h5repack method

Problem: I need to delete specific groups from a file, which contain large datasets. For that, I’m doing:

with h5py.File(‘file_name’, 'a') as f:
    del f['group_name']

This indeed removes the group, but the file size remains unchanged.

In scattered answers online, I found a possible solution would be using h5repack. I can’t find, however, if this method is implemented anywhere in h5py. I’m opening this issue in the hope that it’s a reasonable request.

• Operating System: Windows 10
• Python version 3.7
• Anaconda 3
• h5py version 2.9.0
• HDF5 version 1.10.4

[scopatz]

Pytables has repack functionality that I think you can import. Also, PRs welcome here. Also, note that repack creates a new, smaller file. It doesn’t change the size of the existing file.

[tacaswell]

Automatically calling h5repack on close is not something safe to do. If you remove something small from the start of a very big file, you could incur a huge amount of I/O as you re-write multiple gigabytes of data to reclaim a few kb.

@danschef I think we would happily merge a PR to the docs adding the text that would have saved you time/confusion!