Polars Data Types: How to cast and Convert Columns

When you load a CSV, Polars (and Pandas) often guesses the data types. Sometimes, it guesses wrong, loading a number column (like 5.0) as a string ("5.0"). In these situations, you’ll need to use Polars cast functionality to fix or convert types.

To do math or save memory, you must have the correct types. In Polars, the fast and easy way to fix this is with .cast().

The .cast() Method

cast() is a Polars Expression that you use inside with_columns.

The Setup

Let’s create a DataFrame with the wrong types.

import polars as pl
df = pl.DataFrame({
    "id": ["1", "2", "3"],          # This is a string
    "sales": ["100.50", "200.00", "99.25"], # This is a string
    "category": ["A", "B", "A"]     # This is a string
})
print(df.dtypes)
# Output: [str, str, str]

How to Convert (Cast) Types

Let’s fix all three columns at once.

df_clean = df.with_columns(
    # Cast "id" to a 32-bit Integer
    pl.col("id").cast(pl.Int32),
    
    # Cast "sales" to a 64-bit Float (for decimals)
    pl.col("sales").cast(pl.Float64),
    
    # Cast "category" to Categorical for memory savings
    pl.col("category").cast(pl.Categorical)
)
print(df_clean.dtypes)
# Output: [i32, f64, cat]

This is the fundamental first step of data cleaning. By setting the correct types, your math will work, your memory usage will drop, and your join and groupby operations will be significantly faster.