NXcomponent as a parent base class

[lukaspie]

Disclaimer: This is a draft and not meant to be voted on yet. The idea here is to perform an exploration/survey of the consequences of what having an NXcomponent base class would mean.

This implements the discussion items in #1519 (comment).

NXcomponent is meant as a parent base class that many base classes describing devices/instrument components should derive from. As requested in the comment above, this is meant as a survey of base classes to identify those which should and those which should not derive from NXcomponent.

These are the base classes that are now inheriting from NXcomponent:
in base_classes:

• NXaperture
• NXattenuator
• NXbeam_stop
• NXbending_magnet
• NXcapillary
• NXcollimator
• NXcrystal
• NXdetector
• NXdisk_chopper
• NXelectron_detector (through NXdetector)
• NXfermi_chopper
• NXfilter
• NXflipper
• NXfresnel_zone_plate
• NXgrating
• NXguide
• NXinsertion_device
• NXmirror
• NXmoderator
• NXmonitor
• NXmonochromator
• NXpinhole
• NXpolarizer
• NXpositioner
• NXsample
• NXsample_component
• NXsensor
• NXslit
• NXvelocity_selector
• NXxraylens

in contributed_definitions:

• NXaperture_em
• NXbeam_splitter
• NXchamber
• NXcollectioncolumn
• NXcontainer
• NXcorrector_cs
• NXdeflector
• NXebeam_column
• NXelectronanalyser
• NXenergydispersion
• NXibeam_column
• NXlens_em
• NXlens_opt
• NXmagnetic_kicker
• NXmanipulator
• NXpid
• NXpolarizer_opt
• NXpulser_apm
• NXpump
• NXquadrupole_magnet
• NXreflectron
• NXscanbox_em
• NXseparator
• NXsolenoid_magnet
• NXspin_rotator
• NXspindispersion
• NXstage_lab
• NXwaveplate

In addition, the default attribute is added to NXobject so that it doesn't have to be repeated in every class. As a consequence, it can be removed This affects almost all of the classes above, but additionally the default attribute is removed from these classes (which are still inheriting from NXobject, not from NXcomponent):

• NXbeam
• NXcite
• NXdetector_group
• NXdetector_module
• NXevent_data
• NXgeometry
• NXinstrument
• NXlog
• NXnote
• NXoff_geometry
• NXorientation
• NXparameters
• NXprocess
• NXreflections
• NXshape
• NXtransformations
• NXtranslation
• NXuser

further changes:

• NXcylindrical_geometry: rename "beamline component" to "component" in docs
• NXoff_geometry: rename "beamline component" to "component" in docs

Open questions:

• Should any of these also inherit from NXcomponent?
  • NXbeam
  • NXdetector_module
  • NXinstrument

[lukaspie]

Quick comment that I missed to mention in today's telco: are there any tools that would expect these classes to extend NXobject and not NXcomponent? For example, conditional checks like if group.attribute(NX_class) == "NXobject".

We had a similar problem before when we tried to subclass NXdata (e.g. to NXdata_mpes) that some of the visualization tools (H5web in particular) are only checking for NXdata and not for subclassed versions, so the default visualization breaks.

Since base class inheritance is now supported, such checks will anyway break on a smaller scale, e.g. for NXelectron_detector extending NXdetector (already part of the standard), so maybe it would be even good to do this once for many of the fundamental classes (like here) so that such errors are seen in most places and can be corrected immediately.

[rayosborn]

The question of whether there are tools to implement inheritance is impossible to answer. I can modify the nexusformat package to explicitly include this inheritance, since all the base classes are really classes (i.e., not just objects with an NX_class attribute). However, inheritance is meaningless if someone is using, say, h5py or H5web directly, since HDF5 groups have no class. The NXvalidate package explicitly extend the NXobject class and will, I believe, automatically extend the NXcomponent class once it is approved, but again that is up to the package developer.

That is one reason I have argued in the past for a package like NeXpy to be officially supported or at least recommended, i.e., so that there is a generic NeXus viewer that is guaranteed to follow changes in the standard, but I lost that battle years ago.