linkding/bookmarks/services/website_loader.py
Taku Izumi 937858cf58
Fix website scraper decoding content incorrectly (#126)
* Avoid stall on web scraping

This patch fixes stall on web scraping.
I encountered a stall (scraping never ends) when adding
a bookmark of some site.
To avoid this case, adding a timeout parameter at requests.get()
function is a solution.

Signed-off-by: Taku Izumi <admin@orz-style.com>

* Avoid character corruption of scraping some Japanese sites

This patch fixes character corruption of scraping some Japanese
sites. To avoid character corruption, I use r.content instead
of r.text in load_page function.

The reason of character corruption is encoding problem, I think.
r.text handles data as unicode encoded text, so if scraping
web site's charset is not unicode encoded, character corruption
occurs. r.content handles data as str[], we can avoid encoding
problem.

Signed-off-by: Taku Izumi <admin@orz-style.com>

* use charset_normalizer to determine response encoding

Co-authored-by: Taku Izumi <admin@orz-style.com>
Co-authored-by: Sascha Ißbrücker <sascha.issbruecker@googlemail.com>
2021-08-25 10:16:23 +02:00

44 lines
1.3 KiB
Python

from dataclasses import dataclass
import requests
from bs4 import BeautifulSoup
from charset_normalizer import from_bytes
@dataclass
class WebsiteMetadata:
url: str
title: str
description: str
def to_dict(self):
return {
'url': self.url,
'title': self.title,
'description': self.description,
}
def load_website_metadata(url: str):
title = None
description = None
try:
page_text = load_page(url)
soup = BeautifulSoup(page_text, 'html.parser')
title = soup.title.string if soup.title is not None else None
description_tag = soup.find('meta', attrs={'name': 'description'})
description = description_tag['content'] if description_tag is not None else None
finally:
return WebsiteMetadata(url=url, title=title, description=description)
def load_page(url: str):
r = requests.get(url, timeout=10)
# Use charset_normalizer to determine encoding that best matches the response content
# Several sites seem to specify the response encoding incorrectly, so we ignore it and use custom logic instead
# This is different from Response.text which does respect the encoding specified in the response first,
# before trying to determine one
results = from_bytes(r.content)
return str(results.best())