Add conversion from regex syntax to BNF syntax (#803)

Author: milseman

Package.swift

@@ -135,6 +135,13 @@ let package = Package(
                 "_RegexParser",
                 "_StringProcessing"
             ]),
+        .executableTarget(
+            name: "Regex2BNF",
+            dependencies: [
+                .product(name: "ArgumentParser", package: "swift-argument-parser"),
+                "_RegexParser"
+            ],
+            swiftSettings: [availabilityDefinition]),
         .executableTarget(
             name: "RegexTester",
             dependencies: [

Sources/Regex2BNF/Regex2BNF.swift

@@ -0,0 +1,91 @@
+//===----------------------------------------------------------------------===//
+//
+// This source file is part of the Swift.org open source project
+//
+// Copyright (c) 2021-2025 Apple Inc. and the Swift project authors
+// Licensed under Apache License v2.0 with Runtime Library Exception
+//
+// See https://swift.org/LICENSE.txt for license information
+//
+//===----------------------------------------------------------------------===//
+
+import ArgumentParser
+import _RegexParser
+
+@main
+@available(SwiftStdlib 5.8, *)
+struct Regex2BNF: ParsableCommand {
+  @Argument(help: "The regex pattern to convert to BNF.")
+  var pattern: String
+
+  @Flag(
+    name: [.customShort("e"), .customLong("examples")],
+    help: "Run several examples")
+  var runExamples = false
+
+  func convert(_ pattern: String) throws {
+    print("/\(pattern)/\n")
+    let bnf = try convertRegexToBNF(
+        regex: pattern, namespace: "RE2BNF", version: 0)
+    print(bnf)
+  }
+
+  mutating func run() throws {
+    if runExamples {
+      // TODO: Turn into test cases
+//      print("[Examples")
+
+//      print("Single-scalar character literals:")
+      try convert("a")
+      try convert("Z")
+      try convert("あ")
+      try convert("日")
+      try convert("\u{301}")
+
+
+//      print("Multi-scalar character literals")
+      try convert("🧟‍♀️")
+      try convert("e\u{301}")
+
+//      print("Simple alternations")
+      try convert("a|b")
+      try convert("a|b|c|d")
+      try convert("a|🧟‍♀️\u{301}日|z")
+
+//      print("Simple quantifications")
+      try convert("a*")
+      try convert("a+")
+      try convert("a?")
+      try convert("a{2,10}")
+      try convert("a{,10}")
+      try convert("a{2,}")
+
+//      print("Grouping")
+      try convert("a(b|c)d")
+      try convert("a(?:b|c)d")
+      try convert("a(bcd|def(g|h)+)z")
+
+//      print("Dot")
+      try convert(".*")
+      try convert("(a|b)*.{3}(a|b)")
+
+//      print("Bultin character classes")
+      try convert(#"\(\d{3}\)\d{3}-\d{4}"#)
+      try convert(#"\s+keyword\s+"#)
+
+
+//      print("[Done]")
+
+      // Look at optimizer output, the quant child is very long
+      try convert("a(123456789)+b")
+
+      try convert("Hi the time right now is (AM|PM)")
+
+      try convert("a(b|c)*d{2,4}e?")
+    }
+    try convert(pattern)
+
+
+
+  }
+}

