np.array() from Lists

NumPy arrays can be created in two ways: from existing data (convert a Python list you already have) or from scratch (let NumPy generate the values — zeros, ranges, etc.). This topic covers the first, and most common, way: np.array(). You will use it every time you type in a team's scores by hand.

Definition

np.array(object, dtype=None) creates a new ndarray from any array-like Python object — a list, a tuple, a nested list, or another array. The nesting depth of the input decides the number of dimensions, and NumPy chooses one dtype that can represent every element (this is called type promotion).

• Array-like — anything NumPy can turn into an array: lists, tuples, range objects, nested lists, other arrays.
• Nesting depth — a flat list gives a 1D array; a list of lists gives 2D; a list of lists of lists gives 3D.
• Type promotion — if the input mixes kinds, NumPy picks the "widest" kind that fits all: bool → int → float → complex → string.

Syntax

np.array(object, dtype=None, *, copy=True, ndmin=0)

Example — 1D, 2D and 3D arrays

import numpy as np

# 1D: one batter, 5 matches
virat = np.array([72, 45, 88, 31, 64])
print(virat)
print("ndim:", virat.ndim, "shape:", virat.shape)
print()
# 2D: 3 batters x 4 matches
team = np.array([
    [72, 45, 88, 31],
    [56,  0, 74, 88],
    [45, 32,  0, 67],
])
print(team)
print("ndim:", team.ndim, "shape:", team.shape)
print()
seasons = np.array([
    [[72, 45], [56, 0]],
    [[65, 80], [39, 55]],
])
print(seasons)
print("ndim:", seasons.ndim, "shape:", seasons.shape)

[72 45 88 31 64]
ndim: 1 shape: (5,)

[[72 45 88 31]
 [56  0 74 88]
 [45 32  0 67]]
ndim: 2 shape: (3, 4)

[[[72 45]
  [56  0]]

 [[65 80]
  [39 55]]]
ndim: 3 shape: (2, 2, 2)

How it works: count the opening brackets at the start of the printed array — [ = 1D, [[ = 2D, [[[ = 3D. In the 2D case each inner list becomes a row. In the 3D case each 2D block (a season) is printed separately with a blank line between blocks.

Example — type promotion

import numpy as np
print(np.array([72, 45, 88]).dtype)
print(np.array([72, 45.5, 88]))
print(np.array([72, 45.5, 88]).dtype)
print(np.array([True, 2, 3]))
print(np.array([True, 2, 3]).dtype)
print(np.array([True, False]).dtype)
print(np.array([1, 2.5, 3+4j]))
print(np.array([72, "Virat"]))
print(np.array([72, "Virat"]).dtype)
print(np.array((10, 20, 30)))
print(np.array(range(5)))

int64
[72.  45.5 88. ]
float64
[1 2 3]
int64
bool
[1. +0.j 2.5+0.j 3. +4.j]
['72' 'Virat']
<U21
[10 20 30]
[0 1 2 3 4]

How it works:

Tuples and range objects work just like lists.

Example — forcing a dtype, ndmin, and copying

import numpy as np
a = np.array([72, 45, 88], dtype=float)
print(a)
b = np.array([72.9, 45.2, 88.6], dtype=int)
print(b)
c = np.array([1, 2, 3], ndmin=2)
print(c, c.shape)
lst = [1, 2, 3]
d = np.array(lst)
d[0] = 100
print(lst, d)
print(np.array(5), np.array(5).ndim, np.array(5).shape)

[72. 45. 88.]
[72 45 88]
[[1 2 3]] (1, 3)
[1, 2, 3] [100   2   3]
5 0 ()

How it works:

• dtype=float turns whole numbers into floats.
• dtype=int on decimals truncates (72.9 → 72, not 73). Round first with np.round if you want rounding.
• ndmin=2 wraps the 1D data into a 1×3 2D array.
• np.array copies — changing d does not touch lst.
• np.array(5) is a 0-dimensional array (a scalar wrapped in an array) with shape ().

The ragged-list error

Every row of a 2D array must have the same length. NumPy refuses "ragged" (uneven) nested lists:

import numpy as np
try:
    np.array([[1, 2, 3], [4, 5]])
except ValueError as e:
    print("ValueError:", e)

ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (2,) + inhomogeneous part.

Fix: make every row the same length (pad missing matches with 0 or np.nan). Passing dtype=object "works" but gives an array of Python lists with none of NumPy's speed — avoid it.

Common mistakes

• ❌ np.array(1, 2, 3) — values as separate arguments (2 is then read as the dtype → error). ✅ np.array([1, 2, 3]).
• ❌ Rows of different lengths. ✅ Every inner list must have the same number of items.
• ❌ Expecting dtype=int to round. ✅ It truncates toward zero; use np.round(x).astype(int) to round.
• ❌ One text value in numeric data turning the whole array into strings. ✅ Clean the data or convert with .astype(int) / .astype(float).

Interview tip

Q: What is the dtype of np.array([1, 2, 3.0]) and why? A: float64. NumPy needs one common type; since one element is a float, all elements are promoted (upcast) to float so no information is lost.