Cleanup

Sources/_StringProcessing/ByteCodeGen.swift

@@ -82,7 +82,10 @@ fileprivate extension Compiler.ByteCodeGen {
     case let .unconverted(astAtom):
       if optimizationsEnabled,
          let cc = astAtom.ast.characterClass?.builtinCC {
-        emitBuiltinCharacterClass(cc)
+        builder.buildMatchBuiltin(
+          cc,
+          cc.isStrict(options: options),
+          isScalar: options.semanticLevel == .unicodeScalar)
         return
       }
       if let consumer = try astAtom.ast.generateConsumer(options) {
@@ -114,15 +117,6 @@ fileprivate extension Compiler.ByteCodeGen {
       throw Unsupported("Backreference kind: \(ref)")
     }
   }
-
-  mutating func emitBuiltinCharacterClass(
-    _ cc: BuiltinCC
-  ) {
-    builder.buildMatchBuiltin(
-      cc,
-      cc.isStrict(options: options),
-      isScalar: options.semanticLevel == .unicodeScalar)
-  }
 
   mutating func emitAssertion(
     _ kind: AST.Atom.AssertionKind

Sources/_StringProcessing/Compiler.swift

@@ -68,6 +68,7 @@ func _compileRegex(
 ) throws -> Executor {
   let ast = try parse(regex, syntax)
   let dsl: DSLTree
+
   switch semanticLevel?.base {
   case .graphemeCluster:
     let sequence = AST.MatchingOptionSequence(adding: [.init(.graphemeClusterSemantics, location: .fake)])

Sources/_StringProcessing/Engine/Instruction.swift

@@ -106,10 +106,6 @@ extension Instruction {
     /// Operand: Ascii bitset register containing the bitset
     case matchBitset
 
-    /// TODO: builtin assertions and anchors
-    case builtinAssertion
-
-    /// TODO: builtin character classes
     case matchBuiltin
 
     // MARK: Extension points

Sources/_StringProcessing/Engine/MEBuilder.swift

@@ -179,12 +179,6 @@ extension MEProgram.Builder {
       .consumeBy, .init(consumer: makeConsumeFunction(p))))
   }
 
-//  mutating func buildAssert(
-//    by p: @escaping MEProgram.AssertionFunction
-//  ) {
-//    instructions.append(.init(
-//      .assertBy, .init(assertion: makeAssertionFunction(p))))
-//  }
   mutating func buildAssert(
     by kind: AST.Atom.AssertionKind,
     _ anchorsMatchNewlines: Bool,

Sources/_StringProcessing/Engine/Processor.swift

@@ -262,8 +262,6 @@ extension Processor {
     var matched: Bool
     var next = input.index(after: currentPosition)
     switch cc {
-      // lily note: when do these `any` cases appear? can they be compiled
-      // into consume instructions at compile time?
     case .any, .anyGrapheme: matched = true
     case .anyScalar:
       matched = true
@@ -320,8 +318,6 @@ extension Processor {
       matched = c.isNewline && (c.isASCII || !isStrictAscii)
     case .newlineSequence:
       matched = c.isNewline && (c.isASCII || !isStrictAscii)
-      // lily note: what exactly is this doing? matching a full cr-lf character
-      // even though its in scalar mode? why?
       if c == "\r" && next != input.endIndex && input.unicodeScalars[next] == "\n" {
         input.unicodeScalars.formIndex(after: &next)
       }
@@ -426,7 +422,6 @@ extension Processor {
         if usesSimpleUnicodeBoundaries {
           // TODO: How should we handle bounds?
           return atSimpleBoundary(usesASCIIWord, semanticLevel)
-          // lily note: there appear to be no test cases that use this option, ping alex to ask what they should look like
         } else {
           return input.isOnWordBoundary(at: currentPosition, using: &wordIndexCache, &wordIndexMaxIndex)
         }
@@ -716,9 +711,6 @@ extension Processor {
       storedCaptures[capNum].registerValue(
         value, overwriteInitial: sp)
       controller.step()
-
-    case .builtinAssertion:
-      builtinAssertion()
     }
   }
 }

Sources/_StringProcessing/_CharacterClassModel.swift

@@ -564,35 +564,6 @@ extension DSLTree.CustomCharacterClass {
   }
 }
 
-extension _CharacterClassModel {
-  // FIXME: Calling on inverted sets wont be the same as the
-  // inverse of a boundary if at the start or end of the
-  // string. (Think through what we want: do it ourselves or
-  // give the caller both options).
-  func isBoundary(
-    _ input: String,
-    at pos: String.Index,
-    bounds: Range<String.Index>,
-    with options: MatchingOptions
-  ) -> Bool {
-    // FIXME: How should we handle bounds?
-    // We probably need two concepts
-    if bounds.isEmpty { return false }
-    if pos == bounds.lowerBound {
-      return self.matches(in: input, at: pos, with: options) != nil
-    }
-    let priorIdx = input.index(before: pos)
-    if pos == bounds.upperBound {
-      return self.matches(in: input, at: priorIdx, with: options) != nil
-    }
-
-    let prior = self.matches(in: input, at: priorIdx, with: options) != nil
-    let current = self.matches(in: input, at: pos, with: options) != nil
-    return prior != current
-  }
-
-}
-
 internal enum BuiltinCC: UInt64 {
   case any = 1
   case anyGrapheme
@@ -622,11 +593,9 @@ extension BuiltinCC {
 
 extension _CharacterClassModel {
   internal var builtinCC: BuiltinCC? {
-    if isInverted { return nil } // lily todo: add another flag to the payload? when is this set? why are there so many weird edge cases in ccm? it feels like it's trying to model both builtins and custom models
-
-    // in that case, should we just convert a ccm to a ccc
-    // if it has these weird flags set?
-    // completely remove ccm from compilation and just emit either a builtincc or a ccc or an advance
+    // Future work: Make CCM always either a BuiltinCC or convertable to a
+    // custom character class
+    if isInverted { return nil }
     switch self.cc {
     case .any:
         return .any