Previous post in the series – TDD – commit by commit String Calculator Kata (II) In the previous post I did a cleanup phase and added one feature to the String Calculator. The kata is not finished yet.

Step VIII - Different Delimiters

String Calculator should support different delimiters. Previous solution was just a hard-coded hack, still it was good enough solution. There wasn't any requirement that would suggest a need for something more sophisticated. Now I need to modify my main function so it accepts input -

//;\n where the first line char ';' is delimiter. The test

var input = "//;\n1;2;3;4";
var expectedSum = 10;

int actualSum = stringCalculator.SumFromString(input);

Assert.That(actualSum, Is.EqualTo(expectedSum));

Commit - Parametrized Delimeter Test
Just an extension to previous tests and simpliest solution below.

var delimeterIndex = input.IndexOf("//");

var delimeters = new List<char>();

if (delimeterIndex >= 0)
    delimeters.Add(input[delimeterIndex + 2]);
    input = input.Substring(4);

Commit - Parametrized Delimeter Implementation
IndexOf just checks if the input string does contains sign of parametrisied delimiter. If that's true, code tries to extract delimiter in a very crude way. List of default delimiters is provided in case of missing parametrisied delimiter. This is hard-coded code, not a beautiful solution but I don't have other requirements. I can assume that I don't have to support anything else. I am assuming that data will be provided in one particular format, that's why solution is pretty naive Still it does work and all the existing tests are green, I don't need anything else.

Step IX - Clean up

So now I have a very naive solution. It's time for my favorite part 'The Cleanup'. The main method has some complexity, for the first time. It seems that it has two responsibilities: extracting the delimiters and summing up the values. In order to make this cleaner, I am going to extract those two behaviors into separate methods.

Extract Delimeters

Commit: Extract Delimeters

Extract Sum

Commit: Sum Method

var delimeters = ExtractDelimeters(ref input);
return this.ParseSumValues(delimeters, input);

It's not perfect, but this is just one step. I don't like 'ref' keyword, it's a huge code-smell. This needs to be changed in the 'future'. Another problem is that the 'ExtractDelimeters' function is returning 'List' collection. This is also a code smell, especially when I am only using this list to enumerate through its values. Why then provide type of List, if I am not using its functions like Sort, Add ? There is even a better question, shouldn't I restrict the user of code and communicate to him that this collection should only be used to enumerate ?. I am going to change it to IEnumerable.

Refactor: List to IEnumerable
So that was the final cleanup, no revolutions here. Still, same creepy code that is haunting my eyes. More changes will come naturally with new requirements.

Step X - Negative Values

Next requirement, all the negative values are not supported and I need to throw exception with all the negative values.

 public void Negative_values_not_supported_throws()
    var calculator = new StringCalculator();
    var value = "1,2,-3,-4";

    var ex = Assert.Throws<ArgumentException>(() => calculator.SumFromString(value));
    Assert.That(ex.Message, Is.EqualTo("values '-3,-4' not supported"));

Commit - Test negative values
Another iteration of previous tests. In order to keep the test simple, I am not using the parametrisied delimiters.


 var negativeValues = values.Where(x => x < 0);

 if (negativeValues.Any())
     throw new ArgumentException(string.Format("values '{0}' not supported", string.Join(",", negativeValues)));

 return values.Sum();

Commit - negative values implementation
I had to move 'int.Parse' step so i could enumerate through the parsed values instead of raw strings. With this I can just collect all the negative values and throw an exception if there are any. Little note, I am using 'Any()' instead of 'Count() > 0' with IEnumerables. Count needs to enumerate whole list, any will stop on first matching element. It also looks better.

Question: - Are you a fan of 'Count() > 0' or 'Any()' ?