Upgrade from older OCCT versions {#occt_dev_guides__upgrade} ================================ @tableofcontents @section upgrade_intro Introduction This document provides technical details on changes made in particular versions of OCCT. It can help to upgrade user applications based on previous versions of OCCT to newer ones. @subsection upgrade_intro_precautions Precautions Back-up your code before the upgrade. We strongly recommend using version control system during the upgrade process and saving one or several commits at each step of upgrade, until the overall result is verified. This will facilitate identification and correction of possible problems that can occur at the intermediate steps of upgrade. It is advisable to document each step carefully to be able to repeat it if necessary. @subsection upgrade_intro_disclaim Disclaimer This document describes known issues that have been encountered during porting of OCCT and some applications and approaches that have helped to resolve these issues in known cases. It does not pretend to cover all possible migration issues that can appear in your application. Take this document with discretion; apply your expertise and knowledge of your application to ensure the correct result. The automatic upgrade tool is provided as is, without warranty of any kind, and we explicitly disclaim any liability for possible errors that may appear due to use of this tool. It is your responsibility to ensure that the changes you made in your code are correct. When you upgrade the code by an automatic script, make sure to carefully review the introduced changes at each step before committing them. @section upgrade_700 Upgrade to OCCT 7.0.0 @subsection upgrade_700_persist Removal of legacy persistence Legacy persistence for shapes and OCAF data based on *Storage_Schema* (toolkits *TKShapeShcema, TLStdLSchema, TKStdSchema* and *TKXCAFSchema*) has been removed in OCCT 7.0.0. The applications that used these data persistence tools need to be updated to use other persistence mechanisms. The existing data files in standard formats can be converted using OCCT 6.9.0 or a previous version, as follows. #### CSFDB files Files in CSFDB format (usually with extension .csfdb) contain OCCT shape data that can be converted to BRep format. The easiest way to do that is to use ImportExport sample provided with OCCT 6.9.0 (or earlier): - Start ImportExport sample; - Select File / New; - Select File / Import / CSFDB... and specify the file to be converted; - Drag the mouse with the right button pressed across the view to select all shapes by the rectangle; - Select File / Export / BREP... and specify the location and name for the resulting file #### OCAF and XCAF documents Files containing OCAF data saved in the old format usually have extensions .std, .sgd or .dxc (XDE documents). These files can be converted to XML or binary OCAF formats using DRAW Test Harness commands available in OCCT 6.9.0 or earlier. For that, start *DRAWEXE* and perform the following commands: * To convert *.std and *.sgd file formats to binary format *.cbf (The created document should be in *BinOcaf* format instead of *MDTV-Standard*): @code Draw[]> pload ALL Draw[]> Open [path to *.std or *.sgd file] Doc Draw[]> Format Doc BinOcaf Draw[]> SaveAs Doc [path to the new file] @endcode * To convert *.dxc file format to binary format *.xbf (The created document should be in *BinXCAF* format instead of *MDTV-XCAF*): @code Draw[]> pload ALL Draw[]> XOpen [path to *.dxc file] Doc Draw[]> Format Doc BinXCAF Draw[]> XSave Doc [path to the new file] @endcode On Windows, it is necessary to replace back slashes in the file path by direct slashes or pairs of back slashes. Use *XmlOcaf* or *XmlXCAF* instead of *BinOcaf* and *BinXCAF*, respectively, to save in XML format instead of binary one. @subsection upgrade_occt700_cdl Removal of CDL and WOK OCCT code has been completely refactored in version 7.0 to get rid of obsolete technologies used since its inception: CDL (Cas.Cade Definition Language) and WOK (Workshop Organization Kit). C++ code previously