Load only the needed word lists

Direct data or a named loader

import {readWordLists,unpack} from "@johnmorrisdotca/kotoba";
import {EN_WORDS} from "@johnmorrisdotca/kotoba/words-en";
import {JA_WORDS_3} from "@johnmorrisdotca/kotoba/kana-3";
console.log(readWordLists(EN_WORDS,5)?.easy.length,unpack(JA_WORDS_3,3).release);

880 2026-09-28

The alphabet lists contain whitespace-separated strings by length. readWordLists returns easy answers, all answers and the allowed-guess set, or null for an absent length. Kana data packs one character per kana; unpack uses its alphabet/code mapping and returns the source release too. /load chooses en/fr/de at 4–6 letters or ja at 3–5 kana and returns null for unsupported sizes.

Keep the source terms

Answer pools and allowed guesses have different jobs. Do not choose a hidden word from the entire allowed set unless that is your host's intended rule. Pop-culture answers and guesses have separate entries. The root rules import carries no dictionary; a bundler can split the lazy loader's reachable lists into chunks.

Keep NOTICE.md and the list's own notices with data you distribute. The package documents SCOWL, French/German sources and JMdict, with MIT code and the data's own terms.