Setting Up Fitnesse for .net

I promised (myself) that this blog would be more of a notebook; a stream of consciousness  describing my own discoveries. I’m not sure if what I’m going over is old hat for the Fitnesse veteran, or if organizing acceptance tests and unit tests in this way is anything of value to anyone other than myself. Although there are tutorials of setting up Fitnesse for .net already out there, I’m putting down my own personal take on the matter as what I found wasn’t exactly what I was looking for all in one place. It’s what I’m involved with at the moment.

This was all sparked when I attended Maykel Suarez’s presentation on Fitnesse in flow mode. I’ve learned a lot from Mike on testing and software in general since I’ve had the privilege of working with him. I was put off from Fitnesse because of some of the misuse and abuse I’ve seen in the wild. My main beef was that folk were using it as a scripting platform. This became a tangled mess of imperative goo, and the textarea provides a poor IDE in my opinion. After attending Mike’s presentation I realized I could use the wiki to document the intention of the test, and that there was already support from like minded people to make this happen.

Checkout the acceptance test I wrote for a hypothetical POS system. Notice, there’s nothing techie in there. In fact, it even looks like user documentation. A sign, that perhaps this is a good acceptance test.

For the Fitnesse guru’s; this particular example could probably have been written more efficiently using a good ol’ fashion ColumnFixture, but I wanted to illustrate the use of plain-text tables and how they can provide a very flexible test structure. You’ll see what the underlying Fitnesse code looks like in a second. I also could have taken the introduction text and put that inside the table to use as the driver for the test.

Getting All the Pieces in Place

1. Go get the latest build of Fitnesse. Just download the jar file like it says (and if you don’t have Java installed you’ll need that as well).
2. Go get the latest fitsharp library. This provides the runner and the hooks for Fitnesse to execute .net code.
3. Create a test project. I use MSTest, but whatever you’re into is cool. I like to name my test project something like GoGoGrocer.Specifications, where GoGoGrocer is the name of the system/application.
4. Create a folder to hold your acceptance tests. I name mine AcceptanceTests. Go figure.
5. Dump the fitnesse.jar file you downloaded directly into the AcceptanceTests folder.
6. Create a little batch file to handle starting up Fitnesse and put these options in it:

   java -jar fitnesse.jar -p 8080 -e 0

   There are a lot of options to start Fitnesse, and a big help was to realize how useful the Quick Reference section was in the User Guide. The options above start the Fitnesse server on port 8080 (-p) and the turn off version history – you know the wiki thing (-e 0).
7. Create a bin folder under AcceptanceTests and put all the fitsharp files in there. Also change the project’s build settings to output into that folder as well.
8. Create additional folders under the project for UnitTests and Fixtures. I’ll explain the Fixtures in a second.
9. Create a file called fitness.config and put that in the bin folder. The contents of the file should be:

   <suiteConfig>
     <ApplicationUnderTest>
       <AddAssembly>
         bin\GoGoGrocer.Specifications.dll
       </AddAssembly>
       <AddNamespace>fitlibrary</AddNamespace>
       <AddNamespace>system</AddNamespace>
       <AddNamespace>
         GoGoGrocer.Specifications.Fixtures
       </AddNamespace>
     </ApplicationUnderTest>
     <fitSharp.Fit.Application.FileExclusions>
       <Add>^\.svn$</Add>
     </fitSharp.Fit.Application.FileExclusions>
   </suiteConfig>

   This adds the test assembly into the runner, and also provides a place to reference the namespaces of your fixtures. This makes it easy for you to reference them in your Fitnesse code. UPDATE: If your tests hang, make sure this file is encoded as UTF-8 (XML...sigh)
10. One more file to go. Open up the command prompt and run the startFitnesse.bat file. Among other things, this will install your FitnesseRoot and build the initial structure. Edit the content.txt file directly under the newly created FitnesseRoot folder.

    !*> using fitsharp
    !define COMMAND_PATTERN {%m -c bin\fitnesse.config -r fitnesse.fitserver.FitServer,bin\fit.dll %p}
    !define TEST_RUNNER {bin\Runner.exe}
    
    *!
    
    !*> defines
    !define COLLAPSE_SETUP {true}
    !define COLLAPSE_TEARDOWN {true}
    
    *!

    Hint: use the “Show All Files” feature in the Solution Explorer and add all the relevant folders to your project.

Now you’re ready to start using Fitnesse. When you’re all done with setup your project structure should look something like this:

Writing a Test in Flow Mode

The following content file defines the “Total Items” test shown above:

!2 Total Items Purchased

The customer purchases items. Those items are scanned. The price of the item is put on the 
bill along with it's description. See figure 1. The subtotal is calculated as the sum of
the price of all items on the bill and displayed at the bottom.

!img-l http://sites.google.com/site/futureturnip/home/images/sub-total-receipt.png

![ TotalItemsPurchased
The cashier scans "Wonder Bread" costing $3.49
The description "Wonder Bread" is displayed on the bill
The bill shows that the "Wonder Bread" costs $3.49
The sub total shows $3.49
The cashier scans "Gallon Milk 2%" costing $3.89
The description "Gallon Milk 2%" is displayed on the bill
The bill shows the cost of "Gallon Milk 2%" is $3.89
The sub total shows $7.38
The cashier scans "Bubbly Yummy" costing $0.79
The description "Bubbly Yummy" is displayed on the bill
The bill shows the cost of "Bubbly Yummy" is $0.79
The sub total shows $8.17
]!

