Click here to Skip to main content
Click here to Skip to main content
Go to top

Calculating Metrics and Searching with a CodeDOM (Part 8)

, 6 Mar 2013
Rate this:
Please Sign up or sign in to vote.
Calculating metrics on and searching a CodeDOM.

Introduction

This article is about basic analysis of code in a CodeDOM, such as calculating metrics on it and searching it. Sources are included. This is Part 8 of a series on CodeDOMs. In the previous parts, I’ve discussed CodeDOMs, provided a C# CodeDOM, a WPF UI and IDE, a C# parser, solution/project CodeDOM classes, covered loading type metadata, and handled resolving symbolic references in a CodeDOM.

The Visitor Pattern 

Calculating metrics on or searching a CodeDOM tree (or a sub-tree) both require traversing the objects in the tree. This requires specialized knowledge of each type of object, because different types have different fields representing children objects – a BinaryOperator has Left and Right children, an If statement has a child Expression, etc. In order to avoid writing specific code to visit children objects more than once, the Visitor design pattern has been employed. This was done by creating an IVisitor interface and adding an abstract void Accept(IVisitor visitor) method to CodeObject which is overridden by all subclasses as necessary (all of them that define child objects). To visit nodes in a tree for a specific purpose, a class is declared that implements IVisitor, and this class is instantiated and passed to the Accept() method of the top-most object of the desired tree or sub-tree.

The IVisitor interface defines about 55 methods for the more important types in the codeDOM. Some leaf types are not represented separately (such as concrete operators, references, and doc comment types) in order to keep the method count closer to 50 than 300. If specific types must be checked that don’t have their own method, logic will need to be added to the method for the nearest base type that uses is or as to check the specific type. All methods of the interface must always be implemented, but types which do not need special visiting logic can have their methods left blank or just call a default handling routine. Internally generated (hidden) code objects aren’t visited (such as generated default constructors, delegate constructors and BeginInvoke/EndInvoke methods, the global extern alias of CodeUnits, etc). Hidden symbolic references using the HiddenRef property are optionally visited, depending upon the value of the bool VisitHiddenRefs property of the interface. 

A new example, VisitTree(), has been added to the Nova.Examples project which demonstrates how to implement your own visitor using the class MyVisitor.

Calculating Metrics

The most basic metrics on a codebase are just counts of all of the different types of objects in the code.  For example, projects, files, lines, SLOC (source lines of code – lines with actual code on them), statements, loops, local variables, literals, etc. More advanced metrics require some calculations, such as average SLOC per method/class, average conditions per method/class, average methods per class, average code objects per method/class, etc.

To implement metrics capability, a Metrics class has been added which contains many fields for counting different types of objects in a CodeDOM, and also some fields for calculated metrics. A MetricsVisitor class has been created which implements the IVisitor interface and contains a Metrics instance which is used to count the various types of objects as a tree is visited, and is activated by calling ‘CalculateMetrics(CodeObject)’, passing in the starting node. This method calls Accept() on the object, which updates all object counts, and then it also calls Calculate() on the Metrics instance to update the calculated metrics.  To make this process even easier, a ‘Metrics CalculateMetrics()’ method has been added to CodeObject which creates an instance of MetricsVisitor, calls CalculateMetrics() on it, and returns the resulting Metrics object – thus, metrics can be calculated on any object in a codeDOM simply by calling CalculateMetrics() on it.

When a Solution, Project, or CodeUnit is loaded, if the LoadOptions.LogMetrics option is specified, the Load() method will automatically call CalculateMetrics() on the object and log the total projects (for solutions), files (for solutions/projects), lines, SLOC, types, and code objects. The fields of the Metrics class also have Description attributes that describe each metric in detail, and can be used by a UI to display to the user (and Nova Studio now does this).  A new example, CalculateMetrics(), has been added to the Nova.Examples project which demonstrates calculating metrics. The usage is very simple, for example:

// Calculate some metrics for an entire Solution:
Metrics metrics = solution.CalculateMetrics();
Log.WriteLine(string.Format("Solution '{0}': {1} projects; {2} files; {3} lines; "
    + "{4} SLOC; {5} types; {6} code objects; {7:N2} code objects per SLOC",
    solution.Name, metrics.Projects, metrics.Files, metrics.Lines, metrics.SLOC,
    metrics.Types, metrics.CodeObjects, metrics.CodeObjectsPerSLOC));

Searching the CodeDOM 

Searching for particular references, types, or text in the code can now be accomplished with the new classes FindReferences, FindByType,  and FindByText, all of which implement the IVisitor interface. A new class, Result, is also provided that represents a found CodeObject and its associated CodeUnit. All three of the find classes work on a given scope (starting CodeObject), and produce a collection of Result objects. The FindReferences class has various options, such as including references to members when searching for references to a type, including derived types, including overrides of virtual members, or including overloads when searching for a method. The FindByText class has options to be case-sensitive, match whole words only, use regular expressions, or to match only declarations, references, literals, comments, or messages.

