I have spent the last couple of hours reading about the singleton pattern and why not to use it, amongst others those really good sites:
Singleton I love you, but you're bringing me down
How to Think About the "new" Operator with Respect to Unit Testing
Where have all the Singletons Gone?
I guess quite a lot of you know these already.
Looking at my code after reading that, I clearly am one of the maybe 95% of programmers that misunderstood and misused the singleton pattern.
For some cases, I can clearly remove the pattern, but there are cases where I am unsure what to do:
I know singletons for logging are accepted, one reason for that being that information only flows into them but not back into the application (just into the log file or console etc of course).
What about other classes which do not meet that criteria but are required by a lot of classes?
For example, I have a settings object which is required by a lot of classes. By a lot, I mean more than 200.
I have read into some other SO questions like "Singletons: good design or a crutch?", and all of them pointed out why using singletons is discouraged.
I understand the reasons for that, but I still have one major question:
How do I design a class which needs a single instance, accessible from everywhere, if not using the Singleton pattern?
The options I can think of would be:
Use a static class instead (though I don't see how this would be any better, looking at OOP and unit testing).
Have it created in an ApplicationFactory and perform dependency injection on every single class that needs it (keep in mind it's 200+ for some cases).
Use a singleton anyway, as the global access bonus outweighs the disadvantages for that case.
Something completely different.
It will depend on exactly what you mean by a settings object.
Do all 200 classes need all the settings; if not why do they have access to the unused settings?
Where do the settings come from and is there a good reason why each class can't load its settings as and when required?
Most importantly though, don't make changes to working code just because the code uses a pattern which is frowned upon. I've only used the singleton pattern once but I'd use it again.
EDIT:
I don't know your constraints but I wouldn't worry about multiple access from a file until it had been shown to be an issue. I would split up the configuration into different files for different classes/ groups of classes or, preferably, use a DB instead of files with different tables providing data for each class.
As an aside I've noticed that once you put the data in a db people seem to stop worrying about accessing it multiple times even though you're still going to the file system in the end.
PS: If other options aren't suitable I'd use a singleton... you want data to be globally available, you're not willing to use dependency injection, you only want the file to be read once; you've limited your options and a singleton isn't that bad.
Isn't this already discussed extensively and exhaustingly?
There is no misuse of the pattern. If your software works as expected (inlcuding maintainability and testablility) you are right with singletons.
The thing about people complain is that the singleton pattern has more impact than only restrict a class to have a single instance.
you introduce a global variable
you cannot build a subclass
you cannot reset the instance
If all this is not a problem for you: Use singletons all over the place. The pattern discussion is academic and hairsplitting.
And - to answer your question - checkout the monostate vs singleton thread: Monostate vs. Singleton
Related
I've had a certain feeling these last couple of days that dependency-injection should really be called "I can't make up my mind"-pattern. I know this might sound silly, but really it's about the reasoning behind why I should use Dependency Injection (DI). Often it is said that I should use DI, to achieve a higher level of loose-coupling, and I get that part. But really... how often do I change my database, once my choice has fallen on MS SQL or MySQL .. Very rarely right?
Does anyone have some very compelling reasons why DI is the way to go?
Two words, unit testing.
One of the most compelling reasons for DI is to allow easier unit testing without having to hit a database and worry about setting up 'test' data.
DI is very useful for decoupling your system. If all you're using it for is to decouple the database implementation from the rest of your application, then either your application is pretty simple or you need to do a lot more analysis on the problem domain and discover what components within your problem domain are the most likely to change and the components within your system that have a large amount of coupling.
DI is most useful when you're aiming for code reuse, versatility and robustness to changes in your problem domain.
How relevant it is to your project depends upon the expected lifespan of your code. Depending on the type of work you're doing zero reuse from one project to the next for the majority of code you're writing might actually be quite acceptable.
An example for use the use of DI is in creating an application that can be deployed for several clients using DI to inject customisations for the client, which could also be described as the GOF Strategy pattern. Many of the GOF patterns can be facilitated with the use of a DI framework.
DI is more relevant to Enterprise application development in which you have a large amount of code, complicated business requirements and an expectation (or hope) that the system will be maintained for many years or decades.
