[NeXus-committee] Event-mode

[Peterson, Peter F.]

Names that are unknown to NeXus are not invalid. I think they get warnings
in the validation tools. Older SNS files were written with NXgroup, newer
ones have NXcollection. People aren¹t yelling about the legacy files so
we¹ve just left them alone. Mantid has a check for the name, opens the
group, and carries on.

On the topic of mantid, there is a LoadEventNexus (for event nexus),
LoadRaw (legacy raw file format), and LoadISISNexus (ISIS histogram nexus
files). This is more on point with what people were saying about things
being written against specific schema. Which is really what we are trying
to reduce. 

The format of the events out of the DAS (as in electronics and live
streaming) is different, but the file written to disk should be the same
for event nexus files from ISIS and SNS. If your facility want to encode
the events on disk in a different format, then write a definition that
describes it, and expect that existing code will not (without
modification) be able to read it and understand it correctly. This isn¹t
saying that you aren¹t still writing nexus files. It is just that you are
creating a new definition with its own semantics. The idea behind nexus in
general is to use what exists as much as you can to reduce the amount of
diversity in the code that reads things in. If you want to just use HDF5
and not follow any nexus conventions, then you will have similar amounts
of reuse.

To really understand what nexus is, reread this email with the word
³ascii² in place of ³nexus². Šand are we using utf-8 or utf-16?

P^2

On 7/16/14, 10:04 AM, "Joachim Wuttke" <j.wuttke at fz-juelich.de> wrote:

>On 07/16/2014 03:34 PM, Peterson, Peter F. wrote:
>> Both SNS and ISIS write out event data that validate using
>> the definition and can be read by the same code without specialization.
>
>How can it be that SNS event-mode files validate if they use
>the invalid keyword NXgroup, as discussed a few days ago?
>
>Why does the data-reduction program Mantid provide different
>loaders for SNS and ISIS data?
>
>I was told that the event-data format at these two facilities
>is different because the electronics is different.
>
>My facility has a third type of electronics, with zero pulses
>stored as events instead of being used to re-zero the timer
>for the neutronic events. It needs only 6 bytes per event,
>compared to 64 bytes in SNS-style NeXus. Even if technically
>feasible, it would seem foolish to map our compact binaries
>upon SNS-style NeXus events.
>
>By embedding our binaries in a NXcollection we can ensure that
>our raw-data files are valid NeXus. But we will still need a
>loader that knows about our electronics.
>
>- Joachim
>