131

on structured programming

"In the light of all these difficul-

ties the simple approach may be the
best."
W.F.C. Purser (1976)

Software for local multi-mini proc-

essor systems.
In: 'Multiprocessor Systems', Info-
tech State of the Art Report, pg. 408.
132

Contents Chapter Three

1. Introduction pg. 133

1. The origination of structured programming 133

2. The software crisis 135
3. Structured programming and problem solving 136

2. Survey of Structured Programming Techniques 139

1. Restrictions 139
2. Abstraction 141
3. Stepwise refinement 142
4. Notation 143

• Data representation 143

• Flowcharts 145
• Documentation 147
• Programming Languages 148

3. Discussion: Criticisms on Structured Programming 150

1. Pragmatics 150
2. Complexity 150
3. Rigidity 151
4. Methodology 155

4. Conclusion 157 ,

5. References Chapter Three 158

Appendix B: Graph Generation and Complexity Measures 161

References Appendix A 167
 133

CHAPTER THREE
ON STRUCTURED PROGRAMMING

3.1. INTRODUCTION

One conclusion we can draw from the discussion in chapter two is that correct-
ness proving is by no means a trivial task, neither for sequential programs nor
for concurrent programs. The available analysis tools are not powerful enough
to allow for the verification of just any program. By necessity one must re-
strict oneself to simple and well-structured programs.
But, one may ask, what is a well-structured program?
Fortunately, there is a wealth of literature on this subject. The larger part
of this literature is primarily concerned with sequential programs, but still
we should take careful notice of these theories if we want to be able to apply
the underlying principles to the multiprocessing cases.

In this chapter we will study the techniques which have becOme known as 'struc-
ured programming' techniques. Before doing so we will dicuss the origination
of the theory and in that context we take a brief look at the so-called 'soft-
ware crisis' which stimulated this origination in many ways. We will also con-
sider how the structured programming methodology is related to more general
problem solving techniques. We will then survey the principal techniques from
structured programming, while concentrating ourselves on these four major as-
pects: restrictions, abstraction, stepwise refinement, and clear notation.
We conclude with a discussion of some of the major criticisms that have been
raised against the techniques.
It is not our intention to give, within the scope of a single chapter, a de-
tailed overview of all techniques that have been brought in connection with
structured programming theory. Such overviews can be found elsewhere (see, for
instance, Infotech '78). Our present goal is to describe and motivate the main
principles, in preparation of the studies in part 2 and 3 of this thesis.

3.1.1. The origination of structured programming

The term 'structured programming' was coined by Dijkstra '69b. Though Dijkstra
is not the sole originator of the fundamental ideas behind structured program-
ming, he certainly was the first who succeeded in formulating his ideas on the
subject in a way which was understood and which initiated a wide discussion.
Without exaggerating, it can be said that Dijkstra created a new paradigm in
the computer sciences (Kuhn '62).
In retrospect, the occurrence of a 'paradigm shift' at that moment (i.e. the
late sixties) is easily understood. One was in the middle of the 'software
crisis'. Computers had become faster and bigger than anyone could have fore-
seen, and programmers were confronted with unmanageably large and complex soft-
ware. The construction of a reliable and understandable program of some size
had become an art, rather than a teachable skill. It was noted that the com-
plexity of a program grew almost exponentially with its size.

The negative effect of the use of certain language constructs on the under-
standability of a computer program had been debated on a small scale for some
134

years. Some initial thoughts on the subject can be found already in Dijkstra
'61 (especially pg. 4).
In Dijkstra '68a a reference is made to remarks of Zemanek in 1959.
Knuth '74 (pg. 264 and 265) refers to discussions with Schorre in 1963, to
an article of Naur in '63 and to papers of Landin and Naur presented at an
ACM Conference in 1966.
The discussion was however only really started after the publication of Dijk-
stra's provoking letter 'GO TO Considered Harmful'. The opening sentences of
that letter clearly illustrate its purport and style:
"For a number of years I have been familiar with the observation that
the quality of programmers is a decreasing function of the density
of GO TO statements in the programs they produce. More recently I
discovered why the use of the GO TO statement has such disastrous ef-
fects, and I became convinced that the GO TO statement should be
abolished from all 'higher level' programming languages (i.e. every-
thing except, perhaps, plain machine code)."
(Dijkstra '68a, pg. 147).

Phis letter unleashed a vast amount of articles, conference papers, and emotion-
al discussions. Another influential article of Dijkstra's was published a few
months later (Dijkstra '68b). In that paper Dijkstra illustrated how a large
complex software system could be designed and programmed in 'layers', by sys-
tematic abstraction. One year later, in 1969, the 'Notes on Structured Program-
ming' appeared as a report of the University of Technology in Eindhoven
(Dijkstra '69b). Despite the fact that these 'notes' were not distributed on a
wide scale, they had a large influence. In 1971 Wirth elucidated one of the
design techniques, 'stepwise refinement', to a wider audience (Wirth '71).
One year later, in 1972, Dijkstra's notes were re-published in book-form, sup-
plemented with papers by Hoare and Dahl (Dahl, Dijkstra & Hoare '72).
Two other important articles must be mentioned here. Both appeared in 1973.
The first one is the article by Wulf and Shaw titled 'Global Variables Consider-
ed Harmful' (Wulf & Shaw '73). (The titles of Dijkstra's articles have been
paraphrazed in many ways.) Wulf and Shaw argued that the use of variables with
a large scope can frustrate the analysis of a program. These ideas were later
elaborated by Dijkstra (Dijkstra '76, pg. 79-93).
The second article is the one from Nassi and Shneiderman who introduced a new
flow-charting technique for structured programs (Nassi & Shneiderman '73).
This technique, which was aptly named 'Structogramme' by Schnupp '74, derives
its value from the fact that it forces a programmer to think in terms of only
a few basic control-flow structures and prevents the (undisciplined) use of
GO TO statements.

We can summarize the principal ideas from the papers mentioned up to now in
these four concepts: restrictions, abstraction, stepwise refinement, and clear
notation. Each of these items was heavily debated. The ideas were extended,
modified and reduced, they were defended, ignored and attacked in large waves
of articles.

Of considerable influence upon the eventual acceptance of the ideas were the
many publications on the successful application of the techniques in large in-
dustrial programming projects. Already in 1973 there appeared an article en-
titled 'Revolution in P rogramming', from which we quote:
"Then came the IBM work for the New York Times, with reports of great-
ly increased programmer productivity and very greatly reduced coding
error rates (one detected error per 10.000 lines of coding, or one
per man year)! Absolutely incredible, but these were the facts."
(McCracken '73, pg. 51).
 135

;idently, the reportings were often called in question. For instance, one
'ted that successes are published sooner than failures. Recently, a large soft-
.are institute wrote skeptically in a customer letter:
"- Many of the early successful projects were run by experienced or-
ganizations using top-level staff - how do the new techniques work
in practice in everyday installations?
- The very fact that an organization attempts to introduce new tech-
niques suggests a commitment to improving software production -
how much of the reported improvement is due to that commitment,
and how much to the new techniques?"
(Infotech, letter R377, Nov. 1978).

