Image DM changes suggestions and questions

[Douglas Tody]

On Thu, 31 Oct 2013, Louys Mireille wrote:

> Daer Doug, mark , DM WG members,
>
> Here are a few more suggestion for changes or precisions to ImageDM:
>
> Dataset.Image.Nsubarrays
> Dataset.Image.Naxes
> Dataset.Image.Naxis
>
> These are dimensions specific for Image, cubes , sparse cubes :
> I would prefer Dataset.Dimensions to replace prefix: Dataset.Image ;this is 
> more general.

Dataset.Image holds summary metadata specific to Image datasets (where
we mean Image in the general sense, regardless of dimensionality).
Dataset.Dimensions would only work if Dimensions applies to any type
of Dataset, but it doesn't; these attributes are specific to image data.

> We know we are dealing with Image instance because:
> Dataset.dataproductType='image' and we refer to DataModel.name='ImageDM'

Yes we know it is an Image instance by inference, but it is better to
be explicit.  Something like Image.Naxes is very clear.

> Provenance :
> more preferable to use ObsCore Utypes , because they reflect better possible 
> extensions
> on the Instrument and Facility class.
> Up to know Obscore has an embryonic part of ObsConfig that can be re-usedand 
> developped further .

Ok.

> Provenance.ObsConfig.Bandpass   not clear what to do ?
> Provenance.ObsConfig.DataSource  more explanation needed

These just need elaboration; they carry over from Spectrum and now
Spectral.

> Complete ImageDM Items with ObsCore fields :
> refval but also min and max values if defined in Obscore . Could be defined 
> in ImageDM but optional .

Agree; ImageDM already uses refval in most places.

> Data.Naxes
> Data.Naxis
> isn't it redundant with  Dataset.Image.Naxes, Dataset.Image.Naxis ?
> Are these the particular sizes of each chunk of data?
> Then Dataset.Image.Naxes and Dataset.Image.Naxis could be the overall
> size of the combination in case of complex , or assembled data chunks.
> Is it clearly defined in Image DM?
>
>From mdittmar at cfa.harvard.edu Thu Oct 31 09:44:16 2013
> Doug can correct me if I'm wrong, but I believe that Dataset.Image is in
> the model primarily for query response.. it should not be part of the
> core Image model.  It duplicates information found in the Data element..
> as you note below.  This is definitely one of the pieces I'd like to
> talk about in detail.

Dataset.Image describes the overall Image dataset; Data.Naxes etc.
describe an individual Data element, and these attributes may differ for
individual Data elements.  We have to have something here to describe
the overall Image dataset; as Mark notes this is required for the query
response, but it is useful in any case to provide this important
information in a uniform way regardless of the substructure of the Image
dataset in terms of subarrays.  While there is some duplication of
information in that the Dataset.Image attributes may be derived from the
aggregation of Data elements, it is not necessarily mere duplication,
e.g, Image.Naxes may in principle exceed Data.Naxes for an individual
subarray.  If we were always including Data in an ImageDM instance we
might want to do something different, but since Data is always either a
separate submodel instance or there may be multiple instances, we need
something like this.

In more practical terms: users *will* want this summary information.  We
might actually want to add some more here, e.g., some better indication
of whether the image is sparse.

> Is it clearly defined in Image DM?

More or less - see 3.2.1.1.