Content uploaded by Peter Van Roy
Author content
All content in this area was uploaded by Peter Van Roy
Content may be subject to copyright.
Programming Paradigms for
Dummies: What Every
Programmer Should Know
Peter Van Roy
This chapter gives an introduction to all the main programming paradigms, their un-
derlying concepts, and the relationships between them. We give a broad view to help
programmers choose the right concepts they need to solve the problems at hand. We
give a taxonomy of almost 30 useful programming paradigms and how they are related.
Most of them differ only in one or a few concepts, but this can make a world of difference
in programming. We explain briefly how programming paradigms influence language
design, and we show two sweet spots: dual-paradigm languages and a definitive lan-
guage. We introduce the main concepts of programming languages: records, closures,
independence (concurrency), and named state. We explain the main principles of data
abstraction and how it lets us organize large programs. Finally, we conclude by focus-
ing on concurrency, which is widely considered the hardest concept to program with.
We present four little-known but important paradigms that greatly simplify concurrent
programming with respect to mainstream languages: declarative concurrency (both ea-
ger and lazy), functional reactive programming, discrete synchronous programming, and
constraint programming. These paradigms have no race conditions and can be used in
cases where no other paradigm works. We explain why for multi-core processors and we
give several examples from computer music, which often uses these paradigms.
More is not better (or worse) than less, just different.
– The paradigm paradox.
1 Introduction
Programming is a rich discipline and practical programming languages are usually quite
complicated. Fortunately, the important ideas of programming languages are simple.
This chapter is intended to give readers with some programming experience a running
start for these ideas. Although we do not give formal definitions, we give precise intu-
itions and good references so that interested readers can quickly get started using the
concepts and languages that implement them. We mention all important paradigms but
we favor some little-known paradigms that deserve to be more widely used. We have
deliberately left out detailed explanations of some of the more well-known paradigms
9
Peter Van Roy
(such as functional and object-oriented programming), since they already have a huge
literature.
Solving a programming problem requires choosing the right concepts. All but the
smallest toy problems require different sets of concepts for different parts. This is why
programming languages should support many paradigms. A programming paradigm is an
approach to programming a computer based on a mathematical theory or a coherent set of
principles. Each paradigm supports a set of concepts that makes it the best for a certain
kind of problem. For example, object-oriented programming is best for problems with a
large number of related data abstractions organized in a hierarchy. Logic programming
is best for transforming or navigating complex symbolic structures according to logical
rules. Discrete synchronous programming is best for reactive problems, i.e., problems
that consist of reactions to sequences of external events. Languages that support these
three paradigms are respectively Java, Prolog, and Esterel.
Popular mainstream languages such as Java or C++ support just one or two separate
paradigms. This is unfortunate, since different programming problems need different
programming concepts to solve them cleanly, and those one or two paradigms often do
not contain the right concepts. A language should ideally support many concepts in
a well-factored way, so that the programmer can choose the right concepts whenever
they are needed without being encumbered by the others. This style of programming
is sometimes called multiparadigm programming, implying that it is something exotic
and out of the ordinary. On the contrary, in our experience it is clear that it should be
the normal way of programming. Mainstream languages are far from supporting this.
Nevertheless, understanding the right concepts can help improve programming style even
in languages that do not directly support them, just as object-oriented programming is
possible in C with the right programmer attitude.
This chapter is partly based on the book [50], familiarly known as CTM, which gives
much more information on many of the paradigms and concepts presented here. But this
chapter goes further and presents ideas and paradigms not covered in CTM. The code
examples in this chapter are written in the Oz language, which is also used in CTM.
Oz has the advantage that it supports multiple paradigms well, so that we do not have
to introduce more than one notation. The examples should be fairly self-explanatory;
whenever anything out of the ordinary occurs we explain it in the text.
Contents of this chapter
Languages, paradigms, and concepts Section 2 explains what programming
paradigms are and gives a taxonomy of the main paradigms. If your experience is limited
to one or just a few programming languages or paradigms (e.g., object-oriented program-
ming in Java), then you will find a much broader viewpoint here. We also explain how we
organize the paradigms to show how they are related. We find that it is certainly not true
that there is one “best” paradigm, and a fortiori this is not object-oriented programming!
On the contrary, there are many useful paradigms. Each paradigm has its place: each
has problems for which it gives the best solution (simplest, easiest to reason about, or
most efficient). Since most programs have to solve more than one problem, it follows
that they are best written in different paradigms.
10
Programming Paradigms for Dummies
Designing a language and its programs Section 3 explains how to design languages
to support several paradigms. A good language for large programs must support several
paradigms. One approach that works surprisingly well is the dual-paradigm language:
a language that supports one paradigm for programming in the small and another for
programming in the large. Another approach is the idea of designing a definitive language.
We present an example design that has proved itself in four different areas. The design
has a layered structure with one paradigm in each layer. Each paradigm is carefully
chosen to solve the successive problems that appear. We explain why this design is good
for building large-scale software.
Programming concepts Section 4 explains the four most important concepts in pro-
gramming: records, lexically scoped closures, independence (concurrency), and named
state. Each of these concepts gives the programmer an essential expressiveness that
cannot be obtained in any other way. These concepts are often used in programming
paradigms.
Data abstraction Section 5 explains how to define new forms of data with their oper-
ations in a program. We show the four kinds of data abstractions: objects and abstract
data types are the two most popular, but there exist two others, declarative objects and
stateful abstract data types. Data abstraction allows to organize programs into under-
standable pieces, which is important for clarity, maintenance, and scalability. It allows
to increase a language’s expressiveness by defining new languages on top of the existing
language. This makes data abstraction an important part of most paradigms.
Deterministic concurrent programming Section 6 presents deterministic concur-
rent programming, a concurrent model that trades expressiveness for ease of program-
ming. It is much easier to program in than the usual concurrent paradigms, namely
shared-state concurrency and message-passing concurrency. It is also by far the easiest
way to write parallel programs, i.e., programs that run on multiple processors such as
multi-core processors. We present three important paradigms of deterministic concur-
rency that deserve to be better known. The price for using deterministic concurrency is
that programs cannot express nondeterminism, i.e., where the execution is not completely
determined by the specification. For example, a client/server application with two clients
is nondeterministic since the server does not know from which client the next command
will come. The inability to express nondeterminism inside a program is often irrelevant,
since nondeterminism is either not needed, comes from outside the program, or can be
limited to a small part of the program. In the client/server application, only the com-
munication with the server is nondeterministic. The client and server implementations
can themselves be completely deterministic.
Constraint programming Section 7 presents the most declarative paradigm of our
taxonomy, in the original sense of declarative: telling the computer what is needed in-
stead of how to calculate it. This paradigm provides a high level of abstraction for solving
problems with global conditions. This has been used in the past for combinatorial prob-
lems, but it can also be used for many more general applications such as computer-aided
composition. Constraint programming has achieved a high degree of maturity since its
11
Peter Van Roy
Figure 1. Languages, paradigms, and concepts
origins in the 1970s. It uses sophisticated algorithms to find solutions that satisfy global
conditions. This means that it genuinely delivers on its ambitious claims.
Conclusions and suggestions for going further Section 8 concludes by reiterating
why programming languages should support several paradigms. To understand the “soul”
of each paradigm and to gain experience programming with different paradigms, we
recommend the use of a multiparadigm language. A multiparadigm language permits
programming in each paradigm without interference from other paradigms. The two
most extensive multiparadigm languages are the dynamically typed language Oz [50]
and the statically typed language Alice [38].
2 Languages, paradigms, and concepts
This section gives the big picture of programming paradigms, the languages that realize
them, and the concepts they contain. There are many fewer programming paradigms
than programming languages. That is why it is interesting to focus on paradigms rather
than languages. From this viewpoint, such languages as Java, Javascript, C#, Ruby, and
Python are all virtually identical: they all implement the object-oriented paradigm with
only minor differences, at least from the vantage point of paradigms.
Figure 1 shows the path from languages to paradigms and concepts. Each program-
ming language realizes one or more paradigms. Each paradigm is defined by a set of
programming concepts, organized into a simple core language called the paradigm’s ker-
nel language. There are a huge number of programming languages, but many fewer
paradigms. But there are still a lot of paradigms. This chapter mentions 27 different
paradigms that are actually used. All have good implementations and practical applica-
tions. Fortunately, paradigms are not islands: they have a lot in common. We present a
taxonomy that shows how paradigms are related.
12
Programming Paradigms for Dummies
No
Named stateUnnamed state (seq. or conc.)
More
Expressiveness of state
Less
nondeterminism?
Observable
Yes
functional
programming
Descriptive
declarative
programming
Imperative
search
programming
Event−loop
programming
Multi−agent
programming
Message−passing
concurrent
programming
Data structures only
+ unification
Dataflow and
Oz, Alice, Curry Oz, Alice, Curry
CLU, OCaml, Oz
E in one vat
Continuation
programming
Logic and
constraints message passing
Message passing Shared state
+ nondeterministic
(channel)
Oz, Alice, Curry, Excel,
AKL, FGHC, FCP
+ synch. on partial termination
FrTime, Yampa
Discrete synchronous
programming
Esterel, Lustre, Signal
Functional reactive
programming (FRP)
Continuous synchronous
programming
Pipes, MapReduce
Nondet. state
Erlang, AKL
CSP, Occam,
E, Oz, Alice,
publish/subscribe,
tuple space (Linda)
+ clocked computation
Dijkstra’s GCL
+ cell (state)
+ nondet. choice
programming
Imperative
Pascal, C
programming
Guarded
command
choice
Nonmonotonic
dataflow
programming
Concurrent logic
programming
Oz, Alice, AKL
+ port
Multi−agent
dataflow
programming
+ local cell
Active object
programming
Object−capability
programming
Turing complete
Java, OCaml
+ closure
embeddings
+ solver
LIFE, AKL
CLP, ILOG Solver
+ thread
+ single assignment
+ thread
Smalltalk, Oz,
+ thread
Java, Alice
+ log
+ cell
(state)
Functional
SQL embeddings
Prolog, SQL
+ search
record
XML,
S−expression
Haskell, ML, E
(unforgeable constant)
+ cell
Scheme, ML
+ procedure
+ closure
SNOBOL, Icon, Prolog
+ search
(channel)
+ port
Scheme, ML
(equality)
+ name
+ by−need synchronization
+ by−need
synchronization
+ thread
+ continuation
Lazy concurrent
object−oriented
Concurrent
programming
Shared−state
concurrent
programming
Software
transactional
memory (STM)
Sequential
object−oriented
programming
Stateful
functional
programming
Lazy
declarative
concurrent
programming
programming
Lazy
dataflow
Concurrent
constraint
programming
constraint
programming
Constraint (logic)
programming
Relational & logic
programming
Deterministic
logic programming
synchron.
+ by−need + thread
+ single assign.
