HiSPEET

Hi Scot, thank you for pointing to that interesting approach. Indeed I was not aware of the variable length features of HDF5. Unfortunately, the suggested approach requires to introduce additional...

A trick to fix the corrupted allocation status in Fortran is described here: https://stackoverflow.com/questions/64891525/reset-deallocate-nullify-a-fortran-allocatable-array-that-has-been-corrupted This could solve our problem. But, of course I would prefer a clean solution in HDF5....

> You can either use [...], or use an array of integers with unlimited dimension, pack all the variables in the array and store another array with element lengths. >...

To illustrate the workaround with CFI_deallocate consider the appended example. The basic idea is as follows ``` interface ! interface to CFI_deallocate from ISO Fortran binding integer function CFI_deallocate(a) bind(C,...

Hi Scot, > All the data (including nd) you allocate is written in the hdf5 file I believe there is still an misunderstanding: a dump of the HDF5 file produced...

> h5dump will only report registered members in the compound type, so it does not report nb values. So the descriptors are written and then read, which explains the observed...