How-to

Odd input and pieces that will not load

How a text is made plain before it is looked up, what a lookup does with input that is empty or odd, and what WordsLoadError names.

Odd input, and a piece that will not load

A text is made plain before it is looked up (normalizeText does it, and you can too): composed and compatibility forms are folded together by NFKC, so full-width letters and digits are half-width and half-width kana full-width; whitespace is collapsed and a space beside Japanese removed; zero-width and control characters are removed; a dash between kana, or ending kana, is the long vowel mark ー. Hiragana and katakana are folded as before. A written form JMdict writes in full-width characters (１人) is found by either spelling, and shown as JMdict writes it.

Whatever the text is, a lookup does not throw. A text that is empty, is not a string, or is far too long to be a word (over 100 characters) is nothing to load and has no words. wordsOf still throws a RangeError for a text that was not loaded, because that is a mistake in the caller.

A piece that cannot be fetched is asked for once more, and if that fails too, loadWords, loadCommonWords, findByMeaning and wordsWithKanji reject with a WordsLoadError. It says which piece (error.piece is "the readings piece 12 of 318", error.kind and error.index say it for a program, error.cause is what the fetch rejected with), and nothing is remembered of the failure, so asking again tries again. Lookups at the same moment for the same pieces share one fetch.

import {loadWords,WordsLoadError} from "@johnmorrisdotca/jisho/words";
try {
  await loadWords("かみ");
} catch (error) {
  if (error instanceof WordsLoadError) console.log(error.kind,error.index);
}

Odd input and pieces that will not load

Odd input, and a piece that will not load

See also