Merge pull request #1964 from ahoppen/ahoppen/present-making-formatter

Author: ahoppen

Sources/SwiftBasicFormat/BasicFormat.swift

@@ -10,7 +10,7 @@
 //
 //===----------------------------------------------------------------------===//
 
-import SwiftSyntax
+@_spi(RawSyntax) import SwiftSyntax
 
 /// A rewriter that performs a "basic" format of the passed tree.
 ///
@@ -48,6 +48,14 @@ open class BasicFormat: SyntaxRewriter {
   /// been visited yet.
   private var previousToken: TokenSyntax? = nil
 
+  /// The number of ancestors that are `StringLiteralExprSyntax`.
+  private var stringLiteralNestingLevel = 0
+
+  /// Whether we are currently visiting the subtree of a `StringLiteralExprSyntax`.
+  private var isInsideStringLiteral: Bool {
+    return stringLiteralNestingLevel > 0
+  }
+
   public init(
     indentationWidth: Trivia? = nil,
     initialIndentation: Trivia = [],
@@ -86,6 +94,9 @@ open class BasicFormat: SyntaxRewriter {
   }
 
   open override func visitPre(_ node: Syntax) {
+    if node.is(StringLiteralExprSyntax.self) {
+      stringLiteralNestingLevel += 1
+    }
     if requiresIndent(node) {
       if let firstToken = node.firstToken(viewMode: viewMode),
         let tokenIndentation = firstToken.leadingTrivia.indentation(isOnNewline: false),
@@ -101,6 +112,9 @@ open class BasicFormat: SyntaxRewriter {
   }
 
   open override func visitPost(_ node: Syntax) {
+    if node.is(StringLiteralExprSyntax.self) {
+      stringLiteralNestingLevel -= 1
+    }
     if requiresIndent(node) {
       decreaseIndentationLevel()
     }
@@ -274,7 +288,7 @@ open class BasicFormat: SyntaxRewriter {
       (.keyword(.`init`), .leftParen),  // init()
       (.keyword(.self), .period),  // self.someProperty
       (.keyword(.Self), .period),  // self.someProperty
-      (.keyword(.set), .leftParen),  // var mYar: Int { set(value) {} }
+      (.keyword(.set), .leftParen),  // var mVar: Int { set(value) {} }
       (.keyword(.subscript), .leftParen),  // subscript(x: Int)
       (.keyword(.super), .period),  // super.someProperty
       (.leftBrace, .rightBrace),  // {}
@@ -351,6 +365,26 @@ open class BasicFormat: SyntaxRewriter {
     return true
   }
 
+  /// Change the text of a token during formatting.
+  ///
+  /// This allows formats to e.g. replace missing tokens by placeholder tokens.
+  ///
+  /// - Parameter token: The token whose text should be changed
+  /// - Returns: The new text or `nil` if the text should not be changed
+  open func transformTokenText(_ token: TokenSyntax) -> String? {
+    return nil
+  }
+
+  /// Change the presence of a token during formatting.
+  ///
+  /// This allows formats to e.g. replace missing tokens by placeholder tokens.
+  ///
+  /// - Parameter token: The token whose presence should be changed
+  /// - Returns: The new presence or `nil` if the presence should not be changed
+  open func transformTokenPresence(_ token: TokenSyntax) -> SourcePresence? {
+    return nil
+  }
+
   // MARK: - Formatting a token
 
   open override func visit(_ token: TokenSyntax) -> TokenSyntax {
@@ -360,6 +394,8 @@ open class BasicFormat: SyntaxRewriter {
     let isInitialToken = self.previousToken == nil
     let previousToken = self.previousToken ?? token.previousToken(viewMode: viewMode)
     let nextToken = token.nextToken(viewMode: viewMode)
+    let transformedTokenText = self.transformTokenText(token)
+    let transformedTokenPresence = self.transformTokenPresence(token)
 
     /// In addition to existing trivia of `previousToken`, also considers
     /// `previousToken` as ending with whitespace if it and `token` should be
@@ -375,7 +411,7 @@ open class BasicFormat: SyntaxRewriter {
         || (requiresWhitespace(between: previousToken, and: token) && isMutable(previousToken))
     }()
 