The test is human-readable text, even in it’s wiki markup form. The trick is of course tying it to a Fixture. The fixture is a bridge between the Fitnesse runner and your code. The plain-text table ties itself to the TotalItemsPurchased class on it’s opening declaration. In that class a method is defined for every line in the text; replacing spaces with underscores and generally ignoring special characters like quotes or the percent sign (it did take a little experimentation to find that out). Here’s what fixture code might look like:

namespace GoGoGrocer.Specifications.Fixtures
{
  public class TotalItemsPurchased
  {
    public bool The_cashier_scans_Wonder_Bread_costing_3_49()
    {
      // This method is just the hook.
      // Fire up your API and test the SUT.
    }

    public bool The_description_Wonder_Bread_is_displayed_on_the_bill()
    {
      return true;
    }

The methods return true for a passing test. That’s a little quirky, and there are opportunities to improve the reporting, but that’s the bridge into your test code.

Summary

I still have some little issues with Fitnesse, and I still think the Ruby folk have us beat with Cucumber, but for .net Fitnesse offers a pretty-good solution for acceptance tests.

I also tried to “Export Template…” this, but the resulting template doesn’t install itself into Visual Studio. Haven’t figured that out yet, but I’ll share if I ever do.

Functional Patterns for Refactoring in C#

Trying to work on a little filler presentation for my brown-bag sessions at work on a couple of new code smells that can be solved thanks to the small dose of functional programming afforded us by LINQ and C# 3.5 and greater.

I’m working from the refactoring example given in Agile Principles, Patterns, and Practices in C# for the Sieve of Eratosthenes. When this book was published LINQ wasn’t around, and in a 1.x world this is Uncle Bob’s final listing for the problem:

public static int[] GeneratePrimeNumbers(int maxValue)
  {
      if(maxValue < 2)
          return new int[0];
      else
      {
          UncrossIntegersUpTo(maxValue);
          CrossOutMultiples();
          PutUncrossedIntegersIntoResult();
          return result;
      }
  }

This is fine. It reads well. I’ll get into the implementation of some of the methods in a second, but here’s my final implementation of the same method:

public static int[] GeneratePrimeNumbers(int maxValue)
  {
    return 
      Primes
        .TakeWhile(prime => prime <= maxValue)
        .ToArray();
  }

When Generating a Sequence Return IEnumerable<T>

I would like to drop the call to ToArray, but then I would break the interface from the original problem. However, whenever you see a sequence being generated, use IEnumerable<T> unless you need the indexing, and use an iterator if you can get away with it. Take my implementation of Primes for example:

public static IEnumerable<int> Primes
  {
    get
    {
      var remaining = Integers.Where(i => i >= 2);
      while (true)
      {
        int prime = remaining.First();
        yield return prime;
        remaining = remaining.Where(num => num % prime != 0);
      }
    }
  }

There’s no need for any result arrays or other collectors. Just generate the sequence and let the caller tell stop pulling values when they’re ready. This is all about deferred execution and lazy evaluation.

Likewise, my implementation of Integers is:

public static IEnumerable<int> Integers
  {
    get
    {
      int count = 0;
      do
      {
        yield return count++;
      } while (true);
    }
  }

Replace Loops with Filters, and Use Aggregates

Something you ended up doing a lot in 1.x code was writing a function like this (again from Uncle Bob’s original example):

private static int NumberOfUncrossedIntegers()
  {
      int count = 0;
      for (int i = 2; i < crossedOut.Length; i++)
      {
          if (NotCrossed(i))
              count++; // bump count
      }
      return count;
  }
  

Basically, this counts the number of elements in the boolean array crossedOut that are false. In the case of this algorithm, we have to skip the first two. In LINQ this becomes:

private static int NumberOfUncrossedIntegers()
  {
    return crossedOut.Skip(2).Where(x => x == false).Count();
  }

The where statement provides a filter and the count provides the counting. I in-lined the NotCrossed function in the lambda.

Summary

I’m not knocking Uncle Bob’s original implementation. I think it’s groovy. I’m just saying that now that LINQ is around it can let us leverage some pretty powerful syntax to refactor our loops and generators.

Oh no! It’s you Again.

So it’s January, and my New Year’s resolution was to suck less than I did the year before. My blogging, and my desire to blog have waned over the last year or couple of years. So, new blog; fresh start; time to rekindle the love of programming and creation of software in general. So in the spirit of my resolution, I’m committing to blogging again.

The old blog had to die. I was paying a nominal fee to keep it alive and entertain the 17 readers I had collected. The content attempted too much at times. I was put off from posting frequently because most of the posts were articles and not log entries. Some of the content was useful (to me at least), but it’s best to make a clean break.

The new blog is a place to store my eclectic bits of techie-software related nuggets. It will be more of a journal. Things I’m working on, as I’m working on them. I’ll try to post the whitepapers somewhere else (but probably not at all).

//TODO: Insert Blog Post Here