This release fixes the lack of optimization support when iterating over
`b2h5py.B2Dataset` insta...

And adapt the related tests. Also remove the readme notice about the
operation not being optimi...

This release adds a new way to use Blosc2 optimized slicing explicitly on a
single dataset. You...

No need for this as it is already done elsewhere downstream.

This adds a `b2h5py.B2Dataset` class to wrap `h5py.Dataset` instances
explicitly. Slicing opera...

For consistency with the rest of code.

As `h5py.Dataset.__iter__` calls back `__getitem__`.

Also minor change to main package docstring for consistency.

This avoids repeating the checks, and also fixes the old code not checking
whether optimization ...

Reuse in `B2Dataset` tests too.

This release changes the package interface to require explicitly enabling
optimizations and use ...

The monkey-patching was placing the property under a name which differred from
that of the real ...

This moves all dataset class patching-related code into its own module,
separate from Blosc2-spe...

So that it is easier to handle programmatically.

So that the Blosc2-specific module contains no code related with
monkey-patching.

This changes the default behaviour of importing `b2h5py` to need explicit
patching of the `h5py....

Otherwise the exception that should be raised never is.

Since `import b2h5py` no longer enables optimization implicitly.

Now the patching needs to be explicit.

This provides a very simple interface to enabling the optimizations, while
allowing the plain im...

For explicit temporary optimization support.

This renames patching-related functions to instead refer to enabling and
disabling fast slicing,...

Instead of `b2h5py.patching_dataset_class()`. Update docstrings to lessen the
relevance of the ...

Instead of `b2h5py.is_dataset_class_patched()`. Update docstrings to lessen
the relevance of th...

Instead of `b2h5py.patch_dataset_class()`. Update docstrings to lessen the
relevance of the imp...

Instead of `b2h5py.unpatch_dataset_class()`. Update docstrings to lessen the
relevance of the i...

This also avoids referring to the wrong variable later on.

For consistency with other passages. Also fix print calls referring to the
wrong variable.

It is no longer mentioned in the readme nor in the example; however, a note is
kept in the packa...

For consistency with other function mentions. Also remove in-sentence
newline.

This release adds the context manager `b2h5py.patching_dataset_class()` that
patches the `h5py.D...

Although `from b2h5py import *` does not make much sense, it is useful for
documentation purpose...

This adds the context manager `b2h5py.patching_dataset_class()` that patches
the dataset class o...

It takes care of requiring a test exception or not, and using the class
patching context manager.

This adds GitHub actions to create distributions files and use them to run
several tests on push...

Since additional dependencies are required by h5py tests, a package extra has
been added to requ...

This release only includes minor fixes to allow unit tests to check whether
optimizations are ac...

No need to continue following h5py's code styling rules.

This adds support for testing when a slicing operation which should be able to
use Blosc2 optimi...

And a test to check that the machinery itself works.

To allow testing code to force an exception when they are not used.

Otherwise it may read as `pip install ..`.

They come from PyTables' `bench/b2nd_compare_getslice_plot.py` (narrowed down
to h5py-related re...

Since it is not part of the public API.

No need for PyTest or anything else than `unittest`: tests are very simple for
the moment.