There are two data file formats. One especially for adaptive data and one for uniform data. For reasons of disk space savings all data is stored binary. To maintain a simple data transfer between machines with different architectures, all values are automatically casted from/to BigEndian order. The library user does not need to care about this. Cast to/from float is supported for additional storage savings.
|unified data format (UDF)||WriteUDF(.) and ReadUDF(.)||The UDF agrees with VTK's rectilinear grid format.|
|Additionally, the comment (second text line in the file)|
|MAY contain information about e.g. boundary conditions.|
|Read/WriteUDF are implemented not for all multivariate|
|data structures and the arguments may vary,|
|see the Programming Reference|
|sparse file format||WriteSparse(.) and ReadSparse(.)||files consist of a header and lines of <D> integers and one or more|
|floats or doubles. The D integers contain the index and the|
|floats/doubles the numerical data.|
|It is possible and recommended to store several|
|AdaptiveDatas linked to the same adaptive grid in one file.|
|Then, the index information needs to be stored only once.|
Except for some programs to plot 2D and 3D adaptive grid point distributions, there are no tools to directly visualize sparse data. Therefore, one has to use WriteUDF. This has two drawbacks: In most cases there are large levels active in adaptive grids. It is, however, not possible to write the data w.r.t. the respective very fine grid. Therefore some coarser grid must be used which leads to information loss. On the other hand, even the coarse grid may have significantly more degrees of freedom than the original adaptive grid. Darn it!