xkcd / scrapper.py
odehaene
feat: Add dataset
2dc6bb9
import aiohttp
import asyncio
import re
import pandas as pd
from pathlib import Path
from aiolimiter import AsyncLimiter
from typing import Dict, List
from bs4 import BeautifulSoup
from bs4.element import Tag
LIST_COMICS_500_URL = (
"https://www.explainxkcd.com/wiki/index.php/List_of_all_comics_(1-500)"
)
LIST_COMICS_FULL_URL = (
"https://www.explainxkcd.com/wiki/index.php/List_of_all_comics_(full)"
)
def walk_tag(initial_tag: Tag, end_tag_name: str) -> str:
"""
Walk the HTML tree and aggregates all text between an
initial tag and an end tag.
Parameters
----------
initial_tag: BeautifulSoup
end_tag_name: str
Returns
-------
aggregated_text: str
"""
result = []
current_tag = initial_tag
# Walk the HTML
while True:
if current_tag.name in ["p", "dl"]:
result.append(current_tag.get_text(separator=" ", strip=True))
elif current_tag.name == end_tag_name:
# We reached the end tag, break
break
current_tag = current_tag.next_sibling
return "\n".join(result)
async def parse_url_html(
session: aiohttp.ClientSession, url: str, throttler: AsyncLimiter
) -> BeautifulSoup:
"""
Parse the HTML content of a given URL.
The request is sent asynchronously and using a provided request throttler.
If the request fails, we retry up to 5 times.
Parameters
----------
session: aiohttp.ClientSession
url: str
throttler: AsyncLimiter
Returns
-------
BeautifulSoup
"""
for _ in range(5):
try:
# prevent issues with rate limiters
async with throttler:
async with session.get(url, raise_for_status=True) as resp:
html = await resp.text()
return BeautifulSoup(html, "html.parser")
# request failed
except aiohttp.ClientError:
continue
async def scrap_comic(
session: aiohttp.ClientSession, explained_url: str, throttler: AsyncLimiter
) -> Dict[str, str]:
"""
Try to scrap all information for a given XKCD comic using its `explainxkcd.com` URL
Parameters
----------
session: aiohttp.ClientSession
explained_url: str
throttler: AsyncLimiter
Returns
-------
Dict[str, str]
"""
soup = await parse_url_html(session, explained_url, throttler)
# Parse id and title
title_splits = soup.find("h1").text.split(":")
if len(title_splits) > 1:
id = title_splits[0]
title = "".join(title_splits[1:]).strip()
else:
id = None
title = "".join(title_splits).strip()
# Parse explanation
explanation_soup = soup.find("span", {"id": "Explanation"})
try:
explanation = walk_tag(explanation_soup.parent, "span")
except:
explanation = None
# Parse transcript
transcript_soup = soup.find("span", {"id": "Transcript"})
try:
transcript = walk_tag(transcript_soup.parent, "span")
except:
transcript = None
xkcd_url = f"https://www.xkcd.com/{id}"
xkcd_soup = await parse_url_html(session, xkcd_url, throttler)
# Parse image title
try:
image = xkcd_soup.find("div", {"id": "comic"}).img
if title in image:
image_title = image["title"]
else:
image_title = image["alt"]
except:
image_title = None
# Parse image url
try:
image_url = xkcd_soup.find(text=re.compile("https://imgs.xkcd.com"))
except:
image_url = None
return dict(
id=id,
title=title,
image_title=image_title,
url=xkcd_url,
image_url=image_url,
explained_url=explained_url,
transcript=transcript,
explanation=explanation,
)
async def scap_comic_urls(
session: aiohttp.ClientSession, comic_list_url: str
) -> List[str]:
"""
Scrap all XKCD comic URLs from the `explainxkcd.com` website.
Parameters
----------
session: aiohttp.ClientSession
comic_list_url: str
Returns
-------
urls: List[str]
"""
async with session.get(comic_list_url) as resp:
html = await resp.text()
soup = BeautifulSoup(html, "html.parser")
# Hack to easily find comics
create_spans = soup.find_all("span", {"class": "create"})
return [
"https://www.explainxkcd.com" + span.parent.a["href"] for span in create_spans
]
async def main():
"""
Scrap XKCD dataset
"""
# Throttle to 10 requests per second
throttler = AsyncLimiter(max_rate=10, time_period=1)
async with aiohttp.ClientSession() as session:
# Get all comic urls
comic_urls = await scap_comic_urls(
session, LIST_COMICS_500_URL
) + await scap_comic_urls(session, LIST_COMICS_FULL_URL)
# Scrap all comics asynchronously
data = await asyncio.gather(
*[scrap_comic(session, url, throttler) for url in comic_urls]
)
df = (
pd.DataFrame.from_records(data)
.dropna(subset=["id"])
.astype({"id": "int32"})
.sort_values("id")
)
df.to_json(Path(__file__).parent / "dataset.jsonl", orient="records", lines=True)
if __name__ == "__main__":
asyncio.run(main())