❗ API change: a mandatory parameter for the number of time points is dropped (redundancy), and the function
name is shortened: BdvWriter.write_xml(...) instead of BdvWriter.write_xml_file(ntimes=...).
💎 Lazy downsampling and compression is added: BdvWriter.create_pyramids().
This allows high-speed writing of raw data during acquisition runtime, followed by slower
generation of downsampled and compressed image pyramids (when the time is less critical).
💎 Multiple XML files pointing to the same H5 data file are supported. This allows having several versions of registration or other processing (defined in XML files). See PR #9.
💎 Set labels for view attributes that will be visible in BDV/BigSticher,
e.g. .set_attribute_labels('channel', ('488', '561')).
❗ API change: The BdvReader class was replaced with BdvEditor, which allows e.g. streamlined reading and cropping views in H5 and XML files.
💎 new function BdvWriter.append_substack() to write substacks into a virtual stack.
This allows saving virtual stacks with multi-resolution pyramids downsampled in z as well as (x,y).
💎 BdvWriter and BdvEditor both have method .read_affine(...), which reads an affine transformation of
a view from the XML file as numpy (3,4) array.
💎 BdvWriter and BdvEditor both have method .append_affine(...), which appends affine transformation matrix
to an existing dataset.
This allows e.g. defining unshearing transform and tile position (via translation transform) after the data has been written.
Before, only one transformation could be defined when creating a new dataset in BdvWriter.append_view().
🔍 Test coverage: The tests became much more comprehensive and cover most of the functionality.
🐛 Bugfix: affine transformations of different views were mixed due to using mutable argument m_affine in append_view()
💎 New example: writing multiple tiles on a grid, with (optionally) some tiles missing.
📖 API documentation is added.