4.2 stacking vs concatenating

This file illustrates difference between stack, vstack, hstack, column_stack, row_stack and concatenate

import numpy as np

print(np.__version__)

1.26.3

stack

all arrays must have same shape

both 1d arrays

a = np.random.random(10)
b = np.random.random(10)

print(np.stack([a,b]).shape)

(2, 10)

both 2D

a = np.random.random((10, 1))
b = np.random.random((10, 1))

print(np.stack([a,b]).shape)

(2, 10, 1)

print(np.stack([a,b], axis=1).shape)

(10, 2, 1)

print(np.stack([a,b], axis=2).shape)

(10, 1, 2)

a = np.random.random((10, 2))
b = np.random.random((10, 2))

print(np.stack([a,b]).shape)

(2, 10, 2)

print(np.stack([a,b], axis=0).shape)

(2, 10, 2)

print(np.stack([a,b], axis=1).shape)

(10, 2, 2)

print(np.stack([a,b], axis=2).shape)

(10, 2, 2)

# print(np.stack([a,b], axis=3).shape) # np.AxisError

different shapes

a = np.random.random((10, 2))
b = np.random.random((10, 1))

# print(np.stack([a,b]).shape)  # ValueError

# print(np.stack([a,b], axis=0).shape)  # ValueError

# print(np.stack([a,b], axis=1).shape)  # ValueError

concatenate

a = np.random.random(10)
b = np.random.random(10)

print(np.concatenate([a,b]).shape)

(20,)

print(np.concatenate([a,b], axis=0).shape)

(20,)

# print(np.concatenate([a,b], axis=1).shape)  # Error

a = np.random.random((10, 1))
b = np.random.random((10, 1))

print(np.concatenate([a,b]).shape)

(20, 1)

print(np.concatenate([a,b], axis=1).shape)

(10, 2)

a = np.random.random((10, 2))
b = np.random.random((10, 1))

# print(np.concatenate([a,b], axis=0)) # Error

print(np.concatenate([a,b], axis=1).shape)

(10, 3)

a = np.random.random((10, 2))
b = np.random.random((10, 2))

print(np.concatenate([a,b]).shape)

(20, 2)

print(np.concatenate([a,b], axis=1).shape)

(10, 4)

# print(np.concatenate([a,b], axis=2).shape)  # AxisError

a = np.random.random((10, 5, 3))
b = np.random.random((10, 5, 3))

print(np.concatenate([a,b]).shape)

(20, 5, 3)

print(np.concatenate([a,b], axis=1).shape)

(10, 10, 3)

print(np.concatenate([a,b], axis=2).shape)

(10, 5, 6)

vstack

a = np.random.random(10)
b = np.random.random(10)

print(np.vstack([a,b]).shape)

(2, 10)

a = np.random.random((10, 1))
b = np.random.random((10, 1))

print(np.vstack([a,b]).shape)

(20, 1)

a = np.random.random((10, 2))
b = np.random.random((10, 2))

print(np.vstack([a,b]).shape)

(20, 2)

a = np.random.random((10, 2))
b = np.random.random((10, 1))

print(np.hstack([a,b]).shape)

(10, 3)

a = np.random.random((10, 5, 3))
b = np.random.random((10, 5, 3))

print(np.hstack([a,b]).shape)

(10, 10, 3)

hstack

a = np.random.random(10)
b = np.random.random(10)

print(np.hstack([a,b]).shape)

(20,)

a = np.random.random((10, 1))
b = np.random.random((10, 1))

print(np.hstack([a,b]).shape)

(10, 2)

a = np.random.random((10, 2))
b = np.random.random((10, 2))

print(np.hstack([a,b]).shape)

(10, 4)

a = np.random.random((10, 2))
b = np.random.random((10, 1))

print(np.hstack([a,b]).shape)

(10, 3)

a = np.random.random((10, 5, 3))
b = np.random.random((10, 5, 3))

print(np.hstack([a,b]).shape)

(10, 10, 3)

column stack

a = np.random.random(10)
b = np.random.random(10)

print(np.column_stack([a,b]).shape)

(10, 2)

a = np.random.random((10, 1))
b = np.random.random((10, 1))

print(np.column_stack([a,b]).shape)

(10, 2)

a = np.random.random((10, 2))
b = np.random.random((10, 2))

print(np.column_stack([a,b]).shape)

