问题
Is it generally acceptable to allow a Visitor to modify state of the Receiver, or should that be a Command pattern instead?
回答1:
The purpose of the visitor pattern is to allow new operations to be added to a class heirarchy without modification to that heirarchy. I've never seen anyone suggesting that only read-only operations are acceptable. The only limitation is that the added operations should only use the public interface of the class heirarchy.
回答2:
I don't think you can make a blanket statement whether it's good or bad to modify state of anything. I would think it's ok to modify the states as long as it does not affect the visiting logic itself. For example, you might write a visitor that visits all files under folder structure and renames the file name to upper case.
回答3:
Microsoft's example of a visitor modifying the receiver is the ExpressionVisitor. The purpose of the ExpressionVisitor class is to modify an Expression tree. So I guess Microsoft at least thinks it's acceptable.
回答4:
Each pattern has it's own pros, cons and use cases.
You can use Command pattern to
Decouple the invoker & receiver of command
Implement callback mechanism
Implement undo and redo functionality
Maintain a history of commands
Use Visitor pattern in below scenarios:
- Similar operations have to be performed on objects of different types grouped in a structure
- You need to execute many distinct and unrelated operations. It separates Operation from objects Structure
- New Operations have to be added without change in object structure
Related posts:
Using Command Design pattern
When should I use the Visitor Design Pattern?
来源：https://stackoverflow.com/questions/2857880/command-pattern-vs-visitor-pattern