A more practical problem was that in many cases the programmers involved in the
industrial projects simply refused to alter their style of programming (see
Canning '74, pg. 3; Gries '74). The structured programming principles were
criticized both on practical grounds and for more philosophical reasons.
We take a closer look into the concrete criticisms in a later section (see sec-
tion 3.4). First we take a closer look into the nature of the 'software crisis'.
and on general problem solving techniques.

3.1.2. The software crisis

In1966 Naur wrote:

"The stress has been on always larger, and, allegedly, more powerful
systems, in spite of the fact that the available programmer compe-
tence often is unable to cope with their complexities.
However, a reaction is bound to come. We cannot indefinitely continue
to build on sand. When this is realized there will be an increased
interest in the less glamorous, but more solid basic principles."
(Naur '66, pg. 310).

Within a few years Naur's prediction came true. In 1973 Donaldson already de-
scribed 'a major shift of emphasis' in software engineering.
"The primary requirement to be met in software engineering has always
been to perform the function specified for the software. But, where
at one time secondary emphasis was placed only on software efficien-
cy, that is in core and time required, today three other factors are
recognized as requiring special emphasis. These factors are reliabili-
ty, maintainability, and extensibility."
(Donaldson '73, pg. 52).

We can point out four major reasons for the occurrence of the change described
by Donaldson:
First of all, the machine is today no longer the most expensive 'link' in the
programming chain. Computer time has become relatively cheap. (See chapter 1,
section 1.1.)
Secondly, the considerable growth of computers and programs, both in size and
complexity, has made their reliability more and more important. Software sim-
plicity has become more important than hardware efficiency.
Thirdly, the average life time'of software has increased. Existing programs
must be adaptable to changing requirements, and not only by the people who ac-
tually wrote the programs in question. There are many ways to write a program
which no one understands (Kernighan & Plauger '74), butto write programs which
are transparent to any programmer is much more difficult. It requires another
attitude from programmers and a style of programming which Weinberg once de-
scribed as 'ego-less programming' (Weinberg '71).
The fourth reason concerns the multiprocessing systems. The advantage of a mul-
136

tiprocessing_system is no longer the fact that it is more 'economical' to share

hardware equipment among many users, but it is the possibility of sharing in-
formation. To guarantee the integrity of this shared information has become a
major problem (Update, Vol. 2, No. 6, June 1977).
These observations lead unavoidably to the conclusion that concerns for machine
efficiency should no longer be a guiding principle in software design. Garren
formulated it even more sharply, when he wrote:
"As a general rule local efficiency, i.e. bit twiddling, is in con-
flict with the need for understandability (-)."
(Garren '73, pg. 37).

This does however not imply that the concerns for efficiency are to be entire-
ly neglected. Garren continued:
"By emphasizing ease of modification in system design and implemen-
tation it is possible to quickly build and debug the system and then
measure it to discover where optimization is actually needed."
"Local optimizations can contribute to a relatively small fractional
increase in program efficiency, whereas changes in global algorithm
can result in savings of orders of magnitude."

Concerns for program efficiency are thus merely moved to the final stages in
program design (see especially Knuth '74).

3.1.3. Structured programming and problem solving

After the software crisis it became clear that the real bottle-neck in software-
development was not 'efficient coding' but structuredness, or 'efficient prob-
lem solving'. However, as Gries aptly remarked:
.. almost none of the elementary programming books say anything
about problem solving."
(Gries '74).

It is one of the merits of structured programming that it filled this gap by

constructing an analogon of known problem solving techniques in program design
theory. Dijkstra explicitly described the task of writing a correct computer
program as an intellectual problem:
"The art of programming is the art of organizing complexity, of
mastering multitude and avoiding its bastard chaos as effectively as
possible."
(Dijkstra, in: Dahl, Dijkstra & Hoare '72, pg. 6).

To design a program is to determine its structure. But, not just any structure
will make a program understandable. Especially, the . GO TO statement allows a
programmer to make highly complicated control-flow patterns.
Dijkstra suggested therefore that the programmer restrict himself mainly to
program structures which correspond closely to standard reasoning patterns.
When one deletes the GO TO statement, the programmer has only the following
'tools' for the construction of such patterns: concatenation, conditional se-
lection, iteration, procedure-calls, and recursion. Procedure-calls can (con-
ceptually) be replaced by the code they represent and they are therefore not a
fundamentally different type of tool. A recursion can be represented or under-
stood as an iteration, and it is therefore usually not considered separately
either. That leaves us with: concatenation, selection, and iteration. Liskov
writes:
"... each permissible control structure is associated with a well-
 137

known rule of inference: concatenation with linear reasoning, itera-

tion with induction, and conditional selection with case analysis."
(Liskov '72, pg. 194).

A well-structured program then should be decomposable into these three basic

structures. Ideally this decomposition should be equivalent to a decomposition
in 'logical units' of the problem being solved by the program.
In this manner even large programs should be understandable with an effort
which is at most directly proportional to their sizes, instead of exponentially
proportional. The major problem in programming is, however, not to write a pro-
gram with only the three basic control-flow structures mentioned, but the prob-
lem is to make the structure of the program 'match' the structure of the prob-
lem being solved. To solve that problem the programmer is advised to use ele-
mentary problem solving techniques like abstraction and stepwise refinement.
Knuth wrote:
"We understand complex things by systematically breaking them into
successively simpler parts and understanding how these parts fit
together locally."
(Knuth '74, pg. 291).

By systematic abstraction one can attempt to distinguish minor aspects (imple-

mentation details) from major lines, and concentrate on the latter. By system-
atically splitting up major problems into minor ones, one can derive a solution
in a stepwise manner.

The ideas are not new (Knuth '74, pg. 291/292; Lecarme '74). It has even been
reported that some of the techniques had been in use for some time in 'arti-
ficial intelligence' projects (VandenBrug '74).

Top down or bottom up.

- -

One of the more heavily debated questions with respect to the structured pro-
gramming methodology is whether programs should be designed 'top-down' or 'bot-
tom-up' (e.g. McClure '75). One of the early writers on problem solving, Polya,
says:
"We have before us an unsolved problem, an open question. We have to
find a connection between the data and the unknown.
We may represent our unsolved problem as an open space between the
data and the unknown, as a gap across which we have to construct a
bridge. We can start from either side, from the unknown or from the
data ..."
(Polya '48, pg. 73).

Polya did not express any principal preferences for either of these approaches,
and indeed it would be unwise to do so. In problem solving and in program de-
sign both analysis and synthesis play an important role. It has often be noted
that problem solving is essentially an iterative process. (See, for instance,
Davies '71, '76). It consists of repeated attempts to match the results of
analysis and synthesis. Similarly, the authors of the book 'Structured Program-
ming' (Dahl, Dijkstra & Hoare '72) explicitly stated in their preface that the:
.. structured programming principles can be equally applied in
'bottom-up' as in 'top-down' design."

Still, structured programming is often identified with purely 'top-down' design.

The confusion is perhaps caused by the fact that one of the requirements of
structured programming is 'top-down documentation'. Documentation, however,
concerns completed programs and not primarily programs which are still being
designed. For a further discussion of these points we refer to the section on
138

'criticisms' (section 3.4). We will now give a survey of the principal struc-
tured programming techniques. We will do this under the headings: Restrictions,
Abstraction, Stepwise refinement, and Notation.
 139

3.2. SURVEY OF STRUCTURED PROGRAMMING TECHNIQUES

3.2.1. Restrictions

The most important restriction is that the programmer is asked to use only
three basic control-flow structures: concatenation, conditional selection and
iteration. These three basic structures are to be used as the 'building blocks'
of a program. They all have one entry-point and one exit-point, as illustrated
in figure 3.1 1 .

