NumPy full(): Generating Constant-Filled Arrays in Python

1. Syntax & Formal Parameter Specifications 🔝

The numpy.full() function accepts array dimensions, fill value, and optional parameters:

# Formal syntax of numpy.full()
numpy.full(shape, fill_value, dtype=None, order='C', *, like=None)

2. Array Shapes: 1D Vectors & 2D Matrices 🔝

Passing an integer creates a 1D vector; passing a tuple defines multi-dimensional coordinate axes:

# 1. 1D vector and 2D matrix filled with custom scalar
import numpy as np

# 1D vector filled with 7
vec = np.full(4, 7)
print("1D Vector (7): ", vec)
print("Dtype:          ", vec.dtype)

# 2D matrix filled with 3.14
mat = np.full((3, 3), 3.14)
print("\n2D Matrix (3x3 with 3.14):\n", mat)

Terminal Output:

1D Vector (7):  [7 7 7 7]
Dtype:           int64

2D Matrix (3x3 with 3.14):
 [[3.14 3.14 3.14]
 [3.14 3.14 3.14]
 [3.14 3.14 3.14]]

3. Filling with Sentinel Values: NaN and Inf 🔝

In data science pipelines, arrays frequently need initialization with sentinel flags like np.nan or np.inf:

# 2. Filling with Sentinel Values: NaN and Infs
import numpy as np

nan_grid = np.full((2, 3), np.nan)
inf_grid = np.full((2, 3), np.inf)

print("NaN Matrix:\n", nan_grid)
print("\nInf Matrix:\n", inf_grid)

Terminal Output:

NaN Matrix:
 [[nan nan nan]
 [nan nan nan]]

Inf Matrix:
 [[inf inf inf]
 [inf inf inf]]

4. Data Types (dtypes): Enforcing Precision 🔝

By default, dtype is inferred from the fill_value. You can override and coerce types explicitly:

# 3. Specifying dtypes and casting fill_values
import numpy as np

# Integer 42 coerced to float64
f_arr = np.full(3, 42, dtype=float)

# String fill value
s_arr = np.full((2, 2), 'alpha', dtype='<U6')

print("Float-cast array: ", f_arr)
print("String array:     \n", s_arr)

Terminal Output:

Float-cast array:  [42. 42. 42.]
String array:     
 [['alpha' 'alpha']
 ['alpha' 'alpha']]

5. Memory Layout Order: Row-Major ('C') vs. Column-Major ('F') 🔝

The order parameter specifies row-major ('C') or column-major ('F') physical memory layout:

# 4. Storage layout: Row-Major ('C') vs. Column-Major ('F')
import numpy as np

c_arr = np.full((3, 3), 9, order='C')
f_arr = np.full((3, 3), 9, order='F')

print("C-contiguous (row-major):   ", c_arr.flags['C_CONTIGUOUS'])
print("F-contiguous (column-major):", f_arr.flags['F_CONTIGUOUS'])

Terminal Output:

C-contiguous (row-major):    True
F-contiguous (column-major): True

6. Mirroring Geometries with numpy.full_like() 🔝

Use numpy.full_like() to allocate an array mirroring an existing array's shape and dtype:

# 5. Mirroring prototype arrays with numpy.full_like()
import numpy as np

prototype = np.zeros((2, 4), dtype=np.int32)
filled = np.full_like(prototype, fill_value=99)

print("Prototype shape & dtype:", prototype.shape, prototype.dtype)
print("Mirrored full array:   \n", filled)

Terminal Output:

Prototype shape & dtype: (2, 4) int32
Mirrored full array:   
 [[99 99 99 99]
 [99 99 99 99]]

7. Comparing Constructors: full vs. zeros vs. ones vs. empty 🔝

NumPy provides specialized constructors depending on your initialization needs:

8. Production Recipes 🔝

Initializing a shortest-path distance matrix with infinity:

# 6. Production Recipe: Baseline Game Grid / Distance Matrix Initializer
import numpy as np

# Initializing a 5x5 shortest-path cost matrix with infinity
cost_matrix = np.full((5, 5), fill_value=np.inf)
np.fill_diagonal(cost_matrix, 0.0)

print("Initial cost matrix:\n", cost_matrix)

Terminal Output:

Initial cost matrix:
 [[ 0. inf inf inf inf]
 [inf  0. inf inf inf]
 [inf inf  0. inf inf]
 [inf inf inf  0. inf]
 [inf inf inf inf  0.]]

Creating a constant-value boundary canvas buffer:

# 7. Production Recipe: Constant Value Boundary Padding
import numpy as np

# Creating an image border canvas filled with neutral gray (128)
gray_canvas = np.full((100, 100), fill_value=128, dtype=np.uint8)
gray_canvas[10:90, 10:90] = 255  # Inner white patch

print("Canvas shape:    ", gray_canvas.shape)
print("Boundary pixel:  ", gray_canvas[0, 0])
print("Inner patch pixel:", gray_canvas[50, 50])

Terminal Output:

Canvas shape:     (100, 100)
Boundary pixel:   128
Inner patch pixel: 255

9. Common Developer Pitfalls 🔝

Watch for dtype truncation when passing float fill values to integer arrays:

# 8. Common traps: Fill value type truncation
import numpy as np

# Trap: Float truncated to int when dtype=int is specified
int_truncated = np.full(3, 3.85, dtype=int)
print("Truncated float: ", int_truncated)  # Values are 3, not 3.85!

# Shape tuple trap
try:
    np.full(3, 3, 5)  # 3 positional args interpreted incorrectly
except TypeError as err:
    print("Caught TypeError:", err)

Terminal Output:

Truncated float:  [3 3 3]
Caught TypeError: 'int' object cannot be interpreted as an integer

10. Interactive Practice Notebook 🔝

Practice constant array creation and buffer allocation in our companion notebook:

Self-Verifying Assertion Suite

Run unit assertions to verify numpy.full() mechanics:

# 9. Self-verifying unit test suite for numpy.full()
import numpy as np

# Test 1: 1D vector shape and fill value
v = np.full(5, 42)
assert v.shape == (5,) and np.all(v == 42)

# Test 2: Float NaN matrix
nan_m = np.full((3, 3), np.nan)
assert nan_m.shape == (3, 3) and np.all(np.isnan(nan_m))

# Test 3: full_like equivalence
proto = np.zeros((2, 4), dtype=np.int16)
f_like = np.full_like(proto, 7)
assert f_like.shape == (2, 4) and f_like.dtype == np.int16 and np.all(f_like == 7)

print("All numpy.full() unit assertions successfully verified!")

Terminal Output:

All numpy.full() unit assertions successfully verified!

11. FAQ 🔝

Further Reading & Related Tutorials 🔝

• NumPy zeros() Method — Generating arrays filled with zeros.
• NumPy ones() Method — Generating arrays filled with ones.
• NumPy empty() Method — Fast uninitialized memory allocation.
• NumPy Array Shape — Inspecting and reshaping multi-dimensional dimensions.
• NumPy ndarray Object — Complete architecture and attributes of NumPy arrays.
• NumPy arange() Function — Creating regularly spaced sequences.
• NumPy Master Overview — Central guide to array computing with NumPy.

Subhendu Mohapatra

Author & Instructor at plus2net

I write and maintain practical tutorials on Python, PHP, SQL, JavaScript, HTML, jQuery, and web development at plus2net. The tutorials focus on clear explanations, working examples, and code that readers can test and adapt while learning.

About the author   Watch on YouTube

Connect: