VRMLTest - Full VRML 1.x parser and partial viewer

This article presents:

• CVRMLScene a "Scene graph" and all the associated classes, including a full VRML 1. x parser.
• COGLCanvas my own YAOWM (yet another OpenGL wrapper for MFC ;)).
• CVRMLCanvas a descendant of COGLCanvas a sample application that renders a limited set of VRML primitives using OpenGL. Samples for the kind of models supported by this canvas can be freely downloaded at http://www.rosl.com (site which I have nothing to do with, at least yet ;)). The .wrl files included in .zip files are from that URL (except allnodes.wrl).
• vrmltest, the sample application of all the above and which screenshots you can see at the top.

Index

• Introduction
• Full VRML 1.x Parser
• MFC OpenGL wrapper
• VRML partial viewer
• Sample application

This article is far from exhaustive but you can peep at the documentation of all the classes, which is autogenerated from the code with Eric Artzt's Autoduck, an incode documentation tool (that's why the weird comments you see in the code). I plan to release an article on this autodocumentation tool some day (or another).

Full VRML 1.x parser

The main class is CVRMLScene, where resides the VRML file parsing and loading capabilities. All the other classes define the structure needed to traverse the built scene graph (and do whatever you want with it).

The parser is done with Bison and Flex, so if you modify the sources, take care not to modify the autogenerated files vrml1.cpp (Flex generated) or vrml1_tab.cpp (Bison generated). Only if you don't have at range Bison or Flex, or if you don't want to deal with those beasts, it would be convenient to modify them directly.

Just to state it clear: you don't need Bison and Flex at all to use and compile this classes, only if you want to modify vrml.l (Flex's token definition file) or vrml.y (Bison's grammar) you will need to regenerate vrml1.cpp and vrml1_tab.cpp. Anyway, only because of bugs you should need to modify those files, as the main functionality of these classes reside in VRML.cpp file.

I've based the design of the parser in SGI's VRML 2.x parser, which is one of the simplest and smartest things I've seen on the matter: The parser only knows about simple types of VRML (SFxxxx and MFxxxx) and definition nodes, and knows nothing about the nodes themselves. Then it loads a library of default nodes (defined in a totally compliant VRML file, which has the peculiarity that it only defines nodes and not instances of them), and presto!, you are ready to load any VRML 1.x compliant file.
The included file allnodes.wrl is that library of standard VRML 1.x nodes. The normal operation is to load that library as if it was a normal VRML file (and it IS!), so standard nodes get defined, and then load your desired VRML file/s (you can load as many VRML files as you want in the same scene).

The error detecting capabilities of this class are quite weak, it only returns FALSE if file is invalid, and I wouldn't find it strange if it hanged with corrupt VRML files :-m.

The parser has two non-stantard properties: the VRML file needn't to be a unique VRML node (there are plenty of VRML files that have more than one root node), and any node (and not only group nodes) can have children nodes. The latter is a side effect of not knowing beforehand the standard nodes (anyway, this is VRML 1.x standard's fault, because user defined nodes can't mark in any way if they have children nodes or not ... VRML 1.x is quite a dumb language, if I may say it: it's not LALR, and some hacks are needed to parse it with Bison).

This wrapper exports the overridable function Render, so descendants only have to implement that function and this class handles everything else:

• interaction with mouse: pan (right button), zoom (left button, moving up) and rotation (both buttons)
• wireframe, solid and smooth rendering
• lighting (it sets what VRML calls a headlight)

Someday it will support switching between orthographic and perspective (is just adding the methods, right now the perspective code is commented) and OpenGL selection modes.
Oh, yes, I almost forget it: the mouse buttons get sticked sometimes.

The viewer CVRMLCanvas is based on the classes COGLCanvas and CVRMLScene. It takes a simple CVRMLScene and renders it the best it can ;).

The supported nodes are Material, IndexedFaceSet and Coordinate3.
As I said above, you can find free sample supported VRML files at the robots library of Robot Simulations Ltd. and some of them are included in .zip files.

In the directory ./wrl of the .zip you'll find sample files. Just load one of them and play with it.