The LINDI group writer create_dataset supports the kwargs chunks, compression, and compression_opts and the compression-related kwargs are allowed only if compression=="gzip". This is the most popular use case.
HDMF supports a few additional kwargs for create_dataset via H5DataIO, e.g., compression that is not gzip, such as through hdf5plugin (see pynwb tutorial), compression_opts, maxshape, fillvalue, shuffle, fletcher32. In practice, fillvalue and shuffle are rarely if ever used. Blosc compression through hdf5plugin is sometimes used because it is usually better than gzip. Maxshape will be used more often, pending an upcoming change https://github.com/hdmf-dev/hdmf/pull/1064. Note that h5py.Datasetsupports even more kwargs that HDMF does not currently support.
I started adding support for additional compression algorithms as a nice exercise and will see how far I can get. I'll update here in a few days. This is not high priority but it would be nice to have more complete support for writing/appending datasets via PyNWB to a LINDI file.
The LINDI group writer
create_datasetsupports the kwargschunks,compression, andcompression_optsand the compression-related kwargs are allowed only ifcompression=="gzip". This is the most popular use case.HDMF supports a few additional kwargs for
create_datasetviaH5DataIO, e.g.,compressionthat is not gzip, such as through hdf5plugin (see pynwb tutorial),compression_opts,maxshape,fillvalue,shuffle,fletcher32. In practice,fillvalueandshuffleare rarely if ever used. Blosc compression through hdf5plugin is sometimes used because it is usually better than gzip.Maxshapewill be used more often, pending an upcoming change https://github.com/hdmf-dev/hdmf/pull/1064. Note thath5py.Datasetsupports even more kwargs that HDMF does not currently support.I started adding support for additional compression algorithms as a nice exercise and will see how far I can get. I'll update here in a few days. This is not high priority but it would be nice to have more complete support for writing/appending datasets via PyNWB to a LINDI file.