languageBaseCode: One canonical form of a language code, for comparing codes written in different ISO 639 forms: "fra", "fre" and "fr" all give "fr", "sw" and "swa" give "sw". A region or script suffix is dropped first ("pt-BR" becomes "pt"), as resolveLanguageName() does. A code neither map knows is returned lowercased as-is.

Comprehensive English-name fallback for ISO 639-2 codes that the app does not have a UI translation for (for example "lat" Latin, "swa" Swahili). Used as tier 2 by resolveLanguageName(): translatable languages always go through translate() first; this only fires when no LanguageX key exists for the resolved base code, ensuring no recognized 3-letter code ever displays raw.

Names are the official ISO 639-2 forms. They are intentionally not routed through the i18n pipeline: these languages are unlikely to be a UI locale for our users, so the maintenance cost of translating "Akkadian" into Polish (etc.) outweighs the UX benefit. Users see the English name in any UI when an audio/subtitle track is in one of these less-common languages.

Cached on the caller's m scope after first build.

Maps a normalized 2-letter base code (or a 3-letter code for languages without a 639-1 equivalent) to the translation key used by translate() for the localized language name. Keys must exist in locale/custom/en_US.json.

Region/script suffixes (for example "_BR", "_Hant") are stripped before lookup by resolveLanguageName(), so en_GB and en_US both land at "en", then LanguageEn.

Cached on the caller's m scope after first build.

languagesMatch: True when two language codes name the same language, whatever ISO 639 form each is written in (see languageBaseCode()). Regional variants count as the same language.

Maps 3-letter ISO 639-2 codes (T = terminological, B = bibliographic) and deprecated ISO 639-1 codes to the 2-letter base codes used by the app's translation system. Only languages the app supports as a UI locale are covered (see scripts/lint/update-translations.cjs LANGUAGE_METADATA).

Why this exists: ffmpeg/Jellyfin pass through whatever the container metadata holds: most modern files use 639-2/T (for example "fra"), older or hand-tagged files use 639-2/B (for example "fre"), and some use 639-1 ("fr" already, no alias needed). MediaStream.Language is therefore not a single standardized format; this map normalizes to one canonical form.

3-letter codes that have no 2-letter ISO 639-1 equivalent (ckb, fil, gsw, jbo, kab) are not aliased here: they're looked up directly in languageTranslationKeys() under their 3-letter form.

Cached on the caller's m scope after first build to avoid reallocating on every audio/subtitle render.

ISO 639-2 codes for the languages mediaLanguageAliases() leaves out (it covers only UI locales), mapped to their ISO 639-1 code. For language MATCHING only: see languageBaseCode(). Kept apart from mediaLanguageAliases() because display depends on that map's gaps: languageEnglishFallbacks() is keyed by the 3-letter code, so aliasing "swa" to "sw" there would show Swahili as the raw "swa".

Source: the ISO 639-2 list (Library of Congress) as Jellyfin 12.0 ships it, Emby.Server.Implementations/Localization/iso6392.txt: every code with an ISO 639-1 equivalent that mediaLanguageAliases() does not already map. lint:language-coverage fails if the two maps overlap.