How to Filter a DataFrame for Numeric Values in Pandas   

To filter a DataFrame for numeric values in Pandas we can:

(1) Use str.isnumeric() with boolean indexing

df[df['col'].str.isnumeric()]

(2) Use pd.to_numeric() with errors='coerce'

df[pd.to_numeric(df['col'], errors='coerce').notna()]

(3) Use regular expressions with str.match()

df[df['col'].str.match(r'^\d+$')]

Step 1: Create a DataFrame

Assume we have a DataFrame with a string column that contains both numeric and non-numeric values:

import pandas as pd

data = {
    'col': ['123', 'abc', '45', 'xyz', '678', 'hello']
}

df = pd.DataFrame(data)

DataFrame looks like:

Step 2: Why df['col'].filter(str.isnumeric) Fails

The original attempt:

df['col'].filter(str.isnumeric)

does not work because filter() is a DataFrame/Series method that filters labels (index or column names), not values. It expects a function that operates on the index labels, not the cell values. Additionally, str.isnumeric is a string method, not a callable that filter() expects for label-based filtering.

Step 3: Filter Numeric Values with str.isnumeric()

We can use boolean indexing with the string accessor str.isnumeric():

df_numeric = df[df['col'].str.isnumeric()]

result:

Step 4: Filter Numeric Values with pd.to_numeric()

Another approach is to attempt conversion and keep only successful conversions:

df_numeric = df[pd.to_numeric(df['col'], errors='coerce').notna()]

result:

Step 5: Filter Numeric Values with Regular Expressions

For more control, use str.match() with a regex pattern:

df_numeric = df[df['col'].str.match(r'^\d+$')]

result:

Step 6: Convert Filtered Results to Numeric Type

After filtering, you may want to convert the remaining values to integers:

df_numeric['col'] = df_numeric['col'].astype(int)

result:

Summary

The key mistake in the original code was confusing filter() (a label-based method) with boolean indexing (a value-based approach). For filtering DataFrame values, always use boolean indexing with df[condition].