numpy.fromrecords() method is a powerful tool in the NumPy library that allows you to create structured arrays from a sequence of tuples or other array-like objects.

Let's understand the help of an example:

import numpy as np

# Define a list of records
records = [(1, 'Alice', 25.5), (2, 'Bob', 30.0), (3, 'Charlie', 28.0)]

# Define the data type
dtype = [('id', 'i4'), ('name', 'U10'), ('age', 'f4')]

# Create the structured array
structured_array = np.fromrecords(records, dtype=dtype)

print(structured_array)

Output:

[(1, 'Alice', 25.5) (2, 'Bob', 30. ) (3, 'Charlie', 28. )]

Syntax of Numpy fromrecords():

numpy.fromrecords(recList, dtype=None, shape=None, aligned=False, byteorder=None)

Parameters:

• recList: A list of tuples or structured data to be converted into a structured NumPy array.
• dtype (optional): The data type of the resulting structured array. If not provided, NumPy will infer the type from the input data.
• shape (optional): Shape of the output array. Defaults to one-dimensional.
• aligned (optional): If True, aligns fields to their natural alignment.
• byteorder (optional): Specifies the byte order of the output array.

Accessing Structured Array Fields

We can access specific fields (columns) in the structured array by their names.

import numpy as np

# Define a list of records
records = [(1, 'Alice', 25.5), (2, 'Bob', 30.0), (3, 'Charlie', 28.0)]

# Define the data type
dtype = [('id', 'i4'), ('name', 'U10'), ('age', 'f4')]

# Create the structured array
structured_array = np.fromrecords(records, dtype=dtype)

# Access the 'name' field
print(structured_array['name'])

# Access the 'age' field
print(structured_array['age'])

Output:

[(1, 'Alice', 25.5) (2, 'Bob', 30. ) (3, 'Charlie', 28. )]