Version: 1.5.5 (using KDE 3.5.5 "release 19.1" , openSUSE ) Compiler: Target: x86_64-suse-linux OS: Linux (x86_64) release 2.6.16.13-4-default Hi again I generate javascript code and the result was not good: When you create a composition between class A and class B, the generator returns this: function A() { .... this.m_B = new B(); .... } function B() { .... this.m_A = new A(); .... } and it was suppose to instance only class B in class A. It also returns a lot of instance of class A itself: function A() { ... this.m_B = new B(); this.m_B = new B(); this.m_A = new A(); this.m_B = new B(); this.m_B = new B(); this.m_A = new A(); ... }
I can't exactly reproduce this bug in version 1.5.6. Here is what I get: Given a class A and a class B, I draw a composition between A and B ---- ------ | A | <>-----| B | ----- ------- The generated javascript code is File A.js: ---------------------------------- ** * class A */ A = function () { this._init (); } /** * _init sets all A attributes to their default value. Make sure to call this * method within your class constructor */ A.prototype._init = function () { /**Aggregations: */ /**Compositions: */ this.m_A = new A (); } ----------------------- File B.js: ----------------------- ** * class A */ A = function () { this._init (); } /** * _init sets all A attributes to their default value. Make sure to call this * method within your class constructor */ A.prototype._init = function () { /**Aggregations: */ /**Compositions: */ this.m_A = new A (); } -------------------------------- According to me, there are two errors: 1) There shouldn't be a composant m_A in class B but a composant m_B in class A (In this case, the C++ generated code doesn't generate any composants in any classes. The C++ generated code generates a composant m_B if only I add a attribute of class B in class A) 2) There shouldn't be a composant m_A in class A. In my example, there is only one composant m_A in class A. That's better than many but still incorrect PS : The generated code is identical in trunk
SVN commit 634263 by okellogg: writeClass (aggregations, compositions): Do not generate code when `c´ is at role B. Generate code for the role B object. BUG:135527 M +1 -0 ChangeLog M +10 -4 umbrello/codegenerators/jswriter.cpp --- branches/KDE/3.5/kdesdk/umbrello/ChangeLog #634262:634263 @@ -3,6 +3,7 @@ * Bugs fixed from http://bugs.kde.org: * %date% and %time% not being parsed (96612) * Relationships for entities do not live outside of an entity relationship diagram (125146) +* Javascript wrong Code Generation (135527) * Javascript Code Generation creates bad format methods (135540) * Crash when deleting the link between a package and a class (141602) * Ada95 Code Generation Errors for Aggregation (141644) --- branches/KDE/3.5/kdesdk/umbrello/umbrello/codegenerators/jswriter.cpp #634262:634263 @@ -155,8 +155,11 @@ js << m_endl << m_indentation << "/**Aggregations: */" << m_endl; for (UMLAssociation* a = aggregations.first(); a; a = aggregations.next()) { - QString nm(cleanName(a->getObject(Uml::A)->getName())); - if (a->getMulti(Uml::A).isEmpty()) + UMLObject *b = a->getObject(Uml::B); + if (b == c) + continue; // we need to be at the "A" side and the other guy at "B" + QString nm(cleanName(b->getName())); + if (a->getMulti(Uml::B).isEmpty()) js << m_indentation << "this.m_" << nm << " = new " << nm << " ();" << m_endl; else js << m_indentation << "this.m_" << nm.lower() << " = new Array ();" << m_endl; @@ -168,8 +171,11 @@ js << m_endl << m_indentation << "/**Compositions: */" << m_endl; for (UMLAssociation *a = compositions.first(); a; a = compositions.next()) { - QString nm(cleanName(a->getObject(Uml::A)->getName())); - if(a->getMulti(Uml::A).isEmpty()) + UMLObject *b = a->getObject(Uml::B); + if (b == c) + continue; // we need to be at the "A" side and the other guy at "B" + QString nm(cleanName(b->getName())); + if (a->getMulti(Uml::B).isEmpty()) js << m_indentation << "this.m_" << nm << " = new "<< nm << " ();" << m_endl; else js << m_indentation << "this.m_" << nm.lower() << " = new Array ();" << m_endl;