Make the Regex type Sendable (apple#457)

This change makes `Regex`, `RegexComponent`, and its component types `Sendable`.

Regex stores a `Program` instance, which lazily lowers the DSLTree into a 
compiled program. Without synchronization, this lazy compilation is unsafe 
under concurrency. This change uses atomic initialization for the compiled 
program.

Package.swift

@@ -62,7 +62,10 @@ let package = Package(
             dependencies: []),
         .target(
             name: "_StringProcessing",
-            dependencies: ["_RegexParser", "_CUnicode"],
+            dependencies: [
+              "_RegexParser",
+              "_CUnicode",
+            ],
             swiftSettings: publicStdlibSettings),
         .target(
             name: "RegexBuilder",

Sources/_RegexParser/Regex/AST/AST.swift

@@ -30,7 +30,7 @@ extension AST {
 extension AST {
   /// A node in the regex AST.
   public indirect enum Node:
-    Hashable, _TreeNode //, _ASTPrintable ASTValue, ASTAction
+    Hashable, _TreeNode, Sendable //, _ASTPrintable ASTValue, ASTAction
   {
     /// ... | ... | ...
     case alternation(Alternation)
@@ -143,7 +143,7 @@ extension AST.Node {
 
 extension AST {
 
-  public struct Alternation: Hashable, _ASTNode {
+  public struct Alternation: Hashable, Sendable, _ASTNode {
     public let children: [AST.Node]
     public let pipes: [SourceLocation]
 
@@ -162,7 +162,7 @@ extension AST {
     }
   }
 
-  public struct Concatenation: Hashable, _ASTNode {
+  public struct Concatenation: Hashable, Sendable, _ASTNode {
     public let children: [AST.Node]
     public let location: SourceLocation
 
@@ -172,7 +172,7 @@ extension AST {
     }
   }
 
-  public struct Quote: Hashable, _ASTNode {
+  public struct Quote: Hashable, Sendable, _ASTNode {
     public let literal: String
     public let location: SourceLocation
 
@@ -182,7 +182,7 @@ extension AST {
     }
   }
 
-  public struct Trivia: Hashable, _ASTNode {
+  public struct Trivia: Hashable, Sendable, _ASTNode {
     public let contents: String
     public let location: SourceLocation
 
@@ -197,7 +197,7 @@ extension AST {
     }
   }
 
-  public struct Interpolation: Hashable, _ASTNode {
+  public struct Interpolation: Hashable, Sendable, _ASTNode {
     public let contents: String
     public let location: SourceLocation
 
@@ -207,7 +207,7 @@ extension AST {
     }
   }
 
-  public struct Empty: Hashable, _ASTNode {
+  public struct Empty: Hashable, Sendable, _ASTNode {
     public let location: SourceLocation
 
     public init(_ location: SourceLocation) {
@@ -219,15 +219,15 @@ extension AST {
   ///
   /// This is used to model a pattern which should
   /// not be matched against across varying scopes.
-  public struct AbsentFunction: Hashable, _ASTNode {
-    public enum Start: Hashable {
+  public struct AbsentFunction: Hashable, Sendable, _ASTNode {
+    public enum Start: Hashable, Sendable {
       /// `(?~|`
       case withPipe
 
       /// `(?~`
       case withoutPipe
     }
-    public enum Kind: Hashable {
+    public enum Kind: Hashable, Sendable {
       /// An absent repeater `(?~absent)`. This is equivalent to `(?~|absent|.*)`
       /// and therefore matches as long as the pattern `absent` is not matched.
       case repeater(AST.Node)
@@ -261,8 +261,8 @@ extension AST {
     }
   }
 
-  public struct Reference: Hashable {
-    public enum Kind: Hashable {
+  public struct Reference: Hashable, Sendable {
+    public enum Kind: Hashable, Sendable {
       // \n \gn \g{n} \g<n> \g'n' (?n) (?(n)...
       // Oniguruma: \k<n>, \k'n'
       case absolute(Int)
@@ -304,7 +304,7 @@ extension AST {
   }
 
   /// A set of global matching options in a regular expression literal.
-  public struct GlobalMatchingOptionSequence: Hashable {
+  public struct GlobalMatchingOptionSequence: Hashable, Sendable {
     public var options: [AST.GlobalMatchingOption]
 
     public init?(_ options: [AST.GlobalMatchingOption]) {

Sources/_RegexParser/Regex/AST/Atom.swift

@@ -10,7 +10,7 @@
 //===----------------------------------------------------------------------===//
 
 extension AST {
-  public struct Atom: Hashable, _ASTNode {
+  public struct Atom: Hashable, Sendable, _ASTNode {
     public let kind: Kind
     public let location: SourceLocation
 