Haskell
Lazy
functional
programming
Monotonic
dataflow
programming
Declarative
concurrent
programming
ADT
functional
programming
ADT
imperative
programming
Functional
programming
First−order
Figure 2. Taxonomy of programming paradigms
2.1 Taxonomy of programming paradigms
Figure 2 gives a taxonomy of all major programming paradigms, organized in a graph
that shows how they are related [55]. This figure contains a lot of information and re-
wards careful examination. There are 27 boxes, each representing a paradigm as a set
of programming concepts. Of these 27 boxes, eight contain two paradigms with different
names but the same set of concepts. An arrow between two boxes represents the concept
or concepts that have to be added to go from one paradigm to the next. The concepts
are the basic primitive elements used to construct the paradigms. Often two paradigms
that seem quite different (for example, functional programming and object-oriented pro-
gramming) differ by just one concept. In this chapter we focus on the programming
concepts and how the paradigms emerge from them. With n concepts, it is theoretically
possible to construct 2
n
paradigms. Of course, many of these paradigms are useless in
practice, such as the empty paradigm (no concepts)
1
or paradigms with only one concept.
A paradigm almost always has to be Turing complete to be practical. This explains why
functional programming is so important: it is based on the concept of first-class function,
1
Similar reasoning explains why Baskin-Robbins has exactly 31 flavors of ice cream. We postulate
that they have only 5 flavors, which gives 2
5
− 1 = 31 combinations with at least one flavor. The 32
nd
combination is the empty flavor. The taste of the empty flavor is an open research question.
13
Peter Van Roy
or closure, which makes it equivalent to the λ-calculus which is Turing complete. Of the
2
n
possible paradigms, the number of practically useful paradigms is much smaller. But
it is still much larger than n.
When a language is mentioned under a paradigm in Figure 2, it means that part of
the language is intended (by its designers) to support the paradigm without interference
from other paradigms. It does not mean that there is a perfect fit between the language
and the paradigm. It is not enough that libraries have been written in the language to
support the paradigm. The language’s kernel language should support the paradigm.
When there is a family of related languages, usually only one member of the family is
mentioned to avoid clutter. The absence of a language does not imply any kind of value
judgment. There are just too many good languages to mention them all.
Figure 2 shows two important properties of the paradigms: whether or not they have
observable nondeterminism and how strongly they support state. We now discuss each
of these properties in turn.
Observable nondeterminism
The first key property of a paradigm is whether or not it can express observable nonde-
terminism. This is identified in Figure 2 by boxes with a heavy or light border. We recall
that nondeterminism is when the execution of a program is not completely determined
by its specification, i.e., at some point during the execution the specification allows the
program to choose what to do next. During the execution, this choice is made by a part
of the run-time system called the scheduler. The nondeterminism is observable if a user
can see different results from executions that start at the same internal configuration.
This is highly undesirable. A typical effect is a race condition, where the result of a
program depends on precise differences in timing between different parts of a program
(a “race”). This can happen when the timing affects the choice made by the scheduler.
But paradigms that have the power to express observable nondeterminism can be used
to model real-world situations and to program independent activities.
We conclude that observable nondeterminism should be supported only if its expres-
sive power is needed. This is especially true for concurrent programming. For exam-
ple, the Java language can express observable nondeterminism since it has both named
state and concurrency (see below). This makes concurrent programming in Java quite
difficult [29]. Concurrent programming is much easier with the declarative concurrent
paradigm, in which all programs are deterministic. Sections 6 and 7 present four impor-
tant concurrent paradigms that do not have observable nondeterminism.
Named state
The second key property of a paradigm is how strongly it supports state. State is the
ability to remember information, or more precisely, to store a sequence of values in time.
Its expressive power is strongly influenced by the paradigm that contains it. We distin-
guish three axes of expressiveness, depending on whether the state is unnamed or named,
deterministic or nondeterministic, and sequential or concurrent. This gives eight combi-
nations in all. Later in this chapter we give examples of many of these combinations. Not
all of the combinations are useful. Figure 3 shows some useful ones arranged in a lattice;
14
Programming Paradigms for Dummies
unnamed, deterministic, sequential!
unnamed, deterministic, concurrent!named, deterministic, sequential!
unnamed, nondeterministic, concurrent!
named, nondeterministic, sequential!
named, nondeterministic, concurrent!
Declarative paradigms (relational and functional)!
Deterministic concurrency!
Concurrent logic programming!Guarded command programming!
Imperative programming !
Message-passing and shared-state concurrency!
Less!
Expressiveness of state!
More!
Figure 3. Different levels of support for state
adjacent boxes differ in one coordinate.
2
One intriguing box shown is Dijkstra’s guarded
command language (GCL) [14]. It has named state and nondeterministic choice in a
sequential language. It uses nondeterministic choice to avoid overspecifying algorithms
(saying too much about how they should execute).
The paradigms in Figure 2 are classified on a horizontal axis according to how strongly
they support state. This horizontal axis corresponds to the bold line in Figure 3. Let us
follow the line from top to bottom. The least expressive combination is functional pro-
gramming (threaded state, e.g., DCGs in Prolog and monads in functional programming:
unnamed, deterministic, and sequential). Adding concurrency gives declarative concur-
rent programming (e.g., synchrocells: unnamed, deterministic, and concurrent). Adding
nondeterministic choice gives concurrent logic programming (which uses stream mergers:
unnamed, nondeterministic, and concurrent). Adding ports or cells, respectively, gives
message passing or shared state (both are named, nondeterministic, and concurrent).
Nondeterminism is important for real-world interaction (e.g., client/server). Named state
is important for modularity (see Section 4.4).
Both observable nondeterminism and named state are cases where it is important to
choose a paradigm that is expressive enough, but not too expressive (see epigram at the
head of the chapter). Each of these two concepts is sometimes needed but should be left
out if not needed. The point is to pick a paradigm with just the right concepts. Too few
and programs become complicated. Too many and reasoning becomes complicated. We
will give many examples of this principle throughout this chapter.
2.2 Computer programming and system design
Figure 4 gives a view of computer programming in the context of general system design.
This figure adds computer programming to a diagram taken from Weinberg [56]. The
two axes represent the main properties of systems: complexity (the number of basic
interacting components) and randomness (how nondeterministic the system’s behavior
is). There are two kinds of systems that are understood by science: aggregates (e.g., gas
2
Two of the eight possible combinations are not shown in the figure. We leave it to the reader to
discover them and find out if they make any sense!
15
Peter Van Roy
Computer programming
Computer programming
Figure 4. Computer programming and system design (adapted from Weinberg [56])
molecules in a box, understood by statistical mechanics) and machines (e.g., clocks and
washing machines, a small number of components interacting in mostly deterministic
fashion). The large white area in the middle is mostly not understood. The science of
computer programming is pushing inwards the two frontiers of system science: computer
programs can act as highly complex machines and also as aggregates through simulation.
Computer programming permits the construction of the most complex systems.
Modern programming languages have evolved over more than five decades of expe-
rience in constructing programmed solutions to complex, real-world problems. Modern
programs can be quite complex, reaching sizes measured in millions of lines of source
code, written by large teams of programs over many years. In our view, languages that
scale to this level of complexity are successful in part because they model some essential
factors of how to construct complex systems. In this sense, these languages are not just
arbitrary constructions of the human mind. They explore the limits of complexity in a
more objective way. We would therefore like to understand them in a scientific way, i.e.,
by understanding the basic concepts that compose the underlying paradigms and how
these concepts are designed and combined. This is the deep justification of the creative
extension principle explained below.
2.3 Creative extension principle
Concepts are not combined arbitrarily to form paradigms. They can be organized ac-
cording to the creative extension principle. This principle was first defined by Felleisen
[18] and independently rediscovered in [50]. It gives us a guide for finding order in the
vast set of possible paradigms. In a given paradigm, it can happen that programs be-
come complicated for technical reasons that have no direct relationship to the specific
problem that is being solved. This is a sign that there is a new concept waiting to be
discovered. To show how the principle works, assume we have a simple sequential func-
tional programming paradigm. Then here are three scenarios of how new concepts can
be discovered and added to form new paradigms:
16
Programming Paradigms for Dummies
Figure 5. How adding exceptions to a language can simplify programs
• If we need to model several independent activities, then we will have to implement
several execution stacks, a scheduler, and a mechanism for preempting execution
from one activity to another. All this complexity is unnecessary if we add one
concept to the language: concurrency.
• If we need to model updatable memory, that is, entities that remember and update
their past, then we will have to add two arguments to all function calls relative to
that entity. The arguments represent the input and output values of the memory.
This is unwieldy and it is also not modular because the memory travels throughout
the whole program. All this clumsiness is unnecessary if we add one concept to the
language: named state.
• If we need to model error detection and correction, in which any function can detect
an error at any time and transfer control to an error correction routine, then we
need to add error codes to all function outputs and conditionals to test all function
calls for returned error codes. All this complexity is unnecessary if we add one
concept to the language: exceptions. Figure 5 shows how this works.
The common theme in these three scenarios (and many others!) is that we need to do
pervasive (nonlocal) modifications of the program in order to handle a new concept. If
the need for pervasive modifications manifests itself, we can take this as a sign that there
is a new concept waiting to be discovered. By adding this concept to the language we no
longer need these pervasive modifications and we recover the simplicity of the program.
The only complexity in the program is that needed to solve the problem. No additional
complexity is needed to overcome technical inadequacies of the language. Both Figure 2
and [50] are organized according to the creative extension principle.
3 Designing a language and its programs
A programming language is not designed in a vacuum, but for solving certain kinds of
problems. Each problem has a paradigm that is best for it. No one paradigm is best for all
problems. That is why it is important to choose carefully the paradigms supported by the
17
Peter Van Roy
language. We will look at two interesting cases: languages that support two paradigms
(Section 3.1) and layered languages (Section 3.2). The layered language we present is a
particularly interesting one because almost the same layered structure appears in four
different areas.
3.1 Languages that support two paradigms
Many languages support two paradigms, typically one for programming in the small
and another for programming in the large. The first paradigm is chosen for the kind of
problem most frequently targeted by the language. The second paradigm is chosen to
support abstraction and modularity and is used when writing large programs. Here are
a few examples:
• Prolog: The first paradigm is a logic programming engine based on unification and
depth-first search. The second paradigm is imperative: the assert and retract op-
erations which allow a program to add and remove program clauses. Prolog dates
from 1972, which makes it an old language. Recent developments in modeling lan-
guages based on advanced search algorithms advance both the logic programming
and imperative programming sides. Modern Prolog implementations have added
some of these advances, e.g., support for constraint programming and a module
system.
• Modeling languages (e.g., Comet, Numerica [48]): The first paradigm is a solver:
constraint programming (see Section 7), local search (see the chapter by Philippe
Codognet [8]), satisfiability (SAT solvers), and so forth. The second paradigm is
object-oriented programming.
• Solving libraries (e.g., Gecode): The first paradigm is a solver library based on
advanced search algorithms, such as Gecode [43, 47]. The second paradigm is added
by the host language, e.g., C++ and Java support object-oriented programming.
