Record arrays in NumPy

Introduction

Record arrays let you store different types of data together in one array. This helps when you want to keep related information like names, ages, and scores in one place.

You want to store a list of people with their name, age, and height together.

You have data with different types like strings and numbers mixed in one table.

You want to access columns by name instead of by position.

You need to save structured data that looks like a spreadsheet but in an array.

You want to perform fast operations on mixed-type data using numpy.

Syntax

NumPy

import numpy as np

# Define a record array with fields and types
record_array = np.rec.array(
    [(value1, value2, ...), (value1, value2, ...)],
    dtype=[('field1', type1), ('field2', type2), ...]
)

The dtype defines the name and type of each field.

You can access fields by name like record_array.field1.

Examples

This creates an empty record array with fields 'name' and 'age'. Useful to start with no data.

NumPy

import numpy as np

# Empty record array with defined fields
empty_records = np.rec.array([], dtype=[('name', 'U10'), ('age', 'i4')])
print(empty_records)

Shows a record array with a single entry. You can access the fields by name.

NumPy

import numpy as np

# Record array with one element
one_record = np.rec.array([('Alice', 30)], dtype=[('name', 'U10'), ('age', 'i4')])
print(one_record)
print(one_record.name)
print(one_record.age)

This example has multiple records. You can get all names or ages as arrays.

NumPy

import numpy as np

# Record array with multiple elements
people = np.rec.array([
    ('Bob', 25),
    ('Carol', 40),
    ('Dave', 35)
], dtype=[('name', 'U10'), ('age', 'i4')])
print(people)
print(people.name)
print(people.age)

Shows how to access the last record and its fields.

NumPy

import numpy as np

# Accessing last element
print(people[-1])
print(people[-1].name)
print(people[-1].age)

Sample Program

This program creates a record array with three people, prints it, adds a new person, and prints the updated array. Then it shows how to access each field by name.

NumPy

import numpy as np

# Create a record array with three people
people = np.rec.array([
    ('Alice', 28, 5.5),
    ('Bob', 34, 6.0),
    ('Carol', 22, 5.7)
], dtype=[('name', 'U10'), ('age', 'i4'), ('height', 'f4')])

print("Before adding new record:")
print(people)

# Add a new record by creating a new array with one more element
new_person = np.rec.array([('Dave', 30, 5.9)], dtype=people.dtype)
people = np.concatenate((people, new_person))

print("\nAfter adding new record:")
print(people)

# Access fields by name
print("\nNames:", people.name)
print("Ages:", people.age)
print("Heights:", people.height)

Important Notes

Time complexity for accessing a field is O(1) because fields are stored separately internally.

Space complexity is similar to normal numpy arrays but slightly more due to field names.

Common mistake: Trying to add records by appending directly to the record array. Instead, create a new array and concatenate.

Use record arrays when you want structured data with named fields and mixed types. Use pandas DataFrame if you need more complex table operations.

Summary

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

You can access data by field names like array.field.

They are useful for simple structured data and fast access in numpy.