Creating structured arrays in NumPy

Introduction

Structured arrays let you store different types of data together in one array. This helps when you want to keep related information organized, like a table with columns of different types.

You want to store data like a spreadsheet with columns of names, ages, and scores.

You need to handle data with mixed types, such as strings and numbers, in one array.

You want to access data by column names instead of just by position.

You want to save memory by using arrays instead of lists of dictionaries.

You want to perform fast operations on complex data with NumPy.

Syntax

NumPy

import numpy as np

# Define the data type with field names and types
dtype = [('name', 'U10'), ('age', 'i4'), ('score', 'f4')]

# Create an empty structured array
structured_array = np.zeros(3, dtype=dtype)

# Or create from a list of tuples
data = [('Alice', 25, 88.5), ('Bob', 30, 92.0), ('Cathy', 22, 79.5)]
structured_array = np.array(data, dtype=dtype)

The dtype defines the structure: each field has a name and a data type.

Use 'U10' for a string of max length 10, 'i4' for 4-byte integer, 'f4' for 4-byte float.

Examples

This creates an empty structured array with no rows.

NumPy

import numpy as np

dtype = [('name', 'U10'), ('age', 'i4'), ('score', 'f4')]

# Empty array with 0 rows
empty_array = np.zeros(0, dtype=dtype)
print(empty_array)

This creates a structured array with just one row.

NumPy

import numpy as np

dtype = [('name', 'U10'), ('age', 'i4'), ('score', 'f4')]

# Array with one element
one_element = np.array([('Diana', 28, 85.0)], dtype=dtype)
print(one_element)

You can get all values of a field by using the field name.

NumPy

import numpy as np

dtype = [('name', 'U10'), ('age', 'i4'), ('score', 'f4')]

# Accessing fields
students = np.array([('Eve', 21, 90.0), ('Frank', 24, 87.5)], dtype=dtype)
print(students['name'])
print(students['score'])

Sample Program

This program creates a structured array of students with name, age, and score. It prints the array, updates Bob's score, and prints the updated array.

NumPy

import numpy as np

# Define the structured data type
student_dtype = [('name', 'U10'), ('age', 'i4'), ('score', 'f4')]

# Create an array of students
students = np.array([
    ('Alice', 25, 88.5),
    ('Bob', 30, 92.0),
    ('Cathy', 22, 79.5)
], dtype=student_dtype)

print("Before update:")
print(students)

# Update Bob's score
for index, student in enumerate(students):
    if student['name'] == 'Bob':
        students[index]['score'] = 95.0

print("\nAfter update:")
print(students)

Important Notes

Time complexity for accessing or updating a field is O(n) if you loop, but O(1) if you use vectorized operations.

Space complexity is efficient because all data is stored in a single NumPy array.

Common mistake: forgetting to specify dtype or mismatching data types causes errors.

Use structured arrays when you want fast access by field names and compact storage, instead of lists of dictionaries.

Summary

Structured arrays store mixed data types in one array with named fields.

You define the structure using a dtype with field names and types.

They help organize data like tables and allow easy access by column names.

Practice

1. What is the main purpose of creating a structured array in numpy?

easy

A. To create arrays with only one data type

B. To store data with different types in named fields within one array

C. To speed up numerical calculations on large arrays

D. To visualize data using plots

Solution

1. Step 1: Understand structured arrays

   Structured arrays allow storing mixed data types in one array with named fields.

2. Step 2: Compare options

   Only To store data with different types in named fields within one array correctly describes this purpose; others describe unrelated features.

3. Final Answer:

   To store data with different types in named fields within one array -> Option B

4. Quick Check:

   Structured arrays = mixed types + named fields [OK]

Hint: Structured arrays hold mixed data types with names [OK]

Common Mistakes:

• Thinking structured arrays only hold one data type
• Confusing structured arrays with plotting functions
• Assuming structured arrays speed up all calculations

2. Which of the following is the correct way to define a structured array dtype with fields 'name' (string) and 'age' (integer)?

easy

A. dtype = [('name', 'U10'), ('age', 'i4')]

B. dtype = ['name': str, 'age': int]

C. dtype = {'name': 'string', 'age': 'int'}

D. dtype = [('name', str), ('age', float)]

Solution

1. Step 1: Recall dtype syntax for structured arrays

   Structured array dtypes are defined as a list of tuples: (field_name, data_type).

