Software Construction

Transcription

1 Lecture 2: Decomposition and Java Methods Software Construction in Java for HSE Moscow Tom Verhoeff Eindhoven University of Technology Department of Mathematics & Computer Science Software Engineering & Technology Group wstomv/edu/sc-hse c 2014, T. TUE.NL 1/48 Software Construction: Lecture 2

2 Overview Feedback on Series 1 Manage complexity by Divide and Conquer Java methods Procedural abstraction Functional decomposition Assignments for Series 2 c 2014, T. TUE.NL 2/48 Software Construction: Lecture 2

3 Registration in peach students joined the course in peach 3 (expected 103) Please, provide all information about your name This will be used on your certificate! Example: Wernher Magnus Maximilian von Braun First name = Given name: Wernher Initials: W.M.M. Connectives (mostly for Dutch/German names): von Family Name: Braun Fix this, if it is not OK Bad name No grade c 2014, T. TUE.NL 3/48 Software Construction: Lecture 2

4 Automatic Feedback by peach 3 Mostly disabled before the deadline Enabled after the deadline Check peach 3 feedback; read the explanation: Proper files, cut line, etc. Diff with given files Checkstyle audit Javadoc (in the future) Compilation Test runs c 2014, T. TUE.NL 4/48 Software Construction: Lecture 2

5 Assignments Are to Be Made Individually All assignments must be made individually Do not copy each other s work peach 3 compares all submitted work peach 3 discovered some (nearly) identical copies c 2014, T. TUE.NL 5/48 Software Construction: Lecture 2

6 Statistics for Series 1 Candy 99 submissions 14 open test run 0 secret test runs 83 with 14 points 6 with 13 points 2 with 12 points 6 with 11 points 1 with 3 points 0 rejected DiceGame 97 submissions 6 open test run 0 secret test runs 86 with 6 points 11 with 4 points 0 rejected All test runs are now open c 2014, T. TUE.NL 6/48 Software Construction: Lecture 2

7 General Remarks and Observations Deadlines are hard and are enforced by peach 3 Be disciplined: Set yourself an earlier deadline Place your code at indicated comments only, and nowhere else No package statement allowed; preserve the --8<-- Submit your *.java source file, not *.class, *.rar Use the given class name and file name (= class name +.java) Unless otherwise stated, you may assume that input is valid There was no need to check input validity Suppress your built-in testing code : delete or comment out c 2014, T. TUE.NL 7/48 Software Construction: Lecture 2

8 Coding Standard Not all submissions adhered to the coding standard Some were very sloppy Why adhere to a coding standard? You will make fewer mistakes: sloppy coding = sloppy thinking If you make mistakes, you find them more easily and quickly If you cannot find them yourself, others can help you better Business perspective: When sued, you have a better court case Review the Coding Standard for 2IP15 c 2014, T. TUE.NL 8/48 Software Construction: Lecture 2

9 Coding Idiom Condition b == false is better written as! b (say: not b) Use else to make intention explicit: if (a < b) { S1; } if (a == b) { S2; } if (a < b) { S1; } else if (a == b) { S2; } else { // a > b // do nothing (on purpose) } What if, later, S1 is adapted to modify variables a, b? Be prepared for future evolution of software! c 2014, T. TUE.NL 9/48 Software Construction: Lecture 2

10 Candy Avoid floating-point arithmetic; instead, use % (modulo) operator Avoid division by zero The case divide(0, 0): Consider the equation c == k * q with unknown q Consider two cases: k == 0 and k!= 0 If k == 0, then the equation reduces to c == 0 What does that mean? There are two subcases! Also k!= 0 has two subcases c 2014, T. TUE.NL 10/48 Software Construction: Lecture 2

11 DiceGame Use java.util.random random 0 <= random.nextint(k) < k How to roll two dice? How to roll a dodecahedron? How to determine the round outcome? How to administrate winning counts? Sum of counts = number of rounds (peach 3 tests for this) c 2014, T. TUE.NL 11/48 Software Construction: Lecture 2

