Writing a VTK ASCII Legacy File to draw contours in VisIt - c

Question

I'm trying to write a legacy .vtk file to be read into VisIt using C. Unfortunately my installed VisIt program refuses to render the VTK file that I have writing, reading: 'local host failed'
Below is the code used to read data from one file and convert it to a legacy VTK file. I use the macros XPIX, YPIX, and ZPIX to describe the dimensions of a pixel grid. Each pixel contains a scalar density value. I have listed the pixels in a 'grid-file' using row-major ordering: i.e.
int list_index(x,y,z) = YPIX * ZPIX * x + ZPIX * y + z;
Every entry in this pixel list is read into an array called grid[] of type double, and written to outfile beneath the legacy VTK header data:
/*Write vtk header */
fprintf(outfile,"# vtk DataFile Version 3.0\n");
fprintf(outfile,"Galaxy density grid\nASCII\nDATASET STRUCTURED_POINTS\n");
fprintf(outfile,"DIMENSIONS %d %d %d \n", (XPIX+1), (YPIX+1), (ZPIX+1));
fprintf(outfile,"ORIGIN 0 0 0\n");
fprintf(outfile,"SPACING 1 1 1\n");//or ASPECT_RATIO
fprintf(outfile,"CELL_DATA %d\n", totalpix);
fprintf(outfile,"SCALARS cell_density float 1\n");
fprintf(outfile, "LOOKUP_TABLE default\n");
/*Create Memory Space to store Pixel Grid*/
double *grid;
grid = malloc(XPIX * YPIX * ZPIX * sizeof(double));
if (grid == NULL ){
fprintf(stderr, "Pixel grid of type double failed to initialize\n");
exit(EXIT_FAILURE);
}
fprintf(stderr,"Pixel grid has been initialized.\n Now reading infile\n");
/*Read infile contents into double grid[], using Row-Major Indexing*/
double rho;
char newline;
int i, j, k;
for(i = 0; i < XPIX; i++){
for(j = 0; j < YPIX; j++){
for(k = 0; k < ZPIX; k++){
fscanf(infile, "%lf", &rho);
grid[getindex(i,j,k)] = rho;
}
}
fprintf(stderr,"%d\n", i);
}
fprintf(stderr,"Finished reading\n");
#if !DEBUG
/*Write out grid contents in Row major order*/
fprintf(stderr,"Now writing vtk file");
for(i = 0; i < XPIX; i++){
for(j = 0; j < YPIX; j++){
for(k = 0; k < ZPIX; k++){
fprintf(outfile, "%lf ", grid[getindex(i,j,k)]);
}
fprintf(outfile,"\n");
}
}
fprintf(stderr,"Finished Writing to outfile\n");
#endif
After running the grid data list through this routine, I have XPIX*YPIX lines in the lookup_table, each with ZPIX entries. Is this an incorrect format? VisIt continues to fail reading the input file. I'm aware of the fact that structured_points may use column major indexing, but my first goal of course is to get some sort of result from VisIt. I would like to draw a contour using the scalar cell_density eventually. Is my data set simply too large?

Answer 1

Have you seen the accepted answer to the question vtk data format error? The question is debugging a VTK writer in C++ but it is very similar to your code (and of course should yield the same results).
The key point from the accepted answer is that data is written in column major order, not row major order (you seem to hint at this in your question: "I'm aware of the fact that structured_points may use column major indexing").
Also, it is always helpful (if you can) to compare your code with something which you know works. For example, VisIt provides a small C library for writing legacy VTK file formats called VisItWriterLib. Compare the output from your code and the VisItWriterLib to see where your data files differ. I would recommend using VisItWriterLib for your VTK IO rather than writing your own routines - no need to reinvent the wheel.
Edit: To answer a couple of your other questions:
After running the grid data list through this routine, I have
XPIX*YPIX lines in the lookup_table, each with ZPIX entries. Is this
an incorrect format?
This is not the correct format. LOOKUP_TABLE should be a list of XPIX*YPIX*ZPIX lines, with one element per line (or alternatively, VisIt will accept one line with XPIX*YPIX*ZPIX elements). See the section Dataset Attribute Format in the VTK File Formats document (www.vtk.org/VTK/img/file-formats.pdf).
Is my data set simply too large?
I doubt it. VisIt is designed to handle huge datasets and, AFAIK, can render petabyte data sets. I would be very surprised if your data is that large.
However, if you are concerned about having large files, you can split your data into multiple files and tell VisIt to read these files is parallel. To do this, write a bit a your data into separate files, e.g. domain1.vtk, domain2.vtk, ... domainN.vtk etc. Then write a .visit master file, which has the structure
!NBLOCKS N
domain1.vtk
domain2.vtk
...
domainN.vtk
Save this as, for example, mydata.visit and then open this .visit file, rather than the .vtk files, in VisIt.