Structured Design

• Structured Design
– Fundamentals of a Discipline of Computer Program and Systems
Design
• Edward Yourdon / Larry L. Constantine
– Prentice-Hall, 1979
• Purpose
– Make methodical the process of designing software systems
• Mainly business systems
• Approach
– Defines properties of a good procedural design
– Defines a step-by-step method for transforming a data flow graph
into a procedural design
• N.B. calls “procedures” (possibly with associated static data)
“modules”, which differs from Parnas’ use of the term as a grouping
of multiple procedures and related data

04 - Structured Design CSC407 1

Structured Design Significance

• Very popular in business circles.
• Never caught on in academic circles
– Ideas are somewhat half-baked
• “theorems” with silly or no proofs
• Ill-described concepts (no firm definitions)
– At a time when predicate logic to describe programming semantics
was in vogue
• Nonetheless, full of useful concepts.
• Somewhat dated, as it applies best to data-flow oriented
software (not interactive, real-time, or database oriented)
– E.g., read update records from tape, merge them into a master file,
and print a report.

04 - Structured Design CSC407 2

1
 Modules and Connections
• Module
– A lexically contiguous sequence of program statements,
bounded by boundary elements, having an aggregate
identifier.
• i.e., a “function” or “procedure” or ? “method” ?

• Connections

normal
pathological

04 - Structured Design CSC407 3

Limitations on Dealing with Complexity

• Errors: 7 ± 2 rule
– Based on work of psychologist George Miller
• Now questioned in the HCI community
40
35
30
25
Errors

20
15
10
5
0
0 2 4 6 8 10 12
Things to consider at once

04 - Structured Design CSC407 4

2
 Total Errors in a System
• Two opposing forces:
– Intra-module complexity: Complexity within one module
– Inter-module complexity: Complexity of modules interacting
with one another

Total errors is a combination

Errors
Inter-module effect grows as
the number of modules grow

Intra-module effect decrease as

the modules become smaller

# of modules

04 - Structured Design CSC407 5

Overall Cost
• The cost of developing most systems
– ≈ cost of debugging them (+ cost of changing them nowadays)
• These costs are directly related to the overall complexity
– Complexity injects more errors and makes them harder to fix
– Complexity requires more changes and makes them harder to effect.
• Complexity can be decreased by breaking the problem into
smaller pieces
– So long as those pieces are relatively independent of one another
• Eventually, the process of breaking pieces into smaller pieces
creates more complexity than it eliminates.
– 1970s: Happens later than most designers would like to believe.
– 2000s: Happens sooner than most designers would like to believe.

04 - Structured Design CSC407 6

3
 Design Approach
• Therefore, there is some optimal level of sub-division that
minimizes complexity
– Use your judgment
• Once you know the right level, then must choose how to
sub-divide:
– Minimize coupling between modules
• Reduces the complexities of interaction
– Maximize cohesion within modules
• Keeps changes from propagating
– Duals of one another.

04 - Structured Design CSC407 7

Coupling
• Two modules are independent if each can function
completely without the presence of the other.
– They are decoupled or uncoupled.
• Highly coupled modules are joined by many
interconnections/dependencies
• Loosely coupled modules are joined by few
interconnections/dependencies

• Wish to minimize coupling between modules in a system

– Coupling = probability that in coding/modifying/debugging
module A we will have to take into account something about
module B

04 - Structured Design CSC407 8

4
 Influences on Coupling
• Type of connection
– Minimally connected: parameters to a subroutine
– Pathologically connected: non-parameter data references
• Interface complexity
– Number of parameters/returns
– Difficulty of usage. e.g., A = sqrt(x,y,z)
• Information flow
– Data flow
• Passing a of data to be acted upon in a uniform fashion
– Control flow
• Passing of flags that govern how other data is processed
• Binding time
– More static = more complex
• e.g., literal ’80’ versus pervasive constant N_STUDENTS, versus
execution-time parameter.
04 - Structured Design CSC407 9

Common-Environment Coupling
• A module writes into global data
• A different module reads from it (data or, worse, control).

R S

T
U X
V
W

04 - Structured Design CSC407 10

5
 Coupling Example 1

main()

getChar()

• Alternate interfaces:
– void getChar(bool& eof, char& c)
– char getChar(bool& eof);
– char getChar();
• Either way:
– 1 data coupling
– 1 control coupling

04 - Structured Design CSC407 11

Coupling Example 2

main()

getLine() parseCmd() execCmd()

getChar()

04 - Structured Design CSC407 12

6
 Interfaces
• void getChar(char& c, bool& eof);
• void getLine(char* &line, bool& eof);
• void parseCmd(char* line, Command& cmd);
• void execCmd(Command cmd);

• 3 data couplings, 4 command couplings

04 - Structured Design CSC407 13

Example

• Go to your tutorial!
• I will give you a similar question
on the exam.

04 - Structured Design CSC407 14

7
 Cohesion
• While minimizing coupling, we must also
maximize cohesion.
– How well a particular module “holds together”.
– The cement that holds a module together
– Answers the questions:
• Does this make sense as a distinct module?
• Do these things belong together?
– Best cohesion is when the cohesion comes from the
problem space, not the solution space
• Echoed years later in OOA/OOD

04 - Structured Design CSC407 15

Elements of Processing
• A module is composed of processing elements.
– ill-defined
– roughly corresponds to flowchart steps
• Cohesion is a measure of how well the processing
elements hang together as a module

• Cohesion of a module is
– approximately the highest level of cohesion which is applicable to
all elements of processing in the module

04 - Structured Design CSC407 16

8
 Levels of Lack of Cohesion
• Coincidental
– No rhyme or reason for doing 2 things in the same sub-
routine
• void computeAndRead(double x, double& sqrtX, char& c);
• Logical
– Similar class of things
• char input(bool fromFile, bool fromStdin);
• Temporal
– Things that happen one after the other
• void initSimulationAndPrepareFirst()

04 - Structured Design CSC407 17

Levels of Lack of Cohesion (cont’d)

• Procedural
– Operation are together because they are in the same loop or
decision process (but no higher cohesion exists)
– typeDecide(m)
• Decide type of plant being simulated and perform simulation part 1.
• Communicational
– All operations are on the same set of input data, or produce the
same set of output data
• void printReport(data x, data y, data z)
• Sequential
– A sequence of steps that take the output of the previous step and
process it into input for the next step.
• string compile(String program) {
parse, semantic analysis, code generation }

04 - Structured Design CSC407 18

9
 Cohesion (cont’d)
• Functional
– That which is none of the above
• double sqrt(double x);
– Does one and only one conceptual thing.
– Equivalent to Information Hiding

04 - Structured Design CSC407 19

Implementation and Cohesion

• Consider module FG that does two things: F and G
• When doing these things in the same module, chances are
there is some common code than can be shared.
• If F and G have high cohesion, that's ok.
• Otherwise it becomes difficult to work with

F F

G
G

04 - Structured Design CSC407 20

10
 Data Flow Diagrams (DFDs)

employee
skill
records
department
skill
summary
valid employee
Check skill records
Skill Summarize
Validity

bogus skills

04 - Structured Design CSC407 21

Structured Design Methodology

• Transform Analysis
– Restate the problem as a data flow graph
– Identify Afferent and Efferent data elements
• afferent: high-level input data, furthest removed from the physical
input, which are still considered inputs
• efferent: high-level output data, furthest removed from the physical
output, which are still considered outputs
– Factor Afferent, Efferent, and Transform branches

04 - Structured Design CSC407 22

11