ani-web/server/providers/hianime.py

"""providers/hianime.py — hianime.at source (pystardust/ani-cli#1897 flow).

Flow: search page -> ajax episode list (numeric anime id) -> ajax server list
(sub/dub groups) -> HD-1 hash (base64 embed url) -> embed page
(window.__P = base64(json XOR "otaku-embed-v1")) -> master m3u8.

The CDN hard-403s without the embed origin as Referer, so _master_m3u8
records it in REFERER and registers the playlist hosts in STREAM_HOSTS;
the /stream proxy reads both via stream_meta().
"""
import base64
import json
import re
from urllib.parse import urljoin, urlparse

from ..upstream import _cached, _get

BASE = "https://hianime.at"
KEY = b"otaku-embed-v1"

REFERER = None          # embed origin, set when a master playlist is resolved
STREAM_HOSTS = set()    # playlist/segment hosts the /stream proxy may fetch


def _search(q):
    html = _get(f"{BASE}/search", params={"keyword": q}).text
    # the top-10 sidebar repeats the result markup; cut it off (ani-cli#1897)
    html = html.split('id="main-sidebar"')[0]
    out, seen = [], set()
    for m in re.finditer(
            r'<h3 class="film-name">\s*<a href="[^"]*/([^"/]+)"[^>]*title="([^"]*)"',
            html):
        slug, name = m.group(1), m.group(2)
        if slug in seen:
            continue
        seen.add(slug)
        img = re.search(r'<img[^>]+src="([^"]+)"', html[m.start():m.start() + 3000])
        out.append({"id": int(slug.rsplit("-", 1)[1]) if slug.rsplit("-", 1)[1].isdigit() else 0,
                    "slug_id": slug, "name": name,
                    "poster": img.group(1) if img else None})
    return out


def _details(slug):
    t = _get(f"{BASE}/watch/{slug}").text
    # no details/synopsis page: og tags carry name + poster
    title = re.search(r'property="og:title" content="([^"]+)"', t)
    post = re.search(r'property="og:image" content="([^"]+)"', t)
    name = re.sub(r'\s*Episode\s*\d+.*$', "", title.group(1)) if title else None
    name = re.sub(r'\s*\|.*$', "", name).strip() if name else None
    return {"id": int(slug.rsplit("-", 1)[1]) if slug.rsplit("-", 1)[1].isdigit() else 0,
            "name": name, "synopsis": None, "mal_id": None,
            "poster": post.group(1) if post else None,
            "seasons": []}  # hianime has no seasons block (ani-cli#1897 re-searches)


def _episodes(anime_id):
    t = _get(f"{BASE}/api/theme/episode/list/{anime_id}").text.replace("\\", "")
    out = []
    for num, epid in re.findall(r'data-number="([^"]*)"[\s\S]*?data-id="(\d+)"', t):
        out.append({"id": int(epid), "number": int(num) if num.isdigit() else num,
                    "filler": False})
    return out


def _servers(ep_id):
    return _get(f"{BASE}/api/theme/episode/servers",
                params={"episodeId": ep_id}).text.replace("\\", "")


def _languages(ep_id):
    t = _servers(ep_id)
    out = []
    if re.search(r'data-type="sub"[\s\S]*?data-server-name="HD-1"', t):
        out.append({"code": "jpn", "name": "Sub", "embed_url": "sub"})
    if re.search(r'data-type="dub"[\s\S]*?data-server-name="HD-1"', t):
        out.append({"code": "eng", "name": "Dub", "embed_url": "dub"})
    return out


def _deobfuscate(blob):
    raw = base64.b64decode(blob)
    return json.loads(bytes(b ^ KEY[i % len(KEY)] for i, b in enumerate(raw)))


def _master_m3u8(ep_id, lang="eng"):
    global REFERER, STREAM_HOSTS
    t = _servers(ep_id)
    want = "sub" if lang in ("jpn", "sub") else "dub"
    for typ in (want, "sub") if want == "dub" else (want,):
        m = re.search(rf'data-type="{typ}"[\s\S]*?data-server-name="HD-1"'
                      r'[\s\S]*?data-hash="([^"]+)"', t)
        if not m:
            continue
        embed = base64.b64decode(m.group(1)).decode()
        REFERER = re.match(r'https?://[^/]+', embed).group(0) + "/"
        j = _deobfuscate(re.search(r'window\.__P="([^"]+)"',
                                   _get(embed, timeout=15).text).group(1))
        master = j.get("src")
        if not master:
            continue
        # prove the CDN serves us (it 403s without the referer) and register
        # the hosts so the /stream proxy allowlist survives CDN rotation
        r = _get(master, headers={"Referer": REFERER}, timeout=10)
        hosts = {urlparse(master).netloc}
        for line in r.text.splitlines():
            if line and not line.startswith("#"):
                hosts.add(urlparse(urljoin(master, line)).netloc)
        STREAM_HOSTS |= hosts
        return master
    return None


# Provider interface. Cache keys are namespaced with "hianime:".
search = lambda q: _cached(f"hianime:search:{q.lower()}", 300, _search, q)
details = lambda s: _cached(f"hianime:details:{s}", 3600, _details, s)
episodes = lambda i: _cached(f"hianime:episodes:{i}", 300, _episodes, i)  # 5m: airing shows gain eps
languages = lambda e: _cached(f"hianime:languages:{e}", 3600, _languages, e)
master_m3u8 = lambda e, l="eng": _cached(f"hianime:master:{e}:{l}", 3600, _master_m3u8, e, l)


def stream_meta(url):
    """{Referer} for our own playlist/segment hosts, else None."""
    if REFERER and urlparse(url).netloc in STREAM_HOSTS:
        return {"Referer": REFERER}
    return None