YouTube channels (1/4): channel resolver in core/video/youtube.py

First slice of 'follow a YouTube channel as a show'. Pure, testable resolver:
- parse_channel_url() normalizes any channel reference (@handle, /channel/UC..,
  /c/.., /user/.., bare handle, with/without scheme or tab suffix) to a
  canonical /videos uploads URL, and rejects non-channels (watch/playlist/
  shorts/home/non-youtube).
- shape_channel() maps yt-dlp's flat-extract dict to our shape (channel meta +
  recent uploads), picking best-res thumbs, parsing sparse publish dates
  (timestamp or upload_date → ISO, None when flat mode omits them), filtering
  null/id-less entries.
- resolve_channel() ties them together via an injectable YoutubeDL factory.

yt-dlp only (already a pinned dep), extract_flat for cheap listing. 28 seam
tests, no network. Schema bridge + API + UI are the next slices.
This commit is contained in:
BoulderBadgeDad 2026-06-17 00:25:34 -07:00
parent b7cafb8b72
commit 68383a16b3
2 changed files with 380 additions and 0 deletions

192
core/video/youtube.py Normal file
View file

@ -0,0 +1,192 @@
"""Resolve a YouTube *channel* reference into a followable channel + its recent
uploads the data source behind "follow a YouTube channel" on the video side.
yt-dlp only (no API key). The upload list is fetched with ``extract_flat`` so it
stays cheap even for channels with thousands of videos; the trade-off is that
per-video publish dates are sparse in flat mode, so they get backfilled later
(same pattern as the wishlist art backfill). Real *downloads* will additionally
need Deno + ffmpeg; flat *listing* does not.
The URL parsing and the yt-dlp-dict our-shape mapping are pure functions so
they're unit-testable without touching the network; the one network call goes
through ``_extract`` which accepts an injectable factory for tests.
"""
from __future__ import annotations
import logging
import re
from datetime import datetime, timezone
from urllib.parse import urlparse
try:
import yt_dlp
except Exception: # pragma: no cover - yt-dlp is a hard dep, but stay import-safe
yt_dlp = None
logger = logging.getLogger(__name__)
# Path prefixes that identify a channel (everything after is the channel key).
# Tabs like /videos, /streams, /shorts, /featured, /playlists are stripped.
_CHANNEL_TABS = {"videos", "streams", "shorts", "featured", "playlists", "community", "about"}
_HANDLE_RE = re.compile(r"^@?[A-Za-z0-9._-]{2,}$")
_UA = ("Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 "
"(KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36")
def parse_channel_url(raw):
"""Normalize a pasted channel reference to a canonical uploads URL, or return
None if it isn't a channel (a /watch video, a playlist, the YT home, etc.).
Accepts: ``https://www.youtube.com/@PlayStation`` (+ /videos etc.),
``/channel/UC...``, ``/c/Name``, ``/user/Name``, a bare ``@handle``, or a
bare ``handle``. Returns ``https://www.youtube.com/<base>/videos``.
"""
if not raw or not str(raw).strip():
return None
raw = str(raw).strip()
# Bare handle (no scheme, no slash) → treat as @handle.
if "/" not in raw and "." not in raw and " " not in raw:
if _HANDLE_RE.match(raw):
handle = raw if raw.startswith("@") else "@" + raw
return "https://www.youtube.com/" + handle + "/videos"
return None
# Give bare "youtube.com/..." a scheme so urlparse populates netloc.
parse_target = raw if "://" in raw else "https://" + raw
try:
u = urlparse(parse_target)
except Exception:
return None
host = (u.netloc or "").lower()
if host.startswith("www."):
host = host[4:]
if host not in ("youtube.com", "m.youtube.com", "music.youtube.com"):
return None
segs = [s for s in (u.path or "").split("/") if s]
if not segs:
return None
first = segs[0]
if first.startswith("@"):