generated by WOK from CDL declarations is now included directly in OCCT sources. This modification did not change names, API, and behavior of existing OCCT classes, thus in general the code based on OCCT 6.x should compile and work fine with OCCT 7.0. However, due to redesign of basic mechanisms (CDL generic classes, Handles and RTTI) using C++ templates, some changes may be necessary in the code when porting to OCCT 7.0, as described below. WOK is not necessary anymore for building OCCT from sources, though it still can be used in a traditional way -- auxiliary files required for that are preserved. The recommended method for building OCCT 7.x is CMake, see @ref occt_dev_guides__building_cmake. The alternative solution is to use legacy generator of project files (extracted from WOK), see @ref occt_dev_guides__building_wok. @subsubsection upgrade_occt700_cdl_auto Automatic upgrade Most of typical changes required for upgrading code for OCCT 7.0 can be done automatically using the *upgrade* tool included in OCCT 7.0. This tool is a Tcl script, thus Tcl should be available on your workstation to run it. Example: ~~~~~ $ tclsh % source /adm/upgrade.tcl % upgrade -recurse -all -src= ~~~~~ On Windows, the helper batch script *upgrade.bat* can be used, provided that Tcl is either available in *PATH*, or configured via *custom.bat* script (for instance, if you use OCCT installed from Windows installer package). Start it from the command prompt: ~~~~~ cmd> \upgrade.bat -recurse -all -inc=\inc -src= [options] ~~~~~ Run the upgrade tool without arguments to see the list of available options. The upgrade tool performs the following changes in the code. 1. Replaces macro *DEFINE_STANDARD_RTTI* by *DEFINE_STANDARD_RTTIEXT*, with second argument indicating base class for the main argument class (if inheritance is recognized by the script): ~~~~~ DEFINE_STANDARD_RTTI(Class) -> DEFINE_STANDARD_RTTIEXT(Class, Base) ~~~~~ @note If macro *DEFINE_STANDARD_RTTI* with two arguments (used in intermediate development versions of OCCT 7.0) is found, the script will convert it to either *DEFINE_STANDARD_RTTIEXT* or *DEFINE_STANDARD_RTTI_INLINE*. The former case is used if current file is header and source file with the same name is found in the same folder. In this case, macro *IMPLEMENT_STANDARD_RTTI* is injected in the corresponding source file. The latter variant defines all methods for RTTI as inline, and does not require *IMPLEMENT_STANDARD_RTTIEXT* macro. 2. Replaces forward declarations of collection classes previously generated from CDL generics (defined in *TCollection* package) by inclusion of the corresponding header: ~~~~~ class TColStd_Array1OfReal; -> #include ~~~~~ 3. Replaces underscored names of *Handle* classes by usage of a macro: ~~~~~ Handle_Class -> Handle(Class) ~~~~~ This change is not applied if the source or header file is recognized as containing the definition of Qt class with signals or slots, to avoid possible compilation errors of MOC files caused by inability of MOC to recognize macros (see http://doc.qt.io/qt-4.8/signalsandslots.html). The file is considered as defining a Qt object if it contains strings *Q_OBJECT* and either *slots:* or *signals:*. 4. Removes forward declarations of classes with names Handle(C) or *Handle_C*, replacing them either by forward declaration of its argument class, or (for files defining Qt objects) \#include statement for a header with the name of the argument class and extension .hxx: ~~~~~ class Handle(TColStd_HArray1OfReal); -> #include ~~~~~ 5. Removes \#includes of files Handle_...hxx that have disappeared in OCCT 7.0: ~~~~~ #include -> ~~~~~ 6. Removes *typedef* statements that use *Handle* macro to generate the name: ~~~~~ typedef NCollection_Handle Handle(Message_Msg); -> ~~~~~ 7. Converts C-style casts applied to Handles into calls to DownCast() method: ~~~~~ ((Handle(A)&)b) -> Handle(A)::DownCast(b) (Handle(A)&)b -> Handle(A)::DownCast(b) (*((Handle(A)*)&b)) -> Handle(A)::DownCast(b) *((Handle(A)*)&b) -> Handle(A)::DownCast(b) (*(Handle(A)*)&b) -> Handle(A)::DownCast(b) ~~~~~ 