(a) (b) (c)

Concatenation Conditional Selection, 'Iteration
Figure 3.1.
To form larger structures, the three basic blocks may be nested in any order,
as illustrated in figure 3.2.
The program which is designed in accordance
with these restrictions consists of a hier-
archical nesting of control-flow structures,
and has precisely one entry-point and one
exit-point on each level. It can then always
be established, with little effort, when a
certain program part is executed and why.

It is of course not true that every goto-less

program is by necessity transparent, nor that
every program with goto-statements is unclear.
Wulf argued:
.. it is certainly possible to find
programs which use the goto and are
not difficult to understand - partic-
ularly if the programs are small, the
targets of the goto's are close to
the jump points, and the control paths
established by the goto's use are
standard ones."
(Wulf '73, pg. 28).
Nestings
Two 'standard cases' in which the use of a
Figure 3.2. goto statement may be hard to avoid, are the
'search-loop' and the 'error-exit' (see also
section 3.4).
(1) We will derive a complexity measure for control-flow structures in
appendix B.
190

Wirth, however, commented-:

.. often the need for an exit in the middle construct is based on
a preconceived notion rather than a real necessity, (-) sometimes
an even better solution is found when sticking to the fundamental
constructs."
(Wirth '74, pg. 255).

There have been many proposals for the inclusion of a special construct in the
'set of basic control-flow constructs' which would cover the error-exit and/or
the search-loop (e.g. Knuth & Floyd '71; Friedman '74; Evans '74; Goodenough
'75). There is, however, no consensus on this point. More liberal authors have
suggested not eliminating the goto statements, but restricting their usage.
Three alternatives are then:

(1) to restrict to only forward goto-jumps;

(2) to restrict to goto's used for error-exits only;
(3) to restrict to goto-jumps within the innermost
hierarchical level.
Each variant (and combination) has found its advocates (see, for instance,
Leavenworth '72; Canning '74; Newton '75).
While some authors have suggested relaxing the restrictions of structured pro-
gramming, others have proposed still more stringent rules.
The most important of these are Wulf and Shaw (Wulf & Shaw '73), and Hoare (in:
Dahl, Dijkstra & Hoare '72).
Hoare stressed that not only the control-flow of a program should be well-
structured but also its data. In the process of stepwise refinement the program-
mer should choose a plausible data representation and formalize the necessary
data manipulation techniques. For instance, the programmer can choose to repre-
sent his data in a structured record, a queue or a list, and define the ap-
propriate amending, rearranging, and deleting operations for it. As special
data-types Hoare discussed the 'cartesian product', the 'discriminate set
union', the 'power set', and the 'sequence'. We consider some other ideas on
data structuring more closely in the section on Notation (3.5.4).

Wulf and Shaw suggested restricting the scope of variables, and requiring that
this scope is always made explicit in a program text via declarations. The rea-
sons for these further restrictions are:
.. the non-local variable is a major contributing factor in pro-
grams which are difficult to understand."
.. the complexity of [the task to prove that variables are always
accessed in the intended way] increases exponentially with the
distance of the last previous use of each variable along each con-
trol path."
.. the complexity of the abstraction is directly related to the
number of non-local variables used."
(Wulf & Shaw '73, pg. 28/29).

Wulf and Shaw further argued that an access right to a variable which has been
declared on a high level in a nesting hierarchy should not be 'inherited'
(exported) automatically by (to) all lower levels . Upon entry to a new level
(a 'block') all variable declarations that are to be 'inherited' (imported)
must then declared be explicitly. Dijkstra elaborated similar ideas in
Dijkstra '76.
An inherited (imported) variable is called 'global' by Dijkstra, a new vari-
able is called 'local'. The scope of a local variable extends from the begin-
 141

ning to thp end of the block in which it is declared, with the exception of
those inner blocks that do not explicitly inherit it via a 'global' declaration
(Dijkstra '76, pg. 85). All variables must be initialized before their first
use. A non-initialized variable is called a 'virgin' variable. Via the declara-
tions one can declare whether the variables will be used as references (con-
stants) or will be changed in the corresponding block. Dijkstra then defines 6
different types of variables which can occur in a block (local, global or vir-
gin variables, used as constants or as 'true' variables) with special 'rules of
inheritance'. (Dijkstra '76, pg. 91).
As a possible extension of these ideas both Dijkstra and Wulf & Shaw have men-
tioned the possibility of making explicit and restricting also the ways in
which variables can be modified.

Another type of additional restriction was suggested by Bloom '75 and Weinberg
at al. '75. They suggested eliminating the 'conditional selection' construct,
and replacing it by a 'conditional execution (i.e. replacing the 'if-then-else-
end' by 'if-then-end'). This suggestion has not found much support, and indeed
it seems that it would be too restrictive. It would make program structures
more simple, but it can be questioned whether it would make them more under-
standable.
An article which is often quoted in discussions on structured programming tech-
niques is Bohm & Jacopini '66.
Bohm & Jacopini '66 showed that every flow diagram with goto-statements can be
transformed in a goto-less one. (See also Mills '72 who presented a more rigor-
ous proof of their statements. See also Ashcroft & Manna '71.) The programs are
then equipped with some additional state variables' which formalize part of the
flow-of-control. It will be obvious, however, that such transformations have
no relevance to structured programming whatsoever. Dijkstra stated:
"The exercise to translate an arbitrary flow diagram more or less
mechanically into a jumpless one, however, is not to be recommend-
ed. Then the resulting flow diagram cannot be expected to be more
transparent than the original."
(Dijkstra '68a, pg. 148).

The same point was also clearly illustrated by Cooper who demonstrated, in a
reaction to Bohm & Jacopini's paper, that any flow diagram can trivially be
transformed into a single iteration loop containing one large case-selection
structure. The resulting program is 'simple' indeed, but certainly not under-
standable (Cooper '66).

3.2.2. Abstraction

In the initial stages of program designing the main problem is to find the ma-
jor lines of the solution. Abstraction (or 'generalization' as Polya called it)
is a powerful mental aid to discover what is detail and what is not.
"Executional abstraction is so basic to the whole notion of 'an al-
gorithm' that it is usually taken for granted and left unmentioned.
(-) ... it refers to the way in which we can get a specific compu-
tation within our intellectual grip by considering it as a member
of a large class of different computations ..."
(Dijkstra '76, pg. 1).

The abstraction levels can be expressed in a program structure in several ways.

First of all, one can express it in the orde ring of the control-flow struc-
tures in a nesting hierarchy. The flow-structures on the highest (outer) levels
in the nesting hierarchy should then express the major lines in the argument.
Secondly, one can express it in the subdividing of the basic computational
 142

functions used in the program, and in their relationships.

