] [ true true true false ]
is_not_null() → Series[source]
Returns a boolean Series indicating which values are not null.
Returns:
Series
Series of data type Boolean
.
Examples
s = pl.Series("a", [1.0, 2.0, 3.0, None]) s.is_not_null() shape: (4,) Series: 'a' [bool] [ true true true false ]
is_null() → Series[source]
Returns a boolean Series indicating which values are null.
Returns:
Series
Series of data type Boolean
.
Examples
s = pl.Series("a", [1.0, 2.0, 3.0, None]) s.is_null() shape: (4,) Series: 'a' [bool] [ false false false true ]
is_sorted(*, descending: bool = False, nulls_last: bool = False) → bool[source]
Check if the Series is sorted.
Parameters:
descending
Check if the Series is sorted in descending order
nulls_last
Set nulls at the end of the Series in sorted check.
Examples
s = pl.Series([1, 3, 2]) s.is_sorted() False
s = pl.Series([3, 2, 1]) s.is_sorted(descending=True) True
is_unique() → Series[source]
Get mask of all unique values.
Returns:
Series
Series of data type Boolean
.
Examples
s = pl.Series("a", [1, 2, 2, 3]) s.is_unique() shape: (4,) Series: 'a' [bool] [ true false false true ]
item(index: int | None = None) → Any[source]
Return the Series as a scalar, or return the element at the given index.
If no index is provided, this is equivalent to s[0]
, with a check that the shape is (1,). With an index, this is equivalent to s[index]
.
Examples
s1 = pl.Series("a", [1]) s1.item() 1 s2 = pl.Series("a", [9, 8, 7]) s2.cum_sum().item(-1) 24
kurtosis(*, fisher: bool = True, bias: bool = True) → float | None[source]
Compute the kurtosis (Fisher or Pearson) of a dataset.
Kurtosis is the fourth central moment divided by the square of the variance. If Fisher’s definition is used, then 3.0 is subtracted from the result to give 0.0 for a normal distribution. If bias is False then the kurtosis is calculated using k statistics to eliminate bias coming from biased moment estimators
See scipy.stats for more information
Parameters:
fisherbool, optional
If True, Fisher’s definition is used (normal ==> 0.0). If False, Pearson’s definition is used (normal ==> 3.0).
biasbool, optional
If False, the calculations are corrected for statistical bias.
Examples
s = pl.Series("grades", [66, 79, 54, 97, 96, 70, 69, 85, 93, 75]) s.kurtosis() -1.0522623626787952 s.kurtosis(fisher=False) 1.9477376373212048 s.kurtosis(fisher=False, bias=False) 2.1040361802642717
last() → PythonLiteral | None[source]
Get the last element of the Series.
Returns None
if the Series is empty.
le(other: Any) → Series | Expr[source]
Method equivalent of operator expression series <= other
.
Return the number of elements in the Series.
Null values count towards the total.
Examples
s = pl.Series("a", [1, 2, None]) s.len() 3
limit(n: int = 10) → Series[source]
Get the first n
elements.
Alias for Series.head().
Parameters:
n
Number of elements to return. If a negative value is passed, return all elements except the last abs(n)
.
Examples
s = pl.Series("a", [1, 2, 3, 4, 5]) s.limit(3) shape: (3,) Series: 'a' [i64] [ 1 2 3 ]
Pass a negative value to get all rows except
the last abs(n)
.
s.limit(-3) shape: (2,) Series: 'a' [i64] [ 1 2 ]
log(base: float = 2.718281828459045) → Series[source]
Compute the logarithm to a given base.
Examples
s = pl.Series([1, 2, 3]) s.log() shape: (3,) Series: '' [f64] [ 0.0 0.693147 1.098612 ]
log10() → Series[source]
Compute the base 10 logarithm of the input array, element-wise.
Examples
s = pl.Series([10, 100, 1000]) s.log10() shape: (3,) Series: '' [f64] [ 1.0 2.0 3.0 ]
log1p() → Series[source]
Compute the natural logarithm of the input array plus one, element-wise.
Examples
s = pl.Series([1, 2, 3]) s.log1p() shape: (3,) Series: '' [f64] [ 0.693147 1.098612 1.386294 ]
lower_bound() → Self[source]
Return the lower bound of this Series’ dtype as a unit Series.
See also
return the upper bound of the given Series’ dtype.
Examples
s = pl.Series("s", [-1, 0, 1], dtype=pl.Int32) s.lower_bound() shape: (1,) Series: 's' [i32] [ -2147483648 ]
s = pl.Series("s", [1.0, 2.5, 3.0], dtype=pl.Float32) s.lower_bound() shape: (1,) Series: 's' [f32] [ -inf ]
lt(other: Any) → Series | Expr[source]
Method equivalent of operator expression series < other
.
map_elements(
function: Callable[[Any], Any],
return_dtype: PolarsDataType | None = None,
*,
skip_nulls: bool = True,
) → Self[source]
Map a custom/user-defined function (UDF) over elements in this Series.
Warning
This method is much slower than the native expressions API. Only use it if you cannot implement your logic otherwise.
Suppose that the function is: x ↦ sqrt(x)
:
- For mapping elements of a series, consider:
s.sqrt()
. - For mapping inner elements of lists, consider:
s.list.eval(pl.element().sqrt())
. - For mapping elements of struct fields, consider:
s.struct.field("field_name").sqrt()
.
If the function returns a different datatype, the return_dtype arg should be set, otherwise the method will fail.
Implementing logic using a Python function is almost always _significantly_slower and more memory intensive than implementing the same logic using the native expression API because:
- The native expression engine runs in Rust; UDFs run in Python.
- Use of Python UDFs forces the DataFrame to be materialized in memory.
- Polars-native expressions can be parallelised (UDFs typically cannot).
- Polars-native expressions can be logically optimised (UDFs cannot).
Wherever possible you should strongly prefer the native expression API to achieve the best performance.
Parameters:
function
Custom function or lambda.
return_dtype
Output datatype. If not set, the dtype will be inferred based on the first non-null value that is returned by the function.
skip_nulls
Nulls will be skipped and not passed to the python function. This is faster because python can be skipped and because we call more specialized functions.
Returns:
Series
Warning
If return_dtype
is not provided, this may lead to unexpected results. We allow this, but it is considered a bug in the user’s query.
Notes
If your function is expensive and you don’t want it to be called more than once for a given input, consider applying an @lru_cache
decorator to it. If your data is suitable you may achieve significant speedups.
Examples
s = pl.Series("a", [1, 2, 3]) s.map_elements(lambda x: x + 10, return_dtype=pl.Int64)
shape: (3,) Series: 'a' [i64] [ 11 12 13 ]
max() → PythonLiteral | None[source]
Get the maximum value in this Series.
Examples
s = pl.Series("a", [1, 2, 3]) s.max() 3
mean() → PythonLiteral | None[source]
Reduce this Series to the mean value.
Examples
s = pl.Series("a", [1, 2, 3]) s.mean() 2.0
median() → PythonLiteral | None[source]
Get the median of this Series.
Examples
s = pl.Series("a", [1, 2, 3]) s.median() 2.0
min() → PythonLiteral | None[source]
Get the minimal value in this Series.
Examples
s = pl.Series("a", [1, 2, 3]) s.min() 1
mode() → Series[source]
Compute the most occurring value(s).
Can return multiple Values.
Examples
s = pl.Series("a", [1, 2, 2, 3]) s.mode() shape: (1,) Series: 'a' [i64] [ 2 ]
Get the number of chunks that this Series contains.
Examples
s = pl.Series("a", [1, 2, 3]) s.n_chunks() 1 s2 = pl.Series("a", [4, 5, 6])
Concatenate Series with rechunk = True
pl.concat([s, s2], rechunk=True).n_chunks() 1
Concatenate Series with rechunk = False
pl.concat([s, s2], rechunk=False).n_chunks() 2
Count the number of unique values in this Series.
