TypesBCP47

TypeScript/Node.js library that parses the IANA Language Subtag Registry and exposes it as typed objects, to validate, normalize and describe BCP 47 language tags.

Installation

npm install typesbcp47

How it works

The package bundles a copy of language-subtag-registry.txt (the raw IANA registry file) and parses it at runtime with RegistryParser — there are no network calls. The registry's File-Date can be read with RegistryParser.getRegistryDate().

Only the language, region, script and variant record types are loaded into memory. extlang, grandfathered and redundant records (over 350 entries combined) are parsed but not indexed, so tags that rely on them (e.g. grandfathered tags like art-lojban, or extlang subtags like ar-afb) are not recognized by isValidLanguageTag/normalizeCode/getTagDescription.

A second, independent data source is used for locale-aware language names: extendedLanguageList_{locale}.xml (~866 entries, includes language+region combinations such as af-NA) and languageList_{locale}.xml (~336 "common" entries). These are curated/translated name lists, not derived from the IANA registry, and are currently shipped only for en, es and fr — requesting any other locale throws. LanguageUtils.isBiDi() also reads its data from these files, not from the registry.

Runtime notes

• Node.js only. Resources are read from disk relative to the compiled module's location (node:fs/node:path/node:url); the package is not meant to be bundled directly for the browser.
• ESM only — package.json sets "type": "module" and the exports map has no require condition.
• Depends on typesxml (same author) for XML parsing of the resource files.
• LanguageUtils keeps static, process-wide caches (registryParser singleton, languagesCache per locale, bidiCodes). Data is parsed once, lazily, on first use, and reused for the life of the process — it is not reloaded or invalidated afterwards.

Exported classes

Class	Purpose
LanguageUtils	Static facade — the main entry point, see table below.
RegistryParser	Parses language-subtag-registry.txt into languages/regions/scripts/variants maps and private-use ranges. Exposes getRegistryDate(). LanguageUtils creates one lazily on first use; you can also instantiate it directly for lower-level access to the parsed registry.
RegistryEntry	One raw %%-delimited record from the registry file, exposed as a Type/Subtag/Description/... key-value map via get(type).
Language	code + description, plus getSuppressedScript(), isCJK(), isBiDi() (the last two delegate to LanguageUtils).
Region, Script, Variant	code + description wrappers (Variant also carries getPrefix(), the language it's registered under).

LanguageUtils API

Method	Description
isValidLanguageTag(tag: string, caseSensitive?: boolean): boolean	Validates a tag against the registry. x-... and ...-x-... private-use tags are always accepted without a registry lookup. If caseSensitive is true, the tag must already be in its canonical case (see normalizeCode).
normalizeCode(code: string): string | undefined	Returns the canonical form of a tag: language lowercase, region uppercase (or the registry's UN M.49 numeric code as-is), script title-case, variant lowercase. Returns undefined if the tag doesn't resolve against the registry.
getTagDescription(tag: string): string | undefined	Builds a human-readable description by composing subtag descriptions from the registry, e.g. en-GB → English (United Kingdom). Handles at most 3 subtags (language + one of script/region/variant + one more of region/variant); does not compose descriptions for tags with 4+ subtags or more than one variant. Returns undefined for an unknown tag, and also when the script subtag matches the language's Suppress-Script (e.g. en-Latn, since it's redundant).
getLanguage(code: string, locale: string): Language | undefined	Looks up a Language by code, normalizing it first; description is in locale.
getLanguages(locale: string): Array<Language>	All ~866 entries from extendedLanguageList_{locale}.xml, sorted by description using locale-aware collation.
getCommonLanguages(locale: string): Array<Language>	The smaller ~336-entry curated subset from languageList_{locale}.xml.
isCJK(code: string): boolean	Pure prefix check on the code — true if it starts with zh, ja, ko, vi, ain or aib. Not a registry lookup.
isBiDi(code: string): boolean	true if code is one of the languages flagged bidi="true" in the locale XML (currently: ar, ckb, fa, he, ks, lrc, mzn, ps, sd, ug, ur, yi). Lazily loads the en list on first call if no locale data has been loaded yet.

Note: locale parameters accept en, es or fr only.

Example

import { LanguageUtils } from 'typesbcp47';

const tag = 'en-GB';

if (LanguageUtils.isValidLanguageTag(tag)) {
  console.log(LanguageUtils.getTagDescription(tag)); // "English (United Kingdom)"
  console.log(LanguageUtils.normalizeCode('EN-gb'));  // "en-GB"
  console.log(LanguageUtils.isBiDi('ar'));            // true
  console.log(LanguageUtils.isCJK('ja'));              // true
}

const language = LanguageUtils.getLanguage('fr-CA', 'es');
if (language) {
  console.log(language.getDescription()); // Spanish-language description of "fr-CA"
}