The technique spf dividing the logic required of a program into pieces (modules)
which are small, simple and independent is often called 'modular programming'
(Harding '72). One can then construct a 'module hierarchy' in the structural
relationships of these modules. Major lines are treated in 'high level' modules,
details are treated in the lower level ones. It is however not intended that
the entire program should be constructed as an arbitrary collection of modules
and module-interfaces, at least not in structured programming. In Dijkstra's
view a large program should be constructed in 'layers' (Dijkstra '68b, '69a,
'72). Each layer represents an abstraction level. On a high level of abstrac-
tion one works with abstract operations and functions which are elaborated in
the lower levels (see chapter 1).
The concepts of a 'module' and a 'layer' are closely related though. Dijkstra
explained the differences as follows:
"If we succeed in building up our program along the lines just given,
we have arranged our program in layers. Each program is to be under-
stood by itself, under the assumption of a suitable machine to exe-
cute it, while the function of each layer is to simulate the machine
that is assumed to be available on the level immediately above it."
"The picture of a layered hierarchy of machines provides a counter
poison to one of the dangers evoked by ruthless application of the
principle 'Divide and Rule', viz. that the different components
are programmed so independently of each other that duplication of
work (or worse) takes place.
The fact that a layer contains 'a bunch of programs' to be executed
by some conceptual machine stresses the fact that the programs of
this bunch are invited to share the same primitives. Separation of
tasks is a good thing, on the other hand we have to tie the loose
ends together again."
(Dijkstra, in: Dahl, Dijkstra & Hoare '72, pg. 49/50).

Another concept which is related to the principles of 'abstraction' and 'layer-

ing' is 'information hiding'. Here one makes use of the hierarchical structure
of a well designed program. The basic idea is then to hide information about
the implementation details of the lower levels from the higher levels. This in-
formation hiding serves two purposes: integrity and adaptability. By hiding
information about the way in which a memory-element is listed in an abstract
queue, for instance, one can prevent the information contained in that queue
from being mutilated by an erroneous usage. Similarly, if one hides information
on the way in which specific design decisions are locally solved to the larger
part of a program, it will be a trivial task to adapt that program to alter-
' native design decisions (Parnas '72; Dennis '75; Ross at al. '75).

3.2.3. Stepwise refinement

The process of stepwise refinement can be described in general terms as follows:

A programming problem is first analyzed thoroughly to get an overview of all
the implications of the problem, without however going into specifics.
An attempted solution is then notated in informal code, at a high level of ab-
straction, as if there were a machine that would be capable of interpreting
that code and executing it. Then the refinement process begins. Step by step
the solution is made more specific. At each new level of abstraction one takes
new design decisions, but not more than strictly necessary. The design deci-
sions are taken at such a level that they convey precisely the substructure to
which they apply, not more and not less. The refinement of the program can go
hand in hand with the closer inspection of the problem being solved:
"As the problem analysis proceeds, so does the further refinement of
 143

my program."
(Dijkstra, in: Dahl, Dijkstra & Hoare '72, pg. 27).

Clearly, this is an idealized description of a design process. The programmer

is bound to make mistakes. He will discover in refinement-step x that he made
a mistake in refinement step x 1, or earlier. One then returns to the point
- -

where the mistake was made and continues with a different approach. The design
process is thus essentially an iterative process, though in its idealized form
one is tempted to define it as a one-pass procedure.

Each completed level of abstraction can represent an entire class of problems.

In the process of refinement the class of problems is, however, narrowed down
to precisely the problem that is to be solved by the program desired. If one
day the requirements are reformulated, one merely returns to the appropriate
level of abstraction and refines the solution anew, to satisfy the new require-
ments. Dijkstra writes:
"... when a program has been built up to an intermediate stage of
refinement, what has then been written down is in fact a suitable
'common ancestor' for all possible programs produced by further re-
finements."
(Dijkstra, in: Dahl, Dijkstra & Hoare '72, pg. 40).

Together with the refinement of the program one refines the structure of the
data. Decisions about the implementation of high level data structures are
postponed for as long as can be.
The high level, or pseudo-code (the latter term is from Melekian '76) can be
regarded as a form of 'structured comment' to the code of the leVels immediate-
ly below it in the design hierarchy. The design process naturally stops when
then lowest level has been reached, that is a description in the code of the
programming language used. Gries, therefore, aptly remarked that with struc-
tured programming:
"... one doesn't program in a programming language but into it."
(Gries '74, pg. 656).

3.2.4. Notation

We discuss four items which are related to a clear notation of problems and
solutions: data representations, flow-charts, documentation, and programming
languages.

Data representations
Many articles have been published on the question of what type of data repre-
sentation would lend itself best to the structured programming techniques (for
instance, Hoare, in; Dahl, Dijkstra & Hoare '72; Shneiderman & Scheuerman '74;
see also Jackson '75).
Two important articles have not gained the attention they deserve. One of them
was published just before the wide discussions on programming style began
(1967), the other a few years later (1971). We consider these two articles
briefly below.

The first article is Balzer '67. Balzer proposed to keep the data manipulations
in the design phase independent of specific data representations. The program-
mer then uses informal operations in his program text, like 'delete record', or
'replace', leaving the refinements to a later stage in program design. The meth-
od fits well in the stepwise refinement procedures outlined earlier. We quote:
... programming should be simplified because it can be constructed
top-down in terms of the logical processing required. The problem
144

of data representation can be left until this programming has been

completed; thus, a more rational decision can be made concerning an
optimal representation. Because of this separation, the programmer
should be able to think through his problem better."
(Balzer '67, pg. 542).

Balzer suggested using an informal hierarchical data reference method, in which

data can be accessed at the desired level of abstraction by statements of the
type:
name (index) of namelevel n-1 (index) of ...
level n
The level of formal parameters need not match the level of the corresponding
actual parameters (see also Ross at al. '75.)

The second article is by Eaily '71. Early described the 'V-graph model' for
data representations. These V-graphs represent the set of data on an arbitrary
level of abstraction, and not (necessarily) the way in which the data is im-
plemented (Early '71, pg. 618).
An example of a V-graph is given in figure 3.3. Early explained the meaning as
follows:
"Each node of the graph represents a part of the data structure.
The arrows between nodes represent access paths in the structure,
and arrows from nodes to 'atoms' represent the fact that the atom
is stored at the node."
"NIL is a reserved object which can be used for such things as the
end of a list."

10

Figure 3.3.
EarZy's V-graph model

The annotations on the edges of the graph indicate subscript variables (or
'selectors' as Early calls them). In the example of figure 3.3 the elements
10, 20 and 30 can be accessed by subscripts as well as via the standard opera-
tion NEXT. The names NEXT and CONT are not reserved words though; one can use
any other suitable notation (PRIOR, FOLLOW, ADJACENT, IMAGPART, REALPART, AGE,
YEAR, etc.).

The importance of a clear notation is easily underestimated. An obtuse notation

 145

however can make a problem practically unsolvable. An adequate data represen-

tation can facilitate the solution of the related problem in many ways. These
ideas have been stressed by many, among others by: Polya '48, and later by
Hoare (in: Dahl, Dijkstra & Hoare '72), and Baker '75.
The latter especially stressed the importance of 'intelligent data names'.