12 How to create something big and complex? Engineering, technology Science How to be in control? How to improve probability of success? c 2014, T. TUE.NL 12/48 Software Construction: Lecture 2

13 How to create something big and complex? There are two ways of constructing a software design: one way is to make it so simple that there are obviously no deficiencies, and the other is to make it so complicated that there are no obvious deficiencies. C. A. R. Hoare (winner of Turing Award 1980) c 2014, T. TUE.NL 13/48 Software Construction: Lecture 2

14 The Essence of Engineering [I]t is the essence of modern engineering not only to be able to check one s own work, but also to have one s work checked and to be able to check the work of others. In order for this to be done, the work must follow certain conventions, conform to certain standards, and be an understandable piece of technical communication. H. Petroski. To Engineer is Human: The Role of Failure in Successful Design. Vintage Books, c 2014, T. TUE.NL 14/48 Software Construction: Lecture 2

15 Procedural Abstraction See Chapter 4 (Subroutines) of Eck, especially 4.1 through 4.4 c 2014, T. TUE.NL 15/48 Software Construction: Lecture 2

16 Monolithic Programs The imperative core of Java, consisting of types, values, variables, expressions, assignment, selection, repetition allows you to express every computation! However, in that case, programs are monolithic, without explicit structure, unmanageable c 2014, T. TUE.NL 16/48 Software Construction: Lecture 2

17 Manage Complexity Technique: Divide and Conquer General problem solving technique: 1. Split problem into subproblems ( Decomposition ) 2. Solve subproblems independently ( Recursion possible) 3. Combine subproblem solutions into total solution ( Integration ) c 2014, T. TUE.NL 17/48 Software Construction: Lecture 2

18 Problems and Solutions in Java Wrapped-up solutions in Java can take the form of methods classes packages... c 2014, T. TUE.NL 18/48 Software Construction: Lecture 2

19 Divide and Conquer with Java Methods Specify a subproblem Divide Syntax: method name, (formal) parameters, return type Semantics: Contract with pre/post assumptions/obligations Create (design and implement) subproblem solution Conquer Method body Use subproblem solutions and Rule Method invocations with arguments (actual parameters) Hierarchical (recursive) subdivision: Create subproblem solution by using subsubproblem solutions c 2014, T. TUE.NL 19/48 Software Construction: Lecture 2

20 Java Methods modifiers return type name ( parameters ) { statements } procedure : return type void, used as statement (a.k.a. command) function : return type not void, used as expression (a.k.a. query, inspector) parameters : name, type, final or not, modified or not order of parameters variables : local versus global aliasing : different names referring to the same entity c 2014, T. TUE.NL 20/48 Software Construction: Lecture 2

21 Manage Complexity Why: Intellectual limits of human beings, concerning productivity, memory, communication, accuracy,... Some related concepts: Divide & Conquer Separation of Concerns Decomposition Modularization Encapsulation Abstraction c 2014, T. TUE.NL 21/48 Software Construction: Lecture 2

22 Abstraction To abstract from something Ignore/suppress distinctions that are considered irrelevant Separate relevant from irrelevant details Example: Identify fractions a/b and c/d (b 0 d) when ad = bc Set Theory: equivalence relation, equivalence class, quotient set Abstract from how a subproblem is solved, through a specification Abstract from which subproblem is solved, through parameters Abstract from how a solution will be used c 2014, T. TUE.NL 22/48 Software Construction: Lecture 2

23 Abstraction Example: Rectangle Problem On input are 6 integers, representing three points in a Cartesian coordinate system, for each point first the x- and then the y-coordinate. The first two points designate a rectangle with sides parallel to the axes. The first point is the top left corner of the rectangle, the second point is the bottom right corner. The program has to decide whether the third point is inside the rectangle (including the edges) or outside. The program should output an error message if the rectangle is ill-defined, i.e., if the second point is not to the right of and below the first point. The program may assume that no other errors occur in the input. c 2014, T. TUE.NL 23/48 Software Construction: Lecture 2

