Polars · Module

df.cast

Converts selected DataFrame columns, or every column, to specified data types. Use mappings or selectors for targeted casts and strict=False to turn invalid conversions into nulls.

How it works

DataFrame.cast returns a DataFrame with converted column types. Pass one dtype to convert every column, or provide a mapping whose keys are column names, selectors, or dtypes; strict controls whether invalid conversions raise or become null.

Syntax

DataFrame.cast(dtypes: Mapping[ColumnNameOrSelector | PolarsDataType, PolarsDataType | PythonDataType] | PolarsDataType | Schema, *, strict: bool = True) → DataFrame

Parameters

Name	Type	Description
`dtypes`	`Mapping[ColumnNameOrSelector \| PolarsDataType, PolarsDataType \| PythonDataType] \| PolarsDataType \| Schema`	Mapping of column names or selectors to target dtypes, one dtype for all columns, or a Schema.
`strict`	`bool = True`	Controls invalid conversions; when false, invalid casts produce null values.

Returns

DataFrame

Key points

Use cast to convert the types of one or more DataFrame columns.
A mapping enables per-column or selector-based targets, while one dtype applies across the whole frame.
Column-specific mappings let different columns receive different destination types in one operation.
Selectors provide a concise way to address groups of columns without listing each name.
Passing a dtype directly converts every column to that destination type.
Strict conversion is enabled by default, so invalid casts are not silently accepted.

Common mistakes

With strict=False, values that cannot be converted are replaced with null instead of raising an error.

Example: Cast selected columns

import polars as pl
import polars.selectors as cs

df = pl.DataFrame({
    "count": [1, 2, 3],
    "price": [6.0, 7.0, 8.0],
    "date": ["2020-01-02", "2021-03-04", "2022-05-06"],
})

result = df.cast({cs.numeric(): pl.UInt32, "date": pl.Date})
print(result.dtypes)

Casts every numeric column to UInt32 and converts the date strings to Date.

When to use it

Normalize numeric dtypes

A DataFrame contains numeric columns with inconsistent widths and needs a common representation.

Convert a complete frame

A pipeline must convert all columns to a uniform textual representation.

Handle invalid values

Input data may contain values that cannot be parsed into the requested type.

Interview questions

QWhat does DataFrame.cast do?

DataFrame.cast changes the dtype of DataFrame columns and returns a new DataFrame. A single dtype applies to all columns, while a mapping targets columns or selectors individually.

Follow-up: How would you cast only numeric columns? Use a selector such as cs.numeric() as a mapping key, for example df.cast({cs.numeric(): pl.UInt32}).

QWhat forms can the dtypes argument take?

The dtypes argument accepts a mapping from column names or selectors to target dtypes, a single target dtype, or a Schema. This lets one call express either targeted or whole-frame conversion.

Follow-up: How can temporal and numeric columns receive different targets? Supply multiple selector keys in one mapping, such as cs.numeric() and cs.temporal().

QHow does strict affect casting?

strict defaults to True. When a conversion is invalid, strict mode raises; setting it to False makes invalid converted values null.

Follow-up: When should strict=False be used? Use it when malformed input should be retained as nulls rather than aborting the cast operation.

polars.DataFrame.with_columns
polars.Expr.cast
polars.selectors.numeric
polars.selectors.temporal

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