-    /// This method does not consider any posssible mutations to `previousToken`
+    /// This method does not consider any possible mutations to `previousToken`
     /// because newlines should be added to the next token's leading trivia.
     let previousTokenWillEndWithNewline: Bool = {
       guard let previousToken = previousToken else {
@@ -419,6 +455,14 @@ open class BasicFormat: SyntaxRewriter {
       if nextToken.leadingTrivia.startsWithNewline {
         return true
       }
+      if nextToken.leadingTrivia.isEmpty {
+        if nextToken.text.first?.isNewline ?? false {
+          return true
+        }
+        if nextToken.text.isEmpty && nextToken.trailingTrivia.startsWithNewline {
+          return true
+        }
+      }
       if requiresNewline(between: token, and: nextToken),
         isMutable(nextToken),
         !token.trailingTrivia.endsWithNewline,
@@ -439,6 +483,19 @@ open class BasicFormat: SyntaxRewriter {
       return trailingTrivia + Trivia(pieces: nextTokenLeadingWhitespace)
     }()
 
+    /// Whether the leading trivia of the token is followed by a newline.
+    let leadingTriviaIsFollowedByNewline: Bool = {
+      if (transformedTokenText ?? token.text).isEmpty && token.trailingTrivia.startsWithNewline {
+        return true
+      } else if token.text.first?.isNewline ?? false {
+        return true
+      } else if (transformedTokenText ?? token.text).isEmpty && token.trailingTrivia.isEmpty && nextTokenWillStartWithNewline {
+        return true
+      } else {
+        return false
+      }
+    }()
+
     if requiresNewline(between: previousToken, and: token) {
       // Add a leading newline if the token requires it unless
       //  - it already starts with a newline or
@@ -458,9 +515,12 @@ open class BasicFormat: SyntaxRewriter {
       }
     }
 
-    if leadingTrivia.indentation(isOnNewline: isInitialToken || previousTokenWillEndWithNewline) == [] {
+    if leadingTrivia.indentation(isOnNewline: isInitialToken || previousTokenWillEndWithNewline) == [] && !token.isStringSegment {
       // If the token starts on a new line and does not have indentation, this
-      // is the last non-indented token. Store its indentation level
+      // is the last non-indented token. Store its indentation level.
+      // But never consider string segments as anchor points since you can’t
+      // indent individual lines of a multi-line string literals without breaking
+      // their integrity.
       anchorPoints[token] = currentIndentationLevel
     }
 
@@ -488,14 +548,17 @@ open class BasicFormat: SyntaxRewriter {
     var leadingTriviaIndentation = self.currentIndentationLevel
     var trailingTriviaIndentation = self.currentIndentationLevel
 
-    // If the trivia contain user-defined indentation, find their anchor point
+    // If the trivia contains user-defined indentation, find their anchor point
     // and indent the token relative to that anchor point.
-    if leadingTrivia.containsIndentation(isOnNewline: previousTokenWillEndWithNewline),
+    // Always indent string literals relative to their anchor point because
+    // their indentation has structural meaning and we just want to maintain
+    // what the user wrote.
+    if leadingTrivia.containsIndentation(isOnNewline: previousTokenWillEndWithNewline) || isInsideStringLiteral,
       let anchorPointIndentation = self.anchorPointIndentation(for: token)
     {
       leadingTriviaIndentation = anchorPointIndentation
     }
-    if combinedTrailingTrivia.containsIndentation(isOnNewline: previousTokenWillEndWithNewline),
+    if combinedTrailingTrivia.containsIndentation(isOnNewline: previousTokenWillEndWithNewline) || isInsideStringLiteral,
       let anchorPointIndentation = self.anchorPointIndentation(for: token)
     {
       trailingTriviaIndentation = anchorPointIndentation
@@ -504,18 +567,36 @@ open class BasicFormat: SyntaxRewriter {
     leadingTrivia = leadingTrivia.indented(indentation: leadingTriviaIndentation, isOnNewline: previousTokenIsStringLiteralEndingInNewline)
     trailingTrivia = trailingTrivia.indented(indentation: trailingTriviaIndentation, isOnNewline: false)
 
-    leadingTrivia = leadingTrivia.trimmingTrailingWhitespaceBeforeNewline(isBeforeNewline: false)
+    leadingTrivia = leadingTrivia.trimmingTrailingWhitespaceBeforeNewline(isBeforeNewline: leadingTriviaIsFollowedByNewline)
     trailingTrivia = trailingTrivia.trimmingTrailingWhitespaceBeforeNewline(isBeforeNewline: nextTokenWillStartWithNewline)
 
