How do I access the ith column of a NumPy multidimensional array?

To access the ith column of a NumPy multidimensional array (specifically a 2D array), use slicing syntax with a colon (:) for rows and the column index i for columns. Here’s a detailed explanation with examples:

Basic Syntax

import numpy as np

# Create a 2D array
arr = np.array([[1, 2, 3],
                [4, 5, 6],
                [7, 8, 9]])

# Access the i-th column (0-based index)
column_i = arr[:, i]  # : selects all rows, i selects the column

Key Points

Indexing starts at 0: The first column is i=0, the second is i=1, etc.
Result is a 1D array: The returned column is a flattened view (not a 2D sub-array).
Modifications affect the original array: Since the result is a view (not a copy), changes propagate to the original array.

Examples

1. Accessing Columns

import numpy as np

arr = np.array([[1, 2, 3],
                [4, 5, 6],
                [7, 8, 9]])

# First column (i=0)
col0 = arr[:, 0]  # [1, 4, 7]

# Second column (i=1)
col1 = arr[:, 1]  # [2, 5, 8]

# Last column (i=-1)
col_last = arr[:, -1]  # [3, 6, 9]

2. Modifying a Column

# Double the values in the second column (i=1)
arr[:, 1] *= 2

# Result:
# [[1,  4, 3],
#  [4, 10, 6],
#  [7, 16, 9]]

3. Keeping Column as a 2D Array (Use `reshape` or `newaxis`)

By default, arr[:, i] returns a 1D array. To retain a 2D shape (e.g., column vector):

col0_2d = arr[:, [0]]      # [[1], [4], [7]] (shape: 3x1)
# OR
col0_2d = arr[:, 0:1]     # [[1], [4], [7]]

# Using np.newaxis
col0_2d = arr[:, 0][:, np.newaxis]  # [[1], [4], [7]]

4. Accessing Multiple Columns

# Select columns 0 and 2
subset = arr[:, [0, 2]]  # [[1, 3],
                         #  [4, 6],
                         #  [7, 9]]

Edge Cases & Pitfalls

Invalid Column Index: Using i >= number of columns causes IndexError.

  # Example with 3 columns (indices: 0, 1, 2)
  arr[:, 3]  # ❌ IndexError: index 3 is out of bounds

Empty Arrays: If the array has 0 columns, accessing any column fails.
1D Arrays: “Columns” don’t exist in 1D arrays. Use arr[i] to access elements.

Why This Works

NumPy uses row-major order (C-style), so contiguous elements in a row are stored adjacently. When accessing a column, NumPy jumps between rows using strides.
The syntax [:, i] leverages basic slicing, which returns a view (not a copy) for memory efficiency.

Practical Use Case

# Extract column names from a structured dataset
data = np.array([["Alice", 30, "Engineer"],
                 ["Bob", 25, "Designer"]])

# Get the "Age" column (i=1)
ages = data[:, 1].astype(int)  # [30, 25]

By mastering this slicing technique, you efficiently manipulate columns in NumPy arrays for data analysis, machine learning, and scientific computing.