moonbit
deps moonbitlang/regexp path .

🔍 regexp.mbt

⚠️ API STABILITY NOTICE
This is a alpha release with a stabilizing API. While core functionality is complete and well-tested,
API changes may occur in future versions as we refine the implementation.

Regular expression engine for MoonBit — inspired by Russ Cox's regex series.

⚡ Quick Start

test {
  // Compile once, use everywhere
  let regexp = @regexp.compile("a(bc|de)f")
  guard regexp.match_("xxabcf") is Some(result)
  inspect(
    result.results(),
    content=(
      #|[Some("abcf"), Some("bc")]
    ),
  )

  // Write a simple split with regexp
  fn split(
    regexp : @regexp.Regexp,
    target : @string.View
  ) -> Array[@string.View] {
    let result = []
    loop target {
      "" => ()
      str => {
        let res = regexp.execute(str)
        result.push(res.before())
        continue res.after()
      }
    }
    result
  }

  let re = @regexp.compile("_+")
  inspect(
    split(re, "1_2__3__4__5_____6"),
    content=(
      #|["1", "2", "3", "4", "5", "6"]
    ),
  )
}

🎯 Core API

Build & Execute

• compile(pattern) → Creates an Engine
• engine.execute(text) → Returns MatchResult

Inspect Results

• result.matched() → Bool
• result.get(index) → Capture group content
• result.results() → Iterator over all matches

Named Groups & Advanced

• engine.group_by_name(name) → Find group index by name
• engine.group_count() → Total capture groups
• result.groups() → Get named group content

🎪 Syntax Playground

Feature	Example	What it does
Literals	abc	Match exact text
Wildcards	a.c	. matches any character
Quantifiers	a+, b*, c?	One or more, zero or more, optional
Ranges	a{2,5}	Between 2-5 repetitions
Classes	[a-z], [^0-9]	Character sets, negated sets
Groups	(abc), (?:xyz)	Capturing, non-capturing
Named	(?<word>abc)	Named capture groups
Choice	cat|dog	Match either option
Anchors	^start, end$	Line boundaries
Escapes	\\u{41}, \\u0041	Unicode escapes, standard escapes
Unicode Props	\\p{L}, \\p{Nd}	Unicode general categories
Backrefs ⚠️	(.)\\1	Reference previous captures

🌍 Unicode Property Support

Match characters by their Unicode general categories:

test "unicode properties" {
  // Matching gc=L
  let regex = @regexp.compile("\\p{Letter}+")
  inspect(
    regex.execute("Hello 世界").results(),
    content=(
      #|[Some("Hello")]
    ),
  )

  // Matching gc=N
  let regex = @regexp.compile("\\p{Number}+")
  inspect(
    regex.execute("123 and 456").results(),
    content=(
      #|[Some("123")]
    ),
  )
}

Supported Propertes:

• General Category

🔄 Backreferences

⚠️ Performance Warning: Backreferences can cause exponential time complexity in worst cases!

test "backreferences" {
  // Palindrome detection (simple)
  let palindrome = @regexp.compile("^(.)(.)\\2\\1")
  inspect(
    palindrome.execute("abba").results(),
    content=(
      #|[Some("abba"), Some("a"), Some("b")]
    ),
  )

  // HTML tag matching
  let html_regex = @regexp.compile("<([a-zA-Z]+)[^>]*>(.*?)</\\1>")
  let result = html_regex.execute("<div class='test'>content</div>")
  inspect(
    result.results(),
    content=(
      #|[Some("<div class='test'>content</div>"), Some("div"), Some("content")]
    ),
  )
}

💡 Real Examples

test "character classes" {
  // Email validation (simplified)
  let email = @regexp.compile(
    (
      #|[\w-]+@[\w-]+\.\w+
    ),
  )
  let email_result = email.execute("[email protected]").results()
  inspect(
    email_result,
    content=(
      #|[Some("[email protected]")]
    ),
  )
  // Extract numbers
  let numbers = @regexp.compile(
    (
      #|\d+\.\d{2}
    ),
  )
  let result = numbers.execute("Price: $42.99").results()
  inspect(
    result,
    content=(
      #|[Some("42.99")]
    ),
  )

  // Named captures for parsing
  let parser = @regexp.compile(
    (
      #|(?<year>\d{4})-(?<month>\d{2})-(?<day>\d{2})
    ),
  )
  let date_result = parser.execute("2024-03-15")
  inspect(
    date_result.groups(),
    content=(
      #|{"year": "2024", "month": "03", "day": "15"}

    ),
  )
}

🚨 Error Handling

test {
  try {
    let _ = @regexp.compile("a(b")  // Oops! Missing )
  } catch {
    RegexpError(err=MissingParenthesis, source_fragment=_) => println("Fix your regex! 🔧")
    _ => ()
  }
}

⚡ Performance Characteristics

• Predictable complexity — Designed to avoid catastrophic backtracking (except with backreferences)
• VM-based — Structured interpreter design
• Unicode support — Character set and property support

Built with reliability and correctness as primary goals.

🔍 Implementation Notes

Behavior Differences from Other Engines

This implementation has some behavior differences compared to other popular regex engines:

1. Empty Character Class Handling:

   • In JavaScript: [][] is parsed as two character classes with no characters
   • In Golang: [][] is parsed as one character class containing ] and [
   • In MoonBit: we follow the JavaScript interpretation

2. Empty Alternatives Behavior:

   • Expressions like (|a)* and (|a)+ have specific behavior that may differ from other implementations
   • See Golang issue #46123 for related discussion

3. Backreferences:

   • Backreferences are supported but may impact the complexity guarantees of the engine