How to reshape a NumPy array

Tabular and batched code often receives values in a flat sequence even though later operations expect rows, columns, or records. A NumPy reshape assigns those dimensions without changing the sequence of elements.

The ndarray.reshape() method accepts only dimensions whose product matches the array's size. One dimension can be -1 when NumPy should infer its length from the remaining dimensions.

A reshape returns a view when the existing memory layout can represent the requested dimensions and a copy otherwise. The contiguous source used here shares memory with both reshaped results, so code that needs independent data should call .copy() before making in-place edits.

1. Define the 12-value source array in array-reshape.py.

   array-reshape.py

   import numpy as np
    
   values = np.arange(12)

2. Append the two-dimensional grid reshape to array-reshape.py.

   grid = values.reshape(3, 4)
    
   print("values shape:", values.shape)
   print("grid shape:", grid.shape)
   print(grid)

3. Extend array-reshape.py with an inferred record axis and computed preservation checks.

   records = grid.reshape(2, -1, 3)
   values_preserved = np.array_equal(grid.reshape(-1), values)
   shares_memory = np.shares_memory(values, records)
    
   print("records shape:", records.shape)
   print("values preserved:", values_preserved)
   print("shares memory:", shares_memory)
    
   assert grid.shape == (3, 4)
   assert records.shape == (2, 2, 3)
   assert values_preserved
   assert shares_memory

   The -1 dimension becomes 2 because the other dimensions already account for six positions and the source contains 12 values.

4. Compare the constructed sections with the completed array-reshape.py file.

   array-reshape.py

   import numpy as np
    
   values = np.arange(12)
    
   grid = values.reshape(3, 4)
    
   records = grid.reshape(2, -1, 3)
   values_preserved = np.array_equal(grid.reshape(-1), values)
   shares_memory = np.shares_memory(values, records)
    
   print("values shape:", values.shape)
   print("grid shape:", grid.shape)
   print(grid)
   print("records shape:", records.shape)
   print("values preserved:", values_preserved)
   print("shares memory:", shares_memory)
    
   assert grid.shape == (3, 4)
   assert records.shape == (2, 2, 3)
   assert values_preserved
   assert shares_memory

5. Run array-reshape.py to verify both target shapes, value preservation, and the view relationship.

   $ python3 array-reshape.py
   values shape: (12,)
   grid shape: (3, 4)
   [[ 0  1  2  3]
    [ 4  5  6  7]
    [ 8  9 10 11]]
   records shape: (2, 2, 3)
   values preserved: True
   shares memory: True

   The two reshapes retain all 12 values. An incompatible request such as (5, 3) needs 15 values and raises ValueError instead of returning a partial array.