/// HTML entity decoding for the markdown subset. library; const Map _namedEntities = { 'amp': '&', 'lt': '<', 'gt': '>', 'quot': '"', 'apos': "'", 'nbsp': '\u00A0', 'hellip': '\u2026', 'mdash': '\u2014', 'ndash': '\u2013', 'lsquo': '\u2018', 'rsquo': '\u2019', 'ldquo': '\u201C', 'rdquo': '\u201D', }; final RegExp _entity = RegExp( r'&(#x[0-9A-Fa-f]+|#[0-9]+|[A-Za-z][A-Za-z0-9]*);', ); final RegExp _hexReference = RegExp(r'^#x[0-9A-Fa-f]+$'); final RegExp _decimalReference = RegExp(r'^#[0-9]+$'); /// Decodes the HTML entities a message may carry in one left-to-right pass, so /// `&lt;` becomes `<` and is never decoded twice. Unknown or malformed /// references stay exactly as written. String decodeHtmlEntities(String text) { if (!text.contains('&')) return text; return text.replaceAllMapped(_entity, (match) { final body = match.group(1)!; final named = _namedEntities[body]; if (named != null) return named; return _numericReference(body) ?? match.group(0)!; }); } String? _numericReference(String body) { int? codePoint; if (body.startsWith('#x') || body.startsWith('#X')) { if (_hexReference.hasMatch(body)) { // `tryParse` returns null for a value too large for an int; the web // reference parses such a value as an out-of-range number instead, and // both paths keep the literal. Never let an entity throw. codePoint = int.tryParse(body.substring(2), radix: 16); } } else if (body.startsWith('#')) { if (_decimalReference.hasMatch(body)) { codePoint = int.tryParse(body.substring(1)); } } if (codePoint == null || codePoint < 1 || codePoint > 0x10FFFF) return null; return String.fromCharCode(codePoint); }