Flow charts
-

Judged by the standards of structured programming, the conventional flow-chart-

ing techniques fall short on three major points:
(1) There is no clear representation of the three basic control-flow struc-
tures; the representation of iteration structures is inadequate.
(2) The programmer is in no way restricted in his use of arbitrary GO TO
jumps, on the contrary.
(3) There is no clear representation of hierarchical levels in a program.
For these reasons one has found that the use of conventional flow-charts, es-
pecially in the design stages, conflicts with the aims of structured program-
ming. Several. attempts have been made to improve the flow-charting technique.
One of the additional rules that have been suggested is:
• all jumps (including the iteration jumps) must be drawn on one side
of the flow-chart, while the
• crossing of lines in the flow-chart is forbidden.
This rule would compensate for the second shortcoming mentioned above; but
leaves the first and the third.
Two of the proposals for alternative representations of iteration constructs in
flow-charts are given in figure 3.4. Neither has gained much support. Observe
that the third objection against the traditional flow-charts is still valid.

(a) (b) (c)

Conventional Flowchart- Alternative Suggestion Alternative Suggestion
representation of Iter- of Schmitz '75. of Weiderman et al. '75.
ation.

In 1973 Nassi & Shneiderman suggested quite another approach to the problem,
and their attempt was more successful. Nassi & Shneiderman introduced a new
type of flow-chart which removes all objections mentioned, to a large extent.
In addition, they noted that:
146

.. the scope of local and global variables is immediately obvious."

.. complete thought structures can and should fit on no more than
one page (i.e. no off-page connectors)."
.. recursion has a trivial representation."
(Nassi & Shneiderman '73, pg. 15).

The three basic control-flow-structures are represented in the new model as in-
dicated in figure 3.5. The representation of a block or hierarchical level is
indicated in figure 3.6. The representation of a recursive subroutine is illus-
trated in figure 3.7.

A a

(a) (b)
Concatenation Selection

DOWHILE

Figure 3.5.
Structogramme
Notations
(c) (d)
Case Selection Iteration
A:

a
A

Figure 3.6. Figure 3.7,

Block or Level Recursive Subroutine

Loop-exit or error-exit jumps are more difficult to represent, as could be ex-

pected. Nassi & Shneiderman suggested using the key-word BREAK in such cases.
The type of flow-charting suggested by Nassi & Shneiderman is called 'struc-
tured flow-charting' or the 'structogramme notation'. The term 'structogramme'
stems from Schnupp '75 (see further Moos & Steinbuch '76).

Still, another approach to the problem of representing a program structure was

suggested by Pazel '75. Pazel suggested a Boolean-like notation for the three
basic structures.
The concatenation of three statements a, b, and c is then indicated as a 'dot-
product': a.b.c .
A selection between statements a and b is indicated as: (a + b), where IS is
the 'selection criterion'.
An iteration on statement a with iteration predicate a is indicated with so-
 147

called 'jump-indicators' (J) and 'receptor-indicators' (R): R.a (1 aJ).

The latter notation is rather inelegant, and could perhaps be replaced by a
notation of the type (aa), while avoiding the J- and R-indicators.

Unfortunately, the resemblence with a Boolean algebra is not valid for the ma-
nipulations of the 'formula's'. For instance, it can be shown that the struc-
ture which corresponds to the formula: (a+ fib) .c is equivalent to the struc-
ture corresponding to the formula: (a.c+p.c) . In other words: it can be shown
that the dot-operation is 'right-distributive' over the +. To show that the
structure given in figure 3.8.a is equivalent to that of figure 3.8.b, or that
the relation:
R1 (R2.a+J1).b.(1+ 2) R2.111 (Si+ .b. (a+ „J2)
Pj a
holds, is however much more complicated.

Figure 3.8.(a) Figure 3.8.(b)

Pazel Notation: Pazel Notation:
R1 .(R2 .a +a J1).b.(1 J2 ) ,.1:(1.(5.1 +a J1).b.(a 1 J2)

The approach suggested by Pazel deserves more attention then can be paid to it
in the present context. We leave this as a recommendation for further research.

Documentation
We have noted that the abstract solutions which are drafted in the process of
stepwise refinement can serve as structured comments to the derived program.
Linger & Mills have suggested using still more stringent rules for the inclu-
sion of comments in a program text. They have stated that all data structures
and all actions taken in a program should be explained explicitly in the form
of comments to the code.
In general, every data object introduced should have a data defini-
tion giving its contextual meaning. An action definition should
precede the program part it refers to, and a status definition should
follow the program part which produced the status."
(Linger & Mills '75, pg. 179).

Linger & Mills further suggested preceding each iteration loop with a specifi-
cation of the corresponding loop-invariant assertion.

In Bulow '74 it is suggested to concentrate all comments to a program text in

a sort of 'table of contents' at the beginning of the program. The only com-
148

ments which appear in the program text itself are then the 'chapter headings'
etc. Similar proposals were published by Berry '75.

One other important method of making a program text more understandable is the
hierarchical indentation of that text. Clearly, this indentation is merely an
additional tool to make structured programs understandable. It will, however,
not make a poor design much stronger. An amusing example of this was given by
Van Amstel '77. Van Amstel quoted a person who had the impression that struc-
tured programming boiled down to requiring that sentences like:
'The one who denounces the one who pulled down the pole that stood
near the bridge in the road that leads to Worms which was recently
opened over the canal of ten feet tall within ten days will be re-
warded.'
should,be written as follows:

'The one
who denounces the one
who pulled down the pole
that stood near the bridge
in the road
that leads to Worms
which was recently opened
over the canal
of ten feet tall
within ten days
will be rewarded.'

A well-structured 'sentence', however, should have quite another structure. In

Van Amstel's opinion the well-structured equivalent of the sentence given above
would be more ,like:
'In the road that leads to Worms there is a bridge over a canal.
This bridge was recently opened. Near that bridge there stood a
pole of ten feet tall. This pole has been pulled down. The one who
reports the responsible person within ten days, will be rewarded.'
(The original Dutch example is more concise, but it could not be translated
literally without losing the characteristic structure.)

Programming languages
The question of whether the existing programming languages are suited to struc-
tured programming practice has been raised repeatedly. Already in 1962 McCarthy
wrote:
"At present, programming languages are constructed in a very unsys-
tematic way. (-) A better understanding of the structure of compu-
tations and of data spaces will make it easier to see what features
are really desirable."
(McCarthy '62).

More than once the language FORTRAN has been criticized for its lack of struc-
ture, and the omnipresence of goto-jumps. But, FORTRAN is not really an excep-
tion: in fact no language seems to meet the standards completely.
Lecarme posed the rhetorical question to his readers:
"Is it possible to write structured programs (-) in any language?"

Lecarme answered promptly:

"The answer is certainly no, and this has already been said for
several times."
(Lecarme '74, pg. 16).
 149

The subject was discussed among others by Wulf at aZ. '71; Peck '72; Clark at
al. '73; Wirth '74, and in Update, Vol. 2, No. 5, pg. 8-10, May 1977.
In Lecarme '72 the following requirement was formulated:
"... conciseness, structure and clarity (-) are probably the most
important qualities of any program. We intentionally omit efficien-
cy, because we believe that a program that is concise, structured
and clear must be efficient otherwise the programming language is
bad."

The programming languages which seem to answer this requirement best include
ALGOL-68, ALGOL-W, and PASCAL.
 150

3.3. DISCUSSION

The structured programming methodologies have been critized quite often:

Jones noted that:
"There is often, especially initially, some consumer resistance to
the idea when it is first preached to practising programmers. (-)
But, after some oscillations between 'it's nonsense' and 'we've
always done that' most programmers would be prepared to use such a
framework for recording their work."
(Jones '77, pg. 315).

