ml4ai/automates-v2

AutoMATES: Automated Model Assembly from Text, Equations and Software

automates-v2's Issues

Some loose ends

In creating a (de)serializer for Scala (in https://github.com/kwalcock/skema), I noticed these issues, which are probably known, but here are some notes just in case:

• Several uids are still strings and don't have format uuid.
• The values of LitervalValue and TypedValue are just objects with no properties.

gromet_FN_v0.1.5.yaml is not validating

metadata_collection of gromet_FN needs polymorphism

It looks like

        metadata_collection:
          description: |
            Table (array) of lists (arrays) of metadata,
            where each list in the Table-array represents the collection
            of metadata associated with a GroMEt object.
          type: array  # The Table of metadata_collection(s) (lists)
          items:
            type: array  # Collection (list) of metadata associated with a particular GroMEt object.
            items:
              $ref: "#/components/schemas/Metadata"

means that anything that includes an

      allOf:
        - $ref: '#/components/schemas/Metadata'

is allowed. In order for code generation to work, some things need to be added, including a discriminator and probably oneOf. See also https://swagger.io/docs/specification/data-models/oneof-anyof-allof-not/.

metadata_collection output includes array of array of array

This isn't exactly the right place to record this, but the file dict1--Gromet-FN-auto.json includes in its metadata_collection several times a "source_code_reference" that is wrapped in an array. The computer is telling me that's not right. It's one too many arrays down.

Code generation and need for discriminator

Java/Scala code generation doesn't work from files like gromet_FN_v0.1.4.yaml and gromet_metadata_v0.1.4.yaml for lack of a discriminator in superclasses. For example,

    GrometObject:
      description: Generic base class for any GroMEt object
      type: object
      properties:
        metadata:
          description: Index (integer) into the metadata_collection table in GrometFNModule.
          type: integer
# Lines like the ones below need to be added     
        gromet_object_type: # added
          type: string #
      discriminator: # added
        propertyName: gromet_object_type

When subclass objects are written out, a field should then appear as such:

{
    "gromet_object_type": "GrometFNModule",
    "schema": "FN",
    "schema_version": "0.1.4",

This allows the generated Java/Scala code to figure out how to deserialize the object.

I don't (yet) know how the Python code is working without this, but will probably notice soon. I am hand editing the files like cond1--Gromet-FN-auto.json to add this information to make sure the updates to the yaml file that I'm about to suggest work.

Perhaps this issue has been encountered before, so I'm filing this slightly prematurely in case there's more I should know before continuing.