feat: support YouTube VTT subtitle selection

This commit is contained in:
2026-10-27 13:05:56 +01:00
parent a2c8cc8e4d
commit 0641859d95
12 changed files with 449 additions and 25 deletions
+107
View File
@@ -167,6 +167,112 @@ struct YouTubeResolverChecks {
try require(try decode(emptyHeaders).video.headers.isEmpty, "absent headers serialize to an empty block")
}
static func subtitles() throws {
func vtt(_ url: String, name: Any = "English", headers: Any? = nil) -> [String: Any] {
var row: [String: Any] = ["ext": "vtt", "url": url, "name": name]
if let headers { row["http_headers"] = headers }
return row
}
let original = "https://www.youtube.com/api/timedtext?lang=en&kind=asr&fmt=vtt"
var object = try fixture(combinedJSON)
try require(try decode(object).subtitles.isEmpty, "absent captions default to Off with no optional tracks")
for payload in [NSNull(), [], true, 17, "captions"] as [Any] {
object["subtitles"] = payload
object["automatic_captions"] = payload
try require(try decode(object).subtitles.isEmpty,
"malformed optional caption containers never block playback")
}
object["subtitles"] = [
"broken": true,
"en": [NSNull(), 7, ["ext": false], ["ext": "vtt", "url": 9],
["ext": "srt", "url": "https://subs.example/convert-me"],
vtt("file:///unsafe.vtt"),
vtt("https://subs.example/unsafe.vtt", headers: ["X-Unsafe": "bad\nvalue"]),
vtt("https://subs.example/bad-headers.vtt", headers: ["X-Bad": 3]),
vtt("https://subs.example/en.vtt?token=first", headers: ["X-Track": "manual"]),
vtt("https://subs.example/duplicate.vtt")],
"fr": [vtt("https://subs.example/fr.vtt", name: ["malformed"])],
"srt-only": [["ext": "srt", "url": "https://subs.example/only.srt"]]
] as [String: Any]
object["automatic_captions"] = [
"en-orig": [vtt(original, headers: ["X-Track": "auto"])],
"en": [vtt(original)],
"fr": [vtt(original + "&tlang=fr", name: "French")],
"de": [vtt("https://subs.example/de.vtt", name: "German (Original)")],
"bad-orig": [NSNull(), vtt("ftp://subs.example/bad.vtt")]
] as [String: Any]
let tracks = try decode(object).subtitles
try require(tracks.count == 3 && tracks.map(\.language) == ["en", "fr", "en"],
"manual VTT and one original auto survive malformed siblings; translations and aliases do not")
try require(tracks.map(\.selectionID) == ["youtube:manual:en", "youtube:manual:fr", "youtube:auto:en"],
"manual and auto identities depend on language metadata, not signed URLs or labels")
try require(tracks[0].url == "https://subs.example/en.vtt?token=first"
&& tracks[0].label == "English" && tracks[1].label == "fr"
&& tracks[2].label == "English (auto-generated)",
"one valid supplied VTT per language preserves name and query; malformed name falls back to language")
try require(tracks.allSatisfy { !$0.isDefault } && ResolvedSource.validSubtitles(tracks),
"optional captions preserve transport safeguards and never enable themselves")
try require(tracks[0].headers == "X-Track: manual\r\n"
&& tracks[1].headers == "User-Agent: combined-agent\r\n"
&& tracks[2].headers == "X-Track: auto\r\n",
"variant headers override the complete root map; absent variant headers inherit root")
var changed = object
changed["automatic_captions"] = ["en": [vtt(original + "&token=renewed", name: "Renamed")]]
let alias = try decode(changed).subtitles.last!
try require(alias.selectionID == tracks.last?.selectionID && alias.url != tracks.last?.url,
"explicit untranslated ASR alias retains original identity after signed URL and label refresh")
changed["automatic_captions"] = [
"en-orig": [vtt(original)],
"en": [vtt(original + "&token=alias")],
"other-orig": [vtt(original)]
]
try require(try decode(changed).subtitles.filter { $0.selectionID?.hasPrefix("youtube:auto:") == true }.count == 1,
"original auto aliases deduplicate by language and by URL")
changed["automatic_captions"] = ["en-orig": [vtt(original)]]
changed["subtitles"] = ["en": [vtt("https://subs.example/empty.vtt", headers: [:])]]
try require(try decode(changed).subtitles.first?.headers == "",
"an explicitly empty caption header map does not inherit root headers")
for unsafe in ["https://user:pass@subs.example/c.vtt", "https://subs.example/c.vtt#fragment",
"https://subs.example:0/c.vtt", "https://subs.example/c vtt"] {
changed["subtitles"] = ["en": [vtt(unsafe)]]
try require(try decode(changed).subtitles.count == 1,
"unsafe optional subtitle URL is skipped without discarding a good auto sibling")
}
var fragmented = vtt("https://subs.example/chunks")
fragmented["protocol"] = "m3u8_native"
changed["subtitles"] = ["en": [fragmented]]
try require(try decode(changed).subtitles.count == 1,
"caption playlists are not treated as immutable VTT documents")
var many: [String: Any] = [:]
for index in 0..<40 {
let key = String(format: "lang-%02d", index)
many[key] = [vtt("https://subs.example/\(key).vtt", name: key)]
}
changed["subtitles"] = many
let capped = try decode(changed).subtitles
try require(capped.count == 32 && capped.last?.selectionID == "youtube:auto:en"
&& capped.first?.language == "lang-00" && capped[30].language == "lang-30",
"deterministic 32-track cap reserves original auto instead of crowding it out with manual languages")
changed["automatic_captions"] = [:]
try require(try decode(changed).subtitles.count == 32,
"manual-only metadata uses the entire track allowance")
changed["automatic_captions"] = ["en-orig": [vtt(original + "&tlang=fr")]]
let translated = try decode(changed).subtitles
try require(translated.count == 32
&& translated.allSatisfy { $0.selectionID?.hasPrefix("youtube:manual:") == true },
"translation query metadata is excluded even under an original-looking key")
for (key, value) in [("is_live", true), ("live_status", "is_live")] as [(String, Any)] {
var broadcast = object
broadcast[key] = value
try require(try decode(broadcast).subtitles.isEmpty,
"live captions are disabled instead of freezing a partial live transcript")
}
changed["url"] = "file:///unsafe-video"
try reject("optional caption tolerance does not weaken video validation") { _ = try decode(changed) }
}
static func chapters() throws {
var object = try fixture(combinedJSON)
try require(try decode(object).chapters.isEmpty, "absent chapters remain playable")
@@ -683,6 +789,7 @@ struct YouTubeResolverChecks {
try videoURLs()
try metadata()
try chapters()
try subtitles()
try malformedMetadata()
try qualities()
try bundledHelpers(root: root)