How to delete an existing dataset?

[aghaeifar]

Hi all,

Is it feasible to replace an existing dataset with data that may have a different shape?
I can verify the existence of the dataset name using HighFive::File::exist(...), but I'm uncertain how to delete it and then create a new one using createDataSet(...).
Could you please advise on the approach to take when replacing old data is necessary?
It's important to note that there are other datasets in the HDF5 file, so overwriting the entire file is not a viable solution.

Thanks.

[1uc]

You can "delete" a dataset with:

grp.unlink(dset_name)

and create a new dataset. The issue is that unlinking will not automatically shrink the size of the file. I think it'll (sometimes) mark the space as unused and then reuse it if possible. You can use h5repack to copy the file, but eliminating all gaps in the original file.

Note that the special case of a growing dataset is typically solved differently. One creates a DataSet that's unlimited in the direction in which it'll grow. Then one can later resize the dataset in the unlimited direction and overwrite data as needed.