24 Abstraction Example: Monolithic Solution, Variables 1 // Monolithic solution 2 3 import java.util.scanner; // for input 4 5 class Rectangle1 { 6 7 static Scanner scanner; // input source 8 static int tlx; // top left x 9 static int tly; // top left y 10 static int brx; // bottom right x 11 static int bry; // bottom right y 12 static int px; // point x 13 static int py; // point y c 2014, T. TUE.NL 24/48 Software Construction: Lecture 2

25 Abstraction Example: Monolithic Solution, Method 15 static void dorectangle() { 16 scanner = new Scanner(System.in); 17 tlx = scanner.nextint(); 18 tly = scanner.nextint(); 19 brx = scanner.nextint(); 20 bry = scanner.nextint(); 21 px = scanner.nextint(); 22 py = scanner.nextint(); if (tlx > brx tly > bry) { 25 System.out.println("error"); 26 return; 27 } 28 // tlx <= brx && tly <= bry c 2014, T. TUE.NL 25/48 Software Construction: Lecture 2

26 Abstraction Example: Monolithic Solution, Method 29 if (tlx <= px && px <= brx && tly <= py && py <= bry) { 30 System.out.println("inside"); 31 } else { 32 System.out.println("outside"); 33 } 34 } c 2014, T. TUE.NL 25/48 Software Construction: Lecture 2

27 Abstraction Example: Separate Methods, Top Level + Output 15 static void dorectangle() { 16 scanner = new Scanner(System.in); 17 readinput(); if (! isvalidrectangle()) { 20 System.out.println("error"); 21 return; 22 } 23 // input rectangle is valid 24 if (ispointinsiderectangle()) { 25 System.out.println("inside"); 26 } else { 27 System.out.println("outside"); 28 } 29 } c 2014, T. TUE.NL 26/48 Software Construction: Lecture 2

28 Abstraction Example: Separate Methods, Input 31 // Read input into tlx,..., py via scanner 32 static void readinput() { 33 tlx = scanner.nextint(); 34 tly = scanner.nextint(); 35 brx = scanner.nextint(); 36 bry = scanner.nextint(); 37 px = scanner.nextint(); 38 py = scanner.nextint(); 39 } Note that readinput() operates on the global variables. c 2014, T. TUE.NL 27/48 Software Construction: Lecture 2

29 Abstraction Example: Separate Methods, Calculations 1 // Returns whether tlx,..., bry is a rectangle 2 static boolean isvalidrectangle() { 3 return tlx <= brx && tly <= bry; 4 } 5 6 // Returns whether px, py is inside rectangle tlx,..., bry 7 static boolean ispointinsiderectangle() { 8 return tlx <= px && px <= brx && tly <= py && py <= bry; 9 } Note that these methods also operate on the global variables. c 2014, T. TUE.NL 28/48 Software Construction: Lecture 2

30 Abstraction Example: Dependence Diagram 1 dorectangle() Scanner() readinput() isvalidrectangle() ispointinsiderectangle() System.out.println() scanner, tlx, tly, brx, bry, px, py c 2014, T. TUE.NL 29/48 Software Construction: Lecture 2

31 Dependence Diagrams Thinking in terms of dependencies is key to good design. A B means A depends on B When code of B changes, code of A may also have to change Code of A can change, without affecting code of B When reusing code of A, also code of/like B is needed When reusing code of B, code of A is not needed c 2014, T. TUE.NL 30/48 Software Construction: Lecture 2

32 Abstraction Example: Now with Parameters 14 static void dorectangle() { 15 readinput(new Scanner(System.in)); if (! isvalidrectangle(tlx, tly, brx, bry)) { 18 System.out.println("error"); 19 return; 20 } 21 // input rectangle is valid 22 if (ispointinsiderectangle(tlx, tly, brx, bry, 23 px, py)) { 24 System.out.println("inside"); 25 } else { 26 System.out.println("outside"); 27 } 28 } c 2014, T. TUE.NL 31/48 Software Construction: Lecture 2

