Lecture content. Course goals. Course Introduction. TDDA69 Data and Program Structure Introduction

Transcription

1 Lecture content TDDA69 Data and Program Structure Introduction Cyrille Berger Course Introduction Introduction to the different Programming Paradigm The different programming paradigms Why different paradigms? Introduction to Functional Programming Expressions Functions Control Recursion Advantages and inconvenients of functional programming 2 / 73 Course goals Course Introduction Describe aspects of evaluation and execution in different language models Explain and demonstrate how design choices affect the expressiveness and efficacy of a programming language Analyze and value programming languages based on their evaluation and compilation strategies Implement programming languages in the form of an interpreter and a compiler 4

2 Why do you need to know how program are interpreted? It will help you understand why programming language works a certain way and what are the limits New programming languages and interpreters are constantly being developed Existing interpreters are constantly being developed to improve performance, security, add new features... Programming languages General purposes: C, C++, Java, Python... Special purposes: Prolog, Matlab, R, Agent0... Scripting: JavaScript, VBA... Historical: Fortran, Lisp Evolution of programming languages How is a program interpreted? Source code Parser Parser Abstract Syntax Tree Tree visitor Generator Source code... Bytecode Virtual Machine Assembler Assembly Operating System CPU 7 8

3 Facebook's work on the PHP intepreter Why did Facebook need to develop their own PHP interpreter? Facebook started with PHP in 2004 Back at the time, PHP was the gold standard for website programming and prototyping But this is causing problems and for practical reasons they cannot change programming language start 9 10 What did Facebook do? The standard PHP interpreter is using a virtual machine (Zend) They developed a tool to convert PHP to C++ Then they developed a new interpreter that do Just-In-Time (JIT) compilation, called HHVM They introduced Hack, a variant of PHP with a typing system And other examples... Google with Java, Dalvik, Python with CPython vs Qt's JavaScript, switching from AST Interpretation to JIT and to a mix of JIT and AST Interpretation

4 List of lectures 1Introduction and Functional Programming 2Imperative Programming and Data Structures 3Parsing 4Evaluation 5Object Oriented Programming and types system 6Macros and 7Virtual Machines and Bytecode 8Garbage Collection and Native Code 9Concurrent Computing 10Declarative Programming 11Logic 12Summary Book(s) Structure and Interpretation of Computer Programs in Python by Hal Abelson, Jerry Sussman, Julie Sussman and John Denero Structure and Interpretation of Computer Programs by Hal Abelson, Jerry Sussman and Julie Sussman List of labs 1Functional Programming 2Supporting class for an interpreter 3ECMAScript Interpreter 4Macros 5Bytecode 6Garbage collector 7SQpy Division of time 24h lectures (in 12 sessions) 40h labs (in 20 sessions) 8h tutorials (in 4 88h homework 15 16

5 Last year evaluation and improvments Too much boilerplatting, Labs and lectures are not well connected Two new labs Quizzes are nice but teacher wait for too long Introduce a timeout Teacher is shy, slides are mostly used by the teacher to know what to say Introduce more interractivity The exam is worth more than two credits Introduction to the different Programming Paradigm 17 The different programming paradigms 20

6 Programming Paradigm (2/2) Declarative declarative Symbolic functional Logic Imperative Object-Oriented Expresses logic of computation without control flow: What should be computed and not how it should be computed. Examples: XML/HTML, antlr4/yacc/ regular expressions, make/ants, SQL, Declarative - Examples <b>hello world!</b> SELECT name FROM student WHERE course eq 'TDDA69' grammar Hello; r : 'hello' ID; ID : [a-z]+ ; WS : [' '\t\r\n]+ -> skip ; Functional Computation are treated as mathematical function without changing any internal state Examples: Lisp, Scheme, 23 24

7 Functional - Examples (print "Hello World") (take 25 (squares-of -> ( ) Imperative Express how computation are executed Describes computation in term of statements that change the internal state Examples: C/C++, Pascal, Java, Python, JavaScript Imperative - Examples Object-Oriented for(var i = 1; i < 26; ++i) var sq = i*i; console.log(sq) #include <stdio.h> int main() char ch; printf("enter a character\n"); scanf("%c", &ch); if (ch == 'a' ch == 'A' ch == 'e' ch == 'E' ch == 'i' ch == 'I' ch =='o' ch=='o' ch == 'u' ch == 'U') printf("%c is a vowel.\n", ch); else printf("%c is not a vowel.\n", ch); return 0; Based on the concept of objects, which are data structures containing fields and methods Programs are designed by making objects interact with each others Examples: C++, Java, C#, Python, Ruby, JavaScript

