Move AssertionKind onto the DSL

This enum will start including cases that only the
DSL can use, so move it off the AST.

Author: hamishknight

Sources/RegexBuilder/Anchor.swift

@@ -37,16 +37,30 @@ public struct Anchor {
 
 @available(SwiftStdlib 5.7, *)
 extension Anchor: RegexComponent {
-  var baseAssertion: DSLTree._AST.AssertionKind {
+  var baseAssertion: DSLTree.Atom.Assertion {
     switch kind {
-    case .startOfSubject: return .startOfSubject(isInverted)
-    case .endOfSubjectBeforeNewline: return .endOfSubjectBeforeNewline(isInverted)
-    case .endOfSubject: return .endOfSubject(isInverted)
-    case .firstMatchingPositionInSubject: return .firstMatchingPositionInSubject(isInverted)
-    case .textSegmentBoundary: return .textSegmentBoundary(isInverted)
-    case .startOfLine: return .startOfLine(isInverted)
-    case .endOfLine: return .endOfLine(isInverted)
-    case .wordBoundary: return .wordBoundary(isInverted)
+    case .startOfSubject:
+      // FIXME: Inverted?
+      return .startOfSubject
+    case .endOfSubjectBeforeNewline:
+      // FIXME: Inverted?
+      return .endOfSubjectBeforeNewline
+    case .endOfSubject:
+      // FIXME: Inverted?
+      return .endOfSubject
+    case .firstMatchingPositionInSubject:
+      // FIXME: Inverted?
+      return .firstMatchingPositionInSubject
+    case .textSegmentBoundary:
+      return isInverted ? .notTextSegment : .textSegment
+    case .startOfLine:
+      // FIXME: Inverted?
+      return .caretAnchor
+    case .endOfLine:
+      // FIXME: Inverted?
+      return .dollarAnchor
+    case .wordBoundary:
+      return isInverted ? .notWordBoundary : .wordBoundary
     }
   }
 

Sources/_RegexParser/Regex/AST/Atom.swift

@@ -511,67 +511,6 @@ extension AST.Atom.CharacterProperty {
   }
 }
 
-extension AST.Atom {
-  /// Anchors and other built-in zero-width assertions.
-  public enum AssertionKind: String, Hashable {
-    /// \A
-    case startOfSubject = #"\A"#
-
-    /// \Z
-    case endOfSubjectBeforeNewline = #"\Z"#
-
-    /// \z
-    case endOfSubject = #"\z"#
-
-    /// \K
-    case resetStartOfMatch = #"\K"#
-
-    /// \G
-    case firstMatchingPositionInSubject = #"\G"#
-
-    /// \y
-    case textSegment = #"\y"#
-
-    /// \Y
-    case notTextSegment = #"\Y"#
-
-    /// ^
-    case caretAnchor = #"^"#
-
-    /// $
-    case dollarAnchor = #"$"#
-
-    /// \b (from outside a custom character class)
-    case wordBoundary = #"\b"#
-
-    /// \B
-    case notWordBoundary = #"\B"#
-
-  }
-
-  public var assertionKind: AssertionKind? {
-    switch kind {
-    case .caretAnchor:  return .caretAnchor
-    case .dollarAnchor: return .dollarAnchor
-
-    case .escaped(.wordBoundary):    return .wordBoundary
-    case .escaped(.notWordBoundary): return .notWordBoundary
-    case .escaped(.startOfSubject):  return .startOfSubject
-    case .escaped(.endOfSubject):    return .endOfSubject
-    case .escaped(.textSegment):     return .textSegment
-    case .escaped(.notTextSegment):  return .notTextSegment
-    case .escaped(.endOfSubjectBeforeNewline):
-      return .endOfSubjectBeforeNewline
-    case .escaped(.firstMatchingPositionInSubject):
-      return .firstMatchingPositionInSubject
-
-    case .escaped(.resetStartOfMatch): return .resetStartOfMatch
-
-    default: return nil
-    }
-  }
-}
-
 extension AST.Atom {
   public enum Callout: Hashable {
     /// A PCRE callout written `(?C...)`

Sources/_RegexParser/Regex/Printing/PrintAsCanonical.swift

@@ -237,9 +237,6 @@ extension AST.Atom.Number {
 
 extension AST.Atom {
   var _canonicalBase: String {
-    if let anchor = self.assertionKind {
-      return anchor.rawValue
-    }
     if let lit = self.literalStringValue {
       // FIXME: We may have to re-introduce escapes
       // For example, `\.` will come back as "." instead
@@ -248,6 +245,10 @@ extension AST.Atom {
       return lit
     }
     switch self.kind {
+    case .caretAnchor:
+      return "^"
+    case .dollarAnchor:
+      return "$"
     case .escaped(let e):
       return "\\\(e.character)"
     case .backreference(let br):

Sources/_StringProcessing/ByteCodeGen.swift

@@ -68,7 +68,7 @@ fileprivate extension Compiler.ByteCodeGen {
       try emitScalar(s)
 
     case let .assertion(kind):
-      try emitAssertion(kind.ast)
+      try emitAssertion(kind)
 
     case let .backreference(ref):
       try emitBackreference(ref.ast)
@@ -114,7 +114,7 @@ fileprivate extension Compiler.ByteCodeGen {
   }
 
   mutating func emitAssertion(
-    _ kind: AST.Atom.AssertionKind
+    _ kind: DSLTree.Atom.Assertion
   ) throws {
     // FIXME: Depends on API model we have... We may want to
     // think through some of these with API interactions in mind