Examples
s = pl.Series("a", [1, 2, 2, 3]) s.n_unique() 3
Get the name of this Series.
Examples
s = pl.Series("a", [1, 2, 3]) s.name 'a'
nan_max() → int | float | date | datetime | timedelta | str[source]
Get maximum value, but propagate/poison encountered NaN values.
This differs from numpy’s nanmax
as numpy defaults to propagating NaN values, whereas polars defaults to ignoring them.
Examples
s = pl.Series("a", [1, 3, 4]) s.nan_max() 4
s = pl.Series("a", [1.0, float("nan"), 4.0]) s.nan_max() nan
nan_min() → int | float | date | datetime | timedelta | str[source]
Get minimum value, but propagate/poison encountered NaN values.
This differs from numpy’s nanmax
as numpy defaults to propagating NaN values, whereas polars defaults to ignoring them.
Examples
s = pl.Series("a", [1, 3, 4]) s.nan_min() 1
s = pl.Series("a", [1.0, float("nan"), 4.0]) s.nan_min() nan
ne(other: Any) → Series | Expr[source]
Method equivalent of operator expression series != other
.
ne_missing(other: Any) → Series | Expr[source]
Method equivalent of equality operator series != other
where None == None
.
This differs from the standard ne
where null values are propagated.
Parameters:
other
A literal or expression value to compare with.
Examples
s1 = pl.Series("a", [333, 200, None]) s2 = pl.Series("a", [100, 200, None]) s1.ne(s2) shape: (3,) Series: 'a' [bool] [ true false null ] s1.ne_missing(s2) shape: (3,) Series: 'a' [bool] [ true false false ]
new_from_index(index: int, length: int) → Self[source]
Create a new Series filled with values from the given index.
Examples
s = pl.Series("a", [1, 2, 3, 4, 5]) s.new_from_index(1, 3) shape: (3,) Series: 'a' [i64] [ 2 2 2 ]
not_() → Series[source]
Negate a boolean Series.
Returns:
Series
Series of data type Boolean
.
Examples
s = pl.Series("a", [True, False, False]) s.not_() shape: (3,) Series: 'a' [bool] [ false true true ]
Count the null values in this Series.
Examples
s = pl.Series([1, None, None]) s.null_count() 2
pct_change(n: int | IntoExprColumn = 1) → Series[source]
Computes percentage change between values.
Percentage change (as fraction) between current element and most-recent non-null element at least n
period(s) before the current element.
Computes the change from the previous row by default.
Parameters:
n
periods to shift for forming percent change.
Examples
pl.Series(range(10)).pct_change() shape: (10,) Series: '' [f64] [ null inf 1.0 0.5 0.333333 0.25 0.2 0.166667 0.142857 0.125 ]
pl.Series([1, 2, 4, 8, 16, 32, 64, 128, 256, 512]).pct_change(2) shape: (10,) Series: '' [f64] [ null null 3.0 3.0 3.0 3.0 3.0 3.0 3.0 3.0 ]
peak_max() → Self[source]
Get a boolean mask of the local maximum peaks.
Examples
s = pl.Series("a", [1, 2, 3, 4, 5]) s.peak_max() shape: (5,) Series: 'a' [bool] [ false false false false true ]
peak_min() → Self[source]
Get a boolean mask of the local minimum peaks.
Examples
s = pl.Series("a", [4, 1, 3, 2, 5]) s.peak_min() shape: (5,) Series: 'a' [bool] [ false true false true false ]
property plot_: SeriesPlot_[source]
Create a plot namespace.
Warning
This functionality is currently considered unstable. It may be changed at any point without it being considered a breaking change.
Changed in version 1.6.0: In prior versions of Polars, HvPlot was the plotting backend. If you would like to restore the previous plotting functionality, all you need to do is add import hvplot.polars
at the top of your script and replacedf.plot
with df.hvplot
.
Polars does not implement plotting logic itself, but instead defers to Altair:
s.plot.hist(**kwargs)
is shorthand foralt.Chart(s.to_frame()).mark_bar(tooltip=True).encode(x=alt.X(f'{s.name}:Q', bin=True), y='count()', **kwargs).interactive()
s.plot.kde(**kwargs)
is shorthand foralt.Chart(s.to_frame()).transform_density(s.name, as_=[s.name, 'density']).mark_area(tooltip=True).encode(x=s.name, y='density:Q', **kwargs).interactive()
- for any other attribute
attr
,s.plot.attr(**kwargs)
is shorthand foralt.Chart(s.to_frame().with_row_index()).mark_attr(tooltip=True).encode(x='index', y=s.name, **kwargs).interactive()
For configuration, we suggest readingChart Configuration. For example, you can:
- Change the width/height/title with
.properties(width=500, height=350, title="My amazing plot")
. - Change the x-axis label rotation with
.configure_axisX(labelAngle=30)
. - Change the opacity of the points in your scatter plot with
.configure_point(opacity=.5)
.
Examples
Histogram:
s = pl.Series([1, 4, 4, 6, 2, 4, 3, 5, 5, 7, 1]) s.plot.hist()
KDE plot:
Line plot:
pow(
exponent: int | float | Series,
) → Series[source]
Raise to the power of the given exponent.
If the exponent is float, the result follows the dtype of exponent. Otherwise, it follows dtype of base.
Parameters:
exponent
The exponent. Accepts Series input.
Examples
Raising integers to positive integers results in integers:
s = pl.Series("foo", [1, 2, 3, 4]) s.pow(3) shape: (4,) Series: 'foo' [i64] [ 1 8 27 64 ]
In order to raise integers to negative integers, you can cast either the base or the exponent to float:
s.pow(-3.0) shape: (4,) Series: 'foo' [f64] [ 1.0 0.125 0.037037 0.015625 ]
product() → int | float[source]
Reduce this Series to the product value.
Examples
s = pl.Series("a", [1, 2, 3]) s.product() 6
qcut(
quantiles: Sequence[float] | int,
*,
labels: Sequence[str] | None = None,
left_closed: bool = False,
allow_duplicates: bool = False,
include_breaks: bool = False,
) → Series[source]
Bin continuous values into discrete categories based on their quantiles.
Warning
This functionality is considered unstable. It may be changed at any point without it being considered a breaking change.
Parameters:
quantiles
Either a list of quantile probabilities between 0 and 1 or a positive integer determining the number of bins with uniform probability.
labels
Names of the categories. The number of labels must be equal to the number of cut points plus one.
left_closed
Set the intervals to be left-closed instead of right-closed.
allow_duplicates
If set to True
, duplicates in the resulting quantiles are dropped, rather than raising a DuplicateError
. This can happen even with unique probabilities, depending on the data.
include_breaks
Include a column with the right endpoint of the bin each observation falls in. This will change the data type of the output from aCategorical
to a Struct
.
Returns:
Series
Series of data type Categorical
if include_breaks
is set toFalse
(default), otherwise a Series of data type Struct
.
