Databases Reference
In-Depth Information
• Compatible with all lossless filters
Third-Party Filters
Many other filters exist for HDF5. For example, the BLOSC compressor used by the
PyTables project
is highly tuned for speed. Other filters also exist that are based on the
LZO compression system, BZIP2, and more. You can find the most recent list (and
contact information for the filter developers) at the
HDF Group website
.
There's one more thing to mention, and it's a very new feature (introduced in HDF5
1.8.11): dynamically loaded filters. For the filters listed previously, you have to man‐
ually “register” them with the HDF5 library before they can be used. For example, h5py
registers the LZF filter when it starts up. Newer versions of HDF5 will automatically
load filter modules from disk when they encounter an unknown filter type in a dataset.
Dynamic loading is a new technology in HDF5 and support for it in h5py and PyTables
is still evolving. Your best bet is to check the h5py and PyTables websites for the most
up-to-date information.
And one final reminder: when you use a filter, be
sure
the people you intend to share
the data with also have access to it.
