[Nexus] Nexpy

[Osborn, Raymond]

Steve,
The NeXus format has, since its inception, had the concept that there is a parent object for each link, defined by the attribute ‘target’, even though linked data are stored in HDF5 files as hard links. NeXpy, or rather the underlying nexusformat API, is designed to make that association clear, whereas h5py makes no such distinction. I’m not sure if there is even a way for h5py to tell what groups an object has been added to without checking object ids.

If you have a recent version of NeXpy installed, you can open the example file, ‘chopper.nxs’, from the Help menu. I will use it to discuss why I believe it handles links appropriately. In ‘chopper.nxs’, the data group contains two links. Here are edited selections from the file tree:

entry:NXentry
  data:NXdata
    polar_angle -> /entry/instrument/detector/polar_angle
  instrument:NXinstrument
    detector:NXdetector
      polar_angle = float32(148)
        @target = '/entry/instrument/detector/polar_angle'

So ‘/entry/data/polar_angle’ and ‘/entry/instrument/detector/polar_angle’ are both paths to the same HDF5 object, but the parent object is deemed by the NeXus format to be the latter, as specified by its target attribute. NeXpy chooses to display the other instance, in ‘/entry/data’ as a link to make this clear.

>>> chopper.entry.data.polar_angle
NXlink('/entry/instrument/detector/polar_angle')
>>> chopper.entry.instrument.detector.polar_angle
NXfield(array([  -7.19999981,   -6.5999999 ,   -6.        , ...,  116.3999939 ,
        116.99999237,  117.59999084], dtype=float32))

If you want to know what the parent object is, you can use the NXlink’s ‘nxlink’ attribute, which returns the parent.

>>> chopper.entry.data.polar_angle.nxlink
NXfield(array([  -7.19999981,   -6.5999999 ,   -6.        , ...,  116.3999939 ,
        116.99999237,  117.59999084], dtype=float32))

Within the NeXpy GUI, you can right-click on the link, which has a chain link icon by it, to display the parent. There is no way to do this in a generic HDF5 viewer.

In practice, the fact that ‘/entry/data/polar_angle’ is an NXlink object should make no difference in your use of the data. For example, you can check its dtype or shape, and use fields in arithmetic expressions on either the parent or the link.

>>> chopper.entry.data.polar_angle + 100.0
NXfield(array([  92.80000305,   93.40000153,   94.        , ...,  216.3999939 ,
        217.        ,  217.59999084], dtype=float32))
>>> chopper.entry.instrument.detector.polar_angle + 100.0
NXfield(array([  92.80000305,   93.40000153,   94.        , ...,  216.3999939 ,
        217.        ,  217.59999084], dtype=float32))

I think this is the right way to handle links. In h5py, you could update a dataset’s value without knowing that it has the side effect of changing its value in another group. In NeXpy, the existence of the link is clear but it shouldn’t impede your use of its values or attributes. I would welcome further comments if you think these are not the right calls.

With best regards,
Ray

On Jun 14, 2017, at 6:13 AM, steve.collins at diamond.ac.uk<mailto:steve.collins at diamond.ac.uk> wrote:

Hello,

I’d like to subscribe to this email list. I’m not sure if it is appropriate to ask a technical question here, but here is one anyway (I hope it’s OK):

I have a question about the Python Nexusformat/NeXpy nxload method.

I seem to have problems reading data entries that have been defined using links. That is, the method returns an NXlink object rather than the actual data:

n.entry1.default.ic1monitor
Out[143]: NXlink('['/entry1/instrument/ic1monitor/ic1monitor']')

If I use h5py directly then I get the dataset:

h['/entry1/default/ic1monitor']
Out[144]: <HDF5 dataset "ic1monitor": shape (10001,), type "<f8">

Is this expected or am I doing something wrong? I read in the manual that all data are specified via links so there is no distinction between the 'original' data and a link. Is that right?

Thank you for your help.

Regards,
Steve Collins (Diamond Light Source)

P.S. I can send you the data files and Python code of course, if that helps.

--
Ray Osborn, Senior Scientist
Materials Science Division
Argonne National Laboratory
Argonne, IL 60439, USA
Phone: +1 (630) 252-9011
Email: ROsborn at anl.gov<mailto:ROsborn at anl.gov>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.nexusformat.org/pipermail/nexus/attachments/20170614/3d22f627/attachment-0001.html>