8. Moves Handle() macro out of namespace scope: ~~~~~ Namespace::Handle(Class) -> Handle(Namespace::Class) ~~~~~ 9. Converts local variables of reference type, which are initialized by a temporary object returned by call to DownCast(), to the variables of non-reference type (to avoid using references to destroyed memory): ~~~~~ const Handle(A)& a = Handle(B)::DownCast (b); -> Handle(A) a (Handle(B)::DownCast (b)); ~~~~~ 10. Adds \#include for all classes used as argument to macro STANDARD_TYPE(), except for already included ones; 11. Removes uses of obsolete macros *IMPLEMENT_DOWNCAST* and *IMPLEMENT_STANDARD_*..., except *IMPLEMENT_STANDARD_RTTIEXT*. @note If you plan to keep compatibility of your code with older versions of OCCT, add option -compat to avoid this change. See also @ref upgrade_occt700_cdl_compat. . As long as the upgrade routine runs, some information messages are sent to the standard output. In some cases the warnings or errors like the following may appear: ~~~~~ Error in {HEADER_FILE}: Macro DEFINE_STANDARD_RTTI used for class {CLASS_NAME} whose declaration is not found in this file, cannot fix ~~~~~ Be sure to check carefully all reported errors and warnings, as the corresponding code will likely require manual corrections. In some cases these messages may help you to detect errors in your code, for instance, cases where *DEFINE_STANDARD_RTTI* macro is used with incorrect class name as an argument. @subsubsection upgrade_occt700_cdl_compiler Possible compiler errors Some situations requiring upgrade cannot be detected and / or handled by the automatic procedure. If you get compiler errors or warnings when trying to build the upgraded code, you will need to fix them manually. The following paragraphs list known situations of this kind. #### Missing header files The use of handle objects (construction, comparison using operators == or !=, use of function STANDRAD_TYPE() and method DownCast()) now requires the type of the object pointed by Handle to be completely known at compile time. Thus it may be necessary to include header of the corresponding class to make the code compilable. For example, the following lines will fail to compile if *Geom_Line.hxx* is not included: ~~~~~ Handle(Geom_Line) aLine = 0; if (aLine != aCurve) {...} if (aCurve->IsKind(STANDARD_TYPE(Geom_Line)) {...} aLine = Handle(Geom_Line)::DownCast (aCurve); ~~~~~ Note that it is not necessary to include header of the class to declare Handle to it. However, if you define a class *B* that uses Handle(*A*) in its fields, or contains a method returning Handle(*A*), it is advisable to have header defining *A* included in the header of *B*. This will eliminate the need to include the header *A* in each source file where class *B* is used. #### Ambiguity of calls to overloaded functions This issue appears in the compilers that do not support default arguments in template functions (known cases are Visual C++ 10 and 11): the compiler reports an ambiguity error if a handle is used in the argument of a call to the function that has two or move overloaded versions, accepting handles to different types. The problem is that operator const handle& is defined for any type *T2*, thus the compiler cannot make the right choice. Example: ~~~~~ void func (const Handle(Geom_Curve)&); void func (const Handle(Geom_Surface)&); Handle(Geom_TrimmedCurve) aCurve = new Geom_TrimmedCurve (...); func (aCurve); // ambiguity error in VC++ 10 ~~~~~ Note that this problem does not appear if macro *OCCT_HANDLE_NOCAST* is used, see @ref upgrade_occt700_cdl_nocast "below". To resolve this ambiguity, change your code so that argument type should correspond exactly to the function signature. In some cases this can be done by using the relevant type for the corresponding variable, like in the example above: ~~~~~ Handle(Geom_Curve) aCurve = new Geom_TrimmedCurve (...); ~~~~~ Other variants consist in assigning the argument to a local variable of the correct type and using the direct cast or constructor: ~~~~~ const Handle(Geom_Curve)& aGCurve (aTrimmedCurve); func (aGCurve); // OK - argument has exact type func (static_cast(aCurve)); // OK - direct cast func (Handle(Geom_Curve)(aCurve)); // OK - temporary handle is constructed ~~~~~ Another possibility consists in defining additional template variant of the overloaded function causing ambiguity, and using *SFINAE* to resolve the ambiguity. This technique can be illustrated by the definition of the template variant of method IGESData_IGESWriter::Send(). #### Lack of implicit cast to base type As the cast of a handle to the reference to another handle to the base type has become a user-defined operation, the conversions that require this cast together with another user-defined cast will not be resolved automatically by the compiler. For example: ~~~~~ Handle(Geom_Geometry) aC = GC_MakeLine (p, v); // compiler error ~~~~~ The problem is that the class *GCE2d_MakeSegment* has a user-defined conversion to const Handle(Geom_TrimmedCurve)&, which is not the same as the type of the local variable *aC*. To resolve this, use method Value(): ~~~~~ Handle(Geom_Geometry) aC = GC_MakeLine (p, v).Value(); // ok ~~~~~ or use variable of the appropriate type: ~~~~~ Handle(Geom_TrimmedCurve) aC = GC_MakeLine (p, v); // ok ~~~~~ With GCC compiler, similar problem appears when const handle to derived type is used to construct handle to base type via assignment (and in some cases in return statement), for instance: ~~~~~ const Handle(Geom_Line) aLine; Handle(Geom_Curve) c1 = aLine; // GCC error Handle(Geom_Curve) c2 (aLine); // ok ~~~~~ This problem is specific to GCC and it does not appear if macro *OCCT_HANDLE_NOCAST* is used, see @ref upgrade_occt700_cdl_nocast "below". #### Incorrect use of STANDARD_TYPE and Handle macros You might need to clean your code from incorrect use of macros *STANDARD_TYPE*() and *Handle*(). 1. Explicit definitions of static functions with names generated by macro *STANDARD_TYPE()*, which are artifacts of old implementation of RTTI, should be removed. Example: ~~~~~ const Handle(Standard_Type)& STANDARD_TYPE(math_GlobOptMin) { static Handle(Standard_Type) _atype = new Standard_Type ("math_GlobOptMin", sizeof (math_GlobOptMin)); return _atype; } ~~~~~ 2. Incorrect location of closing parenthesis of *Handle()* macro that was not detectable in OCCT 6.x will cause a compiler error and must be corrected. Example (note misplaced closing parenthesis): ~~~~~ aBSpline = Handle( Geom2d_BSplineCurve::DownCast(BS->Copy()) ); ~~~~~ #### Use of class Standard_AncestorIterator Class *Standard_AncestorIterator* has been removed; use method *Parent()* of *Standard_Type* class to parse the inheritance chain. #### Absence of cast to Standard_Transient* Handles in OCCT 7.0 do not have the operator of conversion to Standard_Transient*, which was present in earlier versions. This is done to prevent possible unintended errors like this: ~~~~~ Handle(Geom_Line) aLine = ...; Handle(Geom_Surface) aSurf = ...; ... if (aLine == aSurf) {...} // will cause a compiler error in OCCT 7.0, but not OCCT 6.x ~~~~~ The places where this implicit cast has been used should be corrected manually. The typical situation is when Handle is passed to stream: ~~~~~ Handle(Geom_Line) aLine = ...; os << aLine; // in OCCT 6.9.0, resolves to operator << (void*) ~~~~~ Call method get() explicitly to output the address of the Handle. @subsubsection upgrade_occt700_cdl_runtime Possible runtime problems Here is the list of known possible problems at run time after the upgrade to OCCT 7.0. #### References to temporary objects In previous versions, the compiler was able to detect the situation when a local variable of a "reference to a Handle" type is initialized by temporary object, and ensured that lifetime of that object is longer than that of the variable. In OCCT 7.0 with default options, it will not work if types of the temporary object and variable are different (due to involvement of user-defined type cast), thus such temporary object will be destroyed immediately. This problem does not appear if macro *OCCT_HANDLE_NOCAST* is used during compilation, see