[MAHID] Discussion at the recent SPring8/ESRF/APS/Petra meeting

[Chris Jacobsen]

Hi - there was a discussion on data formats at the recent SPring8/ESRf/
APS/Petra meeting that was so pleasantly hosted by SPring8 (April
12-14). There was certainly a desire expressed for unified data
formats, and fairly uniform support for NeXus/HDF5 provided it is a
finished, usable system for a variety of programming languages. There
was especially strong support from some at the management level of
some facilities. Pete Jemian then made the point that this requires
resources, in terms of programmer time, since NeXuS has been a part-
time activity for the programming team members.

Pete and I also talked about direct calls to HDF5 routines versus the
NeXuS API. An API is nice if it is easy to install, and debugged and
up-to-date on a wide variety of platforms and programming languages.
Unfortunately this might require more effort than is available right
now. The alternative is to read and write NeXuS-compliant files using
straight HDF5 calls, so that one defines the format by the structure
of HDF5 files. This has the advantage that HDF5 support is built-in
to many scripted languages like IDL, Matlab, and Python, and trivially
downloadable for C on nearly any computing platform. This doesn't
preclude having a NeXuS API also be available, but can make life
simpler in terms of avoiding the need to maintain and update
additional code needed for the blizzard of varieties of programming
languages and environments that exist.

CJ

--
You received this message because you are subscribed to the Google Groups "Methods for the analysis of hyperspectral image data" group.
To post to this group, send email to ma...@googlegroups.com.
To unsubscribe from this group, send email to mahid+un...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/mahid?hl=en.

["V. Armando Solé"]

Hi Chris,

Thanks for the feedback. As a matter of fact, I really like the
"pythonic" access offered by the standard python HDF5 bindings (h5py and
PyTables).

My opinion is that it requires much less effort from the NeXus
developers to keep an up-to-date stand-alone set of validation tools. In
that way, if one generates an HDF5 file with whatever tools, one can
always check for NeXus comformity and act in consequence. To maintain a
complete set of "official" NeXus bindings for C, Fortran, C++, Matlab,
IDL, Java, Python, and so on seems to me a lot of effort.

Best regards,

Armando