-    if leadingTrivia == token.leadingTrivia && trailingTrivia == token.trailingTrivia {
-      return token
+    var result = token.detached
+    if leadingTrivia != result.leadingTrivia {
+      result = result.with(\.leadingTrivia, leadingTrivia)
     }
-
-    return token.detached.with(\.leadingTrivia, leadingTrivia).with(\.trailingTrivia, trailingTrivia)
+    if trailingTrivia != result.trailingTrivia {
+      result = result.with(\.trailingTrivia, trailingTrivia)
+    }
+    if let transformedTokenText {
+      let newKind = TokenKind.fromRaw(kind: token.tokenKind.decomposeToRaw().rawKind, text: transformedTokenText)
+      result = result.with(\.tokenKind, newKind).with(\.presence, .present)
+    }
+    if let transformedTokenPresence {
+      result = result.with(\.presence, transformedTokenPresence)
+    }
+    return result
   }
 }
 
 fileprivate extension TokenSyntax {
+  var isStringSegment: Bool {
+    if case .stringSegment = self.tokenKind {
+      return true
+    } else {
+      return false
+    }
+  }
+
   var isStringSegmentWithLastCharacterBeingNewline: Bool {
     switch self.tokenKind {
     case .stringSegment(let segment):

Sources/SwiftParserDiagnostics/DiagnosticExtensions.swift

@@ -12,7 +12,7 @@
 
 import SwiftDiagnostics
 import SwiftBasicFormat
-import SwiftSyntax
+@_spi(RawSyntax) import SwiftSyntax
 
 extension FixIt {
   /// A more complex set of changes that affects multiple syntax nodes and thus
@@ -108,11 +108,46 @@ extension FixIt.MultiNodeChange {
 
 // MARK: - Make present
 
-class MissingNodesBasicFormatter: BasicFormat {
+class PresentMakingFormatter: BasicFormat {
+  init() {
+    super.init(viewMode: .fixedUp)
+  }
+
   override func isMutable(_ token: TokenSyntax) -> Bool {
     // Assume that all missing nodes will be made present by the Fix-It.
     return token.isMissing
   }
+
+  /// Change the text of all missing tokens to a placeholder with their
+  /// name for diagnostics.
+  override func transformTokenText(_ token: TokenSyntax) -> String? {
+    guard token.isMissing else {
+      return nil
+    }
+
+    let (rawKind, text) = token.tokenKind.decomposeToRaw()
+
+    guard let text = text else {
+      // The token has a default text that we cannot change.
+      return nil
+    }
+
+    if text.isEmpty && rawKind != .stringSegment {
+      // String segments are allowed to have empty text. Replace all other empty
+      // tokens (e.g. missing identifiers) by a placeholder.
+      return "<#\(token.tokenKind.nameForDiagnostics)#>"
+    }
+
+    return nil
+  }
+
+  /// Make all tokens present.
+  override func transformTokenPresence(_ token: TokenSyntax) -> SourcePresence? {
+    guard token.isMissing else {
+      return nil
+    }
+    return .present
+  }
 }
 
 extension FixIt.MultiNodeChange {
@@ -123,8 +158,7 @@ extension FixIt.MultiNodeChange {
     leadingTrivia: Trivia? = nil,
     trailingTrivia: Trivia? = nil
   ) -> Self {
-    var presentNode = MissingNodesBasicFormatter(viewMode: .fixedUp).rewrite(node, detach: true)
-    presentNode = PresentMaker().rewrite(presentNode)
+    var presentNode = PresentMakingFormatter().rewrite(node, detach: true)
 
     if let leadingTrivia {
       presentNode = presentNode.with(\.leadingTrivia, leadingTrivia)

Sources/SwiftParserDiagnostics/PresenceUtils.swift

@@ -43,29 +43,6 @@ extension SyntaxProtocol {
   }
 }
 
-/// Transforms a syntax tree by making all missing tokens present.
-class PresentMaker: SyntaxRewriter {
-  init() {
-    super.init(viewMode: .fixedUp)
-  }
-
-  override func visit(_ token: TokenSyntax) -> TokenSyntax {
-    if token.isMissing {
-      let presentToken: TokenSyntax
-      let (rawKind, text) = token.tokenKind.decomposeToRaw()
-      if let text = text, (!text.isEmpty || rawKind == .stringSegment) {  // string segments can have empty text
-        presentToken = token.with(\.presence, .present)
-      } else {
-        let newKind = TokenKind.fromRaw(kind: rawKind, text: rawKind.defaultText.map(String.init) ?? "<#\(token.tokenKind.nameForDiagnostics)#>")
-        presentToken = token.with(\.tokenKind, newKind).with(\.presence, .present)
-      }
-      return presentToken
-    } else {
-      return token
-    }
-  }
-}
-
 /// Transforms a syntax tree by making all present tokens missing.
 class MissingMaker: SyntaxRewriter {
   init() {