Generating Programs with Grammars and ASTGen

Transcription

1 Generating Programs with Grammars and ASTGen 598DM Guest Lecture Brett Daniel Tuesday, March 11,

2 Overview Grammar AST Nodes language for composed of Program Text parsed by Parser produces Parse Tree 2

3 Overview ASTGen generates Grammar (roughly) corresponds to AST Nodes generates language for composed of Program Text parsed by Parser produces Parse Tree acts on Refactoring Engine 3

4 Grammar-Based Testing Produce many programs from grammar Used to test parser/compiler Grammar (roughly) corresponds to AST Nodes generates language for composed of Program Text parsed by Parser produces Parse Tree 4

5 Example: Fields int f; Field Declaration void m(int i) { f = i * f; Field References 5

6 Field Declaration Grammar FieldDeclaration ::= Modifier Type Identifier ; Modifer ::= public protected private ε Type ::= BasicType Identifier BasicType ::= int boolean Identifier ::= [a-za-z_]+ 6

7 Terminal Symbol Coverage The set of test inputs contains each terminal symbol FieldDeclaration ::= Modifier Type Identifier ; Modifer ::= public protected private ε Type ::= BasicType Identifier BasicType ::= int boolean Identifier ::= [a-za-z_]+ 7

8 Terminal Symbol Coverage The set of test inputs contains each terminal symbol FieldDeclaration ::= Modifier Type Identifier ; Modifer ::= public protected private ε Type ::= BasicType Identifier BasicType ::= int boolean Identifier ::= [a-za-z_]+ 8

9 Production Coverage The set of test inputs contains each production FieldDeclaration ::= Modifier Type Identifier ; Modifer ::= public protected private ε Type ::= BasicType Identifier BasicType ::= int boolean Identifier ::= [a-za-z_]+ 9

10 Derivation Coverage The set of test inputs contains every possible string Modifier Type Identifier public protected private ε int boolean + <id1>... <idn> <id1>... <idn> 10

11 Grammars and ASTs Grammar roughly corresponds to AST class structure Grammar (roughly) corresponds to AST Nodes language for composed of Program Text parsed by Parser produces Parse Tree 11

12 Grammars and ASTs Grammar (roughly) corresponds to AST Nodes FieldDeclaration ::= Modifier Type Identifier ; Modifer ::= public protected private ε Type ::= BasicType Identifier? BasicType ::= int boolean Identifier ::= [a-za-z_]+ 12

13 Grammars and ASTs Grammar (roughly) corresponds to AST Nodes FieldDeclaration ::= Modifier Type Identifier ; Modifer ::= public protected private ε Type ::= BasicType Identifier BasicType ::= int boolean Identifier ::= [a-za-z_]+ 13

14 Abstract Grammars Grammar used to generate AST class and composition structure FieldDeclaration ::= Modifier Type Identifier; Modifer; abstract Type; BasicType : Type; SimpleType : Type ::= Identifier; Identifier ::= <Name:String>; 14

15 What About Field References? Difficult to define grammar for all types of field references int f; void m(int i) { f = i * f; boolean f; class B { new A().f = true; int f; class B extends A { super.f = 0; int f; class B { int i = f; 15

16 What About Field References? Difficult to define grammar for all types of field references Want both read and write references Expressions depend on field name Productions spread across language grammar Multiple classes, inheritance relationships, containment, etc. Alternative: Extended Grammars 16

17 Generating ASTs ASTGen generators produce many AST Nodes Combining generators produces many parse trees ASTGen generates Grammar (roughly) corresponds to AST Nodes language for composed of Program Text parsed by Parser produces Parse Tree 17

18 Generators Normal Java Classes ModifierGenerator modgen = new ModifierGenerator( Modifier.PUBLIC, Modifier.PRIVATE); Produce values when prompted Modifier = modgen.next(); 18

19 Generator Characteristics Iterative Act like collection of AST fragments Internally, produce fragments lazily Composable Tester creates complex generators by reusing simpler parts Generators contain code to combine AST fragments Bounded-Exhaustive Achieves derivation coverage within bounds

