CSC/ECE 517 Fall 2009/wiki3 18 301

Topic: Stable Dependencies Principle

Introduction

Package Design is a challenging area of Software Engineering. An inefficiently designed system can be a root cause of many problems, thus, it is vital to have a solid understanding of various design approaches. This article discusses Stable Dependencies Principle, one of a few key package design principles that software engineers should be aware of.

Definition of Instability

First, let's discuss why certain packages are considered stable, and others are not. The instability of any package can be calculated using the following formula.

I (instability) = Ce / ( Ca + Ce )

In this formula:

Ce – the number of other modules that our package requires in order to operate correctly. In other words, this is an indicator of package independence.

Ca – the number of packages that can be viewed as “dependents“, i.e., “depend upon classes within the package”. This is an indicator of package responsibilities.

Therefore, in order for a package to be completely independent and also stable, it should not be dependent upon other packages.

For example, let’s look at the following component diagram:

The Wheel class in this example is shown as not dependent on any other module.

Using the above formula:

I(W) = 0 / ( 3 + 0 ) = 0

I(W) has the instability of zero. In other words, even though the Wheel class is “responsible” for three classes (Ca=3), it is not dependent upon any of them (Ce = 0), thus resulting in a stable package.

Respectively, instable or irresponsible packages would have the instability of 1, as shown below:

In this case, the Car class does not have any dependent classes (Ca = 0), yet itself depends upon three classes – Wheel, Engine, and Suspension (Ce = 3).

Next, let’s calculate the instability:

I(C) = 3 / ( 0 + 3 ) = 3 / 3 = 1

The instability of 1 indicates a high level of volatility of the Car module.

Stable Dependencies Principle

The Stable Dependencies Principle (SDP) states that “a package should always depend upon packages that are more stable than it is.” Ideally, the direction of dependencies should go from high volatile to more stable packages.

So how do we know whether a certain component design follows the SDP?

The following is a more complex diagram outlining several Car components:

The instability of each of the modules would be calculated as follows:

I(C) = 3 / (0 + 3) = 1 (the least stable)
I(W) = 2 / (1 + 2) = 0.66
I(E) = 2 / (1 + 2) = 0.66
I(Su) = 1 / (1 + 1) = 0.5
I(T) = 0 / (1 + 0) = 0 (the most stable)
I(R) = 0 / (1 + 0) = 0 
I(Cyl) = 0 / (1 + 0) = 0
I(O) = 0 / (1 + 0) = 0
I(St) = 0 / (1 + 0) = 0

After replacing each class with its corresponding instability value, the diagram would look as follows:

The lower the value, the more stable the package is considered to be. In our case, all dependencies progress in the direction of stability, thus, according to the SDP, our design is proven to effective.

Another important consideration when it comes to stability is that more stable packages should not be changed often. On the other hand, a highly volatile class can be changed easily, without affecting other packages in the system.

Links

[1] The Tipping Point: Stability and Instability in OO Design, http://www.ddj.com/architect/184415285

[2] On Package Design, https://prof.hti.bfh.ch/fileadmin/home/due1/uml_dp/udp_package_design_200906.pdf