Skip to main content

​
Overview

Scrapling is built on a modular, layered architecture that separates concerns between fetching, parsing, and data extraction. The framework is designed around three core pillars:

Fetchers

Handle HTTP requests and browser automation

Parser

Process and navigate HTML/XML documents

Sessions

Manage persistent connections and state

​
Core Components

​
Fetcher Layer

The fetcher layer provides a unified interface for making web requests, abstracting away the differences between various HTTP clients and browser automation tools. 
from scrapling import Fetcher, StealthyFetcher, DynamicFetcher

# Simple HTTP requests
response = Fetcher.fetch('https://example.com')

# Browser-based with JavaScript execution
response = DynamicFetcher.fetch('https://example.com')

# Stealth mode with anti-detection
response = StealthyFetcher.fetch('https://example.com')
All fetchers return a unified Response object that inherits from the Selector class, providing immediate parsing capabilities.

​
Parser Layer

At the heart of Scrapling is the Selector class, built on top of lxml for high-performance HTML/XML parsing. Key Features:
  • CSS and XPath selector support
  • Adaptive element relocation (survives page structure changes)
  • Rich text extraction with TextHandler
  • Attribute handling with AttributesHandler
  • Tree navigation (parent, children, siblings)
from scrapling import Selector

# From HTML string
html = '<div class="product"><h2>Item</h2></div>'
selector = Selector(html)

# CSS selectors
title = selector.css('.product h2::text').get()

# XPath selectors
title = selector.xpath('//div[@class="product"]/h2/text()').get()

# Find methods with filters
product = selector.find('div', class_='product')

​
Response System

The Response class extends Selector with HTTP-specific metadata: 
response = Fetcher.fetch('https://httpbin.org/get')

# HTTP metadata
print(response.status)          # 200
print(response.headers)         # Response headers
print(response.cookies)         # Cookies dict
print(response.request_headers) # Request headers

# Parsing capabilities (inherited from Selector)
title = response.css('title::text').get()
links = response.css('a::attr(href)').getall()

# JSON responses
data = response.json()

​
Architecture Diagram

Request Flow

User Code
   ↓
Fetcher (Fetcher/DynamicFetcher/StealthyFetcher)
   ↓
Engine Layer (curl_cffi/Playwright)
   ↓
Response (extends Selector)
   ↓
Selector (lxml-based parser)

Core Inheritance

Selector (base parser)
   ↑
Response (adds HTTP metadata)
   ↑
Spider Results (optional spider framework)

​
Engine Layer

Scrapling uses different engines depending on the fetcher type:

​
Static Engine (curl_cffi)

Used by Fetcher and AsyncFetcher for fast HTTP requests with browser impersonation: 
# Located in: scrapling/engines/static.py
from scrapling.fetchers import Fetcher

response = Fetcher.fetch(
    'https://httpbin.org/get',
    impersonate='chrome',  # Browser fingerprint
    stealthy_headers=True, # Auto-generate realistic headers
    timeout=30,
    retries=3
)
Features:
  • Browser impersonation (Chrome, Firefox, Safari, Edge)
  • HTTP/2 and HTTP/3 support
  • Automatic header generation
  • Connection pooling with sessions

​
Browser Engine (Playwright)

Used by DynamicFetcher and StealthyFetcher for JavaScript-heavy sites: 
# Located in: scrapling/engines/_browsers/
from scrapling.fetchers import DynamicFetcher

response = DynamicFetcher.fetch(
    'https://example.com',
    headless=True,
    disable_resources=True,  # Block images, fonts, etc.
    network_idle=True,       # Wait for network idle
    load_dom=True           # Wait for DOM ready
)
Features:
  • Real browser automation (Chromium-based)
  • JavaScript execution
  • Network interception and resource blocking
  • Page pooling for sessions
  • Stealth mode with anti-detection evasion

​
Data Flow

​
1. Request Initialization

# User makes a request
response = Fetcher.fetch('https://example.com')

​
2. Configuration Merging

The fetcher merges default settings with request-specific parameters: 
# From scrapling/engines/static.py
class _ConfigurationLogic:
    def _merge_request_args(self, **method_kwargs):
        # Merges session defaults with request params
        final_args = {
            'headers': self._headers_job(...),
            'timeout': self._get_param(kwargs, 'timeout', self._default_timeout),
            'impersonate': _select_random_browser(impersonate),
            # ... more args
        }

​
3. Engine Execution

The appropriate engine executes the request: 
# HTTP requests use curl_cffi
session.request(method, **request_args)

# Browser requests use Playwright
page.goto(url, referer=referer)

​
4. Response Creation

Raw responses are converted to unified Response objects: 
# From scrapling/engines/toolbelt/convertor.py
class ResponseFactory:
    @staticmethod
    def from_http_request(response, selector_config, meta):
        return Response(
            url=str(response.url),
            content=response.content,
            status=response.status_code,
            # ... more fields
        )