base = first # /@handle
elif first in ("channel", "c", "user") and len(segs) >= 2:
base = first + "/" + segs[1] # /channel/UC.., /c/Name, /user/Name
else:
return None # /watch, /playlist, /shorts/<id>, home…
return "https://www.youtube.com/" + base + "/videos"
def _best_thumb(thumbs):
"""Pick the highest-resolution thumbnail URL from a yt-dlp thumbnails list."""
if not thumbs:
return None
best, best_area = None, -1
for t in thumbs:
if not isinstance(t, dict) or not t.get("url"):
continue
area = (t.get("width") or 0) * (t.get("height") or 0)
# preference value as a tie-breaker when dimensions are absent
score = area if area else (t.get("preference") or 0)
if score >= best_area:
best, best_area = t.get("url"), score
return best
def _entry_date(e):
"""ISO date (YYYY-MM-DD) for a flat entry, or None — flat mode often omits it."""
ts = e.get("timestamp")
if isinstance(ts, (int, float)):
try:
return datetime.fromtimestamp(ts, tz=timezone.utc).date().isoformat()
except Exception:
pass
ud = e.get("upload_date") # 'YYYYMMDD'
if isinstance(ud, str) and len(ud) == 8 and ud.isdigit():
return f"{ud[0:4]}-{ud[4:6]}-{ud[6:8]}"
return None
def shape_channel(info, limit=30):
"""Map a yt-dlp channel info dict to our followable-channel shape."""
info = info or {}
entries = [e for e in (info.get("entries") or []) if isinstance(e, dict)]
videos = []
for e in entries[:limit]:
vid = e.get("id")
if not vid:
continue
dur = e.get("duration")
videos.append({
"youtube_id": vid,
"title": e.get("title") or "",
"published_at": _entry_date(e),
"duration_seconds": int(dur) if isinstance(dur, (int, float)) else None,
"thumbnail_url": _best_thumb(e.get("thumbnails")) or e.get("thumbnail"),
"view_count": e.get("view_count"),
"description": e.get("description"),
})
handle = info.get("uploader_id") or info.get("channel_id_handle")
if handle and not str(handle).startswith("@"):
handle = None # uploader_id is sometimes the UC id, not a handle
return {
"youtube_id": info.get("channel_id") or info.get("id"),
"title": info.get("channel") or info.get("uploader") or info.get("title") or "",
"handle": handle,
"avatar_url": _best_thumb(info.get("thumbnails")),
"description": info.get("description"),
"subscriber_count": info.get("channel_follower_count"),
"video_count": info.get("playlist_count") or len(videos),
"videos": videos,
}
def _ydl_opts(limit, db=None):
opts = {
"quiet": True,
"no_warnings": True,
"extract_flat": True, # fast: enumerate uploads without per-video format probing
"skip_download": True,
"playlistend": int(limit),
"user_agent": _UA,
}
# Reuse the music client's cookie convention so age/region-gated channels work.
try:
from config.settings import config_manager
cb = config_manager.get("youtube.cookies_browser", "")
if cb:
opts["cookiesfrombrowser"] = (cb,)
except Exception:
pass
return opts
def _extract(url, opts, ydl_factory=None):
"""Run yt-dlp's extract_info; isolated for test injection. Returns a dict or None."""
factory = ydl_factory or (yt_dlp.YoutubeDL if yt_dlp else None)
if factory is None:
logger.warning("yt-dlp unavailable; cannot resolve YouTube channel")
return None
try:
with factory(opts) as ydl:
return ydl.extract_info(url, download=False)
except Exception as e:
logger.info("YouTube channel resolve failed for %s: %s", url, e)
return None
def resolve_channel(raw, limit=30, ydl_factory=None, db=None):
"""Resolve a pasted channel reference to ``{youtube_id, title, handle,
avatar_url, videos:[...], ...}``, or None if it isn't a resolvable channel."""
url = parse_channel_url(raw)
if not url:
return None
info = _extract(url, _ydl_opts(limit, db), ydl_factory)
if not info:
return None
shaped = shape_channel(info, limit)
return shaped if shaped.get("youtube_id") else None