Even if you don't change the structure of your program during development phases you will find out you need to access several subsystems from different parts of your program. With DI each of your classes just needs to ask for services and you're free of having to provide all the wiring manually.
This really helps me on concentrating on the interaction of things in the software design and not on "who needs to carry what around because someone else needs it later".
Additionally it also just saves a LOT of work writing boilerplate code. Do I need a singleton? I just configure a class to be one. Can I test with such a "singleton"? Yes, I still can (since I just CONFIGURED it to exist only once, but the test can instantiate an alternative implementation).
But, by the way before I was using DI I didn't really understand its worth, but trying it was a real eye-opener to me: My designs are a lot more object-oriented as they have been before.
By the way, with the current application I DON'T unit-test (bad, bad me) but I STILL couldn't live with DI anymore. It is so much easier moving things around and keeping classes small and simple.
While I semi-agree with you with the DB example, one of the large things that I found helpful to use DI is to help me test the layer I build on top of the database.
Here's an example...
You have your database.
You have your code that accesses the database and returns objects
You have business domain objects that take the previous item's objects and do some logic with them.
If you merge the data access with your business domain logic, your domain objects can become difficult to test. DI allows you to inject your own data access objects into your domain so that you don't depend on the database for testing or possibly demonstrations (ran a demo where some data was pulled in from xml instead of a database).
Abstracting 3rd party components and frameworks like this would also help you.
Aside from the testing example, there's a few places where DI can be used through a Design by Contract approach. You may find it appropriate to create a processing engine of sorts that calls methods of the objects you're injecting into it. While it may not truly "process it" it runs the methods that have different implementation in each object you provide.
I saw an example of this where the every business domain object had a "Save" function that the was called after it was injected into the processor. The processor modified the component with configuration information and Save handled the object's primary state. In essence, DI supplemented the polymorphic method implementation of the objects that conformed to the Interface.
Dependency Injection gives you the ability to test specific units of code in isolation.
Say I have a class Foo for example that takes an instance of a class Bar in its constructor. One of the methods on Foo might check that a Property value of Bar is one which allows some other processing of Bar to take place.
public class Foo
{
private Bar _bar;
public Foo(Bar bar)
{
_bar = bar;
}
public bool IsPropertyOfBarValid()
{
return _bar.SomeProperty == PropertyEnum.ValidProperty;
}
}
Now let's say that Bar is instantiated and it's Properties are set to data from some datasource in it's constructor. How might I go about testing the IsPropertyOfBarValid() method of Foo (ignoring the fact that this is an incredibly simple example)? Well, Foo is dependent on the instance of Bar passed in to the constructor, which in turn is dependent on the data from the datasource that it's properties are set to. What we would like to do is have some way of isolating Foo from the resources it depends upon so that we can test it in isolation
This is where Dependency Injection comes in. What we want is to have some way of faking an instance of Bar passed to Foo such that we can control the properties set on this fake Bar and achieve what we set out to do, test that the implementation of IsPropertyOfBarValid() does what we expect it to do, i.e. return true when Bar.SomeProperty == PropertyEnum.ValidProperty and false for any other value.
There are two types of fake object, Mocks and Stubs. Stubs provide input for the application under test so that the test can be performed on something else. Mocks on the other hand provide input to the test to decide on pass\fail.
Martin Fowler has a great article on the difference between Mocks and Stubs
I think that DI is worth using when you have many services/components whose implementations must be selected at runtime based on external configuration. (Note that such configuration can take the form of an XML file or a combination of code annotations and separate classes; choose what is more convenient.)
Otherwise, I would simply use a ServiceLocator, which is much "lighter" and easier to understand than a whole DI framework.
For unit testing, I prefer to use a mocking API that can mock objects on demand, instead of requiring them to be "injected" into the tested unit from a test. For Java, one such library is my own, JMockit.
Aside from loose coupling, testing of any type is achieved with much greater ease thanks to DI. You can put replace an existing dependency of a class under test with a mock, a dummy or even another version. If a class is created with its dependencies directly instantiated it can often be difficult or even impossible to "stub" them out if required.
I just understood tonight.
For me, dependancy injection is a method for instantiate objects which require a lot of parameters to work in a specific context.
When should you use dependancy injection?
