问题
The discussion comes up here:
Changing visibility of method in inherited class
question is: is really "BTNode extends GraphNode" design a violation of Liskov's Substitution Princeple? As an "similar" example it was shown that case: Is deriving square from rectangle a violation of Liskov's Substitution Principle?
but I cannot really see why that is similar. I am very new to design, could someone explain me why (if) that's the case?
回答1:
In Is deriving square from rectangle a violation of Liskov's Substitution Principle?, it basically says that Square cannot inherit from Rectangle because there are things that you can do with Rectangles but not with Squares - setting its width to a different number from its height.
You can't inherit BTNode from GraphNode because according to the original post you linked, GraphNode has a method called addChild. BTNode on the other hand, can only have two children. Inheriting from GraphNode will also inherit the addChild method. This allows you to add multiple children to BTNode, which a BTNode cannot handle.
Therefore, BTNode cannot inherit from GraphNode because there are things that you can do with GraphNodes but not with BTNodes - adding multiple children.
For completeness, here is the Liskov's Substitution principle from Wikipedia
Subtype Requirement: Let ϕ(x) be a property provable about objects x of type T. Then ϕ ( y ) should be true for objects y of type S where S is a subtype of T.
In simple terms,
If you can do action X on type T, you should also be able to do action X on any subclasses of T.
来源：https://stackoverflow.com/questions/45897182/is-deriving-binarytreenode-from-graphnode-a-violation-of-liskovs-substitution-p