如何创建包含不同数据类型的结构化NumPy数组

Got it, let's fix this to make your code cleaner and reusable!

The key is to define a structured NumPy dtype that matches your desired columns, then convert your byte-encoded data directly into this structured array in one go. Here's how to do it step by step:

Step 1: Define the structured dtype #

First, specify the data types for each field you care about—id (integer), datetime (we'll use NumPy's efficient datetime64 type), and anomaly_length (integer):

import numpy as np

# Define the structured data type
structured_dtype = [
    ('id', int),
    ('datetime', 'datetime64[ns]'),
    ('anomaly_length', int)
]

Step 2: Decode byte data and create the structured array #

Your raw data has byte strings (like b'1'), so we first decode each element to regular strings, then feed this into np.array with our defined dtype:

# Decode byte columns to strings for all rows (skip the header row)
decoded_rows = [[col.decode('utf-8') for col in row] for row in data[1:, 0:3]]

# Create the structured array in one line
structured_data = np.array(decoded_rows, dtype=structured_dtype)

Step 3: Use the structured array in your function #

Now you can pass the structured array (or its individual fields) directly to special_find without repeating conversions. For example:

# Access individual fields by name (no more separate arrays!)
special_find(structured_data['id'], structured_data['datetime'], structured_data['anomaly_length'])

Bonus: Even simpler if reading directly from CSV #

If you haven't already loaded the CSV into data, you can skip the manual byte handling entirely by reading directly into a structured array with np.genfromtxt:

structured_data = np.genfromtxt(
    'your_file.csv',
    delimiter=',',
    skip_header=1,  # Skip the first row (headers)
    dtype=structured_dtype,
    usecols=(0, 1, 2)  # Only load the columns we need
)

This approach keeps your code concise, avoids redundant conversions, and makes special_find calls much cleaner—perfect for repeated use!

内容的提问来源于stack exchange，提问作者wake-0