below. Example: ~~~~~ // note that DownCast() returns new temporary object! const Handle(Geom_BoundedCurve)& aBC = Handle(Geom_TrimmedCurve)::DownCast(aCurve); aBC->Transform (T); // access violation in OCCT 7.0 ~~~~~ @subsubsection upgrade_occt700_cdl_nocast Option to avoid cast of handle to reference to base type In OCCT 6.x and earlier versions the handle classes formed a hierarchy echoing hierarchy of corresponding object classes. This automatically enabled possibility to use handle to derived class in all contexts where handle to base class was needed, e.g. pass it in function by reference without copying: ~~~~ Standard_Boolean GetCurve (Handle(Geom_Curve)& theCurve); .... Handle(Geom_Line) aLine; if (GetCurve (aLine)) { // use aLine, unsafe } ~~~~ This feature was used in multiple places in OCCT and dependent projects. However it is potentially unsafe: in the above example no checks are done at compile time or at run time to ensure that argument handle is assigned a type compatible with the type of handle passed as argument. If object of incompatible type (e.g. Geom_Circle) is assigned to *theCurve*, the behavior will be unpredictable. For compatibility with existing code, by default OCCT 7.0 keeps this possibility, providing operators of type cast to handle to base type. Besides being unsafe, in specific situations this feature may cause compile-time or run-time errors as described above. In order to provide safer behavior, this feature can be disabled by defining a compile-time macro *OCCT_HANDLE_NOCAST*. When it is defined, constructors and assignment operators are defined (instead of type cast operators) to convert from handle to defived type to handle to base type. This implies creation of temporary objects and hence may be more expensive at run time in some circumstances, however this way is more standard, safer, and in general recommended. The code that relies on possibility of casting to base should be amended so that handle of argument type is always used in function call, and to use DownCast() to safely convert the result to desired type. For instance, the code from the example below can be changed as follows: ~~~~~ Handle(Geom_Line) aLine; Handle(Geom_Curve) aCurve; if (GetCurve (aCure) && !(aLine = Handle(Geom_Line)::DownCast (aCurve)).IsNull()) { // use aLine safely } ~~~~~ @subsubsection upgrade_occt700_cdl_compat Preserving compatibility with OCCT 6.x If you like to preserve the compatibility of your application code with OCCT versions 6.x even after the upgrade to 7.0, consider the following suggestions: 1. If your code used sequences of macros *IMPLEMENT_STANDARD_*... generated by WOK, replace them by single macro *IMPLEMENT_STANDARD_RTTIEXT* 2. When running automatic upgrade tool, add option -compat. 3. Define macros *DEFINE_STANDARD_RTTIEXT* and *DEFINE_STANDARD_RTTI_INLINE* when building with previous versions of OCCT, resolving to *DEFINE_STANDARD_RTTI* with single argument Example: ~~~~~ #if OCC_VERSION_HEX < 0x070000 #define DEFINE_STANDARD_RTTIEXT(C1,C2) DEFINE_STANDARD_RTTI(C1) #define DEFINE_STANDARD_RTTI_INLINE(C1,C2) DEFINE_STANDARD_RTTI(C1) #endif ~~~~~ @subsubsection upgrade_occt700_cdl_wok Applications based on CDL and WOK If your application is essentially based on CDL, and you need to upgrade it to OCCT 7.0, you will very likely need to convert your application code to non-CDL form. This is a non-trivial effort; the required actions would depend strongly on the structure of the code and used CDL features. The upgrade script and sources of a specialized WOK version used for OCCT code upgrade can be found in WOK Git repository in branch [CR0_700_2](http://git.dev.opencascade.org/gitweb/?p=occt-wok.git;a=log;h=refs/heads/CR0_700_2). [Contact us](http://www.opencascade.com/contact/) if you need more help. @subsection upgrade_occt700_bspline Separation of BSpline cache Implementation of NURBS curves and surfaces has been revised: the cache of polynomial coefficients, which is used to accelerate calculate values of B-spline, has been separated from data objects *Geom2d_BSplineCurve, Geom_BSplineCurve* and *Geom_BSplineSurface* into the dedicated classes *BSplCLib_Cache* and *BSplSLib_Cache*. The benefits of this change are: * Reduced memory footprint of OCCT shapes (up to 20% on some cases) * Possibility to evaluate the same B-Spline concurrently in parallel threads without data races and mutex locks The drawback is that direct evaluation of B-Splines using methods of curves and surfaces becomes slower due to the absence of cache. The slow-down can be avoided by using adaptor classes *Geom2dAdaptor_Curve, GeomAdaptor_Curve* and *GeomAdaptor_Surface*, which now use cache when the curve or surface is a B-spline. OCCT algorithms have been changed to use adaptors for B-spline calculations instead of direct methods for curves and surfaces. The same changes (use of adaptors instead of direct call to curve and surface methods) should be implemented in relevant places in the applications based on OCCT to get the maximum performance. @subsection upgrade_occt700_booleanresult Structural result of Boolean operations The result of Boolean operations became structured according to the structure of the input shapes. Therefore it may impact old applications that always iterate on direct children of the result compound assuming to obtain solids as iteration items, regardless of the structure of the input shapes. In order to get always solids as iteration items it is recommended to use TopExp_Explorer instead of TopoDS_Iterator. @subsection upgrade_occt700_brepextrema BRepExtrema_ExtCC finds one solution only Extrema computation between non-analytical curves in shape-shape distance calculation algorithm has been changed in order to return only one solution. So, if e.g. two edges are created on parallel b-spline curves the algorithm BRepExtrema_DistShapeShape will return only one solution instead of enormous number of solutions. There is no way to get algorithm working in old manner. @subsection upgrade_occt700_sorttools Removal of SortTools package Package *SortTools* has been removed. The code that used the tools provided by that package should be corrected manually. The recommended approach is to use sorting algorithms provided by STL. For instance: ~~~~~ #include #include #include ... TCollection_Array1OfReal aValues = ...; ... TCollection_CompareOfReal aCompReal; SortTools_StraightInsertionSortOfReal::Sort(aValues, aCompReal); ~~~~~ can be replaced by: ~~~~~ #include ... TCollection_Array1OfReal aValues = ...; ... std::stable_sort (aValues.begin(), aValues.end()); ~~~~~ @subsection upgrade_occt700_2dlayers On-screen objects and ColorScale The old mechanism for rendering Underlay and Overlay on-screen 2D objects based on *Visual3d_Layer* and immediate drawing model (uncached and thus slow) has been removed. Classes *Aspect_Clayer2d, OpenGl_GraphicDriver_Layer, Visual3d_Layer, Visual3d_LayerItem, V3d_LayerMgr* and *V3d_LayerMgrPointer* have been deleted. General AIS interactive objects with transformation persistence flag *Graphic3d_TMF_2d* can be used as a replacement of *Visual3d_LayerItem*. The anchor point specified for transformation persistence defines the window corner of (or center in case of (0, 0) point). To keep on-screen 2D objects on top of the main screen, they can be assigned to the appropriate Z-layer. Predefined Z-layers *Graphic3d_ZLayerId_TopOSD* and *Graphic3d_ZLayerId_BotOSD* are intended to replace Underlay and Overlay layers within the old API. *ColorScale* object previously implemented using *Visual3d_LayerItem* has been moved to a new class *AIS_ColorScale*, with width and height specified explicitly. The property of *V3d_View* storing the global *ColorScale* object has been removed with associated methods *V3d_View::ColorScaleDisplay(), V3d_View::ColorScaleErase(), V3d_View::ColorScaleIsDisplayed()* and *V3d_View::ColorScale()* as well as the classes *V3d_ColorScale, V3d_ColorScaleLayerItem* and *Aspect_ColorScale*. Here is an example of creating *ColorScale* using the updated API: ~~~~~ Handle(AIS_ColorScale) aCS = new AIS_ColorScale(); // configuring Standard_Integer