Resampling HDF5 Outputs

The Python script athena/vis/python/uniform.py can be used to rewrite .athdf HDF5 files with a single, uniform MeshBlock. This is useful for reading into VisIt (whose performance suffers when thousands of MeshBlocks are present), or for any other analysis tool that would be better served without dealing with the MeshBlock structure.

Note: It is a known issue in VisIt that the number of MeshBlocks (stored as HDF5 data containers) in an open database is not updated when the timeslice is changed. This may result in missing/empty patches of the domain when AMR derefines a MeshBlock; reopening the database will fix the missing data.

The script is used as

python uniform.py <input_filename> <output_filename> <start> <end> <stride> [options]

The necessary arguments are:

• input_filename: base name of files to be rewritten, including the path but excluding the 5-digit number. Files are not modified unless output_filename is the same.
• output_filename: base name of output files to be written. This can be same as input_filename.
• start: integer of first file to rewrite, corresponding to 5-digit identifier (does not need to include leading zeros).
• end: integer of last file to rewrite, corresponding to 5-digit identifier (does not need to include leading zeros).
• stride: stride in file identifiers, useful for only rewriting some of them.

Optional arguments are:

• -m: run script in parallel in an MPI environment. Files are distributed equally among the processes.
• -x: do not write corresponding XDMF (.athdf.xdmf) files
• -l <level>, --level <level>: force refinement level of output to be this, relative to root grid in input. Must be nonnegative integer. If omitted, each file is upsampled to its own maximum refinement level.
• -q <quantity_1> [<quantity_2> ...], --quantities <quantity_1> [<quantity_2> ...]: list of quantities to include in rewriting. These should be variable names, not dataset names. If omitted, all cell-centered quantities are included.

For example, to create 10 new files that include only density and the first component of magnetic field, one might use the script as

python uniform.py data/old.prim data/new.prim 0 9 1 --quantities rho Bcc1

This would use the data in data/old.prim.0000{0..9}.athdf to create data/new.prim.0000{0..9}.athdf and data/new.prim.0000{0..9}.athdf.xdmf.

If a refinement level is forced with --level to be less than the maximum refinement of a file, the subsample=True option will be used in the internal call to the Python HDF5 data reader.

The new HDF5 file will internally consist of a single MeshBlock covering the entire domain. The format is the same as the HDF5 format produced by Athena++, with the exception that datasets are not preserved if the --quantities option is used. In this case the only cell-centered dataset present will be called "quantities", and all variables will be inside this dataset.