Reject .NET subtraction and quote range operands

Tighten up validation of character class range
operands such that we reject quotes and custom
character classes. This includes rejecting syntax
that would be a subtraction in .NET. We throw a
custom error that suggests using `--` instead.

Author: hamishknight

Sources/_RegexParser/Regex/AST/CustomCharClass.swift

@@ -62,6 +62,10 @@ extension AST {
         self.rhs = rhs
         self.trivia = trivia
       }
+
+      public var location: SourceLocation {
+        lhs.location.union(with: rhs.location)
+      }
     }
     public enum SetOp: String, Hashable {
       case subtraction = "--"
@@ -108,6 +112,25 @@ extension CustomCC.Member {
   public var isSemantic: Bool {
     !isTrivia
   }
+
+  public var location: SourceLocation {
+    switch self {
+    case let .custom(c): return c.location
+    case let .range(r):  return r.location
+    case let .atom(a):   return a.location
+    case let .quote(q):  return q.location
+    case let .trivia(t): return t.location
+    case let .setOperation(lhs, dash, rhs):
+      var loc = dash.location
+      if let lhs = lhs.first {
+        loc = loc.union(with: lhs.location)
+      }
+      if let rhs = rhs.last {
+        loc = loc.union(with: rhs.location)
+      }
+      return loc
+    }
+  }
 }
 
 extension AST.CustomCharacterClass {

Sources/_RegexParser/Regex/Parse/Diagnostics.swift

@@ -94,6 +94,7 @@ enum ParseError: Error, Hashable {
   case invalidNamedReference(String)
   case duplicateNamedCapture(String)
   case invalidCharacterClassRangeOperand
+  case unsupportedDotNetSubtraction
   case invalidQuantifierRange(Int, Int)
   case invalidCharacterRange(from: Character, to: Character)
   case notQuantifiable
@@ -174,7 +175,9 @@ extension ParseError: CustomStringConvertible {
     case .expectedCustomCharacterClassMembers:
       return "expected custom character class members"
     case .invalidCharacterClassRangeOperand:
-      return "invalid character class range"
+      return "invalid bound for character class range"
+    case .unsupportedDotNetSubtraction:
+      return "subtraction with '-' is unsupported; use '--' instead"
     case .emptyProperty:
       return "empty property"
     case .unknownProperty(let key, let value):

Sources/_RegexParser/Regex/Parse/LexicalAnalysis.swift

@@ -1245,6 +1245,25 @@ extension Source {
     return nil
   }
 
+  /// Check to see if we can lex a .NET subtraction. Returns the
+  /// location of the `-`.
+  ///
+  ///     DotNetSubtraction -> Trivia* '-' Trivia* CustomCharClass
+  ///
+  func canLexDotNetCharClassSubtraction(
+    context: ParsingContext
+  ) -> SourceLocation? {
+    lookahead { src in
+      // We can lex '-' as a .NET subtraction if it precedes a custom character
+      // class.
+      while (try? src.lexTrivia(context: context)) != nil {}
+      guard let dashLoc = src.tryEatWithLoc("-") else { return nil }
+      while (try? src.lexTrivia(context: context)) != nil {}
+      guard src.lexCustomCCStart() != nil else { return nil }
+      return dashLoc
+    }
+  }
+
   private mutating func lexPOSIXCharacterProperty(
   ) throws -> Located<AST.Atom.CharacterProperty>? {
     try recordLoc { src in

Sources/_RegexParser/Regex/Parse/Parse.swift

@@ -557,7 +557,7 @@ extension Parser {
   mutating func parsePotentialCCRange(
     into members: inout [CustomCC.Member]
   ) throws {
-    guard case .atom(let lhs)? = members.last else { return }
+    guard let lhs = members.last, lhs.isSemantic else { return }
 
     // Try and see if we can parse a character class range. Each time we parse
     // a component of the range, we append to `members` in case it ends up not
@@ -580,15 +580,49 @@ extension Parser {
     guard let rhs = try parseCCCMember() else { return }
     members.append(rhs)
 
-    guard case let .atom(rhs) = rhs else { return }
+    func makeOperand(_ m: CustomCC.Member, isLHS: Bool) throws -> AST.Atom {
+      switch m {
+      case .atom(let a):
+        return a
+      case .custom:
+        // Not supported. While .NET allows `x-[...]` to spell subtraction, we
+        // require `x--[...]`. We also ban `[...]-x` for consistency.
+        if isLHS {
+          throw Source.LocatedError(
+            ParseError.invalidCharacterClassRangeOperand, m.location)
+        } else {
+          throw Source.LocatedError(
+            ParseError.unsupportedDotNetSubtraction, m.location)
+        }
+      case .quote:
+        // Currently unsupported, we need to figure out what the semantics
+        // would be for grapheme/scalar modes.
+        throw Source.LocatedError(
+          ParseError.unsupported("range with quoted sequence"), m.location)
+      case .trivia:
+        throw Unreachable("Should have been lexed separately")
+      case .range, .setOperation:
+        throw Unreachable("Parsed later")
+      }
+    }
+    let lhsOp = try makeOperand(lhs, isLHS: true)
+    let rhsOp = try makeOperand(rhs, isLHS: false)
 
     // We've successfully parsed an atom LHS and RHS, so form a range,
     // collecting the trivia we've parsed, and replacing the members that
     // would have otherwise been added to the custom character class.
     let rangeMemberCount = members.count - membersBeforeRange
     let trivia = members.suffix(rangeMemberCount).compactMap(\.asTrivia)
     members.removeLast(rangeMemberCount)
-    members.append(.range(.init(lhs, dash, rhs, trivia: trivia)))
+    members.append(.range(.init(lhsOp, dash, rhsOp, trivia: trivia)))
+
+    // We need to specially check if we can lex a .NET character class
+    // subtraction here as e.g `[a-c-[...]]` is allowed in .NET. Otherwise we'd
+    // treat the second `-` as literal.
+    if let dashLoc = source.canLexDotNetCharClassSubtraction(context: context) {
+      throw Source.LocatedError(
+        ParseError.unsupportedDotNetSubtraction, dashLoc)
+    }
   }
 
   mutating func parseCCCMembers(

Tests/RegexTests/ParseTests.swift

@@ -515,10 +515,36 @@ extension RegexTests {
 
     parseTest(
       "[a-b-c]", charClass(range_m("a", "b"), "-", "c"))
+    parseTest(
+      "[a-b-c-d]", charClass(range_m("a", "b"), "-", range_m("c", "d")))
+
+    parseTest("[a-c---]", charClass(
+      setOp(range_m("a", "c"), op: .subtraction, "-")
+    ))
+
+    parseTest("(?x)[a-c -- -]", concat(
+      changeMatchingOptions(matchingOptions(adding: .extended)),
+      charClass(setOp(range_m("a", "c"), op: .subtraction, "-"))
+    ))
+
+    parseTest("(?x)[a-c - - -]", concat(
+      changeMatchingOptions(matchingOptions(adding: .extended)),
+      charClass(range_m("a", "c"), range_m("-", "-"))
+    ))
 
     parseTest("[-a-]", charClass("-", "a", "-"))
     parseTest("[[a]-]", charClass(charClass("a"), "-"))
-    parseTest("[[a]-b]", charClass(charClass("a"), "-", "b"))
+    parseTest("[-[a]]", charClass("-", charClass("a")))
+
+    parseTest(#"(?x)[ -[b]]"#, concat(
+      changeMatchingOptions(matchingOptions(adding: .extended)),
+      charClass("-", charClass("b"))
+    ))
+
+    parseTest(#"[ - [ ]]"#, charClass(range_m(" ", " "), charClass(" ")))
+    parseTest(#"[ - [ ] ]"#, charClass(range_m(" ", " "), charClass(" "), " "))
+
+    parseTest(#"[a-c-\Qd\E]"#, charClass(range_m("a", "c"), "-", quote_m("d")))
 
     parseTest("[a-z]", charClass(range_m("a", "z")))
     parseTest("[a-a]", charClass(range_m("a", "a")))
@@ -2688,6 +2714,32 @@ extension RegexTests {
     diagnosticTest("[[:=:]]", .emptyProperty)
 
     diagnosticTest(#"|([\d-c])?"#, .invalidCharacterClassRangeOperand)
+    diagnosticTest("[[a]-b]", .invalidCharacterClassRangeOperand)
+
+    // .NET subtraction is banned, we require explicit '--'.
+    diagnosticTest("[a-[b]]", .unsupportedDotNetSubtraction)
+    diagnosticTest(#"[abc-[def]]"#, .unsupportedDotNetSubtraction)
+    diagnosticTest(#"[abc-[^def]]"#, .unsupportedDotNetSubtraction)
+    diagnosticTest(#"[\d\u{0}[a]-[b-[c]]]"#, .unsupportedDotNetSubtraction)
+    diagnosticTest("[a-z-[d-w-[m-o]]]", .unsupportedDotNetSubtraction)
+    diagnosticTest(#"[a-[:b]]"#, .unsupportedDotNetSubtraction)
+    diagnosticTest(#"[[a]-[b]]"#, .invalidCharacterClassRangeOperand)
+    diagnosticTest(#"[ -[ ]]"#, .unsupportedDotNetSubtraction)
+    diagnosticTest(#"(?x)[a  -  [b]  ]"#, .unsupportedDotNetSubtraction)
+
+    diagnosticTest(#"[a-[]]"#, .expectedCustomCharacterClassMembers)
+    diagnosticTest(#"[-[]]"#, .expectedCustomCharacterClassMembers)
+    diagnosticTest(#"(?x)[ - [ ] ]"#, .expectedCustomCharacterClassMembers)
+    diagnosticTest(#"(?x)[a-[ ] ]"#, .expectedCustomCharacterClassMembers)
+    diagnosticTest(#"[a-[:digit:]]"#, .invalidCharacterClassRangeOperand)
+
+    diagnosticTest("[--]", .expectedCustomCharacterClassMembers)
+    diagnosticTest("[---]", .expectedCustomCharacterClassMembers)
+    diagnosticTest("[----]", .expectedCustomCharacterClassMembers)
+
+    // Quoted sequences aren't currently supported as range operands.
+    diagnosticTest(#"[a-\Qbc\E]"#, .unsupported("range with quoted sequence"))
+    diagnosticTest(#"[\Qbc\E-de]"#, .unsupported("range with quoted sequence"))
 
     diagnosticTest(#"[_-A]"#, .invalidCharacterRange(from: "_", to: "A"))
     diagnosticTest(#"(?i)[_-A]"#, .invalidCharacterRange(from: "_", to: "A"))
@@ -2873,6 +2925,17 @@ extension RegexTests {
       /#
       """#, .quoteMayNotSpanMultipleLines)
 
+    // .NET subtraction
+    diagnosticWithDelimitersTest(#"""
+      #/
+      [
+        a # interesting
+        -   #a
+         [ b] # comment
+        ]
+      /#
+      """#, .unsupportedDotNetSubtraction)
+
     // MARK: Group specifiers
 
     diagnosticTest(#"(*"#, .unknownGroupKind("*"))