• Language embedding (e.g., SQL): SQL already supports two paradigms: a relational
programming engine for logical queries of a database and a transactional interface
for concurrent updates of the database. The host language complements this by
supporting object-oriented programming, for organization of large programs. This
example goes beyond two paradigms to show a design with three complementary
paradigms.
3.2 A definitive programming language
At some point in time, language research will give solutions that are good enough that
researchers will move on to work at higher levels of abstraction. This has already arrived
for many subareas of language design, such as assembly languages and parsing algo-
rithms. In the 1970s, compiler courses were built around a study of parsing algorithms.
Today, parsing is well understood for most practical purposes and compiler design has
moved on. Today’s compiler courses are built around higher level topics such as dataflow
analysis, type systems, and language concepts. We postulate that this kind of evolution
is happening with language design as well.
18
Programming Paradigms for Dummies
Layer Language project
Erlang [6, 5] E [32, 31] Distrib. Oz [10] Didactic Oz [50]
Functional
programming
(see Section 4.2)
A process is a re-
cursive function
in its own thread,
employing clo-
sures for hot code
update
An object is a
recursive func-
tion with a local
state
Functions, pro-
cedures, classes,
and components
are closures with
efficient distrib.
protocols
Closures are the
foundation of all
paradigms
Deterministic
concurrency
(see Section 6)
(not supported) Deterministic
execution of all
objects in one
vat (process)
Dataflow concur-
rency with effi-
cient protocol for
dataflow variables
Concurrency is as
easy as functional
programming, no
race conditions
Message-passing
concurrency
(see Section 4.3)
Fault tolerance
by isolation, fault
detection with
messages
Security by iso-
lation, messages
between objects
in different vats
Asynchronous
message pro-
tocols to hide
latency
Multi-agent pro-
gramming is ex-
pressive and easy
to program
Shared-state
concurrency
(see Section 4.4)
Global database
(Mnesia) keeps
consistent states
(not supported) Coherent global
state protocols;
transactions for
latency and fault
tolerance
Named state for
modularity
Table 1. Layered structure of a definitive programming language
This section presents the structure of one possible definitive language [52]. We study
four research projects that were undertaken to solve four very different problems. The
solutions achieved by all four projects are significant contributions to their respective
areas. All four projects considered language design as a key factor to achieve success. The
surprise is that all four projects ended up using languages with very similar structures.
Table 1 shows the common properties of the programming language invented in each
of the four projects. The common language has a layered structure with four layers: a
strict functional core, followed by declarative concurrency, then asynchronous message
passing, and finally global named state. This layered structure naturally supports four
paradigms. We briefly summarize the four projects:
1. Erlang Programming highly available embedded systems for telecommunications.
This project was undertaken by Joe Armstrong and his colleagues at the Ericsson
Computer Science Laboratory starting in 1986. The Erlang language was designed
and a first efficient and stable implementation was completed in 1991 [5, 6]. An
Erlang program consists of isolated named lightweight processes that send each
other messages. Because of the isolation, Erlang programs can be run almost
unchanged on distributed systems and multi-core processors. The Erlang system
has a replicated database, Mnesia, to keep global coherent states. Erlang and its
programming platform, the OTP (Open Telecom Platform) system, are being used
successfully in commercial systems by Ericsson and other companies [57, 17].
2. E Programming secure distributed systems with multiple users and multiple se-
curity domains. This project was undertaken over many years by different insti-
tutions. It started with Dennis and Van Horn’s capability model in 1965 [13] and
Carl Hewitt’s Actor model in 1973 [24] and it led via concurrent logic program-
ming to the E language designed by Doug Barnes, Mark Miller, and their colleagues
[32, 31]. Predecessors of E have been used to implement various multiuser virtual
19
Peter Van Roy
environments. An E program consists of isolated single-threaded vats (processes)
hosting active objects that send each other messages. Deterministic concurrency is
important in E because nondeterminism can support a covert channel.
3. Distributed Oz Making network-transparent distributed programming practical.
This project started in 1995 in the PERDIO project at the DFKI with the re-
alization that the well-factored design of the Oz language, first developed by Gert
Smolka and his students in 1991 as an outgrowth of the ACCLAIM project, was
a good starting point for making network transparent distribution practical [45].
This resulted in the Mozart Programming System which implements Distributed
Oz and was first released in 1999 [22, 34]. Recent work has both simplified Mozart
and increased its power for building fault-tolerance abstractions [10].
4. Didactic Oz Teaching programming as a unified discipline covering all popular
programming paradigms. This project started in 1999 with the realization by the
author and Seif Haridi that Oz is well-suited to teaching programming because it
has many programming concepts in a well-factored design, it has a simple semantics,
and it has a high-quality implementation. The textbook [50], published in 2004,
“reconstructs” the Oz design according to a principled approach (see Section 2.3).
The book is the basis of programming courses now being taught at several dozen
universities worldwide. The author has been using it at UCL since 2003 for his
second-year programming course given to all engineering students and his third-
year concurrent programming course. The second-year course (since 2005 called
FSAB1402) is particularly interesting since it covers the three most important
paradigms, functional, object-oriented, and dataflow concurrent programming, with
many practical techniques and a formal semantics [51].
From the common structure of these designs, one can infer several plausible consequences
for language design. First, that the notion of declarative programming is at the very core
of programming languages. This is already well-known; our study reinforces this con-
clusion. Second, that declarative programming will stay at the core for the foreseeable
future, because distributed, secure, and fault-tolerant programming are essential topics
that need support from the programming language. Third, that deterministic concur-
rency is an important form of concurrent programming that should not be ignored. We
remark that deterministic concurrency is an excellent way to exploit the parallelism of
multi-core processors because it is as easy as functional programming and it cannot have
race conditions (see also Section 6) [53]. A final conclusion is that message-passing con-
currency is the correct default for general-purpose concurrent programming instead of
shared-state concurrency.
3.3 Architecture of self-sufficient systems
We have presented some preliminary conclusions about a definitive language; let us now
be ambitious and widen our scope to software systems. The ultimate software system is
one that does not require any human assistance, i.e., it can provide for every software
modification that it needs, including maintenance, error detection and correction, and
adaptation to changing requirements. Such a system can be called self sufficient [44].
Self-sufficient systems can be very robust; for example peer-to-peer networks can manage
20
Programming Paradigms for Dummies
Subsystem
Monitoring agentActuating agent
Calculate corrective action
Figure 6. A single feedback loop
themselves to survive in extremely hostile environments by doing reversible phase transi-
tions [44, 54]. Let us leave aside for now the artificial intelligence required to build such
a system, and investigate just the language mechanisms it needs. The system may ask
for human assistance in some cases, but in principle it should contain all the mechanisms
it needs to accomplish its tasks.
What is a reasonable architecture for designing self-sufficient systems? From the
conclusions of the previous section and our experience in building distributed systems,
we can propose an architecture. In terms of programming paradigms, what we need
first is components as first-class entities (specified by closures) that can be manipulated
through higher-order programming. Above this level, the components behave as isolated
concurrent agents that communicate through message passing. Finally, we need named
state and transactions for system reconfiguration and system maintenance. Named state
allows us to manage the content of components and change their interconnections. This
gives us a language that has a layered structure similar to the previous section.
With this language we can program our system. To allow the program to adapt
itself to its environment, we take inspiration from biological systems and organize its
components as feedback loops. The system then consists of a set of interacting feedback
loops. A single feedback loop consists of three concurrent components that interact with
a subsystem (see Figure 6): a monitoring agent, a correcting agent, and an actuating
agent. Realistic systems consist of many feedback loops. Since each subsystem must be
as self-sufficient as possible, there must be feedback loops at all levels. These feedback
loops can interact in two fundamental ways:
• Stigmergy: Two loops share one subsystem.
• Management: One loop controls another loop directly.
Figure 8 gives a real-world example from biology: the human respiratory system [49].
This system contains four loops. Three loops form a tower connected by management.
The fourth loop interacts with the others through stigmergy.
The style of system design illustrated by the human respiratory system can be applied
to programming. A program then consists of a set of feedback loops interacting through
stigmergy and management. Figure 7 shows part of the Transmission Control Protocol
as a feedback loop structure [49]. The inner loop implements reliable transfer of a byte
stream using a sliding window protocol. The outer loop does congestion control: if too
many packets are lost, it reduces the transfer rate of the inner loop by reducing the
window size. In our view, the large-scale structure of software will more and more be
done in this self-sufficient style. If it is not done in this way, the software will simply be
too fragile and collapse with any random error or problem.
21
Peter Van Roy
Send
Inner loop (reliable transfer)
Outer loop (congestion control)
Calculate policy modification
Actuator
(send packet)
Monitor Monitor
throughput
Calculate bytes to send
(modify throughput)
(sliding window protocol)
destination and receives ack)
(network that sends packet to
Subsystem
(receive ack)
Send
stream
acknowledgement
Figure 7. TCP as a feedback loop structure
Render unconscious
reflex
Breathing
Laryngospasm
(seal air tube)
in blood
CO2
Measure
in blood
O2
Measure
Trigger breathing reflex
when CO2 increases to threshold
Increase or decrease breathing rate
Conscious control
when O2 falls to threshold
Trigger unconsciousness
(and reduce CO2 threshold to base level)
(maximum is breath!hold breakpoint)
of body and breathing
and change CO2 threshold
Trigger laryngospasm temporarily
in airways
Detect
obstruction
when sufficient obstruction in airways
breathing
Monitor
Other inputs
in human body
Breathing apparatus
Actuating agents Monitoring agents
Figure 8. The human respiratory system as a feedback loop structure
22
Programming Paradigms for Dummies
4 Programming concepts
Programming paradigms are built out of programming concepts. In this section we
present the four most important programming concepts, namely records, lexically scoped
closures, independence (concurrency), and named state. We explain the concepts and
why they are important for programming.
4.1 Record
A record is a data structure: a group of references to data items with indexed access to
each item. For example:
R=chanson(nom:"Le Roi des Aulnes"
artiste:"Dietrich Fischer-Dieskau"
compositeur:"Franz Schubert"
langue:allemand)
The record is referenced by the identifier R. Members can be references through the dot
operation, e.g., R.nom returns a reference to the string "Le Roi des Aulnes". The
record is the foundation of symbolic programming. A symbolic programming language is
able to calculate with records: create new records, decompose them, and examine them.
Many important data structures such as arrays, lists, strings, trees, and hash tables can
be derived from records. When combined with closures (see next section), records can
be used for component-based programming.
4.2 Lexically scoped closure
Definition of
procedure P
Context D
Context D
x
state
Call of
procedure P
x
Context C
Context C
1. Definition 2. Call
Figure 9. Definition and call of a closure
The lexically scoped closure is an enormously powerful concept that is at the heart
of programming. Functional programming, which is programming with closures, is a
23
Peter Van Roy
central paradigm (see Figure 2). From an implementation viewpoint, a closure combines
a procedure with its external references (the references it uses at its definition). From
the programmer’s viewpoint, a closure is a “packet of work”: a program can transform
any instructions into a closure at one point in the program, pass it to another point,
and decide to execute it at that point. The result of its execution is the same as if the
instructions were executed at the point the closure was created.
