microblog.pub/app/templates.py

226 lines
5.5 KiB
Python

import base64
from datetime import datetime
from datetime import timezone
from functools import lru_cache
from typing import Any
from urllib.parse import urlparse
import bleach
import timeago # type: ignore
from bs4 import BeautifulSoup # type: ignore
from fastapi import Request
from fastapi.templating import Jinja2Templates
from loguru import logger
from sqlalchemy.orm import Session
from starlette.templating import _TemplateResponse as TemplateResponse
from app import models
from app.actor import LOCAL_ACTOR
from app.ap_object import Attachment
from app.ap_object import Object
from app.config import BASE_URL
from app.config import DEBUG
from app.config import VERSION
from app.config import generate_csrf_token
from app.config import session_serializer
from app.database import now
from app.media import proxied_media_url
from app.utils.highlight import HIGHLIGHT_CSS
from app.utils.highlight import highlight
_templates = Jinja2Templates(directory="app/templates")
def _filter_domain(text: str) -> str:
hostname = urlparse(text).hostname
if not hostname:
raise ValueError(f"No hostname for {text}")
return hostname
def _media_proxy_url(url: str | None) -> str:
if not url:
return "/static/nopic.png"
if url.startswith(BASE_URL):
return url
encoded_url = base64.urlsafe_b64encode(url.encode()).decode()
return f"/proxy/media/{encoded_url}"
def is_current_user_admin(request: Request) -> bool:
is_admin = False
session_cookie = request.cookies.get("session")
if session_cookie:
try:
loaded_session = session_serializer.loads(
session_cookie,
max_age=3600 * 12,
)
except Exception:
pass
else:
is_admin = loaded_session.get("is_logged_in")
return is_admin
def render_template(
db: Session,
request: Request,
template: str,
template_args: dict[str, Any] = {},
) -> TemplateResponse:
is_admin = False
is_admin = is_current_user_admin(request)
return _templates.TemplateResponse(
template,
{
"request": request,
"debug": DEBUG,
"microblogpub_version": VERSION,
"is_admin": is_admin,
"csrf_token": generate_csrf_token() if is_admin else None,
"highlight_css": HIGHLIGHT_CSS,
"notifications_count": db.query(models.Notification)
.filter(models.Notification.is_new.is_(True))
.count()
if is_admin
else 0,
"local_actor": LOCAL_ACTOR,
"followers_count": db.query(models.Follower).count(),
"following_count": db.query(models.Following).count(),
**template_args,
},
)
# HTML/templates helper
ALLOWED_TAGS = [
"a",
"abbr",
"acronym",
"b",
"br",
"blockquote",
"code",
"pre",
"em",
"i",
"li",
"ol",
"strong",
"sup",
"sub",
"del",
"ul",
"span",
"div",
"p",
"h1",
"h2",
"h3",
"h4",
"h5",
"h6",
"table",
"th",
"tr",
"td",
"thead",
"tbody",
"tfoot",
"colgroup",
"caption",
"img",
]
ALLOWED_ATTRIBUTES = {
"a": ["href", "title"],
"abbr": ["title"],
"acronym": ["title"],
"img": ["src", "alt", "title"],
}
@lru_cache(maxsize=256)
def _update_inline_imgs(content):
soup = BeautifulSoup(content, "html5lib")
imgs = soup.find_all("img")
if not imgs:
return content
for img in imgs:
if not img.attrs.get("src"):
continue
img.attrs["src"] = _media_proxy_url(img.attrs["src"])
return soup.find("body").decode_contents()
def _clean_html(html: str, note: Object) -> str:
try:
return _replace_custom_emojis(
bleach.clean(
_update_inline_imgs(highlight(html)),
tags=ALLOWED_TAGS,
attributes=ALLOWED_ATTRIBUTES,
strip=True,
),
note,
)
except Exception:
raise
def _timeago(original_dt: datetime) -> str:
dt = original_dt
if dt.tzinfo:
dt = dt.astimezone(timezone.utc).replace(tzinfo=None)
return timeago.format(dt, now().replace(tzinfo=None))
def _has_media_type(attachment: Attachment, media_type_prefix: str) -> bool:
return attachment.media_type.startswith(media_type_prefix)
def _format_date(dt: datetime) -> str:
return dt.strftime("%b %d, %Y, %H:%M")
def _pluralize(count: int, singular: str = "", plural: str = "s") -> str:
if count > 1:
return plural
else:
return singular
def _replace_custom_emojis(content: str, note: Object) -> str:
idx = {}
for tag in note.tags:
if tag.get("type") == "Emoji":
try:
idx[tag["name"]] = proxied_media_url(tag["icon"]["url"])
except KeyError:
logger.warning(f"Failed to parse custom emoji {tag=}")
continue
for emoji_name, emoji_url in idx.items():
content = content.replace(
emoji_name,
f'<img class="custom-emoji" src="{emoji_url}" title="{emoji_name}" alt="{emoji_name}">', # noqa: E501
)
return content
_templates.env.filters["domain"] = _filter_domain
_templates.env.filters["media_proxy_url"] = _media_proxy_url
_templates.env.filters["clean_html"] = _clean_html
_templates.env.filters["timeago"] = _timeago
_templates.env.filters["format_date"] = _format_date
_templates.env.filters["has_media_type"] = _has_media_type
_templates.env.filters["pluralize"] = _pluralize