How to Reshape NumPy Arrays with reshape()

Table Of Contents

• Shape-Shifting Arrays Without Breaking a Sweat
• Reshaping in Action
• Reshape Rules and Tricks
• Common Patterns
• Performance Tips
• Next Level NumPy

Shape-Shifting Arrays Without Breaking a Sweat

Ever had a flat array that needs to become a matrix? Or a 3D tensor that should be 2D? NumPy's reshape() is your dimensional Swiss Army knife.

Reshaping in Action

import numpy as np

# Basic reshaping
flat = np.array([1, 2, 3, 4, 5, 6])
matrix = flat.reshape(2, 3)
print(matrix)
# [[1 2 3]
#  [4 5 6]]

# Use -1 to infer dimension
auto_shape = flat.reshape(-1, 2)
print(auto_shape)
# [[1 2]
#  [3 4]
#  [5 6]]

# 3D reshaping
tensor = np.arange(24).reshape(2, 3, 4)
print(tensor.shape)  # (2, 3, 4)

# Flatten back to 1D
flattened = tensor.reshape(-1)
print(flattened.shape)  # (24,)

# Order matters: C vs Fortran
arr = np.arange(6).reshape(2, 3)
print(arr.reshape(3, 2, order='C'))   # Row-major (default)
print(arr.reshape(3, 2, order='F'))   # Column-major

# View vs Copy
original = np.array([1, 2, 3, 4])
view = original.reshape(2, 2)
view[0, 0] = 99
print(original)  # [99  2  3  4] - Original changed!

Reshape Rules and Tricks

• Size must match: Total elements must remain the same
• Use -1 wisely: Let NumPy calculate one dimension
• Views when possible: Reshape usually returns a view, not a copy
• Contiguous memory: Some reshapes require copying data

Common Patterns

# Image data: (height, width, channels) to (pixels, channels)
image = np.random.rand(28, 28, 3)
pixels = image.reshape(-1, 3)

# Batch processing
data = np.arange(100)
batches = data.reshape(10, 10)  # 10 batches of 10

Performance Tips

• Reshape returns a view when possible (no data copying)
• Use ravel() for guaranteed view when flattening
• flatten() always returns a copy

Next Level NumPy

Explore NumPy broadcasting magic, master advanced array indexing, and dive into NumPy's performance tricks.