How to Convert Float64 Columns with NaNs to Int64 in Pandas

Data Requirement

The task is to convert columns containing floating‑point numbers with missing values (None, np.nan, pd.NA) into integer type.

Understanding NaN, None and pd.NA

In Python, None represents a generic null, np.nan is a floating‑point NaN that does not compare equal to itself, and pd.NA is pandas’ nullable scalar that displays as <NA>. Because np.nan is a float, pandas cannot cast a column containing it directly to an integer.

Conversion Attempts

Using astype(int) on columns A‑D raises TypeError or ValueError because of the non‑finite values. Column E, which contains only valid numbers, converts successfully.

Working Solutions

Apply map(lambda x: pd.NA if pd.isna(x) else int(x)) to each column; the result is an object dtype with integer values and pd.NA for missing entries.

Use pandas’ nullable integer dtype: astype('Int64'). This preserves integers and stores missing values as pd.NA.

Summary

When a DataFrame contains missing values, converting float64 columns to int64 requires either replacing np.nan with pd.NA and using the nullable Int64 dtype, or applying a custom mapping function that handles nulls before casting.