7. Function Application

Apply functions row-wise, column-wise, element-wise, or via a clean pipe.

• df.apply(func, axis=0) → Apply function along an axis (0 = columns, 1 = rows).
• df.applymap(func) → Apply function element-wise.
• df.agg() or df.aggregate() → Aggregate using one or more operations.
• df.transform() → Transform rows/columns; shape is preserved.
• df.pipe(func) → Pipe DataFrame through a custom function.

Pandas Function Application - Comprehensive Guide

Table of Contents

1. df.apply() - Row-wise and Column-wise Operations
2. df.applymap() - Element-wise Operations
3. df.agg() / df.aggregate() - Aggregation Operations
4. df.transform() - Shape-preserving Transformations
5. df.pipe() - Function Chaining
6. Practical Examples with Multiple Datasets
7. Performance Tips

Setup and Data Loading

import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt

# Load sample datasets
tips = sns.load_dataset('tips')
flights = sns.load_dataset('flights')
titanic = sns.load_dataset('titanic')
iris = sns.load_dataset('iris')

print("Datasets loaded:")
print(f"Tips shape: {tips.shape}")
print(f"Flights shape: {flights.shape}")
print(f"Titanic shape: {titanic.shape}")
print(f"Iris shape: {iris.shape}")

Summary

This guide covered the five main function application methods in pandas:

1. df.apply(): Apply functions along rows or columns
2. df.applymap(): Apply functions element-wise to entire DataFrame
3. df.agg()/df.aggregate(): Aggregate data using one or more operations
4. df.transform(): Transform data while preserving DataFrame shape
5. df.pipe(): Chain custom functions for clean, readable data pipelines

When to Use Each Method:

• apply(): When you need row-wise or column-wise operations that can't be vectorized
• applymap(): When you need to transform every single element in the DataFrame
• agg(): When you need to compute summary statistics or aggregations
• transform(): When you need to modify data but keep the same shape (e.g., standardization)
• pipe(): When you want to chain multiple operations for clean, readable code

Performance Hierarchy (fastest to slowest):

1. Vectorized operations (direct arithmetic)
2. Built-in pandas methods
3. transform() and agg()
4. apply() with NumPy functions
5. apply() with custom Python functions
6. applymap() for element-wise operations

Remember: Always prefer vectorized operations when possible, and use these methods when vectorization isn't feasible or when you need specific functionality they provide.