Spike results — yt-dlp¶

S11 — flat playlist fields under extract_flat (closed 2026-09-20)¶

yt-dlp version: resolved by uv run --with yt-dlp on 2026-09-20. Playlist used: https://www.youtube.com/playlist?list=PLFgquLnL59alCl_2TQvOiD5Vgm1hCaGSI ("Popular Music Videos", 183 items) — a large, stable, public playlist.

Command¶

uv run --with yt-dlp python -c "
import yt_dlp
url = 'https://www.youtube.com/playlist?list=PLFgquLnL59alCl_2TQvOiD5Vgm1hCaGSI'
with yt_dlp.YoutubeDL({'extract_flat':'in_playlist','quiet':True,'skip_download':True}) as ydl:
    info = ydl.extract_info(url, download=False)
print(sorted(info['entries'][0]))
"

Result — playlist level¶

TOP-LEVEL KEYS: ['_type', 'availability', 'channel', 'channel_follower_count', 'channel_id',
 'channel_url', 'description', 'epoch', 'extractor', 'extractor_key', 'id', 'modified_date',
 'original_url', 'playlist_count', 'release_year', 'tags', 'thumbnails', 'title', 'uploader',
 'uploader_id', 'uploader_url', 'view_count', 'webpage_url', 'webpage_url_basename',
 'webpage_url_domain']

playlist title      : Popular Music Videos
playlist id         : PLFgquLnL59alCl_2TQvOiD5Vgm1hCaGSI
playlist_count      : 183
channel/uploader    : Music / Music
channel_id          : UC-9-kyTW8ZkZNDHQJ6FgpwQ
n entries           : 183

playlist_count matches len(entries), and the playlist carries its own title, description, channel_id and thumbnails — enough to fill PlaylistMeta and the owning ChannelMeta from a single call.

Result — entry level¶

ENTRY[0] KEYS: ['__x_forwarded_for_ip', '_type', 'availability', 'channel', 'channel_id',
 'channel_url', 'creators', 'duration', 'id', 'ie_key', 'live_status', 'thumbnails',
 'timestamp', 'title', 'uploader', 'uploader_url', 'url', 'view_count']

playlist_index present in any entry: False
timestamp values (first 5)         : [None, None, None, None, None]
description present in any entry   : False
duration None count                : 0 / 183
channel_id None count              : 0 / 183
keys missing from some entries     : ['uploader_id']

dict.get() returns None for a missing key and for a key whose value is None, so the two were measured apart explicitly — the distinction decides whether P2.2 can rely on a .get() default or must branch:

$ uv run --with yt-dlp python -c "... for k in (...): present = sum(1 for e in es if k in e) ..."
description      ABSENT from all entries
timestamp        present on all, value always None
availability     present on all, value always None
playlist_index   ABSENT from all entries
duration         present on all, 183 non-null
uploader_id      present on 27/183, 27 non-null
view_count       present on all, 183 non-null
live_status      present on all, value always None

Result — full extract of one video, for comparison¶

$ uv run --with yt-dlp python -c "...YoutubeDL({'quiet':True,'skip_download':True})..."
{
 "id": "fOT0BUpITw8",
 "title": "BELLAKEO (Video Oficial) - Peso Pluma, Anitta",
 "duration": 234,
 "channel_id": "UCzrM_068Odho89mTRrrxqbA",
 "upload_date": "20231208",
 "timestamp": 1701993610,
 "description": "BELLAKEO (Video Oficial) - Peso Pluma, Anitta\nDouble P Records\n\n...",
 "view_count": 761508498,
 "live_status": "not_live",
 "availability": "public",
 "thumbnail": "https://i.ytimg.com/vi/fOT0BUpITw8/maxresdefault.jpg"
}

A full extract also emits a warning worth knowing about, though it does not affect metadata:

WARNING: [youtube] No supported JavaScript runtime could be found. Only deno is enabled by
default ... YouTube extraction without a JS runtime has been deprecated, and some formats may
be missing.

Formats are irrelevant here — thumbforge never downloads media — but the warning must be suppressed or allow-listed so it does not pollute fetch output.

Conclusions¶

1. playlist_index does not exist under extract_flat. The spike as written in OPEN_QUESTIONS.md assumed it would. PlaylistItemMeta.position must therefore be derived from enumeration order — enumerate(info["entries"], start=1) — and the P2.1 docstring claiming it "mirrors yt-dlp's 1-based playlist_index" is corrected in this PR. Enumeration order is the playlist order, so the derived value is the intended one; it is the provenance that was wrong, not the semantics.
2. A flat playlist extract cannot fill description or published_at. description is not merely None — the key is absent from every entry, while timestamp is present and always None. Either way VideoMeta keeps its "" / None defaults for playlist-sourced rows; a full per-video extract is required to populate them. P2.2 must not fan out 183 full extracts during fetch — the flat pass stays the default and --refresh on a single video does the deep one. PLAN.md §5.2's video show is therefore allowed to display an empty description for a video that was only ever seen through a playlist.
3. duration disagrees by ±1s between flat and full extracts (235 flat vs 234 full for the same video). A refresh that switches extraction mode will flip the stored value; that is upstream rounding, not a bug, and the upsert must not treat it as a meaningful change.
4. channel_id on an entry is the video's own channel, not the playlist owner's. The playlist's channel_id is on the top-level dict. Both are needed: playlist.channel_id for ownership, video.channel_id for the per-video row.
5. thumbnails in a flat extract top out at 336×188. A 1280×720 source thumbnail needs either the full extract's thumbnail (maxresdefault.jpg) or the conventional i.ytimg.com/vi/<id>/maxresdefault.jpg URL. Relevant to P5.1, which fits a source thumbnail to 1280×720.

Fixture recorder consequence (P2.2)¶

The recorder must store the whole info dict for a playlist, including the top-level keys — not just entries — because PlaylistMeta and the owning ChannelMeta are built from them. Fixtures must cover both shapes: one flat playlist extract and one full single-video extract.