Discussion Questions for Riel, Chapters 3-4
God Classes
-
Heuristic 3.2 says that one needs to be suspicious of classes whose names
contain "Driver, Manager, System, or Subsystem". Why is this so? What
problems arise from such nomenculature? And what if one was designing an
operating system/embedded system, and there needs to be a Driver Manager
class? (Dimitar)
-
In heuristic 3.3 Riel says that we should "beware of classes that have many
accessor methods defined in their public interface." However, isn't it
impossible to avoid using accessor methods when a class is as simple as a
point? (Jonathan)
-
Can you give an example of when the God Class problem can't be solved? Also
an example where the real world can't be modeled because of it? (David D)
-
The four schemes for checking if students have the necessary prerequisites
to get into classes Riel lists on pages 34 and 35 all seem to have some
sort of problem: the first two violate Jacobson's heuristics, and the
second two reduce student and class to "mindless clumps of data." Riel
doesn't provide any other solutions, though. What's an object oriented
designer to do when facing such a situation? (John)
-
Rieler says that the number of classes that should be used should be
minimized, and suggests that some classes can be reduced into one. Isn't
this moving towards a god like model? (Elijah)
System Design
-
What is the difference between Heuristic 3.7 and Heuristic 3.8? (Gaurav)
-
How is the solution to adapting a legacy system described on pages 40-41
not a violation of the earlier heuristics 2.6 and 2.10?
-
C++ can be compared to OO programming as essentially functioning under the
category of a god class (one piece does everything). Is there a language
that follows as an ineficient model for OO's "prolifertion of classes"
category? (Eric)
-
When talking about proliferation of classes, Riel describes an ATM with a
needless "network" class, but then later a "network wrapper" class is
created to separate the ATM objects from an outside data source. Is this a
good design technique, or is it part of the proliferation problem?
(Lorelei)
-
Rieler describes agent classes as often being used under the 'guise of
reuse', but that using agents is seen as bad seems more like a failing of
OOP programming. Wouldn't it be ideal to be able to seperate action from
data, IF we could keep track of the relationships between the two? (Elijah)
-
Rieler talks about minimizing classes and then gives an example of a class
that has 30 methods, only some of which are used. He says this is ok. But
shouldn't we try to minimize the number and complexity of our methods as
well? (Elijah)
-
How do you decide what is a useless class? (Monica)
Agents
-
What is the role of an agent class if Heuristic 3.10 indicates that many of
them are irrelevant and should be removed? What are the qualities that
make an agent essential to a design vs. those that should be removed?
(Mike C)
Uses Relationship
-
In heuristic 4.5 Riel argues that a "narrow and deep containment hierarchy"
is better than a "broad and shallow containment hierarchy." Wouldn't
following this heuristic lead to the "proliferation of classes" problem?
(Jonathan)
-
What are the disadvantages of violating Heuristic 4.6?
-
I do not understand the concept of Contained Classes expressed in Heuristic
4.13: "A class must know what it contains, but it should not know who
contains it." What does this Heuristic really try to convey? (Dimitar)
-
Rieler says a class should know what it contains but not what contains
it. can this seperation be made by using an agent class? Would Rieler
aprove of this? (Elijah)
-
Could you talk more about Classed-based, Containment, Object-based and Uses
relationship together? And please clarify the difference among these
relationships. Consistent examples will be appreciated (Sheng).
-
While it is likely true that narrow and deep containment hierarchies are
better than wide and shallow, I don't find Riel's argument on pages 62 and
63 convincing. While it would be troublesome to remember costs for
individual food objects, we could easily store the costs in the objects
themselves, and then calculating the total would simply amount to adding up
the return values of all the individual cost() operations. Only if
individual food items have different behavior would it make sense to me to
break down the meal class. Not really a question, as much as an
observation, I guess. (John)
-
Riel mentions on pg. 70 that "there are three and not two, relationships
invovling uses and containment." How do weak containment and strong
containment affect a model's implementation? Is it possible to have
something that isn't really strong or weak? (Mike C)
-
Why do we have to distinguish attribute class and contained class? How do
we deal with the class that need to have both properties (not necessary a
god class)? (Chayapan)
General Issues
-
Can you relate the Heuritics of Chapter 3 and 4 to various UML diagrams?
(David)
-
What are the pitfalls to binding particular objects and classes to one
domain? (Eric)
-
What's the significant difference between Action-Oriented and
Object-Oriented (Sheng).
-
When do you use containment, when do you use inheritance? (Monica)
-
What is the difference between "object-based versus class-based"? (Chayapan)
This document is available on the World Wide Web as
http://www.cs.grinnell.edu/~walker/courses/223.sp05/readings/riel-3-4.shtml
