[QUANTITY]Re: Quantity.owl

[Ed Shaya]

DIDELON Pierre wrote:

>Hi,
>
>is this the simple quantity class/concept that Ray want to subclass?
>I must admit I am afraid, and for me it seems more related to
>the analysis of a global structure concerning the more general
>concept of VOData, related to [OBSERVATIONS] theme.
>  
>
There was indeed a choice to be made on the design of this.  We could 
have started with a very simple Quantity, ie very few attributes 
(properties in OWL) and then subclass these with more fully featured 
specialized Quantities.  For example start with a Quanitity that can 
only be single valued and has no error and no units.  Then subclass this 
into an Array and then into ArraywithError and then to 
ArraywithErrorandUnits, etc.  So one builds things up slowly so one can 
associate specific astronomical UCD with specific QuantityTypes.

The other way is to create a fully featured Quantity and then, as 
needed, one defines subclasses as restrictions on the general case.  For 
instance a VectorQuantity can be defined as a Quantity that takes a 
DataFormat of VectorDeclaration  only.  Then each of its components are 
ScalarQuantities which can take DataFormat of Float only.

We chose the second way because the number of features that are commonly 
needed is fairly large.  If the model has all of the permutations of 
all  features there  are simply too many  subclasses and it  is  
unnecessarily complicated.  One runs the risk of creating many subclasses
that are not needed or, in the end, not creating the specific case that 
really is need so one
falls back on restriction again anyway.   So we found, the hard way, I 
might add, that it is best for each property range to be an abstract 
class that holds various possibilities and as one needs
one can restrict the range to a particular (set of) concrete class.

>Marginally; repeating data member inhereted from the parent class
>overload the schema without bringing in any add. information.
>  
>
Are you referring to QuantitySet?  We have the particular case that a 
Quantity (say a density map) will be a data hypercube with Argument 
Quantities (x,y,z say) and a QuantitySet (say a table) is also a 
dataCube with Argument Quantities (row, column, perhaps a 3 dimension).  
So it makes sense for the Set of Quantities to be a Quantity itself.  
Perhaps QuantitySet should
inherit from a VO:Set and a QuantityType to make this clearer.  However, 
you make a good point in that the QuantitySet is not exactly like a 
Quantity because it does not make sense to put Units or Values on it, so 
perhaps Quantity and QuantitySet should inherit from a QuantityType 
without those things.  I will make this change.

>--
>Pierre
>-------------------------------------------------------------------------------
>DIDELON                               e-mail : pdidelon_at_cea.fr
>CEA SACLAY - Service d'Astrophysique  W3 : http://www-dapnia.cea.fr/Sap/
>91191 Gif-Sur-Yvette Cedex            Phone : 33 (0)1 69 08 58 89
>-------------------------------------------------------------------------------
>
>
>
>  
>