You can use dependancy injection if you instanciate in a static way an object. For example, if you use a class which can convert objects into XML file or JSON file and if you need only the XML file. You will have to instanciate the object and configure a lot of thing if you don't use dependancy injection.
When should you not use depandancy injection?
If an object is instanciated with request parameters (after a submission form), you should not use depandancy injection because the object is not instanciated in a static way.
My app contains several singletons (following from this tutorial). I've noticed however, when the app crashes because of a singleton, it becomes nearly impossible to figure out where it came from. The app breakpoints at the main function giving an EXEC_BAD_ACCESS even though the problem lies in one of the Singleton objects. Is there a guide to how would I debug my singleton objects if they were problematic?
if you don't want to change your design (as recommended in my other post), then consider the usual debugging facilities: assertions, unit tests, zombie tests, memory tests (GuardMalloc, scribbling), etc. this should identify the vast majority of issues one would encounter.
of course, you will have some restrictions regarding what you can and cannot do - notably regarding what cannot be tested independently using unit tests.
as well, reproducibility may be more difficult in some contexts when/if you are dealing with a complex global state because you have created several enforced singletons. when the global state is quite large and complex - testing these types independently may not be fruitful in all cases since the bug may appear only in a complex global state found in your app (when 4 singletons interact in a specific manner). if you have isolated the issue to interactions of multiple singleton instances (e.g. MONAudioFileCache and MONVideoCache), placing these objects in a container class will allow you to introduce coupling, which will help diagnose this. although increasing coupling is normally considered a bad thing; this does't really increase coupling (it already exists as components of the global state) but simply concentrates existing global state dependencies -- you're really not increasing it as much as you are concentrating it when the state of these singletons affect other components of the mutable global state.
if you still insist on using singletons, these may help:
either make them thread safe or add some assertions to verify mutations happen only on the main thread (for example). too many people assume an object with atomic properties implies the object is thread safe. that is false.
encapsulate your data better, particularly that which mutates. for example: rather than passing out an array your class holds for the client to mutate, have the singleton class add the object to the array it holds. if you truly must expose the array to the client, then return a copy. ths is just basic ood, but many objc devs expose the majority of their ivars disregarding the importance of encapsualtion.
if it's not thread safe and the class is used in a mutithreaded context, make the class (not the client) implement proper thread safety.
design singletons' error checking to be particularly robust. if the programmer passes an invalid argument or misuses the interface - just assert (with a nice message about the problem/resolution).
do write unit tests.
detach state (e.g. if you can remove an ivar easily, do it)
reduce complexity of state.
if something is still impossible to debug after writing/testing with thorough assertions, unit tests, zombie tests, memory tests (GuardMalloc, scribbling), etc,, you are writing programs which are too complex (e.g. divide the complexity among multiple classes), or the requirements do not match the actual usage. if you're at that point, you should definitely refer to my other post. the more complex the global variable state, the more time it will take to debug, and the less you can reuse and test your programs when things do go wrong.
good luck
I scanned the article, and while it had some good ideas it also had some bad advice, and it should not be taken as gospel.
And, as others have suggested, if you have a lot of singleton objects it may mean that you're simply keeping too much state global/persistent. Normally only one or two of your own should be needed (in addition to those that other "packages" of one sort or another may implement).
As to debugging singletons, I don't understand why you say it's hard -- no worse than anything else, for the most part. If you're getting EXEC_BAD_ACCESS it's because you've got some sort of addressing bug, and that's nothing specific to singleton schemes (unless you're using a very bad one).
Macros make debugging difficult because the lines of code they incorporate can't have breakpoints put in them. Deep six macros, if nothing else. In particular, the SYNTHESIZE_SINGLETON_FOR_CLASS macro from the article is interfering with debugging. Replace the call to this macro function with the code it generates for your singleton class.
ugh - don't enforce singletons. just create normal classes. if your app needs just one instance, add them to something which is created once, such as your app delegate.
most cocoa singleton implementations i've seen should not have been singletons.
then you will be able to debug, test, create, mutate and destroy these objects as usual.
the good part is course that the majority of your global variable pains will disappear when you implement these classes as normal objects.
When do you encourage programming against an interface and not directly to a concrete class?
