Paradigms following: 1. Model of computation 2. Concepts - - PowerPoint PPT Presentation

28

Paradigms

Programming paradigm: – An abstract concept that covers the following:

• 1. Model of computation
• 2. Concepts (Primitives of the

programming language)

• 3. Tools (data structures and

algorithms)

• 4. The "human mindset" (what is

important, perspective to programming)

29

Paradigms (two ways

• f

classifying)

• Imperative (procedural)
• Object oriented
• Functional
• Logical
• Generative
• Script
• Model of programming

– sequential – parallel

Division 1: Division 2:

• Imperative

– procedural – object oriented – distributed (parallel)

• Declarative

– functiona – logical – data base languages

30

Examples

• f

paradigms

• Imperative/procedural paradigm
• destructive assignment, explicit control

structures, e.g. C, Ada

• Object oriented paradigm (object languages)
• objects, creating objects, message passing,

inheritance, e.g. Java, (C++)

• Functional paradigm (functional languages)
• immutable variables, no side-effects, e.g.

Haskell, Scala, Clojure

• Logic paradigm
• Facts & rules, automatic deduction of

results, e.g. Prolog, Curry

• Parallel paradigm
• processes, synchronisation, message

passing, e.g. Go, Erlang

31

Why paradigm shift now?

32

Why paradigm shift now?

• Real CPUs very complex
• Performance difficult to predict
• Concurrency/parallelism complicates

things

• Programming more abstract,

detached from hardware

• Compilers need more freedom to
• ptimize code
• → Programmer cannot be in 100%

control

33

Declarative Programming

• Functional and Logic programming
• Problems in imperative programming:

– Implementing algorithms

• How to solve a problem
• Tied to a specific computational model

(abstract view of the real computer)

• Forces a specific way of thinking
• Limiting compiler's/interpreter's

freedom (to optomize) – Problem definition needs its own formalism Programmer: what needs to be done Language: how it is done

34

Problems with imperative paradigm

• Mutual exclusion
• Aliasing
• Cache/data coherence
• Instruction reordering
• (Distributed programs and time
• rdering)

35

Functional Programming

• (Based on theory of functions in

mathematics)

• Principles

– expression (function call) value depends

• nly on the values of subexpressions

(arguments) – No side effects – No assignment, immutable data – Functions as "first-class citizens", as data

• Many functional programming languages

are ”impure” (contain assignments etc.) – Functional part is however dominating

36

Imperative vs. declarative

• Declarative "tells" something,

imperative "orders" something

• Declaration:

int f(int i) { return 2*i; }

• A sequence of statements

(commands):

++i; i = f(i); cout << i;

• Is this a declaration or a statement?:

int j = f(i);

37

Benefits of functional programming

• Gives more freedom to the compiler

(optimization)

• Makes data flow in program clearer

(only parameters & return value)

• Make parallelization easier (immutable

data & clear data flow -> easier to add concurrency)

• Forces programmer to think more

exactly(?)

• Encourages programmers to structure

their code better(?)

38

Concurrency

• No side effects, no mutable data
• → sharing data between threads ok
• More freedom in execution order
• → automatic parallelization easier
• Synchronization and exchanging

information between threads more challenging

• Transactional memory one promising

route

39

Problems with functional programming

• Initially "strange" for those used to

imperative programming

• Performance sometimes difficult to

predict (exact execution steps not clear, especially with laziness)

• Interaction with outside world not as

simple as in imperative programming

40

Other claims about functional programming

• Claims to produce less errors,

because forces programmers to think more

• Claims to encourage programmers to

structure their code better