Examples
Divide a column into three categories according to pre-defined quantile probabilities.
s = pl.Series("foo", [-2, -1, 0, 1, 2]) s.qcut([0.25, 0.75], labels=["a", "b", "c"]) shape: (5,) Series: 'foo' [cat] [ "a" "a" "b" "b" "c" ]
Divide a column into two categories using uniform quantile probabilities.
s.qcut(2, labels=["low", "high"], left_closed=True) shape: (5,) Series: 'foo' [cat] [ "low" "low" "high" "high" "high" ]
Create a DataFrame with the breakpoint and category for each value.
cut = s.qcut([0.25, 0.75], include_breaks=True).alias("cut") s.to_frame().with_columns(cut).unnest("cut") shape: (5, 3) ┌─────┬────────────┬────────────┐ │ foo ┆ breakpoint ┆ category │ │ --- ┆ --- ┆ --- │ │ i64 ┆ f64 ┆ cat │ ╞═════╪════════════╪════════════╡ │ -2 ┆ -1.0 ┆ (-inf, -1] │ │ -1 ┆ -1.0 ┆ (-inf, -1] │ │ 0 ┆ 1.0 ┆ (-1, 1] │ │ 1 ┆ 1.0 ┆ (-1, 1] │ │ 2 ┆ inf ┆ (1, inf] │ └─────┴────────────┴────────────┘
quantile(
quantile: float,
interpolation: RollingInterpolationMethod = 'nearest',
Get the quantile value of this Series.
Parameters:
quantile
Quantile between 0.0 and 1.0.
interpolation{‘nearest’, ‘higher’, ‘lower’, ‘midpoint’, ‘linear’}
Interpolation method.
Examples
s = pl.Series("a", [1, 2, 3]) s.quantile(0.5) 2.0
rank(
method: RankMethod = 'average',
*,
descending: bool = False,
) → Series[source]
Assign ranks to data, dealing with ties appropriately.
Parameters:
method{‘average’, ‘min’, ‘max’, ‘dense’, ‘ordinal’, ‘random’}
The method used to assign ranks to tied elements. The following methods are available (default is ‘average’):
- ‘average’ : The average of the ranks that would have been assigned to all the tied values is assigned to each value.
- ‘min’ : The minimum of the ranks that would have been assigned to all the tied values is assigned to each value. (This is also referred to as “competition” ranking.)
- ‘max’ : The maximum of the ranks that would have been assigned to all the tied values is assigned to each value.
- ‘dense’ : Like ‘min’, but the rank of the next highest element is assigned the rank immediately after those assigned to the tied elements.
- ‘ordinal’ : All values are given a distinct rank, corresponding to the order that the values occur in the Series.
- ‘random’ : Like ‘ordinal’, but the rank for ties is not dependent on the order that the values occur in the Series.
descending
Rank in descending order.
seed
If method="random"
, use this as seed.
Examples
The ‘average’ method:
s = pl.Series("a", [3, 6, 1, 1, 6]) s.rank() shape: (5,) Series: 'a' [f64] [ 3.0 4.5 1.5 1.5 4.5 ]
The ‘ordinal’ method:
s = pl.Series("a", [3, 6, 1, 1, 6]) s.rank("ordinal") shape: (5,) Series: 'a' [u32] [ 3 4 1 2 5 ]
rechunk(*, in_place: bool = False) → Self[source]
Create a single chunk of memory for this Series.
Parameters:
in_place
In place or not.
Examples
s1 = pl.Series("a", [1, 2, 3]) s1.n_chunks() 1 s2 = pl.Series("a", [4, 5, 6]) s = pl.concat([s1, s2], rechunk=False) s.n_chunks() 2 s.rechunk(in_place=True) shape: (6,) Series: 'a' [i64] [ 1 2 3 4 5 6 ] s.n_chunks() 1
reinterpret(*, signed: bool = True) → Series[source]
Reinterpret the underlying bits as a signed/unsigned integer.
This operation is only allowed for 64bit integers. For lower bits integers, you can safely use that cast operation.
Parameters:
signed
If True, reinterpret as pl.Int64
. Otherwise, reinterpret as pl.UInt64
.
Examples
s = pl.Series("a", [-(2**60), -2, 3]) s shape: (3,) Series: 'a' [i64] [ -1152921504606846976 -2 3 ] s.reinterpret(signed=False) shape: (3,) Series: 'a' [u64] [ 17293822569102704640 18446744073709551614 3 ]
rename(name: str) → Series[source]
Rename this Series.
Alias for Series.alias().
Parameters:
name
New name.
Examples
s = pl.Series("a", [1, 2, 3]) s.rename("b") shape: (3,) Series: 'b' [i64] [ 1 2 3 ]
repeat_by(by: int | IntoExprColumn) → Self[source]
Repeat the elements in this Series as specified in the given expression.
The repeated elements are expanded into a List.
Parameters:
by
Numeric column that determines how often the values will be repeated. The column will be coerced to UInt32. Give this dtype to make the coercion a no-op.
Returns:
Expr
Expression of data type List, where the inner data type is equal to the original data type.
replace(old: IntoExpr | Sequence[Any] | Mapping[Any, Any], new: IntoExpr | Sequence[Any] | NoDefault = <no_default>, *, default: IntoExpr | NoDefault = <no_default>, return_dtype: PolarsDataType | None = None) → Self[source]
Replace values by different values of the same data type.
Parameters:
old
Value or sequence of values to replace. Also accepts a mapping of values to their replacement as syntactic sugar forreplace(old=Series(mapping.keys()), new=Series(mapping.values()))
.
new
Value or sequence of values to replace by. Length must match the length of old
or have length 1.
default
Set values that were not replaced to this value. Defaults to keeping the original value. Accepts expression input. Non-expression inputs are parsed as literals.
Deprecated since version 0.20.31: Use replace_all()
instead to set a default while replacing values.
return_dtype
The data type of the resulting expression. If set to None
(default), the data type is determined automatically based on the other inputs.
Deprecated since version 0.20.31: Use replace_all()
instead to set a return data type while replacing values.
Notes
The global string cache must be enabled when replacing categorical values.
Examples
Replace a single value by another value. Values that were not replaced remain unchanged.
s = pl.Series([1, 2, 2, 3]) s.replace(2, 100) shape: (4,) Series: '' [i64] [ 1 100 100 3 ]
Replace multiple values by passing sequences to the old
and new
parameters.
s.replace([2, 3], [100, 200]) shape: (4,) Series: '' [i64] [ 1 100 100 200 ]
Passing a mapping with replacements is also supported as syntactic sugar.
mapping = {2: 100, 3: 200} s.replace(mapping) shape: (4,) Series: '' [i64] [ 1 100 100 200 ]
The original data type is preserved when replacing by values of a different data type. Use replace_strict() to replace and change the return data type.
s = pl.Series(["x", "y", "z"]) mapping = {"x": 1, "y": 2, "z": 3} s.replace(mapping) shape: (3,) Series: '' [str] [ "1" "2" "3" ]
replace_strict(old: IntoExpr | Sequence[Any] | Mapping[Any, Any], new: IntoExpr | Sequence[Any] | NoDefault = <no_default>, *, default: IntoExpr | NoDefault = <no_default>, return_dtype: PolarsDataType | None = None) → Self[source]
Replace all values by different values.
Parameters:
old
Value or sequence of values to replace. Also accepts a mapping of values to their replacement as syntactic sugar forreplace_all(old=Series(mapping.keys()), new=Series(mapping.values()))
.
new
Value or sequence of values to replace by. Length must match the length of old
or have length 1.
default
Set values that were not replaced to this value. If no default is specified, (default), an error is raised if any values were not replaced. Accepts expression input. Non-expression inputs are parsed as literals.
return_dtype
The data type of the resulting Series. If set to None
(default), the data type is determined automatically based on the other inputs.
Raises:
InvalidOperationError
If any non-null values in the original column were not replaced, and nodefault
was specified.
Notes
The global string cache must be enabled when replacing categorical values.
