Despite the many threads on that topic, I am still unclear as to when to choose which approach.
I am hoping that by discussing a specific example, I will finally “get it.”
Note: My language here is Cocoa though the general issue is not language-specific.
I have a class TaskQueue that I want to use to:
- access from anywhere in my code in order to add or remove scheduled tasks
- process automatically the scheduled tasks at regular intervals
When TaskQueue is first used, I want TaskQueue to initiate a thread that will then wake up at regular intervals to process the tasks.
Obviously, I will need at a minimum two variables:
- an array to store the tasks
- an instance of the thread processing the tasks
Since I only want one queue of tasks and one thread to process these tasks, I have two choices:
-
Make TaskQueue a singleton class (using for example CWL_DECLARE_SINGLETON_FOR_CLASS_WITH_ACCESSOR as described in http://www.cocoawithlove.com/2008/11/singletons-appdelegates-and-top-level.html, which I believe I will have to modify the CWLSynthesizeSingleton.h file to start the thread at init time.)
-
Have the array of tasks and the thread instance both be static (following the approach suggested here: How do I declare class-level properties in Objective-C?)
Is there clearly one approach that’s better than the other one in this specific case? If so, why?
The main differences are simple things like:
If you don’t need any of them, as with global functionality that must be accessed all around your code then you can go with static methods.
I personally prefer using static methods unless I have an explicit reason to use a singleton instance (such as having a common interface but different implementations).
Mind the fact that refactoring static methods to a singleton instance is quite a straightforward process so if you ever find the need for the latter you will refactor it easily (then you have the C preprocessor, a single
#definewould be almost enough).