1 / 65

Unit 1 General Introduction

Unit 1 General Introduction. Summary prepared by Kirk Scott. Design Patterns in Java Part V Extension Patterns Chapter 26 Introducing Extensions. Summary prepared by Kirk Scott. The Introduction Before the Introduction.

frayne
Download Presentation

Unit 1 General Introduction

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Unit 1General Introduction Summary prepared by Kirk Scott

  2. Design Patterns in JavaPart VExtension PatternsChapter 26Introducing Extensions Summary prepared by Kirk Scott

  3. The Introduction Before the Introduction • It seems to me that this chapter really doesn’t make clear what the extension patterns are • In part, the book uses it to do a little review of specific patterns that have been introduced so far • That isn’t of too much value • What is interesting, though, is the introduction of some more object-oriented design principles

  4. Concepts like encapsulation come up early in OO design, and recur in this book • In addition, we have seen many practical applications of interfaces, polymorphism, and dynamic binding in this book • There was a whole section on the topic of responsibility, and how distributed responsibility is the norm, but responsibility can also be usefully concentrated in some cases

  5. In numerous patterns, the related topic of coupling also arose • Loose coupling is generally desirable • Ideally, objects are independent of each other • Sometimes a change in one will trigger a change in another

  6. Loose coupling might be thought of as meaning fewer connections • This would mean fewer places where you had to worry about a change in one propagating to another • As usual, there are also cases where tighter coupling is desirable

  7. This chapter adds two major items to the list of object-oriented design principles: • The Liskov Substitution Principle • The Law of Demeter • It’s not immediately apparent that these concepts are directly related to any one of the design patterns under consideration • However, they are interesting in their own right, and draw together some important insights that apply to OO design in general, and might apply to the development of any one design pattern

  8. Introducing Extensions • Book material: • In general, extension is based on this idea: • You rarely write code in a vacuum • You have an existing code base and your task is to add new features, classes, applications to it • In this sense, the term extension basically just means programming within the context of given code

  9. The book’s motivation for introducing general object-oriented design principles at this point is the following: • As a programmer, extending a code base, it would be helpful to have some rules of thumb to decide whether or not the code you are adding is any good

  10. Extension may also refer to specific techniques for adding features to a code base • In addition to looking at general principles, the book does two other things in this chapter: • It looks back at some earlier patterns that have an element of extension to them • It previews some of the extension design patterns that will be covered in the rest of the section

  11. Principles of Object-Oriented Design • For people interested in the topic of patterns, the book cites this Web site: www.c2.com • There is no need to go there for the purposes of this course • From that page there is a link to a Wiki with a collection of patterns that goes beyond the ones covered in the book

  12. The LiskovSubsitution Principle (LSP) • One of the aspects of object-orientation that should be familiar to you is the following: • A subclass is in an “is-a-kind-of” relationship with its superclass • In general, a subclass is a more specific kind of the superclass • At the very least, it has the same set of instance variables, since instance variables can’t (or shouldn’t) be overridden

  13. The subclass is made more specific by having more instance variables • It is also made more specific by overriding methods, if necessary • In Java, it’s always syntactically possible to have a superclass reference to a subclass object

  14. The superclass reference won’t have access to the instance variables and unique methods of the subclass • However, the superclass reference will have access to any methods inherited by the subclass • It will also have access to any methods overridden in the subclass

  15. The ability to have superclass references is important when defining method parameters in an inheritance hierarchy • A method defined in a superclass with a superclass explicit parameter doesn’t have to be overridden in the subclass, with the explicit parameter re-typed to the subclass, in order to be used with a subclass explicit parameter

  16. The book states that subclasses should be logical, consistent extensions of their superclasses • Trying to define the meaning of logical and consistent leads to the Liskov Substitution Principle • The book paraphrases the principle as follows: • “An instance of a class should function as an instance of its superclass.”