(10, 4)

a = np.random.random((10, 2))
b = np.random.random((10, 1))

print(np.column_stack([a,b]).shape)

(10, 3)

a = np.random.random((10, 5, 3))
b = np.random.random((10, 5, 3))

print(np.column_stack([a,b]).shape)

(10, 10, 3)

row stack

a = np.random.random(10)
b = np.random.random(10)

print(np.row_stack([a,b]).shape)

(2, 10)

a = np.random.random((10, 1))
b = np.random.random((10, 1))

print(np.row_stack([a,b]).shape)

(20, 1)

a = np.random.random((10, 2))
b = np.random.random((10, 2))

print(np.row_stack([a,b]).shape)

(20, 2)

a = np.random.random((10, 2))
b = np.random.random((10, 1))

# print(np.row_stack([a,b]).shape) ValueError

a = np.random.random((10, 5, 3))
b = np.random.random((10, 5, 3))

print(np.row_stack([a,b]).shape)

(20, 5, 3)

dstack

depth wise stacking

a = np.random.random(10)
b = np.random.random(10)

print(np.dstack([a,b]).shape)

(1, 10, 2)

a = np.random.random((10, 1))
b = np.random.random((10, 1))

print(np.dstack([a,b]).shape)

(10, 1, 2)

a = np.random.random((10, 2))
b = np.random.random((10, 2))

print(np.dstack([a,b]).shape)

(10, 2, 2)

a = np.random.random((10, 2))
b = np.random.random((10, 1))

# print(np.dstack([a,b]).shape)  # ValueError

a = np.random.random((10, 5, 3))
b = np.random.random((10, 5, 3))

print(np.dstack([a,b]).shape)

(10, 5, 6)

comparison

a = np.random.random(10)
b = np.random.random(10)

print(np.concatenate([a,b]).shape)

print(np.stack([a,b]).shape)

print(np.vstack([a,b]).shape)

print(np.hstack([a,b]).shape)

print(np.row_stack([a,b]).shape)

print(np.column_stack([a,b]).shape)

print(np.dstack([a,b]).shape)

(20,)
(2, 10)
(2, 10)
(20,)
(2, 10)
(10, 2)
(1, 10, 2)

a = np.random.random((10, 1))
b = np.random.random((10, 1))

print(np.concatenate([a,b]).shape)

print(np.stack([a,b]).shape)

print(np.vstack([a,b]).shape)

print(np.hstack([a,b]).shape)

print(np.row_stack([a,b]).shape)

print(np.column_stack([a,b]).shape)

print(np.dstack([a,b]).shape)

(20, 1)
(2, 10, 1)
(20, 1)
(10, 2)
(20, 1)
(10, 2)
(10, 1, 2)

a = np.random.random((10, 2))
b = np.random.random((10, 2))

print(np.concatenate([a,b]).shape)

print(np.stack([a,b]).shape)

print(np.vstack([a,b]).shape)

print(np.hstack([a,b]).shape)

print(np.row_stack([a,b]).shape)

print(np.column_stack([a,b]).shape)

print(np.dstack([a,b]).shape)

(20, 2)
(2, 10, 2)
(20, 2)
(10, 4)
(20, 2)
(10, 4)
(10, 2, 2)

a = np.random.random((10, 2))
b = np.random.random((10, 1))

# print(np.concatenate([a,b]).shape)  # ValueError

# print(np.stack([a,b]).shape)  # ValueError

# print(np.vstack([a,b]).shape) # ValueError

print(np.hstack([a,b]).shape)

# print(np.row_stack([a,b]).shape) # ValueError

print(np.column_stack([a,b]).shape)

# print(np.dstack([a,b]).shape)  # ValueError

(10, 3)
(10, 3)

a = np.random.random((10, 5, 3))
b = np.random.random((10, 5, 3))

print(np.concatenate([a,b]).shape)

print(np.stack([a,b]).shape)

print(np.vstack([a,b]).shape)

print(np.hstack([a,b]).shape)

print(np.row_stack([a,b]).shape)

print(np.column_stack([a,b]).shape)

print(np.dstack([a,b]).shape)

(20, 5, 3)
(2, 10, 5, 3)
(20, 5, 3)
(10, 10, 3)
(20, 5, 3)
(10, 10, 3)
(10, 5, 6)