​
5. Parsing Layer Access

The Response inherits all Selector parsing methods: 
# User extracts data using Selector methods
title = response.css('title::text').get()
links = response.find_all('a', href=True)

​
Design Principles

​
Lazy Imports

Scrapling uses lazy imports for faster startup times: 
# From scrapling/__init__.py
_LAZY_IMPORTS = {
    "Fetcher": ("scrapling.fetchers", "Fetcher"),
    "Selector": ("scrapling.parser", "Selector"),
    # Only imported when accessed
}

def __getattr__(name: str):
    if name in _LAZY_IMPORTS:
        module_path, class_name = _LAZY_IMPORTS[name]
        module = __import__(module_path, fromlist=[class_name])
        return getattr(module, class_name)

​
Unified Response Interface

All fetchers return the same Response type, making it easy to switch between different fetching strategies: 
# All return Response objects
response1 = Fetcher.fetch(url)          # HTTP client
response2 = DynamicFetcher.fetch(url)   # Browser automation  
response3 = StealthyFetcher.fetch(url)  # Stealth browser

# Same API for all
data = response1.css('.product::text').get()
data = response2.css('.product::text').get()
data = response3.css('.product::text').get()

​
Separation of Concerns

Each layer has a clear responsibility:
  • Fetchers: Network communication and browser control
  • Engines: Low-level HTTP/browser implementation
  • Parser: HTML/XML processing and navigation
  • Sessions: State management and connection pooling
  • Toolbelt: Shared utilities (proxy rotation, fingerprints, etc.)

​
Performance Optimizations

Cached Properties: 
# From scrapling/parser.py
class Selector:
    @property
    def tag(self) -> str:
        if not self.__tag:
            self.__tag = str(self._root.tag)
        return self.__tag  # Computed once, cached
Pre-compiled XPath: 
# Pre-compiled for efficiency
_find_all_elements = XPath(".//*")
_find_all_elements_with_spaces = XPath(".//*[normalize-space(text())]")
Element Conversion: 
def __elements_convertor(self, elements):
    # Store config once, reuse for all elements
    url = self.url
    encoding = self.encoding
    adaptive = self.__adaptive_enabled
    
    return Selectors(
        Selector(root=el, url=url, encoding=encoding, adaptive=adaptive)
        for el in elements
    )

​
Extension Points

Scrapling is designed to be extensible:

​
Custom Storage System

Implement custom storage for adaptive element relocation: 
from scrapling.core.storage import StorageSystemMixin
from functools import lru_cache

@lru_cache(maxsize=128)
class RedisStorage(StorageSystemMixin):
    def save(self, element, identifier):
        # Custom save logic
        pass
    
    def retrieve(self, identifier):
        # Custom retrieve logic
        pass

# Use custom storage
selector = Selector(
    html,
    adaptive=True,
    storage=RedisStorage,
    storage_args={'host': 'localhost'}
)

​
Custom Fetcher

Extend BaseFetcher for custom fetching logic: 
from scrapling.engines.toolbelt.custom import BaseFetcher

class CustomFetcher(BaseFetcher):
    @classmethod
    def fetch(cls, url: str, **kwargs):
        # Custom fetching logic
        # Must return Response object
        pass

​
File Structure

scrapling/
├── __init__.py              # Lazy imports, main exports
├── parser.py                # Selector and Selectors classes
├── fetchers/
│   ├── __init__.py         # Fetcher exports
│   ├── requests.py         # HTTP-based fetchers
│   ├── chrome.py           # DynamicFetcher
│   └── stealth_chrome.py   # StealthyFetcher
├── engines/
│   ├── __init__.py
│   ├── static.py           # curl_cffi engine
│   ├── constants.py        # Browser arguments
│   ├── _browsers/          # Playwright engine
│   │   ├── _base.py       # Session base classes
│   │   ├── _controllers.py # DynamicSession
│   │   ├── _stealth.py    # StealthySession
│   │   └── _page.py       # Page pooling
│   └── toolbelt/           # Shared utilities
│       ├── custom.py      # Response, BaseFetcher
│       ├── convertor.py   # ResponseFactory
│       ├── fingerprints.py # Header generation
│       └── proxy_rotation.py # ProxyRotator
└── core/
    ├── custom_types.py     # TextHandler, AttributesHandler
    ├── storage.py          # Adaptive storage system
    ├── translator.py       # CSS to XPath conversion
    └── utils/              # Utilities

​
Next Steps

Fetchers

Learn about different fetcher types

Parsing

Deep dive into the parsing system

Sessions

Understand session management

Build docs developers (and LLMs) love

Get started for freeTalk to us