Cube/ImageDM - new diagram set.

[CresitelloDittmar, Mark]

Francois,

Thank you for looking at these so promptly!


On Fri, Apr 4, 2014 at 10:13 AM, François Bonnarel <
francois.bonnarel at astro.unistra.fr> wrote:

>  Hi Mark, all,
>
>     Very nice diagrams and huge work !!!
>
Thank you


>     I have some very general questions at this stage:
>
>        - HyperCube has NDpoints "data" (à la Spectrum ? ) while Image has
> ImageData (basically an Array "à la FITS image")
>          To me HyperCUbe was also an Array with dimension higher than 3.
> Why this change ?
>

My intent is:
  HyperCube = sparse data (e.g. event list), no pixelized coordinates,
filling ND data space.
     * In some discussion and thought since the diagrams went up, I think
there may be a need to modify this hierarchy a bit to:
          HyperCube contains PointData (extended from Data)
          PointData contains a collection of NDPoint (extended from <Point>)
           <Point> Not extended from <Data>

       Which would provide more symmetry between HyperCube and Image, and
(I think) will set things up better
       for Datasets like Echelle spectra or time series.

  Image == NDImage, where there is regular pixelization.
    * I probably should avoid using straight "Image" term since it often
conveys 2D implications.




>        - Observation has always been a little bit ambiguous term : were we
> talking of the "Observing process" (= experimental action) or of the result
> (= data) ?
>           Definitly you seem to have chosen . Observation = some Kind of
> experiment ? Hence the concept of provenance is disapearing "per se". I
> don't see  anymore  "Sofwtare Provenance" related to Dataset.
>

Yes.  I am treating "Observation" as an Experimental action which produces
an Observation Dataset (ObsDataset).
I expect the Observing process would include data reduction/processing.

We discussed earlier whether a Dataset should have a "Provenance" node, and
it was decided no.. there is no logical
head to Provenance.  The elements which had 'Provenance' in the utypes
previously are still there (ObsConfig, etc).

I admit to not having read the Provenance document put up some time ago,
which may help with this undermodeled regime of metadata.
I don't recall seeing 'Software Provenance' related to Dataset.



>         -  I don't really understand the role of the AxisCoordinates
> attached to ImageData ?
>                 * Are they metadata describing the axes ? But in that case
> isn't that redondant with characterisation ?
>                 * Are they there to store real values ? But what does that
> mean for an array ,
>                 * Are they there just to point towards Frames?  I see
> where the extensive WCS information can be stored in modified STC frames,
> but it seems to be really dispersed information.
>
>
The *AxisCoordinates are the values.  A *AxisCoordinate is a grouping of
the value, errors, resolution, binspec, quality.

The Char.CharAxis metadata would be derived from the corresponding one of
these, where there is overlap.
  eg: the CharAxis.Accuracy.SysError would provide the average (eg) of the
corresponding *AxisCoordinate.sysError.refVal set.

The WCS information is stored in CoordSys, these have references to the
corresponding frame in order to execute the Transform(s) if needed.

Also, I am thinking that I need to have ImageData contain a collection of
Pixel
where Pixel holds the 3 AxisCoordinates.. this will again, make for better
symmetry and clarify the roles.


Hope this helps
Mark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.ivoa.net/pipermail/dm/attachments/20140404/7c4381ae/attachment-0001.html>