[Nexus] Nexpy

[Tobias Richter]

The closest thing to an official statement is that there is no official statement what APIs people should use. That choice is for the facility or individual user to make based on their needs and preferences. To determine whether your output is valid you can check with (c)nxvalidate. Validation tools can be improved and not everything is currently caught, but they are a better thing to standardise on than forcing APIs on people.

In the absence of any boundary conditions python/nexusformat/NeXpy is a very fine option. Many NeXus features have appropriate high level abstractions and Ray et al do a very good job supporting this (shown for example in this thread). 

By the way: Both for the new-style plot axes and for the CIF-style geometry NeXus does use attribute arrays, which is even supported in the legacy NAPI. In both cases the arrays are tiny, commonly less than 5 elements.

Best wishes,
Tobias


On 15/06/2017, 11:43, "NeXus on behalf of Carlos Pascual" <nexus-bounces at nexusformat.org on behalf of cpascual at cells.es> wrote:

    First, one clarification: I was wrongly assuming that the proposed conversion 
    of size 1 arrays  to scalars in nexusformat applied to datasets, not 
    attributes. It was under that (wrong) premise that I qualified nexusformat as a 
    "non-trivial" wrapper.
    
    Now that I stand corrected, I would be comfortable with an interpretation such 
    as Armando's : 
    
    > As you see our point of view is that nexusformat is just a convenience
    > tool on top of h5py being the later the standard (Python) tool to access
    > HDF5 files.
    
    So, my question is: could the above be considered the "official" consensus at 
    the NIAC?
    
    (if so, I think it would be nice to state it more explicitly somewhere in the 
    manual)