I fully understand the value of doing this in UML because it is an accepted modeling language. My two problems with it is that for one this should be represented in a machine readable form so that we can verify and ultimately use it to translate this into the
XML and JSON representation as well as potentially implementation in programming languages.
[GME] I agree and the machine readable form is already available. Papyrus stores the visual design in the XML model Interchange language (XMI). In Papyrus, the XMI files have
the file extension “.uml”.
Secondly, I feel like this is not capturing a lot of the "essence" (for a lack of a better term) of CSDL. For example EnumMembers are contained in an
EnumType, are uniquely named (unique in the EnumType), and have a fully qualified name. This "pattern" shows up in many places in CSDL, all model elements are uniquely named through their fully qualified name (which is based on containment) and references
to these elements are expressed via these names (e.g. in the Type property of a Property.
[GME] Of course, the example I showed you is very incomplete and, in some cases, wrong… the example needs work… Additionally, UML has many of those same constraints. Note,
from a UML point of view, EnumerationLiterals (i.e., EnumMembers) are not contained in an Enumeration (EnumType). Instead, they specify the value of an instance of the enumeration.
George
I want to present a custom "meta modeling language" that captures that. It is minimal and can probably be explained in a few pages. But that I also admit imediatly that this is a problem since it only exists for that purpose.
I will work more on this but just to give you an idea, below is my (incomplete) proposal.
