On Jul 13, 2016, at 6:04 PM, Roy Mendelssohn - NOAA Federal
<roy.mendelssohn@xxxxxxxx<mailto:roy.mendelssohn@xxxxxxxx>> wrote:
Since Aleksander has kindly answered my first question, this raises a second
question on HDF virtual files - is each file in the virtual file a "valid" file
in and of itself, that is combined by whoever is the "base" file of the virtual
file, or does it depend on the "base" file.
Let’s first inter-calibrate the HDF and netCDF vocabularies to avoid confusion:
* HDF5 dataset = netCDF variable
* HDF5 file = netCDF dataset.
About HDF5 virtual datasets:
* There are no HDF5 virtual files, only HDF5 virtual datasets inside a
normal HDF5 file.
* The data for a HDF5 virtual dataset comes from source HDF5 datasets
located in other HDF5 files.
* HDF5 virtual datasets can have “local” data in addition to the data
coming from source datasets.
* Every HDF5 file with virtual datasets is a valid HDF5 file. Source
datasets are just ordinary HDF5 datasets so files with them are valid HDF5
files, too.
* How the data from a source dataset is presented in its virtual dataset is
controlled by a mapping between the two dataset's HDF5 dataspaces (start,
stride, count, block).
* These mappings are set on HDF5 dataset creation and cannot be changed.
* There are HDF5 dataset access property settings that define how to deal
with missing source dataset data. (See my previous email.) These can be changed
each time a virtual dataset is opened for access.
* The fill value is used for any missing data in a virtual dataset.
User’s guide for the virtual dataset feature is being prepared; in the
meantime: https://www.hdfgroup.org/HDF5/Tutor/vds.html.
-Aleksandar
