@cto.af/string-width

Get width of a Unicode string in fixed-width display cells, accounting for combining characters, emoji, flags, Hangul, East Asian Width, default ignorable characters, and a few more edge cases.

Installation

npm install @cto.af/string-width

API

import {StringWidth, AMBIGUOUS, POTENTIAL_EMOJI} from '../lib/index.js'

const sw = new StringWidth()
sw.width('foo') // 3
sw.width('\u{1F4A9}') // 2: Emoji take two cells
sw.width('#\ufe0f\u20e3') // 2: More complicated emoji
sw.break('foobar', 3) // [
  //   {string: 'foo', cells: 3, last: false},
  //   {string: 'bar', cells: 3, last: true}
  // ]

const custom = new StringWidth({
  locale: 'ko-KR',
  isCJK: true,
  extraWidths: new Map([
    // This example is not actually useful, but demonstrates how to customize

    // 'K' how has ambiguous East Asian Width
    [0x4b, AMBIGUOUS],
    // 'O' might now start an Emoji sequence
    [0x4f, POTENTIAL_EMOJI],
    // 'R' now has a width of 3 cells
    [0x52, 3]
  ])
})

Approach

• All of the "interesting" characters are put into a Trie in widths.js at build time. POTENTIAL_EMOJI(14) is a sentinel for "possible emoji". AMBIGUOUS(15) is a sentinel for "ambiguous East Asian Width", and all the rest of the values are the width for that code point. The default result from the Trie is 1.
• If a string is all ASCII, just count characters. This happens a lot, so it's worth a performance shortcut
• For each grapheme cluster:
  • Get the width of the first code point from extraWidths or the Trie.
  • If the width is AMBIGUOUS, return 2 if we're in a CJK context, otherwise 1.
  • If the width is POTENTIAL_EMOJI, check if the whole grapheme cluster is an emoji
• Since backspace has a negative width, ensure that the total width is never less than zero.

Chinese, Japanese, or Korean (CJK) contexts

Some code points have ambiguous length, which depends upon whether we are counting in a CJK context or not. By default, StringWidth will look at the locale that is given (or derived from the environment), and use the default script of that locale to decide if this is a Chinese, Japanese, or Korean context. The script identifiers 'Hans', 'Hant', 'Jpan', and 'Kore' signal CJK context. If desired, this detection can be overridden by passing in the isCJK field in the constructor options.

Width breaking

The break(string, N) method slices a string into chunks, each of which is at most N cells. This was so entangled with the width logic that it made sense to be in this library. It is useful for strings that are longer than N that need to have a hyphen inserted between each of the segments, ensuring that the hyphen doesn't go in the middle of a grapheme cluster.

Development

On a new Unicode version being released, delete the tools/*.txt files, then do npm run build to re-generate the Trie.