Polars · Module

df.to_pandas

Converts a Polars DataFrame to pandas. Enable PyArrow-backed extension arrays to preserve nulls and potentially avoid copying data.

How it works

Call this method on a Polars DataFrame to obtain a pandas DataFrame. By default, conversion uses NumPy-backed columns and copies data; enabling PyArrow-backed extension arrays can allow zero-copy conversion and preserve null values. Extra keyword arguments are forwarded to PyArrow's pandas conversion routine.

Syntax

DataFrame.to_pandas(*, use_pyarrow_extension_array: bool = False, **kwargs: Any) → DataFrame

Parameters

Name	Type	Description
`use_pyarrow_extension_array`	`bool = False`	Choose PyArrow-backed extension arrays for pandas columns; this can preserve nulls and enable zero-copy conversion.
`**kwargs`	`Any`	Additional options forwarded to pyarrow.Table.to_pandas().

Returns

pandas.DataFrame

Key points

The method converts the current Polars DataFrame into a pandas DataFrame.
The optional flag selects PyArrow-backed pandas column storage instead of NumPy-backed storage.
PyArrow-backed conversion can avoid copying and retain missing values as pandas nulls.
Additional keyword arguments configure the underlying PyArrow conversion.
On the default conversion path, missing numeric values appear as NaN in pandas.

Common mistakes

The default path copies data, so use_pyarrow_extension_array=True when avoiding an unnecessary copy matters.
PyArrow-backed columns can later convert to NumPy when pandas operations lack PyArrow compute support.
Install both pandas and PyArrow before calling the method.

Example: Preserve nulls during conversion

import polars as pl

frame = pl.DataFrame({"value": [1, None], "label": ["a", "b"]})
pandas_frame = frame.to_pandas(use_pyarrow_extension_array=True)
print(pandas_frame)
print(pandas_frame.dtypes)

Converts a Polars DataFrame while retaining missing values through PyArrow-backed pandas extension arrays.

When to use it

Interop with pandas tooling

Use pandas-only libraries or APIs that accept pandas.DataFrame rather than Polars DataFrame.

Null-preserving conversion

Retain missing values and reduce copying when transferring columns into pandas.

Customize conversion

Pass supported PyArrow conversion settings for specialized pandas output.

Interview questions

QWhat does DataFrame.to_pandas() return, and what happens to nulls by default?

The method returns a pandas.DataFrame created from the Polars DataFrame. Without the PyArrow extension option, the conversion copies data and numeric nulls become NaN.

Follow-up: How can null preservation be improved? Pass use_pyarrow_extension_array=True, which uses PyArrow-backed extension arrays and can also enable zero-copy conversion.

QWhen should use_pyarrow_extension_array be enabled?

Set use_pyarrow_extension_array=True to use PyArrow-backed extension arrays rather than NumPy arrays. This can preserve null values and permit zero-copy conversion.

Follow-up: Is zero-copy guaranteed for all later pandas work? No; unsupported operations may cause conversion to NumPy.

QWhat is the purpose of **kwargs?

The **kwargs parameter forwards additional keyword arguments to pyarrow.Table.to_pandas(). This lets callers configure the underlying PyArrow-to-pandas conversion.

Follow-up: What prerequisite does this imply? The environment must provide PyArrow, as well as pandas, because the method requires both packages.

polars.DataFrame.to_arrow
polars.from_pandas
polars.DataFrame.to_numpy

Source: polars.DataFrame.to_pandas — Polars documentation. Checked 2026-10-15.