Re: OO Concept: Liskov Substitution Principle

howa wrote:
Hi,

Just read an article talking about the LSP in term of OO design:
http://www.objectmentor.com/resources/articles/lsp.pdf

The article said Rectangle class should not be a superclass of Square
class.

Okay, so how you would design the Rectangle & Square class?


The article did not cover that, so what are your opinion?

First things first, I am going to stop talking about a hierarchy whose practical importance is practically nil and bring in a more useful example.

Let us assume we are writing a linear algebra package. Quite naturally, at some point we have a Matrix class. While writing Matrix.determinant(), we realize that it may be to our advantage to refactor said method into a SquareMatrix.determinant() for ease of design. Should we do so, i.e., should we create a SquareMatrix class that descends from Matrix?

I propose the following answer: yes, since Matrix here is assumed to be immutable. The problem with compatibility is that SquareMatrix assumes that its two dimensions are equal; since these dimensions can only be affected at initialization, thereby rendering this contractual incompatibility moot.

However, in the end, we decide that requiring a separate SquareMatrix class is pointless since a 3x2 matrix multiplied by a 2x3 matrix is a square matrix even if its type is merely a banal Matrix, rendering the entire discussion moot.


--
Beware of bugs in the above code; I have only proved it correct, not tried it. -- Donald E. Knuth
.

Relevant Pages