Protected text filters

Keep supported URL and Markdown syntax unchanged while replacing surrounding text.

Attach a filter

import AhoCorasick from 'modern-ahocorasick'
import { replaceChunks } from 'modern-ahocorasick/stream'
import { protectedText } from 'modern-ahocorasick/stream/filters'

const matcher = new AhoCorasick(['cat'])
const result = [...replaceChunks(
  matcher,
  ['cat `cat` https://example.com/cat'],
  'DOG',
  { filter: protectedText({ urls: true, markdown: true }) },
)].join('')
// 'DOG `cat` https://example.com/cat'

Recognized syntax

urls() protects HTTP/HTTPS schemes (case insensitive) through whitespace or EOF. markdown({ heading?, code? }) defaults both flags to true. It recognizes ATX headings at the start of a line with 1–6 hashes, inline backtick spans with equal delimiter lengths, and line-start backtick/tilde fences of length ≥3. Closing fences use the same character, at least the opening length, and optional trailing spaces/tabs. Indented fences, escapes, links and other CommonMark rules are outside this syntax profile. Unclosed inline code becomes ordinary text at EOF; unclosed fenced code stays protected. protectedText({ urls: true, markdown: true }) combines both rules. Protected text passes through unchanged, and a match cannot cross it. If a syntax edge intersects a grapheme, the entire grapheme is protected.

Buffering and EOF

An unfinished syntax span can retain undecided text until EOF and reach maxBufferLength. Always finish the session or consume the iterable to completion. Filters are a documented syntax subset, not a full Markdown parser.

See Session buffer limits.