@@ -19,7 +19,7 @@ extension AST {
       self.location = loc
     }
 
-    public enum Kind: Hashable {
+    public enum Kind: Hashable, Sendable {
       /// Just a character
       ///
       /// A, \*, \\, ...
@@ -113,7 +113,7 @@ extension AST.Atom {
 }
 
 extension AST.Atom {
-  public struct Scalar: Hashable {
+  public struct Scalar: Hashable, Sendable {
     public var value: UnicodeScalar
     public var location: SourceLocation
 
@@ -123,7 +123,7 @@ extension AST.Atom {
     }
   }
 
-  public struct ScalarSequence: Hashable {
+  public struct ScalarSequence: Hashable, Sendable {
     public var scalars: [Scalar]
     public var trivia: [AST.Trivia]
 
@@ -145,7 +145,7 @@ extension AST.Atom {
 
   // Characters, character types, literals, etc., derived from
   // an escape sequence.
-  public enum EscapedBuiltin: Hashable {
+  public enum EscapedBuiltin: Hashable, Sendable {
     // TODO: better doc comments
 
     // Literal single characters
@@ -374,7 +374,7 @@ extension AST.Atom.EscapedBuiltin {
 }
 
 extension AST.Atom {
-  public struct CharacterProperty: Hashable {
+  public struct CharacterProperty: Hashable, Sendable {
     public var kind: Kind
 
     /// Whether this is an inverted property e.g '\P{Ll}', '[:^ascii:]'.
@@ -397,7 +397,7 @@ extension AST.Atom {
 }
 
 extension AST.Atom.CharacterProperty {
-  public enum Kind: Hashable {
+  public enum Kind: Hashable, Sendable {
     /// Matches any character, equivalent to Oniguruma's '\O'.
     case any
 
@@ -453,14 +453,14 @@ extension AST.Atom.CharacterProperty {
     /// Some special properties implemented by Java.
     case javaSpecial(JavaSpecial)
 
-    public enum MapKind: Hashable {
+    public enum MapKind: Hashable, Sendable {
       case lowercase
       case uppercase
       case titlecase
     }
   }
 
-  public enum PCRESpecialCategory: String, Hashable {
+  public enum PCRESpecialCategory: String, Hashable, Sendable {
     case alphanumeric     = "Xan"
     case posixSpace       = "Xps"
     case perlSpace        = "Xsp"
@@ -470,7 +470,7 @@ extension AST.Atom.CharacterProperty {
 
   /// Special Java properties that correspond to methods on
   /// `java.lang.Character`, with the `java` prefix replaced by `is`.
-  public enum JavaSpecial: String, Hashable, CaseIterable {
+  public enum JavaSpecial: String, Hashable, CaseIterable, Sendable {
     case alphabetic             = "javaAlphabetic"
     case defined                = "javaDefined"
     case digit                  = "javaDigit"
@@ -494,7 +494,7 @@ extension AST.Atom.CharacterProperty {
 
 extension AST.Atom {
   /// Anchors and other built-in zero-width assertions.
-  public enum AssertionKind: String {
+  public enum AssertionKind: String, Hashable, Sendable {
     /// \A
     case startOfSubject = #"\A"#
 
@@ -554,10 +554,10 @@ extension AST.Atom {
 }
 
 extension AST.Atom {
-  public enum Callout: Hashable {
+  public enum Callout: Hashable, Sendable {
     /// A PCRE callout written `(?C...)`
-    public struct PCRE: Hashable {
-      public enum Argument: Hashable {
+    public struct PCRE: Hashable, Sendable {
+      public enum Argument: Hashable, Sendable {
         case number(Int)
         case string(String)
       }
@@ -573,8 +573,8 @@ extension AST.Atom {
     }
 
     /// A named Oniguruma callout written `(*name[tag]{args, ...})`
-    public struct OnigurumaNamed: Hashable {
-      public struct ArgList: Hashable {
+    public struct OnigurumaNamed: Hashable, Sendable {
+      public struct ArgList: Hashable, Sendable {
         public var leftBrace: SourceLocation
         public var args: [AST.Located<String>]
         public var rightBrace: SourceLocation
@@ -604,8 +604,8 @@ extension AST.Atom {
     }
 
