Selecting and Modifying Columns with Specific Patterns in Polars

What will you learn?

In this comprehensive guide, you will delve into the world of Polars to master the art of selecting columns ending with specific patterns and creating new columns based on these selections. You will gain insights into dynamic column manipulation, enabling you to efficiently transform your data without compromising the original naming patterns.

Introduction to the Problem and Solution

Working with dataframes in Polars often involves the need to identify and manipulate columns that adhere to a particular pattern. This task becomes crucial when handling datasets with consistent naming conventions or when applying systematic transformations across subsets of data.

The challenge lies in not only selecting these pattern-matching columns but also in modifying them by creating new columns that retain the data essence without including the original pattern. By leveraging Polars’ dynamic capabilities for column selection and manipulation, we can elegantly address this challenge.


import polars as pl

# Sample DataFrame creation
df = pl.DataFrame({
    "sales_2020": [100, 200, 300],
    "profit_2020": [10, 20, 30],
    "sales_2021": [150, 250, 350]

# Define the pattern to search for - here it's "_2020"
pattern = "_2020"

# Selecting columns ending with the specified pattern
selected_columns = [col for col in df.columns if col.endswith(pattern)]

# Creating new columns without the specified pattern and adding them to dataframe
for col in selected_columns:
    new_col_name = col.replace(pattern,"") # Removing the pattern from column name
    df[new_col_name + "_new"] = df[col] * 2 # Example transformation: doubling each value


# Copyright PHD


Step-by-Step Breakdown:

  1. Data Preparation: Initialize a sample DataFrame df containing sales and profit data for two years.
  2. Pattern Matching: Define pattern as _2020 to identify relevant columns.
  3. Column Selection: Dynamically select all column names ending with _2020.
  4. Dynamic Column Creation & Manipulation:
    • Generate new column names by removing _2020.
    • Perform example transformation (doubling values) on selected columns.
  5. Displaying df showcases both original selected and newly added transformed columns.

This process exemplifies Polars’ flexibility in dynamically selecting and transforming dataset features based on specific naming patterns within your data.

  1. How do I install Polars?

  2. To install Polars, use:

  3. pip install polars
  4. # Copyright PHD
  5. Can I use regex for matching patterns in column names?

  6. Yes! Utilize Python’s built-in re module for more complex pattern matching.

  7. What if I want to remove selected original columns after transformation?

  8. Simply use df.drop(selected_columns) post addition of new columns.

  9. Can I apply different transformations based on different patterns?

  10. Certainly! Iterate over distinct patterns applying unique transformations as needed.

  11. Is there support for inplace modifications?

  12. Polars typically returns modified copies; inplace modifications are not common practice.

  13. How can I reverse this operation? Adding back removed patterns into column names.

  14. Adjust logic during iteration to include removed patterns while constructing new column names.


By mastering dynamic selection techniques in Polars, you empower yourself to efficiently manipulate datasets based on naming conventions. This automation not only streamlines repetitive tasks but also enhances code clarity and flexibility within your projects.

Leave a Comment