  17. As noted, Java expects this and supports it syntactically • The real question is whether or not the internal logic of classes for a given problem domain agree with this • The book gives two problematic examples on which are shown on the next overhead

  18. The Machine/UnloadBuffer example illustrates questions that arise in the problem domain • Superficially, you might want to class UnLoadBuffers with Machines • However, the book notes the following: • According to the way their factory is set up, you can’t add tubs to an UnloadBuffer, which implies that you also could not get tubs associated with an UnloadBuffer

  19. If all other machines have this characteristic, you have a design problem • Do you force the implementation of addTub() and getTubs() into each subclass of machine, when all of that code will be duplicate? • Do you implement the UnloadBuffer class outside of the Machine hierarchy? • Do you leave the UnloadBuffer in the hierarchy and leave the methods in Machine?

  20. The UML diagram shows the last choice, with addTub() and getTubs() overridden in UnloadBuffer • What should you do if it is not possible to call these methods on an UnloadBuffer, throw an exception? • This puts the burden on client code to catch exceptions for a faulty design

  21. An alternative would be to have the overridden versions of the method do nothing • This is even worse • What happens to a tub after a call has been made to add it to an UnloadBuffer? • Does it go into outer space?

  22. The point is that this is a bad design, and the problem with the design can be identified as a violation of the Liskov Substitution Principle • The problem is that it is not possible to use an instance of an UnloadBuffer anywhere that you use an instance of a machine • The UnloadBuffer class doesn’t seem to be a proper subclass of the Machine class

  23. The book continues by observing that it may sometimes be useful to violate the principle • This seems to be eternally true in software development • The additional requirement when you do this is to clearly identify the violation • Justify it • And identify and deal with all possible consequences

  24. In the chapter on bridges/drivers, a somewhat analogous situation arose • The choice had to do with how many methods you put into an interface • Did you put a subset of methods in, so that all implementing classes could implement them? • Or did you put a superset of methods in, forcing some of the implementing classes to have bogus implementations?

  25. The second option was more flexible, in the sense that all possible methods were supported in all cases where they were valid • The shortcoming was that they also existed in cases where they weren’t valid • As long as the invalid cases didn’t cause any harm—or it was possible to detect and deal with invalid cases in client code, then the advantages might outweigh the disadvantages

  26. Challenge 26.1 • A circle is certainly a special case of an ellipse, or is it? Say whether the relationship of the Ellipse and Circle classes in Figure 26.1 is a violation of LSP.

  27. Solution 26.1 • In mathematics, a circle is certainly a special case of an ellipse. However in OO programming, an ellipse has certain behaviors that a circle does not. For example, an ellipse may be twice as wide as it is tall; a circle can’t do that. If that behavior is important to your program, a Circle object won’t function as an Ellipse object and will represent a violation of LSP.

  28. Note that if you’re considering immutable objects, this may not be apply. This is simply an area in which naïve mathematics is not a smooth fit for the semantics of standard type hierarchies.

  29. The Law of Demeter (LoD) • Demeter • From Wikipedia, the free encyclopedia

  30. In Greek mythologyDemeter (pronounced /dəˈmiːtər/; də-MEE-tər; Greek: Δημήτηρ, Dēmētēr, probably "earth-mother")[1][2] was the goddess of the harvest, who presided over grains, the fertility of the earth, the seasons (personified by the Hours), and the harvest. Though Demeter is often described simply as the goddess of the harvest, she presided also over the sanctity of marriage, the sacred law, and the cycle of life and death. She and her daughter Persephone were the central figures of the Eleusinian Mysteries that also predated the Olympian pantheon. • Her Roman cognate is Ceres.

  31. The book quotes a paper by Lieberherr and Holland on the Law of Demeter: • “Informally, the law [of Demeter] says that each method can send messages to only a limited set of objects: to argument objects, to the [this] pseudovariable, and to the immediate subparts of [this].”

