builder-method

Creational Design Patterns: Builder Method

Today, our journey into the Guide to Software Design Patterns series takes us to yet another Creational design pattern: the builder method. In this article we’ll cover what the builder method is using both real world examples and functional C# code, so let’s get to it!

In the Real World

The simplest real world example of the builder pattern, which most of us are familiar with, is when we make (or order) a pizza. The pizza toppings cannot be added in any random order or the whole thing is likely to come out a mess. Instead, there is a step-by-step process that is followed. This usually begins by deciding on the size of the pizza, and therefore how much dough to use. Then the sauce is added, then cheese, then maybe some pepperoni and some olives, and so forth. Yet, as any pizza-maker will tell you, not all orders are the same, and therefore, not all pizzas are built in the same way. Each step in the process, and each topping that is added, can be thought of as a single step in the much larger builder pattern that is used to make that delicious pie you’re about to devour.

How It Works In Code

We’ll keep the pizza analogy going with our code example for this article — maybe we’re planning on developing an online ordering application for that new local pizza joint! In any case, to explore just how the builder method can help us, we must first examine a different, more common design pattern that is often used when working with objects that contain many mutable (changeable) properties.

A common technique in C# is to define our class and then define the constructor (the method with the same name as the class) with all the necessary parameters we might need to set properties for our class. For example, here we have defined our TelescopedPizza class that we’ll use to make pizzas, which has a wide range of properties including type of cheese, size, whether to include sauce or pepperoni or ham, and so forth:

Some keen observers may notice that while our TelescopedPizza class has properties for Olives and Mushrooms, we haven’t included a parameter for either in our constructor. This is because we’re trying to deal with the fundamental problem with this type of design pattern, which is indicated by our class name and is known as a telescoping constructor.

To see what we mean by the telescoping constructor pattern, here’s our second constructor for our class:

Now we’ve added parameters to set both the Olives and Mushrooms properties, but immediately it becomes apparent that this practice could quickly spiral out of control. While C# allows us to add as many constructors to our class as we’d like, so long as their parameter signature (the list of parameter types) is the same, if we wish to allow for all possible combinations of toppings for our pizza, it will be difficult to account for all of these with this pattern. The list of parameters starts to expand or telescope outward with each new constructor that is defined, hence the name of this design (anti)pattern.

One common solution to this issue with C# is to use default values. Here, we take our second constructor above, but add default values to each parameter:

This works in practice, since we can now define just about any combination when creating a new instance of our TelescopedPizza class, but it’s rather cumbersome to use in practice. One issue is that we have to retain (and use) the correct parameter order when creating a new instance:

In the above case, we’re able to make exactly the kind of pizza we want by using the constructor that contains default values, but we have to remember the order and specify values even where we don’t want that particular topping. Since every parameter is default, we have to indicate that ham should be false, just to “get access to” the olives argument that we care about and set it to true.

Newer versions of C# allow us to use named arguments, which means we don’t need to remember the positioning of parameters in the constructor list. Thus, we could also perform the same instantiation like so:

While this pattern is the most user-friendly method when using a telescope-style method, it’s still not ideal. For example, we might want to create our pizza object in code much the same way we’d build it in real life: step-by-step, with a bit of order. After all, it (usually) doesn’t make sense to add the sauce after the cheese. Bizarre, although still probably tasty!

This is where the builder method comes into play. The builder method allows us to create our pizza step-by-step, adding each appropriate topping in whatever order we wish. Best of all, as we’ll see below, because each addition is separated from one another in the code, we can easily add unique logic to the step of adding cheese that is different from when we add olives, for example.

To implement the builder method for our pizza example, we first start with our secondary builder class, which we’re calling PizzaBuilder:
public enum Cheese { Cheddar, Mozzarella, Parmesan, Provolone }
public enum Size { Small, Medium, Large, XLarge }

We’re still using the same list of properties that we did before, but the magic comes in when we start to define some of our instance methods. First, our constructor is very basic. We’ve decided that every pizza will have a size of some sort, so the one and only parameter in our constructor is size, which has a default value of Size.Largefrom our enum:

Now we get to add the extra methods to our builder, which allows us to pick and choose which toppings we want, and in what order to add them. Most of them are quite similar, so we’ll just take a look at two, AddCheese and AddSauce:

These are quite similar to normal instance methods, but the key is that we’ve defined the return type value to be the same as the parent class, PizzaBuilder. Then, after all logic is complete within the method, we return this, which returns the current instance of the PizzaBuilder class. You may also notice that for all our properties that are merely boolean values, we just assume that each will default to false and, therefore, we only set the property value to true when we call the appropriate Add... method.

The last critical method in our PizzaBuilder is the Build method, which returns an instance of the Pizza class (that we’ll define in a second), and which also accepts a single parameter: the current PizzaBuilder instance.

From here, defining our actual Pizza class is quite simple. We just ensure the constructor expects a PizzaBuilderinstance for the parameter, then we use the properties taken from that PizzaBuilder to assign values in our Pizza instance:

With everything defined, we can now make use of our new PizzaBuilder class like so:

Here we’ve created a pizza with a specific set of toppings, and we’ve added them in our desired order. This technique of using multiple dot method calls in a row is called method chaining, and is possible in this case because of how we defined all of our Add... methods inside the PizzaBuilder class. Since they each return the current PizzaBuilder instance, chaining them together like this simply executes the first call, then the next, then the next, each one using (and modifying) the current instance of the class.

Once we’ve added all our toppings as desired, we finally call the Build() method which, as we saw above, returns a new instance of Pizza that accepts our fully built PizzaBuilder and assigns all appropriate properties therein. The end result of our log output is a Pizza with exactly the property values we want:

Best of all, since each Add... method is self-contained with its own parameters and logic, we can easily take advantage of our default values for cheese and pizza size by using the following call:

Again, this builds a pizza just how we want it, including the default values of Large size and Mozzarella cheese: