NumPy reshape(): Change Array Shape Without Data Copying

import numpy as np

ar = np.array([1, 2, 3, 4, 5, 6])
result = ar.reshape(2, 3)
print(result)  # [[1 2 3] [4 5 6]]

numpy.reshape(a, newshape, order='C')
reshape() is a universal method/function that gives a new shape to an array without changing its data elements, returning a memory-efficient view whenever possible.

The reshape() function is fundamental for transforming 1D flat sequences into 2D matrices, 3D image tensors, or vice-versa. Total element count must remain identical (e.g., 12 elements can be shaped as (3, 4), (2, 6), or (2, 2, 3)). For structural cloning without initialization, see empty_like(). For array slicing and indexing, review Indexing & Slicing.

Parameter	Description
`a`	Array to be reshaped
`newshape`	Int or tuple of ints defining the new shape. One dimension can be -1 for automatic inference
`order`	Memory layout order: `'C'` (row-major C-style), `'F'` (column-major Fortran-style), or `'A'` (aligned)

Show Table of Contents

Syntax and Parameters 🔝

numpy.reshape(a, newshape, order='C')
# Or as an ndarray method:
a.reshape(newshape, order='C')

Code Block Modernization Comparison 🔝

The table below highlights how legacy NumPy reshape snippets compare against the modernized, structured implementation:

Aspect	Legacy Implementation	Modernized Implementation
Structure & Markup	Basic echo output with unformatted pre tags and missing metadata.	Structured HTML5 layout, Schema `TechArticle` markup, and responsive tables.
Code Presentation	Requires scrolling past introductory text to find code examples.	Core code snippet placed right at the top for instant visitor access.
Verification	Unchecked print statements with manual result inspection comments.	Automated unit test assertions (`assert`) verifying shape and dimensions.
Ecosystem Integration	Isolated code blocks with sparse navigation links.	Comprehensive cross-linking, quick-jump navigation, and runnable Colab notebooks.

Basic 1D to 2D Reshaping 🔝

Transform a flat 1D array into a multi-dimensional matrix:

import numpy as np

ar = np.arange(12)
ar_reshaped = ar.reshape(3, 4)
print("Reshaped 3x4 matrix:
", ar_reshaped)

assert ar_reshaped.shape == (3, 4)

Expected output

Reshaped 3x4 matrix:
 [[ 0  1  2  3]
 [ 4  5  6  7]
 [ 8  9 10 11]]

Automatic Inference with -1 🔝

You can let NumPy automatically calculate one dimension by passing -1:

ar = np.arange(12)
# Automatically infer column count for 4 rows
auto_mat = ar.reshape(4, -1)
print("Inferred 4x3 matrix:
", auto_mat)

assert auto_mat.shape == (4, 3)

Expected output

Inferred 4x3 matrix:
 [[ 0  1  2]
 [ 3  4  5]
 [ 6  7  8]
 [ 9 10 11]]

C vs Fortran Memory Order 🔝

The order parameter controls how elements are read from and written to memory. 'C' uses row-major order (C-style), while 'F' uses column-major order (Fortran-style):

ar = np.arange(6)
c_order = ar.reshape(2, 3, order='C')
f_order = ar.reshape(2, 3, order='F')

print("C order (row-major):
", c_order)
print("F order (column-major):
", f_order)

assert c_order[0, 1] == 1
assert f_order[0, 1] == 2

Expected output

C order (row-major):
 [[0 1 2]
 [3 4 5]]
F order (column-major):
 [[0 2 4]
 [1 3 5]]

Flattening with ravel() vs flatten() 🔝

To collapse multi-dimensional arrays back into 1D sequences, use ravel() (returns a view when possible) or flatten() (always returns a copy):

mat = np.array([[1, 2], [3, 4]])
rav = mat.ravel()
flat = mat.flatten()

print("Ravel (view):", rav)
print("Flatten (copy):", flat)

assert rav.ndim == 1
assert flat.ndim == 1

Expected output

Ravel (view): [1 2 3 4]
Flatten (copy): [1 2 3 4]

Common Pitfalls & ValueError 🔝

Total Size Mismatch: Attempting to reshape an array of 12 elements into shape (3, 5) raises ValueError: cannot reshape array of size 12 into shape (3,5) because 3 × 5 = 15 ≠ 12.
Multiple -1 Dimensions: You can only omit one dimension using -1. Specifying multiple unknown dimensions (e.g., arr.reshape(-1, -1)) raises a ValueError.

Practice in Google Colab 🔝

Test 1D/2D reshaping, automatic dimension inference, C vs Fortran order, flattening, and run verification test suites in our interactive Google Colab companion notebook:

Interactive Practice: Colab & GitHub

Launch the notebook directly in Google Colab to run live code cells or inspect the source repository on GitHub:

View on GitHub

Frequently Asked Questions (FAQ) 🔝

Does `reshape()` create a copy or a view of the array?

reshape() tries to return a memory-efficient view whenever possible. However, if the source array is non-contiguous in memory, it may return a copied array. Use flags.owndata or check base to verify whether data is shared.

What happens if I specify multiple `-1` dimensions in `reshape()`?

Python will raise a ValueError because NumPy can only automatically infer one unknown dimension at a time. The remaining dimensions must be explicitly specified so the total element count can be calculated.

What is the difference between `reshape()` and `resize()`?

reshape() returns a new view or copy with a different shape without altering the original array data or dimensions, whereas resize() modifies the array shape in-place and can optionally zero-fill or truncate data.

NumPy reshape(): Change Array Shape Without Data Copying

Syntax and Parameters 🔝

Code Block Modernization Comparison 🔝

Basic 1D to 2D Reshaping 🔝

Automatic Inference with -1 🔝

C vs Fortran Memory Order 🔝

Flattening with ravel() vs flatten() 🔝

Common Pitfalls & ValueError 🔝

Practice in Google Colab 🔝

Interactive Practice: Colab & GitHub

Frequently Asked Questions (FAQ) 🔝

Does `reshape()` create a copy or a view of the array?

What happens if I specify multiple `-1` dimensions in `reshape()`?

What is the difference between `reshape()` and `resize()`?

Further Reading & Related Tutorials 🔝

Subhendu Mohapatra