polars.Expr.map_elements#

Expr.map_elements(
function: Callable[[Series], Series] | Callable[[Any], Any],
return_dtype: PolarsDataType | None = None,
*,
skip_nulls: bool = True,
pass_name: bool = False,
strategy: MapElementsStrategy = 'thread_local',
) → Self[source]#

Map a custom/user-defined function (UDF) to each element of a column.

Warning

This method is much slower than the native expressions API. Only use it if you cannot implement your logic otherwise.

The UDF is applied to each element of a column. Note that, in a GroupBy context, the column will have been pre-aggregated and so each element will itself be a Series. Therefore, depending on the context, requirements for function differ:

  • Selection

    Expects function to be of type Callable[[Any], Any]. Applies a Python function to each individual value in the column.

  • GroupBy

    Expects function to be of type Callable[[Series], Any]. For each group, applies a Python function to the slice of the column corresponding to that group.

Parameters:
function

Lambda/function to map.

return_dtype

Dtype of the output Series. If not set, the dtype will be pl.Unknown.

skip_nulls

Don’t map the function over values that contain nulls (this is faster).

pass_name

Pass the Series name to the custom function (this is more expensive).

strategy{‘thread_local’, ‘threading’}

This functionality is considered experimental and may be removed/changed.

  • ‘thread_local’: run the python function on a single thread.

  • ‘threading’: run the python function on separate threads. Use with care as this can slow performance. This might only speed up your code if the amount of work per element is significant and the python function releases the GIL (e.g. via calling a c function)

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

  • Using map_elements is strongly discouraged as you will be effectively running python “for” loops, which will be very slow. Wherever possible you should prefer the native expression API to achieve the best performance.

  • 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.

  • Window function application using over is considered a GroupBy context here, so map_elements can be used to map functions over window groups.

Examples

>>> df = pl.DataFrame(
...     {
...         "a": [1, 2, 3, 1],
...         "b": ["a", "b", "c", "c"],
...     }
... )

The function is applied to each element of column 'a':

>>> df.with_columns(  
...     pl.col("a").map_elements(lambda x: x * 2).alias("a_times_2"),
... )
shape: (4, 3)
┌─────┬─────┬───────────┐
│ a   ┆ b   ┆ a_times_2 │
│ --- ┆ --- ┆ ---       │
│ i64 ┆ str ┆ i64       │
╞═════╪═════╪═══════════╡
│ 1   ┆ a   ┆ 2         │
│ 2   ┆ b   ┆ 4         │
│ 3   ┆ c   ┆ 6         │
│ 1   ┆ c   ┆ 2         │
└─────┴─────┴───────────┘

Tip: it is better to implement this with an expression:

>>> df.with_columns(
...     (pl.col("a") * 2).alias("a_times_2"),
... )

In a GroupBy context, each element of the column is itself a Series:

>>> (
...     df.lazy().group_by("b").agg(pl.col("a")).collect()
... )  
shape: (3, 2)
┌─────┬───────────┐
│ b   ┆ a         │
│ --- ┆ ---       │
│ str ┆ list[i64] │
╞═════╪═══════════╡
│ a   ┆ [1]       │
│ b   ┆ [2]       │
│ c   ┆ [3, 1]    │
└─────┴───────────┘

Therefore, from the user’s point-of-view, the function is applied per-group:

>>> (
...     df.lazy()
...     .group_by("b")
...     .agg(pl.col("a").map_elements(lambda x: x.sum()))
...     .collect()
... )  
shape: (3, 2)
┌─────┬─────┐
│ b   ┆ a   │
│ --- ┆ --- │
│ str ┆ i64 │
╞═════╪═════╡
│ a   ┆ 1   │
│ b   ┆ 2   │
│ c   ┆ 4   │
└─────┴─────┘

Tip: again, it is better to implement this with an expression:

>>> (
...     df.lazy()
...     .group_by("b", maintain_order=True)
...     .agg(pl.col("a").sum())
...     .collect()
... )

Window function application using over will behave as a GroupBy context, with your function receiving individual window groups:

>>> df = pl.DataFrame(
...     {
...         "key": ["x", "x", "y", "x", "y", "z"],
...         "val": [1, 1, 1, 1, 1, 1],
...     }
... )
>>> df.with_columns(
...     scaled=pl.col("val").map_elements(lambda s: s * len(s)).over("key"),
... ).sort("key")
shape: (6, 3)
┌─────┬─────┬────────┐
│ key ┆ val ┆ scaled │
│ --- ┆ --- ┆ ---    │
│ str ┆ i64 ┆ i64    │
╞═════╪═════╪════════╡
│ x   ┆ 1   ┆ 3      │
│ x   ┆ 1   ┆ 3      │
│ x   ┆ 1   ┆ 3      │
│ y   ┆ 1   ┆ 2      │
│ y   ┆ 1   ┆ 2      │
│ z   ┆ 1   ┆ 1      │
└─────┴─────┴────────┘

Note that this function would also be better-implemented natively:

>>> df.with_columns(
...     scaled=(pl.col("val") * pl.col("val").count()).over("key"),
... ).sort(
...     "key"
... )