Gries gave a plausible explanation for this initial resistance:

"The defensive attitude about structured programming (-) occurs part-
ly because people object to being told that they don't know how to
do their jobs efficiently."
(Gries '74, pg. 656).

More fundamental objections have however been raised against the methodology,
which we will consider here. The criticisms can be divided into four topics:
• pragmatics,
• complexity,
• rigidity, and
• methodology.

3.3.1. Pragmatics

A very practical objection raised against the structured programming techniques

is that many programming languages lack the adequate facilities for it
(Abrahams '75). The most widely used programming languages (FORTRAN, COBOL,
PL/1, ALGOL 60) are not really suited for the techniques. They even tend to
make the structured programs less efficient. The rebuttal to this objection can
clearly be that if no adequate languages are available at the moment, they
should be made available on short notice. Lately, several attempts have been
made to define more suitable languages (ALGOL-W, PASCAL, etc.). The chance that
they will eventually repress the existing languages is however not always esti-
mated as high.

3.3.2. Complexity

Another criticism is that the structured design techniques are in fact just as
•complicated as the traditional methods (Abrahams '75). To choose the right data
representation, the right program structure, is still far from trivial.
It is then argued that 'poor programmers' will still write 'poor programs' even
if they apply the structured programming techniques. This criticism can however
be refuted. Structured programming cannot pretend to be a 'magic wand' which
alters 'poor programmers' into first-class programmers. It does, however, offer
the instruments for good program design. As always these instruments can be
used correctly and incorrectly. Denning explained:
"The understandability of the product remains basically a matter of
style: some programmers have good style, others do not. Good program-
ming style is not automatically introduced by the rules of struc-
tured programming, any more than good English prose style is guaran-
teed by following the ten famous rules listed in Strunk & White's:
The Elements of Style ... though these rules can go a long way to-
ward influencing programmers or writers in the direction of good
style, ..."
(Denning '74, pg. 6).
 151

3.3.3. Rigidity

Many authors have 'complained' that the rules of structured programming are too
strict. Some say that their strict application can lead to inferior programs.
Some even claim that their strict application is impossible.

Abrahams, for instance, questions whether each program can be given a 'tree-
structure' (hierarchical structure), whether global variables can always be
avoided in large programs, and whether goto-jumps can always be avoided. Johnson
stated much earlier:
"In some cases, it is difficult or impossible to write a program
without crossing one branch line by another."
(Johnson '70, pg. 122).

Berzheim '75 further objected that:

(1) the iteration construct is sometimes too strong, for instance, for program
segments that must be repeated just once;
(2) overlapping iteration structures can be of value, but are forbidden;
(3) there is a need for the representation of 'error-exits'.
The need for a clear notation of 'error-exits' has been stressed repeatedly,
and as such it would of course not contradict the goals of structured program-
ming to include such a special construct in a language. The flow-of-control in
a program should however be prevented from being diffused by such special con-
structs. One obvious requirement is then that it should not be possible to
'jump' from the error-routines back into the program at some intermediate point.

Another item in what has been called the 'goto-controversy' is the use of loop-
exit structures. The classic example is the 'search-loop'. A search-loop has
two or more possible outcomes: the search is successful (in one or more ways)
or it is not. Each outcome may require a different response. Without special
constructs for the programming of such loop-exit structures one is tempted to
use goto-statements instead.
The represenation of the loop-exit structures is greatly facilitated if one
decouples the declaration of iteration conditions and loop-structures.

In the traditional loop constructs WHILE(...)D0 ... OD and REPEAT ... UNTIL(..)
the iteration condition must be evaluated either at the start or at the end of
the loop-body; they can however not be evaluated within the loop-body.
An alternative construct, much in line with the proposals of e.g. Knuth & Floyd
'71, and Wirth '67 (pg. 29) would be to use the key-word REPEAT in combination
with the exit-statements EXITIF(condition; DOING: statements). The exit should
then be from the innermost block or loop. With a still more liberal approach
we could allow for exits from any level in the nesting hierarchy. The level to
be exited must then be specified: EXITIF(condition; FROM: level; DOING : state-
ments).
In figure 3.9 we have given a complex non-decomposable flow-chart discussed by
Peterson et al. '74. In figure 3.10 the description of that flow-chart in a
structogramme, extended with EXITIF statements, is given. The structogramme was
obtained by a redrawing of the flow-chart into the more structured one of
figure 3.11 in which a few blocks of code have been duplicated. For comparison
we reproduce the notation of the same flow-chart with the aid of a WHILE-UNTIL
construct (without code-duplications), taken from Friedman & Shapiro '74 (pg.
11), below.
The difference in clarity of the control flow in the programs need hardly be
stressed.
 S

a1 a2

b1

C l C2

d2 d1

Figure 3.0.
Non-Decomnoeable Flow Chart
 153

REPEAT (I)

REPEAT (:1)
EXITIF(A El)
a2
EXITIF( i:):I; DOING:d2)

(Lb di
a 1

EXITIF (B: I; DOING:b2)

bi

REPEAT ( ilL)

EXITIF( C: TTT)

c2
EXITIF(T3- :I; DOING: b2)
hi

IP.
CI
EXITIF(D:I; DOING: d2 )
di

(I)

Figure 3.10.
Structogramme With REPEAT/EXITIF Construct
154

ID,

Figure 3.11.
Redrawn Flowchart for Fig. 3.9.
(with some code-duplications)
 155

S; while (if A
then (if (while true repeat al until true)
then (if (while B repeat bi until
(if C then true
else not(while true repeat c2 until true)))
then (while true repeat ci until true)
else not(while true repeat b 2 until true))
else false)
else (while true repeat a 2 until true))
repeat until (if D then not(while true repeat d1 until true)
else (while true repeat d2 until true)); T

3.3.4. Methodology

The most fundamental criticisms on structured programming refer to the actual

design process and the relation to problem solving techniques.
In the first place there is still confusion about the true nature of the design
process. The question is whether structured program designing implies pure
'top-down' programming or not.
Denning says:
"I am not sure what 'top-down programming' means; I have seen the
phrase used in at least two ways:
a. It is a method of decomposition into modules by stepwise refine-
ment.
b. It is the goal of evolving a program so that its final structure
can be presented or described as a hierarchy of tasks or blocks."
(Denning '74a, pg. 7).

Many authors identify structured programming with top-down programming in the

first sense (a). Peter Naur, for instance, wrote:
"Thus, as one example, the insistence of Dijkstra and Wirth on step-
wise refinement could well get in conflict with the first two pro-
les solving precepts of Hyman and Anderson (Hyman & Anderson '65):
Precept I: 'Run over the elements of the problem in rapid succession
several times, until a pattern emerges which encompasses
all these elements simultaneously';
Precept II: 'Suspend judgement. Don't jump to conclusions.' "
(Naur '72, pg. 348).

The conflict described by Naur is however only apparent. The first precept Naur
gives resembles Polya's advice to 'discover the major lines of the argument'
and the rule of abstraction from structured programming.
The second precept is completely in agreement with the design rule of stepwise
refinement, and confirms Polya's remark:
"... it is not reasonable to check minor details before we have good
reasons to believe that the major steps of the argument are sound."
(Polya '48, pg. 69).

