Versioned HDF5 Change Log¶

1.2 (2020-11-17)¶

Add support for sparse datasets (data=None).
Store the chunks on an attribute of the dataset.
versioned-hdf5 is currently pinned to h5py<3. h5py 3 support will be added in a future version.
VersionedHDF5File[timestamp] now returns the closest version before timestamp if there is no version at timestamp.

Added support for shape-0 datasets.
Fix a memory leak where in-memory datasets would not be garbage collected.
Added support for empty datasets (size 0).
Make sure versioned data is read-only after closing and reopening the file.
Allow deleting groups and datasets between versions. Note that currently dataset metadata cannot change between versions, even if they are deleted in between.

Most tests now use a temporary directory instead of writing a file in the current directory.
Fix logic for handling trailing slashes with in.
Automatically create intermediate groups when creating a dataset.
Make indices that should give a scalar object do so instead of giving a shape () array.