Replacement and tokenization

Replace selected ranges or split text into literal and matched tokens.

replace(text, replacement, options?)

Replaces non-overlapping original ranges once. Default strategy is leftmost-longest; leftmost-first and longest-first are also accepted, all is rejected. A callback receives (match, originalSubstring) and must return a string. String replacements are literal: $& has no special meaning. Inserted text is not searched again. Selected matches are consumed incrementally, although output text and intermediate string pieces still require memory.

import AhoCorasick from 'modern-ahocorasick'

new AhoCorasick(['cat']).replace('cat cat', '$&') // '$& $&'

tokenize(text, options?)

tokenize(text, options?) defaults to leftmost-longest and rejects all. It returns Token<T>[]: either { type: 'text', text, start, end } or { type: 'match', text, start, end, match }. No empty text tokens are generated. Concatenating token.text reconstructs the original input. The library emits no HTML.

import AhoCorasick from 'modern-ahocorasick'

const matcher = new AhoCorasick(['cat'])
const tokens = matcher.tokenize('a cat!')
tokens.map(token => [token.type, token.text])
// [['text', 'a '], ['match', 'cat'], ['text', '!']]
tokens.map(token => token.text).join('') // 'a cat!'

Selection and output

Both methods accept the three non-overlapping strategies. longest-first buffers candidates; the leftmost strategies settle candidates with bounded lookahead. Output strings and token arrays still consume memory. Query ranges filter matches while preserving text outside the range.

See Safe highlighting; see Replacement helpers.