Every programmer knows from his own experience that the task of designing pro-
gram, or the task of solving a problem, is hardly ever executed as a pure top-
down process. Denning (citing Wilkes) made the following illustrative compari-
son:
"[compare] writing a textbook to writing a program: (-) the process
of creating, and of evolving, that structure can hardly be charac-
156

terized as a top-down process (-) it is a matter of individual

creativity and style."
(Denning '74b, pg. 6).

That program designing or problem solving is no pure top-down process does how-
ever not imply that it lacks all structure. Oddly enough Polya gives his read-
ers that impression when he gives the following, playful description of the
'rules of discovery':
"The first rule of discovery is to have brains and good luck.
The second rule of discovery is to sit tight and wait till you get
a bright idea."
(Polya '48, pg. 172).

Some problems are however so inherently complex that indeed no method other
than the ones described,here - by Polya may be of use. One such a problem is the
problem of the 'Tower of Hanoi'. Peck '72 made a comparison of the abilities of
a number of programming languages to give a clear representation of the program
which solves this problem. None of the examples is very understandable though.
These observations do, however, not really undermine the applicability of the
structured programming techniques, especially because there are no alternative
techniques which have proved to be more powerful.
 157

3.4. CONCLUSION

The four principles of structured programming (abstraction, restriction, step-

wise refinement, and clear notation) are powerful instruments for the discovery
of transparent solutions of programming problems. The principles can be used as
a guideline in program design; they should however not be interpreted as
'dogma's'. The principles are intended as a means to realize the specific goal
of understandability; they are not an independent goal in themselves.
158

3.5. REFERENCES

Abrahams, P. (1975), Structured programming considered harmful, Sigplan Notices,

Vol. 10, No. 4, pp. 13 24, April 1975.
-

Ashcroft, E. & Manna, Z. (1971), The translation of 'goto' programs into 'while'
programs, Computer Science Dept., Report CS-188, Stanford University, also
presented at IFIP '71, Jan. 1971.
Baker, F.T. (1975), The pros and cons of structured programming, Data Manage-
ment, Vol. 13, No. 9, pp. 60-71, Sept. 1975.
Balzer, R.M. (1967), Dataless programming, Proc. AFIPS 1967 FJCC, AFIPS Press,
Montvale (N.J.), pp. 535-544, 1967.
Berry, D.M. (1975), Structured documentation, Sigplan Notices, Vol. 10, No. 11,
pp. 7-12, Nov. 1975.
Berzheim, H. (1975), Fernwirkungsfreie, strukturierte Programmierung, Online,
Vol. 13, No.. 4, pp. 234,, 237-240, 1975.
Bloom, A.M. (1975), The 'else' must go, too, Datamation (U.S.A.), Vol. 21, No.
5, pp. 123-128.
Bohm, C. & Jacopini, G. (1966), Flow diagrams, Turing machines and languages
with only two formation rules, Comm. ACM, Vol. 9, No. 5, pp. 366-371, May
1966.
Bulow, K. (1974), Programming in book format, Datamation, Vol. 20, No. 10, Oct.
1974.
Canning, R.G. (ed.) (1974), The advent of structured programming, EDP Analyzer
(U.S.A.), Vol. 12, No. 6, pp. 1-14, June 1974.
Clark, B.L. et al. (1973), Reflections on a language design to write an oper-
ating system, Sigplan Notices, Vol. 8, No. 9, 1973.
Cooper, D.C. (1967), Bohm and Jacopini's reduction of flow charts, Comm. ACM,
Vol. 10, No. 8, pp. 463, 1967.
Dahl, 0.J., Dijkstra, E.W. &.Hoare, C.A.R. (1972), Structured programming,
Academic Press, London, 1972, 220 pgs.
Davies, I.K. (1971), The management of learning, McGraw-Hill, London, 1971.
Davies, I.K. (1976), Objectives in curriculum design, McGraw-Hill, London, 1976.
Denning, P.J. (1974a), Is it not time to define structured programming?, Oper-
ating Systems Review, Vol. 8, No. 1, pp. 6 7, Jan. 1974.
-

Denning, P.J. (1974b), Is structured programming any longer the right term?,
Operating Systems Review, Vol. 8, No. 4, pp. 4-6, April 1974.
Dennis, J.B. (1975), Modularity, Lect. Notes in Computer Science, Vol. 30, pp.
128-182, Springer Verlag, 1975.
Donaldson, J.R. (1973), Structured programming, Datamation, Vol. 19, No. 12,
pp. 52-54, 1973.
Dijkstra, E.W. (1961), On the design of machine independent programming lan-
guages, Report MR34-Math. Centre, Amsterdam, Oct. 1961, 18 pgs.
Dijkstra, E.W. (1968a), GO TO statement considered harmful, Comm. ACM, Vol. 11,
No. 3, pp. 147-148, 538, 541, March 1968.
Dijkstra, E.W. (1968b), The structure of the 'THE' multiprogramming system,
Comm. ACM, Vol. 11, No. 5, pp. 341 346, May 1968.
-

Dijkstra, E.W. (1969a), Complexity controlled by hierarchical ordering of func-

tion and variability, In: Naur, P. & Randell, B. (eds.), 'Software Engineer-
ing', Sc. Aff. Div., NATO, Brussels, pp. 114-116, 1969.
Dijkstra, E.W. (1969b), Notes on structured programming, THE Report, Univ. of
Techn. Eindhoven, The Netherlands, EWD-249, 70-Wsk-03, August 1969, see
also Dahl et al. '72.
Dijkstra, E.W. (1972), Hierarchical ordering of sequential processes, Operating
Systems Techniques, New York, Academic Press, 1972.
Dijkstra, E.W. (1975), Correctness concerns and, among other things, why they
are resented, Sigplan Notices, Vol. 10, No. 6, pp. 546-550, 1975.
Dijkstra, E.W. (1976), A discipline of programming, Englewood Cliffs, Prentice
Hall, 1976, 220 pgs.
 159

Early, J. (1971), Toward an understanding of data structures, Comm. ACM, Vol.

14, No.. 10, pp. 617 627, Oct. 1971.
-

Evans, R.V. (1974), Multiple exits from loops using neither goto nor labels,
comm. ACM, Vol. 17, No. 11, pg. 650, Nov. 1974.
Flon, L. (1975), On research in structured programming, Sigplan Notices, Vol.
10, No. 10, pp. 16-17, 1975.
Friedman, D.P. & Shapiro, S.C. (1974), A case for Sigplan
Notices, Vol. 9, No. 7, pp. 7-14, July 1974.
Garren, S. (1973), Structured programming, DECUS Fall 1973, Symposium, San
Fransisco, Cal., U.S.A., Nov. 1973, pp. 37-39.
Goodenough, J.B. (1975), Structured exception handling, In: Conference Record
of the 2nd ACM Symposium on the 'Principles of programming languages', pp.
204 224, Jan. 1975.
-

Gries, D. (1974), On structured programming, a reply to Smoliar, Comm. ACM,

Vol. 17, No. 11, pp. 655-657, Nov. 1974.
Harding, D.A. (1972), Modular programming why not!, Austr. Comput. J., Vol.
-

4, No. 4, pp. 150-156, Nov. 1972.

