/`.
Scene detail page (g1/bimber WordPress theme) zawiera:
- `Name…`
— lista tagów (kategorie + performerzy wymieszani, tak jak porndish je pokazuje
jako „#" hashtagi). Bierzemy wszystkie jako RawTag (resolver dedupuje; performer
z query i tak dochodzi osobno).
- prozę opisu w `
` wewnątrz `.entry-content` (przed `entry-tags`, po embed-JS).
Bez `_fetch_scene_metadata` overrides scena z samego porndish miała 0 tagów i brak
description (bug-report od Jana 2026-06-06: „nie ma tagów (# na stronie) ani description").
"""
from __future__ import annotations
import html as html_mod
import logging
import re
from app.connectors.base import RawPerformer, RawStudio, RawTag
from app.connectors.direct_scrapers._search_base import BaseSearchScraper
from app.extractors import browser_get
log = logging.getLogger(__name__)
_ENTRY_TAG_RE = re.compile(
r']+href="[^"]*/video2/(?P[^"/]+)/"[^>]*class="[^"]*entry-tag[^"]*"[^>]*>'
r'(?P[^<]+)',
re.IGNORECASE,
)
_ENTRY_CONTENT_RE = re.compile(
r']*class="[^"]*entry-content[^"]*"[^>]*>(?P.*?)',
re.IGNORECASE | re.DOTALL,
)
_SCRIPT_STYLE_RE = re.compile(r"|", re.IGNORECASE | re.DOTALL)
_P_RE = re.compile(r"]*>(?P.*?)
", re.IGNORECASE | re.DOTALL)
_TAG_STRIP_RE = re.compile(r"<[^>]+>")
_WS_RE = re.compile(r"\s+")
_SLUG_RE = re.compile(r"[^a-z0-9]+")
def _slugify(name: str) -> str:
return _SLUG_RE.sub("-", name.lower()).strip("-") or "tag"
def _clean_text(fragment: str) -> str:
txt = _TAG_STRIP_RE.sub(" ", fragment)
txt = html_mod.unescape(txt)
return _WS_RE.sub(" ", txt).strip()
class PornDishScraper(BaseSearchScraper):
sitetag = "porndishcom"
_search_url_template = "https://porndish.com/page/{page}/?s={query}"
_scene_url_re = re.compile(
r'href="(?Phttps://porndish\.com/(?P[a-z0-9][a-z0-9\-]+))/"',
re.IGNORECASE,
)
def _fetch_scene_metadata(
self, scene_url: str
) -> tuple[RawStudio | None, list[RawPerformer], list[RawTag], str | None] | None:
"""Fetch scene page → (studio=None, performers=[], tags, description).
4-elementowy zwrot (base unpacka opcjonalny `description`). porndish nie
wyróżnia studia, a performer z query dochodzi w base — tu tylko tagi + opis.
"""
try:
r = browser_get(scene_url, timeout=self._timeout)
except Exception as e:
log.debug("porndish meta fetch failed for %s: %s", scene_url, e)
return None
if r.status_code != 200 or not r.text:
return None
html = r.text
# Tagi: entry-tag anchors (slug z /video2// + display name).
tags: list[RawTag] = []
seen: set[str] = set()
for m in _ENTRY_TAG_RE.finditer(html):
name = html_mod.unescape(m.group("name")).strip()
slug = (m.group("slug") or "").strip().lower() or _slugify(name)
if not name or len(name) > 40 or slug in seen:
continue
seen.add(slug)
tags.append(RawTag(external_id=f"porndishcom:tag:{slug}", name=name, slug=slug))
# Description: najdłuższy prozowy w .entry-content (bez entry-tags / embed-JS).
description: str | None = None
mc = _ENTRY_CONTENT_RE.search(html)
body = mc.group("body") if mc else html
body = _SCRIPT_STYLE_RE.sub(" ", body)
best = ""
for pm in _P_RE.finditer(body):
inner = pm.group("inner")
if "entry-tag" in inner:
continue
txt = _clean_text(inner)
# Pomijamy resztki JS / boilerplate „Watch … porn video" / przyciski serwerów.
if not txt or "getElementById" in txt or "addEventListener" in txt:
continue
low = txt.lower()
if low.startswith("watch ") and low.endswith("porn video"):
continue
if len(txt) > len(best):
best = txt
# Strip wiodące etykiety przycisków embedu („Video Player 1 Video Player 2 …",
# czasem „Server N") które wpadają na początek prozy.
best = re.sub(r"^(?:Video Player \d+\s*|Server \d+\s*|Download\s*)+", "", best, flags=re.IGNORECASE).strip()
if len(best) >= 40:
description = best
if not tags and description is None:
return None
return (None, [], tags, description)