Figure 9 shows schematically what happens when a closure is defined and when it is
called. The procedure P is implemented by a closure. At the definition (context D), P
stores the references from the definition context. For example, it keeps the reference x
to some named state. We say that the environment (set of references) of P is closed over
its definition context. At the call (context C), P uses the references from context D.
Figure 10 shows one possible use for a closure: creating a control structure. At the
left, we execute the instruction <stmt>. At the right, instead of executing <stmt>, we
place it inside a procedure (closure) referenced by P (the example uses Oz syntax). Any
time later on in the program, we can decide to call P. We have separated the definition
of <stmt> from its execution. With this ability we can define control structures such as
an if statement or while loop.
…
…
<stmt>
…
…
…
…
P=proc {$} <stmt> end
…
…
…
{P}
Instruction
Define procedure referenced by P
(context D)
Call the procedure
(context C)
Figure 10. Example: modifying a program to separate creation and execution
The example of Figures 9 and 10 can easily be generalized to procedures with ar-
guments. The closed environment exists as before. The arguments are passed during
each call. The closure therefore has two sets of references: a closed environment (from
the definition) and the arguments (from each call). Almost all programming languages
(except for a few venerable ancestors such as Pascal and C) use this kind of closure:
• functions are closures;
• procedures are closures;
• objects are closures;
• classes are closures;
• software components are closures.
Many abilities normally associated with specific paradigms are based on closures:
• Instantiation and genericity, normally associated with object-oriented program-
ming, can be done easily by writing functions that return other functions. In
object-oriented programming the first function is called a “class” and the second is
called an “object”.
24
Programming Paradigms for Dummies
• Separation of concerns, normally associated with aspect-oriented programming, can
be done easily by writing functions that take other functions as arguments. For ex-
ample, Erlang has a function that implements a generic fault-tolerant client/server.
It is called with a function argument that defines the server’s behavior. Aspect-
oriented programming in object-oriented languages is explained in the chapter by
Pierre Cointe [9]. It is usually done by syntactic transformations (called “weaving”)
that add aspect code to the original source. The AspectJ language is a good exam-
ple of this approach. Weaving is difficult to use because it is fragile: it is easy to
introduce errors in the program (changing the source code changes the semantics
of the program). Using closures instead makes it easier to preserve correctness
because the source code is not changed.
• Component-based programming is a style of programming in which programs are
organized as components, where each component may depend on other components.
A component is a building block that specifies part of a program. An instance of a
component is called a module, which is a record containing closures. A new module
is created by a function that takes its dependent modules as inputs. The component
is the function.
The Erlang language implements all these abilities directly with closures. This is practical
and scalable: successful commercial products with more than one million lines of Erlang
code have been developed (e.g., the AXD-301 ATM switch [57]). In most other languages,
though, the use of closures is hidden inside the language’s implementation and is not
available directly to the programmer. If done carefully this can be an advantage, since
the implementation can guarantee that the closures are used correctly.
4.3 Independence (concurrency)
Another key concept is independence: constructing a program as independent parts. This
is not as simple as it may seem. For example, consider a program that consists of instruc-
tions executing one after the other. The instructions are not independent since they are
ordered in time. To implement independence we need a new programming concept called
concurrency. When two parts do not interact at all, we say they are concurrent.
3
(When
the order of execution of two parts is given, we say they are sequential.) Concurrent parts
can be extended to have some well-defined interaction, which is called communication.
Concurrency should not be confused with parallelism. Concurrency is a language
concept and parallelism is a hardware concept. Two parts are parallel if they execute
simultaneously on multiple processors. Concurrency and parallelism are orthogonal: it is
possible to run concurrent programs on a single processor (using preemptive scheduling
and time slices) and to run sequential programs on multiple processors (by parallelizing
the calculations). Parallel execution on multi-core processors is explained on page 38.
The real world is concurrent: it consists of activities that evolve independently. The
computing world is concurrent as well. It has three levels of concurrency:
3
Technically, a program’s execution consists of a partial order of state transition events and two events
are concurrent if there is no order between them.
25
Peter Van Roy
• Distributed system: a set of computers connected through a network. A concurrent
activity is called a computer. This is the basic structure of the Internet.
• Operating system: the software that manages a computer. A concurrent activity
is called a process. Processes have independent memories. The operating system
handles the task of mapping the process execution and memory to the computer.
For example, each running application typically executes in one process.
• Activities inside one process. A concurrent activity is called a thread. Threads
execute independently but share the same memory space. For example, the different
windows in a Web browser typically execute in separate threads.
The fundamental difference between processes and threads is how resource allocation
is done. Process-level concurrency is sometimes called competitive concurrency: each
process tries to acquire all the system’s resources for itself. The operating system’s chief
role is to arbitrate the resource requests done by all the processes and to allocate resources
in a fair way. Thread-level concurrency is sometimes called cooperative concurrency:
threads in a process share resources and collaborate to achieve the result of the process.
Threads run in the same application and so are guided by the same program.
There are two popular paradigms for concurrency. The first is shared-state concur-
rency: threads access shared data items using special control structures called monitors
to manage concurrent access. This paradigm is by far the most popular. It used by
almost all mainstream languages, such as Java and C#. Another way to do shared-state
concurrency is by means of transactions: threads atomically update shared data items.
This approach is used by databases and by software transactional memory. The second
paradigm is message-passing concurrency: concurrent agents each running in a single
thread that send each other messages. The languages CSP (Communicating Sequential
Processes) [25] and Erlang [6] use message passing. CSP processes send synchronous
messages (the sending process waits until the receiving process has taken the message)
and Erlang processes send asynchronous messages (the sending process does not wait).
Despite their popularity, monitors are the most difficult concurrency primitive to
program with [29]. Transactions and message passing are easier, but still difficult. All
three approaches suffer from their expressiveness: they can express nondeterministic
programs (whose execution is not completely determined by their specifications), which
is why it is hard to reason about their correctness. Concurrent programming would be
much simpler if the nondeterminism were controlled in some way, so that it is not visible
to the programmer. Sections 6 and 7 present four important paradigms that implement
this idea to make concurrent programming much simpler.
4.4 Named state
The final key concept we will introduce is named state. State introduces an abstract
notion of time in programs. In functional programs, there is no notion of time. Functions
are mathematical functions: when called with the same arguments, they always give the
same results. Functions do not change. In the real world, things are different. There are
few real-world entities that have the timeless behavior of functions. Organisms grow and
learn. When the same stimulus is given to an organism at different times, the reaction
will usually be different. How can we model this inside a program? We need to model
an entity with a unique identity (its name) whose behavior changes during the execution
26
Programming Paradigms for Dummies
of the program. To do this, we add an abstract notion of time to the program. This
abstract time is simply a sequence of values in time that has a single name. We call
this sequence a named state. Unnamed state is also possible (monads and DCGs, see
Section 2.1), but it does not have the modularity properties of named state.
Figure 11 shows two components, A and B, where component A has an internal named
state (memory) and component B does not. Component B always has the same behavior:
whenever it is called with the same arguments, it gives the same result. Component A
can have different behaviors each time it is called, if it contains a different value in its
named state. Having named state is both a blessing and a curse. It is a blessing because
it allows the component to adapt to its environment. It can grow and learn. It is a curse
because a component with named state can develop erratic behavior if the content of the
named state is unknown or incorrect. A component without named state, once proved
correct, always stays correct. Correctness is not so simple to maintain for a component
with named state. A good rule is that named state should never be invisible: there
should always be some way to access it from the outside.
Component A
Memory
Component B
No memory
Figure 11. A component with named state and a component without named state
Named state and modularity
Named state is important for a system’s modularity. We say that a system (function,
procedure, component, etc.) is modular if updates can be done to part of the system
without changing the rest of the system. We give a scenario to show how we can design
a modular system by using named state. Without named state, this is not possible.
Assume that we have three developers, P, U1, and U2. P has developed a module M
that contains two functions F and G. U1 and U2 are users of M: their own programs used
module M. Here is one possible definition of M:
fun {ModuleMaker}
fun {F ...}
... % Definition of F
end
fun {G ...}
... % Definition of G
end
in
themodule(f:F g:G)
end
M={ModuleMaker} % Creation of M
27
Peter Van Roy
The function ModuleMaker is a software component, i.e., it defines the behavior of
part of a system. We create instances of this component by calling ModuleMaker. One
such instance is the module M. Note that a module’s interface is simply a record, where
each field is one of the module’s operations. The module M has two operations F and G.
Now assume that developer U2 has an application that consumes a huge amount of
calculation time. U2 would like to investigate where all this time is being spent, so that
he can rewrite his application to be less costly. U2 suspects that F is being called too
many times and he would like to verify this. U2 would like a new version of M that counts
the number of times F is called. So U2 contacts P and asks him to create a new version
of M that does this, but without changing the interface (that defines the operations of
M and how they are called) since otherwise U2 would have to change all of his program
(not to mention U1!).
Surprise! This is not possible without named state. If F does not have named state
then it cannot change its behavior. In particular, it cannot keep a counter of how many
times it is called. The only solution in a program without named state is to change F’s
interface (its arguments):
fun {F ... Fin Fout}
Fout=Fin+1
...
end
We add two arguments to F, namely Fin and Fout. When calling F, Fin gives the count
of how many times F was called, and F calculates the new count in Fout by adding one
to Fin. When calling F, we have to link all these new arguments together. For example,
three successive calls to F would look like this:
A={F ... F1 F2}
B={F ... F2 F3}
C={F ... F3 F4}
F1 is the initial count. The first call calculates F2, which is passed to the second call,
and so forth. The final call returns the count F4. We see that this is a very bad solution,
since U2 has to change his program wherever F is called. It gets worse: U1 also has to
change his program, even though U1 never asked for any change. All users of M, even U1,
have to change their programs, and they are very unhappy for this extra bureaucratic
overhead.
The solution to this problem is to use named state. We give an internal memory
to the module M. In Oz, this internal memory is called a cell or a variable cell. This
corresponds simply to what many languages call a variable. Here is the solution:
28
Programming Paradigms for Dummies
fun {ModuleMaker}
X={NewCell 0} % Create cell referenced by X
fun {F ...}
X:=@X+1 % New content of X is old plus 1
... % Original definition of F
end
fun {F ...}
... % Original definition of G
end
fun {Count} @X end % Return content of X
in
themodule(f:F g:G c:Count)
end
M={ModuleMaker}
The new module M contains a cell inside. Whenever F is called, the cell is incremented.
The additional operation Count (accessed by M.c) returns the current count of the cell.
The interfaces of F and G are unchanged. Now everybody is happy: U2 has his new
module and nobody has to change their programs at all since F and G are called in the
same way. This shows how named state solves the problem of modularity.
S
1
Functional
program
(no state)
S
2
Input state Output state
Figure 12. A program as state transformer
The main advantage of named state is that the program becomes modular. The main
disadvantage is that a program can become incorrect. It seems that we need to have and
not have named state at the same time. How do we solve this dilemma?