Examples
Replace values by passing sequences to the old
and new
parameters.
s = pl.Series([1, 2, 2, 3]) s.replace_strict([1, 2, 3], [100, 200, 300]) shape: (4,) Series: '' [i64] [ 100 200 200 300 ]
Passing a mapping with replacements is also supported as syntactic sugar.
mapping = {1: 100, 2: 200, 3: 300} s.replace_strict(mapping) shape: (4,) Series: '' [i64] [ 100 200 200 300 ]
By default, an error is raised if any non-null values were not replaced. Specify a default to set all values that were not matched.
mapping = {2: 200, 3: 300} s.replace_strict(mapping)
Traceback (most recent call last): ... polars.exceptions.InvalidOperationError: incomplete mapping specified forreplace_strict
s.replace_strict(mapping, default=-1) shape: (4,) Series: '' [i64] [ -1 200 200 300 ]
The default can be another Series.
default = pl.Series([2.5, 5.0, 7.5, 10.0]) s.replace_strict(2, 200, default=default) shape: (4,) Series: '' [f64] [ 2.5 200.0 200.0 10.0 ]
Replacing by values of a different data type sets the return type based on a combination of the new
data type and the default
data type.
s = pl.Series(["x", "y", "z"]) mapping = {"x": 1, "y": 2, "z": 3} s.replace_strict(mapping) shape: (3,) Series: '' [i64] [ 1 2 3 ] s.replace_strict(mapping, default="x") shape: (3,) Series: '' [str] [ "1" "2" "3" ]
Set the return_dtype
parameter to control the resulting data type directly.
s.replace_strict(mapping, return_dtype=pl.UInt8) shape: (3,) Series: '' [u8] [ 1 2 3 ]
reshape(dimensions: tuple[int, ...]) → Series[source]
Reshape this Series to a flat Series or an Array Series.
Parameters:
dimensions
Tuple of the dimension sizes. If a -1 is used in any of the dimensions, that dimension is inferred.
Returns:
Series
If a single dimension is given, results in a Series of the original data type. If a multiple dimensions are given, results in a Series of data typeArray
with shape dimensions
.
Examples
s = pl.Series("foo", [1, 2, 3, 4, 5, 6, 7, 8, 9]) square = s.reshape((3, 3)) square shape: (3,) Series: 'foo' [array[i64, 3]] [ [1, 2, 3] [4, 5, 6] [7, 8, 9] ] square.reshape((9,)) shape: (9,) Series: 'foo' [i64] [ 1 2 3 4 5 6 7 8 9 ]
reverse() → Series[source]
Return Series in reverse order.
Examples
s = pl.Series("a", [1, 2, 3], dtype=pl.Int8) s.reverse() shape: (3,) Series: 'a' [i8] [ 3 2 1 ]
rle() → Series[source]
Compress the Series data using run-length encoding.
Run-length encoding (RLE) encodes data by storing each run of identical values as a single value and its length.
Returns:
Series
Series of data type Struct
with fields len
of data type UInt32
and value
of the original data type.
Examples
s = pl.Series("s", [1, 1, 2, 1, None, 1, 3, 3]) s.rle().struct.unnest() shape: (6, 2) ┌─────┬───────┐ │ len ┆ value │ │ --- ┆ --- │ │ u32 ┆ i64 │ ╞═════╪═══════╡ │ 2 ┆ 1 │ │ 1 ┆ 2 │ │ 1 ┆ 1 │ │ 1 ┆ null │ │ 1 ┆ 1 │ │ 2 ┆ 3 │ └─────┴───────┘
rle_id() → Series[source]
Get a distinct integer ID for each run of identical values.
The ID starts at 0 and increases by one each time the value of the column changes.
Returns:
Series
Series of data type UInt32
.
Notes
This functionality is especially useful for defining a new group for every time a column’s value changes, rather than for every distinct value of that column.
Examples
s = pl.Series("s", [1, 1, 2, 1, None, 1, 3, 3]) s.rle_id() shape: (8,) Series: 's' [u32] [ 0 0 1 2 3 4 5 5 ]
rolling_kurtosis(
window_size: int,
*,
fisher: bool = True,
bias: bool = True,
min_samples: int | None = None,
center: bool = False,
) → Series[source]
Compute a rolling kurtosis.
Warning
This functionality is considered unstable. It may be changed at any point without it being considered a breaking change.
The window at a given row will include the row itself, and the window_size - 1
elements before it.
Parameters:
window_size
Integer size of the rolling window.
fisherbool, optional
If True, Fisher’s definition is used (normal ==> 0.0). If False, Pearson’s definition is used (normal ==> 3.0).
biasbool, optional
If False, the calculations are corrected for statistical bias.
min_samples
The number of values in the window that should be non-null before computing a result. If set to None
(default), it will be set equal to window_size
.
center
Set the labels at the center of the window.
Examples
pl.Series([1, 4, 2, 9]).rolling_kurtosis(3) shape: (4,) Series: '' [f64] [ null null -1.5 -1.5 ]
rolling_map(
function: Callable[[Series], Any],
window_size: int,
weights: list[float] | None = None,
*,
min_samples: int | None = None,
center: bool = False,
) → Series[source]
Compute a custom rolling window function.
Warning
This functionality is considered unstable. It may be changed at any point without it being considered a breaking change.
Parameters:
function
Custom aggregation function.
window_size
The length of the window in number of elements.
weights
An optional slice with the same length as the window that will be multiplied elementwise with the values in the window.
min_samples
The number of values in the window that should be non-null before computing a result. If set to None
(default), it will be set equal to window_size
.
center
Set the labels at the center of the window.
Warning
Computing custom functions is extremely slow. Use specialized rolling functions such as Series.rolling_sum() if at all possible.
Examples
from numpy import nansum s = pl.Series([11.0, 2.0, 9.0, float("nan"), 8.0]) s.rolling_map(nansum, window_size=3) shape: (5,) Series: '' [f64] [ null null 22.0 11.0 17.0 ]
rolling_max(
window_size: int,
weights: list[float] | None = None,
*,
min_samples: int | None = None,
center: bool = False,
) → Series[source]
Apply a rolling max (moving max) over the values in this array.
A window of length window_size
will traverse the array. The values that fill this window will (optionally) be multiplied with the weights given by theweight
vector. The resulting values will be aggregated to their max.
The window at a given row will include the row itself and the window_size - 1
elements before it.
Parameters:
window_size
The length of the window in number of elements.
weights
An optional slice with the same length as the window that will be multiplied elementwise with the values in the window.
min_samples
The number of values in the window that should be non-null before computing a result. If set to None
(default), it will be set equal to window_size
.
center
Set the labels at the center of the window.
Examples
s = pl.Series("a", [100, 200, 300, 400, 500]) s.rolling_max(window_size=2) shape: (5,) Series: 'a' [i64] [ null 200 300 400 500 ]
rolling_mean(
window_size: int,
weights: list[float] | None = None,
*,
min_samples: int | None = None,
center: bool = False,
) → Series[source]
Apply a rolling mean (moving mean) over the values in this array.
A window of length window_size
will traverse the array. The values that fill this window will (optionally) be multiplied with the weights given by theweight
vector. The resulting values will be aggregated to their mean.
The window at a given row will include the row itself and the window_size - 1
elements before it.
Parameters:
window_size
The length of the window in number of elements.
weights
An optional slice with the same length as the window that will be multiplied elementwise with the values in the window.
min_samples
The number of values in the window that should be non-null before computing a result. If set to None
(default), it will be set equal to window_size
.
center
Set the labels at the center of the window.
Examples
s = pl.Series("a", [100, 200, 300, 400, 500]) s.rolling_mean(window_size=2) shape: (5,) Series: 'a' [f64] [ null 150.0 250.0 350.0 450.0 ]
rolling_median(
window_size: int,
weights: list[float] | None = None,
*,
min_samples: int | None = None,
center: bool = False,
) → Series[source]
Compute a rolling median.
Warning
This functionality is considered unstable. It may be changed at any point without it being considered a breaking change.
The window at a given row will include the row itself and the window_size - 1
elements before it.
Parameters:
window_size
The length of the window in number of elements.
weights
An optional slice with the same length as the window that will be multiplied elementwise with the values in the window.
min_samples
The number of values in the window that should be non-null before computing a result. If set to None
(default), it will be set equal to window_size
.
center
Set the labels at the center of the window.