Hyman, R. & Anderson, B. (1965), Solving problems, Int. Science and Technology,
Sept. 1965, pp. 36-41.
Infotech (1978), Structured programming, practice and experience, Infotech
State of the Art Report, 1978, 900 pgs., 5 Volumes, Infotech Int. Ltd.,
Berkshire, U.K.
Jackson, M.A. (1975), Principles of program design, AFIC Studies in Data
Processing, No. 12, Hoare, C.A.R. (ed.), Academic Press, 1975.
Johnson, L.R. (1970), System structure in data, programs, and computers,
Prentice Hall, Englewood Cliffs (N.J.), 1970.
Jones, C.B. (1977), Structured design and coding: theory versus practice, In-
formatie, Vol. 19, No. 6, pp. 311-319, June 1977.
Kernighan, B.W. & Plauger, P.J. (1974), Programming style: examples and coun-
ter examples, Computing Surveys, Vol. 6, No. 4, pp. 303-319, 1974.
Knuth, D.E. & Floyd, R.W. (1971), Notes on avoiding goto statements, Inf. Proc.
Letters, pp. 23-31, 1971.
Knuth, D.E. (1974), Structured programming with 'goto' statements, Computing
Surveys, Vol. 6, No. 4, pp. 261 301, 1974.
-

Kuhn, T. (1962), The structure of scientific revolutions, Univ. of Chicago

Press, Chicago & London, 1962.
Leavenworth, B.M. (ed.) (1972), The 'goto' controversy, a debate, (Programming
with(out) the 'goto' ) Proc. 1972 ACM, National Conference, Vol. 2, pp.
782-797.
Lecarme, O. (1973), What programming language?, Proc. IFIP-TC-2, Working Conf.
on 'Programming Teaching Techniques', Poland, Sept. 1972, North-Holland
Publ. Co., Amsterdam 1973, pp. 61-74.
Lecarme, O. (1974), Structured programming, programming teaching and the lan-
guage PASCAL, Sigplan Notices, Vol. 9, No. 7, pp. 15 21, July 1974.
-

Linger, R.C. & Mills, H.D. (1975), Definitional text in structured programming,
Proc. of the 14th Annual Technical Symposium on: 'Computing. in the mid
70's: and assesment', Gaithersburg (Md.), June 1975. Reprinted in: Proc.
IEEE Compcon '75, Tutorial on Structured, rogramming, pp. 178 183. .
-

Liskov, B.H. (1972), A design methodology for reliable software systems, Proc.
of the 1972 FJCC, AFIPS Press, Montvale (N.J.), pp. 191 199, 1972.
-

McCarthy, J. (1963), Towards a mathematical science of computation, Proc. IFIP

Congress 1962, Munich, North Holland Publ. Co., Amsterdam, pp. 21 28, 1963.
- -

McClure, C.L. (1975), Top down, bottom up, and structured programming, IEEE
- -

Trans. on Software Eng., Vol. SE 1, No. 4, pp. 397 403, Dec. 1975.
- -

McCracken, D. (1973), Revolution in programming, an overview, Datamation, Vol.

19, No. 12, pp. 50 52, Dec. 1973.
-

Melekian, N. (1976), New methods and techniques of programming, (in German),

IBM Nachrichten, Vol. 26, No. 299, pp. 48-55, 1976.
160

Mills, H.D. (1972), Mathematical foundations of structured programming, IBM,

Federal Sys.tems Div., Gaithersburg, Maryland, 1972.
Moos, A. & Steinbuch, P.A. (1976), Structogramme and structured programming,
(in German), Buerotechnik, Vol. 24, No. 9, pp. 39 43, Sept. 1976.
-

Nassi, J. & Shneiderman, B. (1973), Flow chart techniques for structured pro-
gramming, Sigplan Notices, Vol. 8, No. 8, pp. 12 26, 1973.
-

Naur, P. (1966), Proof of algorithms by general snapshots, BIT, Vol. 6, pp.

310-316, 1966.
Naur, P. (1972), An experiment on program development, BIT, Vol. 12, pp. 347-
365, 1972.
Newton, G.E. (1975), A partially annotated bibliography of top-down and goto-
less programming, In: Proc. 'Proving and improving programs', Arc et Senans,
France, July 1975, Rocquencourt, IRIA, pp. 435 473, 1975.
-

Parnas, D.L. (1972), On the criteria to be used in decomposing systems into

modules, Comm. ACM, Vol.,15, No. 12, pp. 1053 1058, Dec. 1972.
-

Pazel, D.P. (1975), Mathematical construct for program reorganization, IBM

Journal of Research & Development, Vol. 19, No. 11, pp. 575 581, Nov. 1975.
-

Peck, J.E.L. (1973), Comparison of languages, In: Programming Teaching Tech-

niques, Proc. IFIP 1972, North Holland Publ. Co., Amsterdam 1973.
Peterson, W.W., Kasami, T. &•okura, N. (1973), On the capabilities of '1Dhile',
'repeat', and 'exit' statements, Comm. ACM, Vol. 16, No. 8, pp. 503-512, '73.
Polya, G. (1971), How to solve it, Doubleday Anchor, New York, 1948 and 1957,
Reprinted by: Princeton Press, 1971.
Ross, D.T. et al. (1975), Software engineering, process, principles and goals,
Computer, Vol. 8, pp. 17 27, May 1975.
-

Schmitz, A. (1975), Das strukturierte Ablaufdiagram zur graphischen Inter-

pretation von Programm-Abldufen, (in German), Online, Vol. 13, No. 10, pp.
665-667, 1975.
Schnupp, P. (1974), Structogramme - a new method of system planning, (in German)
Online, Vol. 12, No. 11, pp. 736-743, 1974.
Shneiderman, B. & Scheuerman, P. (1974), Structured data structures, Comm. ACM,
Vol. 17, No. 10, pp. 566-574, 1974.
Van Amstel, J.J. (1977), Methods for structured programming, (in Dutch), Infor-
matie, Vol. 19, No. 6, pp. 325-338, June 1977.
VanderBrug, G.J. (1974), On structured programming and problem reductson, Re-
port NSF-OCA-GJ32258X-29, Univ. of Maryland, College Park, U.S.A., Jan.
1974, 11 pgs.
Weiderman, N.H. et al. (1975), Flow-charting loops without cycles, Sigplan
Notices, Vol. 10, No. 4, pp. 37 46, 1975.
-

Weinberg, G.M. (1971), The psychology of computer programming, New York, Van
Nostrand Reinhold Co., 1971.
Weinberg, G.M., Geller, D.P. & Plum, T.W.S. (1975), 'If then else' considered
- -

harmful, Sigplan Notices, Vol. 10, No. 8, pp. 34-44, August 1975.
Wirth, N. (1967), On certain basic concepts of programming languages, Report
Stanford Univ.,Comp. Sc. Dept., May 1967, 30 pgs.
Wirth, N. (1971), Program development by stepwise refinement, Comm. ACM, Vol.
14, No. 4, pp. 221 227, April 1971.
-

Wirth, N. (1974), On the composition of well structured programs, Computing

Surveys, Vol. 6, No. 4, pp. 247 259, 1974.
-

Wulf, W.A. et'al. (1971), Reflections on a systems programming language, Sig-

plan Notices, Vol. 6, No. 9, 1971.
Wulf, W.A. & Shaw, M. (1973), Global variables considered harmful, Sigplan
Notices, Vol. 8, No. 2, pp. 28-34, Febr. 1973.