4
One solution
is to concentrate the use of named state in one part of the program and to avoid named
state in the rest. Figure 12 shows how this design works. The bulk of the program is a
pure function without named state. The rest of the program is a state transformer: it
calls the pure function to do the actual work. This concentrates the named state in a
small part of the program.
5 Data abstraction
A data abstraction is a way to organize the use of data structures according to precise
rules which guarantee that the data structures are used correctly. A data abstraction
has an inside, an outside, and an interface between the two. All data structures are kept
on the inside. The inside is hidden from the outside. All operations on the data must
pass through the interface. Figure 13 shows this graphically. There are three advantages
to this organization:
4
This kind of dilemma is at the heart of invention. It is called a technical contradiction in Altshuller’s
Theory of Inventive Problem Solving (TRIZ), which provides techniques for its solution [2].
29
Peter Van Roy
Op1 Op2
Op3
Inside
Outside
Interface
Figure 13. A data abstraction
1. First, there is a guarantee that the data abstraction will always work correctly.
The interface defines the authorized operations on the data structures and no other
operations are possible.
2. Second, the program is easier to understand. A user of the data abstraction does
not need to understand how the abstraction is implemented. The program can
be partitioned into many abstractions, implemented independently, which greatly
reduces the program’s complexity. This can be further improved by adding the
property of compositionality: allowing data abstractions to be defined inside of
other data abstractions.
3. Third, it becomes possible to develop very large programs. We can divide the
implementation among a team of people. Each abstraction has one person who is
responsible for it: he implements it and maintains it. That person has to know just
the interfaces used by his abstraction.
In the rest of this section we first explain the four different ways to organize data ab-
stractions. We then introduce two principles, polymorphism and inheritance, that greatly
increase the power of data abstraction to organize programs. Object-oriented program-
ming, as it is usually understood, is based on data abstraction with polymorphism and
inheritance.
5.1 Objects and abstract data types
There are four main ways to organize data abstractions, organized along two axes. The
first axis is state: does the abstraction use named state or not. The second axis is
bundling: does the abstraction fuse data and operations into a single entity (this is called
an object or a procedural data abstraction (PDA)), or does the abstraction keep them
separate (this is called an abstract data type (ADT)). Multiplying the two axes gives four
possibilities, which are shown in Figure 14.
Two of these four possibilities are especially popular in modern programming lan-
guages. We give examples of both in the Java language. Integers in Java are represented
as values (1, 2, 3, etc.) and operations (+, -, *, etc.). The values are passed as arguments
to the operations, which return new values. This is an example of an abstract data type
without named state. Objects in Java combine the data (their attributes) and the oper-
ations (their methods) into a single entity. This is an example of an object with named
state.
30
Programming Paradigms for Dummies
Bundling
State
Object
(bundled)
Abstract data type
(unbundled)
Stateless
Stateful
“Pure” object
“Pure” ADT
Declarative object
Stateful ADT
Very popular!
For example, an integer
Figure 14. The four ways to organize a data abstraction
The two other possibilities, the abstract data type with named state and the declar-
ative object, can also be useful. But they are less used in current languages.
5.2 Polymorphism and the responsability principle
The most important principle of object-oriented programming, after data abstraction
itself, is polymorphism. In everyday language, we say an entity is polymorphic if it can
take on different forms. In computer programming, we say an entity is polymorphic if
it can take arguments of different types. This ability is very important for organizing
large programs so that the responsibilities of the program’s design are concentrated in
well-defined places instead of being spread out over the whole program. To explain this,
we use a real-world example. A sick patient goes to see a doctor. The patient does not
need to be a doctor, but just to tell the doctor one message: “Cure me!”. The doctor
understands this message and does the right thing depending on his speciality. The
program “GetCured” run by the patient is polymorphic: it takes a doctor as argument
and works with all different kinds of doctors. This is because all doctors understand the
message “Cure me!”.
For programming the idea of polymorphism is similar: if a program works with one
data abstraction as argument, it can work with another, if the other has the same inter-
face. All four kinds of data abstractions we saw before support polymorphism. But it
is particularly simple for objects, which is one reason for the success of object-oriented
programming.
Figure 15 gives an example. Consider a graphics package which includes routines for
drawing different kinds of figures. We define this using class declarations. Look at the
definition of CompoundFigure. This defines figures that consist of a list of other figures
(even other compound figures!). The method draw in CompoundFigure does not know
how to draw any of these figures. But since it is polymorphic, it can call draw in the
other figures. Each figure knows how to draw itself. This is a correct distribution of
responsibilities.
31
Peter Van Roy
class Figure
...
end
class Circle
attr x y r
meth draw ... end
...
end
class Line
attr x1 y1 x2 y2
meth draw ... end
...
end
class CompoundFigure
attr figlist
meth draw
for F in @figlist do {F draw} end
end
...
end
Figure 15. An example of polymorphism in a graphics package
5.3 Inheritance and the substitution principle
The second important principle of object-oriented programming is inheritance. Many
abstractions have a lot in common, in what they do but also in their implementations.
It can be a good idea to define abstractions to emphasize their common relationship and
without repeating the code they share. Repeated code is a source of errors: if one copy
is fixed, all copies have to be fixed. It is all too easy to forget some copies or to fix them
in the wrong way.
Inheritance allows to define abstractions incrementally. Definition A can inherit from
another definition B: definition A takes definition B as its base and shows how it is
modified or extended. The incremental definition A is called a class. However, the
abstraction that results is a full definition, not a partial one.
Inheritance can be a useful tool, but it should be used with care. The possibility of
extending a definition B with inheritance can be seen as another interface to B. This
interface needs to be maintained throughout the lifetime of B. This is an extra source of
bugs. Our recommendation is to use inheritance as little as possible. When defining a
class, we recommend to define it as nonextensible if at all possible. In Java this is called
a final class.
Instead of inheritance, we recommend to use composition instead. Composition is
a natural technique: it means simply that an attribute of an object refers to another
object. The objects are composed together. In this way, it is not necessary to extend
a class with inheritance. We use the objects as they are defined to be used. Figure 16
illustrates inheritance and composition side by side.
32
Programming Paradigms for Dummies
B
A
O
A
!!instance of
B
A
O
A
O
B
!!instance of
!!instance of
attr b: O
B
inherits from
Inheritance Composition
Figure 16. Inheritance versus composition
If you must use inheritance, then the right way to use it is to follow the substitution
principle. Suppose that class A inherits from class B and we have two objects, O
A
and
O
B
. The substitution principle states that any procedure that works with objects O
B
of
class B must also work with objects O
A
of class A. In other words, inheritance should
not break anything. Class A should be a conservative extension of class B.
We end our discussion of inheritance with a cautionary tale. In the 1980s, a very
large multinational company
5
initiated an ambitious project based on object-oriented
programming. Despite a budget of several billion dollars, the project failed miserably.
One of the principal reasons for this failure was a wrong use of inheritance. Two main
errors were committed:
• Violating the substitution principle. A procedure that worked with objects of a class
no longer worked with objects of a subclass. As a result, many almost-identical
procedures needed to be written.
• Using subclasses to mask bugs. Instead of correcting bugs, subclasses were created
to mask bugs, i.e., to test for and handle those cases where the bugs occurred. As
a result, the class hierarchy was very deep, complicated, slow, and filled with bugs.
6 Deterministic concurrent programming
One of the major problems of concurrent programming is nondeterminism. An execution
of a program is nondeterministic if at some point during the execution there is a choice
of what to do next. Nondeterminism appears naturally when there is concurrency: since
two concurrent activities are independent, the program’s specification cannot say which
executes first. If there are several threads ready to run, then in each execution state the
system has to choose which thread to execute next. This choice can be done in different
ways; typically there is a part of the system called the scheduler that makes the choice.
Nondeterminism is very hard to handle if it can be observed by the user of the
program. Observable nondeterminism is sometimes called a race condition. For example,
if each of two threads assigns a variable cell to a different value, then each of the two
values can be observed:
5
Which shall remain anonymous.
33
Peter Van Roy
declare C={NewCell 0}
thread C:=1 end
thread C:=2 end
The variable cell C can contain the value 1 or 2 after both threads execute. This is a
simple case of a race condition. Much trickier cases are possible, when two threads share
several variable cells and do more calculation with them. Debugging and reasoning about
programs with race conditions is very difficult.
6.1 Avoiding nondeterminism in a concurrent language
The easiest way to eliminate race conditions is to design a language that does not have
nondeterminism. But this would be throwing the baby out with the bathwater since
concurrency naturally implies nondeterminism. How can we avoid the ill effects of non-
determinism and still have concurrency?
6
We can solve this problem by making a clear
distinction between nondeterminism inside the system, which cannot be avoided, and
observable nondeterminism, which may be avoidable. We solve the problem in two steps:
• First, we limit observable nondeterminism to those parts of the program that really
need it. The other parts should have no observable nondeterminism.
• Second, we define the language so that it is possible to write concurrent programs
without observable nondeterminism.
Concurrent paradigm Races Inputs can be Example languages
possible? nondeterm.?
Declarative
concurrency
No No Oz [34], Alice [38]
Constraint
programming
No No Gecode [43], Numerica [48]
Functional reactive
programming
No Yes FrTime [12], Yampa [27]
Discrete synchronous
programming
No Yes Esterel [7], Lustre [21], Signal [26]
Message-passing
concurrency
Yes Yes Erlang [6], E [32]
Table 2. Four deterministic concurrent paradigms and one that is not
Is it possible to have a concurrent language without observable nondeterminism?
A superficial examination of popular programming languages might lead one to say no:
Java and C# use shared-state concurrency and Erlang uses message-passing concurrency,
all of which have observable nondeterminism. Fortunately, this superficial impression
is completely wrong. There are at least four useful programming paradigms that are
concurrent but have no observable nondeterminism (no race conditions). Table 2 lists
these four together with message-passing concurrency. Let us explain them in more
detail.
6
This is another example of a technical contradiction. See footnote on page 29.
34
Programming Paradigms for Dummies
Declarative concurrency (also called monotonic dataflow) In this paradigm, de-
terministic inputs are received and used to calculate deterministic outputs. This paradigm
lives completely in a deterministic world. If there are multiple input streams, they must
be deterministic, i.e., the program must know exactly what input elements to read to
calculate each output (for example, there could be a convention that exactly one ele-
ment is read from each input stream). Two languages that implement this paradigm
are Oz [50, 34] and Alice [38]. This paradigm can be made lazy without losing its good
properties. The paradigm and its lazy extension are explained in more detail in Sec-
tion 6.2. Constraint programming is related to declarative concurrency and is explained
in Section 7.
There exists also a nonmonotonic dataflow paradigm, in which changes on any input
are immediately propagated through the program. The changes can be conceptualized as
dataflow tokens traveling through the program. This paradigm can accept nondetermin-
istic input, but it has the disadvantage that it sometimes adds its own nondeterminism
that does not exist in the input (called a “glitch” below). That is why we do not discuss
this paradigm further in this chapter. Functional reactive programming is similar to
nonmonotonic dataflow but without the glitches.