Sources/_StringProcessing/PrintAsPattern.swift

@@ -623,7 +623,7 @@ extension String {
   }
 }
 
-extension AST.Atom.AssertionKind {
+extension DSLTree.Atom.Assertion {
   // TODO: Some way to integrate this with conversion...
   var _patternBase: String {
     switch self {
@@ -811,7 +811,7 @@ extension AST.Atom {
   ///
   /// TODO: Some way to integrate this with conversion...
   var _patternBase: (String, canBeWrapped: Bool) {
-    if let anchor = self.assertionKind {
+    if let anchor = self.dslAssertionKind {
       return (anchor._patternBase, false)
     }
 
@@ -1124,7 +1124,7 @@ extension DSLTree.Atom {
       }
 
     case .assertion(let a):
-      return (a.ast._patternBase, false)
+      return (a._patternBase, false)
 
     case .backreference(_):
       return ("/* TOOD: backreferences */", false)

Sources/_StringProcessing/Regex/ASTConversion.swift

@@ -208,10 +208,38 @@ extension AST.CustomCharacterClass {
   }
 }
 
+extension AST.Atom.EscapedBuiltin {
+  var dslAssertionKind: DSLTree.Atom.Assertion? {
+    switch self {
+    case .wordBoundary:                   return .wordBoundary
+    case .notWordBoundary:                return .notWordBoundary
+    case .startOfSubject:                 return .startOfSubject
+    case .endOfSubject:                   return .endOfSubject
+    case .textSegment:                    return .textSegment
+    case .notTextSegment:                 return .notTextSegment
+    case .endOfSubjectBeforeNewline:      return .endOfSubjectBeforeNewline
+    case .firstMatchingPositionInSubject: return .firstMatchingPositionInSubject
+    case .resetStartOfMatch:              return .resetStartOfMatch
+    default: return nil
+    }
+  }
+}
+
+extension AST.Atom {
+  var dslAssertionKind: DSLTree.Atom.Assertion? {
+    switch kind {
+    case .caretAnchor:    return .caretAnchor
+    case .dollarAnchor:   return .dollarAnchor
+    case .escaped(let b): return b.dslAssertionKind
+    default: return nil
+    }
+  }
+}
+
 extension AST.Atom {
   var dslTreeAtom: DSLTree.Atom {
-    if let kind = assertionKind {
-      return .assertion(.init(ast: kind))
+    if let kind = dslAssertionKind {
+      return .assertion(kind)
     }
 
     switch self.kind {

Sources/_StringProcessing/Regex/DSLTree.swift

@@ -253,7 +253,7 @@ extension DSLTree {
     /// newlines unless single line mode is enabled.
     case dot
 
-    case assertion(_AST.AssertionKind)
+    case assertion(Assertion)
     case backreference(_AST.Reference)
     case symbolicReference(ReferenceID)
 
@@ -263,6 +263,44 @@ extension DSLTree {
   }
 }
 
+extension DSLTree.Atom {
+  @_spi(RegexBuilder)
+  public enum Assertion: Hashable {
+    /// \A
+    case startOfSubject
+
+    /// \Z
+    case endOfSubjectBeforeNewline
+
+    /// \z
+    case endOfSubject
+
+    /// \K
+    case resetStartOfMatch
+
+    /// \G
+    case firstMatchingPositionInSubject
+
+    /// \y
+    case textSegment
+
+    /// \Y
+    case notTextSegment
+
+    /// ^
+    case caretAnchor
+
+    /// $
+    case dollarAnchor
+
+    /// \b (from outside a custom character class)
+    case wordBoundary
+
+    /// \B
+    case notWordBoundary
+  }
+}
+
 extension Unicode.GeneralCategory {
   var extendedGeneralCategory: Unicode.ExtendedGeneralCategory? {
     switch self {
@@ -779,40 +817,6 @@ extension DSLTree {
       internal var ast: AST.AbsentFunction
     }
 
-    @_spi(RegexBuilder)
-    public struct AssertionKind {
-      internal var ast: AST.Atom.AssertionKind
-      
-      public static func startOfSubject(_ inverted: Bool = false) -> Self {
-        .init(ast: .startOfSubject)
-      }
-      public static func endOfSubjectBeforeNewline(_ inverted: Bool = false) -> Self {
-        .init(ast: .endOfSubjectBeforeNewline)
-      }
-      public static func endOfSubject(_ inverted: Bool = false) -> Self {
-        .init(ast: .endOfSubject)
-      }
-      public static func firstMatchingPositionInSubject(_ inverted: Bool = false) -> Self {
-        .init(ast: .firstMatchingPositionInSubject)
-      }
-      public static func textSegmentBoundary(_ inverted: Bool = false) -> Self {
-        inverted
-          ? .init(ast: .notTextSegment)
-          : .init(ast: .textSegment)
-      }
-      public static func startOfLine(_ inverted: Bool = false) -> Self {
-        .init(ast: .caretAnchor)
-      }
-      public static func endOfLine(_ inverted: Bool = false) -> Self {
-        .init(ast: .dollarAnchor)
-      }
-      public static func wordBoundary(_ inverted: Bool = false) -> Self {
-        inverted
-          ? .init(ast: .notWordBoundary)
-          : .init(ast: .wordBoundary)
-      }
-    }
-    
     @_spi(RegexBuilder)
     public struct Reference {
       internal var ast: AST.Reference

Sources/_StringProcessing/Utility/RegexFactory.swift

@@ -40,7 +40,7 @@ public struct _RegexFactory {
   @_spi(RegexBuilder)
   @available(SwiftStdlib 5.7, *)
   public func assertion<Output>(
-    _ kind: DSLTree._AST.AssertionKind
+    _ kind: DSLTree.Atom.Assertion
   ) -> Regex<Output> {
     .init(node: .atom(.assertion(kind)))
   }