8 Object-Oriented - Programming Others paradigm languages #include <iostream> class Character : public Symbol public: Character(char _c) : m_c(_c) bool isvowel() const return ch == 'a' ch == 'A' ch == 'e' ch == 'E' ch == 'i' ch == 'I' ch =='o' ch=='o' ch == 'u' ch == 'U'; private: char m_c; ; int main() char c; std::cout << "Enter a character:\n"; std::cin >> c; Character ch(c); if(ch.isvowel()) std::cout << c << " is a vowel.\n"; else std::cout << c << " is not a vowel.\n"; Logic Based on Formal logic: expressing facts and rules Symbolic A program can manipulate its own formulas and components as if they are data Example: prolog Logic programming Symbolic programming likes(mary,food). likes(mary,wine). likes(john,wine). likes(john,mary).?- likes(mary,food). yes.?- likes(john,wine). yes.?- likes(john,food). no. 31 d( X, X, 1 ):-!. /* d(x) w.r.t. X is 1 */ d( C, X, 0 ):- atomic(c). /* If C is a constant */ /* then d(c)/dx is 0 */ d( U+V, X, R ):- /* d(u+v)/dx = A+B where */ d( U, X, A ), /* A = d(u)/dx and */ d( V, X, B ), R = A + B.... d( sin(w), X, Z*cos(W) ):- /* d(sin(w))/dx = Z*cos(W) */ d( W, X, Z). /* where Z = d(w)/dx */ d( exp(w), X, Z*exp(W) ):- /* d(exp(w))/dx = Z*exp(W) */ d( W, X, Z). /* where Z = d(w)/dx */...?- d(cos(2*x+1), X, what = 2*sin(2*X 32

9 Can you do everything in imperative programming? Why different paradigms? start 34 Is there a paradigm to rule them all? In theory you can program everything in C/C++ and imperative programming, or functional programming... But is that And is that Functional vs Imperative Double all the numbers in an array var numbers = Imperative: var doubled = [] for(var i = 0; i < numbers.length; i++) var newnumber = numbers[i] * 2 doubled.push(newnumber) Functional: var doubled = numbers.map(function(n) return n * 2 ) 35 36

10 Declarative vs Imperative Functional vs Imperative Select all the dogs that belongs to a specific Declarative: SELECT * from dogs INNER JOIN owners WHERE dogs.owner_id = owners.id Imperative: var dogswithowners = [] var dog, owner for(var dog in dogs) for(var owner in owners) if (owner && dog.owner_id == owner.id) dogswithowners.push( dog: dog, owner: owner ) Imperative language (C/C++, Java...) Basic constructs are imperative statements Change existing values, states x = x + y = while(x>0)... Functional Basic constructs are declarative Declare new values function f(x) return x + 1; Computations are primarily done by evaluating expressions Pure if all constructs are declarative Introduction to Functional Programming Expressions

11 Expressions (1/2) Primitive expressions to capture the simplest elements we want to describe numbers, arithmetic expressions... 2 plus 'hello' number operator string Call expressions: max ( 2, 3 ) operator operand operand Expressions (2/2) Means of combining simple elements into coumpound ones max(min(pow(3, 5), -4), min(1, -2) * 2) Names Means of abstracting elements by naming and manipulating them 2 plus 'hello' name a plus b name name name max ( 2, 3 ) name Assignment Binds names to a := 2 Now a has the value 2 a plus 2 evaluates to

12 What is a function? Functions Assignment is a simple means of abstraction: binds names to values Function definition is a more powerful means of abstraction: binds names to expressions 46 Function definition A function definition contains: A signature which defines how many arguments a function takes A body which defines the computation performed when the function is called function <name> ( parameters list ) return <return expression> ; Pure and non-pure function Pure functions: just return values: Math.abs(-2) -> Math.pow(2, 100) -> Non-pure functions: have side effects: print(-2) -> But print '-2' in the A side effect is not a value, it is anything that happens as a consequence of calling a 47 48

