Assemble model descriptors from graph

[jmakeig]

As a data steward, I want to be able to organize my entities into descriptor docs as I see fit, for example, to group related entities into their own descriptor documents or separate entities into one descriptor per entity or to have just one descriptor doc for all entities. Each grouping should distill into the same entity type graph. Today, the downstream artifact generators assume one in-memory map structure that represents the entire type model. This is a direct mapping from a single descriptor doc that represents all entities. It should be possible to assemble the in-memory descriptors needed from code gen regardless of how the input entity definitions are grouped.

One implementation idea is to provide a function that would operate on the triples to assemble a map/object. Someone extending the model would also use this same technique to build custom in-memory representations of their extensions.

Some things to consider:

• Versioning today is declared at the descriptor. Thus all entities declared in a single descriptor will have the same semver. I’d expect stewards to want to version entities at different rates.
• (Now I can’t remember the other thing)

[jmakeig]

@grechaw @damonfeldman @paxtonhare

[grechaw]

I consider this a high-priority item for vnext. Creating entity models out of external, say, RDF-based models requires a method for making models out of triples. Just made #241 as a recipe that would depend on this functionality.

[grechaw]

needs a little design spec/proposal next

[grechaw]

Added labels for BT and planning purposes. I'll have some prototyping to show when we return from London, or before.