Files
Priceflow/app/services/parsers/eleclerc_parser.py
T

76 lines
2.7 KiB
Python

from bs4 import BeautifulSoup
from app.services.parsers.base_parser import BaseParser, ProductResult
import logging
logger = logging.getLogger(__name__)
class ELeclercParser(BaseParser):
def __init__(self):
super().__init__("e-leclerc.com", "https://www.e.leclerc")
def parse_search_results(self, html: str, query: str, search_url: str) -> list[ProductResult]:
soup = BeautifulSoup(html, "html.parser")
results = []
# E.Leclerc products
# Based on config: a.product-card-link
# We look for the card container
cards = soup.select("div[class*='product-card'], div[class*='product-item']")
# If no cards found with div, try looking for the links directly and finding parents
if not cards:
links = soup.select("a.product-card-link")
cards = [link.parent for link in links]
for card in cards:
try:
# Link
link_el = card.select_one("a.product-card-link") or card if card.name == 'a' else card.find('a')
if not link_el:
continue
href = link_el.get('href')
if not href:
continue
url = self.make_absolute_url(href)
# Title
title = None
# Try specific title classes
title_el = card.select_one("[class*='product-title'], [class*='product-label']")
if title_el:
title = title_el.get_text(strip=True)
else:
title = link_el.get_text(strip=True)
if not title:
continue
# Image
img_url = self.extract_image_url(card)
# Price
price = None
price_el = card.select_one("[class*='price'], [class*='amount']")
if price_el:
price = self.parse_price_text(price_el.get_text())
results.append(ProductResult(
title=title,
url=url,
source="E.Leclerc",
price=price,
currency="EUR",
in_stock=True,
image_url=img_url,
snippet=f"Product from E.Leclerc"
))
except Exception as e:
logger.error(f"Error parsing E.Leclerc product: {e}")
continue
logger.info(f"ELeclercParser found {len(results)} results")
return results