What are drawbacks or disadvantages of singleton pattern?

design-patternssingleton

The singleton pattern is a fully paid up member of the GoF's patterns book, but it lately seems rather orphaned by the developer world. I still use quite a lot of singletons, especially for factory classes, and while you have to be a bit careful about multithreading issues (like any class actually), I fail to see why they are so awful.

Stack Overflow especially seems to assume that everyone agrees that Singletons are evil. Why?

Please support your answers with "facts, references, or specific expertise"

Best Answer

Paraphrased from Brian Button:

They are generally used as a global instance, why is that so bad? Because you hide the dependencies of your application in your code, instead of exposing them through the interfaces. Making something global to avoid passing it around is a code smell.
They violate the single responsibility principle: by virtue of the fact that they control their own creation and lifecycle.
They inherently cause code to be tightly coupled. This makes faking them out under test rather difficult in many cases.
They carry state around for the lifetime of the application. Another hit to testing since you can end up with a situation where tests need to be ordered which is a big no no for unit tests. Why? Because each unit test should be independent from the other.

Model-View-Presenter

In MVP, the Presenter contains the UI business logic for the View. All invocations from the View delegate directly to the Presenter. The Presenter is also decoupled directly from the View and talks to it through an interface. This is to allow mocking of the View in a unit test. One common attribute of MVP is that there has to be a lot of two-way dispatching. For example, when someone clicks the "Save" button, the event handler delegates to the Presenter's "OnSave" method. Once the save is completed, the Presenter will then call back the View through its interface so that the View can display that the save has completed.

MVP tends to be a very natural pattern for achieving separated presentation in WebForms. The reason is that the View is always created first by the ASP.NET runtime. You can find out more about both variants.

Two primary variations

Passive View: The View is as dumb as possible and contains almost zero logic. A Presenter is a middle man that talks to the View and the Model. The View and Model are completely shielded from one another. The Model may raise events, but the Presenter subscribes to them for updating the View. In Passive View there is no direct data binding, instead, the View exposes setter properties that the Presenter uses to set the data. All state is managed in the Presenter and not the View.

Pro: maximum testability surface; clean separation of the View and Model
Con: more work (for example all the setter properties) as you are doing all the data binding yourself.

Supervising Controller: The Presenter handles user gestures. The View binds to the Model directly through data binding. In this case, it's the Presenter's job to pass off the Model to the View so that it can bind to it. The Presenter will also contain logic for gestures like pressing a button, navigation, etc.

Pro: by leveraging data binding the amount of code is reduced.
Con: there's a less testable surface (because of data binding), and there's less encapsulation in the View since it talks directly to the Model.

Model-View-Controller

In the MVC, the Controller is responsible for determining which View to display in response to any action including when the application loads. This differs from MVP where actions route through the View to the Presenter. In MVC, every action in the View correlates with a call to a Controller along with an action. In the web, each action involves a call to a URL on the other side of which there is a Controller who responds. Once that Controller has completed its processing, it will return the correct View. The sequence continues in that manner throughout the life of the application:

    Action in the View
        -> Call to Controller
        -> Controller Logic
        -> Controller returns the View.

One other big difference about MVC is that the View does not directly bind to the Model. The view simply renders and is completely stateless. In implementations of MVC, the View usually will not have any logic in the code behind. This is contrary to MVP where it is absolutely necessary because, if the View does not delegate to the Presenter, it will never get called.

Presentation Model

One other pattern to look at is the Presentation Model pattern. In this pattern, there is no Presenter. Instead, the View binds directly to a Presentation Model. The Presentation Model is a Model crafted specifically for the View. This means this Model can expose properties that one would never put on a domain model as it would be a violation of separation-of-concerns. In this case, the Presentation Model binds to the domain model and may subscribe to events coming from that Model. The View then subscribes to events coming from the Presentation Model and updates itself accordingly. The Presentation Model can expose commands which the view uses for invoking actions. The advantage of this approach is that you can essentially remove the code-behind altogether as the PM completely encapsulates all of the behavior for the view. This pattern is a very strong candidate for use in WPF applications and is also called Model-View-ViewModel.

There is a MSDN article about the Presentation Model and a section in the Composite Application Guidance for WPF (former Prism) about Separated Presentation Patterns

Inversion of Control

The Inversion-of-Control (IoC) pattern, is about providing any kind of callback (which controls reaction), instead of acting ourself directly (in other words, inversion and/or redirecting control to external handler/controller). The Dependency-Injection (DI) pattern is a more specific version of IoC pattern, and is all about removing dependencies from your code.

Every DI implementation can be considered IoC, but one should not call it IoC, because implementing Dependency-Injection is harder than callback (Don't lower your product's worth by using general term "IoC" instead).

For DI example, say your application has a text-editor component, and you want to provide spell checking. Your standard code would look something like this:

public class TextEditor {

    private SpellChecker checker;

    public TextEditor() {
        this.checker = new SpellChecker();
    }
}

What we've done here creates a dependency between the TextEditor and the SpellChecker. In an IoC scenario we would instead do something like this:

public class TextEditor {

    private IocSpellChecker checker;

    public TextEditor(IocSpellChecker checker) {
        this.checker = checker;
    }
}

In the first code example we are instantiating SpellChecker (this.checker = new SpellChecker();), which means the TextEditor class directly depends on the SpellChecker class.

In the second code example we are creating an abstraction by having the SpellChecker dependency class in TextEditor's constructor signature (not initializing dependency in class). This allows us to call the dependency then pass it to the TextEditor class like so:

SpellChecker sc = new SpellChecker(); // dependency
TextEditor textEditor = new TextEditor(sc);

Now the client creating the TextEditor class has control over which SpellChecker implementation to use because we're injecting the dependency into the TextEditor signature.