     /// An Oniguruma callout 'of contents', written `(?{...}[tag]D)`
-    public struct OnigurumaOfContents: Hashable {
-      public enum Direction: Hashable {
+    public struct OnigurumaOfContents: Hashable, Sendable {
+      public enum Direction: Hashable, Sendable {
         case inProgress   // > (the default)
         case inRetraction // <
         case both         // X
@@ -652,7 +652,7 @@ extension AST.Atom {
 
 extension AST.Atom.Callout {
   /// A tag specifier `[...]` that can appear in an Oniguruma callout.
-  public struct OnigurumaTag: Hashable {
+  public struct OnigurumaTag: Hashable, Sendable {
     public var leftBracket: SourceLocation
     public var name: AST.Located<String>
     public var rightBracket: SourceLocation
@@ -670,8 +670,8 @@ extension AST.Atom.Callout {
 }
 
 extension AST.Atom {
-  public struct BacktrackingDirective: Hashable {
-    public enum Kind: Hashable {
+  public struct BacktrackingDirective: Hashable, Sendable {
+    public enum Kind: Hashable, Sendable {
       /// (*ACCEPT)
       case accept
 

Sources/_RegexParser/Regex/AST/Conditional.swift

@@ -10,7 +10,7 @@
 //===----------------------------------------------------------------------===//
 
 extension AST {
-  public struct Conditional: Hashable, _ASTNode {
+  public struct Conditional: Hashable, Sendable, _ASTNode {
     public var location: SourceLocation
     public var condition: Condition
 
@@ -32,8 +32,8 @@ extension AST {
 }
 
 extension AST.Conditional {
-  public struct Condition: Hashable {
-    public enum Kind: Hashable {
+  public struct Condition: Hashable, Sendable {
+    public enum Kind: Hashable, Sendable {
       /// Check to see if a certain group was matched.
       case groupMatched(AST.Reference)
 
@@ -65,7 +65,7 @@ extension AST.Conditional {
 }
 
 extension AST.Conditional.Condition {
-  public struct PCREVersionNumber: Hashable {
+  public struct PCREVersionNumber: Hashable, Sendable {
     public var major: Int
     public var minor: Int
     public var location: SourceLocation
@@ -76,8 +76,8 @@ extension AST.Conditional.Condition {
       self.location = location
     }
   }
-  public struct PCREVersionCheck: Hashable {
-    public enum Kind: Hashable {
+  public struct PCREVersionCheck: Hashable, Sendable {
+    public enum Kind: Hashable, Sendable {
       case equal, greaterThanOrEqual
     }
     public var kind: AST.Located<Kind>

Sources/_RegexParser/Regex/AST/CustomCharClass.swift

@@ -11,7 +11,7 @@
 
 
 extension AST {
-  public struct CustomCharacterClass: Hashable {
+  public struct CustomCharacterClass: Hashable, Sendable {
     public var start: Located<Start>
     public var members: [Member]
 
@@ -27,7 +27,7 @@ extension AST {
       self.location = sr
     }
 
-    public enum Member: Hashable {
+    public enum Member: Hashable, Sendable {
       /// A nested custom character class `[[ab][cd]]`
       case custom(CustomCharacterClass)
 
@@ -47,7 +47,7 @@ extension AST {
       /// A binary operator applied to sets of members `abc&&def`
       case setOperation([Member], Located<SetOp>, [Member])
     }
-    public struct Range: Hashable {
+    public struct Range: Hashable, Sendable {
       public var lhs: Atom
       public var dashLoc: SourceLocation
       public var rhs: Atom
@@ -63,12 +63,12 @@ extension AST {
         self.trivia = trivia
       }
     }
-    public enum SetOp: String, Hashable {
+    public enum SetOp: String, Hashable, Sendable {
       case subtraction = "--"
       case intersection = "&&"
       case symmetricDifference = "~~"
     }
-    public enum Start: String {
+    public enum Start: String, Hashable, Sendable {
       case normal = "["
       case inverted = "[^"
     }

Sources/_RegexParser/Regex/AST/Group.swift

@@ -10,7 +10,7 @@
 //===----------------------------------------------------------------------===//
 
 extension AST {
-  public struct Group: Hashable {
+  public struct Group: Hashable, Sendable {
     public let kind: Located<Kind>
     public let child: AST.Node
 
@@ -24,7 +24,7 @@ extension AST {
       self.location = r
     }
 
-    public enum Kind: Hashable {
+    public enum Kind: Hashable, Sendable {
       // (...)
       case capture
 
@@ -116,7 +116,7 @@ extension AST.Group.Kind {
 }
 
 extension AST.Group {
-  public struct BalancedCapture: Hashable {
+  public struct BalancedCapture: Hashable, Sendable {
     /// The name of the group, or nil if the group has no name.
     public var name: AST.Located<String>?