I'm in the process of designing a .NET library and I'm looking for some feedback concerning class organization and namespaces.
Suppose that I have an abstract Message
class that is in the root namespace Foo
. Message
has many concrete implementations for different message types. Given that there will be many concrete implementations (let's assume 20, or even more), is it expected that these concrete types live in a separate namespace, something like Foo.Messages
?
I'm worried about cluttering the namespace with too many concrete Message
classes that users of my library will rarely need to use directly. What are you expectations when using a library?
edit:
Also, if I do group the concrete classes into Foo.Messages
, should the abstract class live there as well, or should that remain in Foo
?
Both interfaces and abstract classes cannot be instantiated, but due to polymorphism, an instance of a class which either implements an interface or inherits from an abstract class can be treated as though it is said interface or abstract class.
Interface Namespace All Superinterfaces: Attribute, XMLEvent, XMLStreamConstants. public interface Namespace extends Attribute. An interface that contains information about a namespace. Namespaces are accessed from a StartElement.
An abstract class permits you to make functionality that subclasses can implement or override whereas an interface only permits you to state functionality but not to implement it. A class can extend only one abstract class while a class can implement multiple interfaces.
Abstract classes do not support multiple inheritance.
If the users of your library won't be constructing your concrete implementation classes (they go through a factory pattern to get instances, and access the instances only through the abstract base type), consider making your concrete implementation classes internal. This will keep them out of the user's code completion choices.
Grouping all the concrete implementation classes under a Foo.Messages namespace sounds fine. You don't want to go too far the other way either - creating a lot of namespaces for minor things is a huge irritation to the user.
Kind of a loaded question, but if the issue is having too many classes hampering discoverability, is part of the solution perhaps making some of these classes internal
rather than public
? If users won't ever instantiate these directly, that might be part of the solution to your clutter problem.
It really does depend quite a lot on the nature of the API you're designing, and to some degree on taste. Often there's a single primary class or set of classes that your users will be going straight to most of the time, like NLog's Logger
, or AutoMapper's Mapper
, or NoRM's Mongo
, or class factories in general. Those should be front and center and discoverable, and you could make a strong case that it's best to have that in your root namespace.
For other classes (ones that are more about the implementation and less about the API), obviously you want to be organized, but you can play with a pretty free hand and organize in a way that feels natural to you, as long as the parts of the API that users need to find are the most visible. Foo.Messages
seems like a totally reasonable way to start. On the other hand, if 90% of your classes are Message
subclasses, but there's an important distinction between Server messages and Client messages, or Purple messages and Plaid messages, maybe Foo.Server
or Foo.Plaid
are the right kinds of namespaces for you.
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With