transform_value_for_tile for resource-instance datatype simplification/enhancement

[whatisgalen]

The ResourceInstanceDataType class supports this kind of value for importing from, for example, csv:

def transform_value_for_tile(self, value, **kwargs):
        try:
            return json.loads(value)
        except ValueError:
            # do this if json (invalid) is formatted with single quotes, re #6390
            try:
                return ast.literal_eval(value)
            except:
                return value
        except TypeError:
            # data should come in as json but python list is accepted as well
            if isinstance(value, list):
                return value

Meaning that each value has to look like this:

{
    "resourceName": "",
    "ontologyProperty": "",
    "inverseOntologyProperty": "",
    "resourceId": ""
}

This datatype already considers data valid if it only contains a value for "resoruceId" and nothing else. It would be nice if a user could import a value that only contained a resourceId uuid string and then the datatype method populates the rest of the boilerplate with empty string values.

[jacobtylerwalls]

Funny, I just hacked out a version of this in #11596 (not that it's factored perfectly), and I'd definitely be up to pull it out into its own PR or collab on your implementation. Say the word.

[whatisgalen]

I'm hacking out a proof of concept right now to import either a single uuid string or a list of uuid strings from a csv cell. I'd be happy to push to a branch to see if there's a collab that makes sense

[jacobtylerwalls]

My hack also took a list of python UUID types (not strings), but doesn't handle a jumbled array of some-of-one-and-some-of-the-other, which would be easy to handle if factored well. Happy to review!

[chiatt]

The ResourceInstanceDataType class supports this kind of value for importing from, for example, csv:

def transform_value_for_tile(self, value, **kwargs):
        try:
            return json.loads(value)
        except ValueError:
            # do this if json (invalid) is formatted with single quotes, re #6390
            try:
                return ast.literal_eval(value)
            except:
                return value
        except TypeError:
            # data should come in as json but python list is accepted as well
            if isinstance(value, list):
                return value

Meaning that each value has to look like this:

{
    "resourceName": "",
    "ontologyProperty": "",
    "inverseOntologyProperty": "",
    "resourceId": ""
}

This datatype already considers data valid if it only contains a value for "resoruceId" and nothing else. It would be nice if a user could import a value that only contained a resourceId uuid string and then the datatype method populates the rest of the boilerplate with empty string values.

It would be important to use the default values for the relationship and inverse relationship if they are available rather than empty strings.

[jacobtylerwalls]

Great point, and that makes this trickier when multiple graphs are allowed. To get the correct ontology props, we'd need the graph, which we don't have from an incoming UUID. We do query the resource table with the incoming UUID in validate(strict=True), so perhaps we should have some temporary lookup state like the concept datatypes do to avoid multiple queries.