A guideline that I follow is to create abstractions whenever code requires to cross a logical/physical boundary, most especially when infrastructure-related concerns are involved.
Another checkpoint would be if a dependency will likely change in the future, due to possible additional concerns code (such as caching, transactional awareness, invoking a webservice instead of in-process execution) or if such dependencies have direct references to infrastructure integration points.
If code depends on something that does not require control to cross a logical/physical boundary, I more or less don't create abstractions to interact with those.
Am I missing anything?
Also, use interfaces when
Multiple objects will need to be acted upon in a particular fashion, but are not fundamentally related. Perhaps many of your business objects access a particular utility object, and when they do they need to give a reference of themselves to that utility object so the utility object can call a particular method. Have that method in an interface and pass that interface to that utility object.
Passing around interfaces as parameters can be very helpful in unit testing. Even if you have just one type of object that sports a particular interface, and hence don't really need a defined interface, you might define/implement an interface solely to "fake" that object in unit tests.
related to the first 2 bullets, check out the Observer pattern and the Dependency Injection. I'm not saying to implement these patterns, but they illustrate types of places where interfaces are really helpful.
Another twist on this is for implementing a couple of the SOLID Principals, Open Closed principal and the Interface Segregation principle. Like the previous bullet, don't get stressed about strictly implementing these principals everywhere (right away at least), but use these concepts to help move your thinking away from just what objects go where to thinking more about contracts and dependency
In the end, let's not make it too complicated: we're in a strongly typed world in .NET. If you need to call a method or set a property but the object you're passing/using could be fundamentally different, use an interface.
I would add that if your code is not going to be referenced by another library (for a while at least), then the decision of whether to use an interface in a particular situation is one that you can responsibly put off. The "extract interface" refactoring is easy to do these days. In my current project, I've got an object being passed around that I'm thinking maybe I should switch to an interface; I'm not stressing about it.
Interfaces abstraction are convenient when doing unit test. It helps for mocking test objects. It very useful in TDD for developing without actually using data from your database.
If you don't need any features of the class that aren't found in the Interface...then why not always prefer the Interface implementation?
It will make your code easier to modify in the future and easier to test (mocking).
you have the right idea, already. i would only add a couple of notes to this...
first, abstraction does not mean 'interface'. for example, a "connection string" is an abstraction, even though it's just a string... it's not about the 'type' of the thing in question, it's about the intention of use for that thing.
and secondly, if you are doing test automation of any kind, look for the pain and friction that are exposed by writing the tests. if you find yourself having to set up too many external conditions for a test, it's a sign that you need a better abstraction between the thing your testing and the things it interacts with.
I think you've said it pretty well. Much of this will be a stylistic thing. There are open source projects I've looked at where everything has an interface and an implementation, and it's kind of frustrating, but it might make iterative development a little easier, since any objects implementation can break but dummies will still work. But honestly, I can dummy any class that doesn't overuse the final keyword by inheritance.
I would add to your list this: anything which can be thought of as a black box should be abstracted. This includes some of the things you've mentioned, but it also includes hairy algorithms, which are likely to have multiple useful implementations with different advantages for different situation.
Additionally, interfaces come in handy very often with composite objects. That's the only way something like java's swing library gets anything done, but it can also be useful for more mundane objects. (I personally like having an interface like ValidityChecker with ways to and-compose or or-compose subordinate ValidityCheckers.)
Most of the useful things that come with the Interface passing have been already said. However I would add:
implementing an interface to an object, or later multiple objects, FORCES all the implementers to follow an IDENTICAL pattern to implement contract with the object. This can be useful in case you have not so OOP-experienced-programmers actually writing the implementation code.
in some languages you can add attributes on the interface itself, which can be different from the actual object implementation attribute as sense and intent
I've inherited a project that stores various parameters either in a config file, the registry and a database. Whoever needs one of these parameters just reads (and in some cases writes) it directly from the store. This, or course, is stupid, so my first thought was to refactor the existing code so that the client doesn't know where the parameter is stored in. I created a classic AppSettings class that has a property for each parameter. Since the store has to have global scope I made a thread-safe singleton. The class doesn't store the parameter values in fields but rather acts as an access point by reading and writing them to and from the actual store, be it config file, registry or database. These days it's hard to avoid all the talk about the dangers of singletons and global state. I will take a proper look at dependency injection and Spring etc later, but for now, I just have a couple of questions.