Sources/_RegexParser/Regex/BNF/BNF.swift

@@ -0,0 +1,186 @@
+//===----------------------------------------------------------------------===//
+//
+// This source file is part of the Swift.org open source project
+//
+// Copyright (c) 2021-2025 Apple Inc. and the Swift project authors
+// Licensed under Apache License v2.0 with Runtime Library Exception
+//
+// See https://swift.org/LICENSE.txt for license information
+//
+//===----------------------------------------------------------------------===//
+
+protocol BNFNode: CustomStringConvertible {
+  func render() -> String
+}
+extension BNFNode {
+  var description: String { render() }
+}
+
+struct BNF: BNFNode {
+  var root: Rule
+  var rules: [Rule]
+
+  func render() -> String {
+    var str = ""// root.render() + "\n"
+    if rules.isEmpty {
+      return str
+    }
+    return str
+      + rules.lazy.map {
+        $0.render()
+      }.joined(separator: "\n")
+      + "\n"
+  }
+}
+
+struct Rule: BNFNode {
+  // The left-hand side
+  var symbol: NonTerminalSymbol
+
+  var expression: Expression
+
+  var predicates: [CharacterPredicate] = []
+
+  func render() -> String {
+    "\(symbol.render()) ::= \(expression.render())"
+  }
+}
+
+struct CharacterPredicate {
+  // TODO: convention c or trivial?
+  let impl: (Unicode.Scalar) -> Bool
+}
+
+struct NonTerminalSymbol: Hashable, BNFNode {
+  var name: String
+
+  func render() -> String {
+    "<\(name)>"
+  }
+}
+
+struct Expression: BNFNode {
+  var choices: [Choice]
+
+  func render() -> String {
+    "\(choices.map({ $0.render() }).joined(separator: " | "))"
+  }
+}
+
+struct Choice: BNFNode {
+  var sequence: [Symbol]
+
+  init(_ symbols: Array<Symbol>) {
+    self.sequence = symbols
+  }
+  init(_ symbols: Symbol...) {
+    self.init(symbols)
+  }
+
+  func render() -> String {
+    "\(sequence.map({ $0.render() }).joined(separator: " "))"
+  }
+}
+
+enum Symbol: BNFNode {
+  case terminal(TerminalSymbol)
+  case terminalSequence([TerminalSymbol])
+  case nonTerminal(NonTerminalSymbol)
+  case builtin(Builtin)
+
+  func render() -> String {
+    switch self {
+    case .terminal(let t):
+      return t.render()
+
+    case .terminalSequence(let s):
+      guard !s.isEmpty else {
+        return "\"\""
+      }
+      return "\(s.map({ $0.render() }).joined(separator: " "))"
+
+    case .nonTerminal(let n):
+      return n.render()
+
+    case .builtin(let b):
+      return b.render()
+    }
+  }
+}
+
+enum Builtin: BNFNode {
+  case any   // NOTE: we map dot to this, not sure if we want non-newline dots
+  case whitespace
+  case notWhitespace
+  case decimalDigit
+  case notDecimalDigit
+  case wordCharacter
+  case notWordCharacter
+
+  func render() -> String {
+    switch self {
+    case .any:
+      return "<ALL_CHARACTERS_EXCEPT_QUOTE_AND_BACKSLASH>"
+    case .whitespace:
+      return "<WHITESPACES_AND_NEWLINES>"
+    case .notWhitespace:
+      fatalError()
+    case .decimalDigit:
+      return "<DECIMAL_DIGITS>"
+    case .notDecimalDigit:
+      fatalError()
+    case .wordCharacter:
+      return "<ALPHANUMERICS>"
+    case .notWordCharacter:
+      fatalError()
+    }
+  }
+}
+
+enum CharacterSet {}
+
+enum TerminalSymbol: BNFNode {
+  case character(Unicode.Scalar)
+  case characterSet(CharacterSet)
+  case utf8CodeUnit(UInt8)
+
+  case characterPredicate(CharacterPredicate)
+
+  func render() -> String {
+    switch self {
+    case .character(let c):
+      return "\"\(c)\""
+    case .characterSet(let _):
+      fatalError()
+    case .utf8CodeUnit(let u):
+      return "\"\(u)\""
+    case .characterPredicate(_):
+      fatalError()
+    }
+  }
+}
+
+extension Expression {
+  init(_ choices: [Choice]) {
+    self.init(choices: choices)
+  }
+  init(_ choices: Choice...) {
+    self.init(choices)
+  }
+}
+
+extension Choice {
+  init(_ elements: [NonTerminalSymbol]) {
+    self.init(elements.map { .nonTerminal($0) })
+  }
+  init(_ elements: NonTerminalSymbol...) {
+    self.init(elements)
+  }
+}
+
+/*
+
+
+node -> choice
+
+ */