Introduction

Improving code quality is always an important issue in software development. In this article, we take a data collection crawler system as an example and explain specifically how to apply design principles and best practices through step-by-step refactoring.

Code before improvement

First, we start with a very simple web scraper with all functionality integrated into one class.

Translated with DeepL.com (free version)

project_root/
├── web_scraper.py
├── main.py
└── requirements.txt

web_scraper.py

import requests
import json
import sqlite3

class WebScraper:
    def __init__(self, url):
        self.url = url

    def fetch_data(self):
        response = requests.get(self.url)
        data = response.text
        parsed_data = self.parse_data(data)
        enriched_data = self.enrich_data(parsed_data)
        self.save_data(enriched_data)
        return enriched_data

    def parse_data(self, data):
        return json.loads(data)

    def enrich_data(self, data):
        # Apply business logic here
        # Example: extract only data containing specific keywords
        return {k: v for k, v in data.items() if 'important' in v.lower()}

    def save_data(self, data):
        conn = sqlite3.connect('test.db')
        cursor = conn.cursor()
        cursor.execute('INSERT INTO data (json_data) VALUES (?)', (json.dumps(data),))
        conn.commit()
        conn.close()

main.py

from web_scraper import WebScraper

def main():
    scraper = WebScraper('https://example.com/api/data')
    data = scraper.fetch_data()
    print(data)

if __name__ == "__main__":
    main()

Points to be improved

Violates the principle of single responsibility: one class is responsible for all data acquisition, analysis, enrichment, and storage
Unclear business logic: business logic is embedded in the enrich_data method, but mixed with other processing
Lack of reusability: functions are tightly coupled, making individual reuse difficult
Testing difficulties: difficult to test individual functions independently
Configuration rigidity: database paths and other settings are embedded directly in the code

Refactoring phase

1. Separation of responsibilities: separation of data acquisition, analysis, and storage

Major Change: Separation of responsibilities for data acquisition, analysis, and storage into separate classes
Objective: Apply single responsibility principle, introduce environmental variables

directory structure

project_root/
├── data_fetcher.py
├── data_parser.py
├── data_saver.py
├── data_enricher.py
├── web_scraper.py
├── main.py
└── requirements.txt

data_enricher.py

class DataEnricher:
    def enrich(self, data):
        return {k: v for k, v in data.items() if 'important' in v.lower()}

web_scraper.py

from data_fetcher import DataFetcher
from data_parser import DataParser
from data_enricher import DataEnricher
from data_saver import DataSaver

class WebScraper:
    def __init__(self, url):
        self.url = url
        self.fetcher = DataFetcher()
        self.parser = DataParser()
        self.enricher = DataEnricher()
        self.saver = DataSaver()

    def fetch_data(self):
        raw_data = self.fetcher.fetch(self.url)
        parsed_data = self.parser.parse(raw_data)
        enriched_data = self.enricher.enrich(parsed_data)
        self.saver.save(enriched_data)
        return enriched_data

This change clarifies the responsibilities of each class and improves reusability and testability. However, the business logic is still embedded in the DataEnricher class.

2. introduction of interfaces and dependency injection

Main change: Introduce interfaces and implement dependency injection.
Purpose: increase flexibility and extensibility, extend environment variables, abstract business logic

directory structure

project_root/
├── interfaces/
│   ├── __init__.py
│   ├── data_fetcher_interface.py
│   ├── data_parser_interface.py
│   ├── data_enricher_interface.py
│   └── data_saver_interface.py
├── implementations/
│   ├── __init__.py
│   ├── http_data_fetcher.py
│   ├── json_data_parser.py
│   ├── keyword_data_enricher.py
│   └── sqlite_data_saver.py
├── web_scraper.py
├── main.py
└── requirements.txt

interfaces/data_fetcher_interface.py

from abc import ABC, abstractmethod

class DataFetcherInterface(ABC):
    @abstractmethod
    def fetch(self, url: str) -> str:
        pass

interfaces/data_parser_interface.py

from abc import ABC, abstractmethod
from typing import Dict, Any

class DataParserInterface(ABC):
    @abstractmethod
    def parse(self, raw_data: str) -> Dict[str, Any]:
        pass

