Feat/s2dm exporter enhancements

[jdacoello]

No description provided.

[sschleemilch]

why not a dict as value, so taking 1:1 what is in the source?

[jdacoello]

This metadata can hold more than only the fields reported by the vss model. For example, I am now annotating when a name was modified. For instance, a leaf name in VSS is PascalCase, whereas in the export it becomes a field inside a type, which in GraphQL conventions is done with camelCase. However, I think I can assign to metadata:

Dict found in VSS model + Dict of extra stuff.

[jdacoello]

Also, a dictionary is not a built-in scalar in GraphQL. There are other community driven scalars specified. But, they don't offer the dictionary scalar. The closest one would be to say it's a JSON.

"""Annotation directive for Vspec mapping information."""
directive @vspec(
  """The element of the Vspec language to which the item in the GraphQL schema maps."""
  element: VspecElement,
  """The Fully Qualified Name (FQN) of the related element in Vspec (aka., path). Example: Vehicle.Cabin.Door.Window.position"""
  fqn: String
  """Additional metadata associated with the mapping."""
  metadata: JSON
) on OBJECT | FIELD_DEFINITION | ENUM | ENUM_VALUE

@vspec(
  element: SENSOR,
  fqn: "Vehicle.Speed",
  metadata: '{"source": "ecu0xAA", "quality": "100"}'
)

Is that what you mean? It seems feasible. But, I see a few limitations with that:

• Less GraphQL-idiomatic (clients must parse JSON string), where the JSON can have multiple nested objects
• No type safety
• GraphQL introspection will provide metadata as a string

[sschleemilch]

that's what I meant but encoding JSON into a string is not that great... then let's stick with the typed key vale pair. Although there exists the same problem that all values are forced to be strings and we are limiting ourselves to non complex metadata values, which in turn means that complex vales could be JSON strings...

[sschleemilch]

sources are gql specific and should be moved into s2dm dir

[jdacoello]

The graphql related utils were moved to the s2dm directory as suggested.

[erikbosch]

@jdacoello - I noticed a regression in VSS master branch as the test there using Makefile gives a file rather than a directory as output, which was changed by your latest PR. No worries, I can fix that. I just want to double check that the help given for s2dm is correct. In the screenshot below it says for -o: "Output file or directory (when using --modular)". That text is understand it obsolete, could you if so possibly change it as part of this PR.