When preparing for this years Advent of Code I am trying to improve my regex skills. A very useful function in the builtin regex module for python is re.sub.

I have been running a lot of clashofcode lately and quite a few of the challenges are about making the shortest code possible.

Here is the story/rant of when I tried to use time in Javascript.

I was working on some code during work today and had a little issue. I had created a internal library for parsing and handling arguments with flags to my C# project.

The idea was to be able to parse arguments like this:

--Person --name Alex --location -x 0 -y 13.4 --height 184

I see the arguments as a tree with nodes and leaves. In the example above there would be a person and location node, and a name, x, y and height leaf. Nodes contain any number of children nodes and there are special types of nodes which have no children called leafs. As you can see in the example above leafs can have values like int, double and bool.

Here is a very much simplified example of what I was working on and where my issue came to be.

public class Node
{
    public string Name;

    public Node(string name)
    {
        Name = name;
    }
}

public class Leaf<Type> : Node
{
    Type Value;

    public Leaf(string name, bool isOptional = false, Type? defaultValue = null) : base(name)
    {
        if (isOptional && defaultValue != null)
        {
            Value = defaultValue.Value;
        }
    }
}

With this code I my plan was to be able to create a Leaf like this:

With a code like this my goal was to be able to handle the arguments like this:

class Program
{
	class Location : Node
	{
		Flag<double> X = new Flag<double("x");
		Flag<double> Y = new Flag<double("y");

		// Methods relating to Nodes

	}

	class Person : Node
	{
		Flag<string> Name = new Flag<string>("name");
		Location location = new Location("location");
		Flag<double> Height = new Flag<double("height");

		// Methods relating to Nodes

	}

	public static void Main(string[] args)
	{
		public Node Person = new Node("person");
		Person.ParseArgs(args);

		Console.WriteLine(Person.Height); // prints 184

	}
}

I created two Nodes, one for location and one for person and then there is methods that handle the parsing and converting from string to the respective types and what not. Now the issue is, this does no in any way compile right now.

The problem is the following piece of code shown in the first listing:

    public Leaf(string name, bool isOptional = false, Type? defaultValue = null) : base(name)

The issue here is Type? (shorthand for Nullable<T>). This is because Nullable<T> only works for value types. This can be seen in how Nullable<T> was implemented, some of that code is shown below:

namespace System
{
    public struct Nullable<T> where T : struct
    {
        public Nullable(T value);
	//...

    }
}

where T : struct is an instance of what is called generic constraints and states that T must be of type struct (which all value types are).

Task for the reader: I strongly recomend pressing F12 on and int and any other type you in your mind belive “work in the same way” in visual studio to see that if all of those are indeed structs!

So back to the issue! To be able to use Nullable in my Leaf class I have to have the same requirement, naimly that all type T must be structs. This is well and good for `bool`, `int` and `double` which are types that I need to be able to have in arguments, but it does not work for strings! Pressing F12 in Visual Studio on a string gets you to its definition and you see that it is actually a class!

So now you are where I was when I realised I had a problem. I however did not really understand my problem at this point. I had a bunch of code. I had at some point decided as a sidethought that I wanted to have optional arguments and that hence default values would make a lot of sense. I decided Type? defaultValue = null was the way to go and when the compiler warned me about not being able to use Nullable just on any type willy-nilly I added a where T : struct and continued on.

I was fine with this for a while until I realised I really did need strings in my Leafs in certain situations. So I tried to solve it.

How did I try to solve it at first?

I imagined two classes:

	class Leaf<T> where T : struct
	class Leaf<T> where T : string

The idea was that I coulde use Type? defaultValue = null in one and type defaultValue = null in the other. But you cannot have two classes like that. I guess it is because the condition does not count as a part of the class definition the same way a name of does. Also even if it were possible having two classes with only this minor change is a lot of code duplication.

Next I thought about just:

	class Leaf<T> where T : struct
	class Leaf

But that felt bad since it would be quite confusing for users of the library, see example below:

Leaf<bool> FileName = new Leaf<bool>("enabled");
Leaf FileName = new Leaf("fileName");

Other leafs have their type clearly marked, but the string version is not which makes the code more difficult to understand. Also this would again require having two almost identical sets of classes which is bad for maintainability.

After some more thinking, some googling around and some thought about “why can’t I have classes like Name<T> and Name<T> where T : struct in the same namespace”. I figured this is the type of question one might ask on Stackoverflow because you would kinda feel smart while asking the question. So I started doing just that, feeling smart, and prepping for writing my question.

I started with a minimal version of my code. I started a new CLI project in Visual Studio and stripped out all the non-essential parts (which amounted to even less than what you have seen here). I did this untill I had my two Leaf classes and looking at those few lines of code in my CLI application I saw what has been clear in this post all along, that the reason I used where T : struct at all was that I wanted default values. I realised that my problem was really not how to have classes with the same name but different generic constraints but how to get a default value for a generic type. That realisation lead me to find the solution in 1 minute!

In the darkest hour.. a hero emerges!

And here default comes to the rescue! Mind you been here for the rescue since C# 2, I just did now know about it!

default(T) (or just default in certain cituaions since C#7.1) returns the default value for the type. 0 for value types and null for reference types.

So for all this the solution was trivial, I just needed this:

public class Leaf<Type> : Node
{
    Type Value;

    public Leaf(string name, bool isOptional = false, Type defaultValue = default) : base(name)
    {
        if (isOptional && defaultValue != null)
        {
            Value = defaultValue.Value;
        }
    }
}

I did end up using default(T) instead since i does not require C# 7.1. Student-Alex would have gone for default but now Im a PROFESSIONAL and that means being a bit more careful with stuff like that and since I have said “But it works on my computer” before when I upgraded to C# 7.0 a while back which caused minor issues for Visual Studio 2015 users. So upgrading to 7.1 to remove 3 chars feelt like a pretty bad ROI.

Conclusion

A real life solution has many parts to keep track of. The process of boiling down your question really helps you to understand what the problem really is. Once you have understood what the problem is the solution is just a google search away!

Also I guess knowing the ins and outs of your programing language has its benefits!

Ok, lets see if this jekyll thing is anything good!