33 Abstraction Example: Separate Methods, Input 30 // Read input into global tlx,..., py via scanner 31 static void readinput(scanner scanner) { 32 tlx = scanner.nextint(); 33 tly = scanner.nextint(); 34 brx = scanner.nextint(); 35 bry = scanner.nextint(); 36 px = scanner.nextint(); 37 py = scanner.nextint(); 38 } Note that readinput() still operates on the global variables. (In Java, this is harder to avoid.) c 2014, T. TUE.NL 32/48 Software Construction: Lecture 2

34 Abstraction Example: Separate Methods, Calculations 1 // Returns whether tlx,..., bry is a rectangle 2 static boolean isvalidrectangle(int tlx, int tly, 3 int brx, int bry) { 4 return tlx <= brx && tly <= bry; 5 } 6 7 // Returns whether x, y is inside rectangle tlx,..., bry 8 static boolean ispointinsiderectangle(int tlx, int tly, 9 int brx, int bry, 10 int x, int y) { 11 return tlx <= x && x <= brx && tly <= y && y <= bry; 12 } Note that these methods now do not involve the global variables. Instead, they operate on parameters only. c 2014, T. TUE.NL 33/48 Software Construction: Lecture 2

35 Abstraction Example: Dependence Diagram 2 dorectangle() Scanner() readinput() isvalidrectangle() ispointinsiderectangle() System.out.println() tlx, tly, brx, bry, px, py c 2014, T. TUE.NL 34/48 Software Construction: Lecture 2

36 Abstraction Example: Separate Methods, Aux Calculation 47 // Returns whether tlx,..., bry is a rectangle 48 static boolean isvalidrectangle(int tlx, int tly, 49 int brx, int bry) { 50 return isdominated(tlx, tly, brx, bry); 51 } // Returns whether x, y is inside rectangle tlx,..., bry 54 static boolean ispointinsiderectangle(int tlx, int tly, 55 int brx, int bry, 56 int x, int y) { 57 return isdominated(tlx, tly, x, y) && 58 isdominated(x, y, brx, bry); 59 } // Returns whether ax, ay is dominated by bx, by 62 static boolean isdominated(int ax, int ay, int bx, int by) { 63 return ax <= bx && ay <= by; 64 } c 2014, T. TUE.NL 35/48 Software Construction: Lecture 2

37 Abstraction Example: Dependence Diagram 3 dorectangle() Scanner() readinput() isvalidrectangle() ispointinsiderectangle() isdominated() System.out.println() tlx, tly, brx, bry, px, py c 2014, T. TUE.NL 36/48 Software Construction: Lecture 2

38 Abstraction Example (cont d) Separate input, calculations, and output Parameterize methods N.B. In this style, it is harder to parameterize readinput further, because Java has no output parameters, only return values. No Java method with parameters v, E equivalent to v := E exists. readinput() could return an array, but then the indexing would be awkward. Separate classes for rectangles and points is the Java way. Doing output could be separated from the top-level coordination, using a string result. c 2014, T. TUE.NL 37/48 Software Construction: Lecture 2

39 Guidelines for Functional Decomposition: Cohesion Maximize cohesion Cohesion = measure of relatedness of things inside a module Keep related things together; separate unrelated things c 2014, T. TUE.NL 38/48 Software Construction: Lecture 2

40 Guidelines for Functional Decomposition: Coupling Minimize coupling, also stated as minimize dependency Coupling = measure of connectedness between modules Interface of a module concerns all its external connections Consider both number and nature of dependencies: GOOD : via parameter OK : use constant, type, (variable,) method defined elsewhere BAD : implicit common knowledge (file name in string literal) BAD : global variables c 2014, T. TUE.NL 39/48 Software Construction: Lecture 2