aWidth, aHeight; aView->Window()->Size (aWidth, aHeight); aCS->SetSize (aWidth, aHeight); aCS->SetRange (0.0, 10.0); aCS->SetNumberOfIntervals (10); // displaying aCS->SetZLayer (Graphic3d_ZLayerId_TopOSD); aCS->SetTransformPersistence (Graphic3d_TMF_2d, gp_Pnt (-1,-1,0)); aCS->SetToUpdate(); theContextAIS->Display (aCS); ~~~~~ To see how 2d objects are implemented in OCCT you can call Draw commands *vcolorscale, vlayerline* or *vdrawtext* (with -2d option). Draw command *vcolorscale* now requires the name of *ColorScale* object as argument. To display this object use command *vdisplay*. For example: ~~~~~ pload VISUALIZATION vinit vcolorscale cs -demo pload MODELING box b 100 100 100 vdisplay b vsetdispmode 1 vfit vlayerline 0 300 300 300 10 vdrawtext t "2D-TEXT" -2d -pos 0 150 0 -color red ~~~~~ Here is a small example in C++ illustrating how to display a custom AIS object in 2d: ~~~~~ Handle(AIS_InteractiveContext) aContext = ...; Handle(AIS_InteractiveObject) anObj =...; // create an AIS object anObj->SetZLayer(Graphic3d_ZLayerId_TopOSD); // display object in overlay anObj->SetTransformPersistence (Graphic3d_TMF_2d, gp_Pnt (-1,-1,0)); // set 2d flag, coordinate origin is set to down-left corner aContext->Display (anObj); // display the object ~~~~~ @subsection upgrade_occt700_userdraw UserDraw and Visual3d #### Visual3d package Package *Visual3d* implementing the intermediate layer between high-level *V3d* classes and low-level OpenGl classes for views and graphic structures management has been dropped. The *OpenGl_View* inherits from the new class *Graphic3d_CView*. *Graphic3d_CView* is an interface class that declares abstract methods for managing displayed structures, display properties and a base layer code that implements computation and management of HLR (or more broadly speaking view-depended) structures. In the new implementation it takes place of the eliminated *Visual3d_View*. As before the instance of *Graphic3d_CView* is still completely managed by *V3d_View* classes. It can be accessed through *V3d_View* interface but normally it should not be required as all its methods are completely wrapped. In more details, a concrete specialization of *Graphic3d_CView* is created and returned by the graphical driver on request. Right after the creation the views are directly used for setting rendering properties and adding graphical structures to be displayed. The rendering of graphics is possible after mapping a window and activating the view. The direct setting of properties obsoletes the use of intermediate structures with display parameter like *Visual3d_ContextView*, etc. This means that the whole package *Visual3d* becomes redundant. The functionality previously provided by *Visual3d* package has been redesigned in the following way : - The management of display of structures has been moved from *Visual3d_ViewManager* into *Graphic3d_StructureManager*. - The class *Visual3d_View* has been removed. The management of computed structures has been moved into the base layer of *Graphi3d_CView*. - All intermediate structures for storing view parameters, e.g. *Visual3d_ContextView*, have been removed. The settings are now kept by instances of *Graphic3d_CView*. - The intermediate class *Visual3d_Light* has been removed. All light properties are stored in *Graphic3d_CLight* structure, which is directly accessed by instances of *V3d_Light* classes. - All necessary enumerations have been moved into *Graphic3d* package. #### Custom OpenGL rendering and UserDraw Old APIs based on global callback functions for creating *UserDraw* objects and for performing custom OpenGL rendering within the view have been dropped. *UserDraw* callbacks are no more required since *OpenGl_Group* now inherits *Graphic3d_Group* and thus can be accessed directly from *AIS_InteractiveObject*: ~~~~~ //! Class implementing custom OpenGL element. class UserDrawElement : public OpenGl_Element {}; //! Implementation of virtual method AIS_InteractiveObject::Compute(). void UserDrawObject::Compute (const Handle(PrsMgr_PresentationManager3d)& thePrsMgr, const