Examples
s = pl.Series("a", [1.0, 2.0, 3.0, 4.0, 6.0, 8.0]) s.rolling_median(window_size=3) shape: (6,) Series: 'a' [f64] [ null null 2.0 3.0 4.0 6.0 ]
rolling_min(
window_size: int,
weights: list[float] | None = None,
*,
min_samples: int | None = None,
center: bool = False,
) → Series[source]
Apply a rolling min (moving min) over the values in this array.
A window of length window_size
will traverse the array. The values that fill this window will (optionally) be multiplied with the weights given by theweight
vector. The resulting values will be aggregated to their min.
The window at a given row will include the row itself and the window_size - 1
elements before it.
Parameters:
window_size
The length of the window in number of elements.
weights
An optional slice with the same length as the window that will be multiplied elementwise with the values in the window.
min_samples
The number of values in the window that should be non-null before computing a result. If set to None
(default), it will be set equal to window_size
.
center
Set the labels at the center of the window.
Examples
s = pl.Series("a", [100, 200, 300, 400, 500]) s.rolling_min(window_size=3) shape: (5,) Series: 'a' [i64] [ null null 100 200 300 ]
rolling_quantile(
quantile: float,
interpolation: RollingInterpolationMethod = 'nearest',
window_size: int = 2,
weights: list[float] | None = None,
*,
min_samples: int | None = None,
center: bool = False,
) → Series[source]
Compute a rolling quantile.
Warning
This functionality is considered unstable. It may be changed at any point without it being considered a breaking change.
The window at a given row will include the row itself and the window_size - 1
elements before it.
Parameters:
quantile
Quantile between 0.0 and 1.0.
interpolation{‘nearest’, ‘higher’, ‘lower’, ‘midpoint’, ‘linear’}
Interpolation method.
window_size
The length of the window in number of elements.
weights
An optional slice with the same length as the window that will be multiplied elementwise with the values in the window.
min_samples
The number of values in the window that should be non-null before computing a result. If set to None
(default), it will be set equal to window_size
.
center
Set the labels at the center of the window.
Examples
s = pl.Series("a", [1.0, 2.0, 3.0, 4.0, 6.0, 8.0]) s.rolling_quantile(quantile=0.33, window_size=3) shape: (6,) Series: 'a' [f64] [ null null 1.0 2.0 3.0 4.0 ] s.rolling_quantile(quantile=0.33, interpolation="linear", window_size=3) shape: (6,) Series: 'a' [f64] [ null null 1.66 2.66 3.66 5.32 ]
rolling_skew(
window_size: int,
*,
bias: bool = True,
min_samples: int | None = None,
center: bool = False,
) → Series[source]
Compute a rolling skew.
Warning
This functionality is considered unstable. It may be changed at any point without it being considered a breaking change.
The window at a given row includes the row itself and thewindow_size - 1
elements before it.
Parameters:
window_size
Integer size of the rolling window.
bias
If False, the calculations are corrected for statistical bias.
min_samples
The number of values in the window that should be non-null before computing a result. If set to None
(default), it will be set equal to window_size
.
center
Set the labels at the center of the window.
Examples
pl.Series([1, 4, 2, 9]).rolling_skew(3) shape: (4,) Series: '' [f64] [ null null 0.381802 0.47033 ]
Note how the values match
pl.Series([1, 4, 2]).skew(), pl.Series([4, 2, 9]).skew() (0.38180177416060584, 0.47033046033698594)
rolling_std(
window_size: int,
weights: list[float] | None = None,
*,
min_samples: int | None = None,
center: bool = False,
ddof: int = 1,
) → Series[source]
Compute a rolling std dev.
A window of length window_size
will traverse the array. The values that fill this window will (optionally) be multiplied with the weights given by theweight
vector. The resulting values will be aggregated to their std dev.
The window at a given row will include the row itself and the window_size - 1
elements before it.
Parameters:
window_size
The length of the window in number of elements.
weights
An optional slice with the same length as the window that will be multiplied elementwise with the values in the window.
min_samples
The number of values in the window that should be non-null before computing a result. If set to None
(default), it will be set equal to window_size
.
center
Set the labels at the center of the window.
ddof
“Delta Degrees of Freedom”: The divisor for a length N window is N - ddof
Examples
s = pl.Series("a", [1.0, 2.0, 3.0, 4.0, 6.0, 8.0]) s.rolling_std(window_size=3) shape: (6,) Series: 'a' [f64] [ null null 1.0 1.0 1.527525 2.0 ]
rolling_sum(
window_size: int,
weights: list[float] | None = None,
*,
min_samples: int | None = None,
center: bool = False,
) → Series[source]
Apply a rolling sum (moving sum) over the values in this array.
A window of length window_size
will traverse the array. The values that fill this window will (optionally) be multiplied with the weights given by theweight
vector. The resulting values will be aggregated to their sum.
The window at a given row will include the row itself and the window_size - 1
elements before it.
Parameters:
window_size
The length of the window in number of elements.
weights
An optional slice with the same length as the window that will be multiplied elementwise with the values in the window.
min_samples
The number of values in the window that should be non-null before computing a result. If set to None
(default), it will be set equal to window_size
.
center
Set the labels at the center of the window.
Examples
s = pl.Series("a", [1, 2, 3, 4, 5]) s.rolling_sum(window_size=2) shape: (5,) Series: 'a' [i64] [ null 3 5 7 9 ]
rolling_var(
window_size: int,
weights: list[float] | None = None,
*,
min_samples: int | None = None,
center: bool = False,
ddof: int = 1,
) → Series[source]
Compute a rolling variance.
A window of length window_size
will traverse the array. The values that fill this window will (optionally) be multiplied with the weights given by theweight
vector. The resulting values will be aggregated to their variance.
The window at a given row will include the row itself and the window_size - 1
elements before it.
Parameters:
window_size
The length of the window in number of elements.
weights
An optional slice with the same length as the window that will be multiplied elementwise with the values in the window.
min_samples
The number of values in the window that should be non-null before computing a result. If set to None
(default), it will be set equal to window_size
.
center
Set the labels at the center of the window.
ddof
“Delta Degrees of Freedom”: The divisor for a length N window is N - ddof
Examples
s = pl.Series("a", [1.0, 2.0, 3.0, 4.0, 6.0, 8.0]) s.rolling_var(window_size=3) shape: (6,) Series: 'a' [f64] [ null null 1.0 1.0 2.333333 4.0 ]
round(decimals: int = 0, mode: RoundMode = 'half_to_even') → Series[source]
Round underlying floating point data by decimals
digits.
Parameters:
decimals
number of decimals to round by.
Examples
s = pl.Series("a", [1.12345, 2.56789, 3.901234]) s.round(2) shape: (3,) Series: 'a' [f64] [ 1.12 2.57 3.9 ]
round_sig_figs(digits: int) → Series[source]
Round to a number of significant figures.
Parameters:
digits
Number of significant figures to round to.
Examples
s = pl.Series([0.01234, 3.333, 3450.0]) s.round_sig_figs(2) shape: (3,) Series: '' [f64] [ 0.012 3.3 3500.0 ]
sample(
*,
fraction: float | None = None,
with_replacement: bool = False,
shuffle: bool = False,
) → Series[source]
Sample from this Series.
Parameters:
n
Number of items to return. Cannot be used with fraction
. Defaults to 1 iffraction
is None.
fraction
Fraction of items to return. Cannot be used with n
.
with_replacement
Allow values to be sampled more than once.
shuffle
Shuffle the order of sampled data points.
seed
Seed for the random number generator. If set to None (default), a random seed is generated for each sample operation.