41 Guidelines for Functional Decomposition (cont d) Each function should solve a single well-defined problem If purpose is difficult to describe, then something is wrong (Describe purpose on the right level of data abstraction!) Limit the size of the interface, i.e., number of parameters Possibly combine parameters (in a new class, representing a new concept, raising the level of abstraction) [NOT THIS WEEK] Each function should have a small implementation No more than a (few) dozen lines, with limited nesting depth If it tends to become larger, consider subdivision Nested control structures should invite further subdivision, especially nested loops c 2014, T. TUE.NL 40/48 Software Construction: Lecture 2

42 Generalization Consider possibilities for generalization : solving a more general problem than (initially) required. Generalization improves (re)usability and insight. It can also simplify solution! Generalize by means of parameters, not via global variables. Parameters abstract from the identity of the data. Code duplication should invite refactoring by parameterization. Decomposing a problem into simpler versions of itself gives rise to recursion. Weigh benefits against costs: too general makes use costly. c 2014, T. TUE.NL 41/48 Software Construction: Lecture 2

43 Generalization Example How many regions are created when space is cut by five arbitrary planes? Arbitrary means: No two planes are parallel, and no three planes have a point in common. c 2014, T. TUE.NL 42/48 Software Construction: Lecture 2

44 Generalization Example How many regions are created when space is cut by five arbitrary planes? Arbitrary means: No two planes are parallel, and no three planes have a point in common. Parameterize on number of planes N c 2014, T. TUE.NL 42/48 Software Construction: Lecture 2

45 Generalization Example How many regions are created when space is cut by five arbitrary planes? Arbitrary means: No two planes are parallel, and no three planes have a point in common. Parameterize on number of planes N Parameterize on dimension D of space to be divided : D = 1 line, D = 2 plane, D = 3 space c 2014, T. TUE.NL 42/48 Software Construction: Lecture 2

46 Generalization Example: Solution R(D, N) = # regions in D-space when cut by N arbitrary D 1-spaces c 2014, T. TUE.NL 43/48 Software Construction: Lecture 2

47 Generalization Example: Solution R(D, N) = # regions in D-space when cut by N arbitrary D 1-spaces R(D, 0) = 1 (when not divided) R(1, N) = N + 1 (a line divided by N points) R(D, N) = R(D, N 1) + R(D 1, N 1) if 2 D and 1 N c 2014, T. TUE.NL 43/48 Software Construction: Lecture 2

48 Generalization Example: Solution R(D, N) = # regions in D-space when cut by N arbitrary D 1-spaces R(D, 0) = 1 (when not divided) R(1, N) = N + 1 (a line divided by N points) R(D, N) = R(D, N 1) + R(D 1, N 1) if 2 D and 1 N R(D, N) N D c 2014, T. TUE.NL 43/48 Software Construction: Lecture 2

49 Typical Purposes of Functions Input, calculate, output Parse, convert, format De-nest (disentangle) control structures Initialize, query, update, finalize Check condition Create, delete Coordinate activities, handle events Abstract from the way data is represented c 2014, T. TUE.NL 44/48 Software Construction: Lecture 2

50 Function Names, Parameter Order Use verb-plus-noun, but avoid overly general words: processinput Do not include the object name that this refers to In class Document do not name a method printdocument Function name should reflect returned value Procedure name should reflect postcondition Short parameter names facilitate writing of specifications Parameter order: 1 input-only, 2 input-and-output, 3 output-only (status and error information last). c 2014, T. TUE.NL 45/48 Software Construction: Lecture 2

51 Assignments Series 2 Improve your Candy program Complete the decomposition for DiceGame Answer some open questions (in preparation of written exam) c 2014, T. TUE.NL 46/48 Software Construction: Lecture 2

52 Before Submitting Work for Assignments in peach 3 Code adheres to Coding Standard for 2IP15 Include author name, group ID, date, above cut line with --8<-- Check your code: read it and test it c 2014, T. TUE.NL 47/48 Software Construction: Lecture 2

53 Summary Why adhere to a coding standard Manage complexity by Divide and Conquer Java methods, with parameters Guidelines for functional decomposition c 2014, T. TUE.NL 48/48 Software Construction: Lecture 2