In the H5MD publication ( https://dx.doi.org/10.1016/j.cpc.2014.01.018 ) it says that HDF5 files support compression and thereby that H5MD files do as well. Using compression with HDF5 files requires chunked storage. These chunks do not have to be 1 frame (especially not to get efficient compression). Chunks of more than 1 frame will result in "empty" frames. It does not say in the H5MD specification how to handle this case. But since it says that other fields should ignore elements that are set to fill value I think that would be the best approach for time dependent data as well.
I would suggest checking if the time stamp or the step number equals the fill value, when reading a frame. If so, that frame should be skipped.
In the H5MD publication ( https://dx.doi.org/10.1016/j.cpc.2014.01.018 ) it says that HDF5 files support compression and thereby that H5MD files do as well. Using compression with HDF5 files requires chunked storage. These chunks do not have to be 1 frame (especially not to get efficient compression). Chunks of more than 1 frame will result in "empty" frames. It does not say in the H5MD specification how to handle this case. But since it says that other fields should ignore elements that are set to fill value I think that would be the best approach for time dependent data as well.
I would suggest checking if the time stamp or the step number equals the fill value, when reading a frame. If so, that frame should be skipped.