NumPy Record Arrays – Mastering Python

Understanding NumPy Record Arrays

A NumPy record array is essentially a structured array where each element contains multiple fields, each with its own data type. This organization facilitates working with data containing diverse attributes, such as sensor readings with timestamps, scientific measurements with units, or even customer data with names, IDs, and purchase amounts. Think of it as a NumPy array of “rows” where each “row” is a structured data point.

Creating Record Arrays

Let’s illustrate how to create and manipulate record arrays. We’ll start by defining a structured data type using numpy.dtype.

import numpy as np

data_type = np.dtype([('name', 'U10'), ('age', 'i4'), ('height', 'f4')])

records = np.array([('Alice', 30, 1.75), ('Bob', 25, 1.80), ('Charlie', 35, 1.70)], dtype=data_type)

print(records)

This code creates a record array named records. The dtype argument specifies the fields: ‘name’ (unicode string of length 10), ‘age’ (4-byte integer), and ‘height’ (4-byte float). The array then populates with sample data.

Accessing Data in Record Arrays

Accessing data within a record array is straightforward. You can access individual fields using attribute-like notation or indexing.

print(records['name'])  # Accesses the 'name' field
print(records['age'])   # Accesses the 'age' field

print(records[0])       # Accesses the first record
print(records[0]['name']) # Accesses the 'name' field of the first record

tall_people = records[records['height'] > 1.75]
print(tall_people)

The example shows how easily you can select and filter data based on specific fields, highlighting the power of combining record array structure with NumPy’s powerful indexing capabilities.

Creating Record Arrays from Existing Data

You can also create record arrays from existing dictionaries or lists of dictionaries.

data = [{'name': 'David', 'age': 28, 'height': 1.85},
        {'name': 'Eve', 'age': 32, 'height': 1.68}]

records2 = np.array(data, dtype=data_type)
print(records2)

This demonstrates creating a record array by directly converting a list of dictionaries. NumPy efficiently handles the type conversion and data structuring.

Beyond the Basics: More Advanced Usage

Record arrays offer significant flexibility. You can perform vectorized operations on specific fields, similar to operations on standard NumPy arrays. For example, you could easily calculate the average age or the standard deviation of heights within the record array. The possibilities are vast, making record arrays indispensable for a wide range of data manipulation tasks involving structured data within the NumPy ecosystem.