2. Step 2: Check each option

   dtype = [('name', 'U10'), ('age', 'i4')] uses correct tuple syntax with numpy string and integer types. Others use invalid syntax or wrong types.

3. Final Answer:

   dtype = [('name', 'U10'), ('age', 'i4')] -> Option A

4. Quick Check:

   Structured dtype = list of (name, type) tuples [OK]

Hint: Use list of (field, type) tuples for dtype [OK]

Common Mistakes:

• Using dictionary syntax instead of list of tuples
• Using Python types instead of numpy dtype strings
• Mixing float type for integer fields

3. What will be the output of the following code?

import numpy as np
dtype = [('id', 'i4'), ('score', 'f4')]
data = np.array([(1, 9.5), (2, 8.0)], dtype=dtype)
print(data['score'])

medium

A. Error: KeyError

B. [1 2]

C. [(1, 9.5) (2, 8.0)]

D. [9.5 8. ]

Solution

1. Step 1: Understand structured array creation

   The array has fields 'id' (int) and 'score' (float), with two records.

2. Step 2: Access the 'score' field

   Accessing data['score'] returns an array of the 'score' values: [9.5, 8.0].

3. Final Answer:

   [9.5 8. ] -> Option D

4. Quick Check:

   Accessing field returns array of that column [OK]

Hint: Access fields by name to get column arrays [OK]

Common Mistakes:

• Expecting full records instead of single field
• Confusing field names causing KeyError
• Thinking output is list of tuples

4. Identify the error in the following code that tries to create a structured array:

import numpy as np
dtype = [('name', 'U5'), ('age', 'i4')]
data = np.array([('Alice', 25), ('Bob', 30)], dtype=dtype)
print(data['age'])

medium

A. The dtype definition is missing field names

B. The tuple elements should be lists, not tuples

C. No error; code runs correctly

D. The string length 'U5' is too short for 'Alice'

Solution

1. Step 1: Check dtype and data compatibility

   'U5' means Unicode string of length 5, 'Alice' has 5 characters, so it fits.

2. Step 2: Verify data structure and code correctness

   Data is a list of tuples matching dtype fields; code runs without error and prints ages.

3. Final Answer:

   No error; code runs correctly -> Option C

4. Quick Check:

   String length matches field size; tuples allowed [OK]

Hint: Check string length matches longest string [OK]

Common Mistakes:

• Assuming string length too short causes error
• Thinking tuples are invalid for data input
• Believing dtype must be dictionary

5. You have a list of employee data: [('John', 28, 50000), ('Jane', 32, 60000), ('Doe', 24, 45000)]. How do you create a structured array with fields 'name' (string, max 10 chars), 'age' (int), and 'salary' (float) to store this data?

hard

A. dtype = [('name', 'U10'), ('age', 'i4'), ('salary', 'f8')]; np.array(data, dtype=dtype)

B. dtype = [('name', str), ('age', int), ('salary', float)]; np.array(data)

C. dtype = {'name': 'U10', 'age': 'i4', 'salary': 'f8'}; np.array(data)

D. dtype = [('name', 'S10'), ('age', 'i4'), ('salary', 'f4')]; np.array(data)

Solution

1. Step 1: Define correct dtype for fields

   Use list of tuples with field names and numpy types: 'U10' for Unicode string max 10 chars, 'i4' for int, 'f8' for float64.

2. Step 2: Create structured array with data and dtype

   Pass data and dtype to np.array to create structured array storing all fields correctly.

3. Final Answer:

   dtype = [('name', 'U10'), ('age', 'i4'), ('salary', 'f8')]; np.array(data, dtype=dtype) -> Option A

4. Quick Check:

   Use list of (field, type) tuples and pass dtype [OK]

Hint: Use Unicode string 'U10' for names, correct numeric types [OK]

Common Mistakes:

• Using Python types instead of numpy dtype strings
• Using dictionary instead of list of tuples for dtype
• Choosing byte string 'S10' instead of Unicode 'U10'