Handle(Prs3d_Presentation)& thePrs, const Standard_Integer theMode) { Graphic3d_Vec4 aBndMin (myCoords[0], myCoords[1], myCoords[2], 1.0f); Graphic3d_Vec4 aBndMax (myCoords[3], myCoords[4], myCoords[5], 1.0f); // casting to OpenGl_Group should be always true as far as application uses OpenGl_GraphicDriver for rendering Handle(OpenGl_Group) aGroup = Handle(OpenGl_Group)::DownCast (thePrs->NewGroup()); aGroup->SetMinMaxValues (aBndMin.x(), aBndMin.y(), aBndMin.z(), aBndMax.x(), aBndMax.y(), aBndMax.z()); UserDrawElement* anElem = new UserDrawElement (this); aGroup->AddElement(anElem); // invalidate bounding box of the scene thePrsMgr->StructureManager()->Update (thePrsMgr->StructureManager()->UpdateMode()); } ~~~~~ To perform a custom OpenGL code within the view, it is necessary to inherit from class *OpenGl_View*. See the following code sample: ~~~~~ //! Custom view. class UserView : public OpenGl_View { public: //! Override rendering into the view. virtual void render (Graphic3d_Camera::Projection theProjection, OpenGl_FrameBuffer* theReadDrawFbo, const Standard_Boolean theToDrawImmediate) { OpenGl_View::render (theProjection, theReadDrawFbo, theToDrawImmediate); if (theToDrawImmediate) { return; } // perform custom drawing const Handle(OpenGl_Context)& aCtx = myWorkspace->GetGlContext(); GLfloat aVerts[3] = { 0.0f, 0,0f, 0,0f }; aCtx->core20->glEnableClientState(GL_VERTEX_ARRAY); aCtx->core20->glVertexPointer(3, GL_FLOAT, 0, aVerts); aCtx->core20->glDrawArrays(GL_POINTS, 0, 1); aCtx->core20->glDisableClientState(GL_VERTEX_ARRAY); } }; //! Custom driver for creating UserView. class UserDriver : public OpenGl_GraphicDriver { public: //! Create instance of own view. virtual Handle(Graphic3d_CView) CreateView (const Handle(Graphic3d_StructureManager)& theMgr) Standard_OVERRIDE { Handle(UserView) aView = new UserView (theMgr, this, myCaps, myDeviceLostFlag, &myStateCounter); myMapOfView.Add (aView); for (TColStd_SequenceOfInteger::Iterator aLayerIt (myLayerSeq); aLayerIt.More(); aLayerIt.Next()) { const Graphic3d_ZLayerId aLayerID = aLayerIt.Value(); const Graphic3d_ZLayerSettings& aSettings = myMapOfZLayerSettings.Find (aLayerID); aView->AddZLayer (aLayerID); aView->SetZLayerSettings (aLayerID, aSettings); } return aView; } }; ~~~~~ @subsection upgrade_occt700_localcontext Deprecation of Local Context The conception of Local Context has been deprecated. The related classes, e.g. *AIS_LocalContext*, and methods ( AIS_InteractiveContext::OpenLocalContext() and others) will be removed in a future OCCT release. The main functionality provided by Local Context - selection of object subparts - can be now used within Neutral Point without opening any Local Context. @subsection upgrade_occt700_separate_caf_visualisation Separation of visualization part from TKCAF Visualization CAF attributes moved into new toolkit TKVCAF. If your application uses the classes from TPrsStd package then add link to TKVCAF library. Verson numbers of BinOCAF and XmlOCAF formats are incremented; new files cannot be read by previous versions of OCCT. For loading OCAF files saved by previous versions and containing attribute TPrsStd_AISPresentation it is necessary that environment variable CSF_MIGRATION_TYPES should be defined, pointing to file src/StdResources/MigrationSheet.txt. When using documents loaded from a file, make sure to call method TPrsStd_AISViewer::New() prior to accessing TPrsStd_AISPresentation attributes in this document (that method will create them). @subsection Correction of interpretation of Euler angles in gp_Quaternion Conversion of gp_Quaternion to and from intrinsic Tait-Bryan angles (including gp_YawPitchRoll) is fixed. Before that fix the sequence of rotation axes was opposite to intended; e.g. gp_YawPitchRoll (equivalent to gp_Intrinsic_ZYX) actually was defining intrinsic rotations around X, then Y, then Z. Now this is fixed, and rotations are made in correct order. Applications that use gp_Quaternion to convert Yaw-Pitch-Roll angles (or other intrinsic Tait-Bryan sequences) may need to be updated to take this change into account.