Code analysis with roslyn - Getting all elements of a given type

Question

Hi

I'm currently playing with roslyn.

whats the easiest way to get all elements of a given type from a source file?

For example getting all "if" from a cs file ?

 

Answer 1

To get all the IfStatementSyntax nodes from a cs file, first get the SyntaxTree for the file and then use the DescendentNodes() method on the root node, followed by the LINQ OfType operator.

var ifStatements = tree.Root.DescendantNodes().OfType<IfStatementSyntax>();

---

Wayward LINQ Lacky

Answer 2

If you wish to 'process' all nodes of type IfStatementSyntax in a SyntaxTree, you could also implement a SyntaxWalker / SyntaxVisitor  -

class Program
{
    static void Main(string[] args)
    {
        var source = @"
class Program
{
    static void Main()
    {
        if (true) Console.WriteLine(""First"");
        if (false) Console.WriteLine(""Second"");
    }
}";
        var tree = SyntaxTree.ParseCompilationUnit(source);
        var walker = new Walker();
        walker.Visit(tree.Root);
    }
}

class Walker : SyntaxWalker
{
    private int count = 0;
    protected override void VisitIfStatement(IfStatementSyntax node)
    {
        count++;
        Console.WriteLine(count + ": " + node.GetText());
        base.VisitIfStatement(node);
    }
}

The above program prints the following output -

1: if (true) Console.WriteLine("First");
2: if (false) Console.WriteLine("Second");

---

Shyam Namboodiripad | Software Development Engineer in Test | Roslyn Compilers Team

Answer 3

Hi

 

the walker was what I'm loogik for.

 

 

What is there correct way to count empty and not empty lines  ?

For examlple to make some LOC and LLOC Metrics ?  Is there a explizit visit method for empty lines or an explicite visit method for non empty lines ?

Should I accumulate the different nodes types like "BlockSyntax" and so on ?

Or should I count the semicolon token?

Many thanks in advances

Answer 4

I think there are different ways to do this depending on your definition of LOC and LLOC :)

You shouldn't need a Walker to count just the total LOC -

 

var tree = SyntaxTree.ParseCompilationUnit(source);
Console.WriteLine("Total LOC: " + tree.Text.LineCount);
// OR
// Console.WriteLine(tree.Root.GetFullTextAsIText().LineCount);
var linesWithNoText = 0;
foreach (var l in tree.Text.Lines)
{
    if (string.IsNullOrEmpty(l.GetText().Trim()))
    {
        ++linesWithNoText;
    }
}

 

If you want to count the number of semi-colons in the code you could write a SyntaxWalker that counts this as follows -

 

static void Main()
{
    var source = @"
    class Program
    {
        // A comment

        /* A multi-line
            comment */
        #if false
            Some disabled code

        #endif

        static void Main()
        {
            if (true) Console.WriteLine(""First"");
            if (false) Console.WriteLine(""Second"");
        }
    }";
    var tree = SyntaxTree.ParseCompilationUnit(source);
    var walker = new Walker();
    walker.Visit(tree.Root);
    Console.WriteLine("SemiColon Count: " + walker.SemiColonCount);
}


class Walker : SyntaxWalker
{
    public Walker()
        : base(visitIntoStructuredTrivia: true)
    {
    }

    public int SemiColonCount { get; private set; }

    protected override void VisitToken(SyntaxToken token)
    {
        base.VisitToken(token);

        if (token.Kind == SyntaxKind.SemicolonToken)
        {
            ++SemiColonCount;
        }
    }
}

 

You could also count the EndOfLineTrivia in the SyntaxTree by overriding VisitTrivia() in your SyntaxWalker - however, this would only count "significant" new lines (for example this wouldn't count new lines within a MultiLineCommentTrivia).

 

class Walker2 : SyntaxWalker
{
    public Walker2()
        : base(visitIntoStructuredTrivia: true)
    {
    }

    public int SignificantEOLCount { get; private set; }

    protected override void VisitTrivia(SyntaxTrivia trivia)
    {
        base.VisitTrivia(trivia);

        if (trivia.Kind == SyntaxKind.EndOfLineTrivia)
        {
            ++SignificantEOLCount;
        }
    }
}

---

Shyam Namboodiripad | Software Development Engineer in Test | Roslyn Compilers Team

Answer 5

To clarify a bit about Roslyn's Syntax model, parts of the source text that are insignificant from the point of view of compilation (e.g. whitespace, new lines, comments etc.) are parsed into objects of type SyntaxTrivia which are attached under SyntaxTokens. Parts of source text that are significant from point of view of compilation (e.g. identifiers, keywords etc.) are parsed into SyntaxTokens and these SyntaxTokens are in turn attached under a hierarchy of SyntaxNodes (e.g. if statements, class declarations etc.). Some SyntaxTrivia such as those for preprocessor directives and xml documentation comments have 'structure' (and can contain child SyntaxNodes and SyntaxTokens) - these special trivia are called structured trivia -http://social.msdn.microsoft.com/Forums/en-US/roslyn/thread/7bcc72a1-e54e-44eb-a488-0a0ace0ed903.

The Roslyn Project Overview Document contains a nice overview regarding the shape and contents of syntax trees (see section 3). This doc also contains a nice overview of the compilation pipeline (Syntax, Symbols, Semantics, Code Generation) and the APIs exposed in Roslyn for each stage in this pipeline.

The Roslyn CTP also ships with a couple of syntax visulaizer tools. You could use 'Syntax Visualizer Tool Window' tool to understand how the tree looks for specific forms of source code and figure out what Kinds / Types you would need special handling for in your Walker...

---

Shyam Namboodiripad | Software Development Engineer in Test | Roslyn Compilers Team