What type of problems, other than testability, can you see with my solution?
What would be a light weight alternative? Creating a factory for each object that uses the parameters is not an option (too much work).
Wouldn't using a singleton serve as an acceptable compromise until I have a chance to do some heavier refactoring?
If the properties in my singleton class only had getters, would that make it OK?
I can anticipate that the store for some of the parameters will change in the future (eg. from registry to database), so that was my motivation for hiding the store behind a singleton class.
This is a bit of a non-answer, but I highly recommend the c2wiki's pages on Singletons as a reference http://c2.com/cgi/wiki?search=Singleton
And also the page http://c2.com/cgi/wiki?GlobalVariablesAreBad
I think the general verdict is that global state creates coupling between vastly different parts of your system that must be thought about and designed around very carefully. The question is, are all of those settings truly global and needed by disparate parts of the system? If not, then is there any way to separate them into smaller parts that can live inside different modules at a lower access level?
If it's a small project I wouldn't worry too much about it, but there is a lot of wisdom on those c2wiki pages about global state and singletons being a pain for larger projects.
I would challenge the assumption that since the config data is global, that you need a global singleton to access it, especially for reading. Consider creating an AppSettings class that can be invoked as needed to read your config settings.
If you need to write in a thread-safe manner, you can create a static (or singleton) private member of the AppSettings class to control writing only. Thus any instance of AppSettings can write, but the "global" access is actually restricted to the AppSettings class.
Thanks, guys. I would consider this a medium-size project (about 200KLOC) and it's C#. The problem is that the project has a long and troubled history and a lot coders have worked on it. As much as I'd like to properly learn dependency injection (as I do understand and subscribe to the concept), the deadline is closing fast so now is not the time for it. After looking at my current singleton class I decided to split it into two instance classes. Some of the parameters are used all over but some only in a single assembly. And like Doug said, I can achieve thread-safety just as easily with an instance class.
As for the various dependency injection frameworks, the problem is there are too many. I've briefly looked at Spring and Unity. I wish I could find a summary of the differences.
Thanks again!
This is a question with many answers - I am interested in knowing what others consider to be "best practice".
Consider the following situation: you have an object-oriented program that contains one or more data structures that are needed by many different classes. How do you make these data structures accessible?
You can explicitly pass references around, for example, in the constructors. This is the "proper" solution, but it means duplicating parameters and instance variables all over the program. This makes changes or additions to the global data difficult.
You can put all of the data structures inside of a single object, and pass around references to this object. This can either be an object created just for this purpose, or it could be the "main" object of your program. This simplifies the problems of (1), but the data structures may or may not have anything to do with one another, and collecting them together in a single object is pretty arbitrary.
You can make the data structures "static". This lets you reference them directly from other classes, without having to pass around references. This entirely avoids the disadvantages of (1), but is clearly not OO. This also means that there can only ever be a single instance of the program.
When there are a lot of data structures, all required by a lot of classes, I tend to use (2). This is a compromise between OO-purity and practicality. What do other folks do? (For what it's worth, I mostly come from the Java world, but this discussion is applicable to any OO language.)
Global data isn't as bad as many OO purists claim!
After all, when implementing OO classes you've usually using an API to your OS. What the heck is this if it isn't a huge pile of global data and services!
If you use some global stuff in your program, you're merely extending this huge environment your class implementation can already see of the OS with a bit of data that is domain specific to your app.
Passing pointers/references everywhere is often taught in OO courses and books, academically it sounds nice. Pragmatically, it is often the thing to do, but it is misguided to follow this rule blindly and absolutely. For a decent sized program, you can end up with a pile of references being passed all over the place and it can result in completely unnecessary drudgery work.
Globally accessible services/data providers (abstracted away behind a nice interface obviously) are pretty much a must in a decent sized app.
I must really really discourage you from using option 3 - making the data static. I've worked on several projects where the early developers made some core data static, only to later realise they did need to run two copies of the program - and incurred a huge amount of work making the data non-static and carefully putting in references into everything.
