mirror of
https://git.sr.ht/~tsileo/microblog.pub
synced 2024-12-22 13:14:28 +00:00
98 lines
2.5 KiB
Python
98 lines
2.5 KiB
Python
import mimetypes
|
|
import re
|
|
from typing import Any
|
|
from urllib.parse import urlparse
|
|
|
|
import httpx
|
|
from bs4 import BeautifulSoup # type: ignore
|
|
from pydantic import BaseModel
|
|
|
|
from app import activitypub as ap
|
|
from app import config
|
|
from app.utils.url import is_url_valid
|
|
|
|
|
|
class OpenGraphMeta(BaseModel):
|
|
url: str
|
|
title: str
|
|
image: str
|
|
description: str
|
|
site_name: str
|
|
|
|
|
|
def _scrap_og_meta(html: str) -> OpenGraphMeta | None:
|
|
soup = BeautifulSoup(html, "html5lib")
|
|
ogs = {
|
|
og.attrs["property"]: og.attrs.get("content")
|
|
for og in soup.html.head.findAll(property=re.compile(r"^og"))
|
|
}
|
|
raw = {}
|
|
for field in OpenGraphMeta.__fields__.keys():
|
|
og_field = f"og:{field}"
|
|
if not ogs.get(og_field):
|
|
return None
|
|
|
|
raw[field] = ogs[og_field]
|
|
|
|
return OpenGraphMeta.parse_obj(raw)
|
|
|
|
|
|
def _urls_from_note(note: ap.RawObject) -> set[str]:
|
|
note_host = urlparse(ap.get_id(note["id"]) or "").netloc
|
|
|
|
tags_hrefs = set()
|
|
for tag in note.get("tag", []):
|
|
if tag_href := tag.get("href"):
|
|
tags_hrefs.add(tag_href)
|
|
|
|
urls = set()
|
|
if "content" in note:
|
|
soup = BeautifulSoup(note["content"], "html5lib")
|
|
for link in soup.find_all("a"):
|
|
h = link.get("href")
|
|
ph = urlparse(h)
|
|
mimetype, _ = mimetypes.guess_type(h)
|
|
if (
|
|
ph.scheme in {"http", "https"}
|
|
and ph.netloc != note_host
|
|
and is_url_valid(h)
|
|
and (
|
|
not mimetype
|
|
or mimetype.split("/")[0] in ["image", "video", "audio"]
|
|
)
|
|
):
|
|
urls.add(h)
|
|
|
|
return urls - tags_hrefs
|
|
|
|
|
|
async def _og_meta_from_url(url: str) -> OpenGraphMeta | None:
|
|
async with httpx.AsyncClient() as client:
|
|
resp = await client.get(
|
|
url,
|
|
headers={
|
|
"User-Agent": config.USER_AGENT,
|
|
},
|
|
follow_redirects=True,
|
|
)
|
|
|
|
resp.raise_for_status()
|
|
|
|
if not (ct := resp.headers.get("content-type")) or not ct.startswith("text/html"):
|
|
return None
|
|
|
|
return _scrap_og_meta(resp.text)
|
|
|
|
|
|
async def og_meta_from_note(note: ap.RawObject) -> list[dict[str, Any]]:
|
|
og_meta = []
|
|
urls = _urls_from_note(note)
|
|
for url in urls:
|
|
try:
|
|
maybe_og_meta = await _og_meta_from_url(url)
|
|
if maybe_og_meta:
|
|
og_meta.append(maybe_og_meta.dict())
|
|
except httpx.HTTPError:
|
|
pass
|
|
|
|
return og_meta
|