Examples
s = pl.Series("a", [1, 2, 3, 4, 5]) s.sample(2, seed=0)
shape: (2,) Series: 'a' [i64] [ 1 5 ]
scatter(
indices: Series | Iterable[int] | int | np.ndarray[Any, Any],
values: Series | Iterable[PythonLiteral] | PythonLiteral | None,
) → Series[source]
Set values at the index locations.
Parameters:
indices
Integers representing the index locations.
values
Replacement values.
Notes
Use of this function is frequently an anti-pattern, as it can block optimization (predicate pushdown, etc). Consider usingpl.when(predicate).then(value).otherwise(self)
instead.
Examples
s = pl.Series("a", [1, 2, 3]) s.scatter(1, 10) shape: (3,) Series: 'a' [i64] [ 1 10 3 ]
It is better to implement this as follows:
s.to_frame().with_row_index().select( ... pl.when(pl.col("index") == 1).then(10).otherwise(pl.col("a")) ... ) shape: (3, 1) ┌─────────┐ │ literal │ │ --- │ │ i64 │ ╞═════════╡ │ 1 │ │ 10 │ │ 3 │ └─────────┘
search_sorted(
element: IntoExpr | np.ndarray[Any, Any] | None,
side: SearchSortedSide = 'any',
Find indices where elements should be inserted to maintain order.
\[a[i-1] < v <= a[i]\]
Parameters:
element
Expression or scalar value.
side{‘any’, ‘left’, ‘right’}
If ‘any’, the index of the first suitable location found is given. If ‘left’, the index of the leftmost suitable location found is given. If ‘right’, return the rightmost suitable location found is given.
Examples
s = pl.Series("set", [1, 2, 3, 4, 4, 5, 6, 7]) s.search_sorted(4) 3 s.search_sorted(4, "left") 3 s.search_sorted(4, "right") 5 s.search_sorted([1, 4, 5]) shape: (3,) Series: 'set' [u32] [ 0 3 5 ] s.search_sorted([1, 4, 5], "left") shape: (3,) Series: 'set' [u32] [ 0 3 5 ] s.search_sorted([1, 4, 5], "right") shape: (3,) Series: 'set' [u32] [ 1 5 6 ]
set(
filter: Series,
value: int | float | str | bool | None,
) → Series[source]
Set masked values.
Parameters:
filter
Boolean mask.
value
Value with which to replace the masked values.
Notes
Use of this function is frequently an anti-pattern, as it can block optimisation (predicate pushdown, etc). Consider usingpl.when(predicate).then(value).otherwise(self)
instead.
Examples
s = pl.Series("a", [1, 2, 3]) s.set(s == 2, 10) shape: (3,) Series: 'a' [i64] [ 1 10 3 ]
It is better to implement this as follows:
s.to_frame().select( ... pl.when(pl.col("a") == 2).then(10).otherwise(pl.col("a")) ... ) shape: (3, 1) ┌─────────┐ │ literal │ │ --- │ │ i64 │ ╞═════════╡ │ 1 │ │ 10 │ │ 3 │ └─────────┘
set_sorted(*, descending: bool = False) → Self[source]
Flags the Series as ‘sorted’.
Enables downstream code to user fast paths for sorted arrays.
Parameters:
descending
If the Series
order is descending.
Warning
This can lead to incorrect results if this Series
is not sorted!! Use with care!
Examples
s = pl.Series("a", [1, 2, 3]) s.set_sorted().max() 3
property shape_: tuple[int]_[source]
Shape of this Series.
Examples
s = pl.Series("a", [1, 2, 3]) s.shape (3,)
shift(n: int = 1, *, fill_value: IntoExpr | None = None) → Series[source]
Shift values by the given number of indices.
Parameters:
n
Number of indices to shift forward. If a negative value is passed, values are shifted in the opposite direction instead.
fill_value
Fill the resulting null values with this value. Accepts scalar expression input. Non-expression inputs are parsed as literals.
Notes
This method is similar to the LAG
operation in SQL when the value for n
is positive. With a negative value for n
, it is similar to LEAD
.
Examples
By default, values are shifted forward by one index.
s = pl.Series([1, 2, 3, 4]) s.shift() shape: (4,) Series: '' [i64] [ null 1 2 3 ]
Pass a negative value to shift in the opposite direction instead.
s.shift(-2) shape: (4,) Series: '' [i64] [ 3 4 null null ]
Specify fill_value
to fill the resulting null values.
s.shift(-2, fill_value=100) shape: (4,) Series: '' [i64] [ 3 4 100 100 ]
shrink_dtype() → Series[source]
Shrink numeric columns to the minimal required datatype.
Shrink to the dtype needed to fit the extrema of this [Series
]. This can be used to reduce memory pressure.
Examples
s = pl.Series("a", [1, 2, 3, 4, 5, 6]) s shape: (6,) Series: 'a' [i64] [ 1 2 3 4 5 6 ] s.shrink_dtype() shape: (6,) Series: 'a' [i8] [ 1 2 3 4 5 6 ]
shrink_to_fit(*, in_place: bool = False) → Series[source]
Shrink Series memory usage.
Shrinks the underlying array capacity to exactly fit the actual data. (Note that this function does not change the Series data type).
shuffle(seed: int | None = None) → Series[source]
Shuffle the contents of this Series.
Parameters:
seed
Seed for the random number generator. If set to None (default), a random seed is generated each time the shuffle is called.
Examples
s = pl.Series("a", [1, 2, 3]) s.shuffle(seed=1) shape: (3,) Series: 'a' [i64] [ 2 1 3 ]
sign() → Series[source]
Compute the element-wise sign function on numeric types.
The returned value is computed as follows:
- -1 if x < 0.
- 1 if x > 0.
- x otherwise (typically 0, but could be NaN if the input is).
Null values are preserved as-is, and the dtype of the input is preserved.
Examples
s = pl.Series("a", [-9.0, -0.0, 0.0, 4.0, float("nan"), None]) s.sign() shape: (6,) Series: 'a' [f64] [ -1.0 -0.0 0.0 1.0 NaN null ]
sin() → Series[source]
Compute the element-wise value for the sine.
Examples
import math s = pl.Series("a", [0.0, math.pi / 2.0, math.pi]) s.sin() shape: (3,) Series: 'a' [f64] [ 0.0 1.0 1.2246e-16 ]
sinh() → Series[source]
Compute the element-wise value for the hyperbolic sine.
Examples
s = pl.Series("a", [1.0, 0.0, -1.0]) s.sinh() shape: (3,) Series: 'a' [f64] [ 1.175201 0.0 -1.175201 ]
skew(*, bias: bool = True) → float | None[source]
Compute the sample skewness of a data set.
For normally distributed data, the skewness should be about zero. For unimodal continuous distributions, a skewness value greater than zero means that there is more weight in the right tail of the distribution. The function skewtest
can be used to determine if the skewness value is close enough to zero, statistically speaking.
See scipy.stats for more information.
Parameters:
biasbool, optional
If False, the calculations are corrected for statistical bias.
Notes
The sample skewness is computed as the Fisher-Pearson coefficient of skewness, i.e.
\[g_1=\frac{m_3}{m_2^{3/2}}\]
where
\[m_i=\frac{1}{N}\sum_{n=1}^N(x[n]-\bar{x})^i\]
is the biased sample \(i\texttt{th}\) central moment, and\(\bar{x}\) is the sample mean. If bias
is False, the calculations are corrected for bias and the value computed is the adjusted Fisher-Pearson standardized moment coefficient, i.e.
\[G_1 = \frac{k_3}{k_2^{3/2}} = \frac{\sqrt{N(N-1)}}{N-2}\frac{m_3}{m_2^{3/2}}\]
Examples
s = pl.Series([1, 2, 2, 4, 5]) s.skew() 0.34776706224699483
slice(offset: int, length: int | None = None) → Series[source]
Get a slice of this Series.
