ExtractXml: Including Examples in Your API Documentation

When using an API it's always nice to have some examples showing how the different classes and their methods and properties are actually used. Examples usually come in two flavors:

  • Full projects (more or less) ready to compile
  • Code snippets in the documentation (e.g. generated by NDoc)

Ok, I have to admit that I tend to ignore the full projects. When I need an example to understand something, I'm often looking for quick answers - not easy to get in an example project sometimes showing several concepts at once.

The small examples usually consist of maybe only a few lines of code, that in many cases are not ready to be compiled (at least not without adding some more code). But often these examples are exactly what is required to get quick help, getting straight to the point.

Unfortunately for the API developer, sample code inside the <example> tag of the XML Doc comments is a pain to create and maintain.

  • The code has to be tested.
    A missing semicolon in an example is definitely not the best way to convince the API users of the quality of the documentation. And if the code shows something that simply is not possible (e.g. a feature that was planned, but later cut from the release), people get really upset.
  • The code has to reflect the current state of the API.
    As the work on the API continues, sample code copied and pasted into your documentation has to be updated over and over again.

With a large number of examples, the only real solution is to take sample code from actually working code (e.g. from some of the NUnit tests for the API) and to automatically insert these code snippets into the doc comments.

The <include> tag of the XML doc comments allows the inclusion of XML fragments stored in an external XML file. I wrote a small command line tool (ExtractXml, download here) that generates an XML file from a ".cs" file. It's pretty much of a quick hack without much planning, but it's good enough for my own work.

This is how it works: the ".cs" file is decorated with "////" comments containing XML processing instructions and tags:

//// <?xml version="1.0" encoding="utf-16" ?>
//// <Documentation><!--
using System;
using System.Diagnostics;
namespace ExtractXml
    /// <summary>
    /// Summary description for Example.
    /// </summary>
    public class Example
        /// <summary>
        /// Does something really cool.
        /// </summary>
        public void DoSomething()
            //// -->
            //// <Example name="HowToAddStrings">
            //// <example>
            //// <b>How add two strings:</b>
            //// <code>
            //// <![CDATA[
            string str1="Hello ";
            string str2="World";
            string str3=str1+str2;
            //// ]]></code>
            //// </example>
            //// </Example><!--
//// --></Documentation>

The tool is called like this:

ExtractXml inputFile.cs outputFile.xml

It removes the "////" comments, loads the resulting text into an XML document (detecting errors in the XML), removes the XML "<!-- ... -->" comments and saves the result:

<?xml version="1.0" encoding="utf-16"?>
<Example name="HowToAddStrings">
<b>How add two strings:</b>
string str1="Hello ";
string str2="World";
string str3=str1+str2;

In your API code you reference the example like this:

/// ...
/// <include file="Example.xml" path="Documentation/Example[@name='HowToAddStrings']/*"/>
/// ...

Note that you are completely free regarding the structure of the XML tags, as long as the result is a valid XML document.


The download contains the executable (for .NET Framework 1.1) and the source (project file for VS.Net 2003).

Recommended Links

No Comments