188
tests/test_video_youtube.py Normal file
View file

@ -0,0 +1,188 @@
"""Seam tests for core/video/youtube.py — the YouTube channel resolver.
Pure URL parsing + yt-dlp-dict our-shape mapping are tested directly; the one
network call is exercised through an injected fake YoutubeDL, so nothing here
touches the network.
"""
import pytest
from core.video import youtube as yt
# ── parse_channel_url ────────────────────────────────────────────────────────
@pytest.mark.parametrize("raw,expected", [
("https://www.youtube.com/@PlayStation", "https://www.youtube.com/@PlayStation/videos"),
("https://www.youtube.com/@PlayStation/videos", "https://www.youtube.com/@PlayStation/videos"),
("https://www.youtube.com/@PlayStation/streams", "https://www.youtube.com/@PlayStation/videos"),
("http://youtube.com/@GoodMythicalMorning", "https://www.youtube.com/@GoodMythicalMorning/videos"),
("youtube.com/@PlayStation", "https://www.youtube.com/@PlayStation/videos"),
("m.youtube.com/@PlayStation", "https://www.youtube.com/@PlayStation/videos"),
("https://www.youtube.com/channel/UCabc123", "https://www.youtube.com/channel/UCabc123/videos"),
("https://www.youtube.com/c/LinusTechTips", "https://www.youtube.com/c/LinusTechTips/videos"),
("https://www.youtube.com/user/PewDiePie", "https://www.youtube.com/user/PewDiePie/videos"),
("@PlayStation", "https://www.youtube.com/@PlayStation/videos"),
("PlayStation", "https://www.youtube.com/@PlayStation/videos"), # bare → @handle
])
def test_parse_channel_url_accepts_channel_forms(raw, expected):
assert yt.parse_channel_url(raw) == expected
@pytest.mark.parametrize("raw", [
"",
" ",
None,
"https://www.youtube.com/watch?v=dQw4w9WgXcQ", # a video, not a channel
"https://www.youtube.com/playlist?list=PL123", # a playlist
"https://www.youtube.com/", # home
"https://youtu.be/dQw4w9WgXcQ", # short video link
"https://vimeo.com/@someone", # not youtube
"https://www.youtube.com/shorts/abc123", # a short, not a channel
])
def test_parse_channel_url_rejects_non_channels(raw):
assert yt.parse_channel_url(raw) is None
# ── shape_channel ────────────────────────────────────────────────────────────
def _flat_info():
return {
"channel_id": "UCPlayStation",
"channel": "PlayStation",
"uploader": "PlayStation",
"uploader_id": "@PlayStation",
"channel_follower_count": 14_000_000,
"playlist_count": 1200,
"thumbnails": [
{"url": "http://img/small.jpg", "width": 88, "height": 88},
{"url": "http://img/big.jpg", "width": 800, "height": 800},
],
"entries": [
{"id": "vid1", "title": "State of Play", "timestamp": 1_700_000_000,
"duration": 3600, "view_count": 50000,
"thumbnails": [{"url": "http://t/1.jpg", "width": 320, "height": 180}]},
{"id": "vid2", "title": "Trailer", "upload_date": "20240115", "duration": 120,
"thumbnail": "http://t/2.jpg"},
{"id": "vid3", "title": "No date video"}, # sparse flat entry
None, # yt-dlp can yield Nones
{"title": "missing id — skip"}, # no id → skipped
],
}
def test_shape_channel_maps_channel_fields():
out = yt.shape_channel(_flat_info())
assert out["youtube_id"] == "UCPlayStation"
assert out["title"] == "PlayStation"
assert out["handle"] == "@PlayStation"
assert out["avatar_url"] == "http://img/big.jpg" # highest-res thumb wins
assert out["subscriber_count"] == 14_000_000
assert out["video_count"] == 1200 # playlist_count, not len(videos)
def test_shape_channel_maps_and_filters_videos():
out = yt.shape_channel(_flat_info())
vids = out["videos"]
# the None and the id-less entry are dropped
assert [v["youtube_id"] for v in vids] == ["vid1", "vid2", "vid3"]
# timestamp → ISO date
assert vids[0]["published_at"] == "2023-11-14"
assert vids[0]["duration_seconds"] == 3600
assert vids[0]["thumbnail_url"] == "http://t/1.jpg"
# upload_date 'YYYYMMDD' → ISO date; plain 'thumbnail' string honored
assert vids[1]["published_at"] == "2024-01-15"
assert vids[1]["thumbnail_url"] == "http://t/2.jpg"
# sparse entry: no date, no crash
assert vids[2]["published_at"] is None
assert vids[2]["duration_seconds"] is None
def test_shape_channel_respects_limit():
out = yt.shape_channel(_flat_info(), limit=1)
assert len(out["videos"]) == 1
assert out["videos"][0]["youtube_id"] == "vid1"
def test_shape_channel_uploader_id_not_a_handle_is_dropped():
info = {"channel_id": "UCx", "channel": "X", "uploader_id": "UCx", "entries": []}
assert yt.shape_channel(info)["handle"] is None
# ── resolve_channel (network call injected) ──────────────────────────────────
class _FakeYDL:
"""Stand-in for yt_dlp.YoutubeDL: records opts, returns a canned info dict."""
last_opts = None
last_url = None
def __init__(self, opts):
_FakeYDL.last_opts = opts
self.opts = opts
def __enter__(self):
return self
def __exit__(self, *a):
return False
def extract_info(self, url, download=False):
_FakeYDL.last_url = url
assert download is False
return _flat_info()
def test_resolve_channel_happy_path_uses_canonical_url_and_limit():
out = yt.resolve_channel("https://www.youtube.com/@PlayStation", limit=2,
ydl_factory=_FakeYDL)
assert out["youtube_id"] == "UCPlayStation"
assert len(out["videos"]) == 2
# resolver normalizes to the /videos uploads URL and passes the limit through
assert _FakeYDL.last_url == "https://www.youtube.com/@PlayStation/videos"
assert _FakeYDL.last_opts["playlistend"] == 2
assert _FakeYDL.last_opts["extract_flat"] is True
def test_resolve_channel_rejects_non_channel_without_network():
called = []
def factory(opts):
called.append(opts)
raise AssertionError("should not be called for a non-channel URL")
assert yt.resolve_channel("https://www.youtube.com/watch?v=abc", ydl_factory=factory) is None
assert called == []
def test_resolve_channel_returns_none_on_extractor_error():
class _Boom:
def __init__(self, opts):
pass
def __enter__(self):
return self
def __exit__(self, *a):
return False
def extract_info(self, url, download=False):
raise RuntimeError("DownloadError: channel not found")
assert yt.resolve_channel("@nope", ydl_factory=_Boom) is None
def test_resolve_channel_none_when_info_has_no_channel_id():
class _NoId:
def __init__(self, opts):
pass
def __enter__(self):
return self
def __exit__(self, *a):
return False
def extract_info(self, url, download=False):
return {"entries": [{"id": "v", "title": "t"}]} # no channel_id/id
assert yt.resolve_channel("@x", ydl_factory=_NoId) is None