Removing redundant semicolons in code with SyntaxRewriter

Question

I am trying to remove redundant semicolons in the code using a custom syntax rewriter.

public class Sample
{
   public void Foo()
   {
      Console.WriteLine("Foo");
      ;
   }
}

The following syntax rewriter covers most scenarios as in the Sample class.

public class EmptyStatementRemoval : CSharpSyntaxRewriter
{
  public override SyntaxNode VisitEmptyStatement(EmptyStatementSyntax node)
  {
    return null;
  }
}

However when the semicolon has a leading or trailing trivia, returning null from the VisitEmptyStatement method removes the trivia, which is unintended.

public class Sample
{
   public void Foo()
   {
      Console.WriteLine("Foo");
      #region SomeRegion
      //Some other code
      #endregion
      ;
   }
}

I couldn't determine how to return a node with only the leading and trailing trivia removing the semicolon. I tried to replace the semicolon token with another token using node.WithSemicolonToken(SyntaxToken) method, which turns out to be accepting only tokens of type SyntaxKind.SemicolonToken or throws an ArgumentException.

Answer 1

One approach that could work is to replace the semicolon token with a missing semicolon token:

public override SyntaxNode VisitEmptyStatement(EmptyStatementSyntax node)
{
    return node.WithSemicolonToken(
        SyntaxFactory.MissingToken(SyntaxKind.SemicolonToken)
            .WithLeadingTrivia(node.SemicolonToken.LeadingTrivia)
            .WithTrailingTrivia(node.SemicolonToken.TrailingTrivia));
}

For your #region example, the result looks like this (notice the line containing only whitespace where the semicolon was):

public class Sample
{
    public void Foo()
    {
        Console.WriteLine("Foo");
        #region SomeRegion
        //Some other code
        #endregion

    }
}