Functional reactive programming (also called continuous synchronous pro-
gramming) In this paradigm, programs are functional but the function arguments
can be changed and the change is propagated to the output. This paradigm can accept
nondeterministic input and does not add any nondeterminism of its own. Semantically,
the arguments are continuous functions of a totally ordered variable (which can corre-
spond to useful magnitudes such as time or size). Implementations typically recompute
values only when they change and are needed. Discretization is introduced only when
results are calculated [16]. This means that arbitrary scaling is possible without losing
accuracy due to approximation. If the changes are propagated correctly, then the func-
tional program does not add any nondeterminism. For example, the simple functional
expression x+(x
*
y) with x=3 and y=4 gives 15. If x is changed to 5, then the expres-
sion’s result changes from 15 to 25. Implementing this naively with a concurrent stream
connecting a times agent to a plus agent is incorrect. This implementation can give a
glitch, for example if the new value of x reaches the addition before the new result of
the multiplication. This gives a temporary result of 17, which is incorrect. Glitches
are a source of nondeterminism that the implementation must avoid, for example by
compile-time preprocessing (doing a topological sort of operations) or thread scheduling
constraints. Some languages that implement this paradigm are Yampa (embedded in
Haskell) [27] and FrTime (embedded in Scheme) [12].
Discrete synchronous programming In this paradigm, a program waits for input
events, does internal calculations, and emits output events. This is called a reactive
system. Reactive systems must be deterministic: the same sequence of inputs produces
the same sequence of outputs. Like functional reactive programming, this paradigm can
accept nondeterministic input and does not add any nondeterminism of its own. The
main difference is that time is discrete instead of continuous: time advances in steps
from one input event to the next. Output events are emitted at the same logical time
35
Peter Van Roy
instants as the input events.
7
All calculations done to determine the next output event are
considered to be part of the same time instant. This is exactly what happens in clocked
digital logic: combinational circuits are “instantaneous” (they happen within one cycle)
and sequential circuits “take time”: they use clocked memory (they happen over several
cycles). The clock signal is a sequence of input events. Using discrete time enormously
simplifies programming for reactive systems. For example, it means that subprograms
can be trivially composed: output events from one subcomponent are instantaneously
available as input events in other subcomponents. Some languages that implement this
paradigm are Esterel [7], Lustre [21], and Signal [26]. Esterel is an imperative language,
Lustre is a functional dataflow language, and Signal is a relational dataflow language. It
is possible to combine the discrete synchronous and concurrent constraint paradigms to
get the advantages of both. This gives the Timed CC model, which is explained in the
chapter by Carlos Olarte et al [35].
All three paradigms have important practical applications and have been realized with
languages that have good implementations. It is beyond the scope of this chapter to go
into detail for all three paradigms. Because of its simplicity and importance, we give just
the basic ideas of the first paradigm, declarative concurrency, in Section 6.2.
Deterministic concurrency and computer music
Deterministic concurrency is omnipresent in computer music. We give four examples:
• The OpenMusic music composition studio provides a set of graphical tools for
composers [1]. It has an interactive visual language to define a dataflow graph
of music components (called “patches”). It has a semantics similar to discrete
synchronous programming. The main difference is explicit triggering. Explicit
triggering is used to interface between the human composer and the system. The
evaluation of a graph is triggered by the composer explicitly requesting the value
of a component. This causes a demand-driven (lazy) chain of calculations: the
component requests evaluation of the components that it depends on, and so forth
transitively until reaching components that have no dependencies. The components
have memory: the result of the evaluation is stored in the component.
• The Antescofo score follower lets music software follow human musicians when
they are playing a piece [11]. It translates clock time (seconds) into human time
(tempo, i.e., beats per minute). Antescofo is extended with a language that lets
the composer annotate the score with control instructions. This language has a
semantics similar to discrete synchronous programming, where input events are
notes and output events are the composer’s instructions. The main difference is
that Antescofo has a tempo oscillator that can schedule events on fractional notes.
This adds a continuous aspect to Antescofo’s execution.
7
Technically, a program in a synchronous language such as Esterel defines a deterministic Mealy
machine, which is a finite state automaton in which each state transition is labeled with an input and
an output.
36
Programming Paradigms for Dummies
• The Faust signal processing tool presented in the chapter by Yann Orlarey et al
provides a visual dataflow language to define audio signal processing plug-ins or
applications [37]. The dataflow language has a discrete synchronous semantics with
a functional flavor, similar to Lustre [36]. Faust is optimized for high performance:
it supports a high clock frequency and efficient compilation to C++.
• The Max/MSP programming and execution environment provides a set of graph-
ical tools for music performance using an interactive visual language to define a
dataflow graph [39]. Max/MSP has three distinct parts: the Max dataflow lan-
guage, which provides the overall control flow, the MSP digital signal processing
library, which generates the audio, and the Jitter library for video and 3D pro-
cessing. The dataflow graph somewhat resembles that of OpenMusic, although the
semantics is quite different. The Max language executes in real-time and in ea-
ger style starting from metronomes or other generators. The language is designed
to make it easy to write functional dataflow programs (respecting the functional
equations at all times, similar to functional reactive programming), although the
implementation does not enforce this.
8
The language has a sequential semantics,
since at most one message can be traversing the dataflow graph at any instant.
The sequentiality is not immediately apparent to the user, but it is important for
deterministic execution.
It is remarkable that these examples exist at three different levels of abstraction: mu-
sic composition (OpenMusic), music performance (human time scale, Max/MSP and
Antescofo), and music performance (signal processing time scale, Max/MSP and Faust).
OpenMusic, Max/MSP, and Antescofo provide a tantalizing confirmation of Table 1.
OpenMusic has a mature language organized as three layers: functional, deterministic
concurrency, and shared state. Max/MSP has a sequential core with a deterministic
concurrent layer on top. Antescofo’s language is still being designed: so far, it has just
a deterministic concurrent layer, but other layers are planned. The Hermes/dl language,
described in the chapter by Alexandre Fran¸cois, also distinguishes between deterministic
and stateful layers [19].
6.2 Declarative concurrency
We explain briefly how to do declarative concurrency, which is the first and simplest form
of deterministic concurrency (see chapter 4 of [50] for more information). Declarative
concurrency has the main advantage of functional programming, namely confluence, in
a concurrent model. This means that all evaluation orders give the same result, or in
other words, it has no race conditions. It adds two concepts to the functional paradigm:
threads and dataflow variables. A thread defines a sequence of instructions, executed
independently of other threads. Threads have one operation:
• {NewThread P}: create a new thread that executes the 0-argument procedure P.
A dataflow variable is a single-assignment variable that is used for synchronization.
Dataflow variables have three primitive operations:
8
Strange errors sometimes appear at inconvenient times during the execution of Max/MSP programs.
Some of these are likely due to program errors that result in temporary nondeterministic behavior, i.e.,
glitches. Such errors could be avoided by changing the language design or its run-time system, in ways
similar to synchronous programming.
37
Peter Van Roy
• X={NewVar}: create a new dataflow variable referenced by X.
• {Bind X V}: bind X to V, where V is a value or another dataflow variable.
• {Wait X}: the current thread waits until X is bound to a value.
Using these primitive operations, we extend all the operations of the language to wait
until their arguments are available and to bind their result. For example, we define the
operation Add in terms of dataflow variables and a primitive addition operation PrimAdd:
proc {Add X Y Z}
{Wait X} {Wait Y}
local R in {PrimAdd X Y R} {Bind Z R} end
end
The call Z={Add 2 3} causes Z to be bound to 5 (the function output is the procedure’s
third argument). We do the same for all operations including the conditional (if) state-
ment (which waits until the condition is bound) and the procedure call (which waits until
the procedure variable is bound). The result is a declarative dataflow language.
Lazy declarative concurrency
We can add lazy execution to declarative concurrency and still keep the good properties
of confluence and determinism. In lazy execution, it is the consumer of a result that
decides whether or not to perform a calculation, not the producer of the result. In a
loop, the termination condition is in the consumer, not the producer. The producer
can even be programmed as an infinite loop. Lazy execution does the least amount of
calculation needed to get the result. We make declarative concurrency lazy by adding
one concept, by-need synchronization, which is implemented by one operation:
• {WaitNeeded X}: the current thread waits until a thread does {Wait X}.
This paradigm adds both lazy evaluation and concurrency to functional programming
and is still declarative. It is the most general declarative paradigm based on functional
programming known so far.
9
With WaitNeeded we can define a lazy version of Add:
proc {LazyAdd X Y Z}
thread {WaitNeeded Z} {Add X Y Z} end
end
This is practical if threads are efficient, such as in Mozart [34]. The call Z={LazyAdd 2
3} delays the addition until the value of Z is needed. We say that it creates a lazy suspen-
sion. If another thread executes Z2={Add Z 4}, then the suspension will be executed,
binding Z to 5. If the other thread executes Z2={LazyAdd Z 4} instead, then two lazy
suspensions are created. If a third thread needs Z2, then both will be executed.
Declarative concurrency and multi-core processors
With the advent of multi-core processors, parallel programming has finally reached the
mainstream. A multi-core processor combines two or more processing elements (called
cores) in a single package, on a single die or multiple dies. The cores share the interconnect
9
Constraint programming is more general but it is based on relational programming.
38
Programming Paradigms for Dummies
to the rest of the system and often share on-chip cache memory. As transistor density
continues to increase according to Moore’s Law (doubling approximately every two years,
which is expected to continue at least until 2020) [33], the number of cores will increase
as well. To use all this processing power we need to write parallel programs.
Decades of research show that parallel programming cannot be completely hidden
from the programmer: it is not possible in general to automatically transform an arbitrary
program into a parallel program. There is no magic bullet. The best that we can do is
to make parallel programming as easy as possible. The programming language and its
libraries should help and not hinder the programmer. Traditional languages such as Java
or C++ are poorly equipped for this because shared-state concurrency is difficult.
Declarative concurrency is a good paradigm for parallel programming [53]. This is
because it combines concurrency with the good properties of functional programming.
Programs are mathematical functions: a correct function stays correct no matter how it
is called (which is not true for objects). Programs have no race conditions: any part of a
correct program can be executed concurrently without changing the results. Any correct
program can be parallelized simply by executing its parts concurrently on different cores.
If the set of instructions to execute is not totally ordered, then this can give a speedup.
Paradigms that have named state (variable cells) make this harder because each variable
cell imposes an order (its sequence of values). A common programming style is to have
concurrent agents connected by streams. This kind of program can be parallelized simply
by partitioning the agents over the cores, which gives a pipelined execution.
7 Constraint programming
In constraint programming, we express the problem to be solved as a constraint satis-
faction problem (CSP). A CSP can be stated as follows: given a set of variables ranging
over well-defined domains and a set of constraints (logical relations) on those variables,
find an assignment of values to the variables that satisfies all the constraints. Constraint
programming is the most declarative of all practical programming paradigms. The pro-
grammer specifies the result and the system searches for it. This use of search harnesses
blind chance to find a solution: the system can find a solution that is completely un-
expected by the programmer. The chapter by Philippe Codognet explains why this is
useful for artistic invention [8].
