Signal and Axis

[Ray Osborn]

on 2000/1/12 10:39 AM, Eric Boucher at boucher at aps.anl.gov wrote:

> 
> The idea of Ray is very similar to the one I have suggested in an earlier
> message to the NeXus list:
> 
>> That would be nice if plotable data sets could name their axis instead.
>> We could achieve that by assigning axis attributes to a plottable data set
> ..
>> Example for a N dimensions data set:
>> axis1: (TAG REF) (TAG REF)     would list all possible X axis
>> axis2: (TAG REF) (TAG REF) (TAG REF)  would list all possible Y axis
>> ...
>> axisN: (TAG REF) (TAG REF) (TAG REF)  would list all possible N'th axis
> 
> 

It was such a long time ago that we last talked about this that I had
forgotten your proposal, but indeed it was obviously the subconscious
inspiration for my own.  At the time, I had forgotten about the fact that
tags were all identical, which is why I wasn't sure the scheme would work.
We can't store NXlink structs as attributes, but we can store int32
attributes. 

I'm not sure about adding too many axis_id attributes to a data set
(axis_id1, axis_id2 etc).  The names rapidly start becoming very inelegant,
and I'm not sure they add much extra functionality.  Actually, if the users
are not going to use the default axes, then they will probably be involved
in making a manual decision about which axes to use (e.g. by clicking a
mouse). As long as they have enough information to make an intelligent
choice, we probably don't have to worry about it.

Thanks again for raising the issue.  I think this could be a very useful
addition to the format unless people identify some serious problems with it.

Regards,
Ray
-- 
Dr Ray Osborn                Tel: +1 (630) 252-9011
Materials Science Division   Fax: +1 (630) 252-7777
Argonne National Laboratory  E-mail: ROsborn at anl.gov
Argonne, IL 60439-4845