Since MPI-POSIX VFD has been removed from hdf5 > 1.8.12, but a test
with stride data should be performed, we use independent IO instead for
hdf5 > 1.8.12
- H5PartSetNumPoints() renamed to H5PartSetNumItems()
- H5PartGetNumPoints() renamed to H5PartGetNumItems()
- Dataset names longer then 64 bytes are handled as error.
- Same for step/iteration names.
* core API
- we use the term 'iteration' instead of 'step'
- we use the term 'item' instead of 'point'
- re-factor function and variable names
- in printing messages/debug output fixed
- do not flush (sync to disk) after writing a dataset by default,
can be controlled by a property
multiple steps:
- the functions for inquiring datasets and attributes do not return
an HDF5 type any more but an enum of type h5_types_t. This change
was required for the Python module.
- bugfix in reading attributes: See https://git.psi.ch/H5hut/src/issues/4
- several consts and macros have been moved from the public C-API to
the core API
- more consitent file naming
- several 'private' function have been moved to their 'private' header
files as 'static inline'.
- minor formatting changes
gsell
Gsell Achim