interfaces/data_enricher_interface.py

from abc import ABC, abstractmethod
from typing import Dict, Any

class DataEnricherInterface(ABC):
    @abstractmethod
    def enrich(self, data: Dict[str, Any]) -> Dict[str, Any]:
        pass

interfaces/data_saver_interface.py

from abc import ABC, abstractmethod
from typing import Dict, Any

class DataSaverInterface(ABC):
    @abstractmethod
    def save(self, data: Dict[str, Any]) -> None:
        pass

implementations/keyword_data_enricher.py

import os
from interfaces.data_enricher_interface import DataEnricherInterface

class KeywordDataEnricher(DataEnricherInterface):
    def __init__(self):
        self.keyword = os.getenv('IMPORTANT_KEYWORD', 'important')

    def enrich(self, data):
        return {k: v for k, v in data.items() if self.keyword in str(v).lower()}

web_scraper.py

from interfaces.data_fetcher_interface import DataFetcherInterface
from interfaces.data_parser_interface import DataParserInterface
from interfaces.data_enricher_interface import DataEnricherInterface
from interfaces.data_saver_interface import DataSaverInterface

class WebScraper:
    def __init__(self, fetcher: DataFetcherInterface, parser: DataParserInterface, 
                 enricher: DataEnricherInterface, saver: DataSaverInterface):
        self.fetcher = fetcher
        self.parser = parser
        self.enricher = enricher
        self.saver = saver

    def fetch_data(self, url):
        raw_data = self.fetcher.fetch(url)
        parsed_data = self.parser.parse(raw_data)
        enriched_data = self.enricher.enrich(parsed_data)
        self.saver.save(enriched_data)
        return enriched_data

The main changes at this stage are

introduction of an interface to facilitate switching to different implementations
dependency injection to make the WebScraper class more flexible
The fetch_data method has been changed to take url as an argument, making URL specification more flexible.
Business logic has been abstracted as DataEnricherInterface and implemented as KeywordDataEnricher.
The business logic has been made more flexible by allowing keywords to be set using environment variables.

These changes have greatly improved the flexibility and extensibility of the system. However, the business logic remains embedded in the DataEnricherInterface and its implementation. The next step is to further separate this business logic and clearly define it as a domain layer.

3. introduction of domain layer and separation of business logic

In the previous step, the introduction of interfaces increased the flexibility of the system. However, the business logic (in this case, data importance determination and filtering) is still treated as part of the data layer. Based on the concept of domain-driven design, treating this business logic as the central concept of the system and implementing it as an independent domain layer provides the following benefits

centralized management of business logic
more expressive code through the domain model
greater flexibility for changing business rules
ease of testing

Updated directory structure:

project_root/
├── domain/
│   ├── __init__.py
│   ├── scraped_data.py
│   └── data_enrichment_service.py
├── data/
│   ├── __init__.py
│   ├── interfaces/
│   │   ├── __init__.py
│   │   ├── data_fetcher_interface.py
│   │   ├── data_parser_interface.py
│   │   └── data_saver_interface.py
│   ├── implementations/
│   │   ├── __init__.py
│   │   ├── http_data_fetcher.py
│   │   ├── json_data_parser.py
│   │   └── sqlite_data_saver.py
├── application/
│   ├── __init__.py
│   └── web_scraper.py
├── main.py
└── requirements.txt

At this stage, the roles of DataEnricherInterface and KeywordDataEnricher will be moved to the ScrapedData model and DataEnrichmentService at the domain layer. Details of this change are provided below.