Constraint programming is at a much higher level of abstraction than all the other
paradigms in this chapter. This shows up in two ways. First, constraint programming can
impose a global condition on a problem: a condition that is true for a solution. Second,
constraint programming can actually find a solution in reasonable time, because it can use
sophisticated algorithms for the implemented constraints and the search algorithm. This
gives the solver a lot of power. For example, path-finding constraints can use shortest
path algorithms, multiplication constraints can use prime factorization algorithms, and
so forth. Because of its power in imposing both local and global conditions, constraint
programming has been used in computer-aided composition [3, 41].
Programming with constraints is very different from programming in the other
paradigms of this chapter. Instead of writing a set of instructions to be executed, the
programmer models the problem: represent the problem using variables with their do-
mains, define the problem as constraints on the variables, choose the propagators that
implement the constraints, and define the distribution and search strategies. For small
39
Peter Van Roy
constraint problems, a naive model works fine. For big problems, the model and heuris-
tics have to be designed with care, to reduce search as much as possible by exploiting
the problem structure and properties. The art of constraint programming consists in
designing a model that makes big problems tractable.
The power and flexibility of a constraint programming system depend on the expres-
siveness of its variable domains, the expressiveness and pruning power of its propagators,
and the smartness of its CSP solver. Early constraint systems were based on simple
domains such as finite trees and integers. Modern constraint systems have added real
numbers and recently also directed graphs as domains.
Constraint programming is closely related to declarative concurrency. Semantically,
both are applications of Saraswat’s concurrent constraint programming framework [42].
Like declarative concurrency, constraint programming is both concurrent and determinis-
tic. It lives a deterministic world: for a given input it calculates a given output. It differs
from declarative concurrency in two main ways. First, it replaces dataflow variables by
general constraints. Binding a dataflow variable, e.g., X=V, can be seen as an equality
constraint: X is equal to V. Second, it has a more flexible control flow: each constraint
executes in its own thread, which makes it into a concurrent agent called a propagator
(see Section 7.2). This allows the constraints to better prune the search space.
7.1 Some applications of constraint programming
Constraint programming has applications in many areas, such as combinatorics, plan-
ning, scheduling, optimization, and goal-oriented programming. The possible applica-
tions depend very much on the variable domains and constraints that are implemented
in the solver. Simple combinatorial problems can be solved with integers. The vari-
able domain corresponding to an integer is called a finite domain because it contains
a finite set of integers. When we say, for example, that x ∈ {0, · · · , 9}, we mean
that the solution for x is an element of the finite set {0, · · · , 9}. If we have eight
variables s, e, n, d, m, o, r, y, all in the set {0, · · · , 9}, then we can model the
SEND+MORE=MONEY puzzle (where each letter represents a digit) with the single
constraint 1000s+100e+10n+d+1000m+100o+10r+e = 10000m+1000o+100n+10e+y.
We add the constraints s > 0 and m > 0 to ensure the first digits are nonzero and the
constraint alldiff({s, e, n, d, m, o, r, y}) to ensure that all digits are different. To solve this
problem intelligently, the constraint solver needs just one more piece of information: a
heuristic known as the distribution strategy (see Section 7.2). For this example, a simple
heuristic called first-fail is sufficient.
Finite domains are a simple example of a discrete domain. Constraint systems have
also been built using continuous domains. For example, the Numerica system uses real
intervals and can solve problems with differential equations [48]. The difference be-
tween Numerica’s techniques and the usual numerical solution of differential equations
(e.g., Runge-Kutta or predictor-corrector methods) is that the constraint solver gives a
guarantee: the solution, if it exists is guaranteed to be in the interval calculated by the
solver. The usual methods give no guarantee but only an approximate error bound.
Graph constraints and computer music
Recent research since 2006 has introduced a very powerful discrete domain, namely di-
rected graphs. Variables range over directed graphs and the constraints define conditions
40
Programming Paradigms for Dummies
on the graphs. These can include simple conditions such as existence or nonexistence of
edges or nodes. But what makes the domain truly interesting is that it can also include
complex conditions such as transitive closure, the existence of paths and dominators,
and subgraph isomorphisms [15, 40, 58]. The complex conditions are implemented by
sophisticated graph algorithms. A Gecode library for graph constraints is in preparation
as part of the MANCOOSI project [20, 43].
Graph constraints can be used in any problem where the solution involves graphs.
The MANCOOSI project uses them to solve the package installability problem for large
open-source software distributions. Spiessens has used graph constraints to reason about
authority propagation in secure systems [46]. The nodes of an authority graph are sub-
jects and objects. An edge in an authority graph describes a permission: an entity has
a right to perform an action on another entity. A path in an authority graph describes
an authority: an entity can perform an action, either directly or indirectly. Authority
propagation problems can be formulated as graph problems. Since constraint programs
are relational, this works in both directions: to find the use conditions for a system with
given security properties or the security properties of a system with given use conditions.
A piece of music has a global order. A music score can be represented as a graph.
Because of these two facts, we hypothesize that graph constraints can be useful primitives
for computer-aided composition. For example, subgraph isomorphism can be used to find
or to impose themes throughout a composition. Probably it will be necessary to design
new graph constraints for computer music. For example, in a music score, the same theme
can often be found in different places and at different time scales, perhaps giving the score
a fractal structure. A global constraint can be designed to enforce this condition.
7.2 How the constraint solver works
In principle, solving a CSP is easy: just enumerate all possible values for all variables and
test whether each enumeration is a solution. This naive approach is wildly impractical.
Practical constraint solvers use much smarter techniques such as local search (explained
in the chapter by Philippe Codognet [8]) or the propagate-distribute algorithm (explained
in this section). The latter reduces the amount of search by alternating propagate and
distribute steps (for more information see [47], which explains the Gecode library):
• Propagate step: Reduce the domains of the variables in size as much as possible
according to the propagators. A propagator is a concurrent agent that implements
a constraint. It is triggered when the domains of any of its arguments change.
It then attempts to further reduce the domains of its arguments according to the
constraint it implements. Propagators can trigger each other through shared argu-
ments. They execute until no more reduction is possible (a fixpoint). This leads to
three possibilities: a solution, a failure (no solution), or an incomplete solution.
• Distribute step: For each incomplete solution, choose a constraint C and split the
problem P into two subproblems P ∧ C and P ∧ ¬C. This increases the number
of problems to solve, but each problem may be easier to solve since it has extra
information (C or ¬C). This step is the most primitive form of search.
The algorithm then continues with propagate steps for the two subproblems. This
creates a binary tree called the search tree. The efficiency of the propagate-distribute
algorithm depends on three factors that can be chosen independently (see Figure 17):
41
Peter Van Roy
Propagation over the
constraint domains
Distribution strategy Search strategy
!!Integers (finite domains)
!!Reals (real intervals)
!!Graphs (graph intervals)
!!…
!!Heuristic for creating the
search tree
!!Choose the constraint
used to split the problem
!!For example: Choose
first value in variable with
smallest domain (first-fail)
!!Algorithm for traversing the
search tree
!!For example: Depth-first
search, breadth-first search,
iterative deepening, limited
discrepancy search, etc.
! !
Figure 17. Constraint solver based on the propagate-distribute algorithm
• Propagation over the constraint domains. This defines how much propagation
(pruning) is done by the propagators. This depends on two factors: the sophistica-
tion of the propagators and the expressiveness of the constraint domains. Propaga-
tors can implement highly sophisticated algorithms that depend on deep theoretical
results. For example, the multiplication propagator A
*
B=:C can use factorization
algorithms to improve propagation. For positive integers A and B, the multiplication
propagator A
*
B=:12 will either reduce to A,B∈ {1, · · · , 12} or A,B∈ {1, 2, 3, 4, 6, 12},
depending on whether the constraint domains can have “holes” or not. Better prop-
agation and more expressive constraint domains reduce the number of distribution
steps (less search) at the cost of more propagation (more inferencing). Depending
on the problem, this may or may not be a good trade-off.
• Distribution strategy. This heuristic defines how the constraint C is chosen for
each distribute step. A good choice of C depends on the structure of the problem
and the distribution of its solutions. For example, the first-fail heuristic finds the
variable with the smallest domain and chooses the first value in this domain.
• Search strategy. This heuristic defines how the search tree is traversed. Typical
traversals are depth-first or breadth-first, but many more sophisticated traversals
exist, such as A*, iterative deepening, and limited discrepancy. A* finds the short-
est path by guiding the search with a heuristic function: actual distance traveled
plus estimated remaining distance to goal. The estimation must not be greater
than the actual remaining distance. Iterative deepening and limited discrepancy
do progressively wider searches, starting with a bound of 1 and incrementing the
bound after each traversal. Iterative deepening uses a depth bound and limited
discrepancy uses a discrepancy bound (for example, the number of differences with
respect to a depth-first path).
8 Conclusions and suggestions for going further
The chapter gives a quick overview of the main programming paradigms and their con-
cepts. Programming languages should support several paradigms because different prob-
lems require different concepts to solve them. We showed several ways to achieve this:
dual-paradigm languages that support two paradigms and a definitive language with four
paradigms in a layered structure. Each paradigm has its own “soul” that can only be un-
derstand by actually using the paradigm. We recommend that you explore the paradigms
42
Programming Paradigms for Dummies
by actually programming in them. Each paradigm has programming languages that sup-
port it well with their communities and champions. For example, we recommend Haskell
for lazy functional programming [28], Erlang for message-passing concurrency [6], SQL
for transactional programming, Esterel for discrete synchronous programming [7], and
Oz for declarative concurrency and constraint programming [50].
If you want to explore how to use different paradigms in one program, we recommend
a multiparadigm language like Oz [34], Alice [38], Curry [4], or CIAO [23]. Each of
these four has its own distinctive approach; pick the one you like best! For Oz there is
a textbook and a Web site that has much material including full courses in English and
French [50, 51]. There is a big difference between a language that is designed from the
start to be multiparadigm (like Oz) and a language that contains many programming
concepts (like Common Lisp). A true multiparadigm language is factored: it is possible
to program in one paradigm without interference from the other paradigms.
Acknowledgements
This chapter was written during the author’s sabbatical at IRCAM. Part of the work re-
ported here is funded by the European Union in the SELFMAN project (sixth framework
programme contract 34084) and the MANCOOSI project (seventh framework programme
grant agreement 214898). The author thanks Arshia Cont, the developer of Antescofo,
and G´erard Assayag and the other members of the RepMus group at IRCAM for inter-
esting discussions that led to some of the conclusions in this chapter.
43
Peter Van Roy
Bibliography
[1] Agon C., Assayag G. and Bresson B., OpenMusic 6.0.6, IRCAM, Music Represen-
tations Research Group, 2008.