  32. Stated more explicitly, the law goes like this: • When writing method code, you may be working with the implicit parameter and an explicit parameter • You can call methods directly on the implicit parameter • You can also call methods on the instance variables of the implicit parameter

  33. Keep in mind that the method you’re writing is in the class of the implicit parameter, so you have direct access to the instance variables of the implicit parameter • You can also call methods directly on the explicit parameter • However, you can’t call methods on the instance variables of the explicit parameter

  34. In summary, you’re allowed to call methods on objects that you have a direct reference to • You shouldn’t be calling methods second-hand • In other words, if you have to call a get method to get a reference to an instance variable of something, you shouldn’t then be calling a method on that object

  35. You may recall that I was complaining about a violation of encapsulation in the last chapter • In CS 202 it was pointed out that you can avoid some encapsulation violation problems if get methods for instance variables that are references return clones • On the other hand, in the last chapter the problem was that client code unavoidably retained a reference to an object that had been sent in as a construction parameter to another object

  36. In the scenario under consideration for the Law of Demeter, there may be no violation of encapsulation • Or a formal violation of encapsulation could be fixed by returning a clone • However, the law still applies, and that’s why it’s not simply a formal violation of encapsulation

  37. The book points out that this can be easier to understand if you consider cases that would violate the Law of Demeter • They set up this scenario • There is a MaterialManager object with a method X that takes a Tub object (named tub in the book) as an explicit parameter • Tub objects have a location instance variable

  38. A location is a machine, for example • A machine has an instance variable that tells whether the machine is up or down • In the code for method X, you could write code like this: • if(tub.getLocation().isUp()){ • … • }

  39. According to the Law of Demeter, it’s OK to call tub.getLocation() • That’s just one level deep into the explicit parameter • That call returns a reference to the location instance variable of the tub • The Law of Demeter says that calling a method on that reference is not OK

  40. Challenge 26.2 • Explain why the expression tub.getLocation().isUp() might be viewed as unhealthy.

  41. Solution 26.2 • The expression tub.getLocation().isUp() might lead to programming errors if there are any subtleties around the value of a tub object’s Location property. For example, the location might be null or might be a Robot object if the tub is in transit. If location is null, evaluating tub.getLocation().isUp() will throw an exception.

  42. If the location is a Robot object, the problem may be even worse, as we try to use a robot to collect a tub from itself. These potential problems are manageable, but do we want the ensuing code to be in the method that uses the tub.getLocation().isUp() expression? No. The necessary code may already be in the Tub class! If not, it belongs there. To prevent us from having to recode around the same subtleties in other methods that interact with tubs.

  43. The book’s challenge solution is complete and correct • It might benefit from a summary • If method X deals with tubs, then it would be nice if it were protected from having to deal with the internals of tubs • The book’s challenge solution suggests a general approach

  44. I offer an alternative approach which is not necessarily good, but which emphasizes the idea of the level to which the code of method X should have to go • If method X deals with tubs, and is ultimately concerned with whether or not a tub’s location is up or down, then maybe a new method should be added to the Tub class

  45. It could be tentatively named something like, isMyLocationUpOrDown() • Formally, at least, method X could be written without violating the Law of Demeter • Instead of this: • tub.getLocation().isUp() • You could write this: • tub.isMyLocationUpOrDown()

  46. The code for method X is protected from having to know what the interface or internals of the Location class are • It only has to know the interface for the Tub class, which is the class that it works with directly

  47. The book goes on to say that the Law of Demeter is more subtle than this: • “Expressions of the form a.b.c are bad.” • The law is based on an understanding of things like coupling and responsibility, where you are actively trying to avoid foreseeable problems that would result if you didn’t follow the law • As always, there may be cases where you know you are violating the law, but some other consideration causes you to do so

  48. Removing Code Smells • The title of this subsection is unfortunate • To the same degree, the content is also unfortunate • The book mentions that some authors have identified certain signs of poorly written code • They mention one source in particular, “Refactoring: Improving the Design of Existing Code” by Fowler et al. • This contains 22 signs that code can be improved by refactoring

More Related