So in my experience, if you do 3), you will eventually end up doing 1) at twice the cost.
Go for 1, and be fine-grained about what data structures you reference from each object. Don't use "context objects", just pass in precisely the data needed. Yes, it makes the code more complicated, but on the plus side, it makes it clearer - the fact that a FwurzleDigestionListener is holding a reference to both a Fwurzle and a DigestionTract immediately gives the reader an idea about its purpose.
And by definition, if the data format changes, so will the classes that operate on it, so you have to change them anyway.
You might want to think about altering the requirement that lots of objects need to know about the same data structures. One reason there does not seem to be a clean OO way of sharing data is that sharing data is not very object-oriented.
You will need to look at the specifics of your application but the general idea is to have one object responsible for the shared data which provides services to the other objects based on the data encapsulated in it. However these services should not involve giving other objects the data structures - merely giving other objects the pieces of information they need to meet their responsibilites and performing mutations on the data structures internally.
I tend to use 3) and be very careful about the synchronisation and locking across threads. I agree it is less OO, but then you confess to having global data, which is very un-OO in the first place.
Don't get too hung up on whether you are sticking purely to one programming methodology or another, find a solution which fits your problem. I think there are perfectly valid contexts for singletons (Logging for instance).
I use a combination of having one global object and passing interfaces in via constructors.
From the one main global object (usually named after what your program is called or does) you can start up other globals (maybe that have their own threads). This lets you control the setting up of program objects in the main objects constructor and tearing them down again in the right order when the application stops in this main objects destructor. Using static classes directly makes it tricky to initialize/uninitialize any resources these classes use in a controlled manner. This main global object also has properties for getting at the interfaces of different sub-systems of your application that various objects may want to get hold of to do their work.
I also pass references to relevant data-structures into constructors of some objects where I feel it is useful to isolate those objects from the rest of the world within the program when they only need to be concerned with a small part of it.
Whether an object grabs the global object and navigates its properties to get the interfaces it wants or gets passed the interfaces it uses via its constructor is a matter of taste and intuition. Any object you're implementing that you think might be reused in some other project should definately be passed data structures it should use via its constructor. Objects that grab the global object should be more to do with the infrastructure of your application.
Objects that receive interfaces they use via the constructor are probably easier to unit-test because you can feed them a mock interface, and tickle their methods to make sure they return the right arguments or interact with mock interfaces correctly. To test objects that access the main global object, you have to mock up the main global object so that when they request interfaces (I often call these services) from it they get appropriate mock objects and can be tested against them.
I prefer using the singleton pattern as described in the GoF book for these situations. A singleton is not the same as either of the three options described in the question. The constructor is private (or protected) so that it cannot be used just anywhere. You use a get() function (or whatever you prefer to call it) to obtain an instance. However, the architecture of the singleton class guarantees that each call to get() returns the same instance.
We should take care not to confuse Object Oriented Design with Object Oriented Implementation. Al too often, the term OO Design is used to judge an implementation, just as, imho, it is here.
Design
If in your design you see a lot of objects having a reference to exactly the same object, that means a lot of arrows. The designer should feel an itch here. He should verify whether this object is just commonly used, or if it is really a utility (e.g. a COM factory, a registry of some kind, ...).
From the project's requirements, he can see if it really needs to be a singleton (e.g. 'The Internet'), or if the object is shared because it's too general or too expensive or whatsoever.
Implementation
When you are asked to implement an OO Design in an OO language, you face a lot of decisions, like the one you mentioned: how should I implement all the arrows to the oft used object in the design?
That's the point where questions are addressed about 'static member', 'global variable' , 'god class' and 'a-lot-of-function-arguments'.
The Design phase should have clarified if the object needs to be a singleton or not. The implementation phase will decide on how this singleness will be represented in the program.
Option 3) while not purist OO, tends to be the most reasonable solution. But I would not make your class a singleton; and use some other object as a static 'dictionary' to manage those shared resources.
I don't like any of your proposed solutions:
You are passing around a bunch of "context" objects - the things that use them don't specify what fields or pieces of data they are really interested in
See here for a description of the God Object pattern. This is the worst of all worlds
Simply do not ever use Singleton objects for anything. You seem to have identified a few of the potential problems yourself