Locates runtime data files used by Text modules, fetching them from upstream sources when permitted.

Most modules in this package ship a small bundled dataset (e.g. Text.Hyphenation ships American English patterns, Text.Lemma ships English lemmatization data) but also accept additional language packs at runtime. This module is the single configuration surface those modules share for finding, caching, and optionally downloading those packs.

Configuration

config :text,
  # Where on-demand-downloaded data is cached.
  # Default: "~/.cache/text".
  data_dir: "~/.cache/text",

  # Per-domain auto-download permissions. All default to false:
  # the package never reaches out to the network without an
  # explicit opt-in for the relevant domain.
  auto_download_hyphenation_data: true,
  auto_download_lemma_data: true,
  auto_download_wordfreq_data: false

Lookup order

When a Text module asks for a data file via fetch/3:

1. The configured :data_dir is checked first. Once a file has been cached there (whether downloaded by this package, copied manually, or shipped by a deployment script) it is reused on every subsequent call.

2. The bundled priv/<domain>/ directory inside the :text package is checked next.

3. If a download URL is known and auto_download_<domain>_data is true, the file is downloaded to the cache directory and the cached path is returned.

4. Otherwise {:error, %ArgumentError{}} is returned with a message describing how to resolve the situation (enable auto-download, or place the file manually).

Domains

Each Text module that needs runtime data uses its own domain atom — :hyphenation, :lemma, :wordfreq. Domains map directly to subdirectory names under both priv/ and :data_dir, and to per-domain permission keys.