VTK Adaptor Framework README


The VTK Adaptor Framework and Generic Filtering

Overview

The classes in this directory are part of a body of work that support the interface of VTK to external simulation packages. These external packages may be finite element programs, CFD systems, or other simulation systems that are based on discretizing a domain into elements, where the element (or in VTK terminology, cell) basis functions are of arbitrary complexity. The purpose of this work is to avoid duplicating data in order to visualize it with VTK, and to take advantage of basis function formulations that are already implemented in existing simulation packages. (For further details, please see the paper "Framework for Visualizing Higher-Order Basis Functions" by Schroeder, Bertel, Malaterre, Thompson, Pebay, O'Bara and Tendulkar, in the Proceedings of IEEE Visualization 2005, October 2005.)

In the past, systems that interface with VTK typically require writing output data in VTK format, or require writing a special VTK reader. Further, cells of higher-order basis (e.g., cubic, quadric, or even mixed-order formulations) had no counterpart in VTK, so the user was responsible for tessellating the cells into linear forms, often at the cost of excessive memory consumption or inaccuracy.

We refer to this work as the "adaptor framework". That is, it adapts VTK to a particular simulation system. It does this by generalizing the VTK dataset, cell, and attribute classes, and providing special filters that operate on these new general classes. Further, because we assume that the cells may be of arbitrary complexity in terms of their basis functions, tessellation functions (with associated error metric) are used to automatically subdivide complex cells into linear primitives. This allows the use of standard linear visualization algorithms: algorithms that have been proven to be robust, fast, and dependable.

The adaptor framework consists of a few basic components: new implementations of data objects (vtkGenericDataSet and vtkGenericAdaptorCell), classes to support the tessellation of cells (vtkGenericCellTessellator and vtkGenericSubdivisionErrorMetric), and classes that filter the new data objects (e.g., vtkGenericContourFilter).

Note that we choose to implement a select number of filters. One filter simply performs tessellation on a vtkGenericDataSet to produce a linearized vtkUnstructuredGrid (vtkGenericDataSetTessellator). The others are common filters used in a wide variety of visualization applications:

At the same time that we were designing the adaptor framework, we also found it necessary to rework the vtkDataSet/vtkCell/vtkDataArray API. We found that the assumptions of contiguous data (manifested in the form of monotonically increasing, contiguous, integral ids) to be overly limiting. Thus we moved to an iterator-based framework. It is possible that in time the vtkGenericDataSet may replace the current vtkDataSet, and that the existing filters that input vtkDataSet may be rewritten. However, this is a future undertaking.

Compatibility Warning

The classes in this directory (and the classes found in VTK/Common/vtkGeneric*) are undergoing rapid development. The names of these classes, and their API, may change. Please use them only if you are prepared to deal with the inevitable changes that will occur over the next six months.

Commercial Expression

Kitware http://www.kitware.com is teaming with Simmetrix, Inc. http://www.simmetrix.com to create commercial tools that integrate the VTK visualization environment to Simmetrix's products, and to integrate Simmetrix's integration framework into ParaView.

If you are interested in building an adapted framework to VTK from your analysis/simulation package, please email sales "at" kitware.com for more information.

Acknowledgements

This work has been supported through the efforts of several organizations and individuals.

The National Science Foundation

The US National Tri-Lab Community (Sandia, Livermore, Los Alamos)

Kitware, Inc.

Simmetrix, Inc.

LeHigh University.