Parameters:
offset
Start index. Negative indexing is supported.
length
Length of the slice. If set to None
, all rows starting at the offset will be selected.
Examples
s = pl.Series("a", [1, 2, 3, 4]) s.slice(1, 2) shape: (2,) Series: 'a' [i64] [ 2 3 ]
sort(
*,
descending: bool = False,
nulls_last: bool = False,
multithreaded: bool = True,
in_place: bool = False,
) → Self[source]
Sort this Series.
Parameters:
descending
Sort in descending order.
nulls_last
Place null values last instead of first.
multithreaded
Sort using multiple threads.
in_place
Sort in-place.
Examples
s = pl.Series("a", [1, 3, 4, 2]) s.sort() shape: (4,) Series: 'a' [i64] [ 1 2 3 4 ] s.sort(descending=True) shape: (4,) Series: 'a' [i64] [ 4 3 2 1 ]
sqrt() → Series[source]
Compute the square root of the elements.
Syntactic sugar for
pl.Series([1, 2]) ** 0.5 shape: (2,) Series: '' [f64] [ 1.0 1.414214 ]
Examples
s = pl.Series([1, 2, 3]) s.sqrt() shape: (3,) Series: '' [f64] [ 1.0 1.414214 1.732051 ]
std(ddof: int = 1) → float | timedelta | None[source]
Get the standard deviation of this Series.
Parameters:
ddof
“Delta Degrees of Freedom”: the divisor used in the calculation is N - ddof, where N represents the number of elements. By default ddof is 1.
Examples
s = pl.Series("a", [1, 2, 3]) s.std() 1.0
Reduce this Series to the sum value.
Notes
Dtypes in {Int8, UInt8, Int16, UInt16} are cast to Int64 before summing to prevent overflow issues.
Examples
s = pl.Series("a", [1, 2, 3]) s.sum() 6
tail(n: int = 10) → Series[source]
Get the last n
elements.
Parameters:
n
Number of elements to return. If a negative value is passed, return all elements except the first abs(n)
.
Examples
s = pl.Series("a", [1, 2, 3, 4, 5]) s.tail(3) shape: (3,) Series: 'a' [i64] [ 3 4 5 ]
Pass a negative value to get all rows except
the first abs(n)
.
s.tail(-3) shape: (2,) Series: 'a' [i64] [ 4 5 ]
tan() → Series[source]
Compute the element-wise value for the tangent.
Examples
import math s = pl.Series("a", [0.0, math.pi / 2.0, math.pi]) s.tan() shape: (3,) Series: 'a' [f64] [ 0.0 1.6331e16 -1.2246e-16 ]
tanh() → Series[source]
Compute the element-wise value for the hyperbolic tangent.
Examples
s = pl.Series("a", [1.0, 0.0, -1.0]) s.tanh() shape: (3,) Series: 'a' [f64] [ 0.761594 0.0 -0.761594 ]
to_arrow(
*,
compat_level: CompatLevel | None = None,
Return the underlying Arrow array.
If the Series contains only a single chunk this operation is zero copy.
Parameters:
compat_level
Use a specific compatibility level when exporting Polars’ internal data structures.
Examples
s = pl.Series("a", [1, 2, 3]) s = s.to_arrow() s
<pyarrow.lib.Int64Array object at ...> [ 1, 2, 3 ]
to_dummies(*, separator: str = '_', drop_first: bool = False) → DataFrame[source]
Get dummy/indicator variables.
Parameters:
separator
Separator/delimiter used when generating column names.
drop_first
Remove the first category from the variable being encoded.
Examples
s = pl.Series("a", [1, 2, 3]) s.to_dummies() shape: (3, 3) ┌─────┬─────┬─────┐ │ a_1 ┆ a_2 ┆ a_3 │ │ --- ┆ --- ┆ --- │ │ u8 ┆ u8 ┆ u8 │ ╞═════╪═════╪═════╡ │ 1 ┆ 0 ┆ 0 │ │ 0 ┆ 1 ┆ 0 │ │ 0 ┆ 0 ┆ 1 │ └─────┴─────┴─────┘
s.to_dummies(drop_first=True) shape: (3, 2) ┌─────┬─────┐ │ a_2 ┆ a_3 │ │ --- ┆ --- │ │ u8 ┆ u8 │ ╞═════╪═════╡ │ 0 ┆ 0 │ │ 1 ┆ 0 │ │ 0 ┆ 1 │ └─────┴─────┘
to_frame(name: str | None = None) → DataFrame[source]
Cast this Series to a DataFrame.
Parameters:
name
optionally name/rename the Series column in the new DataFrame.
Examples
s = pl.Series("a", [123, 456]) df = s.to_frame() df shape: (2, 1) ┌─────┐ │ a │ │ --- │ │ i64 │ ╞═════╡ │ 123 │ │ 456 │ └─────┘
df = s.to_frame("xyz") df shape: (2, 1) ┌─────┐ │ xyz │ │ --- │ │ i64 │ ╞═════╡ │ 123 │ │ 456 │ └─────┘
to_init_repr(n: int = 1000) → str[source]
Convert Series to instantiable string representation.
Parameters:
n
Only use first n elements.
Examples
s = pl.Series("a", [1, 2, None, 4], dtype=pl.Int16) print(s.to_init_repr()) pl.Series('a', [1, 2, None, 4], dtype=pl.Int16) s_from_str_repr = eval(s.to_init_repr()) s_from_str_repr shape: (4,) Series: 'a' [i16] [ 1 2 null 4 ]
to_jax(device: jax.Device | str | None = None) → jax.Array[source]
Convert this Series to a Jax Array.
Added in version 0.20.27.
Warning
This functionality is currently considered unstable. It may be changed at any point without it being considered a breaking change.
Parameters:
device
Specify the jax Device
on which the array will be created; can provide a string (such as “cpu”, “gpu”, or “tpu”) in which case the device is retrieved as jax.devices(string)[0]
. For more specific control you can supply the instantiated Device
directly. If None, arrays are created on the default device.
Examples
s = pl.Series("x", [10.5, 0.0, -10.0, 5.5]) s.to_jax() Array([ 10.5, 0. , -10. , 5.5], dtype=float32)
Convert this Series to a Python list.
This operation copies data.
Examples
s = pl.Series("a", [1, 2, 3]) s.to_list() [1, 2, 3] type(s.to_list()) <class 'list'>
to_numpy(
*,
writable: bool = False,
allow_copy: bool = True,
use_pyarrow: bool | None = None,
zero_copy_only: bool | None = None,
Convert this Series to a NumPy ndarray.
This operation copies data only when necessary. The conversion is zero copy when all of the following hold:
- The data type is an integer, float,
Datetime
,Duration
, orArray
. - The Series contains no null values.
- The Series consists of a single chunk.
- The
writable
parameter is set toFalse
(default).
Parameters:
writable
Ensure the resulting array is writable. This will force a copy of the data if the array was created without copy as the underlying Arrow data is immutable.
allow_copy
Allow memory to be copied to perform the conversion. If set to False
, causes conversions that are not zero-copy to fail.
use_pyarrow
First convert to PyArrow, then call pyarrow.Array.to_numpyto convert to NumPy. If set to False
, Polars’ own conversion logic is used.
Deprecated since version 0.20.28: Polars now uses its native engine by default for conversion to NumPy. To use PyArrow’s engine, call .to_arrow().to_numpy()
instead.
zero_copy_only
Raise an exception if the conversion to a NumPy would require copying the underlying data. Data copy occurs, for example, when the Series contains nulls or non-numeric types.
Deprecated since version 0.20.10: Use the allow_copy
parameter instead, which is the inverse of this one.