20 Field Declaration Generator Iterates children and builds AST node Instances supplied by caller 20

21 Field Declaration Generator // define child generators ModifierGenerator modgen = new ModifierGenerator(Modifier.PUBLIC, Modifier.PRIVATE); IGenerator<Type> typegen = new Chain<Type>(BasicType.INT, new SimpleType("Object")); IdentifierGenerator idgen = new IdentifierGenerator("f", "g"); // compose generator FieldDeclarationGenerator fielddeclgen = new FieldDeclarationGenerator(); fielddeclgen.setmodifiergenerator(modgen); fielddeclgen.settypegenerator(typegen); fielddeclgen.setidentifiergenerator(idgen); // loop over values for(fielddeclaration fielddecl : fielddeclgen) { System.out.println(fieldDecl); 21

22 Complex Generator Double-class field reference generator Produces pairs of classes related by containment and inheritance. One class declares a field, and the other references the field in some way. int f; class B extends A { super.f = 0; boolean f; class B { new A().f = true; int f; class B { int i = f;... 22

23 Complex Generator Produces pairs of classes related by containment and inheritance. One class declares a field, and the other references the field in some way. int f; class B extends A { super.f = 0; boolean f; class B { new A().f = true; int f; class B { int i = f;... 23

24 Complex Generator Produces pairs of classes related by containment and inheritance. One class declares a field, and the other references the field in some way. int f; class B extends A { super.f = 0; boolean f; class B { new A().f = true; int f; class B { int i = f;... 24

25 Complex Generator Produces pairs of classes related by containment and inheritance. One class declares a field, and the other references the field in some way. int f; class B extends A { super.f = 0; boolean f; class B { new A().f = true; int f; class B { int i = f;... 25

26 Complex Generator Produces pairs of classes related by containment and inheritance. One class declares a field, and the other references the field in some way. int f; class B extends A { super.f = 0; boolean f; class B { new A().f = true; int f; class B { int i = f;... 26

27 Composing Generators Containment Generator class B { class B { class B {... Inheritance Generator class B { class B extends A {... Field Declaration Generator Field Reference Generator int f; boolean f; public char f;... f this.f this.a.f new A().f super.f... 27

28 Invalid Combinations Composition may be invalid class B { + class A extends B { class B { = class A extends B { class B { Three solutions Tester writes filter that verifies values Dependent generators super.f depends on int f; Delegate to compiler 28

29 Testing Refactoring Engines ASTGen generates Grammar (roughly) corresponds to AST Nodes generates language for composed of Program Text parsed by Parser produces Parse Tree acts on Refactoring Engine 29

30 Example: Encapsulate Field Replaces all field reads and writes with accesses through getter and setter methods int f; void m(int i) { f = i * f; Encapsulate Field private int f; 5 2 void m(int i) { setf(i * getf()); 4 3 void setf(int f) { this.f = f; 1 int getf() { return f; 30

31 Testing Encapsulate Field String fieldname = "f"; IGenerator<Program> testgen = new DoubleClassFieldReferenceGenerator(fieldName); for (Program in : testgen) { Refactoring r = new EncapsulateFieldRefactoring(); r.settargetfield(fieldname); Program out = r.performrefactoring(in); checkoracles(out);

32 Oracles Check that the program was refactored correctly Challenges Don't know expected output Semantic equivalence is undecidable Need to verify that correct structural changes were made 32

33 Oracles 33

34 Oracles DoesCrash: Engine throws exception DoesNotCompile: Refactored program does not compile WarningStatus: Engine can(not) perform refactoring Inverse: A refactoring is not undone by its inverse Custom Structural: Check for desired structural changes Differential: Perform refactoring in both Eclipse and NetBeans and compare result 34

35 Case Study Tested Eclipse and NetBeans Eight refactorings Target field, method, or class Wrote about 50 generators Reported 47 new bugs Compared effectiveness of oracles 35

36 Results 47 new bugs reported 21 in Eclipse: 20 confirmed by developers 26 in NetBeans: 17 confirmed, 3 fixed, 5 duplicates, 1 won't fix Found others, but did not report duplicate or fixed 36

37 Recap ASTGen generates Grammar (roughly) corresponds to AST Nodes generates language for composed of Program Text parsed by Parser produces Parse Tree acts on Refactoring Engine 37