Parse shebang as a child of SourceFileSyntax

Author: StevenWong12

CodeGeneration/Sources/SyntaxSupport/DeclNodes.swift

@@ -1997,6 +1997,12 @@ public let DECL_NODES: [Node] = [
     parserFunction: "parseSourceFile",
     traits: ["WithStatements"],
     children: [
+      Child(
+        name: "Shebang",
+        kind: .token(choices: [.token(.shebang)]),
+        documentation: "A shebang can specify the path of the compiler when using Swift source file as a script.",
+        isOptional: true
+      ),
       Child(
         name: "Statements",
         kind: .collection(kind: .codeBlockItemList, collectionElementName: "Statement")

CodeGeneration/Sources/SyntaxSupport/TokenSpec.swift

@@ -113,6 +113,7 @@ public enum Token: CaseIterable {
   case rightParen
   case rightSquare
   case semicolon
+  case shebang
   case singleQuote
   case stringQuote
   case stringSegment
@@ -209,6 +210,8 @@ public enum Token: CaseIterable {
       return .punctuator(name: "rightSquare", text: "]")
     case .semicolon:
       return .punctuator(name: "semicolon", text: ";")
+    case .shebang:
+      return .other(name: "shebang", nameForDiagnostics: "shebang")
     case .singleQuote:
       return .punctuator(name: "singleQuote", text: "\'")
     case .stringQuote:

CodeGeneration/Sources/SyntaxSupport/Trivia.swift

@@ -159,11 +159,6 @@ public let TRIVIAS: [Trivia] = [
     ]
   ),
 
-  Trivia(
-    name: "Shebang",
-    comment: #"A script command, starting with '#!'."#
-  ),
-
   Trivia(
     name: "Space",
     comment: #"A space ' ' character."#,

Sources/SwiftIDEUtils/SyntaxClassification.swift

@@ -193,6 +193,8 @@ extension RawTokenKind {
       return .none
     case .semicolon:
       return .none
+    case .shebang:
+      return .none
     case .singleQuote:
       return .stringLiteral
     case .stringQuote:

Sources/SwiftParser/Lexer/Cursor.swift

@@ -905,6 +905,11 @@ extension Lexer.Cursor {
     case UInt8(ascii: "\\"): _ = self.advance(); return Lexer.Result(.backslash)
 
     case UInt8(ascii: "#"):
+      // Try lex shebang.
+      if self.isAtStartOfFile, self.peek(at: 1) == UInt8(ascii: "!") {
+        self.advanceToEndOfLine()
+        return Lexer.Result(.shebang)
+      }
       // Try lex a raw string literal.
       if let delimiterLength = self.advanceIfOpeningRawStringDelimiter() {
         return Lexer.Result(.rawStringPoundDelimiter, stateTransition: .push(newState: .afterRawStringDelimiter(delimiterLength: delimiterLength)))
@@ -1171,12 +1176,6 @@ extension Lexer.Cursor {
         default:
           break
         }
-      case UInt8(ascii: "#"):
-        guard start.isAtStartOfFile, self.advance(if: { $0 == "!" }) else {
-          break
-        }
-        self.advanceToEndOfLine()
-        continue
       case UInt8(ascii: "<"), UInt8(ascii: ">"):
         if self.tryLexConflictMarker(start: start) {
           error = LexingDiagnostic(.sourceConflictMarker, position: start)
@@ -1189,6 +1188,7 @@ extension Lexer.Cursor {
         UInt8(ascii: "}"), UInt8(ascii: "]"), UInt8(ascii: ")"),
         UInt8(ascii: "@"), UInt8(ascii: ","), UInt8(ascii: ";"),
         UInt8(ascii: ":"), UInt8(ascii: "\\"), UInt8(ascii: "$"),
+        UInt8(ascii: "#"),
 
         // Start of integer/hex/float literals.
         UInt8(ascii: "0"), UInt8(ascii: "1"), UInt8(ascii: "2"),

Sources/SwiftParser/Lexer/RegexLiteralLexer.swift

@@ -625,6 +625,10 @@ extension Lexer.Cursor {
     case .arrow, .ellipsis, .period, .atSign, .pound, .backtick, .backslash:
       return false
 
+    // Shebang does not sequence expressions.
+    case .shebang:
+      return false
+
     case .keyword:
       // There are a handful of keywords that are expressions, handle them.
       // Otherwise, a regex literal can generally be parsed after a keyword.

Sources/SwiftParser/TokenPrecedence.swift

@@ -124,8 +124,8 @@ enum TokenPrecedence: Comparable {
       .dollarIdentifier, .identifier,
       // '_' can occur in types to replace a type identifier
       .wildcard,
-      // String segment, string interpolation anchor, pound, and regex pattern don't really fit anywhere else
-      .pound, .stringSegment, .regexLiteralPattern:
+      // String segment, string interpolation anchor, pound, shebang and regex pattern don't really fit anywhere else
+      .pound, .stringSegment, .regexLiteralPattern, .shebang:
       self = .identifierLike
 
     // MARK: Expr keyword

Sources/SwiftParser/TopLevel.swift

@@ -40,10 +40,12 @@ extension Parser {
   /// ``Parser/parse(source:parseTransition:filenameForDiagnostics:languageVersion:enableBareSlashRegexLiteral:)-7tndx``
   /// API calls.
   mutating func parseSourceFile() -> RawSourceFileSyntax {
+    let shebang = self.consume(if: .shebang)
     let items = self.parseTopLevelCodeBlockItems()
     let unexpectedBeforeEndOfFileToken = consumeRemainingTokens()
     let endOfFile = self.consume(if: .endOfFile)!
     return .init(
+      shebang: shebang,
       statements: items,
       RawUnexpectedNodesSyntax(unexpectedBeforeEndOfFileToken, arena: self.arena),
       endOfFileToken: endOfFile,

Sources/SwiftParser/TriviaParser.swift

@@ -81,15 +81,6 @@ public struct TriviaParser {
         }
 
       case UInt8(ascii: "#"):
-        // "#!...": .shebang
-        // NOTE: .shebang appears only if this trivia is at the start of the
-        // file. We don't know if this trivia is at the start of the file, but
-        // we believe that the lexer lexed it accordingly.
-        if position == .leading && pieces.isEmpty && cursor.advance(if: { $0 == "!" }) {
-          cursor.advanceToEndOfLine()
-          pieces.append(.shebang(start.text(upTo: cursor)))
-          continue
-        }
         cursor.advance(while: { $0 == "#" })
         pieces.append(.pounds(start.distance(to: cursor)))
         continue

Sources/SwiftParser/generated/TokenSpecStaticMembers.swift

@@ -187,6 +187,10 @@ extension TokenSpec {
     return TokenSpec(.semicolon)
   }
 
+  static var shebang: TokenSpec {
+    return TokenSpec(.shebang)
+  }
+  
   static var singleQuote: TokenSpec {
     return TokenSpec(.singleQuote)
   }