Examples
Numeric data without nulls can be converted without copying data. The resulting array will not be writable.
s = pl.Series([1, 2, 3], dtype=pl.Int8) arr = s.to_numpy() arr array([1, 2, 3], dtype=int8) arr.flags.writeable False
Set writable=True
to force data copy to make the array writable.
s.to_numpy(writable=True).flags.writeable True
Integer Series containing nulls will be cast to a float type with nan
representing a null value. This requires data to be copied.
s = pl.Series([1, 2, None], dtype=pl.UInt16) s.to_numpy() array([ 1., 2., nan], dtype=float32)
Set allow_copy=False
to raise an error if data would be copied.
s.to_numpy(allow_copy=False)
Traceback (most recent call last): ... RuntimeError: copy not allowed: cannot convert to a NumPy array without copying data
Series of data type Array
and Struct
will result in an array with more than one dimension.
s = pl.Series([[1, 2, 3], [4, 5, 6]], dtype=pl.Array(pl.Int64, 3)) s.to_numpy() array([[1, 2, 3], [4, 5, 6]])
to_pandas(
*,
use_pyarrow_extension_array: bool = False,
**kwargs: Any,
) → pd.Series[Any][source]
Convert this Series to a pandas Series.
This operation copies data if use_pyarrow_extension_array
is not enabled.
Parameters:
use_pyarrow_extension_array
Use a PyArrow-backed extension array instead of a NumPy array for the pandas Series. This allows zero copy operations and preservation of null values. Subsequent operations on the resulting pandas Series may trigger conversion to NumPy if those operations are not supported by PyArrow compute functions.
**kwargs
Additional keyword arguments to be passed topyarrow.Array.to_pandas().
Returns:
Notes
This operation requires that both pandas and pyarrow
are installed.
Examples
s = pl.Series("a", [1, 2, 3]) s.to_pandas() 0 1 1 2 2 3 Name: a, dtype: int64
Null values are converted to NaN
.
s = pl.Series("b", [1, 2, None]) s.to_pandas() 0 1.0 1 2.0 2 NaN Name: b, dtype: float64
Pass use_pyarrow_extension_array=True
to get a pandas Series backed by a PyArrow extension array. This will preserve null values.
s.to_pandas(use_pyarrow_extension_array=True) 0 1 1 2 2 Name: b, dtype: int64[pyarrow]
to_physical() → Series[source]
Cast to physical representation of the logical dtype.
- polars.datatypes.Date() -> polars.datatypes.Int32()
- polars.datatypes.Datetime() -> polars.datatypes.Int64()
- polars.datatypes.Time() -> polars.datatypes.Int64()
- polars.datatypes.Duration() -> polars.datatypes.Int64()
- polars.datatypes.Categorical() -> polars.datatypes.UInt32()
List(inner)
->List(physical of inner)
Array(inner)
->Array(physical of inner)
Struct(fields)
->Struct(physical of fields)
- Other data types will be left unchanged.
Warning
The physical representations are an implementation detail and not guaranteed to be stable.
Examples
Replicating the pandaspd.Series.factorizemethod.
s = pl.Series("values", ["a", None, "x", "a"]) s.cast(pl.Categorical).to_physical() shape: (4,) Series: 'values' [u32] [ 0 null 1 0 ]
to_torch() → torch.Tensor[source]
Convert this Series to a PyTorch Tensor.
Added in version 0.20.23.
Warning
This functionality is currently considered unstable. It may be changed at any point without it being considered a breaking change.
Notes
PyTorch tensors do not support UInt16, UInt32, or UInt64; these dtypes will be automatically cast to Int32, Int64, and Int64, respectively.
Examples
s = pl.Series("x", [1, 0, 1, 2, 0], dtype=pl.UInt8) s.to_torch() tensor([1, 0, 1, 2, 0], dtype=torch.uint8) s = pl.Series("x", [5.5, -10.0, 2.5], dtype=pl.Float32) s.to_torch() tensor([ 5.5000, -10.0000, 2.5000])
top_k(k: int = 5) → Series[source]
Return the k
largest elements.
Non-null elements are always preferred over null elements. The output is not guaranteed to be in any particular order, call sort() after this function if you wish the output to be sorted.
This has time complexity:
\[O(n)\]
Parameters:
k
Number of elements to return.
Examples
s = pl.Series("a", [2, 5, 1, 4, 3]) s.top_k(3) shape: (3,) Series: 'a' [i64] [ 5 4 3 ]
unique(*, maintain_order: bool = False) → Series[source]
Get unique elements in series.
Parameters:
maintain_order
Maintain order of data. This requires more work.
Examples
s = pl.Series("a", [1, 2, 2, 3]) s.unique().sort() shape: (3,) Series: 'a' [i64] [ 1 2 3 ]
unique_counts() → Series[source]
Return a count of the unique values in the order of appearance.
Examples
s = pl.Series("id", ["a", "b", "b", "c", "c", "c"]) s.unique_counts() shape: (3,) Series: 'id' [u32] [ 1 2 3 ]
upper_bound() → Self[source]
Return the upper bound of this Series’ dtype as a unit Series.
See also
return the lower bound of the given Series’ dtype.
Examples
s = pl.Series("s", [-1, 0, 1], dtype=pl.Int8) s.upper_bound() shape: (1,) Series: 's' [i8] [ 127 ]
s = pl.Series("s", [1.0, 2.5, 3.0], dtype=pl.Float64) s.upper_bound() shape: (1,) Series: 's' [f64] [ inf ]
value_counts(
*,
sort: bool = False,
parallel: bool = False,
normalize: bool = False,
) → DataFrame[source]
Count the occurrences of unique values.
Parameters:
sort
Sort the output by count in descending order. If set to False
(default), the order of the output is random.
parallel
Execute the computation in parallel.
Note
This option should likely not be enabled in a group by context, as the computation is already parallelized per group.
name
Give the resulting count column a specific name; if normalize
is True defaults to “proportion”, otherwise defaults to “count”.
normalize
If true gives relative frequencies of the unique values
Returns:
DataFrame
Mapping of unique values to their count.
Examples
s = pl.Series("color", ["red", "blue", "red", "green", "blue", "blue"]) s.value_counts()
shape: (3, 2) ┌───────┬───────┐ │ color ┆ count │ │ --- ┆ --- │ │ str ┆ u32 │ ╞═══════╪═══════╡ │ red ┆ 2 │ │ green ┆ 1 │ │ blue ┆ 3 │ └───────┴───────┘
Sort the output by count and customize the count column name.
s.value_counts(sort=True, name="n") shape: (3, 2) ┌───────┬─────┐ │ color ┆ n │ │ --- ┆ --- │ │ str ┆ u32 │ ╞═══════╪═════╡ │ blue ┆ 3 │ │ red ┆ 2 │ │ green ┆ 1 │ └───────┴─────┘
var(ddof: int = 1) → float | timedelta | None[source]
Get variance of this Series.
Parameters:
ddof
“Delta Degrees of Freedom”: the divisor used in the calculation is N - ddof, where N represents the number of elements. By default ddof is 1.
Examples
s = pl.Series("a", [1, 2, 3]) s.var() 1.0
zip_with(mask: Series, other: Series) → Self[source]
Take values from self or other based on the given mask.
Where mask evaluates true, take values from self. Where mask evaluates false, take values from other.
Parameters:
mask
Boolean Series.
other
Series of same type.
Returns:
Series
Examples
s1 = pl.Series([1, 2, 3, 4, 5]) s2 = pl.Series([5, 4, 3, 2, 1]) s1.zip_with(s1 < s2, s2) shape: (5,) Series: '' [i64] [ 1 2 3 2 1 ] mask = pl.Series([True, False, True, False, True]) s1.zip_with(mask, s2) shape: (5,) Series: '' [i64] [ 1 4 3 2 5 ]