BEAM Lab Languages

View Source

Linguistic metadata for human languages: grammatical gender, writing direction, canonical and native names, and BCP 47 normalization. Curated, compile-time data with zero runtime dependencies.

Sibling library to beamlab_countries — beamlab_countries knows where languages are spoken, beamlab_languages knows what they are like.

What it answers

  • "Does Russian use grammatical gender? If so, what genders?"
  • "Is Arabic written right-to-left?"
  • "What's the canonical English name of fr? The endonym?"
  • "Does the user's locale string en-US collapse to a base I can use as a key?"

Installation

defp deps do
  [
    {:beamlab_languages, "~> 0.1"}
  ]
end

Then mix deps.get.

Quick start

BeamlabLanguages.has_gender?("fr")
# true

BeamlabLanguages.genders("de")
# ["m", "f", "n"]

BeamlabLanguages.direction("ar")
# :rtl

BeamlabLanguages.name("ja")
# "Japanese"

BeamlabLanguages.native_name("ja")
# "日本語"

BeamlabLanguages.normalize("en-US")
# "en"

BeamlabLanguages.get("fr")
# %BeamlabLanguages.Language{
#   code: "fr",
#   name: "French",
#   native_name: "Français",
#   direction: :ltr,
#   has_gender: true,
#   genders: ["m", "f"]
# }

Every function that takes a language code runs normalize/1 internally, so "en-US", "FR", and " fr " all work. Predicates (has_gender?/1, known?/1) return false for nil or unknown input rather than raising — handy in form-validation paths.

Documentation

Full API docs at HexDocs.

Coverage

v1 covers 50+ languages: the top-spoken languages worldwide plus all CEFR / JLPT / HSK targets. The data lives in priv/data/languages.json — open a PR to add more or correct an entry.

Roadmap (planned, not in v1)

These are intentionally deferred so v1 ships small. The v1 API is shaped to leave room for them:

  • Localized language names — BeamlabLanguages.name("fr", in: "es") → "francés"
  • Plural rules (CLDR categories: :zero, :one, :two, :few, :many, :other)
  • Articles (definite/indefinite, by gender)
  • Case marking (Slavic, Finnic, etc.)
  • Noun classes (Bantu)
  • Scripts / writing systems per language
  • IPA inventory
  • Honorific levels (Japanese / Korean)

Non-goals

  • Not a CLDR wrapper. No locale formatting (numbers, dates, currencies). That belongs elsewhere.
  • Not a translation API. Knows what languages are; doesn't translate text.
  • No GenServer / Agent / ETS. All data is compile-time.

Contributing

  1. Fork it
  2. Create a feature branch (git checkout -b my-new-feature)
  3. Edit priv/data/languages.json and/or code
  4. mix test and mix format
  5. Open a PR

License

MIT — see LICENSE.md.