Before change (Section 2)

class DataEnricherInterface(ABC):
    @abstractmethod
    def enrich(self, data: Dict[str, Any]) -> Dict[str, Any]:
        pass

class KeywordDataEnricher(DataEnricherInterface):
    def __init__(self):
        self.keyword = os.getenv('IMPORTANT_KEYWORD', 'important')

    def enrich(self, data):
        return {k: v for k, v in data.items() if self.keyword in str(v).lower()}

After modification (Section 3)

@dataclass
class ScrapedData:
    content: Dict[str, Any]
    source_url: str

    def is_important(self) -> bool:
        important_keyword = os.getenv('IMPORTANT_KEYWORD', 'important')
        return any(important_keyword in str(v).lower() for v in self.content.values())

class DataEnrichmentService:
    def __init__(self):
        self.important_keyword = os.getenv('IMPORTANT_KEYWORD', 'important')

    def enrich(self, data: ScrapedData) -> ScrapedData:
        if data.is_important():
            enriched_content = {k: v for k, v in data.content.items() if self.important_keyword in str(v).lower()}
            return ScrapedData(content=enriched_content, source_url=data.source_url)
        return data

This change improves the following.

business logic has been moved to the domain layer, eliminating the need for a DataEnricherInterface.
the KeywordDataEnricher functionality has been merged into the DataEnrichmentService, centralizing the business logic in one place.
The is_important method has been added to the ScrapedData model. This makes the domain model itself responsible for determining the importance of data and makes the domain concept clearer.
DataEnrichmentService now handles ScrapedData objects directly, improving type safety.

The WebScraper class will also be updated to reflect this change.

from data.interfaces.data_fetcher_interface import DataFetcherInterface
from data.interfaces.data_parser_interface import DataParserInterface
from data.interfaces.data_saver_interface import DataSaverInterface
from domain.scraped_data import ScrapedData
from domain.data_enrichment_service import DataEnrichmentService

class WebScraper:
    def __init__(self, fetcher: DataFetcherInterface, parser: DataParserInterface, 
                 saver: DataSaverInterface, enrichment_service: DataEnrichmentService):
        self.fetcher = fetcher
        self.parser = parser
        self.saver = saver
        self.enrichment_service = enrichment_service

    def fetch_data(self, url: str) -> ScrapedData:
        raw_data = self.fetcher.fetch(url)
        parsed_data = self.parser.parse(raw_data)
        scraped_data = ScrapedData(content=parsed_data, source_url=url)
        enriched_data = self.enrichment_service.enrich(scraped_data)
        self.saver.save(enriched_data)
        return enriched_data

This change completely shifts the business logic from the data layer to the domain layer, giving the system a clearer structure. The removal of the DataEnricherInterface and the introduction of the DataEnrichmentService are not just interface replacements, but fundamental changes in the way business logic is handled.

Summary

This article has demonstrated how to improve code quality and apply design principles specifically through a step-by-step refactoring process for the data collection crawler system. The main areas of improvement are as follows.

Separation of Responsibility: Applying the principle of single responsibility, we separated data acquisition, parsing, enrichment, and storage into separate classes.
Introduction of interfaces and dependency injection: greatly increased the flexibility and scalability of the system, making it easier to switch to different implementations.
Introduction of domain model and services: clearly separated the business logic and defined the core concepts of the system.
Adoption of Layered Architecture: Clearly separated the domain, data, and application layers and defined the responsibilities of each layer. 5.Maintain interfaces: Maintained abstraction at the data layer to ensure flexibility in implementation.

These improvements have greatly enhanced the system's modularity, reusability, testability, maintainability, and scalability. In particular, by applying some concepts of domain-driven design, the business logic became clearer and the structure was more flexible to accommodate future changes in requirements. At the same time, by maintaining the interfaces, we ensured the flexibility to easily change and extend the data layer implementation.

It is important to note that this refactoring process is not a one-time event, but part of a continuous improvement process. Depending on the size and complexity of the project, it is important to adopt design principles and DDD concepts at the appropriate level and to make incremental improvements.

Finally, the approach presented in this article can be applied to a wide variety of software projects, not just data collection crawlers. We encourage you to use them as a reference as you work to improve code quality and design.

Refactoring based on design principles: example of a data collection crawler system

Introduction

Code before improvement

Points to be improved

Refactoring phase

1. Separation of responsibilities: separation of data acquisition, analysis, and storage

2. introduction of interfaces and dependency injection

3. introduction of domain layer and separation of business logic

Summary