[2] Altshuller G., “The Innovation Algorithm: TRIZ, Systematic Innovation and Tech-
nical Creativity,” Technical Innovation Center, Inc. (Translated from the Russian
by Lev Shulyak and Steven Rodman), 2005.
[3] Anders T., Composing Music by Composing Rules: Design and Usage of a Generic
Music Constraint System, Ph.D. dissertation, Queen’s University, Belfast, Northern
Ireland, Feb. 2007.
[4] Hanus A., Hanus S., and Hanus M., Curry: A Tutorial Introduction, Dec. 2007.
See www.curry-language.org.
[5] Armstrong J., Making Reliable Distributed Systems in the Presence of Software Er-
rors, Ph.D. dissertation, Royal Institute of Technology (KTH), Stockholm, Sweden,
Nov. 2003.
[6] Armstrong J., Williams M., Wikstr
¨
om C. and Virding R., Concurrent Programming
in Erlang, Prentice-Hall, 1996. See www.erlang.org.
[7] Berry G., The Esterel v5 Language Primer–Version 5.21 release 2.0,
´
Ecole des
Mines and INRIA, April 1999.
[8] Codognet P. “Combinatorics, Randomness, and the Art of Inventing”, New Com-
putational Paradigms for Computer Music, Assayag G. and Gerzso A. (eds.), IR-
CAM/Delatour France, 2009.
[9] Cointe P. “Designing Open-Ended Languages: an Historical Perspective”, New
Computational Paradigms for Computer Music, Assayag G. and Gerzso A. (eds.),
IRCAM/Delatour France, 2009.
[10] Collet R., The Limits of Network Transparency in a Distributed Programming Lan-
guage, Ph.D. dissertation, Universit´e catholique de Louvain, Louvain-la-Neuve, Bel-
gium, Dec. 2007.
[11] Cont A., “Antescofo: Anticipatory Synchronization and Control of Interactive Pa-
rameters in Computer Music”, ICMC 2008, Belfast, Northern Ireland, Aug. 2008.
[12] Cooper G. H., Integrating Dataflow Evaluation into a Practical Higher-Order Call-
by-Value Language, Ph.D. dissertation, Brown University, Providence, Rhode Is-
land, May 2008.
[13] Dennis J.B. and Van Horn E.C., “Programming Semantics for Multiprogrammed
Computations”, Communications of the ACM, 9(3), March 1966.
[14] Dijkstra E. W., A Discipline of Programming, Prentice-Hall, 1976.
[15] Dooms G., The CP(Graph) Computation Domain in Constraint Programming,
Ph.D. dissertation, Universit´e catholique de Louvain, Louvain-la-Neuve, Belgium,
2006.
44
Programming Paradigms for Dummies
[16] Elliott C., Simply Efficient Functional Reactivity, LambdaPix technical report
2008-01, April 2008.
[17] Ericsson, Open Telecom Platform—User’s Guide, Reference Manual, Installation
Guide, OS Specific Parts, Telefonaktiebolaget LM Ericsson, Stockholm, Sweden,
1996.
[18] Felleisen M., “On the Expressive Power of Programming Languages”, in 3rd Euro-
pean Symposium on Programming (ESOP 1990), May 1990, pp. 134-151.
[19] Fran¸cois, A. R. J., “Time and Perception in Music and Computation”, New Com-
putational Paradigms for Computer Music, Assayag G. and Gerzso A. (eds.), IR-
CAM/Delatour France, 2009.
[20] Guti´errez G., Graph Constraint Library for Gecode, MANCOOSI project deliver-
able, see www.mancoosi.org. In preparation, 2009.
[21] Halbwachs N. and Pascal R., A Tutorial of Lustre, Jan. 2002.
[22] Seif H., Van Roy P., Brand P., and Schulte C., “Programming Languages for Dis-
tributed Applications”, Journal of New Generation Computing, 16(3), pp. 223-261,
May 1998.
[23] Hermenegildo M. V., Bueno F., Carro M., L´opez P., Morales J. F., and Puebla G.,
“An Overview of the Ciao Multiparadigm Language and Program Development
Environment and Its Design Philosophy”, Springer LNCS 5065 (Essays Dedicated
to Ugo Montanari on the Occasion of His 65th Birthday), pp. 209-237, 2008.
[24] Hewitt C., Bishop P. and Steiger R., “A Universal Modular ACTOR Formalism for
Artificial Intelligence”. In 3rd International Joint Conference on Artificial Intelli-
gence (IJCAI), pp. 235-245, Aug. 1973.
[25] Hoare C. A. R., Communicating Sequential Processes, Prentice-Hall, 1985.
[26] Houssais B., The Synchronous Programming Language SIGNAL: A Tutorial, IRISA,
April 2002.
[27] Hudak P., Courtney A., Nilsson H., and Peterson J., “Arrows, Robots, and Func-
tional Reactive Programming”. In summer School on Advanced Functional pro-
gramming, Springer LNCS 2638, pp. 159-187, 2003.
[28] Hudak P., Peterson J. and Fasel J., A Gentle Introduction to Haskell, Version 98,
See www.haskell.org/tutorial.
[29] Lea D., Concurrent Programming in Java: Design Principles and Patterns, Pren-
tice Hall, 1999.
[30] Lienhard M., Schmitt A. and Stefani J.-B., Oz/K: A Kernel Language for
Component-Based Open Programming. In Sixth International Conference on Gen-
erative Programming and Component Engineering (GPCE’07), Oct. 2007.
45
Peter Van Roy
[31] Miller M. S., “Robust Composition: Towards a Unified Approach to Access Control
and Concurrency Control,” Ph.D. dissertation, Johns Hopkins University, Balti-
more, Maryland, May 2006.
[32] Miller M S., Stiegler M., Close T., Frantz B., Yee K-P., Morningstar C., Shapiro J.,
Hardy N., Tribble E .D., Barnes D., Bornstien D., Wilcox-O’Hearn B., Stanley T.,
Reid K. and Bacon D., E: Open Source Distributed Capabilities, 2001.
See www.erights.org.
[33] Moore G. E., Moore’s Law, Wikipedia, the free encyclopedia, 2009.
[34] Mozart Consortium, Mozart Programming System, Version 1.4.0 released July 2008.
See www.mozart-oz.org.
[35] Olarte C., Rueda C. and Valencia F. D., “Concurrent Constraint Calculi: a Declar-
ative Paradigm for Modeling Music Systems”. New Computational Paradigms for
Computer Music, Assayag G. and Gerzso A. (eds.), IRCAM/Delatour France, 2009.
[36] Orlarey Y., Fober D. and Letz S., “Syntactical and Semantical Aspects of Faust”.
Soft Comput. 8(9), Nov. 2004, pp. 623-632.
[37] Orlarey Y., Fober D. and Letz S., “FAUST: an Efficient Functional Approach to
DSP Programming”. New Computational Paradigms for Computer Music, As-
sayag G. and Gerzso A. (eds.), IRCAM/Delatour France, 2009.
[38] Programming Systems Lab, Saarland University, Alice ML Version 1.4.
See www.ps.uni-sb.de/alice.
[39] Puckette M., et al., Max/MSP Version 5.0.6, 2009. See www.cycling74.com.
[40] Quesada L., Solving Constraint Graph Problems using Reachability Constraints
based on Transitive Closure and Dominators, Ph.D. dissertation, Universit´e
catholique de Louvain, Louvain-la-Neuve, Belgium, Nov. 2006.
[41] Rueda C., Alvarez G., Quesada L., Tamura G., Valencia F., D´ıaz J. F. and As-
sayag G., “Integrating Constraints and Concurrent Objects in Musical Applications:
A Calculus and its Visual Language”, Constraints Journal, Kluwer Academic Pub-
lishers, number 6, pp. 21-52, 2001.
[42] Saraswat, V. A. Concurrent Constraint Programming, MIT Press, Cambridge, MA,
1993.
[43] Schulte C., Lagerkvist M. and Tack G., Gecode: Generic Constraint Development
Environment, 2006. See www.gecode.org.
[44] SELFMAN project. Self Management for Large-Scale Distributed Systems based on
Structured Overlay Networks and Components, European Commission 6th Frame-
work Programme, 2006-2009. See www.ist-selfman.org.
[45] Smolka G., Schulte C. and Van Roy P., PERDIO–Persistent and Distributed Pro-
gramming in Oz, BMBF project proposal, DFKI, Saarbr
¨
ucken, Germany. Feb.
1995.
46
Programming Paradigms for Dummies
[46] Spiessens A., Patterns of Safe Collaboration, Ph.D. dissertation, Universit´e
catholique de Louvain, Louvain-la-Neuve, Belgium, Feb. 2007.
[47] Tack G., Constraint Propagation: Models, Techniques, Implementation, Ph.D.
dissertation, Saarland University, Saarbr
¨
ucken, Germany, Jan. 2009.
[48] Van Hentenryck P., “A Gentle Introduction to NUMERICA”, Artif. Intell. 103
(1-2), Aug. 1998, pp. 209-235.
[49] Van Roy P., “Self Management and the Future of Software Design”, Third Interna-
tional Workshop on Formal Aspects of Component Software (FACS 2006), ENTCS
volume 182, June 2007, pages 201-217.
[50] Van Roy P. and Seif H., Concepts, Techniques, and Models of Computer Program-
ming, MIT Press, Cambridge, MA, 2004. See ctm.info.ucl.ac.be.
[51] Van Roy P. and Seif H., Programmation: Concepts, Techniques et Mod`eles (in
French), Dunod
´
Editeur, 2007. See ctm.info.ucl.ac.be/fr.
[52] Van Roy P., “Convergence in Language Design: A Case of Lightning Striking Four
Times in the Same Place”. In 8th International Symposium on Functional and Logic
Programming (FLOPS 2006), Fuji Sosono, Japan, Springer LNCS 3945, April 2006,
pp. 2-12.
[53] Van Roy P.,“The Challenges and Opportunities of Multiple Processors: Why Multi-
Core Processors are Easy and Internet is Hard”, Position statement, in ICMC 2008,
Belfast, Northern Ireland, Aug. 2008.
[54] Van Roy P., “ Overcoming Software Fragility with Interacting Feedback Loops and
Reversible Phase Transitions”, In BCS Symposium on Visions of Computer Science,
London, UK, Sept. 2008.
[55] Van Roy P., The Principal Programming Paradigms, Poster version 1.08. See
www.info.ucl.ac.be/~pvr/paradigms.html, 2008.
[56] Weinberg G., An Introduction to General Systems Thinking, Dorset House Pub-
lishing Co., 1977 (silver anniversary edition 2001).
[57] Wiger U., “Four-Fold Increase in Productivity and Quality–Industrial-Strength
Runctional Programming in Telecom-Class Products”, In Proceedings of the 2001
Workshop on Formal Design of Safety Critical Embedded Systems, 2001.
[58] Zampelli S., A Constraint Programming Approach to Subgraph Isomorphism, Ph.D.
dissertation, Universit´e catholique de Louvain, Louvain-la-Neuve, Belgium, June
2008.
47