13 What is a lambda? start Closure and lambda A lambda is a function with no name var doubled = numbers.map(function(n) return n * 2 ) A closure is a function with captured variables This seem simple, but this is actually rather powerful! function create_function_multiplication(number) return function(x) return x * number; var doubled = numbers.map(create_function_multiplication(2)) Closure in Python Python has limited support for lamdas: single statement: numbers.map(lambda v: v * 2) But support nested functions (and closure): def create_function_multiplication(number): def function_multiplication(x): return x * number return function_multiplication doubled = numbers.map(create_function_multiplication(2)) Control 51

14 Control flow (1/2) In imperative programming, a control flow statement execution result in a choice between two paths Exemple: if, while... In functional programming, it is done using a special function. For instance, in Lisp: (COND (condition1 result1 ) (condition2 result2 )... (T resultn ) ) Control flow (2/2) The functional way in imperative languages: In JavaScript, C++, Java, Ruby...: condition? result1 : result2; In Python: condition if result1 else result What about loops? Loop constructs is imperative How would you implement the equivalent of a loop in functional? function factorial(n) var r = 1; for(var i = 2; i <= n; ++i) r *= i; return r; Recursion 55

15 What is recursion? A function is called recursive if the body of that function calls itself, either directly or indirectly. Factorial: the classical example (1/2) Factorial in factorial :: Integral -> Integral factorial 0 = 1 factorial n = n * factorial (n-1) Factorial in Common LISP: (define (factorial n) (cond ((= n 0) 1) (t (* n (factorial (- n 1)))))) Factorial: the classical example (1/2) Recursion vs loops With a loop: function factorial(n) var r = 1; for(var i = 2; i <= n; ++i) r *= i; return r; With a recursive function factorial(n) return (n===0)? 1 : n * factorial(n-1) 59 while(expression) do_something(); function loop_something(args...) if(expression) return; else do_something(); loop_something(args...); (define (loop_something args...) (cond (expression) value) (t (do_something) (loop_something args...)) 60

16 When to use recursion rather than iteration? (1/4) When to use recursion rather than iteration? (2/4) def factorial(n): if(n == 0): return 1 else: return n * factorial(n-1) Lets try: factorial(10) factorial(1000) In most programming language, the number of function call is limited by the size of the stack sys.getrecursionlimit() sys.setrecursionlimit(1003) factorial(1000) Tail-call optimisation When to use recursion rather than iteration? (3/4) When to use recursion rather than iteration? (4/4) Calling a function is usually more expensive than a loop def factorial2(n): r = 1 for i in range(1, n + 1): r = r * i return r timeit.timeit("factorial(30)", "from main import factorial") timeit.timeit("factorial2(30)", "from main import factorial2") It is a matter of Recursion is a bit more general than loops When walking through a tree 63 64

17 No side-effects pure functional Advantages and inconvenients of functional programming In pure functional, calling a function only return a value The implication is that calling a function with the same arguments will always return the same value Is the withdraw function pure (define balance 100) (define (withdraw amount) (if (>= balance amount) (begin (set! balance (- balance amount)) balance) "Insufficient funds")) enter code start 66 Verification and proving To prove a program correct, we must consider everything a program depends on In pure functional programs, dependence on any data structure is explicit Proving properties in functional programming (define (power b n) (cond ((= n 0) 1) (t (* (power b (- n 1)))))) Claim: for any integer n 0 and any number b, (power b n) = bⁿ Proof: 1) Verify the base case: (power b 2) Assume that (power b (- n 1)) is correct 3) Verify that (power b n) is correct assuming that (power b (- n 1)) is correct 67 68

18 Proving properties in imperative programming function power(b, n) int result = 1; for(int i = 0; i < n; + +i) result *= b; return result; Devise a loop invariant: (n i) (result = bⁱ) Prove that it is true for the first loop iteration Prove that each loop iteration preserves it Assume that (n i) (result = bⁱ) Prove that (n j) (result = bʲ) with j = i + 69 Concurency Concurency is one of the current hot topic in programming The main challenge is data-race Imperative programs are very sensible to data-race because of states There is no data-race in pure functional languages all data is immutable all functions are pure, without side-effects 70 Summary on the upside of functional programming The downside of functional programming The main advantage is no sideeffects Verification and proving Concurency Productivity? Ericsson claims an increase in productivity between 9 and 25 times when using their home-grown version of Erlang In practice, there is very limited need for proving a program Mostly in critical applications: rocket control, hospital... And how do you prove hardware? Performance issues (remember function call are expensive) Very limited support Most programming tasks require states 71 72

19 My Key message about programming paradigms Be pragmatic, there is no one answer! 73