Alphabetical ordering of I/O binary files leads to Paraview visualization issue

[bahardy]

Care should be taken when one chooses names for the binary files in the I/O files (out2d.90,out3d.h90).

The python script write_xdmf.py orders fields to be visualized in the xdmf file based on the name of the binary files. If the 3 components of a vector field (e.g. velocity) are not contiguous in terms of alphabetical ordering, xdmf will not recognize it as a vector field but instead as 3 independent scalar fields, with annoying consequences in further Paraview visualization.

As an example if 2-D slices binary files for the three velocity components are named as
u_slice_fld_*****.bin, v_slice_fld_*****.bin, w_slice_fld_*****.bin
and that, at the same time, we print vorticity components as
vox_slice_fld_*****.bin, voy_slice_fld_*****.bin,voz_slice_fld_*****.bin,
the three vorticity files will be printed in between v_slice_fld_*****.bin and w_slice_fld_*****.bin in the xdmf file, so that the 3 velocity components will be considered as 3 distinct scalar fields.

[p-costa]

Good catch. This needs to be fixed. Again, thanks a lot for finding and reporting it!

The workaround for now is to name the output binary files such that, when all files are ordered alphabetically, three vector field components always appear consecutively in the list.

[p-costa]

Does changing the line in write_xdmf from:

saves = np.sort(saves, order='isave')

to

saves = np.sort(saves, order=['isave', 'variable'])

fix the issue?

And just in case, I should also change the default names of velocity in out2d.h90.

Note to self: Do not forget to change the two .xdmf generators (raw binary and HDF5).

[bahardy]

This fix seems to work perfectly ! 👍

[p-costa]

Great! Thanks for checking!