When would a singleton actually be easier or better than a static class? It seems to me creating a singleton is just extra effort that's not actually needed, but I'm sure there is a good reason. Otherwise, they wouldn't be used, obviously.
Singletons are often preferred to global variables because:
- They don't pollute the global namespace (or, in languages with namespaces, their containing namespace) with unnecessary variables.
- They permit lazy allocation and initialization, whereas global variables in many languages will always consume resources.
EDIT:
One cool use of the singleton is, when combined with the factory method, can be used to create the Flyweight pattern. This is when you create a new object, the Factory (instead of creating a new object) first checks to see a singleton of that object is already made, if it is, it just returns that object, if not, it creates a new singleton and returns that, keeping track of the singletons it creates. Flyweights work because of the immutability of the singleton.
In many languages static classes lack useful features, like inheritance (and polymorphism more generally).
Singletons preserve the conventional class approach, and don't require that you use the static keyword everywhere. They may be more demanding to implement at first, but will greatly simplify the architecture of your program. Unlike static classes, we can use singletons as parameters or objects.
Also,you can use singletons with interfaces just like any other class.
Even though IMHO the singleton pattern is a rather overused pattern, it does sometimes offer benefits, such as:
- The ability to use different kinds of objects (that inherit from the same base) as the instance (think data providers where one uses the filesystem and one an SQL database, for example)
- Serializability. I haven't used a framework that can automatically serialize static classes.
- Using less static fields. For some people this is more of an esthetic feature, but in some cases it has practical benefits.
One good reason for preferring a singleton over a static class (assuming you have no better patterns at your disposal ;) ), is swapping out one singleton instance with another.
For example, if I have a logging class like this:
public static class Logger {
public static void Log(string s) { ... }
}
public class Client {
public void DoSomething() {
Logger.Log("DoSomething called");
}
}
It works really well, but what if Logger writes things to a database, or writes output to the console. If you're writing tests, you probably don't want all those side-effects -- but since the log method is static, you can't do anything except.
Ok, so I want to hot-swap my Log method for testing. Go go gadget singleton!
public class Logger {
private static Logger _instance;
public static Logger Instance
{
get
{
if (_instance == null)
_instance = new Logger();
return _instance;
}
set { _instance = value; }
}
protected Logger() { }
public virtual void Log(string s) { ... }
}
public class Client {
public void DoSomething() {
Logger.Instance.Log("DoSomething called");
}
}
So you can define a TestLogger : Logger
with an empty Log
method, then set an instance of your test logger to the singleton instance for tests. Presto! You can hotswap your logger implementation for tests or production without affecting client code.