valid values and properties now use labels at the same time

[andrewelamb]

Problem:

The Curator extensioncan change property names and enum values to a verison thats more friendly to being annotation keys and column headers for Curator. This includes stripping whitespace and special characters. The problem is these were two separate parameters, so a user could do this for only property names, but leave enum values the same. This could cause issues if the data model were doing conditionals:

{
  "$id": "http://example.com/Cell",
  "$schema": "http://json-schema.org/draft-07/schema#",
  "allOf": [
    {
      "if": {
        "properties": {
          "CellType": {
            "enum": [
              "TCell"
            ]
          }
        }
      },
      "then": {
        "properties": {
          "TCellType": {
            "not": {
              "type": "null"
            }
          }
        },
        "required": [
          "TCellType"
        ]
      }
    }
  ],
  "description": "TBD",
  "properties": {
    "CellType": {
      "description": "TBD",
      "enum": [
        "bCell",
        "tCell"
      ],
      "title": "cellType"
    },
    "TCellType": {
      "description": "TBD",
      "title": "tCellType"
    }
  },
  "required": [
    "CellType"
  ],
  "title": "Cell",
  "type": "object"
}

Solution:

create_json_schema and all its helper functions now only have one parameter: use_display_names. When this is True both properties and enum values aren't modifed, and vice-versa. This ensures that enum values and properties will be consistent in the JSON Schema.

[linglp]

The changes look good to me overall. I like that now we have a single parameter to control if display names are used! One thing that I still think is a little bit confusing is for generating JSON schema, we have either class_label or display_label when generating a JSON schema, but here the option is called use_display_names. Should we make the naming more consistent in a future PR? (use_display_names -> use_display_label)

[linglp]

hmm is this correct for generating json schema? The function of generate_jsonschema looks like this:

def generate_jsonschema(
    data_model_source: str,
    synapse_client: Synapse,
    data_types: Optional[list[str]] = None,
    output: Optional[str] = None,
    data_model_labels: DisplayLabelType = "class_label",
) -> tuple[list[dict[str, Any]], list[str]]:

There's only data_model_labels option, not use_display_names

[linglp]

use__display_names should also be use_display_names

[linglp]

I think this is confusing. With the change to a single use_display_names parameter, you can no longer control property names and valid values independently.

[linglp]

Nit: this should be #9