New examples (with the same names as the classes) have been added to the Nova.Examples project which demonstrate how use these three new classes for searching a CodeDOM. Here are some example usages:

// Find all references to a method in an entire solution
FindReferences findMethodReferences = new FindReferences(methodDecl, solution);
findMethodReferences.Find();
Log.WriteLine("Found " + findMethodReferences.Results.Count
    + " references to method declaration in the solution");


// Find all If statements in a CodeUnit
FindByType findIfs = new FindByType(typeof(If), codeUnit);
findIfs.Find();
Log.WriteLine("Found " + findIfs.Results.Count + " 'if' statements in "
    + codeUnit.Name);

// Find all code objects in a solution that have a name of 'find1' OR 'find2', or
// contain it in their text content.  Match case, whole words, and use reg exp.
FindByText findText = new FindByText("find1|find2", solution, true, true, true);
findText.Find();
Log.WriteLine("Found " + findText.Results.Count + " objects containing the "
    + "text 'find1' OR 'find2' in the solution");

The FindByType class has been used to implement a helper method on CodeObject named GetAllChildren<T>() which returns an IEnumerable<T> and can be used to easily iterate over all child objects of a particular type. For example:

// Query ALL MethodDecls in an entire Solution that have more than 2 'if' statements
var methodDecls = from methodDecl in solution.GetAllChildren<MethodDecl>()
                  where methodDecl.GetAllChildren<If>().Count() > 2
                  select methodDecl;
Log.WriteLine(methodDecls.Count() + " MethodDecls have more than 2 'if' statements");

Nova Studio Improvements

Nova Studio now has a Metrics option on the context menu of the Solution Tree that calculates metrics for the selected solution, project, folder, or individual file, and displays them in a new Metrics dialog. Tooltips provide detailed descriptions of each metric.  There is also a tool bar icon that calculates metrics for any file being viewed. The dialog looks like this: 

The context menu of the code window now has Find options with new dialogs for Find References and Find By Type, and a Find icon on the tool bar brings up a new Find Text dialog.  These new dialogs allow the scope of the search and various options to be specified, and the results show in a new Results window at the bottom of Nova Studio. The dialogs and Results window look like this:

Using the Attached Source Code

A new Analysis folder has been added with the new classes Metrics, MetricsVisitor, FindByText, FindByType, FindReferences, and Result. Various analysis-related methods – such as Accept() – have been added to many of the existing CodeDOM classes, segregated into regions with a comment of “ANALYSIS”. New examples have been added to the Nova.Examples project that make use of the new features: FindReferences(), FindByType(), FindByText(), CalculateMetrics() and VisitTree(), which makes use of a new MyVisitor class. Also, LINQ queries have been added that make use of CodeObject.GetAllChildren<T>() to query all objects of a particular type.  Nova Studio has been improved as mentioned in the previous section. As usual, a separate ZIP file containing binaries is provided so that you can run them without having to build them first.

Summary

It’s now possible to easily calculate metrics for a CodeDOM, create custom code to process a CodeDOM using the Visitor pattern, and search for text/types/references in a CodeDOM.  Nova Studio has added support for doing such things from the UI. Quite a large feature set has been built up over this series of articles, resulting in a useful C# CodeDOM. Something that would be nice to add would be static code analysis to detect possible issues with code and suggest improvements, but I’m not going to tackle that one just yet. In my next article, I’ll take a look at the Roslyn project, and see how it compares with the usability, functionality, and performance of my CodeDOM.

License

This article, along with any associated source code and files, is licensed under The Common Development and Distribution License (CDDL)

Share

About the Author

KenBeckett
Software Developer (Senior)
United States United States
I've been writing software since the late 70's, currently focusing mainly on C#.NET. I also like to travel around the world, and I own a Chocolate Factory (sadly, none of my employees are oompa loompas).

Comments and Discussions

 
QuestionImpressive Pinmemberaralmo2-Apr-13 5:03 

General General    News News    Suggestion Suggestion    Question Question    Bug Bug    Answer Answer    Joke Joke    Rant Rant    Admin Admin   

Use Ctrl+Left/Right to switch messages, Ctrl+Up/Down to switch threads, Ctrl+Shift+Left/Right to switch pages.

| Advertise | Privacy | Mobile
Web03 | 2.8.140921.1 | Last Updated 6 Mar 2013
Article Copyright 2012 by KenBeckett
Everything else Copyright © CodeProject, 1999-2014
Terms of Service
Layout: fixed | fluid